Datasets:

aalst

/

ms_stack

like 0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. from typing import Dict, Optional, Tuple import torch import torch.nn.functional as F from fairseq import utils from fairseq.incremental_decoding_utils import with_incremental_state from fairseq.modules.fairseq_dropout import FairseqDropout from torch import Tensor, nn try: from fairseq.model_parallel.megatron.mpu import ( get_cuda_rng_tracker, get_model_parallel_world_size, ColumnParallelLinear, RowParallelLinear, ) has_megatron_submodule = True except (ImportError, ModuleNotFoundError): has_megatron_submodule = False @with_incremental_state class ModelParallelMultiheadAttention(nn.Module): """Model parallel Multi-headed attention. This performs the Multi-headed attention over multiple gpus. See "Megatron-LM: https://arxiv.org/pdf/1909.08053.pdf" for more details. """ def __init__( self, embed_dim, num_heads, kdim=None, vdim=None, dropout=0.0, bias=True, self_attention=False, encoder_decoder_attention=False, ): super().__init__() if not has_megatron_submodule: raise ImportError( "\n\nPlease install the megatron submodule:" "\n\n git submodule update --init " "fairseq/model_parallel/megatron" ) self.embed_dim = embed_dim self.kdim = kdim if kdim is not None else embed_dim self.vdim = vdim if vdim is not None else embed_dim self.qkv_same_dim = self.kdim == embed_dim and self.vdim == embed_dim self.model_parallel_size = get_model_parallel_world_size() self.num_heads_partition = num_heads // self.model_parallel_size assert ( self.num_heads_partition * self.model_parallel_size == num_heads ), "Number of heads must be divisible by model parallel size" self.dropout_module = FairseqDropout( dropout, module_name=self.__class__.__name__ ) self.head_dim = embed_dim // num_heads assert ( self.head_dim * num_heads == self.embed_dim ), "embed_dim must be divisible by num_heads" self.scaling = self.head_dim ** -0.5 self.self_attention = self_attention self.encoder_decoder_attention = encoder_decoder_attention assert ( not self.self_attention or self.qkv_same_dim ), "Self-attention requires query, key and value to be of the same size" self.k_proj = ColumnParallelLinear( self.kdim, embed_dim, bias=bias, gather_output=False ) self.v_proj = ColumnParallelLinear( self.vdim, embed_dim, bias=bias, gather_output=False ) self.q_proj = ColumnParallelLinear( embed_dim, embed_dim, bias=bias, gather_output=False ) self.out_proj = RowParallelLinear( embed_dim, embed_dim, bias=bias, input_is_parallel=True ) def forward( self, query, key: Optional[Tensor], value: Optional[Tensor], key_padding_mask: Optional[Tensor] = None, incremental_state: Optional[Dict[str, Dict[str, Optional[Tensor]]]] = None, static_kv: bool = False, attn_mask: Optional[Tensor] = None, **unused_kwargs, ) -> Tuple[Tensor, Optional[Tensor]]: """Input shape: Time x Batch x Channel Args: key_padding_mask (ByteTensor, optional): mask to exclude keys that are pads, of shape `(batch, src_len)`, where padding elements are indicated by 1s. attn_mask (ByteTensor, optional): typically used to implement causal attention, where the mask prevents the attention from looking forward in time (default: None). """ tgt_len, bsz, embed_dim = query.size() assert embed_dim == self.embed_dim assert list(query.size()) == [tgt_len, bsz, embed_dim] is_tpu = query.device.type == "xla" if incremental_state is not None: saved_state = self._get_input_buffer(incremental_state) if saved_state is not None and "prev_key" in saved_state: # previous time steps are cached - no need to recompute # key and value if they are static if static_kv: assert self.encoder_decoder_attention and not self.self_attention key = value = None else: saved_state = None if self.self_attention: q = self.q_proj(query) k = self.k_proj(query) v = self.v_proj(query) elif self.encoder_decoder_attention: # encoder-decoder attention q = self.q_proj(query) if key is None: assert value is None k = v = None else: k = self.k_proj(key) v = self.v_proj(key) else: assert key is not None and value is not None q = self.q_proj(query) k = self.k_proj(key) v = self.v_proj(value) q *= self.scaling q = ( q.contiguous() .view(tgt_len, bsz * self.num_heads_partition, self.head_dim) .transpose(0, 1) ) if k is not None: k = ( k.contiguous() .view(-1, bsz * self.num_heads_partition, self.head_dim) .transpose(0, 1) ) if v is not None: v = ( v.contiguous() .view(-1, bsz * self.num_heads_partition, self.head_dim) .transpose(0, 1) ) if saved_state is not None: # saved states are stored with shape (bsz, num_heads_partition, seq_len, head_dim) if "prev_key" in saved_state: _prev_key = saved_state["prev_key"] assert _prev_key is not None prev_key = _prev_key.view( bsz * self.num_heads_partition, -1, self.head_dim ) if static_kv: k = prev_key else: assert k is not None k = torch.cat([prev_key, k], dim=1) if "prev_value" in saved_state: _prev_value = saved_state["prev_value"] assert _prev_value is not None prev_value = _prev_value.view( bsz * self.num_heads_partition, -1, self.head_dim ) if static_kv: v = prev_value else: assert v is not None v = torch.cat([prev_value, v], dim=1) prev_key_padding_mask: Optional[Tensor] = None if "prev_key_padding_mask" in saved_state: prev_key_padding_mask = saved_state["prev_key_padding_mask"] assert k is not None and v is not None key_padding_mask = ( ModelParallelMultiheadAttention._append_prev_key_padding_mask( key_padding_mask=key_padding_mask, prev_key_padding_mask=prev_key_padding_mask, batch_size=bsz, src_len=k.size(1), static_kv=static_kv, ) ) saved_state["prev_key"] = k.view( bsz, self.num_heads_partition, -1, self.head_dim ) saved_state["prev_value"] = v.view( bsz, self.num_heads_partition, -1, self.head_dim ) saved_state["prev_key_padding_mask"] = key_padding_mask # In this branch incremental_state is never None assert incremental_state is not None incremental_state = self._set_input_buffer(incremental_state, saved_state) assert k is not None src_len = k.size(1) # This is part of a workaround to get around fork/join parallelism # not supporting Optional types. if key_padding_mask is not None and key_padding_mask.dim() == 0: key_padding_mask = None if key_padding_mask is not None: assert key_padding_mask.size(0) == bsz assert key_padding_mask.size(1) == src_len attn_weights = torch.bmm(q, k.transpose(1, 2)) assert list(attn_weights.size()) == [ bsz * self.num_heads_partition, tgt_len, src_len, ] if attn_mask is not None: attn_mask = attn_mask.unsqueeze(0) attn_weights += attn_mask if key_padding_mask is not None: # don't attend to padding symbols attn_weights = attn_weights.view( bsz, self.num_heads_partition, tgt_len, src_len ) if not is_tpu: attn_weights = attn_weights.masked_fill( key_padding_mask.unsqueeze(1).unsqueeze(2).to(torch.bool), float("-inf"), ) else: attn_weights = attn_weights.transpose(0, 2) attn_weights = attn_weights.masked_fill(key_padding_mask, float("-inf")) attn_weights = attn_weights.transpose(0, 2) attn_weights = attn_weights.view( bsz * self.num_heads_partition, tgt_len, src_len ) attn_weights_float = utils.softmax(attn_weights, dim=-1) attn_weights = attn_weights_float.type_as(attn_weights) with get_cuda_rng_tracker().fork(): attn_probs = self.dropout_module(attn_weights) assert v is not None attn = torch.bmm(attn_probs, v) assert list(attn.size()) == [ bsz * self.num_heads_partition, tgt_len, self.head_dim, ] embed_dim_partition = embed_dim // self.model_parallel_size attn = attn.transpose(0, 1).contiguous().view(tgt_len, bsz, embed_dim_partition) attn = self.out_proj(attn) # return attn_weights None to keep the return type same as single gpu multihead attention # This will be deprecated. attn_weights: Optional[Tensor] = None return attn, attn_weights @staticmethod def _append_prev_key_padding_mask( key_padding_mask: Optional[Tensor], prev_key_padding_mask: Optional[Tensor], batch_size: int, src_len: int, static_kv: bool, ) -> Optional[Tensor]: # saved key padding masks have shape (bsz, seq_len) if prev_key_padding_mask is not None and static_kv: new_key_padding_mask = prev_key_padding_mask elif prev_key_padding_mask is not None and key_padding_mask is not None: new_key_padding_mask = torch.cat( [prev_key_padding_mask.float(), key_padding_mask.float()], dim=1 ) # During incremental decoding, as the padding token enters and # leaves the frame, there will be a time when prev or current # is None elif prev_key_padding_mask is not None: filler = torch.zeros(batch_size, src_len - prev_key_padding_mask.size(1)) if prev_key_padding_mask.is_cuda: filler = filler.cuda() new_key_padding_mask = torch.cat( [prev_key_padding_mask.float(), filler.float()], dim=1 ) elif key_padding_mask is not None: filler = torch.zeros(batch_size, src_len - key_padding_mask.size(1)) if key_padding_mask.is_cuda: filler = filler.cuda() new_key_padding_mask = torch.cat( [filler.float(), key_padding_mask.float()], dim=1 ) else: new_key_padding_mask = prev_key_padding_mask return new_key_padding_mask def reorder_incremental_state( self, incremental_state: Dict[str, Dict[str, Optional[Tensor]]], new_order ): """Reorder buffered internal state (for incremental generation).""" input_buffer = self._get_input_buffer(incremental_state) if input_buffer is not None: for k in input_buffer.keys(): if input_buffer[k] is not None: input_buffer[k] = input_buffer[k].index_select(0, new_order) incremental_state = self._set_input_buffer(incremental_state, input_buffer) return incremental_state def _get_input_buffer( self, incremental_state: Optional[Dict[str, Dict[str, Optional[Tensor]]]] ) -> Dict[str, Optional[Tensor]]: result = self.get_incremental_state(incremental_state, "attn_state") if result is not None: return result else: empty_result: Dict[str, Optional[Tensor]] = {} return empty_result def _set_input_buffer( self, incremental_state: Dict[str, Dict[str, Optional[Tensor]]], buffer: Dict[str, Optional[Tensor]], ): return self.set_incremental_state(incremental_state, "attn_state", buffer)

COCO-LM/fairseq/fairseq/model_parallel/modules/multihead_attention.py/0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. import logging import math import os import torch import torch.nn as nn import torch.nn.functional as F from fairseq import checkpoint_utils from fairseq.incremental_decoding_utils import with_incremental_state from fairseq.models import ( CompositeEncoder, FairseqDecoder, FairseqEncoder, FairseqEncoderDecoderModel, register_model, register_model_architecture, ) from fairseq.modules import ( DownsampledMultiHeadAttention, FairseqDropout, GradMultiply, LayerNorm, LearnedPositionalEmbedding, LinearizedConvolution, ) logger = logging.getLogger(__name__) @register_model("fconv_self_att") class FConvModelSelfAtt(FairseqEncoderDecoderModel): @classmethod def hub_models(cls): return { "conv.stories.pretrained": { "path": "https://dl.fbaipublicfiles.com/fairseq/models/stories_checkpoint.tar.gz", "checkpoint_file": "pretrained_checkpoint.pt", "tokenizer": "nltk", }, "conv.stories": { "path": "https://dl.fbaipublicfiles.com/fairseq/models/stories_checkpoint.tar.gz", "checkpoint_file": "fusion_checkpoint.pt", "tokenizer": "nltk", "pretrained": "True", "pretrained_checkpoint": "./pretrained_checkpoint.pt", }, # Test set containing dictionaries "data.stories": "https://dl.fbaipublicfiles.com/fairseq/data/stories_test.tar.bz2", } def __init__(self, encoder, decoder, pretrained_encoder=None): super().__init__(encoder, decoder) self.encoder.num_attention_layers = sum( layer is not None for layer in decoder.attention ) self.pretrained_encoder = pretrained_encoder if self.pretrained_encoder is None: encoders = {"encoder": encoder} else: encoders = {"encoder": encoder, "pretrained": self.pretrained_encoder} # for fusion model, CompositeEncoder contains both pretrained and training encoders # these are forwarded and then combined in the decoder self.encoder = CompositeEncoder(encoders) @staticmethod def add_args(parser): """Add model-specific arguments to the parser.""" # fmt: off parser.add_argument('--dropout', type=float, metavar='D', help='dropout probability') parser.add_argument('--encoder-embed-dim', type=int, metavar='N', help='encoder embedding dimension') parser.add_argument('--encoder-layers', type=str, metavar='EXPR', help='encoder layers [(dim, kernel_size), ...]') parser.add_argument('--decoder-embed-dim', type=int, metavar='N', help='decoder embedding dimension') parser.add_argument('--decoder-layers', type=str, metavar='EXPR', help='decoder layers [(dim, kernel_size), ...]') parser.add_argument('--decoder-out-embed-dim', type=int, metavar='N', help='decoder output embedding dimension') parser.add_argument('--decoder-attention', type=str, metavar='EXPR', help='decoder attention [True, ...]') parser.add_argument('--self-attention', type=str, metavar='EXPR', help='decoder self-attention layers, ex: [True] + [False]*5') parser.add_argument('--multihead-attention-nheads', type=int, help='Number of heads to use in attention') parser.add_argument('--multihead-self-attention-nheads', type=int, help='Number of heads to use in self-attention') parser.add_argument('--encoder-attention', type=str, metavar='EXPR', help='encoder attention [True, ...]') parser.add_argument('--encoder-attention-nheads', type=int, help='Number of heads to use in encoder attention') parser.add_argument('--project-input', type=str, metavar='EXPR', help='Use projections in self-attention [True, ...]') parser.add_argument('--gated-attention', type=str, metavar='EXPR', help='Use GLU layers in self-attention projections [True, ...]') parser.add_argument('--downsample', type=str, metavar='EXPR', help='Use downsampling in self-attention [True, ...]') parser.add_argument('--pretrained-checkpoint', metavar='DIR', help='path to load checkpoint from pretrained model') parser.add_argument('--pretrained', type=str, metavar='EXPR', help='use pretrained model when training [True, ...]') # fmt: on @classmethod def build_model(cls, args, task): """Build a new model instance.""" trained_encoder, trained_decoder = None, None pretrained = eval(args.pretrained) if pretrained: logger.info("loading pretrained model") if not os.path.exists(args.pretrained_checkpoint): new_pretrained_checkpoint = os.path.join( args.data, args.pretrained_checkpoint ) if os.path.exists(new_pretrained_checkpoint): args.pretrained_checkpoint = new_pretrained_checkpoint trained_model = checkpoint_utils.load_model_ensemble( filenames=[args.pretrained_checkpoint], task=task, )[0][0] trained_decoder = list(trained_model.children())[1] trained_encoder = list(trained_model.children())[0] # freeze pretrained model for param in trained_decoder.parameters(): param.requires_grad = False for param in trained_encoder.parameters(): param.requires_grad = False encoder = FConvEncoder( task.source_dictionary, embed_dim=args.encoder_embed_dim, convolutions=eval(args.encoder_layers), dropout=args.dropout, max_positions=args.max_source_positions, attention=eval(args.encoder_attention), attention_nheads=args.encoder_attention_nheads, ) decoder = FConvDecoder( task.target_dictionary, embed_dim=args.decoder_embed_dim, convolutions=eval(args.decoder_layers), out_embed_dim=args.decoder_out_embed_dim, attention=eval(args.decoder_attention), dropout=args.dropout, max_positions=args.max_target_positions, selfattention=eval(args.self_attention), attention_nheads=args.multihead_attention_nheads, selfattention_nheads=args.multihead_self_attention_nheads, project_input=eval(args.project_input), gated_attention=eval(args.gated_attention), downsample=eval(args.downsample), pretrained=pretrained, trained_decoder=trained_decoder, ) model = FConvModelSelfAtt(encoder, decoder, trained_encoder) return model @property def pretrained(self): return self.pretrained_encoder is not None class FConvEncoder(FairseqEncoder): """Convolutional encoder""" def __init__( self, dictionary, embed_dim=512, max_positions=1024, convolutions=((512, 3),) * 20, dropout=0.1, attention=False, attention_nheads=1, ): super().__init__(dictionary) self.dropout_module = FairseqDropout( dropout, module_name=self.__class__.__name__ ) self.num_attention_layers = None num_embeddings = len(dictionary) self.padding_idx = dictionary.pad() self.embed_tokens = Embedding(num_embeddings, embed_dim, self.padding_idx) self.embed_positions = PositionalEmbedding( max_positions, embed_dim, self.padding_idx, ) def expand_bool_array(val): if isinstance(val, bool): # expand True into [True, True, ...] and do the same with False return [val] * len(convolutions) return val attention = expand_bool_array(attention) in_channels = convolutions[0][0] self.fc1 = Linear(embed_dim, in_channels, dropout=dropout) self.projections = nn.ModuleList() self.convolutions = nn.ModuleList() self.attention = nn.ModuleList() self.attproj = nn.ModuleList() for i, (out_channels, kernel_size) in enumerate(convolutions): self.projections.append( Linear(in_channels, out_channels) if in_channels != out_channels else None ) self.convolutions.append( ConvTBC(in_channels, out_channels * 2, kernel_size, dropout=dropout) ) self.attention.append( SelfAttention(out_channels, embed_dim, attention_nheads) if attention[i] else None ) in_channels = out_channels self.fc2 = Linear(in_channels, embed_dim) def forward(self, src_tokens, src_lengths): # embed tokens and positions x = self.embed_tokens(src_tokens) + self.embed_positions(src_tokens) x = self.dropout_module(x) input_embedding = x.transpose(0, 1) # project to size of convolution x = self.fc1(x) encoder_padding_mask = src_tokens.eq(self.padding_idx).t() # -> T x B if not encoder_padding_mask.any(): encoder_padding_mask = None # B x T x C -> T x B x C x = x.transpose(0, 1) # temporal convolutions for proj, conv, attention in zip( self.projections, self.convolutions, self.attention ): residual = x if proj is None else proj(x) if encoder_padding_mask is not None: x = x.masked_fill(encoder_padding_mask.unsqueeze(-1), 0) x = self.dropout_module(x) padding_l = (conv.kernel_size[0] - 1) // 2 padding_r = conv.kernel_size[0] // 2 x = F.pad(x, (0, 0, 0, 0, padding_l, padding_r)) x = conv(x) x = F.glu(x, dim=2) if attention is not None: x = attention(x) x = (x + residual) * math.sqrt(0.5) # T x B x C -> B x T x C x = x.transpose(1, 0) # project back to size of embedding x = self.fc2(x) if encoder_padding_mask is not None: encoder_padding_mask = encoder_padding_mask.t() # -> B x T x = x.masked_fill(encoder_padding_mask.unsqueeze(-1), 0) # scale gradients (this only affects backward, not forward) x = GradMultiply.apply(x, 1.0 / (2.0 * self.num_attention_layers)) # add output to input embedding for attention y = (x + input_embedding.transpose(0, 1)) * math.sqrt(0.5) return { "encoder_out": (x, y), "encoder_padding_mask": encoder_padding_mask, # B x T } def reorder_encoder_out(self, encoder_out, new_order): encoder_out["encoder_out"] = tuple( eo.index_select(0, new_order) for eo in encoder_out["encoder_out"] ) if encoder_out["encoder_padding_mask"] is not None: encoder_out["encoder_padding_mask"] = encoder_out[ "encoder_padding_mask" ].index_select(0, new_order) if "pretrained" in encoder_out: encoder_out["pretrained"]["encoder_out"] = tuple( eo.index_select(0, new_order) for eo in encoder_out["pretrained"]["encoder_out"] ) return encoder_out def max_positions(self): """Maximum input length supported by the encoder.""" return self.embed_positions.max_positions @with_incremental_state class FConvDecoder(FairseqDecoder): """Convolutional decoder""" def __init__( self, dictionary, embed_dim=512, out_embed_dim=256, max_positions=1024, convolutions=((512, 3),) * 8, attention=True, dropout=0.1, selfattention=False, attention_nheads=1, selfattention_nheads=1, project_input=False, gated_attention=False, downsample=False, pretrained=False, trained_decoder=None, ): super().__init__(dictionary) self.register_buffer("version", torch.Tensor([2])) self.pretrained = pretrained self.pretrained_decoder = trained_decoder self.dropout_module = FairseqDropout( dropout, module_name=self.__class__.__name__ ) self.need_attn = True in_channels = convolutions[0][0] def expand_bool_array(val): if isinstance(val, bool): # expand True into [True, True, ...] and do the same with False return [val] * len(convolutions) return val attention = expand_bool_array(attention) selfattention = expand_bool_array(selfattention) if not isinstance(attention, list) or len(attention) != len(convolutions): raise ValueError( "Attention is expected to be a list of booleans of " "length equal to the number of layers." ) num_embeddings = len(dictionary) padding_idx = dictionary.pad() self.embed_tokens = Embedding(num_embeddings, embed_dim, padding_idx) self.embed_positions = PositionalEmbedding( max_positions, embed_dim, padding_idx, ) self.fc1 = Linear(embed_dim, in_channels, dropout=dropout) self.projections = nn.ModuleList() self.convolutions = nn.ModuleList() self.attention = nn.ModuleList() self.selfattention = nn.ModuleList() self.attproj = nn.ModuleList() for i, (out_channels, kernel_size) in enumerate(convolutions): self.projections.append( Linear(in_channels, out_channels) if in_channels != out_channels else None ) self.convolutions.append( LinearizedConv1d( in_channels, out_channels * 2, kernel_size, padding=(kernel_size - 1), dropout=dropout, ) ) self.attention.append( DownsampledMultiHeadAttention( out_channels, embed_dim, attention_nheads, project_input=project_input, gated=False, downsample=False, ) if attention[i] else None ) self.attproj.append( Linear(out_channels, embed_dim, dropout=dropout) if attention[i] else None ) self.selfattention.append( SelfAttention( out_channels, embed_dim, selfattention_nheads, project_input=project_input, gated=gated_attention, downsample=downsample, ) if selfattention[i] else None ) in_channels = out_channels self.fc2 = Linear(in_channels, out_embed_dim) self.fc3 = Linear(out_embed_dim, num_embeddings, dropout=dropout) # model fusion if self.pretrained: # independent gates are learned from the concatenated input self.gate1 = nn.Sequential( Linear(out_embed_dim * 2, out_embed_dim), nn.Sigmoid() ) self.gate2 = nn.Sequential( Linear(out_embed_dim * 2, out_embed_dim), nn.Sigmoid() ) # pretrained and trained models are joined self.joining = nn.Sequential( Linear(out_embed_dim * 2, out_embed_dim * 2), LayerNorm(out_embed_dim * 2), nn.GLU(), Linear(out_embed_dim, out_embed_dim * 2), LayerNorm(out_embed_dim * 2), nn.GLU(), Linear(out_embed_dim, out_embed_dim), LayerNorm(out_embed_dim), ) # pretrained model contains an output layer that is nhid -> vocab size # but the models are combined in their hidden state # the hook stores the output of the pretrained model forward self.pretrained_outputs = {} def save_output(): def hook(a, b, output): self.pretrained_outputs["out"] = output return hook self.pretrained_decoder.fc2.register_forward_hook(save_output()) def forward(self, prev_output_tokens, encoder_out): trained_encoder_out = encoder_out["pretrained"] if self.pretrained else None encoder_out = encoder_out["encoder"]["encoder_out"] encoder_a, encoder_b = self._split_encoder_out(encoder_out) # embed positions positions = self.embed_positions(prev_output_tokens) # embed tokens and positions x = self.embed_tokens(prev_output_tokens) + positions x = self.dropout_module(x) target_embedding = x.transpose(0, 1) # project to size of convolution x = self.fc1(x) # B x T x C -> T x B x C x = x.transpose(0, 1) # temporal convolutions avg_attn_scores = None for proj, conv, attention, selfattention, attproj in zip( self.projections, self.convolutions, self.attention, self.selfattention, self.attproj, ): residual = x if proj is None else proj(x) x = self.dropout_module(x) x = conv(x) x = F.glu(x, dim=2) # attention if attention is not None: r = x x, attn_scores = attention( attproj(x) + target_embedding, encoder_a, encoder_b ) x = x + r if not self.training and self.need_attn: if avg_attn_scores is None: avg_attn_scores = attn_scores else: avg_attn_scores.add_(attn_scores) if selfattention is not None: x = selfattention(x) x = (x + residual) * math.sqrt(0.5) # T x B x C -> B x T x C x = x.transpose(0, 1) # project back to size of vocabulary x = self.fc2(x) x = self.dropout_module(x) if not self.pretrained: x = self.fc3(x) # fusion gating if self.pretrained: trained_x, _ = self.pretrained_decoder.forward( prev_output_tokens, trained_encoder_out ) y = torch.cat([x, self.pretrained_outputs["out"]], dim=-1) gate1 = self.gate1(y) gate2 = self.gate2(y) gated_x1 = gate1 * x gated_x2 = gate2 * self.pretrained_outputs["out"] fusion = torch.cat([gated_x1, gated_x2], dim=-1) fusion = self.joining(fusion) fusion_output = self.fc3(fusion) return fusion_output, avg_attn_scores else: return x, avg_attn_scores def max_positions(self): """Maximum output length supported by the decoder.""" return self.embed_positions.max_positions def make_generation_fast_(self, need_attn=False, **kwargs): self.need_attn = need_attn def _split_encoder_out(self, encoder_out): """Split and transpose encoder outputs.""" # transpose only once to speed up attention layers encoder_a, encoder_b = encoder_out encoder_a = encoder_a.transpose(0, 1).contiguous() encoder_b = encoder_b.transpose(0, 1).contiguous() result = (encoder_a, encoder_b) return result class SelfAttention(nn.Module): def __init__( self, out_channels, embed_dim, num_heads, project_input=False, gated=False, downsample=False, ): super().__init__() self.attention = DownsampledMultiHeadAttention( out_channels, embed_dim, num_heads, dropout=0, bias=True, project_input=project_input, gated=gated, downsample=downsample, ) self.in_proj_q = Linear(out_channels, embed_dim) self.in_proj_k = Linear(out_channels, embed_dim) self.in_proj_v = Linear(out_channels, embed_dim) self.ln = LayerNorm(out_channels) def forward(self, x): residual = x query = self.in_proj_q(x) key = self.in_proj_k(x) value = self.in_proj_v(x) x, _ = self.attention( query, key, value, mask_future_timesteps=True, use_scalar_bias=True ) return self.ln(x + residual) def Embedding(num_embeddings, embedding_dim, padding_idx): m = nn.Embedding(num_embeddings, embedding_dim, padding_idx=padding_idx) m.weight.data.normal_(0, 0.1) return m def PositionalEmbedding(num_embeddings, embedding_dim, padding_idx): m = LearnedPositionalEmbedding(num_embeddings, embedding_dim, padding_idx) m.weight.data.normal_(0, 0.1) return m def Linear(in_features, out_features, dropout=0.0): """Weight-normalized Linear layer (input: N x T x C)""" m = nn.Linear(in_features, out_features) m.weight.data.normal_(mean=0, std=math.sqrt((1 - dropout) / in_features)) m.bias.data.zero_() return m def LinearizedConv1d(in_channels, out_channels, kernel_size, dropout=0.0, **kwargs): """Weight-normalized Conv1d layer optimized for decoding""" m = LinearizedConvolution(in_channels, out_channels, kernel_size, **kwargs) std = math.sqrt((4 * (1.0 - dropout)) / (m.kernel_size[0] * in_channels)) m.weight.data.normal_(mean=0, std=std) m.bias.data.zero_() return m def ConvTBC(in_channels, out_channels, kernel_size, dropout=0.0, **kwargs): """Weight-normalized Conv1d layer""" from fairseq.modules import ConvTBC m = ConvTBC(in_channels, out_channels, kernel_size, **kwargs) std = math.sqrt((4 * (1.0 - dropout)) / (m.kernel_size[0] * in_channels)) m.weight.data.normal_(mean=0, std=std) m.bias.data.zero_() return m @register_model_architecture("fconv_self_att", "fconv_self_att") def base_architecture(args): args.dropout = getattr(args, "dropout", 0.1) args.encoder_embed_dim = getattr(args, "encoder_embed_dim", 512) args.encoder_layers = getattr(args, "encoder_layers", "[(512, 3)] * 3") args.decoder_embed_dim = getattr(args, "decoder_embed_dim", 512) args.decoder_layers = getattr(args, "decoder_layers", "[(512, 3)] * 8") args.decoder_out_embed_dim = getattr(args, "decoder_out_embed_dim", 256) args.decoder_attention = getattr(args, "decoder_attention", "True") args.self_attention = getattr(args, "self_attention", "False") args.encoder_attention = getattr(args, "encoder_attention", "False") args.multihead_attention_nheads = getattr(args, "multihead_attention_nheads", 1) args.multihead_self_attention_nheads = getattr( args, "multihead_self_attention_nheads", 1 ) args.encoder_attention_nheads = getattr(args, "encoder_attention_nheads", 1) args.project_input = getattr(args, "project_input", "False") args.gated_attention = getattr(args, "gated_attention", "False") args.downsample = getattr(args, "downsample", "False") args.pretrained_checkpoint = getattr(args, "pretrained_checkpoint", "") args.pretrained = getattr(args, "pretrained", "False") @register_model_architecture("fconv_self_att", "fconv_self_att_wp") def fconv_self_att_wp(args): args.encoder_embed_dim = getattr(args, "encoder_embed_dim", 256) args.encoder_layers = getattr( args, "encoder_layers", "[(128, 3)] * 2 + [(512,3)] * 1" ) args.decoder_embed_dim = getattr(args, "decoder_embed_dim", 256) args.decoder_layers = getattr( args, "decoder_layers", "[(512, 4)] * 4 + [(768, 4)] * 2 + [(1024, 4)] * 1" ) args.decoder_out_embed_dim = getattr(args, "decoder_out_embed_dim", 256) args.self_attention = getattr(args, "self_attention", "True") args.multihead_self_attention_nheads = getattr( args, "multihead_self_attention_nheads", 4 ) args.project_input = getattr(args, "project_input", "True") args.gated_attention = getattr(args, "gated_attention", "True") args.downsample = getattr(args, "downsample", "True") base_architecture(args)

COCO-LM/fairseq/fairseq/models/fconv_self_att.py/0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. import torch from fairseq.utils import new_arange # -------------- Helper Functions --------------------------------------------------- # def load_libnat(): try: from fairseq import libnat_cuda return libnat_cuda, True except ImportError as e: print(str(e) + "... fall back to CPU version") try: from fairseq import libnat return libnat, False except ImportError as e: import sys sys.stderr.write( "ERROR: missing libnat_cuda. run `python setup.py build_ext --inplace`\n" ) raise e def _get_ins_targets(in_tokens, out_tokens, padding_idx, unk_idx): libnat, use_cuda = load_libnat() def _get_ins_targets_cuda(in_tokens, out_tokens, padding_idx, unk_idx): in_masks = in_tokens.ne(padding_idx) out_masks = out_tokens.ne(padding_idx) mask_ins_targets, masked_tgt_masks = libnat.generate_insertion_labels( out_tokens.int(), libnat.levenshtein_distance( in_tokens.int(), out_tokens.int(), in_masks.sum(1).int(), out_masks.sum(1).int(), ), ) masked_tgt_masks = masked_tgt_masks.bool() & out_masks mask_ins_targets = mask_ins_targets.type_as(in_tokens)[ :, 1 : in_masks.size(1) ].masked_fill_(~in_masks[:, 1:], 0) masked_tgt_tokens = out_tokens.masked_fill(masked_tgt_masks, unk_idx) return masked_tgt_masks, masked_tgt_tokens, mask_ins_targets def _get_ins_targets_cpu(in_tokens, out_tokens, padding_idx, unk_idx): in_seq_len, out_seq_len = in_tokens.size(1), out_tokens.size(1) in_tokens_list = [ [t for t in s if t != padding_idx] for i, s in enumerate(in_tokens.tolist()) ] out_tokens_list = [ [t for t in s if t != padding_idx] for i, s in enumerate(out_tokens.tolist()) ] full_labels = libnat.suggested_ed2_path( in_tokens_list, out_tokens_list, padding_idx ) mask_inputs = [ [len(c) if c[0] != padding_idx else 0 for c in a[:-1]] for a in full_labels ] # generate labels masked_tgt_masks = [] for mask_input in mask_inputs: mask_label = [] for beam_size in mask_input[1:-1]: # HACK 1:-1 mask_label += [0] + [1 for _ in range(beam_size)] masked_tgt_masks.append( mask_label + [0 for _ in range(out_seq_len - len(mask_label))] ) mask_ins_targets = [ mask_input[1:-1] + [0 for _ in range(in_seq_len - 1 - len(mask_input[1:-1]))] for mask_input in mask_inputs ] # transform to tensor masked_tgt_masks = torch.tensor( masked_tgt_masks, device=out_tokens.device ).bool() mask_ins_targets = torch.tensor(mask_ins_targets, device=in_tokens.device) masked_tgt_tokens = out_tokens.masked_fill(masked_tgt_masks, unk_idx) return masked_tgt_masks, masked_tgt_tokens, mask_ins_targets if use_cuda: return _get_ins_targets_cuda(in_tokens, out_tokens, padding_idx, unk_idx) return _get_ins_targets_cpu(in_tokens, out_tokens, padding_idx, unk_idx) def _get_del_targets(in_tokens, out_tokens, padding_idx): libnat, use_cuda = load_libnat() def _get_del_targets_cuda(in_tokens, out_tokens, padding_idx): in_masks = in_tokens.ne(padding_idx) out_masks = out_tokens.ne(padding_idx) word_del_targets = libnat.generate_deletion_labels( in_tokens.int(), libnat.levenshtein_distance( in_tokens.int(), out_tokens.int(), in_masks.sum(1).int(), out_masks.sum(1).int(), ), ) word_del_targets = word_del_targets.type_as(in_tokens).masked_fill_( ~in_masks, 0 ) return word_del_targets def _get_del_targets_cpu(in_tokens, out_tokens, padding_idx): out_seq_len = out_tokens.size(1) with torch.cuda.device_of(in_tokens): in_tokens_list = [ [t for t in s if t != padding_idx] for i, s in enumerate(in_tokens.tolist()) ] out_tokens_list = [ [t for t in s if t != padding_idx] for i, s in enumerate(out_tokens.tolist()) ] full_labels = libnat.suggested_ed2_path( in_tokens_list, out_tokens_list, padding_idx ) word_del_targets = [b[-1] for b in full_labels] word_del_targets = [ labels + [0 for _ in range(out_seq_len - len(labels))] for labels in word_del_targets ] # transform to tensor word_del_targets = torch.tensor(word_del_targets, device=out_tokens.device) return word_del_targets if use_cuda: return _get_del_targets_cuda(in_tokens, out_tokens, padding_idx) return _get_del_targets_cpu(in_tokens, out_tokens, padding_idx) def _apply_ins_masks( in_tokens, in_scores, mask_ins_pred, padding_idx, unk_idx, eos_idx ): in_masks = in_tokens.ne(padding_idx) in_lengths = in_masks.sum(1) # HACK: hacky way to shift all the paddings to eos first. in_tokens.masked_fill_(~in_masks, eos_idx) mask_ins_pred.masked_fill_(~in_masks[:, 1:], 0) out_lengths = in_lengths + mask_ins_pred.sum(1) out_max_len = out_lengths.max() out_masks = new_arange(out_lengths, out_max_len)[None, :] < out_lengths[:, None] reordering = (mask_ins_pred + in_masks[:, 1:].long()).cumsum(1) out_tokens = ( in_tokens.new_zeros(in_tokens.size(0), out_max_len) .fill_(padding_idx) .masked_fill_(out_masks, unk_idx) ) out_tokens[:, 0] = in_tokens[:, 0] out_tokens.scatter_(1, reordering, in_tokens[:, 1:]) out_scores = None if in_scores is not None: in_scores.masked_fill_(~in_masks, 0) out_scores = in_scores.new_zeros(*out_tokens.size()) out_scores[:, 0] = in_scores[:, 0] out_scores.scatter_(1, reordering, in_scores[:, 1:]) return out_tokens, out_scores def _apply_ins_words(in_tokens, in_scores, word_ins_pred, word_ins_scores, unk_idx): word_ins_masks = in_tokens.eq(unk_idx) out_tokens = in_tokens.masked_scatter(word_ins_masks, word_ins_pred[word_ins_masks]) if in_scores is not None: out_scores = in_scores.masked_scatter( word_ins_masks, word_ins_scores[word_ins_masks] ) else: out_scores = None return out_tokens, out_scores def _apply_del_words( in_tokens, in_scores, in_attn, word_del_pred, padding_idx, bos_idx, eos_idx ): # apply deletion to a tensor in_masks = in_tokens.ne(padding_idx) bos_eos_masks = in_tokens.eq(bos_idx) | in_tokens.eq(eos_idx) max_len = in_tokens.size(1) word_del_pred.masked_fill_(~in_masks, 1) word_del_pred.masked_fill_(bos_eos_masks, 0) reordering = new_arange(in_tokens).masked_fill_(word_del_pred, max_len).sort(1)[1] out_tokens = in_tokens.masked_fill(word_del_pred, padding_idx).gather(1, reordering) out_scores = None if in_scores is not None: out_scores = in_scores.masked_fill(word_del_pred, 0).gather(1, reordering) out_attn = None if in_attn is not None: _mask = word_del_pred[:, :, None].expand_as(in_attn) _reordering = reordering[:, :, None].expand_as(in_attn) out_attn = in_attn.masked_fill(_mask, 0.0).gather(1, _reordering) return out_tokens, out_scores, out_attn def _skip(x, mask): """ Getting sliced (dim=0) tensor by mask. Supporting tensor and list/dict of tensors. """ if isinstance(x, int): return x if x is None: return None if isinstance(x, torch.Tensor): if x.size(0) == mask.size(0): return x[mask] elif x.size(1) == mask.size(0): return x[:, mask] if isinstance(x, list): return [_skip(x_i, mask) for x_i in x] if isinstance(x, dict): return {k: _skip(v, mask) for k, v in x.items()} raise NotImplementedError def _skip_encoder_out(encoder, encoder_out, mask): if not mask.any(): return encoder_out else: return encoder.reorder_encoder_out( encoder_out, mask.nonzero(as_tuple=False).squeeze() ) def _fill(x, mask, y, padding_idx): """ Filling tensor x with y at masked positions (dim=0). """ if x is None: return y assert x.dim() == y.dim() and mask.size(0) == x.size(0) assert x.dim() == 2 or (x.dim() == 3 and x.size(2) == y.size(2)) n_selected = mask.sum() assert n_selected == y.size(0) if n_selected == x.size(0): return y if x.size(1) < y.size(1): dims = [x.size(0), y.size(1) - x.size(1)] if x.dim() == 3: dims.append(x.size(2)) x = torch.cat([x, x.new_zeros(*dims).fill_(padding_idx)], 1) x[mask] = y elif x.size(1) > y.size(1): x[mask] = padding_idx if x.dim() == 2: x[mask, : y.size(1)] = y else: x[mask, : y.size(1), :] = y else: x[mask] = y return x

COCO-LM/fairseq/fairseq/models/nat/levenshtein_utils.py/0

#!/usr/bin/env python3 import logging import math from typing import Dict, List, Optional, Tuple import torch.nn as nn from fairseq import checkpoint_utils, utils from fairseq.data.data_utils import lengths_to_padding_mask from fairseq.models import ( FairseqEncoder, FairseqEncoderDecoderModel, register_model, register_model_architecture, ) from fairseq.models.transformer import Embedding, TransformerDecoder from fairseq.modules import ( FairseqDropout, LayerNorm, PositionalEmbedding, TransformerEncoderLayer, ) from torch import Tensor logger = logging.getLogger(__name__) class Conv1dSubsampler(nn.Module): """Convolutional subsampler: a stack of 1D convolution (along temporal dimension) followed by non-linear activation via gated linear units (https://arxiv.org/abs/1911.08460) Args: in_channels (int): the number of input channels mid_channels (int): the number of intermediate channels out_channels (int): the number of output channels kernel_sizes (List[int]): the kernel size for each convolutional layer """ def __init__( self, in_channels: int, mid_channels: int, out_channels: int, kernel_sizes: List[int] = (3, 3), ): super(Conv1dSubsampler, self).__init__() self.n_layers = len(kernel_sizes) self.conv_layers = nn.ModuleList( nn.Conv1d( in_channels if i == 0 else mid_channels // 2, mid_channels if i < self.n_layers - 1 else out_channels * 2, k, stride=2, padding=k // 2, ) for i, k in enumerate(kernel_sizes) ) def get_out_seq_lens_tensor(self, in_seq_lens_tensor): out = in_seq_lens_tensor.clone() for _ in range(self.n_layers): out = ((out.float() - 1) / 2 + 1).floor().long() return out def forward(self, src_tokens, src_lengths): bsz, in_seq_len, _ = src_tokens.size() # B x T x (C x D) x = src_tokens.transpose(1, 2).contiguous() # -> B x (C x D) x T for conv in self.conv_layers: x = conv(x) x = nn.functional.glu(x, dim=1) _, _, out_seq_len = x.size() x = x.transpose(1, 2).transpose(0, 1).contiguous() # -> T x B x (C x D) return x, self.get_out_seq_lens_tensor(src_lengths) @register_model("s2t_transformer") class S2TTransformerModel(FairseqEncoderDecoderModel): """Adapted Transformer model (https://arxiv.org/abs/1706.03762) for speech-to-text tasks. The Transformer encoder/decoder remains the same. A trainable input subsampler is prepended to the Transformer encoder to project inputs into the encoder dimension as well as downsample input sequence for computational efficiency.""" def __init__(self, encoder, decoder): super().__init__(encoder, decoder) @staticmethod def add_args(parser): """Add model-specific arguments to the parser.""" # input parser.add_argument( "--conv-kernel-sizes", type=str, metavar="N", help="kernel sizes of Conv1d subsampling layers", ) parser.add_argument( "--conv-channels", type=int, metavar="N", help="# of channels in Conv1d subsampling layers", ) # Transformer parser.add_argument( "--activation-fn", type=str, default="relu", choices=utils.get_available_activation_fns(), help="activation function to use", ) parser.add_argument( "--dropout", type=float, metavar="D", help="dropout probability" ) parser.add_argument( "--attention-dropout", type=float, metavar="D", help="dropout probability for attention weights", ) parser.add_argument( "--activation-dropout", "--relu-dropout", type=float, metavar="D", help="dropout probability after activation in FFN.", ) parser.add_argument( "--encoder-embed-dim", type=int, metavar="N", help="encoder embedding dimension", ) parser.add_argument( "--encoder-ffn-embed-dim", type=int, metavar="N", help="encoder embedding dimension for FFN", ) parser.add_argument( "--encoder-layers", type=int, metavar="N", help="num encoder layers" ) parser.add_argument( "--encoder-attention-heads", type=int, metavar="N", help="num encoder attention heads", ) parser.add_argument( "--encoder-normalize-before", action="store_true", help="apply layernorm before each encoder block", ) parser.add_argument( "--decoder-embed-dim", type=int, metavar="N", help="decoder embedding dimension", ) parser.add_argument( "--decoder-ffn-embed-dim", type=int, metavar="N", help="decoder embedding dimension for FFN", ) parser.add_argument( "--decoder-layers", type=int, metavar="N", help="num decoder layers" ) parser.add_argument( "--decoder-attention-heads", type=int, metavar="N", help="num decoder attention heads", ) parser.add_argument( "--decoder-normalize-before", action="store_true", help="apply layernorm before each decoder block", ) parser.add_argument( "--share-decoder-input-output-embed", action="store_true", help="share decoder input and output embeddings", ) parser.add_argument( "--layernorm-embedding", action="store_true", help="add layernorm to embedding", ) parser.add_argument( "--no-scale-embedding", action="store_true", help="if True, dont scale embeddings", ) parser.add_argument( "--load-pretrained-encoder-from", type=str, metavar="STR", help="model to take encoder weights from (for initialization)", ) @classmethod def build_encoder(cls, args): encoder = S2TTransformerEncoder(args) if getattr(args, "load_pretrained_encoder_from", None): encoder = checkpoint_utils.load_pretrained_component_from_model( component=encoder, checkpoint=args.load_pretrained_encoder_from ) logger.info( f"loaded pretrained encoder from: " f"{args.load_pretrained_encoder_from}" ) return encoder @classmethod def build_decoder(cls, args, task, embed_tokens): return TransformerDecoderScriptable(args, task.target_dictionary, embed_tokens) @classmethod def build_model(cls, args, task): """Build a new model instance.""" # make sure all arguments are present in older models base_architecture(args) def build_embedding(dictionary, embed_dim): num_embeddings = len(dictionary) padding_idx = dictionary.pad() return Embedding(num_embeddings, embed_dim, padding_idx) decoder_embed_tokens = build_embedding( task.target_dictionary, args.decoder_embed_dim ) encoder = cls.build_encoder(args) decoder = cls.build_decoder(args, task, decoder_embed_tokens) return cls(encoder, decoder) def get_normalized_probs( self, net_output: Tuple[Tensor, Optional[Dict[str, List[Optional[Tensor]]]]], log_probs: bool, sample: Optional[Dict[str, Tensor]] = None, ): # net_output['encoder_out'] is a (B, T, D) tensor lprobs = self.get_normalized_probs_scriptable(net_output, log_probs, sample) lprobs.batch_first = True return lprobs def forward(self, src_tokens, src_lengths, prev_output_tokens): """ The forward method inherited from the base class has a **kwargs argument in its input, which is not supported in torchscript. This method overwrites the forward method definition without **kwargs. """ encoder_out = self.encoder(src_tokens=src_tokens, src_lengths=src_lengths) decoder_out = self.decoder( prev_output_tokens=prev_output_tokens, encoder_out=encoder_out ) return decoder_out class S2TTransformerEncoder(FairseqEncoder): """Speech-to-text Transformer encoder that consists of input subsampler and Transformer encoder.""" def __init__(self, args): super().__init__(None) self.dropout_module = FairseqDropout( p=args.dropout, module_name=self.__class__.__name__ ) self.embed_scale = math.sqrt(args.encoder_embed_dim) if args.no_scale_embedding: self.embed_scale = 1.0 self.padding_idx = 1 self.subsample = Conv1dSubsampler( args.input_feat_per_channel * args.input_channels, args.conv_channels, args.encoder_embed_dim, [int(k) for k in args.conv_kernel_sizes.split(",")], ) self.embed_positions = PositionalEmbedding( args.max_source_positions, args.encoder_embed_dim, self.padding_idx ) self.transformer_layers = nn.ModuleList( [TransformerEncoderLayer(args) for _ in range(args.encoder_layers)] ) if args.encoder_normalize_before: self.layer_norm = LayerNorm(args.encoder_embed_dim) else: self.layer_norm = None def forward(self, src_tokens, src_lengths): x, input_lengths = self.subsample(src_tokens, src_lengths) x = self.embed_scale * x encoder_padding_mask = lengths_to_padding_mask(input_lengths) positions = self.embed_positions(encoder_padding_mask).transpose(0, 1) x += positions x = self.dropout_module(x) for layer in self.transformer_layers: x = layer(x, encoder_padding_mask) if self.layer_norm is not None: x = self.layer_norm(x) return { "encoder_out": [x], # T x B x C "encoder_padding_mask": [encoder_padding_mask] if encoder_padding_mask.any() else [], # B x T "encoder_embedding": [], # B x T x C "encoder_states": [], # List[T x B x C] "src_tokens": [], "src_lengths": [], } def reorder_encoder_out(self, encoder_out, new_order): new_encoder_out = ( [] if len(encoder_out["encoder_out"]) == 0 else [x.index_select(1, new_order) for x in encoder_out["encoder_out"]] ) new_encoder_padding_mask = ( [] if len(encoder_out["encoder_padding_mask"]) == 0 else [x.index_select(0, new_order) for x in encoder_out["encoder_padding_mask"]] ) new_encoder_embedding = ( [] if len(encoder_out["encoder_embedding"]) == 0 else [x.index_select(0, new_order) for x in encoder_out["encoder_embedding"]] ) encoder_states = encoder_out["encoder_states"] if len(encoder_states) > 0: for idx, state in enumerate(encoder_states): encoder_states[idx] = state.index_select(1, new_order) return { "encoder_out": new_encoder_out, # T x B x C "encoder_padding_mask": new_encoder_padding_mask, # B x T "encoder_embedding": new_encoder_embedding, # B x T x C "encoder_states": encoder_states, # List[T x B x C] "src_tokens": [], # B x T "src_lengths": [], # B x 1 } class TransformerDecoderScriptable(TransformerDecoder): def extract_features( self, prev_output_tokens, encoder_out: Optional[Dict[str, List[Tensor]]] = None, incremental_state: Optional[Dict[str, Dict[str, Optional[Tensor]]]] = None, full_context_alignment: bool = False, alignment_layer: Optional[int] = None, alignment_heads: Optional[int] = None, ): # call scriptable method from parent class x, _ = self.extract_features_scriptable( prev_output_tokens, encoder_out, incremental_state, full_context_alignment, alignment_layer, alignment_heads, ) return x, None @register_model_architecture(model_name="s2t_transformer", arch_name="s2t_transformer") def base_architecture(args): # Convolutional subsampler args.conv_kernel_sizes = getattr(args, "conv_kernel_sizes", "5,5") args.conv_channels = getattr(args, "conv_channels", 1024) # Transformer args.encoder_embed_dim = getattr(args, "encoder_embed_dim", 512) args.encoder_ffn_embed_dim = getattr(args, "encoder_ffn_embed_dim", 2048) args.encoder_layers = getattr(args, "encoder_layers", 12) args.encoder_attention_heads = getattr(args, "encoder_attention_heads", 8) args.encoder_normalize_before = getattr(args, "encoder_normalize_before", True) args.decoder_embed_dim = getattr(args, "decoder_embed_dim", args.encoder_embed_dim) args.decoder_ffn_embed_dim = getattr( args, "decoder_ffn_embed_dim", args.encoder_ffn_embed_dim ) args.decoder_layers = getattr(args, "decoder_layers", 6) args.decoder_attention_heads = getattr(args, "decoder_attention_heads", 8) args.decoder_normalize_before = getattr(args, "decoder_normalize_before", True) args.decoder_learned_pos = getattr(args, "decoder_learned_pos", False) args.dropout = getattr(args, "dropout", 0.1) args.attention_dropout = getattr(args, "attention_dropout", args.dropout) args.activation_dropout = getattr(args, "activation_dropout", args.dropout) args.activation_fn = getattr(args, "activation_fn", "relu") args.adaptive_softmax_cutoff = getattr(args, "adaptive_softmax_cutoff", None) args.adaptive_softmax_dropout = getattr(args, "adaptive_softmax_dropout", 0) args.share_decoder_input_output_embed = getattr( args, "share_decoder_input_output_embed", False ) args.no_token_positional_embeddings = getattr( args, "no_token_positional_embeddings", False ) args.adaptive_input = getattr(args, "adaptive_input", False) args.decoder_layerdrop = getattr(args, "decoder_layerdrop", 0.0) args.decoder_output_dim = getattr( args, "decoder_output_dim", args.decoder_embed_dim ) args.decoder_input_dim = getattr(args, "decoder_input_dim", args.decoder_embed_dim) args.no_scale_embedding = getattr(args, "no_scale_embedding", False) args.quant_noise_pq = getattr(args, "quant_noise_pq", 0) @register_model_architecture("s2t_transformer", "s2t_transformer_s") def s2t_transformer_s(args): args.encoder_embed_dim = getattr(args, "encoder_embed_dim", 256) args.encoder_ffn_embed_dim = getattr(args, "encoder_ffn_embed_dim", 256 * 8) args.encoder_attention_heads = getattr(args, "encoder_attention_heads", 4) args.decoder_attention_heads = getattr(args, "decoder_attention_heads", 4) args.dropout = getattr(args, "dropout", 0.1) base_architecture(args) @register_model_architecture("s2t_transformer", "s2t_transformer_xs") def s2t_transformer_xs(args): args.encoder_layers = getattr(args, "encoder_layers", 6) args.decoder_layers = getattr(args, "decoder_layers", 3) args.encoder_ffn_embed_dim = getattr(args, "encoder_ffn_embed_dim", 256 * 4) args.dropout = getattr(args, "dropout", 0.3) s2t_transformer_s(args) @register_model_architecture("s2t_transformer", "s2t_transformer_sp") def s2t_transformer_sp(args): args.encoder_layers = getattr(args, "encoder_layers", 16) s2t_transformer_s(args) @register_model_architecture("s2t_transformer", "s2t_transformer_m") def s2t_transformer_m(args): args.encoder_embed_dim = getattr(args, "encoder_embed_dim", 512) args.encoder_ffn_embed_dim = getattr(args, "encoder_ffn_embed_dim", 512 * 4) args.encoder_attention_heads = getattr(args, "encoder_attention_heads", 8) args.decoder_attention_heads = getattr(args, "decoder_attention_heads", 8) args.dropout = getattr(args, "dropout", 0.15) base_architecture(args) @register_model_architecture("s2t_transformer", "s2t_transformer_mp") def s2t_transformer_mp(args): args.encoder_layers = getattr(args, "encoder_layers", 16) s2t_transformer_m(args) @register_model_architecture("s2t_transformer", "s2t_transformer_l") def s2t_transformer_l(args): args.encoder_embed_dim = getattr(args, "encoder_embed_dim", 1024) args.encoder_ffn_embed_dim = getattr(args, "encoder_ffn_embed_dim", 1024 * 4) args.encoder_attention_heads = getattr(args, "encoder_attention_heads", 16) args.decoder_attention_heads = getattr(args, "decoder_attention_heads", 16) args.dropout = getattr(args, "dropout", 0.2) base_architecture(args) @register_model_architecture("s2t_transformer", "s2t_transformer_lp") def s2t_transformer_lp(args): args.encoder_layers = getattr(args, "encoder_layers", 16) s2t_transformer_l(args)

COCO-LM/fairseq/fairseq/models/speech_to_text/s2t_transformer.py/0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. import logging from typing import List, Tuple import torch import torch.nn.functional as F from fairseq.data import Dictionary from torch import nn CHAR_PAD_IDX = 0 CHAR_EOS_IDX = 257 logger = logging.getLogger(__name__) class CharacterTokenEmbedder(torch.nn.Module): def __init__( self, vocab: Dictionary, filters: List[Tuple[int, int]], char_embed_dim: int, word_embed_dim: int, highway_layers: int, max_char_len: int = 50, char_inputs: bool = False, ): super(CharacterTokenEmbedder, self).__init__() self.onnx_trace = False self.embedding_dim = word_embed_dim self.max_char_len = max_char_len self.char_embeddings = nn.Embedding(257, char_embed_dim, padding_idx=0) self.symbol_embeddings = nn.Parameter(torch.FloatTensor(2, word_embed_dim)) self.eos_idx, self.unk_idx = 0, 1 self.char_inputs = char_inputs self.convolutions = nn.ModuleList() for width, out_c in filters: self.convolutions.append( nn.Conv1d(char_embed_dim, out_c, kernel_size=width) ) last_dim = sum(f[1] for f in filters) self.highway = Highway(last_dim, highway_layers) if highway_layers > 0 else None self.projection = nn.Linear(last_dim, word_embed_dim) assert ( vocab is not None or char_inputs ), "vocab must be set if not using char inputs" self.vocab = None if vocab is not None: self.set_vocab(vocab, max_char_len) self.reset_parameters() def prepare_for_onnx_export_(self): self.onnx_trace = True def set_vocab(self, vocab, max_char_len): word_to_char = torch.LongTensor(len(vocab), max_char_len) truncated = 0 for i in range(len(vocab)): if i < vocab.nspecial: char_idxs = [0] * max_char_len else: chars = vocab[i].encode() # +1 for padding char_idxs = [c + 1 for c in chars] + [0] * (max_char_len - len(chars)) if len(char_idxs) > max_char_len: truncated += 1 char_idxs = char_idxs[:max_char_len] word_to_char[i] = torch.LongTensor(char_idxs) if truncated > 0: logger.info( "truncated {} words longer than {} characters".format( truncated, max_char_len ) ) self.vocab = vocab self.word_to_char = word_to_char @property def padding_idx(self): return Dictionary().pad() if self.vocab is None else self.vocab.pad() def reset_parameters(self): nn.init.xavier_normal_(self.char_embeddings.weight) nn.init.xavier_normal_(self.symbol_embeddings) nn.init.xavier_uniform_(self.projection.weight) nn.init.constant_( self.char_embeddings.weight[self.char_embeddings.padding_idx], 0.0 ) nn.init.constant_(self.projection.bias, 0.0) def forward( self, input: torch.Tensor, ): if self.char_inputs: chars = input.view(-1, self.max_char_len) pads = chars[:, 0].eq(CHAR_PAD_IDX) eos = chars[:, 0].eq(CHAR_EOS_IDX) if eos.any(): if self.onnx_trace: chars = torch.where(eos.unsqueeze(1), chars.new_zeros(1), chars) else: chars[eos] = 0 unk = None else: flat_words = input.view(-1) chars = self.word_to_char[flat_words.type_as(self.word_to_char)].type_as( input ) pads = flat_words.eq(self.vocab.pad()) eos = flat_words.eq(self.vocab.eos()) unk = flat_words.eq(self.vocab.unk()) word_embs = self._convolve(chars) if self.onnx_trace: if pads.any(): word_embs = torch.where( pads.unsqueeze(1), word_embs.new_zeros(1), word_embs ) if eos.any(): word_embs = torch.where( eos.unsqueeze(1), self.symbol_embeddings[self.eos_idx], word_embs ) if unk is not None and unk.any(): word_embs = torch.where( unk.unsqueeze(1), self.symbol_embeddings[self.unk_idx], word_embs ) else: if pads.any(): word_embs[pads] = 0 if eos.any(): word_embs[eos] = self.symbol_embeddings[self.eos_idx] if unk is not None and unk.any(): word_embs[unk] = self.symbol_embeddings[self.unk_idx] return word_embs.view(input.size()[:2] + (-1,)) def _convolve( self, char_idxs: torch.Tensor, ): char_embs = self.char_embeddings(char_idxs) char_embs = char_embs.transpose(1, 2) # BTC -> BCT conv_result = [] for conv in self.convolutions: x = conv(char_embs) x, _ = torch.max(x, -1) x = F.relu(x) conv_result.append(x) x = torch.cat(conv_result, dim=-1) if self.highway is not None: x = self.highway(x) x = self.projection(x) return x class Highway(torch.nn.Module): """ A `Highway layer <https://arxiv.org/abs/1505.00387>`_. Adopted from the AllenNLP implementation. """ def __init__(self, input_dim: int, num_layers: int = 1): super(Highway, self).__init__() self.input_dim = input_dim self.layers = nn.ModuleList( [nn.Linear(input_dim, input_dim * 2) for _ in range(num_layers)] ) self.activation = nn.ReLU() self.reset_parameters() def reset_parameters(self): for layer in self.layers: # As per comment in AllenNLP: # We should bias the highway layer to just carry its input forward. We do that by # setting the bias on `B(x)` to be positive, because that means `g` will be biased to # be high, so we will carry the input forward. The bias on `B(x)` is the second half # of the bias vector in each Linear layer. nn.init.constant_(layer.bias[self.input_dim :], 1) nn.init.constant_(layer.bias[: self.input_dim], 0) nn.init.xavier_normal_(layer.weight) def forward(self, x: torch.Tensor): for layer in self.layers: projection = layer(x) proj_x, gate = projection.chunk(2, dim=-1) proj_x = self.activation(proj_x) gate = torch.sigmoid(gate) x = gate * x + (gate.new_tensor([1]) - gate) * proj_x return x

COCO-LM/fairseq/fairseq/modules/character_token_embedder.py/0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. import logging from typing import List, Optional import torch.nn as nn import torch.nn.functional as F logger = logging.getLogger(__name__) class FairseqDropout(nn.Module): def __init__(self, p, module_name=None): super().__init__() self.p = p self.module_name = module_name self.apply_during_inference = False def forward(self, x, inplace: bool = False): if self.p > 0 and (self.training or self.apply_during_inference): return F.dropout(x, p=self.p, training=True, inplace=inplace) else: return x def make_generation_fast_( self, name: str, retain_dropout: bool = False, retain_dropout_modules: Optional[List[str]] = None, **kwargs ): if retain_dropout: if retain_dropout_modules is not None and self.module_name is None: logger.warning( "Cannot enable dropout during inference for module {} " "because module_name was not set".format(name) ) elif ( retain_dropout_modules is None # if None, apply to all modules or self.module_name in retain_dropout_modules ): logger.info( "Enabling dropout during inference for module: {}".format(name) ) self.apply_during_inference = True else: logger.info("Disabling dropout for module: {}".format(name))

COCO-LM/fairseq/fairseq/modules/fairseq_dropout.py/0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. import torch import torch.nn as nn import torch.nn.functional as F from fairseq import utils from fairseq.incremental_decoding_utils import with_incremental_state from fairseq.modules.fairseq_dropout import FairseqDropout from fairseq.modules.unfold import unfold1d def LightweightConv( input_size, kernel_size=1, padding_l=None, num_heads=1, weight_dropout=0.0, weight_softmax=False, bias=False, ): if torch.cuda.is_available(): try: from fairseq.modules.lightconv_layer import LightconvLayer return LightconvLayer( input_size, kernel_size=kernel_size, padding_l=padding_l, num_heads=num_heads, weight_dropout=weight_dropout, weight_softmax=weight_softmax, bias=bias, ) except ImportError as e: print(e) return LightweightConv1dTBC( input_size, kernel_size=kernel_size, padding_l=padding_l, num_heads=num_heads, weight_dropout=weight_dropout, weight_softmax=weight_softmax, bias=bias, ) class LightweightConv1d(nn.Module): """Lightweight Convolution assuming the input is BxCxT This is just an example that explains LightConv clearer than the TBC version. We don't use this module in the model. Args: input_size: # of channels of the input and output kernel_size: convolution channels padding: padding num_heads: number of heads used. The weight is of shape `(num_heads, 1, kernel_size)` weight_softmax: normalize the weight with softmax before the convolution Shape: Input: BxCxT, i.e. (batch_size, input_size, timesteps) Output: BxCxT, i.e. (batch_size, input_size, timesteps) Attributes: weight: the learnable weights of the module of shape `(num_heads, 1, kernel_size)` bias: the learnable bias of the module of shape `(input_size)` """ def __init__( self, input_size, kernel_size=1, padding=0, num_heads=1, weight_softmax=False, bias=False, weight_dropout=0.0, ): super().__init__() self.input_size = input_size self.kernel_size = kernel_size self.num_heads = num_heads self.padding = padding self.weight_softmax = weight_softmax self.weight = nn.Parameter(torch.Tensor(num_heads, 1, kernel_size)) if bias: self.bias = nn.Parameter(torch.Tensor(input_size)) else: self.bias = None self.weight_dropout_module = FairseqDropout( weight_dropout, module_name=self.__class__.__name__ ) self.reset_parameters() def reset_parameters(self): nn.init.xavier_uniform_(self.weight) if self.bias is not None: nn.init.constant_(self.bias, 0.0) def forward(self, input): """ input size: B x C x T output size: B x C x T """ B, C, T = input.size() H = self.num_heads weight = self.weight if self.weight_softmax: weight = F.softmax(weight, dim=-1) weight = self.weight_dropout_module(weight) # Merge every C/H entries into the batch dimension (C = self.input_size) # B x C x T -> (B * C/H) x H x T # One can also expand the weight to C x 1 x K by a factor of C/H # and do not reshape the input instead, which is slow though input = input.view(-1, H, T) output = F.conv1d(input, weight, padding=self.padding, groups=self.num_heads) output = output.view(B, C, T) if self.bias is not None: output = output + self.bias.view(1, -1, 1) return output @with_incremental_state class LightweightConv1dTBC(nn.Module): """Lightweight Convolution assuming the input is TxBxC Args: input_size: # of channels of the input kernel_size: convolution channels padding_l: padding to the left when using "same" padding num_heads: number of heads used. The weight is of shape (num_heads, 1, kernel_size) weight_dropout: the drop rate of the DropConnect to drop the weight weight_softmax: normalize the weight with softmax before the convolution bias: use bias Shape: Input: TxBxC, i.e. (timesteps, batch_size, input_size) Output: TxBxC, i.e. (timesteps, batch_size, input_size) Attributes: weight: the learnable weights of the module of shape `(num_heads, 1, kernel_size)` bias: the learnable bias of the module of shape `(input_size)` """ def __init__( self, input_size, kernel_size=1, padding_l=None, num_heads=1, weight_dropout=0.0, weight_softmax=False, bias=False, ): super().__init__() self.input_size = input_size self.kernel_size = kernel_size self.padding_l = padding_l self.num_heads = num_heads self.weight_dropout_module = FairseqDropout( weight_dropout, module_name=self.__class__.__name__ ) self.weight_softmax = weight_softmax self.weight = nn.Parameter(torch.Tensor(num_heads, 1, kernel_size)) if bias: self.bias = nn.Parameter(torch.Tensor(input_size)) else: self.bias = None self.reset_parameters() self.onnx_trace = False def reset_parameters(self): nn.init.xavier_uniform_(self.weight) if self.bias is not None: nn.init.constant_(self.bias, 0.0) def forward(self, x, incremental_state=None, unfold=False): """Assuming the input, x, of the shape T x B x C and producing an output in the shape T x B x C args: x: Input of shape T x B x C, i.e. (timesteps, batch_size, input_size) incremental_state: A dict to keep the state unfold: unfold the input or not. If not, we use the matrix trick instead """ unfold = unfold or (incremental_state is not None) if unfold: output = self._forward_unfolded(x, incremental_state) else: output = self._forward_expanded(x, incremental_state) if self.bias is not None: output = output + self.bias.view(1, 1, -1) return output def prepare_for_onnx_export_(self): self.onnx_trace = True def _forward_unfolded(self, x, incremental_state): """The conventional implementation of convolutions. Unfolding the input by having a window shifting to the right.""" T, B, C = x.size() K, H = self.kernel_size, self.num_heads R = C // H assert R * H == C == self.input_size weight = self.weight.view(H, K) if incremental_state is not None: input_buffer = self._get_input_buffer(incremental_state) if input_buffer is None: input_buffer = x.new() x_unfold = torch.cat([input_buffer, x.unsqueeze(3)], dim=3) if self.kernel_size > 1: self._set_input_buffer( incremental_state, x_unfold[:, :, :, -self.kernel_size + 1 :] ) x_unfold = x_unfold.view(T * B * H, R, -1) else: # unfold the input: T x B x C --> T' x B x C x K x_unfold = unfold1d(x, self.kernel_size, self.padding_l, 0) x_unfold = x_unfold.view(T * B * H, R, K) if self.weight_softmax: weight = utils.softmax(weight, dim=1, onnx_trace=self.onnx_trace).type_as( weight ) if incremental_state is not None: weight = weight[:, -x_unfold.size(2) :] K = weight.size(1) weight = ( weight.view(1, H, K).expand(T * B, H, K).contiguous().view(T * B * H, K, 1) ) weight = self.weight_dropout_module(weight) output = torch.bmm(x_unfold, weight) # T*B*H x R x 1 output = output.view(T, B, C) return output def _forward_expanded(self, x, incremental_state): """Turn the convolution filters into band matrices and do matrix multiplication. This is faster when the sequence is short, but less memory efficient. This is not used in the decoder during inference. """ T, B, C = x.size() K, H = self.kernel_size, self.num_heads R = C // H assert R * H == C == self.input_size weight = self.weight.view(H, K) if self.weight_softmax: weight = utils.softmax(weight, dim=1, onnx_trace=self.onnx_trace).type_as( weight ) weight = weight.view(1, H, K).expand(T * B, H, K).contiguous() weight = weight.view(T, B * H, K).transpose(0, 1) x = x.view(T, B * H, R).transpose(0, 1) P = self.padding_l if K > T and P == K - 1: weight = weight.narrow(2, K - T, T) K, P = T, T - 1 # turn the convolution filters into band matrices weight_expanded = weight.new_zeros(B * H, T, T + K - 1, requires_grad=False) weight_expanded.as_strided((B * H, T, K), (T * (T + K - 1), T + K, 1)).copy_( weight ) weight_expanded = weight_expanded.narrow(2, P, T) weight_expanded = self.weight_dropout_module(weight_expanded) output = torch.bmm(weight_expanded, x) output = output.transpose(0, 1).contiguous().view(T, B, C) return output def reorder_incremental_state(self, incremental_state, new_order): input_buffer = self._get_input_buffer(incremental_state) if input_buffer is not None: input_buffer = input_buffer.index_select(1, new_order) self._set_input_buffer(incremental_state, input_buffer) def _get_input_buffer(self, incremental_state): return utils.get_incremental_state(self, incremental_state, "input_buffer") def _set_input_buffer(self, incremental_state, new_buffer): return utils.set_incremental_state( self, incremental_state, "input_buffer", new_buffer ) def extra_repr(self): s = "{}, kernel_size={}, padding_l={}, num_heads={}, weight_softmax={}, bias={}".format( self.input_size, self.kernel_size, self.padding_l, self.num_heads, self.weight_softmax, self.bias is not None, ) if self.weight_dropout_module.p > 0.0: s += ", weight_dropout={}".format(self.weight_dropout_module.p) return s

COCO-LM/fairseq/fairseq/modules/lightweight_convolution.py/0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. from .qact import ActivationQuantizer # NOQA from .qconv import IntConv2d # NOQA from .qemb import IntEmbedding # NOQA from .qlinear import IntLinear # NOQA

COCO-LM/fairseq/fairseq/modules/quantization/scalar/modules/__init__.py/0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. """ transpose last 2 dimensions of the input """ import torch.nn as nn class TransposeLast(nn.Module): def __init__(self, deconstruct_idx=None): super().__init__() self.deconstruct_idx = deconstruct_idx def forward(self, x): if self.deconstruct_idx is not None: x = x[self.deconstruct_idx] return x.transpose(-2, -1)

COCO-LM/fairseq/fairseq/modules/transpose_last.py/0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. from collections import defaultdict from itertools import chain import torch from fairseq import optim from omegaconf import DictConfig from .dynamic_loss_scaler import DynamicLossScaler class _FP16OptimizerMixin(object): def __init__(self, *args, **kwargs): # forward __init__ call to the next class in mro(method resolution order) super().__init__(*args, **kwargs) self._multiply_factor = 1.0 @property def has_flat_params(self): return torch.is_tensor(self.fp32_params) or ( isinstance(self.fp32_params, dict) and all(torch.is_tensor(t) for t in self.fp32_params.values()) ) @classmethod def build_fp32_params(cls, args, params, flatten=True): # create FP32 copy of parameters and grads if flatten: is_pipeline_parallel = getattr( args, "pipeline_model_parallel", False ) and getattr(args, "distributed_no_spawn", False) total_param_size = sum(p.data.numel() for p in params) devices = [torch.cuda.current_device()] if is_pipeline_parallel: devices = list(set(args.pipeline_devices)) fp32_params = {} for device in devices: if is_pipeline_parallel: device_param_size = sum( p.data.numel() for p in params if p.device.index == device ) device_params = [p for p in params if p.device.index == device] else: device_param_size = total_param_size device_params = params fp32_params[device] = ( device_params[0].new(0).float().new(device_param_size) ) offset = 0 for p in device_params: numel = p.data.numel() fp32_params[device][offset : offset + numel].copy_(p.data.view(-1)) offset += numel fp32_params[device] = torch.nn.Parameter(fp32_params[device]) fp32_params[device].grad = fp32_params[device].data.new( device_param_size ) return fp32_params else: fp32_params = [] for p in params: p32 = torch.nn.Parameter(p.data.float()) p32.grad = torch.zeros_like(p32.data) if hasattr(p, "param_group"): p32.param_group = p.param_group fp32_params.append(p32) return fp32_params def state_dict(self): """Return the optimizer's state dict.""" state_dict = self.fp32_optimizer.state_dict() if self.scaler is not None: state_dict["loss_scale"] = self.scaler.loss_scale return state_dict def load_state_dict(self, state_dict, optimizer_overrides=None): """Load an optimizer state dict. In general we should prefer the configuration of the existing optimizer instance (e.g., learning rate) over that found in the state_dict. This allows us to resume training from a checkpoint using a new set of optimizer args. """ if "loss_scale" in state_dict and self.scaler is not None: self.scaler.loss_scale = state_dict["loss_scale"] self.fp32_optimizer.load_state_dict(state_dict, optimizer_overrides) def backward(self, loss): """Computes the sum of gradients of the given tensor w.r.t. graph leaves. Compared to :func:`fairseq.optim.FairseqOptimizer.backward`, this function additionally dynamically scales the loss to avoid gradient underflow. """ if self.scaler is not None: loss = self.scaler.scale(loss) loss.backward() self._needs_sync = True def _sync_fp16_grads_to_fp32(self): if self._needs_sync: # copy FP16 grads to FP32 if self.has_flat_params: devices = list(self.fp32_params.keys()) device_params_dict = defaultdict(list) for p in self.fp16_params: if p.requires_grad: device_params_dict[p.device.index].append(p) for device in devices: device_params = device_params_dict[device] offset = 0 for p in device_params: grad_data = ( p.grad.data if p.grad is not None else p.data.new_zeros(p.data.shape) ) numel = grad_data.numel() self.fp32_params[device].grad.data[ offset : offset + numel ].copy_(grad_data.view(-1)) offset += numel else: for p, p32 in zip(self.fp16_params, self.fp32_params): if not p.requires_grad: continue if p.grad is not None: if p32.grad is None: p32.grad = p.grad.data.float() else: p32.grad.data.copy_(p.grad.data) else: p32.grad = torch.zeros_like(p.data, dtype=torch.float) self._needs_sync = False def _sync_fp32_params_to_fp16(self): # copy FP32 params back into FP16 model if self.has_flat_params: devices = list(self.fp32_params.keys()) device_params_dict = defaultdict(list) for p in self.fp16_params: device_params_dict[p.device.index].append(p) for device in devices: device_params = device_params_dict[device] offset = 0 for p in device_params: numel = p.data.numel() p.data.copy_( self.fp32_params[device] .data[offset : offset + numel] .view_as(p.data) ) offset += numel else: for p, p32 in zip(self.fp16_params, self.fp32_params): if not p.requires_grad: continue p.data.copy_(p32.data) def _unscale_grads(self): self._sync_fp16_grads_to_fp32() if ( # Skip the multiplication if it's a no-op (i.e., if _multiply_factor # is 1.0). At the same time, we want to avoid the device-to-host # transfer by comparing it to 1.0. Since _multiply_factor starts as # a Python float, we roughly assume that if it's a tensor then it's # probably not =1.0 anymore and we do the multiplication. Otherwise # we can safely check the value without a D2H transfer. torch.is_tensor(self._multiply_factor) or self._multiply_factor != 1.0 ): self.fp32_optimizer.multiply_grads(self._multiply_factor) self._multiply_factor = 1.0 def multiply_grads(self, c): """Multiplies grads by a constant ``c``.""" self._multiply_factor *= c def clip_grad_norm(self, max_norm, aggregate_norm_fn=None): """Clips gradient norm and updates dynamic loss scaler.""" self._sync_fp16_grads_to_fp32() grad_norm = self._multiply_factor * self.fp32_optimizer.clip_grad_norm( 0, aggregate_norm_fn ) if self.scaler is not None: if grad_norm > max_norm > 0.0: self._multiply_factor *= max_norm / grad_norm self.scaler.check_overflow(grad_norm) elif max_norm > 0.0: clip_coef = (max_norm / (grad_norm + 1e-6)).clamp_(max=1) self._multiply_factor *= clip_coef return grad_norm def step(self, closure=None, groups=None): """Performs a single optimization step.""" self._sync_fp16_grads_to_fp32() if getattr(self, "supports_step_with_scale", False): self.fp32_optimizer.step(closure, scale=(1.0 / self._multiply_factor), groups=groups) else: self._unscale_grads() self.fp32_optimizer.step(closure, groups=groups) if self.scaler is not None: self.scaler.update() self._sync_fp32_params_to_fp16() def zero_grad(self): """Clears the gradients of all optimized parameters.""" for p in self.fp16_params: p.grad = None if self.has_flat_params: if torch.is_tensor(self.fp32_params): self.fp32_params.grad.zero_() elif isinstance(self.fp32_params, dict): for fp32_params in self.fp32_params.values(): fp32_params.grad.zero_() else: raise RuntimeError("self.fp32_params must be a tensor or dict") else: for p32 in self.fp32_params: if p32.grad is not None: p32.grad.zero_() self._needs_sync = False if self.scaler is not None: self._multiply_factor = 1.0 / float(self.scaler.loss_scale) class FP16Optimizer(_FP16OptimizerMixin, optim.FairseqOptimizer): """ Wrap an *optimizer* to support FP16 (mixed precision) training. """ def __init__(self, cfg: DictConfig, params, fp32_optimizer, fp32_params, **kwargs): super().__init__(cfg.optimizer) self.fp16_params = params self.fp32_optimizer = fp32_optimizer self.fp32_params = fp32_params if getattr(cfg.common, "fp16_scale_window", None) is None: if len(cfg.optimization.update_freq) > 1: raise ValueError( "--fp16-scale-window must be given explicitly when using a " "custom --update-freq schedule" ) data_parallel_size = int( cfg.distributed_training.distributed_world_size / cfg.common.model_parallel_size ) scale_window = int( 2 ** 14 / data_parallel_size / cfg.optimization.update_freq[0] ) else: scale_window = cfg.common.fp16_scale_window if not getattr(cfg.common, "bf16", False): self.scaler = DynamicLossScaler( init_scale=cfg.common.fp16_init_scale, scale_window=scale_window, tolerance=cfg.common.fp16_scale_tolerance, threshold=cfg.common.threshold_loss_scale, min_loss_scale=cfg.common.min_loss_scale, ) else: # disable loss scaling for bfloat16 self.scaler = None @classmethod def build_optimizer(cls, cfg: DictConfig, params, **kwargs): """ Args: cfg (omegaconf.DictConfig): fairseq args params (iterable): iterable of parameters to optimize """ flatten = not getattr(cfg.common, "fp16_no_flatten_grads", False) if getattr(cfg.common, "bf16", False): flatten = False # mixed precision is faster on TPUs without flat grads fp32_params = cls.build_fp32_params(cfg.optimizer, params, flatten=flatten) if flatten: fp32_optimizer = optim.build_optimizer(cfg.optimizer, [fp32_params]) else: fp32_optimizer = optim.build_optimizer(cfg.optimizer, fp32_params) if flatten and not fp32_optimizer.supports_flat_params: raise RuntimeError( f"chosen optimizer {fp32_optimizer.__class__.__name__} does not support flat params, please set --fp16-no-flatten-grads" ) return cls(cfg, params, fp32_optimizer, fp32_params, **kwargs) @property def optimizer(self): return self.fp32_optimizer.optimizer @optimizer.setter def optimizer(self, optimizer): self.fp32_optimizer.optimizer = optimizer @property def lr_scheduler(self): return getattr(self.fp32_optimizer, "lr_scheduler", None) @property def optimizer_config(self): return self.fp32_optimizer.optimizer_config def get_lr(self): return self.fp32_optimizer.get_lr() def set_lr(self, lr): self.fp32_optimizer.set_lr(lr) def all_reduce_grads(self, module): self.fp32_optimizer.all_reduce_grads(module) @property def supports_flat_params(self): return self.fp32_optimizer.supports_flat_params class _MemoryEfficientFP16OptimizerMixin(object): def __init__(self, *args, **kwargs): # forward __init__ call to the next class in MRO (method resolution order) super().__init__(*args, **kwargs) self._multiply_factor = 1.0 @property def has_flat_params(self): return False def state_dict(self): """Return the optimizer's state dict.""" state_dict = self.wrapped_optimizer.state_dict() if self.scaler is not None: state_dict["loss_scale"] = self.scaler.loss_scale return state_dict def load_state_dict(self, state_dict, optimizer_overrides=None): """Load an optimizer state dict. In general we should prefer the configuration of the existing optimizer instance (e.g., learning rate) over that found in the state_dict. This allows us to resume training from a checkpoint using a new set of optimizer args. """ if "loss_scale" in state_dict and self.scaler is not None: self.scaler.loss_scale = state_dict["loss_scale"] self.wrapped_optimizer.load_state_dict(state_dict, optimizer_overrides) # Hack: PyTorch automatically casts the optimizer state to match the # type of the current parameters. But with --memory-efficient-fp16 the # params are FP16 while the optimizer state is FP32 and we don't want # to cast. A workaround is to manually copy back the original state # after the optimizer has been loaded. if not getattr(self.optimizer, "disable_mem_eff_fp16_loading_hack", False): groups = self.optimizer.param_groups saved_groups = state_dict["param_groups"] id_map = { old_id: p for old_id, p in zip( chain(*(g["params"] for g in saved_groups)), chain(*(g["params"] for g in groups)), ) } for k, v in state_dict["state"].items(): if k in id_map: param = id_map[k] self.optimizer.state[param] = v def backward(self, loss): """Computes the sum of gradients of the given tensor w.r.t. graph leaves. Compared to :func:`fairseq.optim.FairseqOptimizer.backward`, this function additionally dynamically scales the loss to avoid gradient underflow. """ if self.scaler is not None: loss = self.scaler.scale(loss) loss.backward() def _unscale_grads(self): if ( # Skip the multiplication if it's a no-op (i.e., if _multiply_factor # is 1.0). At the same time, we want to avoid the device-to-host # transfer by comparing it to 1.0. Since _multiply_factor starts as # a Python float, we roughly assume that if it's a tensor then it's # probably not =1.0 anymore and we do the multiplication. Otherwise # we can safely check the value without a D2H transfer. torch.is_tensor(self._multiply_factor) or self._multiply_factor != 1.0 ): self.wrapped_optimizer.multiply_grads(self._multiply_factor) self._multiply_factor = 1.0 def multiply_grads(self, c): """Multiplies grads by a constant *c*.""" self._multiply_factor *= c def clip_grad_norm(self, max_norm, aggregate_norm_fn=None): """Clips gradient norm and updates dynamic loss scaler.""" max_norm = float(max_norm) grad_norm = self._multiply_factor * self.wrapped_optimizer.clip_grad_norm( 0, aggregate_norm_fn ) if self.scaler is not None: grad_norm_cpu = float(grad_norm) if grad_norm_cpu > max_norm > 0.0: self._multiply_factor *= max_norm / grad_norm_cpu # detect overflow and adjust loss scale self.scaler.check_overflow(grad_norm_cpu) elif max_norm > 0.0: clip_coef = (max_norm / (grad_norm + 1e-6)).clamp_(max=1) self._multiply_factor *= clip_coef return grad_norm def step(self, closure=None, groups=None): """Performs a single optimization step.""" if getattr(self, "supports_step_with_scale", False): # NOTE(msb) optimizer divides by scale factor self.wrapped_optimizer.step(closure, scale=(1.0 / self._multiply_factor), groups=groups) else: self._unscale_grads() self.wrapped_optimizer.step(closure, groups=groups) if self.scaler is not None: self.scaler.update() def zero_grad(self): """Clears the gradients of all optimized parameters.""" self.wrapped_optimizer.zero_grad() if self.scaler is not None: self._multiply_factor = 1.0 / float(self.scaler.loss_scale) else: self._multiply_factor = 1.0 @property def supports_flat_params(self): return self.wrapped_optimizer.supports_flat_params class MemoryEfficientFP16Optimizer( _MemoryEfficientFP16OptimizerMixin, optim.FairseqOptimizer ): """ Wrap an *optimizer* to support FP16 (mixed precision) training. Compared to :class:`fairseq.optim.FP16Optimizer`, this version does not maintain an FP32 copy of the model. We instead expect the optimizer to convert the gradients to FP32 internally and sync the results back to the FP16 model params. This significantly reduces memory usage but slightly increases the time spent in the optimizer. Since this wrapper depends on specific functionality in the wrapped optimizer (i.e., on-the-fly conversion of grads to FP32), only certain optimizers can be wrapped. This is determined by the *supports_memory_efficient_fp16* property. """ def __init__( self, cfg: DictConfig, params, optimizer, allow_unsupported=False, **kwargs ): if not allow_unsupported and not optimizer.supports_memory_efficient_fp16: raise ValueError( "Unsupported optimizer: {}".format(optimizer.__class__.__name__) ) super().__init__(cfg.optimizer) self.wrapped_optimizer = optimizer if getattr(cfg.common, "fp16_scale_window", None) is None: if len(cfg.optimization.update_freq) > 1: raise ValueError( "--fp16-scale-window must be given explicitly when using a " "custom --update-freq schedule" ) data_parallel_size = int( cfg.distributed_training.distributed_world_size / cfg.common.model_parallel_size ) scale_window = int( 2 ** 14 / data_parallel_size / cfg.optimization.update_freq[0] ) else: scale_window = cfg.common.fp16_scale_window if not getattr(cfg.common, "bf16", False): self.scaler = DynamicLossScaler( init_scale=cfg.common.fp16_init_scale, scale_window=scale_window, tolerance=cfg.common.fp16_scale_tolerance, threshold=cfg.common.threshold_loss_scale, min_loss_scale=cfg.common.min_loss_scale, ) else: # disable loss scaling for bfloat16 self.scaler = None @classmethod def build_optimizer(cls, cfg: DictConfig, params, **kwargs): """ Args: args (argparse.Namespace): fairseq args params (iterable): iterable of parameters to optimize """ fp16_optimizer = optim.build_optimizer(cfg.optimizer, params) return cls(cfg, params, fp16_optimizer, **kwargs) @property def optimizer(self): return self.wrapped_optimizer.optimizer @optimizer.setter def optimizer(self, optimizer): self.wrapped_optimizer.optimizer = optimizer @property def optimizer_config(self): return self.wrapped_optimizer.optimizer_config @property def lr_scheduler(self): return getattr(self.wrapped_optimizer, "lr_scheduler", None) def get_lr(self): return self.wrapped_optimizer.get_lr() def set_lr(self, lr): self.wrapped_optimizer.set_lr(lr) def all_reduce_grads(self, module): self.wrapped_optimizer.all_reduce_grads(module)

COCO-LM/fairseq/fairseq/optim/fp16_optimizer.py/0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. from typing import Any, Dict from fairseq.distributed import utils try: from fairscale.optim import OSS _has_fairscale = True except ImportError: _has_fairscale = False def shard_(optimizer, group): if not _has_fairscale: raise ImportError( "\n\nPlease install the fairscale package:" "\n\n pip install fairscale" ) class FairseqOSS(OSS): @property def disable_mem_eff_fp16_loading_hack(self): return True def __getattr__(self, name): if name.startswith("supports") and hasattr(self.optim, name): return getattr(self.optim, name) raise AttributeError( "'FairseqOSS' object has no attribute {0!r}".format(name) ) def broadcast_global_state_dict( self, state_dict: Dict[str, Any] ) -> Dict[str, Any]: """ Broadcasts the entire state_dict to all other ranks each rank is responsible to load their own partition of data """ return utils.broadcast_object( state_dict, src_rank=0, group=self.group, ) torch_optimizer = optimizer.optimizer optim_cls = type(torch_optimizer) optimizer.optimizer = FairseqOSS( torch_optimizer.param_groups, optim_cls, group=group, **optimizer.optimizer_config )

COCO-LM/fairseq/fairseq/optim/shard.py/0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. import logging import os from fairseq import utils from fairseq.data import ( AppendTokenDataset, DenoisingDataset, Dictionary, IdDataset, NestedDictionaryDataset, NumelDataset, PadDataset, PrependTokenDataset, StripTokenDataset, TokenBlockDataset, data_utils, ) from fairseq.data.encoders.utils import get_whole_word_mask from fairseq.data.shorten_dataset import maybe_shorten_dataset from fairseq.tasks import LegacyFairseqTask, register_task import numpy as np logger = logging.getLogger(__name__) @register_task("denoising") class DenoisingTask(LegacyFairseqTask): """ Denoising task for applying sequence to sequence denoising. (ie. BART) """ @staticmethod def add_args(parser): """Add task-specific arguments to the parser.""" parser.add_argument("data", help="path to data directory") parser.add_argument( "--tokens-per-sample", default=512, type=int, help="max number of total tokens over all segments" " per sample for dataset", ) parser.add_argument( "--sample-break-mode", default="complete_doc", type=str, help="mode for breaking sentence", ) parser.add_argument( "--mask", default=0.0, type=float, help="fraction of words/subwords that will be masked", ) parser.add_argument( "--mask-random", default=0.0, type=float, help="instead of using [MASK], use random token this often", ) parser.add_argument( "--insert", default=0.0, type=float, help="insert this percentage of additional random tokens", ) parser.add_argument( "--permute", default=0.0, type=float, help="take this proportion of subwords and permute them", ) parser.add_argument( "--rotate", default=0.5, type=float, help="rotate this proportion of inputs", ) parser.add_argument( "--poisson-lambda", default=3.0, type=float, help="randomly shuffle sentences for this proportion of inputs", ) parser.add_argument( "--permute-sentences", default=0.0, type=float, help="shuffle this proportion of sentences in all inputs", ) parser.add_argument( "--mask-length", default="subword", type=str, choices=["subword", "word", "span-poisson"], help="mask length to choose", ) parser.add_argument( "--replace-length", default=-1, type=int, help="when masking N tokens, replace with 0, 1, or N tokens (use -1 for N)", ) parser.add_argument( "--max-source-positions", default=1024, type=int, metavar="N", help="max number of tokens in the source sequence", ) parser.add_argument( "--max-target-positions", default=1024, type=int, metavar="N", help="max number of tokens in the target sequence", ) parser.add_argument( "--shorten-method", default="none", choices=["none", "truncate", "random_crop"], help="if not none, shorten sequences that exceed --tokens-per-sample", ) parser.add_argument( "--shorten-data-split-list", default="", help="comma-separated list of dataset splits to apply shortening to, " 'e.g., "train,valid" (default: all dataset splits)', ) def __init__(self, args, dictionary): super().__init__(args) self.dictionary = dictionary self.seed = args.seed # add mask token self.mask_idx = self.dictionary.add_symbol("<mask>") @classmethod def setup_task(cls, args, **kwargs): """Setup the task.""" dictionary = Dictionary.load(os.path.join(args.data, "dict.txt")) logger.info("dictionary: {} types".format(len(dictionary))) if not hasattr(args, "shuffle_instance"): args.shuffle_instance = False return cls(args, dictionary) def load_dataset(self, split, epoch=1, combine=False, **kwargs): """Load a given dataset split. Args: split (str): name of the split (e.g., train, valid, test) """ paths = utils.split_paths(self.args.data) assert len(paths) > 0 data_path = paths[(epoch - 1) % len(paths)] split_path = os.path.join(data_path, split) dataset = data_utils.load_indexed_dataset( split_path, self.dictionary, self.args.dataset_impl, combine=combine, ) if dataset is None: raise FileNotFoundError( "Dataset not found: {} ({})".format(split, split_path) ) dataset = StripTokenDataset(dataset, self.dictionary.eos()) dataset = maybe_shorten_dataset( dataset, split, self.args.shorten_data_split_list, self.args.shorten_method, self.args.tokens_per_sample, self.args.seed, ) # create continuous blocks of tokens dataset = TokenBlockDataset( dataset, dataset.sizes, self.args.tokens_per_sample - 2, # one less for <s> and one for </s> pad=self.dictionary.pad(), eos=self.dictionary.eos(), break_mode=self.args.sample_break_mode, document_sep_len=0, ) # prepend beginning-of-sentence token (<s>, equiv. to [CLS] in BERT) dataset = PrependTokenDataset(dataset, self.source_dictionary.bos()) dataset = AppendTokenDataset(dataset, self.source_dictionary.eos()) mask_whole_words = ( get_whole_word_mask(self.args, self.source_dictionary) if self.args.mask_length != "subword" else None ) self.datasets[split] = DenoisingDataset( dataset, dataset.sizes, self.dictionary, self.mask_idx, mask_whole_words, shuffle=self.args.shuffle_instance, seed=self.seed, args=self.args, ) logger.info( "Split: {0}, Loaded {1} samples of denoising_dataset".format( split, len(self.datasets[split]), ) ) def build_dataset_for_inference(self, src_tokens, src_lengths, **kwargs): """ Generate batches for inference. We assume that the input begins with a bos symbol (`<s>`) and ends with an eos symbol (`</s>`). """ pad = self.source_dictionary.pad() eos = self.source_dictionary.eos() src_dataset = TokenBlockDataset( src_tokens, src_lengths, block_size=self.args.tokens_per_sample - 2, # for <s> and </s> pad=pad, eos=eos, break_mode=self.args.sample_break_mode, document_sep_len=0, ) prev_output_tokens = PrependTokenDataset( StripTokenDataset(src_dataset, eos), eos ) src_dataset = PadDataset(src_dataset, pad_idx=pad, left_pad=False) return NestedDictionaryDataset( { "id": IdDataset(), "net_input": { "src_tokens": src_dataset, "src_lengths": NumelDataset(src_dataset, reduce=False), "prev_output_tokens": PadDataset( prev_output_tokens, pad_idx=pad, left_pad=False ), }, "target": src_dataset, }, sizes=[np.array(src_lengths)], ) def max_positions(self): """Return the max sentence length allowed by the task.""" return (self.args.max_source_positions, self.args.max_target_positions) @property def source_dictionary(self): """Return the source :class:`~fairseq.data.Dictionary`.""" return self.dictionary @property def target_dictionary(self): """Return the target :class:`~fairseq.data.Dictionary`.""" return self.dictionary

COCO-LM/fairseq/fairseq/tasks/denoising.py/0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. from dataclasses import dataclass, field import torch from fairseq import utils from fairseq.data import LanguagePairDataset from fairseq.dataclass import ChoiceEnum from fairseq.tasks import register_task from fairseq.tasks.translation import TranslationConfig, TranslationTask, load_langpair_dataset from fairseq.utils import new_arange NOISE_CHOICES = ChoiceEnum(["random_delete", "random_mask", "no_noise", "full_mask"]) @dataclass class TranslationLevenshteinConfig(TranslationConfig): noise: NOISE_CHOICES = field( default="random_delete", metadata={ "help": "type of noise" }, ) @register_task("translation_lev", dataclass=TranslationLevenshteinConfig) class TranslationLevenshteinTask(TranslationTask): """ Translation (Sequence Generation) task for Levenshtein Transformer See `"Levenshtein Transformer" <https://arxiv.org/abs/1905.11006>`_. """ cfg: TranslationLevenshteinConfig def load_dataset(self, split, epoch=1, combine=False, **kwargs): """Load a given dataset split. Args: split (str): name of the split (e.g., train, valid, test) """ paths = utils.split_paths(self.cfg.data) assert len(paths) > 0 data_path = paths[(epoch - 1) % len(paths)] # infer langcode src, tgt = self.cfg.source_lang, self.cfg.target_lang self.datasets[split] = load_langpair_dataset( data_path, split, src, self.src_dict, tgt, self.tgt_dict, combine=combine, dataset_impl=self.cfg.dataset_impl, upsample_primary=self.cfg.upsample_primary, left_pad_source=self.cfg.left_pad_source, left_pad_target=self.cfg.left_pad_target, max_source_positions=self.cfg.max_source_positions, max_target_positions=self.cfg.max_target_positions, prepend_bos=True, ) def inject_noise(self, target_tokens): def _random_delete(target_tokens): pad = self.tgt_dict.pad() bos = self.tgt_dict.bos() eos = self.tgt_dict.eos() max_len = target_tokens.size(1) target_mask = target_tokens.eq(pad) target_score = target_tokens.clone().float().uniform_() target_score.masked_fill_( target_tokens.eq(bos) | target_tokens.eq(eos), 0.0 ) target_score.masked_fill_(target_mask, 1) target_score, target_rank = target_score.sort(1) target_length = target_mask.size(1) - target_mask.float().sum( 1, keepdim=True ) # do not delete <bos> and <eos> (we assign 0 score for them) target_cutoff = ( 2 + ( (target_length - 2) * target_score.new_zeros(target_score.size(0), 1).uniform_() ).long() ) target_cutoff = target_score.sort(1)[1] >= target_cutoff prev_target_tokens = ( target_tokens.gather(1, target_rank) .masked_fill_(target_cutoff, pad) .gather(1, target_rank.masked_fill_(target_cutoff, max_len).sort(1)[1]) ) prev_target_tokens = prev_target_tokens[ :, : prev_target_tokens.ne(pad).sum(1).max() ] return prev_target_tokens def _random_mask(target_tokens): pad = self.tgt_dict.pad() bos = self.tgt_dict.bos() eos = self.tgt_dict.eos() unk = self.tgt_dict.unk() target_masks = ( target_tokens.ne(pad) & target_tokens.ne(bos) & target_tokens.ne(eos) ) target_score = target_tokens.clone().float().uniform_() target_score.masked_fill_(~target_masks, 2.0) target_length = target_masks.sum(1).float() target_length = target_length * target_length.clone().uniform_() target_length = target_length + 1 # make sure to mask at least one token. _, target_rank = target_score.sort(1) target_cutoff = new_arange(target_rank) < target_length[:, None].long() prev_target_tokens = target_tokens.masked_fill( target_cutoff.scatter(1, target_rank, target_cutoff), unk ) return prev_target_tokens def _full_mask(target_tokens): pad = self.tgt_dict.pad() bos = self.tgt_dict.bos() eos = self.tgt_dict.eos() unk = self.tgt_dict.unk() target_mask = ( target_tokens.eq(bos) | target_tokens.eq(eos) | target_tokens.eq(pad) ) return target_tokens.masked_fill(~target_mask, unk) if self.cfg.noise == "random_delete": return _random_delete(target_tokens) elif self.cfg.noise == "random_mask": return _random_mask(target_tokens) elif self.cfg.noise == "full_mask": return _full_mask(target_tokens) elif self.cfg.noise == "no_noise": return target_tokens else: raise NotImplementedError def build_generator(self, models, args, **unused): # add models input to match the API for SequenceGenerator from fairseq.iterative_refinement_generator import IterativeRefinementGenerator return IterativeRefinementGenerator( self.target_dictionary, eos_penalty=getattr(args, "iter_decode_eos_penalty", 0.0), max_iter=getattr(args, "iter_decode_max_iter", 10), beam_size=getattr(args, "iter_decode_with_beam", 1), reranking=getattr(args, "iter_decode_with_external_reranker", False), decoding_format=getattr(args, "decoding_format", None), adaptive=not getattr(args, "iter_decode_force_max_iter", False), retain_history=getattr(args, "retain_iter_history", False), ) def build_dataset_for_inference(self, src_tokens, src_lengths, constraints=None): if constraints is not None: # Though see Susanto et al. (ACL 2020): https://www.aclweb.org/anthology/2020.acl-main.325/ raise NotImplementedError( "Constrained decoding with the translation_lev task is not supported" ) return LanguagePairDataset( src_tokens, src_lengths, self.source_dictionary, append_bos=True ) def train_step( self, sample, model, criterion, optimizer, update_num, ignore_grad=False ): model.train() sample["prev_target"] = self.inject_noise(sample["target"]) loss, sample_size, logging_output = criterion(model, sample) if ignore_grad: loss *= 0 optimizer.backward(loss) return loss, sample_size, logging_output def valid_step(self, sample, model, criterion): model.eval() with torch.no_grad(): sample["prev_target"] = self.inject_noise(sample["target"]) loss, sample_size, logging_output = criterion(model, sample) return loss, sample_size, logging_output

COCO-LM/fairseq/fairseq/tasks/translation_lev.py/0

#!/usr/bin/env python3 -u # Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. import logging import os import sys from argparse import Namespace from itertools import chain import torch from fairseq import checkpoint_utils, distributed_utils, options, utils from fairseq.dataclass.utils import convert_namespace_to_omegaconf from fairseq.logging import metrics, progress_bar from omegaconf import DictConfig logging.basicConfig( format="%(asctime)s | %(levelname)s | %(name)s | %(message)s", datefmt="%Y-%m-%d %H:%M:%S", level=os.environ.get("LOGLEVEL", "INFO").upper(), stream=sys.stdout, ) logger = logging.getLogger("fairseq_cli.validate") def main(cfg: DictConfig, override_args=None): if isinstance(cfg, Namespace): cfg = convert_namespace_to_omegaconf(cfg) utils.import_user_module(cfg.common) assert ( cfg.dataset.max_tokens is not None or cfg.dataset.batch_size is not None ), "Must specify batch size either with --max-tokens or --batch-size" use_fp16 = cfg.common.fp16 use_cuda = torch.cuda.is_available() and not cfg.common.cpu if use_cuda: torch.cuda.set_device(cfg.distributed_training.device_id) if cfg.distributed_training.distributed_world_size > 1: data_parallel_world_size = distributed_utils.get_data_parallel_world_size() data_parallel_rank = distributed_utils.get_data_parallel_rank() else: data_parallel_world_size = 1 data_parallel_rank = 0 if override_args is not None: overrides = vars(override_args) overrides.update(eval(getattr(override_args, "model_overrides", "{}"))) else: overrides = None # Load ensemble logger.info("loading model(s) from {}".format(cfg.common_eval.path)) models, saved_cfg, task = checkpoint_utils.load_model_ensemble_and_task( [cfg.common_eval.path], arg_overrides=overrides, suffix=cfg.checkpoint.checkpoint_suffix, ) model = models[0] # Move models to GPU for model in models: if use_fp16: model.half() if use_cuda: model.cuda() # Print args logger.info(saved_cfg) # Build criterion criterion = task.build_criterion(saved_cfg.criterion) criterion.eval() for subset in cfg.dataset.valid_subset.split(","): try: task.load_dataset(subset, combine=False, epoch=1, task_cfg=saved_cfg.task) dataset = task.dataset(subset) except KeyError: raise Exception("Cannot find dataset: " + subset) # Initialize data iterator itr = task.get_batch_iterator( dataset=dataset, max_tokens=cfg.dataset.max_tokens, max_sentences=cfg.dataset.batch_size, max_positions=utils.resolve_max_positions( task.max_positions(), *[m.max_positions() for m in models], ), ignore_invalid_inputs=cfg.dataset.skip_invalid_size_inputs_valid_test, required_batch_size_multiple=cfg.dataset.required_batch_size_multiple, seed=cfg.common.seed, num_shards=data_parallel_world_size, shard_id=data_parallel_rank, num_workers=cfg.dataset.num_workers, data_buffer_size=cfg.dataset.data_buffer_size, ).next_epoch_itr(shuffle=False) progress = progress_bar.progress_bar( itr, log_format=cfg.common.log_format, log_interval=cfg.common.log_interval, prefix=f"valid on '{subset}' subset", default_log_format=("tqdm" if not cfg.common.no_progress_bar else "simple"), ) log_outputs = [] for i, sample in enumerate(progress): sample = utils.move_to_cuda(sample) if use_cuda else sample _loss, _sample_size, log_output = task.valid_step(sample, model, criterion) progress.log(log_output, step=i) log_outputs.append(log_output) if data_parallel_world_size > 1: log_outputs = distributed_utils.all_gather_list( log_outputs, max_size=cfg.common.all_gather_list_size, group=distributed_utils.get_data_parallel_group(), ) log_outputs = list(chain.from_iterable(log_outputs)) with metrics.aggregate() as agg: task.reduce_metrics(log_outputs, criterion) log_output = agg.get_smoothed_values() progress.print(log_output, tag=subset, step=i) def cli_main(): parser = options.get_validation_parser() args = options.parse_args_and_arch(parser) # only override args that are explicitly given on the command line override_parser = options.get_validation_parser() override_args = options.parse_args_and_arch( override_parser, suppress_defaults=True ) distributed_utils.call_main( convert_namespace_to_omegaconf(args), main, override_args=override_args ) if __name__ == "__main__": cli_main()

COCO-LM/fairseq/fairseq_cli/validate.py/0

import torch import fused_xentropy_cuda class SoftmaxCrossEntropyLoss(torch.autograd.Function): @staticmethod def forward(ctx, logits, labels, padding_idx=0, half_to_float=False): losses, max_log_sum_exp = fused_xentropy_cuda.forward( logits, labels, half_to_float) if padding_idx >= 0: losses.masked_fill_(labels==padding_idx, 0) ctx.save_for_backward(logits, max_log_sum_exp, labels, torch.LongTensor([padding_idx])) return losses @staticmethod def backward(ctx, grad_loss): logits, max_log_sum_exp, labels, padding_idx = ctx.saved_tensors if not grad_loss.is_contiguous(): grad_loss = grad_loss.contiguous() if padding_idx >= 0: grad_loss.masked_fill_(labels==padding_idx.item(), 0) grad_logits = fused_xentropy_cuda.backward( grad_loss, logits, max_log_sum_exp, labels) return grad_logits, None, None, None, None

COCO-LM/fairseq/fused_ops/fused_ops/xentropy/softmax_xentropy.py/0

#!/usr/bin/env python3 # # Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. """Extracts random constraints from reference files.""" import argparse import random import sys from sacrebleu import extract_ngrams def get_phrase(words, index, length): assert index < len(words) - length + 1 phr = " ".join(words[index : index + length]) for i in range(index, index + length): words.pop(index) return phr def main(args): if args.seed: random.seed(args.seed) for line in sys.stdin: constraints = [] def add_constraint(constraint): constraints.append(constraint) source = line.rstrip() if "\t" in line: source, target = line.split("\t") if args.add_sos: target = f"<s> {target}" if args.add_eos: target = f"{target} </s>" if len(target.split()) >= args.len: words = [target] num = args.number choices = {} for i in range(num): if len(words) == 0: break segmentno = random.choice(range(len(words))) segment = words.pop(segmentno) tokens = segment.split() phrase_index = random.choice(range(len(tokens))) choice = " ".join( tokens[phrase_index : min(len(tokens), phrase_index + args.len)] ) for j in range( phrase_index, min(len(tokens), phrase_index + args.len) ): tokens.pop(phrase_index) if phrase_index > 0: words.append(" ".join(tokens[0:phrase_index])) if phrase_index + 1 < len(tokens): words.append(" ".join(tokens[phrase_index:])) choices[target.find(choice)] = choice # mask out with spaces target = target.replace(choice, " " * len(choice), 1) for key in sorted(choices.keys()): add_constraint(choices[key]) print(source, *constraints, sep="\t") if __name__ == "__main__": parser = argparse.ArgumentParser() parser.add_argument("--number", "-n", type=int, default=1, help="number of phrases") parser.add_argument("--len", "-l", type=int, default=1, help="phrase length") parser.add_argument( "--add-sos", default=False, action="store_true", help="add <s> token" ) parser.add_argument( "--add-eos", default=False, action="store_true", help="add </s> token" ) parser.add_argument("--seed", "-s", default=0, type=int) args = parser.parse_args() main(args)

COCO-LM/fairseq/scripts/constraints/extract.py/0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. import collections import os import shutil import tempfile import unittest import numpy as np import torch from scripts.average_checkpoints import average_checkpoints from torch import nn class ModelWithSharedParameter(nn.Module): def __init__(self): super(ModelWithSharedParameter, self).__init__() self.embedding = nn.Embedding(1000, 200) self.FC1 = nn.Linear(200, 200) self.FC2 = nn.Linear(200, 200) # tie weight in FC2 to FC1 self.FC2.weight = nn.Parameter(self.FC1.weight) self.FC2.bias = nn.Parameter(self.FC1.bias) self.relu = nn.ReLU() def forward(self, input): return self.FC2(self.ReLU(self.FC1(input))) + self.FC1(input) class TestAverageCheckpoints(unittest.TestCase): def test_average_checkpoints(self): params_0 = collections.OrderedDict( [ ("a", torch.DoubleTensor([100.0])), ("b", torch.FloatTensor([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]])), ("c", torch.IntTensor([7, 8, 9])), ] ) params_1 = collections.OrderedDict( [ ("a", torch.DoubleTensor([1.0])), ("b", torch.FloatTensor([[1.0, 1.0, 1.0], [1.0, 1.0, 1.0]])), ("c", torch.IntTensor([2, 2, 2])), ] ) params_avg = collections.OrderedDict( [ ("a", torch.DoubleTensor([50.5])), ("b", torch.FloatTensor([[1.0, 1.5, 2.0], [2.5, 3.0, 3.5]])), # We expect truncation for integer division ("c", torch.IntTensor([4, 5, 5])), ] ) fd_0, path_0 = tempfile.mkstemp() fd_1, path_1 = tempfile.mkstemp() torch.save(collections.OrderedDict([("model", params_0)]), path_0) torch.save(collections.OrderedDict([("model", params_1)]), path_1) output = average_checkpoints([path_0, path_1])["model"] os.close(fd_0) os.remove(path_0) os.close(fd_1) os.remove(path_1) for (k_expected, v_expected), (k_out, v_out) in zip( params_avg.items(), output.items() ): self.assertEqual( k_expected, k_out, "Key mismatch - expected {} but found {}. " "(Expected list of keys: {} vs actual list of keys: {})".format( k_expected, k_out, params_avg.keys(), output.keys() ), ) np.testing.assert_allclose( v_expected.numpy(), v_out.numpy(), err_msg="Tensor value mismatch for key {}".format(k_expected), ) def test_average_checkpoints_with_shared_parameters(self): def _construct_model_with_shared_parameters(path, value): m = ModelWithSharedParameter() nn.init.constant_(m.FC1.weight, value) torch.save({"model": m.state_dict()}, path) return m tmpdir = tempfile.mkdtemp() paths = [] path = os.path.join(tmpdir, "m1.pt") m1 = _construct_model_with_shared_parameters(path, 1.0) paths.append(path) path = os.path.join(tmpdir, "m2.pt") m2 = _construct_model_with_shared_parameters(path, 2.0) paths.append(path) path = os.path.join(tmpdir, "m3.pt") m3 = _construct_model_with_shared_parameters(path, 3.0) paths.append(path) new_model = average_checkpoints(paths) self.assertTrue( torch.equal( new_model["model"]["embedding.weight"], (m1.embedding.weight + m2.embedding.weight + m3.embedding.weight) / 3.0, ) ) self.assertTrue( torch.equal( new_model["model"]["FC1.weight"], (m1.FC1.weight + m2.FC1.weight + m3.FC1.weight) / 3.0, ) ) self.assertTrue( torch.equal( new_model["model"]["FC2.weight"], (m1.FC2.weight + m2.FC2.weight + m3.FC2.weight) / 3.0, ) ) shutil.rmtree(tmpdir) if __name__ == "__main__": unittest.main()

COCO-LM/fairseq/tests/test_average_checkpoints.py/0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. import unittest from fairseq.data import iterators class TestIterators(unittest.TestCase): def test_counting_iterator(self, ref=None, itr=None): if ref is None: assert itr is None ref = list(range(10)) itr = iterators.CountingIterator(ref) else: assert len(ref) == 10 assert itr is not None self.assertTrue(itr.has_next()) self.assertEqual(itr.n, 0) self.assertEqual(next(itr), ref[0]) self.assertEqual(itr.n, 1) self.assertEqual(next(itr), ref[1]) self.assertEqual(itr.n, 2) itr.skip(3) self.assertEqual(itr.n, 5) self.assertEqual(next(itr), ref[5]) itr.skip(3) self.assertEqual(itr.n, 9) self.assertEqual(next(itr), ref[9]) self.assertFalse(itr.has_next()) def test_grouped_iterator(self): # test correctness x = list(range(10)) itr = iterators.GroupedIterator(x, 1) self.assertEqual(list(itr), [[0], [1], [2], [3], [4], [5], [6], [7], [8], [9]]) itr = iterators.GroupedIterator(x, 4) self.assertEqual(list(itr), [[0, 1, 2, 3], [4, 5, 6, 7], [8, 9]]) itr = iterators.GroupedIterator(x, 5) self.assertEqual(list(itr), [[0, 1, 2, 3, 4], [5, 6, 7, 8, 9]]) # test CountingIterator functionality x = list(range(30)) ref = list(iterators.GroupedIterator(x, 3)) itr = iterators.GroupedIterator(x, 3) self.test_counting_iterator(ref, itr) def test_sharded_iterator(self): # test correctness x = list(range(10)) itr = iterators.ShardedIterator(x, num_shards=1, shard_id=0) self.assertEqual(list(itr), x) itr = iterators.ShardedIterator(x, num_shards=2, shard_id=0) self.assertEqual(list(itr), [0, 2, 4, 6, 8]) itr = iterators.ShardedIterator(x, num_shards=2, shard_id=1) self.assertEqual(list(itr), [1, 3, 5, 7, 9]) itr = iterators.ShardedIterator(x, num_shards=3, shard_id=0) self.assertEqual(list(itr), [0, 3, 6, 9]) itr = iterators.ShardedIterator(x, num_shards=3, shard_id=1) self.assertEqual(list(itr), [1, 4, 7, None]) itr = iterators.ShardedIterator(x, num_shards=3, shard_id=2) self.assertEqual(list(itr), [2, 5, 8, None]) # test CountingIterator functionality x = list(range(30)) ref = list(iterators.ShardedIterator(x, num_shards=3, shard_id=0)) itr = iterators.ShardedIterator(x, num_shards=3, shard_id=0) self.test_counting_iterator(ref, itr) def test_counting_iterator_take(self): ref = list(range(10)) itr = iterators.CountingIterator(ref) itr.take(5) self.assertEqual(len(itr), len(list(iter(itr)))) self.assertEqual(len(itr), 5) itr = iterators.CountingIterator(ref) itr.take(5) self.assertEqual(next(itr), ref[0]) self.assertEqual(next(itr), ref[1]) itr.skip(2) self.assertEqual(next(itr), ref[4]) self.assertFalse(itr.has_next()) def test_counting_iterator_buffered_iterator_take(self): ref = list(range(10)) buffered_itr = iterators.BufferedIterator(2, ref) itr = iterators.CountingIterator(buffered_itr) itr.take(5) self.assertEqual(len(itr), len(list(iter(itr)))) self.assertEqual(len(itr), 5) buffered_itr = iterators.BufferedIterator(2, ref) itr = iterators.CountingIterator(buffered_itr) itr.take(5) self.assertEqual(len(buffered_itr), 5) self.assertEqual(len(list(iter(buffered_itr))), 5) buffered_itr = iterators.BufferedIterator(2, ref) itr = iterators.CountingIterator(buffered_itr) itr.take(5) self.assertEqual(next(itr), ref[0]) self.assertEqual(next(itr), ref[1]) itr.skip(2) self.assertEqual(next(itr), ref[4]) self.assertFalse(itr.has_next()) self.assertRaises(StopIteration, next, buffered_itr) ref = list(range(4, 10)) buffered_itr = iterators.BufferedIterator(2, ref) itr = iterators.CountingIterator(buffered_itr, start=4) itr.take(5) self.assertEqual(len(itr), 5) self.assertEqual(len(buffered_itr), 1) self.assertEqual(next(itr), ref[0]) self.assertFalse(itr.has_next()) self.assertRaises(StopIteration, next, buffered_itr) if __name__ == "__main__": unittest.main()

COCO-LM/fairseq/tests/test_iterators.py/0

# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. import unittest import tests.utils as test_utils import torch from fairseq.data import TokenBlockDataset class TestTokenBlockDataset(unittest.TestCase): def _build_dataset(self, data, **kwargs): sizes = [len(x) for x in data] underlying_ds = test_utils.TestDataset(data) return TokenBlockDataset(underlying_ds, sizes, **kwargs) def test_eos_break_mode(self): data = [ torch.tensor([5, 4, 3, 2, 1], dtype=torch.long), torch.tensor([1], dtype=torch.long), torch.tensor([8, 7, 6, 1], dtype=torch.long), ] ds = self._build_dataset(data, block_size=None, pad=0, eos=1, break_mode="eos") self.assertEqual(ds[0].tolist(), [5, 4, 3, 2, 1]) self.assertEqual(ds[1].tolist(), [1]) self.assertEqual(ds[2].tolist(), [8, 7, 6, 1]) data = [ torch.tensor([5, 4, 3, 2, 1], dtype=torch.long), torch.tensor([8, 7, 6, 1], dtype=torch.long), torch.tensor([1], dtype=torch.long), ] ds = self._build_dataset(data, block_size=None, pad=0, eos=1, break_mode="eos") self.assertEqual(ds[0].tolist(), [5, 4, 3, 2, 1]) self.assertEqual(ds[1].tolist(), [8, 7, 6, 1]) self.assertEqual(ds[2].tolist(), [1]) def test_block_break_mode(self): data = [ torch.tensor([5, 4, 3, 2, 1], dtype=torch.long), torch.tensor([8, 7, 6, 1], dtype=torch.long), torch.tensor([9, 1], dtype=torch.long), ] ds = self._build_dataset(data, block_size=3, pad=0, eos=1, break_mode="none") self.assertEqual(ds[0].tolist(), [5, 4, 3]) self.assertEqual(ds[1].tolist(), [2, 1, 8]) self.assertEqual(ds[2].tolist(), [7, 6, 1]) self.assertEqual(ds[3].tolist(), [9, 1]) def test_complete_break_mode(self): data = [ torch.tensor([5, 4, 3, 2, 1], dtype=torch.long), torch.tensor([8, 7, 6, 1], dtype=torch.long), torch.tensor([9, 1], dtype=torch.long), ] ds = self._build_dataset( data, block_size=6, pad=0, eos=1, break_mode="complete" ) self.assertEqual(ds[0].tolist(), [5, 4, 3, 2, 1]) self.assertEqual(ds[1].tolist(), [8, 7, 6, 1, 9, 1]) data = [ torch.tensor([4, 3, 2, 1], dtype=torch.long), torch.tensor([5, 1], dtype=torch.long), torch.tensor([1], dtype=torch.long), torch.tensor([6, 1], dtype=torch.long), ] ds = self._build_dataset( data, block_size=3, pad=0, eos=1, break_mode="complete" ) self.assertEqual(ds[0].tolist(), [4, 3, 2, 1]) self.assertEqual(ds[1].tolist(), [5, 1, 1]) self.assertEqual(ds[2].tolist(), [6, 1]) def test_4billion_tokens(self): """Regression test for numpy type promotion issue https://github.com/numpy/numpy/issues/5745""" data = [torch.tensor(list(range(10000)), dtype=torch.long)] * 430000 ds = self._build_dataset( data, block_size=6, pad=0, eos=1, break_mode="complete" ) ds[-1] # __getitem__ works start, end = ds.slice_indices[-1] assert end > 4294967295 # data must be sufficiently large to overflow uint32 assert not isinstance( end + 1, float ) # this would also raise, since np.uint64(1) + 1 => 2.0 if __name__ == "__main__": unittest.main()

COCO-LM/fairseq/tests/test_token_block_dataset.py/0

# Copyright (c) Microsoft Corporation. # Licensed under the MIT license. ## The script is largely adapted from the huggingface transformers library. """ GLUE processors and helpers """ import logging import os import csv import sys import copy import json from scipy.stats import pearsonr, spearmanr from sklearn.metrics import matthews_corrcoef, f1_score logger = logging.getLogger(__name__) class InputExample(object): """ A single training/test example for simple sequence classification. Args: guid: Unique id for the example. text_a: string. The untokenized text of the first sequence. For single sequence tasks, only this sequence must be specified. text_b: (Optional) string. The untokenized text of the second sequence. Only must be specified for sequence pair tasks. label: (Optional) string. The label of the example. This should be specified for train and dev examples, but not for test examples. """ def __init__(self, guid, text_a, text_b=None, label=None): self.guid = guid self.text_a = text_a self.text_b = text_b self.label = label def __repr__(self): return str(self.to_json_string()) def to_dict(self): """Serializes this instance to a Python dictionary.""" output = copy.deepcopy(self.__dict__) return output def to_json_string(self): """Serializes this instance to a JSON string.""" return json.dumps(self.to_dict(), indent=2, sort_keys=True) + "\n" class InputFeatures(object): """ A single set of features of data. Args: input_ids: Indices of input sequence tokens in the vocabulary. attention_mask: Mask to avoid performing attention on padding token indices. Mask values selected in ``[0, 1]``: Usually ``1`` for tokens that are NOT MASKED, ``0`` for MASKED (padded) tokens. token_type_ids: Segment token indices to indicate first and second portions of the inputs. label: Label corresponding to the input """ def __init__(self, input_ids, attention_mask=None, token_type_ids=None, label=None): self.input_ids = input_ids self.attention_mask = attention_mask self.token_type_ids = token_type_ids self.label = label def __repr__(self): return str(self.to_json_string()) def to_dict(self): """Serializes this instance to a Python dictionary.""" output = copy.deepcopy(self.__dict__) return output def to_json_string(self): """Serializes this instance to a JSON string.""" return json.dumps(self.to_dict(), indent=2, sort_keys=True) + "\n" class DataProcessor(object): """Base class for data converters for sequence classification data sets.""" @classmethod def _read_tsv(cls, input_file, quotechar=None): """Reads a tab separated value file.""" with open(input_file, "r", encoding="utf-8-sig") as f: reader = csv.reader(f, delimiter="\t", quotechar=quotechar) lines = [] for line in reader: if sys.version_info[0] == 2: line = list(unicode(cell, 'utf-8') for cell in line) lines.append(line) return lines def glue_convert_examples_to_features(examples, tokenizer, max_length=512, task=None, label_list=None, output_mode=None, pad_on_left=False, pad_token_id=0, pad_token_segment_id=0, mask_padding_with_zero=True): """ Loads a data file into a list of ``InputFeatures`` Args: examples: List of ``InputExamples`` or ``tf.data.Dataset`` containing the examples. tokenizer: Instance of a tokenizer that will tokenize the examples max_length: Maximum example length task: GLUE task label_list: List of labels. Can be obtained from the processor using the ``processor.get_labels()`` method output_mode: String indicating the output mode. Either ``regression`` or ``classification`` pad_on_left: If set to ``True``, the examples will be padded on the left rather than on the right (default) pad_token_id: Padding token id pad_token_segment_id: The segment ID for the padding token (It is usually 0, but can vary such as for XLNet where it is 4) mask_padding_with_zero: If set to ``True``, the attention mask will be filled by ``1`` for actual values and by ``0`` for padded values. If set to ``False``, inverts it (``1`` for padded values, ``0`` for actual values) Returns: If the ``examples`` input is a ``tf.data.Dataset``, will return a ``tf.data.Dataset`` containing the task-specific features. If the input is a list of ``InputExamples``, will return a list of task-specific ``InputFeatures`` which can be fed to the model. """ is_tf_dataset = False if task is not None: processor = glue_processors[task]() if label_list is None: label_list = processor.get_labels() logger.info("Using label list %s for task %s" % (label_list, task)) if output_mode is None: output_mode = glue_output_modes[task] logger.info("Using output mode %s for task %s" % (output_mode, task)) label_map = {label: i for i, label in enumerate(label_list)} features = [] for (ex_index, example) in enumerate(examples): if ex_index % 10000 == 0: logger.info("Writing example %d" % (ex_index)) if is_tf_dataset: example = processor.get_example_from_tensor_dict(example) example = processor.tfds_map(example) inputs = tokenizer.encode_plus( example.text_a, example.text_b, add_special_tokens=True, max_length=max_length, ) input_ids = inputs["input_ids"] if "token_type_ids" in inputs: token_type_ids = inputs["token_type_ids"] else: token_type_ids = [] # The mask has 1 for real tokens and 0 for padding tokens. Only real # tokens are attended to. attention_mask = [1 if mask_padding_with_zero else 0] * len(input_ids) # Zero-pad up to the sequence length. padding_length = max_length - len(input_ids) if pad_on_left: input_ids = ([pad_token_id] * padding_length) + input_ids attention_mask = ([0 if mask_padding_with_zero else 1] * padding_length) + attention_mask if len(token_type_ids) == 0: padding_length = max_length token_type_ids = ([pad_token_segment_id] * padding_length) + token_type_ids else: input_ids = input_ids + ([pad_token_id] * padding_length) attention_mask = attention_mask + ([0 if mask_padding_with_zero else 1] * padding_length) if len(token_type_ids) == 0: padding_length = max_length token_type_ids = token_type_ids + ([pad_token_segment_id] * padding_length) assert len(input_ids) == max_length, "Error with input length {} vs {}".format(len(input_ids), max_length) assert len(attention_mask) == max_length, "Error with input length {} vs {}".format(len(attention_mask), max_length) assert len(token_type_ids) == max_length, "Error with input length {} vs {}".format(len(token_type_ids), max_length) if output_mode == "classification": label = label_map[example.label] elif output_mode == "regression": label = float(example.label) else: raise KeyError(output_mode) if ex_index < 5: logger.info("*** Example ***") logger.info("guid: %s" % (example.guid)) logger.info("input_ids: %s" % " ".join([str(x) for x in input_ids])) # logger.info("input_tokens: %s" % " ".join(tokenizer.convert_ids_to_tokens(input_ids))) logger.info("attention_mask: %s" % " ".join([str(x) for x in attention_mask])) logger.info("token_type_ids: %s" % " ".join([str(x) for x in token_type_ids])) logger.info("label: %s (id = %d)" % (example.label, label)) features.append( InputFeatures(input_ids=input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, label=label)) return features class MrpcProcessor(DataProcessor): """Processor for the MRPC data set (GLUE version).""" def get_example_from_tensor_dict(self, tensor_dict): """See base class.""" return InputExample(tensor_dict['idx'].numpy(), tensor_dict['sentence1'].numpy().decode('utf-8'), tensor_dict['sentence2'].numpy().decode('utf-8'), str(tensor_dict['label'].numpy())) def get_train_examples(self, data_dir): """See base class.""" logger.info("LOOKING AT {}".format(os.path.join(data_dir, "train.tsv"))) return self._create_examples( self._read_tsv(os.path.join(data_dir, "train.tsv")), "train") def get_dev_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "dev.tsv")), "dev") def get_labels(self): """See base class.""" return ["0", "1"] def _create_examples(self, lines, set_type): """Creates examples for the training and dev sets.""" examples = [] for (i, line) in enumerate(lines): if i == 0: continue guid = "%s-%s" % (set_type, i) text_a = line[3] text_b = line[4] label = line[0] examples.append( InputExample(guid=guid, text_a=text_a, text_b=text_b, label=label)) return examples class MnliProcessor(DataProcessor): """Processor for the MultiNLI data set (GLUE version).""" def get_example_from_tensor_dict(self, tensor_dict): """See base class.""" return InputExample(tensor_dict['idx'].numpy(), tensor_dict['premise'].numpy().decode('utf-8'), tensor_dict['hypothesis'].numpy().decode('utf-8'), str(tensor_dict['label'].numpy())) def get_train_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "train.tsv")), "train") def get_dev_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "dev_matched.tsv")), "dev_matched") def get_labels(self): """See base class.""" return ["contradiction", "entailment", "neutral"] def _create_examples(self, lines, set_type): """Creates examples for the training and dev sets.""" examples = [] for (i, line) in enumerate(lines): if i == 0: continue guid = "%s-%s" % (set_type, line[0]) text_a = line[8] text_b = line[9] label = line[-1] examples.append( InputExample(guid=guid, text_a=text_a, text_b=text_b, label=label)) return examples class MnliMismatchedProcessor(MnliProcessor): """Processor for the MultiNLI Mismatched data set (GLUE version).""" def get_dev_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "dev_mismatched.tsv")), "dev_matched") class ColaProcessor(DataProcessor): """Processor for the CoLA data set (GLUE version).""" def get_example_from_tensor_dict(self, tensor_dict): """See base class.""" return InputExample(tensor_dict['idx'].numpy(), tensor_dict['sentence'].numpy().decode('utf-8'), None, str(tensor_dict['label'].numpy())) def get_train_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "train.tsv")), "train") def get_dev_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "dev.tsv")), "dev") def get_labels(self): """See base class.""" return ["0", "1"] def _create_examples(self, lines, set_type): """Creates examples for the training and dev sets.""" examples = [] for (i, line) in enumerate(lines): guid = "%s-%s" % (set_type, i) text_a = line[3] label = line[1] examples.append( InputExample(guid=guid, text_a=text_a, text_b=None, label=label)) return examples class Sst2Processor(DataProcessor): """Processor for the SST-2 data set (GLUE version).""" def get_example_from_tensor_dict(self, tensor_dict): """See base class.""" return InputExample(tensor_dict['idx'].numpy(), tensor_dict['sentence'].numpy().decode('utf-8'), None, str(tensor_dict['label'].numpy())) def get_train_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "train.tsv")), "train") def get_dev_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "dev.tsv")), "dev") def get_labels(self): """See base class.""" return ["0", "1"] def _create_examples(self, lines, set_type): """Creates examples for the training and dev sets.""" examples = [] for (i, line) in enumerate(lines): if i == 0: continue guid = "%s-%s" % (set_type, i) text_a = line[0] label = line[1] examples.append( InputExample(guid=guid, text_a=text_a, text_b=None, label=label)) return examples class StsbProcessor(DataProcessor): """Processor for the STS-B data set (GLUE version).""" def get_example_from_tensor_dict(self, tensor_dict): """See base class.""" return InputExample(tensor_dict['idx'].numpy(), tensor_dict['sentence1'].numpy().decode('utf-8'), tensor_dict['sentence2'].numpy().decode('utf-8'), str(tensor_dict['label'].numpy())) def get_train_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "train.tsv")), "train") def get_dev_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "dev.tsv")), "dev") def get_labels(self): """See base class.""" return [None] def _create_examples(self, lines, set_type): """Creates examples for the training and dev sets.""" examples = [] for (i, line) in enumerate(lines): if i == 0: continue guid = "%s-%s" % (set_type, line[0]) text_a = line[7] text_b = line[8] label = line[-1] examples.append( InputExample(guid=guid, text_a=text_a, text_b=text_b, label=label)) return examples class QqpProcessor(DataProcessor): """Processor for the QQP data set (GLUE version).""" def get_example_from_tensor_dict(self, tensor_dict): """See base class.""" return InputExample(tensor_dict['idx'].numpy(), tensor_dict['question1'].numpy().decode('utf-8'), tensor_dict['question2'].numpy().decode('utf-8'), str(tensor_dict['label'].numpy())) def get_train_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "train.tsv")), "train") def get_dev_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "dev.tsv")), "dev") def get_labels(self): """See base class.""" return ["0", "1"] def _create_examples(self, lines, set_type): """Creates examples for the training and dev sets.""" examples = [] for (i, line) in enumerate(lines): if i == 0: continue guid = "%s-%s" % (set_type, line[0]) try: text_a = line[3] text_b = line[4] label = line[5] except IndexError: continue examples.append( InputExample(guid=guid, text_a=text_a, text_b=text_b, label=label)) return examples class QnliProcessor(DataProcessor): """Processor for the QNLI data set (GLUE version).""" def get_example_from_tensor_dict(self, tensor_dict): """See base class.""" return InputExample(tensor_dict['idx'].numpy(), tensor_dict['question'].numpy().decode('utf-8'), tensor_dict['sentence'].numpy().decode('utf-8'), str(tensor_dict['label'].numpy())) def get_train_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "train.tsv")), "train") def get_dev_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "dev.tsv")), "dev_matched") def get_labels(self): """See base class.""" return ["entailment", "not_entailment"] def _create_examples(self, lines, set_type): """Creates examples for the training and dev sets.""" examples = [] for (i, line) in enumerate(lines): if i == 0: continue guid = "%s-%s" % (set_type, line[0]) text_a = line[1] text_b = line[2] label = line[-1] examples.append( InputExample(guid=guid, text_a=text_a, text_b=text_b, label=label)) return examples class RteProcessor(DataProcessor): """Processor for the RTE data set (GLUE version).""" def get_example_from_tensor_dict(self, tensor_dict): """See base class.""" return InputExample(tensor_dict['idx'].numpy(), tensor_dict['sentence1'].numpy().decode('utf-8'), tensor_dict['sentence2'].numpy().decode('utf-8'), str(tensor_dict['label'].numpy())) def get_train_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "train.tsv")), "train") def get_dev_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "dev.tsv")), "dev") def get_labels(self): """See base class.""" return ["entailment", "not_entailment"] def _create_examples(self, lines, set_type): """Creates examples for the training and dev sets.""" examples = [] for (i, line) in enumerate(lines): if i == 0: continue guid = "%s-%s" % (set_type, line[0]) text_a = line[1] text_b = line[2] label = line[-1] examples.append( InputExample(guid=guid, text_a=text_a, text_b=text_b, label=label)) return examples class WnliProcessor(DataProcessor): """Processor for the WNLI data set (GLUE version).""" def get_example_from_tensor_dict(self, tensor_dict): """See base class.""" return InputExample(tensor_dict['idx'].numpy(), tensor_dict['sentence1'].numpy().decode('utf-8'), tensor_dict['sentence2'].numpy().decode('utf-8'), str(tensor_dict['label'].numpy())) def get_train_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "train.tsv")), "train") def get_dev_examples(self, data_dir): """See base class.""" return self._create_examples( self._read_tsv(os.path.join(data_dir, "dev.tsv")), "dev") def get_labels(self): """See base class.""" return ["0", "1"] def _create_examples(self, lines, set_type): """Creates examples for the training and dev sets.""" examples = [] for (i, line) in enumerate(lines): if i == 0: continue guid = "%s-%s" % (set_type, line[0]) text_a = line[1] text_b = line[2] label = line[-1] examples.append( InputExample(guid=guid, text_a=text_a, text_b=text_b, label=label)) return examples glue_tasks_num_labels = { "cola": 2, "mnli": 3, "mrpc": 2, "sst-2": 2, "sts-b": 1, "qqp": 2, "qnli": 2, "rte": 2, "wnli": 2, } glue_processors = { "cola": ColaProcessor, "mnli": MnliProcessor, "mnli-mm": MnliMismatchedProcessor, "mrpc": MrpcProcessor, "sst-2": Sst2Processor, "sts-b": StsbProcessor, "qqp": QqpProcessor, "qnli": QnliProcessor, "rte": RteProcessor, "wnli": WnliProcessor, } glue_output_modes = { "cola": "classification", "mnli": "classification", "mnli-mm": "classification", "mrpc": "classification", "sst-2": "classification", "sts-b": "regression", "qqp": "classification", "qnli": "classification", "rte": "classification", "wnli": "classification", } def simple_accuracy(preds, labels): return (preds == labels).mean() def acc_and_f1(preds, labels): acc = simple_accuracy(preds, labels) f1 = f1_score(y_true=labels, y_pred=preds) return { "acc": acc, "f1": f1, "acc_and_f1": (acc + f1) / 2, } def pearson_and_spearman(preds, labels): pearson_corr = pearsonr(preds, labels)[0] spearman_corr = spearmanr(preds, labels)[0] return { "pearson": pearson_corr, "spearmanr": spearman_corr, "corr": (pearson_corr + spearman_corr) / 2, } def glue_compute_metrics(task_name, preds, labels): assert len(preds) == len(labels) if task_name == "cola": return {"mcc": matthews_corrcoef(labels, preds)} elif task_name == "sst-2": return {"acc": simple_accuracy(preds, labels)} elif task_name == "mrpc": return acc_and_f1(preds, labels) elif task_name == "sts-b": return pearson_and_spearman(preds, labels) elif task_name == "qqp": return acc_and_f1(preds, labels) elif task_name == "mnli": return {"acc": simple_accuracy(preds, labels)} elif task_name == "mnli-mm": return {"acc": simple_accuracy(preds, labels)} elif task_name == "qnli": return {"acc": simple_accuracy(preds, labels)} elif task_name == "rte": return {"acc": simple_accuracy(preds, labels)} elif task_name == "wnli": return {"acc": simple_accuracy(preds, labels)} else: raise KeyError(task_name)

COCO-LM/huggingface/utils_for_glue.py/0

_base_ = [ '../_base_/models/upernet_cswin.py', '../_base_/datasets/ade20k.py', '../_base_/default_runtime.py', '../_base_/schedules/schedule_160k.py' ] model = dict( backbone=dict( type='CSWin', embed_dim=96, depth=[2,4,32,2], num_heads=[4,8,16,32], split_size=[1,2,7,7], drop_path_rate=0.6, use_chk=False, ), decode_head=dict( in_channels=[96,192,384,768], num_classes=150 ), auxiliary_head=dict( in_channels=384, num_classes=150 )) # AdamW optimizer, no weight decay for position embedding & layer norm in backbone optimizer = dict(_delete_=True, type='AdamW', lr=0.00006, betas=(0.9, 0.999), weight_decay=0.01, paramwise_cfg=dict(custom_keys={'absolute_pos_embed': dict(decay_mult=0.), 'relative_position_bias_table': dict(decay_mult=0.), 'norm': dict(decay_mult=0.)})) lr_config = dict(_delete_=True, policy='poly', warmup='linear', warmup_iters=1500, warmup_ratio=1e-6, power=1.0, min_lr=0.0, by_epoch=False) data=dict(samples_per_gpu=2)

CSWin-Transformer/segmentation/configs/cswin/upernet_cswin_base.py/0

[build-system] requires = ["setuptools", "setuptools-scm"] build-backend = "setuptools.build_meta" [project] name = "ClimaX" version = "0.3.1" authors =[ {name="Tung Nguyen", email="[email protected]"}, {name="Jayesh K. Gupta", email="[email protected]"} ] description = "" readme = "README.md" requires-python = ">=3.8" classifiers = [ "Programming Language :: Python :: 3", "License :: OSI Approved :: MIT License", ] dependencies = [ ] [project.urls] "Homepage" = "https://microsoft.github.io/ClimaX/" "Bug Tracker" = "https://github.com/microsoft/ClimaX/issues" [tool.setuptools.packages.find] where = ["src"]

ClimaX/pyproject.toml/0

datadir: /data/CMIP6/HAMMOZ name: 10m_u_component_of_wind cmip_name: uas era_name: u10 run: r1i1p1f1 version: v20190627 res: - 1.40625 # - 5.625

ClimaX/snakemake_configs/HAMMOZ/config_10m_u_component_of_wind.yml/0

datadir: /data/CMIP6/MPI-ESM server_prefix: http://esgf-data1.llnl.gov/thredds/fileServer/css03_data/CMIP6/CMIP name: v_component_of_wind cmip_name: va era_name: v output_type: 6hrPlevPt run: r1i1p1f1 version: v20190815 res: - 1.40625 # - 5.625

ClimaX/snakemake_configs/MPI-ESM/config_v_component_of_wind.yml/0

# Copyright (c) Microsoft Corporation. # Licensed under the MIT license. import math import warnings from typing import List from torch.optim import Optimizer from torch.optim.lr_scheduler import _LRScheduler class LinearWarmupCosineAnnealingLR(_LRScheduler): """Sets the learning rate of each parameter group to follow a linear warmup schedule between warmup_start_lr and base_lr followed by a cosine annealing schedule between base_lr and eta_min.""" def __init__( self, optimizer: Optimizer, warmup_epochs: int, max_epochs: int, warmup_start_lr: float = 0.0, eta_min: float = 0.0, last_epoch: int = -1, ) -> None: """ Args: optimizer (Optimizer): Wrapped optimizer. warmup_epochs (int): Maximum number of iterations for linear warmup max_epochs (int): Maximum number of iterations warmup_start_lr (float): Learning rate to start the linear warmup. Default: 0. eta_min (float): Minimum learning rate. Default: 0. last_epoch (int): The index of last epoch. Default: -1. """ self.warmup_epochs = warmup_epochs self.max_epochs = max_epochs self.warmup_start_lr = warmup_start_lr self.eta_min = eta_min super().__init__(optimizer, last_epoch) def get_lr(self) -> List[float]: """Compute learning rate using chainable form of the scheduler.""" if not self._get_lr_called_within_step: warnings.warn( "To get the last learning rate computed by the scheduler, " "please use `get_last_lr()`.", UserWarning, ) if self.last_epoch == self.warmup_epochs: return self.base_lrs if self.last_epoch == 0: return [self.warmup_start_lr] * len(self.base_lrs) if self.last_epoch < self.warmup_epochs: return [ group["lr"] + (base_lr - self.warmup_start_lr) / (self.warmup_epochs - 1) for base_lr, group in zip(self.base_lrs, self.optimizer.param_groups) ] if (self.last_epoch - 1 - self.max_epochs) % (2 * (self.max_epochs - self.warmup_epochs)) == 0: return [ group["lr"] + (base_lr - self.eta_min) * (1 - math.cos(math.pi / (self.max_epochs - self.warmup_epochs))) / 2 for base_lr, group in zip(self.base_lrs, self.optimizer.param_groups) ] return [ (1 + math.cos(math.pi * (self.last_epoch - self.warmup_epochs) / (self.max_epochs - self.warmup_epochs))) / ( 1 + math.cos( math.pi * (self.last_epoch - self.warmup_epochs - 1) / (self.max_epochs - self.warmup_epochs) ) ) * (group["lr"] - self.eta_min) + self.eta_min for group in self.optimizer.param_groups ] def _get_closed_form_lr(self) -> List[float]: """Called when epoch is passed as a param to the `step` function of the scheduler.""" if self.last_epoch < self.warmup_epochs: return [ self.warmup_start_lr + self.last_epoch * (base_lr - self.warmup_start_lr) / max(1, self.warmup_epochs - 1) for base_lr in self.base_lrs ] return [ self.eta_min + 0.5 * (base_lr - self.eta_min) * (1 + math.cos(math.pi * (self.last_epoch - self.warmup_epochs) / (self.max_epochs - self.warmup_epochs))) for base_lr in self.base_lrs ]

ClimaX/src/climax/utils/lr_scheduler.py/0

# Copyright (c) Microsoft Corporation. # Licensed under the MIT License. import os import sys import random import argparse import pickle import numpy as np import torch import models import data from util import util class BaseOptions(): def __init__(self): self.initialized = False def initialize(self, parser): # experiment specifics parser.add_argument('--name', type=str, default='deepfashionHD', help='name of the experiment. It decides where to store samples and models') parser.add_argument('--gpu_ids', type=str, default='0,1,2,3', help='gpu ids: e.g. 0 0,1,2, 0,2. use -1 for CPU') parser.add_argument('--checkpoints_dir', type=str, default='./checkpoints', help='models are saved here') parser.add_argument('--model', type=str, default='pix2pix', help='which model to use') parser.add_argument('--norm_G', type=str, default='spectralinstance', help='instance normalization or batch normalization') parser.add_argument('--norm_D', type=str, default='spectralinstance', help='instance normalization or batch normalization') parser.add_argument('--norm_E', type=str, default='spectralinstance', help='instance normalization or batch normalization') parser.add_argument('--phase', type=str, default='train', help='train, val, test, etc') # input/output sizes parser.add_argument('--batchSize', type=int, default=4, help='input batch size') parser.add_argument('--preprocess_mode', type=str, default='scale_width_and_crop', help='scaling and cropping of images at load time.', choices=("resize_and_crop", "crop", "scale_width", "scale_width_and_crop", "scale_shortside", "scale_shortside_and_crop", "fixed", "none")) parser.add_argument('--load_size', type=int, default=256, help='Scale images to this size. The final image will be cropped to --crop_size.') parser.add_argument('--crop_size', type=int, default=256, help='Crop to the width of crop_size (after initially scaling the images to load_size.)') parser.add_argument('--aspect_ratio', type=float, default=1.0, help='The ratio width/height. The final height of the load image will be crop_size/aspect_ratio') parser.add_argument('--label_nc', type=int, default=182, help='# of input label classes without unknown class. If you have unknown class as class label, specify --contain_dopntcare_label.') parser.add_argument('--contain_dontcare_label', action='store_true', help='if the label map contains dontcare label (dontcare=255)') parser.add_argument('--output_nc', type=int, default=3, help='# of output image channels') # for setting inputs parser.add_argument('--dataroot', type=str, default='dataset/deepfashionHD') parser.add_argument('--dataset_mode', type=str, default='deepfashionHD') parser.add_argument('--serial_batches', action='store_true', help='if true, takes images in order to make batches, otherwise takes them randomly') parser.add_argument('--no_flip', action='store_true', help='if specified, do not flip the images for data argumentation') parser.add_argument('--nThreads', default=16, type=int, help='# threads for loading data') parser.add_argument('--max_dataset_size', type=int, default=sys.maxsize, help='Maximum number of samples allowed per dataset. If the dataset directory contains more than max_dataset_size, only a subset is loaded.') parser.add_argument('--load_from_opt_file', action='store_true', help='load the options from checkpoints and use that as default') parser.add_argument('--cache_filelist_write', action='store_true', help='saves the current filelist into a text file, so that it loads faster') parser.add_argument('--cache_filelist_read', action='store_true', help='reads from the file list cache') # for displays parser.add_argument('--display_winsize', type=int, default=512, help='display window size') # for generator parser.add_argument('--netG', type=str, default='spade', help='selects model to use for netG (pix2pixhd | spade)') parser.add_argument('--ngf', type=int, default=64, help='# of gen filters in first conv layer') parser.add_argument('--init_type', type=str, default='xavier', help='network initialization [normal|xavier|kaiming|orthogonal]') parser.add_argument('--init_variance', type=float, default=0.02, help='variance of the initialization distribution') # for feature encoder parser.add_argument('--netCorr', type=str, default='NoVGGHPM') parser.add_argument('--nef', type=int, default=32, help='# of gen filters in first conv layer') # for instance-wise features parser.add_argument('--CBN_intype', type=str, default='warp_mask', help='type of CBN input for framework, warp/mask/warp_mask') parser.add_argument('--match_kernel', type=int, default=1, help='correspondence matrix match kernel size') parser.add_argument('--featEnc_kernel', type=int, default=3, help='kernel size in domain adaptor') parser.add_argument('--PONO', action='store_true', help='use positional normalization ') parser.add_argument('--PONO_C', action='store_true', help='use C normalization in corr module') parser.add_argument('--vgg_normal_correct', action='store_true', help='if true, correct vgg normalization and replace vgg FM model with ctx model') parser.add_argument('--use_coordconv', action='store_true', help='if true, use coordconv in CorrNet') parser.add_argument('--video_like', action='store_true', help='useful in deepfashion') parser.add_argument('--amp', action='store_true', help='use torch.cuda.amp') parser.add_argument('--temperature', type=float, default=0.01) parser.add_argument('--iteration_count', type=int, default=5) self.initialized = True return parser def gather_options(self): # initialize parser with basic options if not self.initialized: parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter) parser = self.initialize(parser) # get the basic options opt, unknown = parser.parse_known_args() # modify model-related parser options model_name = opt.model model_option_setter = models.get_option_setter(model_name) parser = model_option_setter(parser, self.isTrain) # modify dataset-related parser options dataset_mode = opt.dataset_mode dataset_option_setter = data.get_option_setter(dataset_mode) parser = dataset_option_setter(parser, self.isTrain) opt, unknown = parser.parse_known_args() # if there is opt_file, load it. # The previous default options will be overwritten if opt.load_from_opt_file: parser = self.update_options_from_file(parser, opt) opt = parser.parse_args() self.parser = parser return opt def print_options(self, opt): message = '' message += '----------------- Options ---------------\n' for k, v in sorted(vars(opt).items()): comment = '' default = self.parser.get_default(k) if v != default: comment = '\t[default: %s]' % str(default) message += '{:>25}: {:<30}{}\n'.format(str(k), str(v), comment) message += '----------------- End -------------------' print(message) def option_file_path(self, opt, makedir=False): expr_dir = os.path.join(opt.checkpoints_dir, opt.name) if makedir: util.mkdirs(expr_dir) file_name = os.path.join(expr_dir, 'opt') return file_name def save_options(self, opt): file_name = self.option_file_path(opt, makedir=True) with open(file_name + '.txt', 'wt') as opt_file: for k, v in sorted(vars(opt).items()): comment = '' default = self.parser.get_default(k) if v != default: comment = '\t[default: %s]' % str(default) opt_file.write('{:>25}: {:<30}{}\n'.format(str(k), str(v), comment)) with open(file_name + '.pkl', 'wb') as opt_file: pickle.dump(opt, opt_file) def update_options_from_file(self, parser, opt): new_opt = self.load_options(opt) for k, v in sorted(vars(opt).items()): if hasattr(new_opt, k) and v != getattr(new_opt, k): new_val = getattr(new_opt, k) parser.set_defaults(**{k: new_val}) return parser def load_options(self, opt): file_name = self.option_file_path(opt, makedir=False) new_opt = pickle.load(open(file_name + '.pkl', 'rb')) return new_opt def parse(self, save=False): # gather options from base, train, dataset, model opt = self.gather_options() # train or test opt.isTrain = self.isTrain self.print_options(opt) if opt.isTrain: self.save_options(opt) # Set semantic_nc based on the option. # This will be convenient in many places opt.semantic_nc = opt.label_nc + (1 if opt.contain_dontcare_label else 0) os.environ['CUDA_VISIBLE_DEVICES'] = opt.gpu_ids str_ids = opt.gpu_ids.split(',') opt.gpu_ids = list(range(len(str_ids))) seed = 1234 random.seed(seed) np.random.seed(seed) torch.manual_seed(seed) torch.random.manual_seed(seed) torch.cuda.manual_seed_all(seed) torch.backends.cudnn.benchmark = True if len(opt.gpu_ids) > 0: torch.cuda.set_device(opt.gpu_ids[0]) self.opt = opt return self.opt

CoCosNet-v2/options/base_options.py/0

# Copyright (c) Microsoft Corporation. # Licensed under the MIT License. import os import cv2 import torch import numpy as np from PIL import Image from skimage import feature from data.pix2pix_dataset import Pix2pixDataset from data.base_dataset import get_params, get_transform class CelebAHQEdgeDataset(Pix2pixDataset): #hair, skin, l_brow, r_blow, l_eye, r_eye, l_ear, r_ear, nose, u_lip, mouth, l_lip, neck, @staticmethod def modify_commandline_options(parser, is_train): parser = Pix2pixDataset.modify_commandline_options(parser, is_train) parser.set_defaults(preprocess_mode='resize_and_crop') parser.set_defaults(no_pairing_check=True) if is_train: parser.set_defaults(load_size=286) else: parser.set_defaults(load_size=256) parser.set_defaults(crop_size=256) parser.set_defaults(display_winsize=256) parser.set_defaults(label_nc=15) parser.set_defaults(contain_dontcare_label=False) parser.set_defaults(cache_filelist_read=False) parser.set_defaults(cache_filelist_write=False) return parser def get_paths(self, opt): if opt.phase == 'train': fd = open(os.path.join(opt.dataroot, 'train.txt')) lines = fd.readlines() fd.close() elif opt.phase == 'test': fd = open(os.path.join(opt.dataroot, 'val.txt')) lines = fd.readlines() fd.close() image_paths = [] label_paths = [] for i in range(len(lines)): image_paths.append(os.path.join(opt.dataroot, 'CelebA-HQ-img', lines[i].strip() + '.jpg')) subfolder = str(int(lines[i].strip()) // 2000) label_paths.append(os.path.join(opt.dataroot, 'CelebAMask-HQ-mask-anno', subfolder, lines[i].strip().zfill(5) + '_{}.png')) return label_paths, image_paths def get_ref(self, opt): extra = '' if opt.phase == 'test': extra = '_test' with open('./data/celebahq_ref{}.txt'.format(extra)) as fd: lines = fd.readlines() ref_dict = {} for i in range(len(lines)): items = lines[i].strip().split(',') key = items[0] if opt.phase == 'test': val = items[1:] else: val = [items[1], items[-1]] ref_dict[key] = val train_test_folder = ('', '') return ref_dict, train_test_folder def get_edges(self, edge, t): edge[:,1:] = edge[:,1:] | (t[:,1:] != t[:,:-1]) edge[:,:-1] = edge[:,:-1] | (t[:,1:] != t[:,:-1]) edge[1:,:] = edge[1:,:] | (t[1:,:] != t[:-1,:]) edge[:-1,:] = edge[:-1,:] | (t[1:,:] != t[:-1,:]) return edge def get_label_tensor(self, path): inner_parts = ['skin', 'l_brow', 'r_brow', 'l_eye', 'r_eye', 'l_ear', 'r_ear', 'nose', 'u_lip', 'mouth', 'l_lip', 'eye_g', 'hair'] img_path = self.labelpath_to_imgpath(path) img = Image.open(img_path).resize((self.opt.load_size, self.opt.load_size), resample=Image.BILINEAR) params = get_params(self.opt, img.size) transform_label = get_transform(self.opt, params, method=Image.NEAREST, normalize=False) transform_img = get_transform(self.opt, params, method=Image.BILINEAR, normalize=False) inner_label = np.ones(img.size, dtype=np.uint8) edges = np.zeros(img.size, dtype=np.uint8) tensors_dist = 0 e = 1 for part in inner_parts: edge = np.zeros(img.size, dtype=np.uint8) #this for distance transform map on each facial part if os.path.exists(path.format(part)): part_label = Image.open(path.format(part)).convert('L').resize((self.opt.load_size, self.opt.load_size), resample=Image.NEAREST) part_label = np.array(part_label) if part == 'hair': inner_label[part_label == 255] = 1 else: inner_label[part_label == 255] = 0 edges = self.get_edges(edges, part_label) edge = self.get_edges(edge, part_label) im_dist = cv2.distanceTransform(255-edge*255, cv2.DIST_L1, 3) im_dist = np.clip((im_dist / 3), 0, 255).astype(np.uint8) tensor_dist = transform_img(Image.fromarray(im_dist)) tensors_dist = tensor_dist if e == 1 else torch.cat([tensors_dist, tensor_dist]) e += 1 # canny edge for background canny_edges = feature.canny(np.array(img.convert('L'))) canny_edges = canny_edges * inner_label edges_all = edges + canny_edges edges_all[edges_all > 1] = 1 tensor_edges_all = transform_label(Image.fromarray(edges_all * 255)) edges[edges > 1] = 1 tensor_edges = transform_label(Image.fromarray(edges * 255)) label_tensor = torch.cat((tensor_edges_all, tensors_dist, tensor_edges), dim=0) return label_tensor, params def imgpath_to_labelpath(self, path): root, name = path.split('CelebA-HQ-img/') subfolder = str(int(name.split('.')[0]) // 2000) label_path = os.path.join(root, 'CelebAMask-HQ-mask-anno', subfolder, name.split('.')[0].zfill(5) + '_{}.png') return label_path def labelpath_to_imgpath(self, path): root= path.replace('\\', '/').split('CelebAMask-HQ-mask-anno/')[0] name = os.path.basename(path).split('_')[0] img_path = os.path.join(root, 'CelebA-HQ-img', str(int(name)) + '.jpg') return img_path # In ADE20k, 'unknown' label is of value 0. # Change the 'unknown' label to the last label to match other datasets. # def postprocess(self, input_dict): # label = input_dict['label'] # label = label - 1 # label[label == -1] = self.opt.label_nc # input_dict['label'] = label # if input_dict['label_ref'] is not None: # label_ref = input_dict['label_ref'] # label_ref = label_ref - 1 # label_ref[label_ref == -1] = self.opt.label_nc # input_dict['label_ref'] = label_ref

CoCosNet/data/celebahqedge_dataset.py/0

# Copyright (c) Microsoft Corporation. # Licensed under the MIT License. #08.09 change pad import numpy as np import torch import torch.nn as nn import torch.nn.functional as F from torch.autograd import Function from models.networks.base_network import BaseNetwork from models.networks.normalization import get_nonspade_norm_layer, equal_lr from models.networks.architecture import ResnetBlock as ResnetBlock from models.networks.architecture import SPADEResnetBlock as SPADEResnetBlock from models.networks.architecture import Attention from models.networks.sync_batchnorm import SynchronizedBatchNorm2d, SynchronizedBatchNorm1d class SPADEGenerator(BaseNetwork): @staticmethod def modify_commandline_options(parser, is_train): parser.set_defaults(norm_G='spectralspadesyncbatch3x3') return parser def __init__(self, opt): super().__init__() self.opt = opt nf = opt.ngf self.sw, self.sh = self.compute_latent_vector_size(opt) ic = 0 + (3 if 'warp' in self.opt.CBN_intype else 0) + (self.opt.semantic_nc if 'mask' in self.opt.CBN_intype else 0) self.fc = nn.Conv2d(ic, 16 * nf, 3, padding=1) if opt.eqlr_sn: self.fc = equal_lr(self.fc) self.head_0 = SPADEResnetBlock(16 * nf, 16 * nf, opt) self.G_middle_0 = SPADEResnetBlock(16 * nf, 16 * nf, opt) self.G_middle_1 = SPADEResnetBlock(16 * nf, 16 * nf, opt) self.up_0 = SPADEResnetBlock(16 * nf, 8 * nf, opt) self.up_1 = SPADEResnetBlock(8 * nf, 4 * nf, opt) if opt.use_attention: self.attn = Attention(4 * nf, 'spectral' in opt.norm_G) self.up_2 = SPADEResnetBlock(4 * nf, 2 * nf, opt) self.up_3 = SPADEResnetBlock(2 * nf, 1 * nf, opt) final_nc = nf self.conv_img = nn.Conv2d(final_nc, 3, 3, padding=1) self.up = nn.Upsample(scale_factor=2) def compute_latent_vector_size(self, opt): num_up_layers = 5 sw = opt.crop_size // (2**num_up_layers) sh = round(sw / opt.aspect_ratio) return sw, sh def forward(self, input, warp_out=None): seg = input if warp_out is None else warp_out # we downsample segmap and run convolution x = F.interpolate(seg, size=(self.sh, self.sw)) x = self.fc(x) x = self.head_0(x, seg) x = self.up(x) x = self.G_middle_0(x, seg) x = self.G_middle_1(x, seg) x = self.up(x) x = self.up_0(x, seg) x = self.up(x) x = self.up_1(x, seg) x = self.up(x) if self.opt.use_attention: x = self.attn(x) x = self.up_2(x, seg) x = self.up(x) x = self.up_3(x, seg) x = self.conv_img(F.leaky_relu(x, 2e-1)) x = F.tanh(x) return x class AdaptiveFeatureGenerator(BaseNetwork): @staticmethod def modify_commandline_options(parser, is_train): parser.set_defaults(norm_G='spectralspadesyncbatch3x3') parser.add_argument('--num_upsampling_layers', choices=('normal', 'more', 'most'), default='normal', help="If 'more', adds upsampling layer between the two middle resnet blocks. If 'most', also add one more upsampling + resnet layer at the end of the generator") return parser def __init__(self, opt): # TODO: kernel=4, concat noise, or change architecture to vgg feature pyramid super().__init__() self.opt = opt kw = 3 pw = int(np.ceil((kw - 1.0) / 2)) ndf = opt.ngf norm_layer = get_nonspade_norm_layer(opt, opt.norm_E) self.layer1 = norm_layer(nn.Conv2d(opt.spade_ic, ndf, kw, stride=1, padding=pw)) self.layer2 = norm_layer(nn.Conv2d(ndf * 1, ndf * 2, opt.adaptor_kernel, stride=2, padding=pw)) self.layer3 = norm_layer(nn.Conv2d(ndf * 2, ndf * 4, kw, stride=1, padding=pw)) if opt.warp_stride == 2: self.layer4 = norm_layer(nn.Conv2d(ndf * 4, ndf * 8, kw, stride=1, padding=pw)) else: self.layer4 = norm_layer(nn.Conv2d(ndf * 4, ndf * 8, opt.adaptor_kernel, stride=2, padding=pw)) self.layer5 = norm_layer(nn.Conv2d(ndf * 8, ndf * 8, kw, stride=1, padding=pw)) self.actvn = nn.LeakyReLU(0.2, False) self.opt = opt nf = opt.ngf self.head_0 = SPADEResnetBlock(8 * nf, 8 * nf, opt, use_se=opt.adaptor_se) if opt.adaptor_nonlocal: self.attn = Attention(8 * nf, False) self.G_middle_0 = SPADEResnetBlock(8 * nf, 8 * nf, opt, use_se=opt.adaptor_se) self.G_middle_1 = SPADEResnetBlock(8 * nf, 4 * nf, opt, use_se=opt.adaptor_se) if opt.adaptor_res_deeper: self.deeper0 = SPADEResnetBlock(4 * nf, 4 * nf, opt) if opt.dilation_conv: self.deeper1 = SPADEResnetBlock(4 * nf, 4 * nf, opt, dilation=2) self.deeper2 = SPADEResnetBlock(4 * nf, 4 * nf, opt, dilation=4) self.degridding0 = norm_layer(nn.Conv2d(ndf * 4, ndf * 4, 3, stride=1, padding=2, dilation=2)) self.degridding1 = norm_layer(nn.Conv2d(ndf * 4, ndf * 4, 3, stride=1, padding=1)) else: self.deeper1 = SPADEResnetBlock(4 * nf, 4 * nf, opt) self.deeper2 = SPADEResnetBlock(4 * nf, 4 * nf, opt) def forward(self, input, seg): x = self.layer1(input) x = self.layer2(self.actvn(x)) x = self.layer3(self.actvn(x)) x = self.layer4(self.actvn(x)) x = self.layer5(self.actvn(x)) x = self.head_0(x, seg) if self.opt.adaptor_nonlocal: x = self.attn(x) x = self.G_middle_0(x, seg) x = self.G_middle_1(x, seg) if self.opt.adaptor_res_deeper: x = self.deeper0(x, seg) x = self.deeper1(x, seg) x = self.deeper2(x, seg) if self.opt.dilation_conv: x = self.degridding0(x) x = self.degridding1(x) return x class ReverseGenerator(BaseNetwork): def __init__(self, opt, ic, oc, size): super().__init__() self.opt = opt self.downsample = True if size == 256 else False nf = opt.ngf opt.spade_ic = ic if opt.warp_reverseG_s: self.backbone_0 = SPADEResnetBlock(4 * nf, 4 * nf, opt) else: self.backbone_0 = SPADEResnetBlock(4 * nf, 8 * nf, opt) self.backbone_1 = SPADEResnetBlock(8 * nf, 8 * nf, opt) self.backbone_2 = SPADEResnetBlock(8 * nf, 8 * nf, opt) self.backbone_3 = SPADEResnetBlock(8 * nf, 4 * nf, opt) self.backbone_4 = SPADEResnetBlock(4 * nf, 2 * nf, opt) self.backbone_5 = SPADEResnetBlock(2 * nf, nf, opt) del opt.spade_ic if self.downsample: kw = 3 pw = int(np.ceil((kw - 1.0) / 2)) ndf = opt.ngf norm_layer = get_nonspade_norm_layer(opt, opt.norm_E) self.layer1 = norm_layer(nn.Conv2d(ic, ndf, kw, stride=1, padding=pw)) self.layer2 = norm_layer(nn.Conv2d(ndf * 1, ndf * 2, 4, stride=2, padding=pw)) self.layer3 = norm_layer(nn.Conv2d(ndf * 2, ndf * 4, kw, stride=1, padding=pw)) self.layer4 = norm_layer(nn.Conv2d(ndf * 4, ndf * 4, 4, stride=2, padding=pw)) self.up = nn.Upsample(scale_factor=2) self.actvn = nn.LeakyReLU(0.2, False) self.conv_img = nn.Conv2d(nf, oc, 3, padding=1) def forward(self, x): input = x if self.downsample: x = self.layer1(input) x = self.layer2(self.actvn(x)) x = self.layer3(self.actvn(x)) x = self.layer4(self.actvn(x)) x = self.backbone_0(x, input) if not self.opt.warp_reverseG_s: x = self.backbone_1(x, input) x = self.backbone_2(x, input) x = self.backbone_3(x, input) if self.downsample: x = self.up(x) x = self.backbone_4(x, input) if self.downsample: x = self.up(x) x = self.backbone_5(x, input) x = self.conv_img(F.leaky_relu(x, 2e-1)) x = F.tanh(x) return x class DomainClassifier(BaseNetwork): def __init__(self, opt): super().__init__() nf = opt.ngf kw = 4 if opt.domain_rela else 3 pw = int((kw - 1.0) / 2) self.feature = nn.Sequential(nn.Conv2d(4 * nf, 2 * nf, kw, stride=2, padding=pw), SynchronizedBatchNorm2d(2 * nf, affine=True), nn.LeakyReLU(0.2, False), nn.Conv2d(2 * nf, nf, kw, stride=2, padding=pw), SynchronizedBatchNorm2d(nf, affine=True), nn.LeakyReLU(0.2, False), nn.Conv2d(nf, int(nf // 2), kw, stride=2, padding=pw), SynchronizedBatchNorm2d(int(nf // 2), affine=True), nn.LeakyReLU(0.2, False)) #32*8*8 model = [nn.Linear(int(nf // 2) * 8 * 8, 100), SynchronizedBatchNorm1d(100, affine=True), nn.ReLU()] if opt.domain_rela: model += [nn.Linear(100, 1)] else: model += [nn.Linear(100, 2), nn.LogSoftmax(dim=1)] self.classifier = nn.Sequential(*model) def forward(self, x): x = self.feature(x) x = self.classifier(x.view(x.shape[0], -1)) return x class ReverseLayerF(Function): @staticmethod def forward(ctx, x, alpha): ctx.alpha = alpha return x.view_as(x) @staticmethod def backward(ctx, grad_output): output = grad_output.neg() * ctx.alpha return output, None class EMA(): def __init__(self, mu): self.mu = mu self.shadow = {} self.original = {} def register(self, name, val): self.shadow[name] = val.clone() def __call__(self, model): for name, param in model.named_parameters(): if param.requires_grad: assert name in self.shadow decay = self.mu new_average = (1.0 - decay) * param.data + decay * self.shadow[name] self.shadow[name] = new_average.clone() def assign(self, model): for name, param in model.named_parameters(): if param.requires_grad: assert name in self.shadow self.original[name] = param.data.clone() param.data = self.shadow[name] def resume(self, model): for name, param in model.named_parameters(): if param.requires_grad: assert name in self.shadow param.data = self.original[name]

CoCosNet/models/networks/generator.py/0

source_lang=python target_lang=python python run.py \ --model_name_or_path microsoft/unixcoder-base \ --query_data_file ../data/code_to_code_search_test.json \ --candidate_data_file ../data/code_to_code_search_test.json \ --trace_file ../saved_models/code_to_code_search/preds.txt \ --query_lang ${source_lang} \ --candidate_lang ${target_lang} \ --code_length 512 \ --eval_batch_size 256

CodeBERT/CodeExecutor/downstream/run.sh/0

import os import torch import logging import argparse import random import json from tqdm import tqdm import multiprocessing import time from itertools import cycle from torch.utils.data import DataLoader, RandomSampler, SequentialSampler from torch.utils.data import ConcatDataset from torch.utils.data.distributed import DistributedSampler from transformers import AdamW, get_linear_schedule_with_warmup from models import build_or_load_gen_model from configs import add_args, set_seed, set_dist from torch.nn.parallel import DistributedDataParallel as DDP import torch.distributed as dist from utils import CommentGenDataset, SimpleGenDataset from evaluator.smooth_bleu import bleu_fromstr logging.basicConfig( format="%(asctime)s - %(levelname)s - %(name)s - %(message)s", datefmt="%m/%d/%Y %H:%M:%S", level=logging.INFO, ) logger = logging.getLogger(__name__) def get_loaders(data_files, args, tokenizer, pool, eval=False): def fn(features): return features global_rank = args.global_rank for data_file in data_files: if args.raw_input: dataset = SimpleGenDataset(tokenizer, pool, args, data_file) else: dataset = CommentGenDataset(tokenizer, pool, args, data_file) data_len = len(dataset) if global_rank == 0: logger.info(f"Data length: {data_len}.") if eval: sampler = SequentialSampler(dataset) else: sampler = DistributedSampler(dataset) dataloader = DataLoader(dataset, sampler=sampler, batch_size=args.train_batch_size if not eval else args.eval_batch_size, \ num_workers=args.cpu_count, collate_fn=fn) yield dataset, sampler, dataloader def eval_bleu_epoch(args, eval_dataloader, model, tokenizer): logger.info(f" ***** Running bleu evaluation on {args.eval_file} *****") logger.info(" Batch size = %d", args.eval_batch_size) model.eval() if hasattr(model, "module"): model = model.module pred_ids, ex_ids = [], [] for step, examples in enumerate(eval_dataloader, 1): source_ids = torch.tensor( [ex.source_ids for ex in examples], dtype=torch.long ).to(args.local_rank) ids = [ex.example_id for ex in examples] source_mask = source_ids.ne(tokenizer.pad_id) preds = model.generate(source_ids, attention_mask=source_mask, use_cache=True, num_beams=args.beam_size, early_stopping=True, max_length=args.max_target_length) top_preds = list(preds.cpu().numpy()) pred_ids.extend(top_preds) # [1:] to remove beginning '<msg>' pred_nls = [tokenizer.decode(id[1:], skip_special_tokens=True, clean_up_tokenization_spaces=False) for id in pred_ids] valid_file = args.dev_filename golds = [] with open(valid_file, "r") as f: for line in f: golds.append(json.loads(line)["msg"]) golds = golds[:len(pred_nls)] bleu = bleu_fromstr(pred_nls, golds, rmstop=False) return bleu def save_model(model, optimizer, scheduler, output_dir, config): if not os.path.exists(output_dir): os.makedirs(output_dir) model_to_save = model.module if hasattr(model, "module") else model config.save_pretrained(output_dir) output_model_file = os.path.join(output_dir, "pytorch_model.bin") torch.save(model_to_save.state_dict(), output_model_file) output_optimizer_file = os.path.join(output_dir, "optimizer.pt") torch.save( optimizer.state_dict(), output_optimizer_file, _use_new_zipfile_serialization=False, ) output_scheduler_file = os.path.join(output_dir, "scheduler.pt") torch.save( scheduler.state_dict(), output_scheduler_file, _use_new_zipfile_serialization=False, ) def main(args): dist.init_process_group(backend="nccl") local_rank = dist.get_rank() % args.gpu_per_node args.global_rank = local_rank + args.node_index * args.gpu_per_node args.local_rank = local_rank args.world_size = dist.get_world_size() logger.warning("Process rank: %s, global rank: %s, world size: %s, bs: %s", args.local_rank, args.global_rank, \ torch.distributed.get_world_size(), \ args.train_batch_size) torch.cuda.set_device(local_rank) set_seed(args) config, model, tokenizer = build_or_load_gen_model(args) model = DDP(model.cuda(), device_ids=[local_rank], output_device=local_rank, find_unused_parameters=True) pool = multiprocessing.Pool(args.cpu_count) # Prepare optimizer and schedule (linear warmup and decay) no_decay = ["bias", "LayerNorm.weight"] optimizer_grouped_parameters = [ { "params": [ p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay) ], "weight_decay": args.weight_decay, }, { "params": [ p for n, p in model.named_parameters() if any(nd in n for nd in no_decay) ], "weight_decay": 0.0, }, ] optimizer = AdamW( optimizer_grouped_parameters, lr=args.learning_rate, eps=args.adam_epsilon ) args.warmup_steps = int(args.train_steps * 0.1) scheduler = get_linear_schedule_with_warmup( optimizer, num_warmup_steps=args.warmup_steps, num_training_steps=args.train_steps, ) if os.path.exists("{}/checkpoints-last/optimizer.pt".format(args.output_dir)): optimizer.load_state_dict( torch.load( "{}/checkpoints-last/optimizer.pt".format(args.output_dir), map_location="cpu", ) ) scheduler.load_state_dict( torch.load( "{}/checkpoints-last/scheduler.pt".format(args.output_dir), map_location="cpu", ) ) global_step = 0 save_steps = args.save_steps train_file = args.train_filename valid_file = args.dev_filename if os.path.isdir(train_file): train_files = [file for file in os.listdir(train_file) if file.startswith("train") and file.endswith(".jsonl")] else: train_files = [train_file] random.seed(args.seed) random.shuffle(train_files) train_files = [os.path.join(train_file, file) for file in train_files] valid_files = [valid_file] # bleu = eval_bleu_epoch(args, valid_dataloader, model, tokenizer) # logger.warning("Initial bleu: {}".format(bleu)) for epoch in range(1, args.train_epochs + 1): # set seed for reproducible data split save_seed = args.seed args.seed += epoch set_seed(args) args.seed = save_seed model.train() nb_tr_examples, nb_tr_steps, tr_loss = 0, 0, 0 for _, _, train_dataloader in get_loaders(train_files, args, tokenizer, pool): # WARNING: this is an iterator, to save memory for step, examples in enumerate(train_dataloader, 1): if step == 1: ex = examples[0] logger.info(f"batch size: {len(examples)}") logger.info(f"example source: {tokenizer.convert_ids_to_tokens(ex.source_ids)}") # logger.info(f"example label: {tokenizer.convert_ids_to_tokens(ex.source_labels)}") logger.info(f"example target: {tokenizer.convert_ids_to_tokens(ex.target_ids)}") source_ids = torch.tensor( [ex.source_ids for ex in examples], dtype=torch.long ).to(local_rank) source_labels = None target_ids = torch.tensor( [ex.target_ids for ex in examples], dtype=torch.long ).to(local_rank) source_mask = source_ids.ne(tokenizer.pad_id) target_mask = target_ids.ne(tokenizer.pad_id) loss = model( input_ids=source_ids, input_labels=source_labels, decoder_input_ids=target_ids, attention_mask=source_mask, decoder_attention_mask=target_mask, encoder_loss=False ) if args.gpu_per_node > 1: loss = loss.mean() # mean() to average on multi-gpu. if args.gradient_accumulation_steps > 1: loss = loss / args.gradient_accumulation_steps tr_loss += loss.item() nb_tr_examples += source_ids.size(0) nb_tr_steps += 1 loss.backward() if nb_tr_steps % args.gradient_accumulation_steps == 0: # Update parameters optimizer.step() optimizer.zero_grad() scheduler.step() global_step += 1 if args.global_rank == 0 and global_step % args.log_steps == 0: train_loss = round( tr_loss * args.gradient_accumulation_steps / nb_tr_steps, 4, ) logger.info( "step {}/{}: Train loss {}".format( global_step, args.train_steps, round(train_loss, 3), ) ) if global_step == args.train_steps and args.global_rank == 0: # end training _, _, valid_dataloader = next(get_loaders(valid_files, args, tokenizer, pool, eval=True)) bleu = eval_bleu_epoch(args, valid_dataloader, model, tokenizer) output_dir = os.path.join(args.output_dir, "checkpoints-last" + "-" + str(bleu)) save_model(model, optimizer, scheduler, output_dir, config) logger.info(f"Reach max steps {args.train_steps}.") time.sleep(5) return if args.global_rank == 0 and \ global_step % save_steps == 0 and \ nb_tr_steps % args.gradient_accumulation_steps == 0: _, _, valid_dataloader = next(get_loaders(valid_files, args, tokenizer, pool, eval=True)) bleu = eval_bleu_epoch(args, valid_dataloader, model, tokenizer) output_dir = os.path.join(args.output_dir, "checkpoints-" + str(global_step) + "-" + str(bleu)) save_model(model, optimizer, scheduler, output_dir, config) logger.info( "Save the {}-step model and optimizer into {}".format( global_step, output_dir ) ) time.sleep(5) if __name__ == "__main__": parser = argparse.ArgumentParser() args = add_args(parser) args.cpu_count = multiprocessing.cpu_count() # remove long tokenization warning. ref: https://github.com/huggingface/transformers/issues/991 logging.getLogger("transformers.tokenization_utils_base").setLevel(logging.ERROR) logger.info(args) main(args) logger.info("Training finished.") # torch.multiprocessing.spawn(main, args=(args,), nprocs=torch.cuda.device_count())

CodeBERT/CodeReviewer/code/run_finetune_msg.py/0

# Clone Detection ## Task Definition Given two codes as the input, the task is to do binary classification (0/1), where 1 stands for semantic equivalence and 0 for others. Models are evaluated by F1 score. ## Updates 2021-9-13: We have update the evaluater script. Since it's a binary classification, we use binary F1 score instead of "macro" F1 score. ## Dataset The dataset we use is [BigCloneBench](https://www.cs.usask.ca/faculty/croy/papers/2014/SvajlenkoICSME2014BigERA.pdf) and filtered following the paper [Detecting Code Clones with Graph Neural Network and Flow-Augmented Abstract Syntax Tree](https://arxiv.org/pdf/2002.08653.pdf). ### Data Format 1. dataset/data.jsonl is stored in jsonlines format. Each line in the uncompressed file represents one function. One row is illustrated below. - **func:** the function - **idx:** index of the example 2. train.txt/valid.txt/test.txt provide examples, stored in the following format: idx1 idx2 label ### Data Statistics Data statistics of the dataset are shown in the below table: | | #Examples | | ----- | :-------: | | Train | 901,028 | | Dev | 415,416 | | Test | 415,416 | You can get data using the following command. ``` unzip dataset.zip ``` ## Evaluator We provide a script to evaluate predictions for this task, and report F1 score ### Example ```bash python evaluator/evaluator.py -a evaluator/answers.txt -p evaluator/predictions.txt ``` {'Recall': 0.25, 'Prediction': 0.5, 'F1': 0.3333333333333333} ### Input predictions A predications file that has predictions in TXT format, such as evaluator/predictions.txt. For example: ```b 13653451 21955002 0 1188160 8831513 1 1141235 14322332 0 16765164 17526811 1 ``` ## Pipeline-GraphCodeBERT We also provide a pipeline that fine-tunes GraphCodeBERT on this task. ### Dependency - pip install torch - pip install transformers - pip install tree_sitter - pip sklearn ### Tree-sitter (optional) If the built file "parser/my-languages.so" doesn't work for you, please rebuild as the following command: ```shell cd parser bash build.sh cd .. ``` ### Fine-tune We use 4*V100-16G to fine-tune and 10% valid data to evaluate. ```shell mkdir saved_models python run.py \ --output_dir=saved_models \ --config_name=microsoft/graphcodebert-base \ --model_name_or_path=microsoft/graphcodebert-base \ --tokenizer_name=microsoft/graphcodebert-base \ --do_train \ --train_data_file=dataset/train.txt \ --eval_data_file=dataset/valid.txt \ --test_data_file=dataset/test.txt \ --epoch 1 \ --code_length 512 \ --data_flow_length 128 \ --train_batch_size 16 \ --eval_batch_size 32 \ --learning_rate 2e-5 \ --max_grad_norm 1.0 \ --evaluate_during_training \ --seed 123456 2>&1| tee saved_models/train.log ``` ### Inference We use full test data for inference. ```shell python run.py \ --output_dir=saved_models \ --config_name=microsoft/graphcodebert-base \ --model_name_or_path=microsoft/graphcodebert-base \ --tokenizer_name=microsoft/graphcodebert-base \ --do_eval \ --do_test \ --train_data_file=dataset/train.txt \ --eval_data_file=dataset/valid.txt \ --test_data_file=dataset/test.txt \ --epoch 1 \ --code_length 512 \ --data_flow_length 128 \ --train_batch_size 16 \ --eval_batch_size 32 \ --learning_rate 2e-5 \ --max_grad_norm 1.0 \ --evaluate_during_training \ --seed 123456 2>&1| tee saved_models/test.log ``` ### Evaluation ```shell python evaluator/evaluator.py -a dataset/test.txt -p saved_models/predictions.txt 2>&1| tee saved_models/score.log ``` ## Result The results on the test set are shown as below: | Method | Precision | Recall | F1 | | ------------- | :-------: | :-------: | :-------: | | Deckard | 0.93 | 0.02 | 0.03 | | RtvNN | 0.95 | 0.01 | 0.01 | | CDLH | 0.92 | 0.74 | 0.82 | | ASTNN | 0.92 | 0.94 | 0.93 | | FA-AST-GMN | **0.96** | 0.94 | 0.95 | | CodeBERT | 0.947 | 0.934 | 0.941 | | GraphCodeBERT | 0.948 | **0.952** | **0.950** |

CodeBERT/GraphCodeBERT/clonedetection/README.md/0

model=../../../../pretrained-model/UniXcoder-base mkdir saved_models CUDA_VISIBLE_DEVICES=0,1,2,3 python run.py \ --output_dir=./saved_models \ --model_type=roberta \ --model_name_or_path=$model \ --do_train \ --train_data_file=../../dataset/train.txt \ --eval_data_file=../../dataset/valid.txt \ --test_data_file=../../dataset/test.txt \ --epoch 1 \ --block_size 512 \ --train_batch_size 16 \ --eval_batch_size 32 \ --learning_rate 5e-5 \ --max_grad_norm 1.0 \ --evaluate_during_training \ --seed 123456 2>&1| tee saved_models/train.log CUDA_VISIBLE_DEVICES=0,1,2,3 python run.py \ --output_dir=./saved_models \ --model_type=roberta \ --model_name_or_path=$model \ --do_eval \ --do_test \ --train_data_file=../../dataset/train.txt \ --eval_data_file=../../dataset/valid.txt \ --test_data_file=../../dataset/test.txt \ --epoch 1 \ --block_size 512 \ --train_batch_size 16 \ --eval_batch_size 32 \ --learning_rate 5e-5 \ --max_grad_norm 1.0 \ --evaluate_during_training \ --seed 123456 2>&1| tee saved_models/test.log python ../evaluator/evaluator.py -a ../../dataset/test.txt -p saved_models/predictions.txt 2>&1| tee saved_models/score.log

CodeBERT/UniXcoder/downstream-tasks/clone-detection/BCB/run.sh/0

# coding=utf-8 # Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team. # Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """ Fine-tuning the library models for language modeling on a text file (GPT, GPT-2, BERT, RoBERTa). GPT and GPT-2 are fine-tuned using a causal language modeling (CLM) loss while BERT and RoBERTa are fine-tuned using a masked language modeling (MLM) loss. """ import sys import argparse import logging import os import pickle import random import torch import json import numpy as np from model import Model from torch.nn import CrossEntropyLoss, MSELoss from torch.utils.data import DataLoader, Dataset, SequentialSampler, RandomSampler,TensorDataset from transformers import (WEIGHTS_NAME, AdamW, get_linear_schedule_with_warmup, RobertaConfig, RobertaModel, RobertaTokenizer) logger = logging.getLogger(__name__) class InputFeatures(object): """A single training/test features for a example.""" def __init__(self, code_tokens, code_ids, nl_tokens, nl_ids, url, ): self.code_tokens = code_tokens self.code_ids = code_ids self.nl_tokens = nl_tokens self.nl_ids = nl_ids self.url = url def convert_examples_to_features(js,tokenizer,args): """convert examples to token ids""" code = ' '.join(js['code_tokens']) if type(js['code_tokens']) is list else ' '.join(js['code_tokens'].split()) code_tokens = tokenizer.tokenize(code)[:args.code_length-4] code_tokens =[tokenizer.cls_token,"<encoder-only>",tokenizer.sep_token]+code_tokens+[tokenizer.sep_token] code_ids = tokenizer.convert_tokens_to_ids(code_tokens) padding_length = args.code_length - len(code_ids) code_ids += [tokenizer.pad_token_id]*padding_length nl = ' '.join(js['docstring_tokens']) if type(js['docstring_tokens']) is list else ' '.join(js['doc'].split()) nl_tokens = tokenizer.tokenize(nl)[:args.nl_length-4] nl_tokens = [tokenizer.cls_token,"<encoder-only>",tokenizer.sep_token]+nl_tokens+[tokenizer.sep_token] nl_ids = tokenizer.convert_tokens_to_ids(nl_tokens) padding_length = args.nl_length - len(nl_ids) nl_ids += [tokenizer.pad_token_id]*padding_length return InputFeatures(code_tokens,code_ids,nl_tokens,nl_ids,js['url'] if "url" in js else js["retrieval_idx"]) class TextDataset(Dataset): def __init__(self, tokenizer, args, file_path=None): self.examples = [] data = [] with open(file_path) as f: if "jsonl" in file_path: for line in f: line = line.strip() js = json.loads(line) if 'function_tokens' in js: js['code_tokens'] = js['function_tokens'] data.append(js) elif "codebase"in file_path or "code_idx_map" in file_path: js = json.load(f) for key in js: temp = {} temp['code_tokens'] = key.split() temp["retrieval_idx"] = js[key] temp['doc'] = "" temp['docstring_tokens'] = "" data.append(temp) elif "json" in file_path: for js in json.load(f): data.append(js) for js in data: self.examples.append(convert_examples_to_features(js,tokenizer,args)) if "train" in file_path: for idx, example in enumerate(self.examples[:3]): logger.info("*** Example ***") logger.info("idx: {}".format(idx)) logger.info("code_tokens: {}".format([x.replace('\u0120','_') for x in example.code_tokens])) logger.info("code_ids: {}".format(' '.join(map(str, example.code_ids)))) logger.info("nl_tokens: {}".format([x.replace('\u0120','_') for x in example.nl_tokens])) logger.info("nl_ids: {}".format(' '.join(map(str, example.nl_ids)))) def __len__(self): return len(self.examples) def __getitem__(self, i): return (torch.tensor(self.examples[i].code_ids),torch.tensor(self.examples[i].nl_ids)) def set_seed(seed=42): random.seed(seed) os.environ['PYHTONHASHSEED'] = str(seed) np.random.seed(seed) torch.manual_seed(seed) torch.cuda.manual_seed(seed) torch.backends.cudnn.deterministic = True def train(args, model, tokenizer): """ Train the model """ #get training dataset train_dataset = TextDataset(tokenizer, args, args.train_data_file) train_sampler = RandomSampler(train_dataset) train_dataloader = DataLoader(train_dataset, sampler=train_sampler, batch_size=args.train_batch_size,num_workers=4) #get optimizer and scheduler optimizer = AdamW(model.parameters(), lr=args.learning_rate, eps=1e-8) scheduler = get_linear_schedule_with_warmup(optimizer, num_warmup_steps = 0, num_training_steps = len(train_dataloader) * args.num_train_epochs) # Train! logger.info("***** Running training *****") logger.info(" Num examples = %d", len(train_dataset)) logger.info(" Num Epochs = %d", args.num_train_epochs) logger.info(" Instantaneous batch size per GPU = %d", args.train_batch_size//args.n_gpu) logger.info(" Total train batch size = %d", args.train_batch_size) logger.info(" Total optimization steps = %d", len(train_dataloader)*args.num_train_epochs) # model.resize_token_embeddings(len(tokenizer)) model.zero_grad() model.train() tr_num,tr_loss,best_mrr = 0,0,0 for idx in range(args.num_train_epochs): for step,batch in enumerate(train_dataloader): #get inputs code_inputs = batch[0].to(args.device) nl_inputs = batch[1].to(args.device) #get code and nl vectors code_vec = model(code_inputs=code_inputs) nl_vec = model(nl_inputs=nl_inputs) #calculate scores and loss scores = torch.einsum("ab,cb->ac",nl_vec,code_vec) loss_fct = CrossEntropyLoss() loss = loss_fct(scores*20, torch.arange(code_inputs.size(0), device=scores.device)) #report loss tr_loss += loss.item() tr_num += 1 if (step+1)%100 == 0: logger.info("epoch {} step {} loss {}".format(idx,step+1,round(tr_loss/tr_num,5))) tr_loss = 0 tr_num = 0 #backward loss.backward() torch.nn.utils.clip_grad_norm_(model.parameters(), args.max_grad_norm) optimizer.step() optimizer.zero_grad() scheduler.step() #evaluate results = evaluate(args, model, tokenizer,args.eval_data_file, eval_when_training=True) for key, value in results.items(): logger.info(" %s = %s", key, round(value,4)) #save best model if results['eval_mrr']>best_mrr: best_mrr = results['eval_mrr'] logger.info(" "+"*"*20) logger.info(" Best mrr:%s",round(best_mrr,4)) logger.info(" "+"*"*20) checkpoint_prefix = 'checkpoint-best-mrr' output_dir = os.path.join(args.output_dir, '{}'.format(checkpoint_prefix)) if not os.path.exists(output_dir): os.makedirs(output_dir) model_to_save = model.module if hasattr(model,'module') else model output_dir = os.path.join(output_dir, '{}'.format('model.bin')) torch.save(model_to_save.state_dict(), output_dir) logger.info("Saving model checkpoint to %s", output_dir) def evaluate(args, model, tokenizer,file_name,eval_when_training=False): query_dataset = TextDataset(tokenizer, args, file_name) query_sampler = SequentialSampler(query_dataset) query_dataloader = DataLoader(query_dataset, sampler=query_sampler, batch_size=args.eval_batch_size,num_workers=4) code_dataset = TextDataset(tokenizer, args, args.codebase_file) code_sampler = SequentialSampler(code_dataset) code_dataloader = DataLoader(code_dataset, sampler=code_sampler, batch_size=args.eval_batch_size,num_workers=4) # Eval! logger.info("***** Running evaluation *****") logger.info(" Num queries = %d", len(query_dataset)) logger.info(" Num codes = %d", len(code_dataset)) logger.info(" Batch size = %d", args.eval_batch_size) model.eval() code_vecs = [] nl_vecs = [] for batch in query_dataloader: nl_inputs = batch[1].to(args.device) with torch.no_grad(): nl_vec = model(nl_inputs=nl_inputs) nl_vecs.append(nl_vec.cpu().numpy()) for batch in code_dataloader: code_inputs = batch[0].to(args.device) with torch.no_grad(): code_vec = model(code_inputs=code_inputs) code_vecs.append(code_vec.cpu().numpy()) model.train() code_vecs = np.concatenate(code_vecs,0) nl_vecs = np.concatenate(nl_vecs,0) scores = np.matmul(nl_vecs,code_vecs.T) sort_ids = np.argsort(scores, axis=-1, kind='quicksort', order=None)[:,::-1] nl_urls = [] code_urls = [] for example in query_dataset.examples: nl_urls.append(example.url) for example in code_dataset.examples: code_urls.append(example.url) ranks = [] for url, sort_id in zip(nl_urls,sort_ids): rank = 0 find = False for idx in sort_id[:1000]: if find is False: rank += 1 if code_urls[idx] == url: find = True if find: ranks.append(1/rank) else: ranks.append(0) result = { "eval_mrr":float(np.mean(ranks)) } return result def main(): parser = argparse.ArgumentParser() ## Required parameters parser.add_argument("--train_data_file", default=None, type=str, help="The input training data file (a json file).") parser.add_argument("--output_dir", default=None, type=str, required=True, help="The output directory where the model predictions and checkpoints will be written.") parser.add_argument("--eval_data_file", default=None, type=str, help="An optional input evaluation data file to evaluate the MRR(a jsonl file).") parser.add_argument("--test_data_file", default=None, type=str, help="An optional input test data file to test the MRR(a josnl file).") parser.add_argument("--codebase_file", default=None, type=str, help="An optional input test data file to codebase (a jsonl file).") parser.add_argument("--model_name_or_path", default=None, type=str, help="The model checkpoint for weights initialization.") parser.add_argument("--config_name", default="", type=str, help="Optional pretrained config name or path if not the same as model_name_or_path") parser.add_argument("--tokenizer_name", default="", type=str, help="Optional pretrained tokenizer name or path if not the same as model_name_or_path") parser.add_argument("--nl_length", default=128, type=int, help="Optional NL input sequence length after tokenization.") parser.add_argument("--code_length", default=256, type=int, help="Optional Code input sequence length after tokenization.") parser.add_argument("--do_train", action='store_true', help="Whether to run training.") parser.add_argument("--do_eval", action='store_true', help="Whether to run eval on the dev set.") parser.add_argument("--do_test", action='store_true', help="Whether to run eval on the test set.") parser.add_argument("--do_zero_shot", action='store_true', help="Whether to run eval on the test set.") parser.add_argument("--do_F2_norm", action='store_true', help="Whether to run eval on the test set.") parser.add_argument("--train_batch_size", default=4, type=int, help="Batch size for training.") parser.add_argument("--eval_batch_size", default=4, type=int, help="Batch size for evaluation.") parser.add_argument("--learning_rate", default=5e-5, type=float, help="The initial learning rate for Adam.") parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.") parser.add_argument("--num_train_epochs", default=1, type=int, help="Total number of training epochs to perform.") parser.add_argument('--seed', type=int, default=42, help="random seed for initialization") #print arguments args = parser.parse_args() #set log logging.basicConfig(format='%(asctime)s - %(levelname)s - %(name)s - %(message)s', datefmt='%m/%d/%Y %H:%M:%S',level=logging.INFO ) #set device device = torch.device("cuda" if torch.cuda.is_available() else "cpu") args.n_gpu = torch.cuda.device_count() args.device = device logger.info("device: %s, n_gpu: %s",device, args.n_gpu) # Set seed set_seed(args.seed) #build model tokenizer = RobertaTokenizer.from_pretrained(args.model_name_or_path) config = RobertaConfig.from_pretrained(args.model_name_or_path) model = RobertaModel.from_pretrained(args.model_name_or_path) model = Model(model) logger.info("Training/evaluation parameters %s", args) model.to(args.device) if args.n_gpu > 1: model = torch.nn.DataParallel(model) # Training if args.do_train: train(args, model, tokenizer) # Evaluation results = {} if args.do_eval: if args.do_zero_shot is False: checkpoint_prefix = 'checkpoint-best-mrr/model.bin' output_dir = os.path.join(args.output_dir, '{}'.format(checkpoint_prefix)) model_to_load = model.module if hasattr(model, 'module') else model model_to_load.load_state_dict(torch.load(output_dir)) model.to(args.device) result = evaluate(args, model, tokenizer,args.eval_data_file) logger.info("***** Eval results *****") for key in sorted(result.keys()): logger.info(" %s = %s", key, str(round(result[key],3))) if args.do_test: if args.do_zero_shot is False: checkpoint_prefix = 'checkpoint-best-mrr/model.bin' output_dir = os.path.join(args.output_dir, '{}'.format(checkpoint_prefix)) model_to_load = model.module if hasattr(model, 'module') else model model_to_load.load_state_dict(torch.load(output_dir)) model.to(args.device) result = evaluate(args, model, tokenizer,args.test_data_file) logger.info("***** Eval results *****") for key in sorted(result.keys()): logger.info(" %s = %s", key, str(round(result[key],3))) if __name__ == "__main__": main()

CodeBERT/UniXcoder/downstream-tasks/code-search/run.py/0

{ "train_micro_batch_size_per_gpu": 8, "gradient_accumulation_steps": 1, "fp16": { "enabled": false }, "zero_optimization": { "stage": 1, "reduce_bucket_size": 5e8 } }

CodeT/DIVERSE/code/src/ds_config.json/0

include CONTRIBUTING.md include LICENSE-IMAGE.md include LICENSE.md include README.md include ThirdPartyNotices.txt

Cognitive-Face-Python/MANIFEST.in/0

#!/usr/bin/env python # -*- coding: utf-8 -*- """ File: test_face.py Description: Unittests for Face section of the Cognitive Face API. """ import unittest import cognitive_face as CF from . import util class TestFace(unittest.TestCase): """Unittests for Face section.""" def test_detect(self): """Unittest for `face.detect`.""" image = '{}detection1.jpg'.format(util.BASE_URL_IMAGE) res = CF.face.detect(image) print(res) self.assertIsInstance(res, list) util.wait() def test_find_similars_face_ids(self): """Unittest for `face.find_similars` with face ids.""" res = CF.face.find_similars( util.DataStore.face_id, face_ids=util.DataStore.face_ids) print(res) self.assertIsInstance(res, list) util.wait() def test_find_similars_face_list(self): """Unittest for `face.find_similars` in face list.""" res = CF.face.find_similars( util.DataStore.face_id, face_list_id=util.DataStore.face_list_id) print(res) self.assertIsInstance(res, list) util.wait() def test_find_similars_large_face_list(self): """Unittest for `face.find_similars` in large face list.""" CF.util.wait_for_large_face_list_training( util.DataStore.large_face_list_id) res = CF.face.find_similars( util.DataStore.face_id, large_face_list_id=util.DataStore.large_face_list_id) print(res) self.assertIsInstance(res, list) util.wait() def test_group(self): """Unittest for `face.group`.""" temp_face_ids = util.DataStore.face_ids temp_face_ids.append(util.DataStore.face_id) temp_face_ids.append(util.DataStore.another_face_id) res = CF.face.group(temp_face_ids) print(res) self.assertIsInstance(res, dict) util.wait() def test_identify_person_group(self): """Unittest for `face.identify` in person gorup.""" CF.util.wait_for_person_group_training(util.DataStore.person_group_id) res = CF.face.identify( util.DataStore.face_ids, person_group_id=util.DataStore.person_group_id) print(res) self.assertIsInstance(res, list) util.wait() def test_identify_large_person_group(self): """Unittest for `face.identify` in large person gorup.""" CF.util.wait_for_large_person_group_training( util.DataStore.large_person_group_id) res = CF.face.identify( util.DataStore.face_ids, large_person_group_id=util.DataStore.large_person_group_id) print(res) self.assertIsInstance(res, list) util.wait() def test_verify_face_ids(self): """Unittest for `face.verify` with face ids.""" res = CF.face.verify( util.DataStore.face_id, another_face_id=util.DataStore.another_face_id) print(res) self.assertIsInstance(res, dict) util.wait() def test_verify_person_group(self): """Unittest for `face.verify` in person group.""" CF.util.wait_for_person_group_training(util.DataStore.person_group_id) res = CF.face.verify( util.DataStore.face_id, person_group_id=util.DataStore.person_group_id, person_id=util.DataStore.person_id['Dad']) print(res) self.assertIsInstance(res, dict) util.wait() def test_verify_large_person_group(self): """Unittest for `face.verify` in large person group.""" CF.util.wait_for_large_person_group_training( util.DataStore.large_person_group_id) res = CF.face.verify( util.DataStore.face_id, large_person_group_id=util.DataStore.large_person_group_id, person_id=util.DataStore.large_person_group_person_id['Dad']) print(res) self.assertIsInstance(res, dict) util.wait() if __name__ == '__main__': unittest.main()

Cognitive-Face-Python/cognitive_face/tests/test_face.py/0

#!/usr/bin/env python # -*- coding: utf-8 -*- """ File: __init__.py Description: View components for Python SDK sample. """ import wx import wx.lib.agw.labelbook as LB from wx.lib.agw.fmresources import INB_FIT_LABELTEXT from wx.lib.agw.fmresources import INB_LEFT from wx.lib.agw.fmresources import INB_NO_RESIZE from view.panel_detection import DetectionPanel from view.panel_subscription import SubscriptionPanel from view.panel_find_similar import FindSimilarPanel from view.panel_group import GroupPanel from view.panel_identification import IdentificationPanel from view.panel_verification import VerificationPanel TITLE = u"Microsoft Cognitive Services Face Samples" class MyLabelBook(LB.LabelBook): """LabelBook part in Main Frame.""" def __init__(self, parent): agw_style = INB_LEFT | INB_FIT_LABELTEXT | INB_NO_RESIZE super(MyLabelBook, self).__init__(parent, agwStyle=agw_style) subscription_panel = SubscriptionPanel(self) subscription_text = u"Subscription Key Management" self.AddPage(subscription_panel, subscription_text, True) self.AddPage(wx.Panel(self), u"Select a scenario:") self.EnableTab(1, False) self.AddPage(DetectionPanel(self), u" - Face Detection") self.AddPage(FindSimilarPanel(self), u" - Face Find Similar") self.AddPage(GroupPanel(self), u" - Face Grouping") self.AddPage(IdentificationPanel(self), u" - Face Identification") self.AddPage(VerificationPanel(self), u" - Face Verification") class MyTitle(wx.Panel): """Title part in Main Frame.""" def __init__(self, parent): super(MyTitle, self).__init__(parent) self.SetBackgroundColour('#00b294') self.SetMinSize((-1, 80)) sizer = wx.BoxSizer() sizer.AddStretchSpacer() family = wx.FONTFAMILY_DEFAULT style = wx.FONTSTYLE_NORMAL weight = wx.FONTWEIGHT_NORMAL font = wx.Font(20, family, style, weight) self.text = wx.StaticText(self, label=TITLE, style=wx.ALIGN_CENTER) self.text.SetFont(font) sizer.Add(self.text, flag=wx.ALIGN_CENTER_VERTICAL) sizer.AddStretchSpacer() self.SetSizer(sizer) class MyFrame(wx.Frame): """Main Frame.""" def __init__(self, parent): super(MyFrame, self).__init__(parent, title=TITLE, size=(1280, 768)) icon_path = 'Assets/Microsoft-logo_rgb_c-gray.png' self.SetIcon(wx.Icon(icon_path)) sizer = wx.BoxSizer(wx.VERTICAL) self.title = MyTitle(self) sizer.Add(self.title, flag=wx.EXPAND) self.book = MyLabelBook(self) sizer.Add(self.book, 1, flag=wx.EXPAND) status_text = ( 'Microsoft will receive the images you upload and may use them to ' 'improve Face API and related services. By submitting an image, ' 'you confirm you have consent from everyone in it.') self.status = wx.StatusBar(self) self.status.SetStatusText(status_text) sizer.Add(self.status, flag=wx.EXPAND) self.SetSizer(sizer) self.Layout() class MyApp(wx.App): """The whole app.""" def OnInit(self): """Show main frame.""" frame = MyFrame(None) frame.Show() return True

Cognitive-Face-Python/sample/view/__init__.py/0

export CUDA_VISIBLE_DEVICES=0 python t5_run_train.py \ --model_name_or_path t5-base \ --subtask Com \ --method MainExp \ --train_file pretrain \ --max_steps 100000 \ --save_steps 100000 \ --batch_size 8 \ --ebatch_size 16 \ --gas 1 \ --seed 1 \ --set set1

ContextualSP/abstraction_probing/code/t5_code/Com_MainExp_pretrain.sh/0

import pdb import subprocess import argparse import os def run_command(bash_command): process = subprocess.Popen(bash_command.split()) output, error = process.communicate() print(error) print(output) if __name__ == "__main__": parser = argparse.ArgumentParser() parser.add_argument("--model_name_or_path", type=str, default="", help="model_name_or_path") parser.add_argument("--output_dir", type=str, default="./checkpoint/", help="output dir") parser.add_argument("--train_file", type=str, default='pretrain', help="train file") parser.add_argument("--validation_file", type=str, default='test', help="validation file") parser.add_argument("--max_steps", type=int, default=100000, help="max_steps") parser.add_argument("--batch_size", type=int, default=8, help="batch_size") parser.add_argument("--ebatch_size", type=int, default=16, help="eval batch_size") parser.add_argument("--learning_rate", type=float, default=1e-5, help="learning_rate") parser.add_argument("--weight_decay", type=float, default=1e-2, help="weight_decay") parser.add_argument("--gas", type=int, default=1, help="gradient_accumulation_steps") parser.add_argument("--save_steps", type=int, default=100000, help="save_steps") parser.add_argument("--device_num", type=int, default=1, help="device_num") parser.add_argument("--method", type=str, default='MainExp', help="method") parser.add_argument("--seed", type=int, default=1, help="seed") parser.add_argument("--init_weights", type=bool, default=False, help="init_weights") parser.add_argument("--subtask", type=str, default='Com', help="subtask") parser.add_argument("--set", type=str, default='set1', help="subtask") args = parser.parse_args() print("START training") run_command("printenv") output_dir = './checkpoint/' + args.subtask + '/' + args.method + '_' + args.train_file + '_' + args.set + '_seed' + str(args.seed) # ./checkpoint/Com/MainExp_pretrain_set1_seed1 train_file = '../../data/' + args.subtask + '/' + args.set + '/' + args.train_file + '.json' # .../data/Com/set1/pretrain.json validation_file = '../../data/' + args.subtask + '/' + args.set + '/' + args.validation_file + '.json' # .../data/Com/set1/test.json cmd = f""" python -m torch.distributed.launch --nproc_per_node {args.device_num} --master_port=12343 t5_train_model.py \ --model_name_or_path {args.model_name_or_path} \ --output_dir {output_dir} \ --do_train \ --do_eval \ --train_file {train_file} \ --validation_file {validation_file} \ --per_device_train_batch_size {args.batch_size} \ --per_device_eval_batch_size {args.ebatch_size} \ --overwrite_output_dir \ --gradient_accumulation_steps {args.gas} \ --max_steps {args.max_steps} \ --logging_steps 10 \ --learning_rate {args.learning_rate} \ --save_steps {args.save_steps} \ --eval_steps {args.save_steps} \ --evaluation_strategy steps \ --freeze_model_parameter False \ --weight_decay {args.weight_decay} \ --label_smoothing_factor 0.1 \ --lr_scheduler_type constant \ --fp16 False \ --predict_with_generate \ --num_beams 5 \ --seed {args.seed} \ --adafactor False \ --max_source_length 1024 \ --max_target_length 1024 \ --gradient_checkpointing False \ --init_weights {args.init_weights} """ print("RUN {}".format(cmd)) run_command(cmd)

ContextualSP/abstraction_probing/code/t5_code/t5_run_train.py/0

# define common functions

ContextualSP/adaptershare/experiments/__init__.py/0

#pretrain config mlm: data_format: MLM enable_san: false metric_meta: - ACC n_class: 30522 task_type: MaskLM loss: MlmCriterion kd_loss: MseCriterion adv_loss: SymKlCriterion

ContextualSP/adaptershare/experiments/mlm/mlm.yml/0

import os import argparse import random from sys import path path.append(os.getcwd()) from experiments.common_utils import dump_rows from data_utils.task_def import DataFormat from data_utils.log_wrapper import create_logger from experiments.glue.glue_utils import * logger = create_logger(__name__, to_disk=True, log_file="xnli_prepro.log") def load_xnli(file, header=True): rows = [] cnt = 0 with open(file, encoding="utf8") as f: for line in f: if header: header = False continue blocks = line.strip().split("\t") if blocks[1] == "-": continue lab = blocks[1] if lab is None: import pdb pdb.set_trace() sample = { "uid": blocks[9], "premise": blocks[6], "hypothesis": blocks[7], "label": lab, "lang": blocks[0], } rows.append(sample) cnt += 1 return rows def parse_args(): parser = argparse.ArgumentParser(description="Preprocessing XNLI dataset.") parser.add_argument("--seed", type=int, default=13) parser.add_argument("--root_dir", type=str, default="data") args = parser.parse_args() return args def main(args): root = args.root_dir assert os.path.exists(root) ###################################### # XNLI/SciTail Tasks ###################################### xnli_dev_path = os.path.join(root, "XNLI/xnli.dev.tsv") xnli_test_path = os.path.join(root, "XNLI/xnli.test.tsv") ###################################### # Loading DATA ###################################### xnli_dev_data = load_xnli(xnli_dev_path) xnli_test_data = load_xnli(xnli_test_path) logger.info("Loaded {} XNLI train samples".format(len(xnli_dev_data))) logger.info("Loaded {} XNLI test samples".format(len(xnli_test_data))) canonical_data_suffix = "canonical_data" canonical_data_root = os.path.join(root, canonical_data_suffix) if not os.path.isdir(canonical_data_root): os.mkdir(canonical_data_root) # BUILD XNLI xnli_dev_fout = os.path.join(canonical_data_root, "xnli_dev.tsv") xnli_test_fout = os.path.join(canonical_data_root, "xnli_test.tsv") dump_rows(xnli_dev_data, xnli_dev_fout, DataFormat.PremiseAndOneHypothesis) dump_rows(xnli_test_data, xnli_test_fout, DataFormat.PremiseAndOneHypothesis) logger.info("done with XNLI") if __name__ == "__main__": args = parse_args() main(args)

ContextualSP/adaptershare/experiments/xnli/xnli_prepro.py/0

# coding=utf-8 # Copyright (c) Microsoft. All rights reserved. # This is a quick hack of adamaxw by xiaodong liu import math import torch from torch.optim import Optimizer class AdamaxW(Optimizer): r"""Implements AdamaxW algorithm. The original Adam algorithm was proposed in `Adam: A Method for Stochastic Optimization`_. The AdamW variant was proposed in `Decoupled Weight Decay Regularization`_. Arguments: params (iterable): iterable of parameters to optimize or dicts defining parameter groups lr (float, optional): learning rate (default: 1e-3) betas (Tuple[float, float], optional): coefficients used for computing running averages of gradient and its square (default: (0.9, 0.999)) eps (float, optional): term added to the denominator to improve numerical stability (default: 1e-8) weight_decay (float, optional): weight decay coefficient (default: 1e-2) amsgrad (boolean, optional): whether to use the AMSGrad variant of this algorithm from the paper `On the Convergence of Adam and Beyond`_ (default: False) .. _Adamax\: A Method for Stochastic Optimization: https://arxiv.org/abs/1412.6980 .. _Decoupled Weight Decay Regularization: https://arxiv.org/abs/1711.05101 .. _On the Convergence of Adam and Beyond: https://openreview.net/forum?id=ryQu7f-RZ """ def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8, weight_decay=1e-2, amsgrad=False): if not 0.0 <= lr: raise ValueError("Invalid learning rate: {}".format(lr)) if not 0.0 <= eps: raise ValueError("Invalid epsilon value: {}".format(eps)) if not 0.0 <= betas[0] < 1.0: raise ValueError("Invalid beta parameter at index 0: {}".format(betas[0])) if not 0.0 <= betas[1] < 1.0: raise ValueError("Invalid beta parameter at index 1: {}".format(betas[1])) defaults = dict(lr=lr, betas=betas, eps=eps, weight_decay=weight_decay, amsgrad=amsgrad) super(AdamaxW, self).__init__(params, defaults) def __setstate__(self, state): super(AdamaxW, self).__setstate__(state) for group in self.param_groups: group.setdefault('amsgrad', False) def step(self, closure=None): """Performs a single optimization step. Arguments: closure (callable, optional): A closure that reevaluates the model and returns the loss. """ loss = None if closure is not None: loss = closure() for group in self.param_groups: for p in group['params']: if p.grad is None: continue # Perform stepweight decay p.data.mul_(1 - group['lr'] * group['weight_decay']) # Perform optimization step grad = p.grad.data if grad.is_sparse: raise RuntimeError('Adam does not support sparse gradients, please consider SparseAdam instead') amsgrad = group['amsgrad'] state = self.state[p] eps = group['eps'] # State initialization if len(state) == 0: state['step'] = 0 # Exponential moving average of gradient values state['exp_avg'] = torch.zeros_like(p.data) # Exponential moving average of inf gradient values state['exp_inf'] = torch.zeros_like(p.data) exp_avg, exp_inf = state['exp_avg'], state['exp_inf'] beta1, beta2 = group['betas'] state['step'] += 1 # Decay the first and inf moment running average coefficient exp_avg.mul_(beta1).add_(grad, alpha=1 - beta1) norm_buf = torch.cat([ exp_inf.mul_(beta2).unsqueeze(0), grad.abs().add_(eps).unsqueeze_(0) ], 0) torch.max(norm_buf, 0, keepdim=False, out=(exp_inf, exp_inf.new().long())) bias_correction = 1 - beta1 ** state['step'] clr = group['lr'] / bias_correction p.data.addcdiv_(exp_avg, exp_inf, value=-clr) return loss

ContextualSP/adaptershare/mt_dnn/optim.py/0

import os import argparse import torch import json from models import * from utils import * from tqdm import tqdm def load_model_and_data_iter(args): ckpt_path = args.checkpoint device = torch.device(args.device) config = json.load(open(os.path.join(os.path.dirname(ckpt_path), 'config.json'), 'r', encoding='utf-8')) config['checkpoint'] = ckpt_path config['device'] = device model = load_model_from_checkpoint(**config) print('load model from {} over.'.format(ckpt_path)) model.eval() print('-------------------Config-------------------') for key, val in config.items(): print(key, val) print('load {} from {} over .'.format(config['model'], ckpt_path)) bert_version = config['bert_version'] tokenizer = BertTokenizer.from_pretrained(bert_version) print('load {} tokenizer over'.format(bert_version)) return config, model, tokenizer # load schemas from database def get_column_names_unique(column_names: List[Tuple[int, str]], table_names: List[str], primary_keys: List[int]) -> List[str]: column_names_dict = defaultdict(int) for tbl_idx, col_name in column_names: column_names_dict[col_name] += 1 column_names_unique = [] for c_idx, (tbl_idx, col_name) in enumerate(column_names): if tbl_idx == -1: column_names_unique.append(col_name) continue if column_names_dict[col_name] == 1: column_names_unique.append(col_name) elif c_idx in primary_keys: column_names_unique.append(col_name) else: tbl_name = table_names[tbl_idx] full_name = '{} . {}'.format(tbl_name, col_name) column_names_unique.append(full_name) assert len(column_names_unique) == len(column_names) return column_names_unique def alt_tbl_name(tbl_name): tbl_name = tbl_name.split() if len(tbl_name) > 1 and tbl_name[0] == 'reference': tbl_name = tbl_name[1:] if len(tbl_name) > 1 and tbl_name[-1] == 'data': tbl_name = tbl_name[:-1] if len(tbl_name) > 1 and tbl_name[-1] == 'list': tbl_name = tbl_name[:-1] return ' '.join(tbl_name) def remove_shared_prefix(col_name: str, tbl_name: str) -> str: col_tokens, tbl_tokens = col_name.split(), tbl_name.split() idx = 0 while idx < len(col_tokens) and idx < len(tbl_tokens) and col_tokens[idx] == tbl_tokens[idx]: idx += 1 return " ".join(col_tokens[idx:]) def get_column_name_normalized(column_lem_names: List[Tuple[int, str]], table_lem_names: List[str], verbose: bool = False): column_norm_names, table_norm_names = [], [] for tbl_name in table_lem_names: table_norm_names.append(alt_tbl_name(tbl_name)) for col_idx, (tbl_idx, col_name) in enumerate(column_lem_names): if col_name == '*': column_norm_names.append('*') continue col_norm_name = remove_shared_prefix(col_name, table_norm_names[tbl_idx]) if col_norm_name != col_name and verbose: logging.info(" {}\t{}\t{}".format(table_norm_names[tbl_idx], col_name, col_norm_name)) column_norm_names.append(col_norm_name) return column_norm_names, table_norm_names def load_schema(obj: Dict) -> SpiderSchema: column_names_lemma = obj['column_names_lemma'] table_names_lemma = obj['table_names_lemma'] column_names_original = [x[1] for x in obj['column_names_original']] column_to_table, table_to_columns = {}, {} for col_idx, (tbl_idx, _) in enumerate(obj['column_names']): if tbl_idx not in table_to_columns: table_to_columns[tbl_idx] = [] table_to_columns[tbl_idx].append(col_idx) column_to_table[col_idx] = tbl_idx col_norm_names, tbl_norm_names = get_column_name_normalized(column_names_lemma, table_names_lemma, True) return SpiderSchema( db_id=obj['db_id'], column_names=col_norm_names, column_types=obj['column_types'], column_names_lemma=[x[1] for x in column_names_lemma], column_names_original=column_names_original, table_names=tbl_norm_names, table_names_lemma=table_names_lemma, table_names_original=obj['table_names_original'], table_to_columns=table_to_columns, column_to_table=column_to_table, primary_keys=obj['primary_keys'], foreign_keys=obj['foreign_keys']) def load_schemas(path: str): databases = json.load(open(path, 'r', encoding='utf-8')) schemas = {} for database in databases: schema = load_schema(database) schemas[schema.db_id] = schema return schemas def process_examples(input_path: str, database_dir: str, tokenizer: BertTokenizer, table_path: str, output_path): schemas = load_schemas(table_path) print('load schemas over.') value_matchers = load_value_matchers(database_dir, schemas) print('load value matchers over') processed_examples = [] for raw_example in tqdm(json.load(open(input_path, 'r', encoding='utf-8'))): db_id = raw_example['db_id'] assert db_id in schemas processed_example = process_slsql_example(raw_example, tokenizer, schemas[db_id], value_matchers[db_id]) processed_examples += [processed_example] save_json_objects(processed_examples, output_path) print('process examples over, save into {}'.format(output_path)) def fix_tok(tok): tok = tok.lower() if tok == '-lrb-': tok = '(' elif tok == '-rrb-': tok = ')' elif tok == '\"': tok = '\'' return tok spider_type_mappings = { 'text': 'text', 'time': 'time', 'number': 'number', 'boolean': 'boolean', 'others': 'text' } def get_data_type(db_data_type: str): if db_data_type.startswith("int") or db_data_type.startswith("bigint") or db_data_type.startswith("mediumint"): return "int" if db_data_type.startswith("smallint") or db_data_type.startswith("tinyint") or db_data_type.startswith("bit") or db_data_type.startswith("bool") : return "int" if db_data_type.startswith("real") or db_data_type.startswith("numeric") or db_data_type.startswith("number"): return "real" if db_data_type.startswith("double") or db_data_type.startswith("decimal") or db_data_type.startswith("float"): return "real" if db_data_type.startswith("text") or db_data_type.startswith("varchar") or db_data_type.startswith("char"): return "text" if db_data_type.startswith("timestamp") or db_data_type.startswith("date") or db_data_type.startswith("year"): return "datetime" if len(db_data_type) == 0 or db_data_type.startswith("blob"): return "text" return 'text' #raise ValueError("not support data type: " + db_data_type) def get_column_with_values(path: str): column_values = defaultdict(list) try: conn = sqlite3.connect(path) conn.text_factory = lambda b: b.decode(errors = 'ignore') cur = conn.cursor() cur.execute("SELECT name FROM sqlite_master WHERE type='table';") tables =[x[0] for x in cur.fetchall()] for table in tables: col_results = cur.execute("PRAGMA table_info('%s')" % table).fetchall() columns = [] for col in col_results: col_name = col[1] data_type = get_data_type(col[2].lower()) columns.append((col_name, data_type)) assert len(columns) > 0 # get rows cur.execute("SELECT * FROM " + table + ";") row_results = cur.fetchall() rows = [] for row in row_results: assert len(row) == len(columns) rows.append(row) for i, (col_name, col_type) in enumerate(columns): values = [row[i] for row in rows] unique_name = '{}.{}'.format(table, col_name).lower() column_values[unique_name] = (unique_name, col_type, values) except: pass return column_values.values() def load_value_matchers(database_dir: str, schemas: Dict[str, SpiderSchema]): db_matchers = {} for schema in schemas.values(): db_id = schema.db_id column_with_values = get_column_with_values(os.path.join(database_dir, db_id, f'{db_id}.sqlite')) db_matchers[db_id] = ValueMatcher(column_with_values) return db_matchers def process_slsql_example(query: Dict, tokenizer: BertTokenizer, schema: SpiderSchema, value_matcher: ValueMatcher) -> Dict: question = query['question'] assert len(query['toks']) == len(query['lemma']) question_utterance = generate_utterance(tokenizer, question, [fix_tok(x) for x in query['toks']], [fix_tok(x) for x in query['lemma']]) # Step 2: process tables & columns processed_tables = [] for tbl_idx, col_indices in schema.table_to_columns.items(): # special column * if tbl_idx == -1: table_json = { 'index': -1, 'utterance': Utterance('*', tokens=[]).to_json(), 'columns': None } processed_tables += [table_json] continue tbl_name = schema.table_names[tbl_idx] table_utterance = generate_utterance(tokenizer, tbl_name) processed_columns = [] for col_idx in col_indices: column_type = schema.column_types[col_idx] column_utterance = generate_utterance(tokenizer, schema.column_names[col_idx]) column_json = { 'index': col_idx, 'utterance': column_utterance.to_json(), 'data_type': spider_type_mappings.get(column_type, 'text') } processed_columns += [column_json] table_json = { 'index': tbl_idx, 'utterance': table_utterance.to_json(), 'columns': processed_columns } processed_tables += [table_json] matched_values = value_matcher.match(question_utterance.text_tokens, 0.8, 3) processed_query = { 'question': question_utterance.to_json(), 'tables': processed_tables, 'schema': schema.to_json(), 'values': [v.to_json() for v in matched_values] } return processed_query def predict_alignments(model: nn.Module, data_iter: DataLoader, saved_path: str, threshold: float): slsql_align_labels = [] model.eval() with torch.no_grad(): for model_input in data_iter: model_output = model(**model_input) example = model_input['example'][0] meta_index: MetaIndex = model_input['meta_index'][0] question: Utterance = Utterance.from_json(example['question']) schema: SpiderSchema = SpiderSchema.from_json(example['schema']) values = [ValueMatch.from_json(v) for v in example['values']] identify_logits = { SQLTokenType.table: model_output['table_logits'][0], SQLTokenType.column: model_output['column_logits'][0], SQLTokenType.value: model_output['value_logits'][0] } tbl_align_weights, col_align_weights, val_align_weights = meta_index.split(model_output['alignment_weights'][0]) align_weights = { SQLTokenType.table: tbl_align_weights, SQLTokenType.column: col_align_weights, SQLTokenType.value: val_align_weights } pred_align_labels = greedy_link_spider(identify_logits, align_weights, question, schema, values, threshold=threshold) assert len(pred_align_labels) == len(question.tokens) sql_align_label = [label.to_slsql(schema) for label in pred_align_labels] slsql_align_labels += [sql_align_label] save_json_objects(slsql_align_labels, saved_path) print('predict alignments over, saved into {}'.format(saved_path)) pass if __name__ == '__main__': parser = argparse.ArgumentParser() parser.add_argument('-ckpt', '--checkpoint', default='baseline/slsql/codalab/saved_models_large/align_model.bin') parser.add_argument('-data', '--data_dir', default='baseline/slsql/codalab/dev_data') parser.add_argument('-db_dir', '--database_dir', default='baseline/slsql/data/database') parser.add_argument('-threshold', '--threshold', default=0.4, type=float) parser.add_argument('-output_dir', '--output_dir', default='output') parser.add_argument('-gpu', '--device', default='cuda:0' if torch.cuda.is_available() else 'cpu') args = parser.parse_args() config, model, tokenizer = load_model_and_data_iter(args) print('loading data iterator ...') processed_path = os.path.join(args.output_dir, 'dev.val_processed.json') process_examples( input_path=os.path.join(args.output_dir, 'dev.processed.json'), table_path=os.path.join(args.output_dir, 'tables.processed.json'), database_dir=args.database_dir, tokenizer=tokenizer, output_path=processed_path ) data_iter = get_data_iterator_func(config['model'])(processed_path, tokenizer, 1, config['device'], False, False, 512, None) predict_alignments(model, data_iter, os.path.join(args.output_dir, 'dev.align.json'), args.threshold) print('Run Alignment Over')

ContextualSP/awakening_latent_grounding/predict.py/0

import numpy as np import torch from utils.data_types import * from utils.nlp_utils import * class GreedyLinker: schema: SpiderSchema question: Utterance matched_values: List[ValueMatch] threshold: float identify_results: Dict[SQLTokenType, List[float]] alignment_dict: Dict[Tuple[SQLTokenType, int, int], Tuple[float, int]] def __init__(self, schema: SpiderSchema, question: Utterance, matched_values: List[ValueMatch], threshold=0.3) -> None: self.schema = schema self.question = question self.matched_values = matched_values self.threshold = threshold pass def link(self, identify_results: Dict[SQLTokenType, List[float]], align_weights: Dict[SQLTokenType, List[List[float]]]) -> List[AlignmentLabel]: self.identify_results = identify_results alignments = self.init_skeleton_alignments(align_weights) alignments = self.fix_missing(alignments) return alignments def fix_missing(self, alignments: List[AlignmentLabel]) -> List[AlignmentLabel]: # fix column suffix missing for i, align_label in enumerate(alignments): if align_label.align_type != SQLTokenType.column: continue column_original_lemma = self.schema.column_names_lemma[self.schema.id_map[align_label.align_value]].split() if i + 1 < len(alignments) and alignments[i+1].align_type == SQLTokenType.null and alignments[i+1].token.lemma in column_original_lemma: alignments[i+1] = AlignmentLabel(token=alignments[i+1].token, align_type=SQLTokenType.column, align_value=align_label.align_value, confidence=align_label.confidence) # fix column prefix missing for i in range(len(alignments)-1, -1, -1): align_label = alignments[i] if align_label.align_type != SQLTokenType.column: continue column_original_lemma = self.schema.column_names_lemma[self.schema.id_map[align_label.align_value]].split() if i - 1 >= 0 and alignments[i-1].align_type == SQLTokenType.null and alignments[i-1].token.lemma in column_original_lemma: alignments[i-1] = AlignmentLabel(token=alignments[i-1].token, align_type=SQLTokenType.column, align_value=align_label.align_value, confidence=align_label.confidence) schema_extact_matches = self._lookup_all_schema_extract_matches() # Fix column missing with value exists for i, align_label in enumerate(alignments): if align_label.align_type != SQLTokenType.value: continue column_name = align_label.align_value.replace("VAL_", "") column_idx = self.schema.id_map[column_name] for match in schema_extact_matches: if match['type'] != SQLTokenType.column or match['id'] != column_idx or not match['is_distinct']: continue is_all_unmatched = True for q_idx in range(match['start'], match['end'] + 1): if alignments[q_idx].align_type != SQLTokenType.null: is_all_unmatched = False break if is_all_unmatched: for q_idx in range(match['start'], match['end'] + 1): alignments[q_idx] = AlignmentLabel(self.question.tokens[q_idx], align_type=SQLTokenType.column, align_value=column_name, confidence=1.0) # Fix table column that occurs multiple times for match in schema_extact_matches: if self.identify_results[match['type']][match['id']] < 0.5 or not match['is_distinct']: continue is_all_unmatched = True for q_idx in range(match['start'], match['end'] + 1): if alignments[q_idx].align_type != SQLTokenType.null: is_all_unmatched = False break if is_all_unmatched: align_value = self.schema.get_identifier_name(type=match['type'].abbr, index=match['id']) for q_idx in range(match['start'], match['end'] + 1): alignments[q_idx] = AlignmentLabel(self.question.tokens[q_idx], align_type=match['type'], align_value=align_value, confidence=1.0) return alignments def _lookup_extract_tokens(self, query: str) -> List[Tuple[int, int]]: ngrams = permutate_ngrams(tokens=[x.lemma for x in self.question.tokens]) matched_spans = [] for i, j, ngram in ngrams: if ngram == query: matched_spans.append((i, j)) return matched_spans def _lookup_all_schema_extract_matches(self, identify_threshold: float=0.5) -> List[Dict]: schema_matches = [] for tbl_idx in range(self.schema.num_tables): if self.identify_results[SQLTokenType.table][tbl_idx] < identify_threshold: continue table_lemma = self.schema.table_names_lemma[tbl_idx] for start, end in self._lookup_extract_tokens(table_lemma): match = { 'type': SQLTokenType.table, 'id': tbl_idx, 'start': start, 'end': end, 'is_distinct': True } schema_matches.append(match) for col_idx in range(self.schema.num_columns): if self.identify_results[SQLTokenType.column][col_idx] < identify_threshold: continue column_lemma = self.schema.column_names_lemma[col_idx] for start, end in self._lookup_extract_tokens(column_lemma): match = { 'type': SQLTokenType.column, 'id': col_idx, 'start': start, 'end': end, 'is_distinct': True } schema_matches.append(match) for i, match in enumerate(schema_matches): is_distinct = match['is_distinct'] for j in range(i + 1, len(schema_matches)): match_j = schema_matches[j] if match_j['start'] > match['end'] or match_j['end'] < match['start']: continue is_distinct = False match_j['is_distinct'] = False match['is_distinct'] = is_distinct return schema_matches def _is_ngram_tokens(self, token1: Token, token2: Token, entity: str): ngram12 = '{} {}'.format(token1.lemma, token2.lemma) return ngram12 in entity def init_skeleton_alignments(self, align_weights: Dict[SQLTokenType, List[List[float]]]): alignments = self._init_alignments(align_weights) question_align_labels = [] for q_idx in range(len(self.question.tokens)): if q_idx not in alignments or len(alignments) == 0: question_align_labels.append(AlignmentLabel(token=self.question.tokens[q_idx], align_type=SQLTokenType.null, align_value=None, confidence=1.0)) continue question_align_labels.append(alignments[q_idx][0]) return question_align_labels def _init_alignments(self, align_weights: Dict[SQLTokenType, List[List[float]]]) -> Dict[int, List[AlignmentLabel]]: alignments = defaultdict(list) threshold = self.threshold low_threshold = max(0.2, 1.0 / self.question.num_tokens) # First is Value as value has span information val_align_weights = align_weights[SQLTokenType.value] col_align_weights = align_weights[SQLTokenType.column] tbl_align_weights = align_weights[SQLTokenType.table] columns_with_value = set([]) for v_idx, value in enumerate(self.matched_values): if self.identify_results[SQLTokenType.value][v_idx] < 0.5: continue confidence = self.identify_results[SQLTokenType.value][v_idx] for q_idx in range(value.start, value.end + 1): alignments[q_idx].append(AlignmentLabel( token=self.question.tokens[q_idx], align_type=SQLTokenType.value, align_value="VAL_{}".format(value.column), confidence=confidence)) columns_with_value.add(value.column) for c_idx in range(1, self.schema.num_columns): # Ignore column * if self.identify_results[SQLTokenType.column][c_idx] < 0.5: continue align_vector = np.array(col_align_weights[c_idx]) # * self.identify_results[SQLTokenType.column][c_idx] ranks = np.argsort(align_vector)[::-1] total_score = 0.0 for rk in range(len(align_vector)): q_idx = ranks[rk] score = align_vector[q_idx] if score < threshold / len(self.schema.column_names_lemma[c_idx].split()): break if total_score >= threshold: break total_score += score alignments[q_idx].append(AlignmentLabel( token=self.question.tokens[q_idx], align_type=SQLTokenType.column, align_value=self.schema.get_col_identifier_name(c_idx), confidence=score)) for t_idx in range(self.schema.num_tables): # Ignore column * if self.identify_results[SQLTokenType.table][t_idx] < 0.5: continue align_vector = np.array(tbl_align_weights[t_idx]) #* self.identify_results[SQLTokenType.table][t_idx] ranks = np.argsort(align_vector)[::-1] total_score = 0.0 for rk in range(len(align_vector)): q_idx = ranks[rk] score = align_vector[q_idx] if score < low_threshold or rk > 4: break if total_score >= threshold: break total_score += score alignments[q_idx].append(AlignmentLabel( token=self.question.tokens[q_idx], align_type=SQLTokenType.table, align_value=self.schema.get_tbl_identifier_name(t_idx), confidence=score )) for q_idx in alignments: alignments[q_idx] = list(sorted(alignments[q_idx], key=lambda x: self.get_alignment_label_sort_weight(x), reverse=True)) return alignments def get_alignment_label_sort_weight(self, align_label: AlignmentLabel) -> float: if align_label.align_type == SQLTokenType.value: return 100.0 + align_label.confidence elif align_label.align_type == SQLTokenType.column: column_idx = self.schema.id_map[align_label.align_value] weight = 1.0 if align_label.token.lemma.lower() in self.schema.column_names_original[column_idx].lower(): weight = 1.5 return align_label.confidence * weight elif align_label.align_type == SQLTokenType.table: table_idx = self.schema.id_map[align_label.align_value] weight = 1.0 if align_label.token.lemma.lower() in self.schema.table_names_original[table_idx].lower(): weight *= 1.5 return align_label.confidence * weight else: print(align_label) raise NotImplementedError() class SpiderGreedyLinker: schema: SpiderSchema question: Utterance matched_values: List[ValueMatch] identify_results: Dict[SQLTokenType, List[float]] alignment_dict: Dict[Tuple[SQLTokenType, int, int], Tuple[float, int]] threshold: float def __init__(self, schema: SpiderSchema, question: Utterance, matched_values: List[ValueMatch], threshold=0.3) -> None: self.schema = schema self.question = question self.matched_values = matched_values self.threshold = threshold ''' Lookup all linking relations with different confidences ''' def search_all(self, identify_results: Dict[SQLTokenType, List[float]], align_weights: Dict[SQLTokenType, List[List[float]]]) -> List[List[Dict]]: assert len(identify_results[SQLTokenType.table]) == len(self.schema.table_names_original) assert len(identify_results[SQLTokenType.column]) == len(self.schema.column_names_original) assert len(identify_results[SQLTokenType.value]) == len(self.matched_values) self.identify_results = identify_results init_alignments = self._init_alignments(align_weights) alignments_with_scores = defaultdict(list) for q_idx, align_labels in init_alignments.items(): for rk, align_label in enumerate(align_labels): slsql_label = align_label.to_slsql(self.schema) if align_label.confidence > self.threshold and rk < 1: slsql_label['confidence'] = 'high' else: slsql_label['confidence'] = 'low' is_added = False for label in alignments_with_scores[q_idx]: if label['type'] == slsql_label['type'] and label['id'] == slsql_label['id']: is_added = True break if not is_added: alignments_with_scores[q_idx].append(slsql_label) pass schema_extact_matches = self._lookup_all_schema_extract_matches() for match in schema_extact_matches: if self.identify_results[match['type']][match['id']] < 0.5: continue align_value = self.schema.get_identifier_name(match['type'].abbr, match['id']) slsql_label = { 'type': match['type'].abbr, 'id': match['id'], 'confidence': 'low', 'value': align_value } is_all_unmatched = True for q_idx in range(match['start'], match['end'] + 1): if q_idx in init_alignments and \ not (len(init_alignments[q_idx]) == 1 and init_alignments[q_idx][0].align_type == match['type'] and self.schema.id_map[init_alignments[q_idx][0].align_value] == match['id']): is_all_unmatched = False break if is_all_unmatched and match['is_distinct']: slsql_label['confidence'] = 'high' for q_idx in range(match['start'], match['end'] + 1): is_added = False for label in alignments_with_scores[q_idx]: if label['type'] == slsql_label['type'] and label['id'] == slsql_label['id']: is_added = True break if not is_added: slsql_label['token'] = self.question.tokens[q_idx].token alignments_with_scores[q_idx].append(slsql_label) all_alignment_labels = [] for q_idx in range(self.question.num_tokens): if q_idx not in alignments_with_scores: all_alignment_labels.append(None) else: sorted_alignments = sorted(alignments_with_scores[q_idx], key=lambda x: x['confidence'] == 'high', reverse=True) alignment_sets = set([]) distinct_labels = [] for alignment in sorted_alignments: if (alignment['type'], alignment['id']) in alignment_sets: continue alignment_sets.add((alignment['type'], alignment['id'])) distinct_labels.append(alignment) all_alignment_labels.append(distinct_labels) return all_alignment_labels def link(self, identify_results: Dict[SQLTokenType, List[float]], align_weights: Dict[SQLTokenType, List[List[float]]]) -> List[AlignmentLabel]: assert len(identify_results[SQLTokenType.table]) == len(self.schema.table_names_original) assert len(identify_results[SQLTokenType.column]) == len(self.schema.column_names_original) assert len(identify_results[SQLTokenType.value]) == len(self.matched_values) self.identify_results = identify_results alignments = self.init_skeleton_alignments(align_weights) alignments = self.fix_missing(alignments) return alignments def fix_missing(self, alignments: List[AlignmentLabel]) -> List[AlignmentLabel]: # fix column suffix missing for i, align_label in enumerate(alignments): if align_label.align_type != SQLTokenType.column: continue column_original_lemma = self.schema.column_names_lemma[self.schema.id_map[align_label.align_value]].split() if i + 1 < len(alignments) and alignments[i+1].align_type == SQLTokenType.null and alignments[i+1].token.lemma in column_original_lemma: alignments[i+1] = AlignmentLabel(token=alignments[i+1].token, align_type=SQLTokenType.column, align_value=align_label.align_value, confidence=align_label.confidence) # fix column prefix missing for i in range(len(alignments)-1, -1, -1): align_label = alignments[i] if align_label.align_type != SQLTokenType.column: continue column_original_lemma = self.schema.column_names_lemma[self.schema.id_map[align_label.align_value]].split() if i - 1 >= 0 and alignments[i-1].align_type == SQLTokenType.null and alignments[i-1].token.lemma in column_original_lemma: alignments[i-1] = AlignmentLabel(token=alignments[i-1].token, align_type=SQLTokenType.column, align_value=align_label.align_value, confidence=align_label.confidence) schema_extact_matches = self._lookup_all_schema_extract_matches() # Fix column missing with value exists for i, align_label in enumerate(alignments): if align_label.align_type != SQLTokenType.value: continue column_name = align_label.align_value.replace("VAL_", "") column_idx = self.schema.id_map[column_name] for match in schema_extact_matches: if match['type'] != SQLTokenType.column or match['id'] != column_idx or not match['is_distinct']: continue is_all_unmatched = True for q_idx in range(match['start'], match['end'] + 1): if alignments[q_idx].align_type != SQLTokenType.null: is_all_unmatched = False break if is_all_unmatched: for q_idx in range(match['start'], match['end'] + 1): alignments[q_idx] = AlignmentLabel(self.question.tokens[q_idx], align_type=SQLTokenType.column, align_value=column_name, confidence=1.0) # Fix table column that occurs multiple times for match in schema_extact_matches: if self.identify_results[match['type']][match['id']] < 0.5 or not match['is_distinct']: continue is_all_unmatched = True for q_idx in range(match['start'], match['end'] + 1): if alignments[q_idx].align_type != SQLTokenType.null: is_all_unmatched = False break if is_all_unmatched: align_value = self.schema.get_identifier_name(type=match['type'].abbr, index=match['id']) for q_idx in range(match['start'], match['end'] + 1): alignments[q_idx] = AlignmentLabel(self.question.tokens[q_idx], align_type=match['type'], align_value=align_value, confidence=1.0) return alignments def _lookup_extract_tokens(self, query: str) -> List[Tuple[int, int]]: ngrams = permutate_ngrams(tokens=[x.lemma for x in self.question.tokens]) matched_spans = [] for i, j, ngram in ngrams: if ngram == query: matched_spans.append((i, j)) return matched_spans def _lookup_all_schema_extract_matches(self, identify_threshold: float=0.5) -> List[Dict]: schema_matches = [] for tbl_idx in range(self.schema.num_tables): if self.identify_results[SQLTokenType.table][tbl_idx] < identify_threshold: continue table_lemma = self.schema.table_names_lemma[tbl_idx] for start, end in self._lookup_extract_tokens(table_lemma): match = { 'type': SQLTokenType.table, 'id': tbl_idx, 'start': start, 'end': end, 'is_distinct': True } schema_matches.append(match) for col_idx in range(self.schema.num_columns): if self.identify_results[SQLTokenType.column][col_idx] < identify_threshold: continue column_lemma = self.schema.column_names_lemma[col_idx] for start, end in self._lookup_extract_tokens(column_lemma): match = { 'type': SQLTokenType.column, 'id': col_idx, 'start': start, 'end': end, 'is_distinct': True } schema_matches.append(match) for i, match in enumerate(schema_matches): is_distinct = match['is_distinct'] for j in range(i + 1, len(schema_matches)): match_j = schema_matches[j] if match_j['start'] > match['end'] or match_j['end'] < match['start']: continue is_distinct = False match_j['is_distinct'] = False match['is_distinct'] = is_distinct return schema_matches def _is_ngram_tokens(self, token1: Token, token2: Token, entity: str): ngram12 = '{} {}'.format(token1.lemma, token2.lemma) return ngram12 in entity def init_skeleton_alignments(self, align_weights: Dict[SQLTokenType, List[List[float]]]): alignments = self._init_alignments(align_weights) question_align_labels = [] for q_idx in range(len(self.question.tokens)): if q_idx not in alignments or len(alignments) == 0: question_align_labels.append(AlignmentLabel(token=self.question.tokens[q_idx], align_type=SQLTokenType.null, align_value=None, confidence=1.0)) continue question_align_labels.append(alignments[q_idx][0]) return question_align_labels def _init_alignments(self, align_weights: Dict[SQLTokenType, List[List[float]]]) -> Dict[int, List[AlignmentLabel]]: alignments = defaultdict(list) threshold = self.threshold low_threshold = max(0.05, 1.0 / self.question.num_tokens) # First is Value as value has span information val_align_weights = align_weights[SQLTokenType.value] col_align_weights = align_weights[SQLTokenType.column] tbl_align_weights = align_weights[SQLTokenType.table] columns_with_value = set([]) for v_idx, value in enumerate(self.matched_values): if self.identify_results[SQLTokenType.value][v_idx] < 0.5: continue confidence = self.identify_results[SQLTokenType.value][v_idx] for q_idx in range(value.start, value.end + 1): alignments[q_idx].append(AlignmentLabel( token=self.question.tokens[q_idx], align_type=SQLTokenType.value, align_value="VAL_{}".format(value.column), confidence=confidence)) columns_with_value.add(value.column) for c_idx in range(1, self.schema.num_columns): # Ignore column * # if self.identify_results[SQLTokenType.column][c_idx] < 0.5: # continue align_vector = np.array(col_align_weights[c_idx]) * self.identify_results[SQLTokenType.column][c_idx] ranks = np.argsort(align_vector)[::-1] for rk in range(len(align_vector)): q_idx = ranks[rk] score = align_vector[q_idx] #* self.identify_results[SQLTokenType.column][c_idx] # if self.schema.get_column_full_name(c_idx) in columns_with_value: # score *= 0.5 if score >= threshold: alignments[q_idx].append(AlignmentLabel( token=self.question.tokens[q_idx], align_type=SQLTokenType.column, align_value=self.schema.get_col_identifier_name(c_idx), confidence=score )) if score >= low_threshold and self.identify_results[SQLTokenType.column][c_idx] > 0.5: if rk < 1 or self.question.tokens[q_idx].lemma in self.schema.column_names_lemma[c_idx].split(' '): alignments[q_idx].append(AlignmentLabel( token=self.question.tokens[q_idx], align_type=SQLTokenType.column, align_value=self.schema.get_col_identifier_name(c_idx), confidence=score )) for t_idx in range(self.schema.num_tables): # if self.identify_results[SQLTokenType.table][t_idx] < 0.5: # continue align_vector = np.array(tbl_align_weights[t_idx]) * self.identify_results[SQLTokenType.table][t_idx] ranks = np.argsort(align_vector)[::-1] for rk in range(len(align_vector)): q_idx = ranks[rk] score = align_vector[q_idx] #* self.identify_results[SQLTokenType.table][t_idx] if score >= threshold: alignments[q_idx].append(AlignmentLabel( token=self.question.tokens[q_idx], align_type=SQLTokenType.table, align_value=self.schema.table_names_original[t_idx].lower(), confidence=score )) if score >= low_threshold and self.identify_results[SQLTokenType.table][t_idx] > 0.5: if rk < 1 or self.question.tokens[q_idx].lemma in self.schema.table_names_lemma[t_idx].split(' '): alignments[q_idx].append(AlignmentLabel( token=self.question.tokens[q_idx], align_type=SQLTokenType.table, align_value=self.schema.get_tbl_identifier_name(t_idx), confidence=score )) for q_idx in alignments: alignments[q_idx] = list(sorted(alignments[q_idx], key=lambda x: self.get_alignment_label_sort_weight(x), reverse=True)) return alignments def get_alignment_label_sort_weight(self, align_label: AlignmentLabel) -> float: if align_label.align_type == SQLTokenType.value: return 100.0 + align_label.confidence elif align_label.align_type == SQLTokenType.column: column_idx = self.schema.id_map[align_label.align_value] weight = 1.0 if align_label.token.lemma.lower() in self.schema.column_names_original[column_idx].lower(): weight = 1.5 return align_label.confidence * weight elif align_label.align_type == SQLTokenType.table: table_idx = self.schema.id_map[align_label.align_value] weight = 1.0 if align_label.token.lemma.lower() in self.schema.table_names_original[table_idx].lower(): weight *= 1.5 return align_label.confidence * weight else: print(align_label) raise NotImplementedError() def greedy_link_spider( identify_logits: Dict[SQLTokenType, torch.Tensor], alignment_weights: Dict[SQLTokenType, torch.Tensor], question: Utterance, schema: SpiderSchema, values: List[ValueMatch], threshold: float = 0.25 ) -> List[AlignmentLabel]: linker = SpiderGreedyLinker(schema=schema, question=question, matched_values=values, threshold=threshold) for align_type in identify_logits: identify_logits[align_type] = torch.softmax(identify_logits[align_type], dim=-1)[:, 1].cpu().tolist() alignment_weights[align_type] = alignment_weights[align_type].cpu().tolist() alignment_labels = linker.link(identify_results=identify_logits, align_weights=alignment_weights) return alignment_labels def greedy_search_all_spider( identify_logits: Dict[SQLTokenType, torch.Tensor], alignment_weights: Dict[SQLTokenType, torch.Tensor], question: Utterance, schema: SpiderSchema, values: List[ValueMatch], threshold: float = 0.3 ) -> List[List[Dict]]: linker = SpiderGreedyLinker(schema=schema, question=question, matched_values=values, threshold=threshold) return linker.search_all(identify_results=identify_logits, align_weights=alignment_weights) def mask_value_alignments(align_weights: torch.Tensor, values: List[ValueMatch]) -> torch.Tensor: for v_i in range(len(align_weights)): start, end = values[v_i].start, values[v_i].end mask = torch.zeros(align_weights.size(1), dtype=torch.bool) mask[start:end+1] = 1 align_weights[v_i].masked_fill_(mask == 0, 0.0) return align_weights def generate_alignments_spider( align_weights: Dict[SQLTokenType, torch.Tensor], question: Utterance, schema: SpiderSchema, values: List[ValueMatch], threshold: float=0.3, ) -> List[AlignmentLabel]: assert len(align_weights[SQLTokenType.table]) == schema.num_tables assert len(align_weights[SQLTokenType.column]) == schema.num_columns assert len(align_weights[SQLTokenType.value]) == len(values) align_weights[SQLTokenType.value] = mask_value_alignments(align_weights[SQLTokenType.value], values) align_matrix = torch.cat([align_weights[SQLTokenType.table], align_weights[SQLTokenType.column], align_weights[SQLTokenType.value]], dim=0) align_matrix = align_matrix.transpose(0, 1) # question_length * num_entities assert len(align_matrix) == question.num_tokens align_labels = [] for q_idx in range(question.num_tokens): max_idx = torch.argmax(align_matrix[q_idx], dim=-1).item() confidence = align_matrix[q_idx, max_idx] if confidence < threshold: align_label = AlignmentLabel(question.tokens[q_idx], SQLTokenType.null, None, 1 - confidence) align_labels.append(align_label) continue if max_idx < schema.num_tables: align_labels.append(AlignmentLabel(question.tokens[q_idx], SQLTokenType.table, schema.get_tbl_identifier_name(max_idx), confidence)) elif max_idx < schema.num_tables + schema.num_columns: column_idx = max_idx - schema.num_tables align_labels.append(AlignmentLabel(question.tokens[q_idx], SQLTokenType.column, schema.get_col_identifier_name(column_idx), confidence)) elif max_idx < schema.num_tables + schema.num_columns + len(values): value_idx = max_idx - schema.num_tables - schema.num_columns align_labels.append(AlignmentLabel(question.tokens[q_idx], SQLTokenType.value, 'VAL_{}'.format(values[value_idx].column), confidence)) else: raise NotImplementedError() return align_labels

ContextualSP/awakening_latent_grounding/utils/schema_linker.py/0

import torch from torch import nn class LstmRnn(nn.Module): def __init__(self, input_dim, hidden_dim): super().__init__() self.i_dim = input_dim self.h_dim = hidden_dim self.lstm = nn.LSTMCell(input_dim, hidden_dim) self.h0 = nn.Parameter(torch.empty(size=(1, hidden_dim), dtype=torch.float32)) self.c0 = nn.Parameter(torch.empty(size=(1, hidden_dim), dtype=torch.float32)) self.reset_parameters() def reset_parameters(self): nn.init.constant_(self.h0, val=0) nn.init.constant_(self.c0, val=0) nn.init.xavier_uniform_(self.lstm.weight_ih) nn.init.orthogonal_(self.lstm.weight_hh) nn.init.constant_(self.lstm.bias_ih, val=0) nn.init.constant_(self.lstm.bias_hh, val=0) def forward(self, x, mask, backward=False): L = x.shape[1] prev_h = self.h0.expand(x.shape[0], -1) prev_c = self.c0.expand(x.shape[0], -1) h = [] for idx in range(L): idx = L - 1 - idx if backward else idx mask_idx = mask[:, idx, None] h_idx, c_idx = self.lstm(x[:, idx], (prev_h, prev_c)) prev_h = h_idx * mask_idx + prev_h * (1. - mask_idx) prev_c = c_idx * mask_idx + prev_c * (1. - mask_idx) h.append(prev_h) return torch.stack(h[::-1] if backward else h, dim=1)

ContextualSP/compositional_generalization/modules/LstmRnn.py/0

{ "random_seed": 42, "numpy_seed": 42, "pytorch_seed": 42, "dataset_reader": { "type": "rewrite", "lazy": false, "super_mode": "before", "joint_encoding": true, "use_bert": true, "language": "zh", "extra_stop_words": ["的", "是", "我", "了", "去"] }, "model": { "type": "rewrite", "word_embedder": { "bert": { "type": "bert-pretrained", "pretrained_model": "bert-base-chinese", "top_layer_only": true, "requires_grad": true }, "allow_unmatched_keys": true, "embedder_to_indexer_map": { "bert": [ "bert", "bert-offsets", "bert-type-ids" ] } }, "text_encoder": { "type": "lstm", "input_size": 768, "hidden_size": 200, "bidirectional": true, "num_layers": 1 }, "inp_drop_rate": 0.2, "out_drop_rate": 0.2, "feature_sel": 83, "loss_weights": [0.2, 0.2, 0.6], "super_mode": "before", "unet_down_channel": 64 }, "iterator": { "type": "basic", "batch_size": 12 }, "validation_iterator": { "type": "basic", "batch_size": 12 }, "trainer": { "num_epochs": 100, "cuda_device": 0, "patience": 10, "validation_metric": "+F3", "optimizer": { "type": "adam", "parameter_groups": [ [ [ ".*word_embedder.*" ], { "lr": 1e-5 } ] ], "lr": 1e-3 }, "learning_rate_scheduler": { "type": "reduce_on_plateau", "factor": 0.5, "mode": "max", "patience": 5 }, "num_serialized_models_to_keep": 10, "should_log_learning_rate": true } }

ContextualSP/incomplete_utterance_rewriting/configs/multi_bert.jsonnet/0

""" Utility functions for reading the standardised text2sql datasets presented in `"Improving Text to SQL Evaluation Methodology" <https://arxiv.org/abs/1806.09029>`_ """ import json import os import sqlite3 from collections import defaultdict from typing import List, Dict, Optional from allennlp.common import JsonDict class TableColumn: """ Representing the column of table """ def __init__(self, name: str, text: str, column_type: str, is_primary_key: bool, refer_table, foreign_key: Optional[List[str]]): self.name = name self.text = text self.column_type = column_type self.is_primary_key = is_primary_key self.foreign_key = foreign_key self.refer_table = refer_table def __str__(self): return f'{self.name}' class Table: """ Representing the table """ def __init__(self, name: str, text: str, columns: List[TableColumn]): self.name = name self.text = text self.columns = columns def read_dataset_schema(schema_path: str): """ Reading all table from `schema_path`. :param schema_path: default from `tables.json` of sparc data folder. :return: """ schemas: Dict[str, Dict[str, Table]] = defaultdict(dict) schema_id_to_table: Dict[str, Dict[int, Table]] = defaultdict(dict) schema_id_to_col: Dict[str, Dict[int, TableColumn]] = defaultdict(dict) dbs_json_blob = json.load(open(schema_path, "r")) for db in dbs_json_blob: db_id = db['db_id'] column_id_to_table = {} column_id_to_column = {} for i, (column, text, column_type) in enumerate(zip(db['column_names_original'], db['column_names'], db['column_types'])): table_id, column_name = column _, column_text = text table_name = db['table_names_original'][table_id] if table_name not in schemas[db_id]: table_text = db['table_names'][table_id] table_obj = Table(table_name, table_text, []) schemas[db_id][table_name] = table_obj # TODO: we cannot add an extra command to handle * problem. # we now use a special embedding for linking * and predicting action # if column_name == '*': # continue table_obj = schemas[db_id][table_name] if column_name == "*": is_primary_key = False else: is_primary_key = i in db['primary_keys'] # allocate new column object column_obj = TableColumn(column_name.lower(), column_text, column_type, is_primary_key, table_obj, None) schemas[db_id][table_name].columns.append(column_obj) column_id_to_column[i] = column_obj for (c1, c2) in db['foreign_keys']: foreign_key = column_id_to_column[c2].refer_table.name + ':' + column_id_to_column[c2].name # TODO: we able multiple foreign keys existing to allow the shortcut join if column_id_to_column[c1].foreign_key is None: column_id_to_column[c1].foreign_key = [] column_id_to_column[c1].foreign_key.append(foreign_key) for i, table_name in enumerate(db['table_names_original']): column_id_to_table[i] = schemas[db_id][table_name] # assign id to column and id to table schema_id_to_table[db_id] = column_id_to_table schema_id_to_col[db_id] = column_id_to_column return {**schemas}, {**schema_id_to_col}, {**schema_id_to_table} def read_dataset_values(db_id: str, dataset_path: str, tables: List[str]): db = os.path.join(dataset_path, db_id, db_id + ".sqlite") try: conn = sqlite3.connect(db) except Exception as e: raise Exception(f"Can't connect to SQL: {e} in path {db}") conn.text_factory = str cursor = conn.cursor() values = {} for table in tables: try: cursor.execute(f"SELECT * FROM {table.name} LIMIT 5000") values[table] = cursor.fetchall() except: conn.text_factory = lambda x: str(x, 'latin1') cursor = conn.cursor() cursor.execute(f"SELECT * FROM {table.name} LIMIT 5000") values[table] = cursor.fetchall() return values def ent_key_to_name(key): parts = key.split(':') if parts[0] == 'table': return parts[1] elif parts[0] == 'column': _, _, table_name, column_name = parts return f'{table_name}@{column_name}' else: return parts[1] def fix_number_value(ex: JsonDict): """ There is something weird in the dataset files - the `query_toks_no_value` field anonymizes all values, which is good since the evaluator doesn't check for the values. But it also anonymizes numbers that should not be anonymized: e.g. LIMIT 3 becomes LIMIT 'value', while the evaluator fails if it is not a number. """ def split_and_keep(s, sep): if not s: return [''] # consistent with string.split() # Find replacement character that is not used in string # i.e. just use the highest available character plus one # Note: This fails if ord(max(s)) = 0x10FFFF (ValueError) p = chr(ord(max(s)) + 1) return s.replace(sep, p + sep + p).split(p) # input is tokenized in different ways... so first try to make splits equal query_toks = ex['query_toks'] ex['query_toks'] = [] for q in query_toks: ex['query_toks'] += split_and_keep(q, '.') i_val, i_no_val = 0, 0 while i_val < len(ex['query_toks']) and i_no_val < len(ex['query_toks_no_value']): if ex['query_toks_no_value'][i_no_val] != 'value': i_val += 1 i_no_val += 1 continue i_val_end = i_val while i_val + 1 < len(ex['query_toks']) and \ i_no_val + 1 < len(ex['query_toks_no_value']) and \ ex['query_toks'][i_val_end + 1].lower() != ex['query_toks_no_value'][i_no_val + 1].lower(): i_val_end += 1 if i_val == i_val_end and ex['query_toks'][i_val] in ["1", "2", "3"] and ex['query_toks'][i_val - 1].lower() == "limit": ex['query_toks_no_value'][i_no_val] = ex['query_toks'][i_val] i_val = i_val_end i_val += 1 i_no_val += 1 return ex

ContextualSP/interactive_text_to_sql/src/context/utils.py/0

Code and dataset are under cleaning. Coming soon.

ContextualSP/knowledge_intensive_text_to_sql/README.md/0

from abc import ABCMeta, abstractmethod, abstractproperty from collections import defaultdict, Counter import numpy as np from numpy.testing import assert_approx_equal def last_k(tokens, k): """Get the last k elements of a list as a tuple.""" if not (0 <= k <= len(tokens)): raise ValueError('k must be between 0 and len(tokens) = {}, got: {}'.format(len(tokens), k)) return tuple(tokens[len(tokens) - k:]) def replace_parens(tokens): """Replace all instances of -LRB- and -RRB- with actual parentheses.""" parens_map = {'-LRB-': '(', '-RRB-': ')'} return [parens_map.get(s, s) for s in tokens] # return identity if not parens symbols def normalize_counts(counts): """Return a normalized Counter object.""" normed = Counter() total = sum(list(counts.values()), 0.0) assert total > 0 # cannot normalize empty Counter for key, ct in counts.items(): normed[key] = ct / total normed.old_total = total # document what the total was before normalization return normed class LM(object, metaclass=ABCMeta): """Language model interface.""" START = '<START>' END = '<END>' @abstractmethod def next_distribution(self, history): """Return a distribution over the next token. Args: history (List): a list of tokens generated so far Returns (Counter): a distribution """ raise NotImplementedError @abstractproperty def max_context_size(self): """Return max allowed history context. Returns (int): maximum size of history context to keep """ raise NotImplementedError class CountLM(LM): """Naive language model. Uses empirical counts from the largest context it has observed. No sophisticated backoff strategy. Examples: lm = CountLM(4) # 'train' the language model for line in lines: tokens = line.split() lm.record_counts(tokens, append_end=True) """ def __init__(self, max_context_size): """Construct a language model. Args: max_context_size (int): maximum # tokens to use as context """ self._max_context_size = max_context_size self.contexts = defaultdict(Counter) @property def max_context_size(self): return self._max_context_size def _get_contexts(self, tokens): """List of contexts, from smallest to largest. Includes empty context. Returns: List[Tuple[str]] """ contexts = [] max_context = min(self._max_context_size, len(tokens)) # cannot go beyond max tokens for k in range(max_context + 1): contexts.append(last_k(tokens, k)) return contexts def record_counts(self, tokens, append_end): """Record counts using `tokens` as a corpus. Args: tokens (List[str]): list of strings """ history = [LM.START] if append_end: tokens = tokens + [LM.END] for tok in tokens: for context in self._get_contexts(history): self.contexts[context][tok] += 1 history.append(tok) # update history def _largest_context(self, history, contexts): """Find the largest context which matches history. Args: history (List[str]): a sequence of tokens contexts (Set[Tuple[str]]): a set of contexts, must include the empty context Returns: Tuple[str]: an item from contexts, which may be the empty context """ assert tuple() in contexts # empty context must be present for context in reversed(self._get_contexts(history)): if context in contexts: return context def _largest_known_context(self, history): """Find the largest recorded context which matches history.""" return self._largest_context(history, self.contexts) def next_distribution(self, history): """Given a history, return a distribution (Counter) over the next token.""" context = self._largest_known_context(history) counts = self.contexts[context] normed = normalize_counts(counts) normed.context = context return normed def sequence_probability(self, tokens): """Return the probability of each token in an article, based on the language model. Args: tokens (List): a list of tokens in the article Returns: List[Tuple[str, float]]: an ordered list of token-probability pairs""" history = [LM.START] probabilities = [] for word in tokens: distr = self.next_distribution(history) if word in distr: probabilities.append((word, distr[word])) else: probabilities.append((word, 0.0)) history.append(word) return probabilities class KNNLM(LM): def __init__(self, article_embeddings, max_context_size, k_nearest): """Construct k-nearest-neighbor language model. Args: article_embeddings (ArticleEmbeddings): embeddings of each article max_context_size (int): max history to consider for CountLM k_nearest (int): # neighbors to consider """ self.article_embeddings = article_embeddings self.k = k_nearest self.lm = CountLM(max_context_size) def record_nearest_counts(self, vec): name_score_pairs = self.article_embeddings.k_nearest_approx(vec, self.k) articles = [self.article_embeddings.name_to_article(name) for name, score in name_score_pairs] for art in articles: self.lm.record_counts(art.tokens, append_end=True) def next_distribution(self, history): return self.lm.next_distribution(history) @property def max_context_size(self): return self.lm.max_context_size def sequence_probability(self, tokens): return self.lm.sequence_probability(tokens) class Generator(object, metaclass=ABCMeta): """Interface for language generator.""" @abstractmethod def init_history(self): """Return a sequence of tokens to initialize the history.""" pass @abstractmethod def get_next(self, history): """Get next token, given history.""" pass @abstractmethod def stop_or_not(self, history): """Given what has been generated, decide whether to stop.""" pass @abstractproperty def max_context_size(self): """Return max allowed history context. Returns (int): maximum size of history context to keep """ raise NotImplementedError def truncate_history(self, history): """Truncate history when it grows much longer than max context size.""" if len(history) > 2 * self.max_context_size: return list(last_k(history, self.max_context_size)) return history def generate(self, history=None): """Generate a sequence of tokens.""" if not history: history = self.init_history() return self.generate_custom(history, self.get_next, self.stop_or_not) def generate_custom(self, history, next_fxn, stop_fxn): """Generate a sequence using a custom next-token function and a custom stopping function. Args: history (List[T]): initial history next_fxn (Callable[[List[T]], T]): given a history, produce the next token stop_fxn (Callable[[List[T]], bool]): given a history, decide whether to stop """ generated = [] history = list(history) # make a copy while True: next = next_fxn(history) history.append(next) history = self.truncate_history(history) if stop_fxn(history): break generated.append(next) return generated class LMSampler(Generator): """Generation by sampling from a language model.""" def __init__(self, lm): """Construct a LM sampler. Args: lm (LM): a language model """ self.lm = lm @property def max_context_size(self): return self.lm.max_context_size def _sample_from_distribution(self, distr): """Sample from a categorical distribution. Args: distr (Counter): values must sum to 1 Returns: one of the keys of distr """ keys, probs = list(zip(*list(distr.items()))) assert_approx_equal(sum(probs), 1.) return np.random.choice(keys, p=probs) def init_history(self): return [self.lm.START] def get_next(self, history): return self._sample_from_distribution(self.lm.next_distribution(history)) def stop_or_not(self, history): return history[-1] == LM.END @staticmethod def format_generation(tokens): return ' '.join(replace_parens(tokens)) class DistributionStats(object): def __init__(self, distr): self.total = distr.old_total self.context = distr.context probs = list(distr.values()) assert_approx_equal(sum(probs), 1.) self.entropy = -1. * sum([p * np.log(p) for p in probs]) def __repr__(self): return '{}:{}:{}'.format(len(self.context), self.total, self.entropy) class LMSamplerWithStats(LMSampler): def init_history(self): return [(LM.START, 0)] def get_next(self, history): token_history, _ = list(zip(*history)) distr = self.lm.next_distribution(token_history) next_token = self._sample_from_distribution(distr) return next_token, DistributionStats(distr) def stop_or_not(self, history): word = lambda pair: pair[0] return word(history[-1]) == LM.END @staticmethod def format_generation(token_stat_pairs): tokens, stats = list(zip(*list(token_stat_pairs))) tokens = replace_parens(tokens) tokens = ['{:20}[{}]'.format(tok, stat) for tok, stat in zip(tokens, stats)] return '\n'.join(tokens)

ContextualSP/lemon/executor/gtd/lm.py/0

from abc import abstractmethod from collections import Sequence, Mapping import numpy as np import pytest import tensorflow as tf from keras.engine import Input from keras.layers import Dense from numpy.testing import assert_array_almost_equal from gtd.ml.framework import Feedable, KerasModel from gtd.ml.utils import guarantee_initialized_variables, clean_session from gtd.utils import Bunch @pytest.yield_fixture def clean_test_session(): with clean_session() as sess: yield sess def assert_array_collections_equal(correct, test, decimal=7): """Assert that two collections of numpy arrays have the same values. Collections can be either a Sequence or a Mapping. """ if type(correct) != type(test): raise ValueError('correct ({}) and test ({}) must have the same type.'.format(type(correct), type(test))) assert_equal = lambda c, t: assert_array_almost_equal(c, t, decimal=decimal) if isinstance(correct, Sequence): assert len(correct) == len(test) for c, t in zip(correct, test): assert_equal(c, t) elif isinstance(correct, Mapping): # same keys assert set(test.keys()) == set(correct.keys()) # same values for key in test: assert_equal(correct[key], test[key]) else: raise TypeError('Inputs must be of type Mapping or Sequence, not {}.'.format(type(correct))) class FeedableTester(object): """A template for testing Feedable classes. Subclass this class and implement all of its abstractmethods. NOTE: You must decorate the implementation of each abstractmethod with a @pytest.fixture decorator. See the `TestFeedable` class below for an example. """ @abstractmethod def model(self): """The Model to be tested.""" pass @abstractmethod def inputs(self): """Inputs to the model. Returns: (list, dict): an args, kwargs pair """ pass @classmethod def as_args_kwargs(cls, *args, **kwargs): return args, kwargs @abstractmethod def feed_dict(self): """Return the correct result of the model's `feed_dict` method.""" pass @abstractmethod def output_tensors(self): """Output tensors to be fetched. Returns: list[np.array] """ pass @abstractmethod def outputs(self): """Return the correct results of running model.compute(fetch=output_tensors, ...) Returns: list[np.array] """ pass @pytest.mark.usefixtures('clean_test_session') def test_inputs_to_feed_dict(self, model, inputs, feed_dict): """Test for correct feed_dict.""" args, kwargs = inputs test_feed_dict = model.inputs_to_feed_dict(*args, **kwargs) assert_array_collections_equal(feed_dict, test_feed_dict) @pytest.mark.usefixtures('clean_test_session') def test_outputs(self, model, inputs, output_tensors, outputs): """Test for correct output.""" sess = tf.get_default_session() guarantee_initialized_variables(sess) args, kwargs = inputs test_outputs = model.compute(output_tensors, *args, **kwargs) assert_array_collections_equal(outputs, test_outputs, decimal=4) class KerasModelTester(FeedableTester): @pytest.fixture def output_tensors(self, model): return model.output_tensors @pytest.mark.usefixtures('clean_test_session') def test_placeholders(self, model, feed_dict): """Test that model.placeholders matches the keys of feed_dict.""" assert set(model.placeholders) == set(feed_dict.keys()) class FeedableExample(Feedable): def __init__(self): x = tf.placeholder(tf.float32, shape=[], name='x') y = tf.get_variable('y', shape=[], initializer=tf.constant_initializer(2.0)) z = x * y self.x = x self.y = y self.z = z def inputs_to_feed_dict(self, batch): return {self.x: batch.x} class TestFeedableExample(FeedableTester): @pytest.fixture def model(self): return FeedableExample() @pytest.fixture def inputs(self): return self.as_args_kwargs(Bunch(x=5.0)) @pytest.fixture def feed_dict(self, model): return {model.x: 5.0} @pytest.fixture def output_tensors(self, model): return [model.z] @pytest.fixture def outputs(self): return [10.0] class KerasLayersModelExample(KerasModel): """A Model that is defined using Keras layers from beginning to end.""" def __init__(self): x = Input([1]) y = np.array([[2.0]]) b = np.array([0.0]) mult = Dense(1, weights=(y, b)) z = mult(x) self.x = x self.mult = mult self.z = z @property def placeholders(self): return [self.x] def inputs_to_feed_dict(self, batch): return {self.x: np.array([[batch.x]])} @property def output_tensors(self): return [self.z] class TestKerasLayersModel(KerasModelTester): @pytest.fixture def model(self): return KerasLayersModelExample() @pytest.fixture def inputs(self): return self.as_args_kwargs(Bunch(x=5.0)) @pytest.fixture def feed_dict(self, model): return {model.x: 5.0} @pytest.fixture def outputs(self): return [10.0]

ContextualSP/lemon/executor/gtd/tests/ml/test_framework.py/0

from collections import namedtuple import numpy as np from gtd.utils import flatten from strongsup.case_weighter import get_case_weighter from strongsup.value_function import get_value_function, ValueFunctionExample class NormalizationOptions(object): """Constants for normalization options""" LOCAL = 'local' GLOBAL = 'global' # used by the Decoder to compute gradients WeightedCase = namedtuple('WeightedCase', ['case', 'weight']) class Decoder(object): """A decoder does two things: - Given a batch of examples, produce a Beam (list of ParsePaths) for each example. Internally it uses an ExplorationPolicy to produce beams, and a ParseModel to score the ParseCases. - Given a batch of Beams, update the model parameters by passing appropriate ParseCases to the TrainParseModel. """ def __init__(self, parse_model, config, domain): """Create a new decoder. Args: parse_model (TrainParseModel) config (Config): The decoder section of the config domain (Domain) """ self._parse_model = parse_model self._value_function = get_value_function( config.value_function, parse_model.parse_model) self._case_weighter = get_case_weighter( config.case_weighter, parse_model.parse_model, self._value_function) self._config = config self._caching = config.inputs_caching self._domain = domain self._path_checker = domain.path_checker # Normalization and update policy self._normalization = config.normalization if config.normalization == NormalizationOptions.GLOBAL: raise ValueError('Global normalization is no longer supported.') # Exploration policy # TODO: Resolve this circular import differently from strongsup.exploration_policy import get_exploration_policy self._test_exploration_policy = get_exploration_policy( self, config.test_exploration_policy, self._normalization, train=False) self._train_exploration_policy = get_exploration_policy( self, config.train_exploration_policy, self._normalization, train=True) @property def parse_model(self): return self._parse_model @property def caching(self): return self._caching @property def domain(self): return self._domain def exploration_policy(self, train): """Returns the train or test exploration policy depending on train Args: train (bool) Returns: ExplorationPolicy """ if train: return self._train_exploration_policy else: return self._test_exploration_policy def path_checker(self, path): """Return False if the ParsePath should be pruned away; True otherwise. Args: path (ParsePath) Returns: bool """ return self._path_checker(path) def get_probs(self, beam): """Return a numpy array containing the probabilities of the paths in the given beam. The entries may not sum to 1 for local normalization since we have pruned away choices that are not executable. Args: beam (Beam) Returns: np.array of length len(beam) containing the probabilities. """ if len(beam) == 0: return np.zeros(0) if self._normalization == NormalizationOptions.LOCAL: return np.exp(np.array([path.log_prob for path in beam])) else: stuff = np.array([path.score for path in beam]) stuff = np.array(stuff - np.min(stuff)) return stuff / np.sum(stuff) ################################ # Prediction def predictions(self, examples, train, verbose=False): """Return the final beams for a batch of contexts. Args: contexts (list[Context]): a batch of Contexts verbose (bool) train (bool): If you're training or evaluating Returns: list[Beam]: a batch of Beams """ exploration_policy = self.exploration_policy(train) beams = exploration_policy.get_beams(examples, verbose) return [beam.get_terminated() for beam in beams] def get_intermediate_beams(self, examples, train, verbose=False): exploration_policy = self.exploration_policy(train) return exploration_policy.get_intermediate_beams(examples, verbose) def score_breakdown(self, paths): """Return the logits for all (parse case, choice, scorer) tuples. Args: paths (list[ParsePath]) Returns: grouped_attentions: a list of length(paths). Each entry is an np.array of shape (>= len(utterance)) containing the attention scores grouped_subscores: a list of length len(paths). Each entry is an np.array of shape (>= number of cases, len(choices), number of scorers) containing the logits of each scorer on each choice. By default there are 3 scorers: basic, attention, and soft copy. """ if len(paths) == 0: return [], [] cumul = [0] # Used to group the results back cases = [] for path in paths: for case in path: cases.append(case) cumul.append(len(cases)) # Get the scores from the model attentions, subscores = self._parse_model.score_breakdown(cases, ignore_previous_utterances=False, caching=False) # Group the scores by paths grouped_attentions, grouped_subscores = [], [] for i in range(len(paths)): grouped_attentions.append(attentions[cumul[i]:cumul[i+1]]) grouped_subscores.append(subscores[cumul[i]:cumul[i+1]]) return grouped_attentions, grouped_subscores ################################ # Training def train_step(self, examples): # sample a beam of logical forms for each example beams = self.predictions(examples, train=True) all_cases = [] # a list of ParseCases to give to ParseModel all_case_weights = [] # the weights associated with the cases for example, paths in zip(examples, beams): case_weights = self._case_weighter(paths, example) case_weights = flatten(case_weights) cases = flatten(paths) assert len(case_weights) == sum(len(p) for p in paths) all_cases.extend(cases) all_case_weights.extend(case_weights) # for efficiency, prune cases with weight 0 cases_to_reinforce = [] weights_to_reinforce = [] for case, weight in zip(all_cases, all_case_weights): if weight != 0: cases_to_reinforce.append(case) weights_to_reinforce.append(weight) # update value function vf_examples = [] for example, paths in zip(examples, beams): vf_examples.extend(ValueFunctionExample.examples_from_paths(paths, example)) self._value_function.train_step(vf_examples) # update parse model self._parse_model.train_step( cases_to_reinforce, weights_to_reinforce, caching=False) @property def step(self): return self._parse_model.step

ContextualSP/lemon/executor/strongsup/decoder.py/0

class EntrySelector(object): """Given a list of Entries, returns single Entry based on some criteria. Args: entries (list[Entry]): the entries """ def __init__(self, entries): self._entries = entries @property def best_any_seed(self): """Returns the Entry with the best ResultValue over any seed.""" if len(self._entries) == 0: return None return max(self._entries, key=lambda entry: entry.best[1]) @property def best_avg(self): """Returns the Entry with the best ResultValue averaged over all seeds.""" if len(self._entries) == 0: return None return max(entries, key=lambda entry: entry.avg)

ContextualSP/lemon/executor/strongsup/results/entry_selector.py/0

import itertools import time from abc import ABCMeta, abstractmethod, abstractproperty from collections import deque import logging from strongsup.parse_case import ParseCase from strongsup.value import check_denotation class StaticCase(object, metaclass=ABCMeta): """Like a ParseCase, but only statically analyzed, never dynamically executed. Primarily used by StaticBatchExploration. """ @abstractmethod def seeds(cls): """Return a list of seed cases to start searching from.""" pass @abstractmethod def extend(self, predicate): """Return a new StaticCase which extends from this one.""" pass @abstractproperty def choices(self): """Choices available from this state.""" pass @abstractproperty def length(self): """Length of episode so far.""" pass @abstractproperty def utterances_read(self): """Number of utterances processed so far in this episode.""" pass @abstractproperty def stack_depth(self): """Depth of execution stack.""" pass @abstractproperty def path(self): """Return a list of StaticCases.""" pass class AlchemyCase(object): __slots__ = ['predicate', 'prev_case', 'length', 'utterances_read', 'execution_stack', 'command_history'] choices = [ 'r', 'y', 'g', 'o', 'p', 'b', '1', '2', '3', '4', '5', '6', '7', '-1', 'X1/1', 'PColor', 'APour', 'AMix', 'ADrain', 'all-objects', 'index', 'H0', 'H1', 'H2', ] def __init__(self, predicate, prev_case, length, utterances_read, execution_stack, command_history): self.predicate = predicate self.prev_case = prev_case self.length = length self.utterances_read = utterances_read self.execution_stack = execution_stack self.command_history = command_history @classmethod def seeds(cls): seeds = [] for p in cls.choices: state = cls._update_state([], [], p) if state is None: continue exec_stack, cmd_history = state case = AlchemyCase(p, None, 1, 0, exec_stack, cmd_history) seeds.append(case) return seeds def extend(self, predicate): state = self._update_state(self.execution_stack, self.command_history, predicate) if state is None: return None # predicate leads to invalid state exec_stack, cmd_history = state utterances_read = self.utterances_read if predicate[0] == 'A' or predicate == 'H0': utterances_read += 1 return AlchemyCase(predicate, self, self.length + 1, utterances_read, exec_stack, cmd_history) @property def stack_depth(self): return len(self.execution_stack) @property def path(self): path = [] current = self while True: path.append(current) current = current.prev_case if current is None: break path.reverse() return path @classmethod def _get_args_from_stack(cls, exec_stack, predicate): if predicate in ('APour', 'ADrain', 'index'): n = 2 elif predicate in ('AMix', 'PColor') or predicate[0] == 'H': n = 1 else: return None if len(exec_stack) < n: # not enough arguments return None return exec_stack[-n:] def __repr__(self): return self.predicate @classmethod def _update_state(cls, exec_stack, command_history, predicate): """ We assume action clears stack. Args: exec_stack command_history predicate Returns: new_exec_stack, new_command_history """ # TYPES COLOR = 'CLR' BEAKER = 'BKR' LIST = 'LST' is_number = lambda s: s in ('1', '2', '3', '4', '5', '6', '7', '-1') # SIMPLE VALUES if predicate in ('r', 'y', 'g', 'o', 'p', 'b'): # abstract to COLOR return exec_stack + [COLOR], list(command_history) if is_number(predicate): # preserve numbers exactly return exec_stack + [predicate], list(command_history) if predicate == 'all-objects': # abstract to LIST return exec_stack + [LIST], list(command_history) # FUNCTIONS args = cls._get_args_from_stack(exec_stack, predicate) if args is None: return None # not enough arguments logging.debug('Args peeked: {}'.format(args)) prefix = predicate[0] # actions if prefix == 'A': logging.debug('Processing action') logging.debug(exec_stack) if len(args) != len(exec_stack): # action must clear stack return None # type check if predicate == 'APour': if args != [BEAKER, BEAKER]: return None if predicate == 'ADrain': if args[0] != BEAKER or not is_number(args[1]): return None if predicate == 'AMix': if args != [BEAKER]: return None new_stack = [] new_command_history = list(command_history) new_command_history.append([predicate] + args) return new_stack, new_command_history if predicate == 'PColor': if args[0] != COLOR: return None new_stack = exec_stack[:-1] new_stack.append(LIST) return new_stack, list(command_history) if predicate == 'index': if args[0] != LIST or not is_number(args[1]): return None new_stack = exec_stack[:-2] new_stack.append(BEAKER) return new_stack, list(command_history) # history referencing predicates if prefix == 'H': arg_pos = int(predicate[1:]) history_idx_str = args[0] if not is_number(history_idx_str): return None if history_idx_str in ('X1/1', '-1'): return None history_idx = int(history_idx_str) - 1 try: referenced = command_history[history_idx][arg_pos] except IndexError: return None # failed to retrieve return cls._update_state(exec_stack, command_history, referenced) raise ValueError('Invalid predicate: {}'.format(predicate)) class StaticBatchExploration(object): def __init__(self, examples, case_type, max_length, max_utterances, max_stack_depth): """Perform BFS to find silver logical forms. Args: examples (list[Example]) case_type: subclass of Case max_length (int): max # predicates in a logical form max_utterances (int): max # utterances processed by a logical form max_stack_depth (int): max depth of execution stack """ # metrics for reporting start_time = time.time() visited = 0 longest_so_far = 0 max_queue_size = 0 queue = deque(case_type.seeds()) # seed the queue complete = [] while len(queue) != 0: case = queue.popleft() # update metrics visited += 1 max_queue_size = max(max_queue_size, len(queue)) if case.length > longest_so_far: now = time.time() print('reached length {} after visiting {} states ({} s)'.format(case.length, visited, now - start_time)) longest_so_far = max(longest_so_far, case.length) if visited % 100000 == 0: print('visited: {}, completed: {}, peak queue size: {}'.format(visited, len(complete), max_queue_size)) # prune if case.stack_depth > max_stack_depth: continue has_terminated = case.utterances_read >= max_utterances if has_terminated: complete.append(case.path) continue if case.length >= max_length: continue # extend for choice in case.choices: new_case = case.extend(choice) if new_case is None: continue queue.append(new_case) self.complete = complete self.complete = complete # Here just for comparison with StaticBatchExploration # Performs a typical search which uses dynamic execution for pruning. def simple_bfs(example, path_checker, max_depth): root = ParseCase.initial(example.context) queue = deque([root]) terminated = [] start_time = time.time() depth = 0 max_queue_size = 0 for i in itertools.count(): if len(queue) == 0: break max_queue_size = max(max_queue_size, len(queue)) case = queue.popleft() zeros = [0.] * len(case.choices) case.choice_logits = zeros # not used case.choice_log_probs = zeros # not used for choice in case.choices: clone = case.copy_with_decision(choice) # set the decision # don't extend cases with invalid denotation denotation = clone.denotation if isinstance(denotation, Exception): continue path = clone.path if len(path) != depth: depth = len(path) now = time.time() print('reached depth {} after visiting {} states ({}s)'.format(depth, i + 1, now - start_time)) print('peak queue size: {}'.format(max_queue_size)) if path.terminated: # terminates when all the utterances have been processed terminated.append(path) continue if len(path) >= max_depth: continue # Path is not complete. Apply pruning to see if we should continue. if not path_checker(path): continue # Decide to extend this path. new_case = path.extend() queue.append(new_case) silver_lfs = [] for path in terminated: try: if check_denotation(example.answer, path.finalized_denotation): silver_lfs.append(path) except Exception: pass return silver_lfs

ContextualSP/lemon/executor/strongsup/static_exploration.py/0

# import pytest import sys sys.path.append('../../../') from strongsup.example import Example, Context from strongsup.rlong.exploration_policy import AlchemyOraclePathFinder from strongsup.rlong.state import RLongAlchemyState from strongsup.rlong.world import RLongAlchemyWorld from strongsup.rlong.value import RLongStateValue class TestAlchemyExplorationPolicy(object): def test_exploration(self): initial_state = RLongAlchemyState.from_raw_string( '1:ggg 2:_ 3:_ 4:r 5:o 6:ooo 7:gggg') final_state = RLongAlchemyState.from_raw_string( '1:ggg 2:_ 3:_ 4:r 5:o 6:ooo 7:_') num_steps = 2 world = RLongAlchemyWorld(initial_state) context = Context(world, [[""], [""]]) ex = Example(context, answer=[RLongStateValue(final_state)]) print() print(' INIT:', initial_state) print('FINAL:', final_state) print('STEPS:', num_steps) path_finder = AlchemyOraclePathFinder(ex, debug=True) found = set() for path in path_finder.all_actual_paths: finalized = ex.context.executor.finalize(path.denotation) assert finalized[0].state == final_state found.add(' '.join(str(x) for x in path.decisions)) assert 'all-objects -1 index 2 ADrain -1 H1 -1 H2 -1 H0' in found assert 'all-objects -1 index all-objects 2 index APour 1 H2 X1/1 ADrain' in found if __name__ == '__main__': tester = TestAlchemyExplorationPolicy() tester.test_exploration()

ContextualSP/lemon/executor/strongsup/tests/rlong/test_exploration_policy.py/0

import abc import sys from collections import namedtuple import numpy as np import tensorflow as tf from gtd.ml.framework import Feedable, Model from keras.layers import Dense from strongsup.utils import OptimizerOptions, get_optimizer from strongsup.value import check_denotation class ValueFunctionExample(namedtuple('ValueFunctionExample', ['case', 'reward'])): """Represents a single training example for StateValueFunction.train_step. Attributes: case (ParseCase) reward (float): typically 0 or 1 """ __slots__ = () @classmethod def examples_from_paths(cls, paths, example): """Return a list of ValueFunctionExamples derived from ParsePaths discovered during exploration. Args: paths (list[ParsePath]) example (strongsup.example.Example) Returns: list[ValueFunctionExample] """ vf_examples = [] for path in paths: reward = 1 if check_denotation(example.answer, path.finalized_denotation) else 0 vf_examples.extend(ValueFunctionExample(case, reward) for case in path) return vf_examples class StateValueFunction(Model, metaclass=abc.ABCMeta): """Defines a value function that associates a value V to each state s as in RL""" @abc.abstractmethod def values(self, cases): """Returns the values for the states corresponding to a list of cases in the same order. Args: cases (list[ParseCase]): the cases Returns: values (list[float]): the values in same order as cases """ raise NotImplementedError @abc.abstractmethod def loss(self, vf_examples): """Compute the loss for which we are performing gradient descent upon. Args: vf_examples (list[ValueFunctionExample]) Returns: float """ raise NotImplementedError @abc.abstractmethod def train_step(self, vf_examples): """Takes a train step based on training examples Args: vf_examples (list[ValueFunctionExample]) """ raise NotImplementedError class ConstantValueFunction(StateValueFunction): """Gives every state the same value""" def __init__(self, constant_value): self._constant_value = constant_value def values(self, cases): return [self._constant_value] * len(cases) @property def constant_value(self): return self._constant_value def loss(self, vf_examples): """Loss in terms of mean squared error.""" if len(vf_examples) == 0: return 0.0 c = self._constant_value diffs = [(c - ex.reward) for ex in vf_examples] return np.mean(np.power(diffs, 2)) def train_step(self, vf_examples): """Is a no-op""" return class LogisticValueFunction(StateValueFunction, Feedable): def __init__(self, parse_model, learning_rate, optimizer_opt): """ Args: parse_model (ParseModel) learning_rate (float) optimizer_opt (OptimizerOptions) """ with tf.name_scope("LogisticValueFunction"): self._rewards = tf.placeholder( tf.float32, shape=[None], name="rewards") # Prevent gradient from updating the stuff that makes up encoding encodings = tf.stop_gradient(parse_model.case_encodings) self._values = tf.squeeze( Dense(1, activation="sigmoid", bias=True)(encodings), axis=[1]) loss = tf.reduce_mean(tf.contrib.losses.log_loss( self._values, labels=self._rewards)) optimizer = get_optimizer(optimizer_opt)(learning_rate) self._take_step = optimizer.minimize(loss) self._parse_model = parse_model # Hold it around for testing purposes self._loss = loss @classmethod def _unpack_vf_examples(cls, vf_examples): cases = [ex.case for ex in vf_examples] rewards = [ex.reward for ex in vf_examples] return cases, rewards def values(self, cases, ignore_previous_utterances=False): if len(cases) == 0: # Should only happen if everything gets pruned off beam. return [] fetch = {"values": self._values} fetched = self.compute( fetch, cases, rewards=None, ignore_previous_utterances=ignore_previous_utterances) return fetched["values"] def loss(self, vf_examples): if len(vf_examples) == 0: return 0.0 cases, rewards = self._unpack_vf_examples(vf_examples) return self.compute(self._loss, cases, rewards, ignore_previous_utterances=False) def train_step(self, vf_examples): # Make sure all rewards are between [0, 1] for log_loss for ex in vf_examples: assert 0 <= ex.reward <= 1 if len(vf_examples) == 0: print(" WARNING: (ValueFunction) Zero cases \033[F", file=sys.stderr) else: print(" Updating (ValueFunction) ({} cases) \033[F".format( len(vf_examples)), file=sys.stderr) cases, rewards = self._unpack_vf_examples(vf_examples) # Always acknowledge previous utterances on train steps self.compute( self._take_step, cases, rewards, ignore_previous_utterances=False) def inputs_to_feed_dict(self, cases, rewards=None, ignore_previous_utterances=False): feed = {} if rewards: feed[self._rewards] = rewards if len(cases) == 0: raise ValueError("No cases") feed.update(self._parse_model.inputs_to_feed_dict( cases, ignore_previous_utterances, caching=False)) return feed def get_value_function(config, parse_model): """Needs to take the Config for ValueFunction""" if config.type == "constant": return ConstantValueFunction(config.constant_value) elif config.type == "logistic": return LogisticValueFunction( parse_model, config.learning_rate, OptimizerOptions(config.optimizer)) else: raise ValueError( "ValueFunction {} not supported.".format(config.type))

ContextualSP/lemon/executor/strongsup/value_function.py/0

The file [dummy-predictions.csv](dummy-predictions.csv) is a valid example prediction file that can be submitted to the [ARC Challenge Leaderboard](https://leaderboard.allenai.org/). This is a prediction that every question's correct answer is the first choice (either `A` or `1`), and scores about 23% correct.

ContextualSP/lemon/propara_evaluator/aristo-leaderboard/arc/data-challenge/README.md/0

import os import evaluator import unittest import tempfile import typing class TestAccuracy(unittest.TestCase): def test_EverythingCorrect(self): qa = {"Q1": "A", "Q2": "A", "Q3": "A"} p = {"Q1": ["A"], "Q2": ["A"], "Q3": ["A"]} self.assertEqual(3.0 / 3.0, evaluator.calculate_accuracy(qa, p)) def test_EverythingWrong(self): qa = {"Q1": "A", "Q2": "A", "Q3": "A"} p = {"Q1": ["B"], "Q2": ["B"], "Q3": ["B"]} self.assertEqual(0.0 / 3.0, evaluator.calculate_accuracy(qa, p)) def test_MixedResults(self): qa = {"Q1": "A", "Q2": "A", "Q3": "A"} p = {"Q1": ["A"], "Q2": ["A"], "Q3": ["B"]} self.assertEqual(2.0 / 3.0, evaluator.calculate_accuracy(qa, p)) def test_PartialGuess(self): qa = {"Q1": "A", "Q2": "A", "Q3": "A"} p = {"Q1": ["A", "B"], "Q2": ["B"], "Q3": ["B"]} self.assertEqual(0.5 / 3, evaluator.calculate_accuracy(qa, p)) def test_ExtraPredictions(self): qa = {"Q1": "A", "Q2": "A", "Q3": "A"} p = {"Q1": ["A"], "Q2": ["A"], "Q3": ["B"], "QExtra": ["X"]} with self.assertRaises(SystemExit) as context: evaluator.calculate_accuracy(qa, p) self.assertEqual(context.exception.code, evaluator.EXIT_STATUS_PREDICTIONS_EXTRA) def test_MissingPredictions(self): qa = {"Q1": "A", "Q2": "A", "Q3": "A"} p = {"Q1": ["A"], "Q2": ["A"]} with self.assertRaises(SystemExit) as context: evaluator.calculate_accuracy(qa, p) self.assertEqual(context.exception.code, evaluator.EXIT_STATUS_PREDICTION_MISSING) def temp_file_with_contents(lines: typing.List[str]) -> str: t = tempfile.NamedTemporaryFile(mode='wt', delete=False) t.writelines(lines) t.close() return t.name class TestReadAnswers(unittest.TestCase): def test_ReadAnswers(self): t = temp_file_with_contents([ '{"id": "Q1", "answerKey": "A"}\n', '{"id": "Q2", "answerKey": "B"}\n', '{"id": "Q3", "answerKey": "C"}\n', ]) answers = evaluator.read_answers(t) os.remove(t) self.assertEqual(answers, {"Q1": "A", "Q2": "B", "Q3": "C"}) def test_ReadAnswersEmpty(self): t = temp_file_with_contents([]) with self.assertRaises(SystemExit) as context: evaluator.read_answers(t) os.remove(t) self.assertEqual(context.exception.code, evaluator.EXIT_STATUS_ANSWERS_MALFORMED) def test_ReadAnswersCorrupted(self): t = temp_file_with_contents(['this is not json']) with self.assertRaises(SystemExit) as context: evaluator.read_answers(t) os.remove(t) self.assertEqual(context.exception.code, evaluator.EXIT_STATUS_ANSWERS_MALFORMED) def test_ReadAnswersRepeated(self): t = temp_file_with_contents([ '{"id": "Q1", "answerKey": "A"}\n', '{"id": "Q1", "answerKey": "B"}\n', ]) with self.assertRaises(SystemExit) as context: evaluator.read_answers(t) os.remove(t) self.assertEqual(context.exception.code, evaluator.EXIT_STATUS_ANSWERS_MALFORMED) class TestReadPredictions(unittest.TestCase): def test_ReadPredictions(self): t = temp_file_with_contents([ 'Q1,A\n', '"Q2",A;B\n', 'Q3,"A;B;C"\n', ]) predictions = evaluator.read_predictions(t) os.remove(t) self.assertEqual(predictions, { "Q1": ["A"], "Q2": ["A", "B"], "Q3": ["A", "B", "C"], }) def test_ReadPredictionsMissingColumn(self): t = temp_file_with_contents([ 'Q1,A\n', '"Q2"\n', ]) with self.assertRaises(SystemExit) as context: evaluator.read_predictions(t) os.remove(t) self.assertEqual(context.exception.code, evaluator.EXIT_STATUS_PREDICTIONS_MALFORMED) def test_ReadPredictionsRepeated(self): t = temp_file_with_contents([ 'Q1,A\n', 'Q1,A\n', ]) with self.assertRaises(SystemExit) as context: evaluator.read_predictions(t) os.remove(t) self.assertEqual(context.exception.code, evaluator.EXIT_STATUS_PREDICTIONS_MALFORMED) def test_ReadPredictionsCorruptedEmptyKey(self): t = temp_file_with_contents([ ',A\n', ]) with self.assertRaises(SystemExit) as context: evaluator.read_predictions(t) os.remove(t) self.assertEqual(context.exception.code, evaluator.EXIT_STATUS_PREDICTIONS_MALFORMED) def test_ReadPredictionsCorruptedEmptyLabels(self): t = temp_file_with_contents([ 'Q1,A;\n', ]) with self.assertRaises(SystemExit) as context: p = evaluator.read_predictions(t) print(p) os.remove(t) self.assertEqual(context.exception.code, evaluator.EXIT_STATUS_PREDICTIONS_MALFORMED) if __name__ == '__main__': unittest.main()

ContextualSP/lemon/propara_evaluator/aristo-leaderboard/arc/evaluator/test_evaluator.py/0

The mapping of chain ids to correct labels for the dev and test splits are in these files: * dev: chainid_to_label_dev.json * test: chainid_to_label_test.json ## Dummy predictions As a convenienece for testing the evaluator, two "dummy" prediction files are provided which give a score of 0.5 to all chains for both splits: * dev: dummy_predictions_dev.jsonl * test: dummy_predictions_test.jsonl These prediction files were created like this: * dev: `cat chainid_to_label_dev.json | jq -c '. | keys[] | {"chain_id":., "score":0.5}' > dummy_predictions_dev.jsonl` * test: `cat chainid_to_label_test.json | jq -c '. | keys[] | {"chain_id":., "score":0.5}' > dummy_predictions_test.jsonl` You can use these as inputs to the predictor, to confirm that the evaluator is working as expected. The scores you should expect from these dummy predictions are: * dev: `{"auc_roc": 0.5, "explainP1": 0.23612622415669204, "explainNDCG": 0.4791226010631029}` * test: `{"auc_roc": 0.5, "explainP1": 0.2174863387978142, "explainNDCG": 0.48617247810718606}`

ContextualSP/lemon/propara_evaluator/aristo-leaderboard/eqasc/data/README.md/0

**/__pycache__

ContextualSP/lemon/propara_evaluator/aristo-leaderboard/propara/evaluator/.dockerignore/0

from typing import List, NamedTuple, Dict from process.constants import NO_LOCATION, CREATE, DESTROY, MOVE class Input(NamedTuple): participants: str class Output(NamedTuple): participants: str class Conversion(NamedTuple): created: str destroyed: str locations: str step_id: str class Move(NamedTuple): participants: str location_before: str location_after: str step_id: str class Process(NamedTuple): process_id: int locations: Dict actions: Dict num_steps: int # Q1: What are the inputs? # - If a participant exists in state1, but does not exist in the end stateN, it's an input. def inputs(self) -> List[Input]: inputs = [] # type: List[Input] for participant in self.locations.keys(): actions = self.actions[participant] if _is_this_action_seq_of_an_input(actions): inputs.append(Input(participants=_summarize_participants(participant))) return inputs # Q2: What are the outputs # - If a participant does not exist in state1, but exists in the end stateN, it's an output. def outputs(self) -> List[Output]: outputs = [] # type: List[Output] for participant in self.locations.keys(): actions = self.actions[participant] if _is_this_action_seq_of_an_output(actions): outputs.append(Output(participants=_summarize_participants(participant))) return outputs # Q3: What is converted? # tuple: (participant-list-from, participant-list-to, loc-list, step-id) # a. For any event with BOTH "D" and "C" in: # The "D" participants are converted to the "C" participants at the union of the D and C locations # b. IF an event has ONLY "D" but no "C" in ("M" is ok - irrelevant) # AND the NEXT event has ONLY "C" but no "D" in ("M" is ok - irrelevant) # THEN the "D" participants are converted to the "C" participants at the union of the D and C locations def conversions(self) -> List[Conversion]: conversions = [] # type: List[Conversion] for step_id in range(1, self.num_steps + 1): (created, c_locations) = self._get_created_at_step(step_id) (destroyed, d_locations) = self._get_destroyed_at_step(step_id) if created and destroyed: conversions.append(Conversion( destroyed=_conjunction(*destroyed), created=_conjunction(*created), locations=_conjunction(*set(c_locations + d_locations)), step_id=str(step_id) )) elif destroyed and step_id < self.num_steps - 1: (created2, c_locations2) = self._get_created_at_step(step_id + 1) (destroyed2, d_locations2) = self._get_destroyed_at_step(step_id + 1) created_but_not_destroyed = set(created2) - set(destroyed) if not destroyed2 and created_but_not_destroyed: conversions.append(Conversion( destroyed=_conjunction(*destroyed), created=_conjunction(*created_but_not_destroyed), locations=_conjunction(*set(c_locations2 + d_locations)), step_id=str(step_id) )) elif created and step_id < self.num_steps - 1: (created2, c_locations2) = self._get_created_at_step(step_id + 1) (destroyed2, d_locations2) = self._get_destroyed_at_step(step_id + 1) destroyed_but_not_created = set(destroyed2) - set(created) if not created2 and destroyed_but_not_created: conversions.append(Conversion( destroyed=_conjunction(*destroyed_but_not_created), created=_conjunction(*created), locations=_conjunction(*set(c_locations + d_locations2)), step_id=str(step_id) )) return conversions # Q4: What is moved? # tuple: (participant, from-loc, to-loc, step-id) # return all moves def moves(self): moves = [] for participant in self.locations.keys(): locations = self.locations[participant] actions = self.actions[participant] for step_id in range(1, len(locations)): is_moved = actions[step_id - 1] == MOVE or ( locations[step_id - 1] != NO_LOCATION and locations[step_id] != NO_LOCATION and locations[step_id - 1] != locations[step_id] ) if not is_moved: continue moves.append(Move( participants=_summarize_participants(participant), location_before=locations[step_id - 1], location_after=locations[step_id], step_id=str(step_id) )) return moves def _get_created_at_step(self, step_id: int): created = [] locations = [] for participant in self.locations.keys(): state_values = self.locations[participant] is_creation = state_values[step_id - 1] == NO_LOCATION \ and state_values[step_id] != NO_LOCATION if is_creation: created.append(_summarize_participants(participant)) locations.append(state_values[step_id]) return created, locations def _get_destroyed_at_step(self, step_id: int): destroyed = [] locations = [] for participant in self.locations.keys(): state_values = self.locations[participant] is_destruction = state_values[step_id - 1] != NO_LOCATION \ and state_values[step_id] == NO_LOCATION if is_destruction: destroyed.append(_summarize_participants(participant)) locations.append(state_values[step_id - 1]) return destroyed, locations def _is_this_action_seq_of_an_output(actions) -> bool: for action_id, _ in enumerate(actions): no_destroy_move_before = DESTROY not in actions[0:action_id] and MOVE not in actions[0:action_id] current_create = actions[action_id] == CREATE no_destroy_later = DESTROY not in actions[action_id + 1:] if no_destroy_move_before and current_create and no_destroy_later: return True return False def _is_this_action_seq_of_an_input(actions) -> bool: for action_id, _ in enumerate(actions): no_create_before = CREATE not in actions[0:action_id] # last action_id must be checked current_destroy = actions[action_id] == DESTROY no_create_move_later = CREATE not in actions[action_id + 1:] and MOVE not in actions[action_id + 1:] if no_create_before and current_destroy and no_create_move_later: return True return False def _split_participants(participant) -> List[str]: return [p.strip() for p in participant.split(';')] def _summarize_participants(participant) -> str: return ' OR '.join(_split_participants(participant)) def _conjunction(*things) -> str: return ' AND '.join(things)

ContextualSP/lemon/propara_evaluator/aristo-leaderboard/propara/evaluator/process/process.py/0

#!/usr/bin/env python # coding=utf-8 # Copyright The HuggingFace Team and The HuggingFace Inc. team. All rights reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """ Fine-tuning the library models for sequence to sequence. """ # You can also adapt this script on your own sequence to sequence task. Pointers for this are left as comments. import logging import os import sys from dataclasses import dataclass, field from typing import Optional import json import datasets datasets.set_caching_enabled(False) import numpy as np from datasets import load_dataset, load_metric from parameters16g_es_corpusb import * from copy import deepcopy import transformers from transformers import ( AutoConfig, AutoModelForSeq2SeqLM, AutoTokenizer, DataCollatorForSeq2Seq, HfArgumentParser, M2M100Tokenizer, MBart50Tokenizer, MBart50TokenizerFast, MBartTokenizer, MBartTokenizerFast, Seq2SeqTrainer, Seq2SeqTrainingArguments, default_data_collator, set_seed, ) from gan_dataset import DataCollatorForGAN from transformers.trainer_utils import get_last_checkpoint from transformers.utils import check_min_version from transformers.utils.versions import require_version from GenTrainer import GenTrainer from modeling_t5_with_loss import T5ForConditionalGeneration ##### Softscore loss # from modeling_bart import BartForConditionalGeneration ##### Softscore loss os.environ["WANDB_DISABLED"] = "true" logger = logging.getLogger(__name__) # A list of all multilingual tokenizer which require src_lang and tgt_lang attributes. MULTILINGUAL_TOKENIZERS = [MBartTokenizer, MBartTokenizerFast, MBart50Tokenizer, MBart50TokenizerFast, M2M100Tokenizer] @dataclass class ModelArguments: """ Arguments pertaining to which model/config/tokenizer we are going to fine-tune from. """ model_name_or_path: str = field( metadata={"help": "Path to pretrained model or model identifier from huggingface.co/models"} ) config_name: Optional[str] = field( default=None, metadata={"help": "Pretrained config name or path if not the same as model_name"} ) tokenizer_name: Optional[str] = field( default=None, metadata={"help": "Pretrained tokenizer name or path if not the same as model_name"} ) cache_dir: Optional[str] = field( default=None, metadata={"help": "Where to store the pretrained models downloaded from huggingface.co"}, ) use_fast_tokenizer: bool = field( default=True, metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."}, ) model_revision: str = field( default="main", metadata={"help": "The specific model version to use (can be a branch name, tag name or commit id)."}, ) use_auth_token: bool = field( default=False, metadata={ "help": "Will use the token generated when running `transformers-cli login` (necessary to use this script " "with private models)." }, ) @dataclass class DataTrainingArguments: """ Arguments pertaining to what data we are going to input our model for training and eval. """ dataset_name: Optional[str] = field( default=None, metadata={"help": "The name of the dataset to use (via the datasets library)."} ) dataset_config_name: Optional[str] = field( default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."} ) train_file: Optional[str] = field(default=None, metadata={"help": "The input training data file (a jsonlines)."}) validation_file: Optional[str] = field( default=None, metadata={ "help": "An optional input evaluation data file to evaluate the metrics (sacreblue) on " "a jsonlines file." }, ) test_file: Optional[str] = field( default=None, metadata={ "help": "An optional input test data file to evaluate the metrics (sacreblue) on " "a jsonlines file." }, ) overwrite_cache: bool = field( default=False, metadata={"help": "Overwrite the cached training and evaluation sets"} ) preprocessing_num_workers: Optional[int] = field( default=None, metadata={"help": "The number of processes to use for the preprocessing."}, ) max_source_length: Optional[int] = field( default=1024, metadata={ "help": "The maximum total input sequence length after tokenization. Sequences longer " "than this will be truncated, sequences shorter will be padded." }, ) max_target_length: Optional[int] = field( default=128, metadata={ "help": "The maximum total sequence length for target text after tokenization. Sequences longer " "than this will be truncated, sequences shorter will be padded." }, ) val_max_target_length: Optional[int] = field( default=None, metadata={ "help": "The maximum total sequence length for validation target text after tokenization. Sequences longer " "than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`." "This argument is also used to override the ``max_length`` param of ``model.generate``, which is used " "during ``evaluate`` and ``predict``." }, ) pad_to_max_length: bool = field( default=False, metadata={ "help": "Whether to pad all samples to model maximum sentence length. " "If False, will pad the samples dynamically when batching to the maximum length in the batch. More " "efficient on GPU but very bad for TPU." }, ) max_train_samples: Optional[int] = field( default=None, metadata={ "help": "For debugging purposes or quicker training, truncate the number of training examples to this " "value if set." }, ) max_eval_samples: Optional[int] = field( default=None, metadata={ "help": "For debugging purposes or quicker training, truncate the number of evaluation examples to this " "value if set." }, ) max_predict_samples: Optional[int] = field( default=None, metadata={ "help": "For debugging purposes or quicker training, truncate the number of prediction examples to this " "value if set." }, ) num_beams: Optional[int] = field( default=None, metadata={ "help": "Number of beams to use for evaluation. This argument will be passed to ``model.generate``, " "which is used during ``evaluate`` and ``predict``." }, ) ignore_pad_token_for_loss: bool = field( default=True, metadata={ "help": "Whether to ignore the tokens corresponding to padded labels in the loss computation or not." }, ) source_prefix: Optional[str] = field( default=None, metadata={"help": "A prefix to add before every source text (useful for T5 models)."} ) forced_bos_token: Optional[str] = field( default=None, metadata={ "help": "The token to force as the first generated token after the :obj:`decoder_start_token_id`." "Useful for multilingual models like :doc:`mBART <../model_doc/mbart>` where the first generated token " "needs to be the target language token.(Usually it is the target language token)" }, ) data_dir: Optional[str] = field( default=None, metadata={"help": "Path for data files"}, ) prediction_mode: Optional[str] = field( ################## default="gen", metadata={"help": "Choose from [gen, ver]. gen-train: self-sampling; ver: create verifier adhoc corpus inference."}, ) batch_example_num: Optional[int] = field( default=6, metadata={ "help": "the number of instance (number pos+neg) per batch, default as 6" "value if set." }, ) gan_alpha: float = field( default=0.9, metadata={ "help": "the ratio of the teacher forcing loss in the gan loss" }, ) def __post_init__(self): if self.dataset_name is None and self.train_file is None and self.validation_file is None: raise ValueError("Need either a dataset name or a training/validation file.") # accepting both json and jsonl file extensions, as # many jsonlines files actually have a .json extension valid_extensions = ["json", "jsonl"] if self.train_file is not None: extension = self.train_file.split(".")[-1] assert extension in valid_extensions, "`train_file` should be a jsonlines file." if self.validation_file is not None: extension = self.validation_file.split(".")[-1] assert extension in valid_extensions, "`validation_file` should be a jsonlines file." if self.val_max_target_length is None: self.val_max_target_length = self.max_target_length def main(): # See all possible arguments in src/transformers/training_args.py # or by passing the --help flag to this script. # We now keep distinct sets of args, for a cleaner separation of concerns. import os parser = HfArgumentParser((ModelArguments, DataTrainingArguments, Seq2SeqTrainingArguments)) if len(sys.argv) == 2 and sys.argv[1].endswith(".json"): # If we pass only one argument to the script and it's the path to a json file, # let's parse it to get our arguments. model_args, data_args, training_args = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1])) else: model_args, data_args, training_args = parser.parse_args_into_dataclasses() # Setup logging logging.basicConfig( format="%(asctime)s - %(levelname)s - %(name)s - %(message)s", datefmt="%m/%d/%Y %H:%M:%S", handlers=[logging.StreamHandler(sys.stdout)], ) log_level = training_args.get_process_log_level() logger.setLevel(log_level) datasets.utils.logging.set_verbosity(log_level) transformers.utils.logging.set_verbosity(log_level) transformers.utils.logging.enable_default_handler() transformers.utils.logging.enable_explicit_format() # Log on each process the small summary: logger.warning( f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}" + f"distributed training: {bool(training_args.local_rank != -1)}, 16-bits training: {training_args.fp16}" ) logger.info(f"Training/evaluation parameters {training_args}") if data_args.source_prefix is None and model_args.model_name_or_path in [ "t5-small", "t5-base", "t5-large", "t5-3b", "t5-11b", ]: logger.warning( "You're running a t5 model but didn't provide a source prefix, which is expected, e.g. with " "`--source_prefix 'translate English to German: ' `" ) # Detecting last checkpoint. last_checkpoint = None if os.path.isdir(training_args.output_dir) and training_args.do_train and not training_args.overwrite_output_dir: last_checkpoint = get_last_checkpoint(training_args.output_dir) if last_checkpoint is None and len(os.listdir(training_args.output_dir)) > 0: raise ValueError( f"Output directory ({training_args.output_dir}) already exists and is not empty. " "Use --overwrite_output_dir to overcome." ) elif last_checkpoint is not None and training_args.resume_from_checkpoint is None: logger.info( f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change " "the `--output_dir` or add `--overwrite_output_dir` to train from scratch." ) # Set seed before initializing model. set_seed(training_args.seed) # Get the datasets: you can either provide your own JSON training and evaluation files (see below) # or just provide the name of one of the public datasets available on the hub at https://huggingface.co/datasets/ # (the dataset will be downloaded automatically from the datasets Hub). # # For translation, only JSON files are supported, with one field named "translation" containing two keys for the # source and target languages (unless you adapt what follows). # # In distributed training, the load_dataset function guarantee that only one local process can concurrently # download the dataset. data_files = {} if training_args.do_train: data_files["train"] = os.path.join(data_args.data_dir, data_args.train_file) if training_args.do_eval: data_files['validation'] = os.path.join(data_args.data_dir, data_args.validation_file) if training_args.do_predict: data_files['test'] = os.path.join(data_args.data_dir, data_args.test_file) # data_files = { # 'train': os.path.join(data_args.data_dir, data_args.train_file) if training_args.do_train else None, # 'validation': os.path.join(data_args.data_dir, data_args.validation_file) if training_args.do_eval else None, # 'test': os.path.join(data_args.data_dir, data_args.test_file) if training_args.do_predict else None, # } print(data_files) raw_datasets = load_dataset('json', data_files=data_files, cache_dir=model_args.cache_dir, download_mode='force_redownload') # See more about loading any type of standard or custom dataset (from files, python dict, pandas DataFrame, etc) at # https://huggingface.co/docs/datasets/loading_datasets.html. # Load pretrained model and tokenizer # # Distributed training: # The .from_pretrained methods guarantee that only one local process can concurrently # download model & vocab. config = AutoConfig.from_pretrained( model_args.config_name if model_args.config_name else model_args.model_name_or_path, cache_dir=model_args.cache_dir, revision=model_args.model_revision, use_auth_token=True if model_args.use_auth_token else None, ) tokenizer = AutoTokenizer.from_pretrained( model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path, cache_dir=model_args.cache_dir, use_fast=model_args.use_fast_tokenizer, revision=model_args.model_revision, use_auth_token=True if model_args.use_auth_token else None, ) model = T5ForConditionalGeneration.from_pretrained( ################# model_args.model_name_or_path, from_tf=bool(".ckpt" in model_args.model_name_or_path), config=config, cache_dir=model_args.cache_dir, revision=model_args.model_revision, use_auth_token=True if model_args.use_auth_token else None, ) special_tokens=['[SEP]','[MASK]'] if training_args.do_train and not all([t in tokenizer.vocab for t in special_tokens]): special_tokens_dict = {'additional_special_tokens': special_tokens} num_added_toks = tokenizer.add_special_tokens(special_tokens_dict) added_tokens = tokenizer.get_added_vocab() logger.info('Added tokens: {}'.format(added_tokens)) model.resize_token_embeddings(len(tokenizer)) # Set decoder_start_token_id if model.config.decoder_start_token_id is None and isinstance(tokenizer, (MBartTokenizer, MBartTokenizerFast)): if isinstance(tokenizer, MBartTokenizer): model.config.decoder_start_token_id = tokenizer.lang_code_to_id[data_args.target_lang] else: model.config.decoder_start_token_id = tokenizer.convert_tokens_to_ids(data_args.target_lang) if model.config.decoder_start_token_id is None: raise ValueError("Make sure that `config.decoder_start_token_id` is correctly defined") prefix = data_args.source_prefix if data_args.source_prefix is not None else "" # Preprocessing the datasets. # We need to tokenize inputs and targets. if training_args.do_train: column_names = raw_datasets["train"].column_names elif training_args.do_eval: column_names = raw_datasets["validation"].column_names elif training_args.do_predict: column_names = raw_datasets["test"].column_names else: logger.info("There is nothing to do. Please pass `do_train`, `do_eval` and/or `do_predict`.") return # Temporarily set max_target_length for training. max_target_length = data_args.max_target_length padding = "max_length" if data_args.pad_to_max_length else False if training_args.label_smoothing_factor > 0 and not hasattr(model, "prepare_decoder_input_ids_from_labels"): logger.warning( "label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for" f"`{model.__class__.__name__}`. This will lead to loss being calculated twice and will take up more memory" ) def preprocess_function(examples): inputs = examples['input'] #[ex for ex in examples["input"]] # scores = examples["ver_prob"] targets = examples['conclusions'] inputs = [prefix + inp for inp in inputs] model_inputs = tokenizer(inputs, max_length=data_args.max_source_length, padding=padding, truncation=True) if not (data_args.prediction_mode == "gen" and training_args.do_train): output_labels = tokenizer([tar[-1] for tar in targets], max_length=max_target_length, padding=padding, truncation=True) if padding == "max_length" and data_args.ignore_pad_token_for_loss: output_labels["input_ids"] = [ [(l if l != tokenizer.pad_token_id else -100) for l in label] for label in labels["input_ids"] ] output_labels = output_labels['input_ids'] else: # Setup the tokenizer for targets output_labels = [] with tokenizer.as_target_tokenizer(): for i in range(len(inputs)): labels = tokenizer(targets[i],max_length=max_target_length, padding=padding, truncation=True) if padding == "max_length" and data_args.ignore_pad_token_for_loss: labels["input_ids"] = [ [(l if l != tokenizer.pad_token_id else -100) for l in label] for label in labels["input_ids"] ] output_labels.append(labels['input_ids']) # labels = tokenizer(targets, max_length=max_target_length, padding=padding, truncation=True) model_inputs["labels"] = output_labels if data_args.prediction_mode == "gen" and training_args.do_train: model_inputs["is_gold"] = examples['is_gold'] model_inputs["ver_prob"] = examples["ver_prob"] return model_inputs if training_args.do_train: if "train" not in raw_datasets: raise ValueError("--do_train requires a train dataset") train_dataset = raw_datasets["train"] if data_args.max_train_samples is not None: np.random.seed(training_args.seed) indices = np.random.choice(len(train_dataset["input"]), data_args.max_train_samples, replace=False) # magic "input" train_dataset = train_dataset.select(indices) # Mine # train_dataset = train_dataset.select(range(data_args.max_train_samples)) ####### Original # train_dataset = train_dataset.select(np.random.choice(len(train_dataset["input"]), int(0.12 * len(train_dataset["input"])), replace=False)) with training_args.main_process_first(desc="train dataset map pre-processing"): # print(train_dataset.column_names) # print([len(train_dataset[column]) for column in train_dataset.column_names]) # exit() column_names = train_dataset.column_names train_dataset = train_dataset.map( preprocess_function, batched=True, num_proc=data_args.preprocessing_num_workers, remove_columns=column_names, load_from_cache_file=not data_args.overwrite_cache, desc="Running tokenizer on train dataset" ) if training_args.do_eval: max_target_length = data_args.val_max_target_length if "validation" not in raw_datasets: raise ValueError("--do_eval requires a validation dataset") eval_dataset = raw_datasets["validation"] if data_args.max_eval_samples is not None: # eval_dataset = eval_dataset.select(range(data_args.max_eval_samples)) # Original np.random.seed(training_args.seed) indices = np.random.choice(len(eval_dataset ["input"]), data_args.max_train_samples, replace=False) # magic "input" eval_dataset = eval_dataset.select(indices) # Mine with training_args.main_process_first(desc="validation dataset map pre-processing"): eval_dataset = eval_dataset.map( preprocess_function, batched=True, num_proc=data_args.preprocessing_num_workers, remove_columns=column_names, load_from_cache_file=not data_args.overwrite_cache, desc="Running tokenizer on validation dataset", ) if training_args.do_predict: max_target_length = data_args.val_max_target_length if "test" not in raw_datasets: raise ValueError("--do_predict requires a test dataset") predict_dataset = raw_datasets["test"] predict_dataset_copy = deepcopy(predict_dataset) if data_args.max_predict_samples is not None: # predict_dataset = predict_dataset.select(range(data_args.max_predict_samples)) np.random.seed(training_args.seed) indices = np.random.choice(len(predict_dataset["input"]), data_args.max_predict_samples, replace=False) # magic "input" predict_dataset_copy = predict_dataset_copy.select(indices) predict_dataset = predict_dataset.select(indices) # Mine with training_args.main_process_first(desc="prediction dataset map pre-processing"): predict_dataset = predict_dataset.map( preprocess_function, batched=True, num_proc=data_args.preprocessing_num_workers, remove_columns=column_names, load_from_cache_file=not data_args.overwrite_cache, desc="Running tokenizer on prediction dataset", ) batch_size = gen_per_device_examples_num # Data collator label_pad_token_id = -100 if data_args.ignore_pad_token_for_loss else tokenizer.pad_token_id # if data_args.pad_to_max_length: # data_collator = default_data_collator # else: if data_args.prediction_mode == "gen" and training_args.do_train: data_collator = DataCollatorForGAN( tokenizer, model=model, label_pad_token_id=label_pad_token_id, pad_to_multiple_of=8 if training_args.fp16 else None, max_instance_num=batch_size ) else: data_collator = DataCollatorForSeq2Seq( tokenizer, model=model, label_pad_token_id=label_pad_token_id, pad_to_multiple_of=8 if training_args.fp16 else None ) # Metric metric = load_metric("sacrebleu") def postprocess_text(preds, labels): preds = [pred.strip() for pred in preds] labels = [[label.strip()] for label in labels] return preds, labels def compute_metrics(eval_preds): preds, labels = eval_preds if isinstance(preds, tuple): preds = preds[0] decoded_preds = tokenizer.batch_decode(preds, skip_special_tokens=True) if data_args.ignore_pad_token_for_loss: # Replace -100 in the labels as we can't decode them. labels = np.where(labels != -100, labels, tokenizer.pad_token_id) decoded_labels = tokenizer.batch_decode(labels, skip_special_tokens=True) # Some simple post-processing decoded_preds, decoded_labels = postprocess_text(decoded_preds, decoded_labels) result = metric.compute(predictions=decoded_preds, references=decoded_labels) result = {"bleu": result["score"]} prediction_lens = [np.count_nonzero(pred != tokenizer.pad_token_id) for pred in preds] result["gen_len"] = np.mean(prediction_lens) result = {k: round(v, 4) for k, v in result.items()} return result # Initialize our Trainer trainer = GenTrainer( ######################### model=model, args=training_args, train_dataset=train_dataset if training_args.do_train else None, eval_dataset=eval_dataset if training_args.do_eval else None, tokenizer=tokenizer, data_collator=data_collator, compute_metrics=compute_metrics if training_args.predict_with_generate else None, num_return_seq=num_return_seq, num_beams=gen_num_beams, gan_alpha=data_args.gan_alpha ) # Training if training_args.do_train: checkpoint = None if training_args.resume_from_checkpoint is not None: checkpoint = training_args.resume_from_checkpoint elif last_checkpoint is not None: checkpoint = last_checkpoint train_result = trainer.train(resume_from_checkpoint=checkpoint) trainer.save_model() # Saves the tokenizer too for easy upload metrics = train_result.metrics max_train_samples = ( data_args.max_train_samples if data_args.max_train_samples is not None else len(train_dataset) ) metrics["train_samples"] = min(max_train_samples, len(train_dataset)) trainer.log_metrics("train", metrics) trainer.save_metrics("train", metrics) trainer.save_state() # Evaluation results = {} max_length = ( training_args.generation_max_length if training_args.generation_max_length is not None else data_args.val_max_target_length ) num_beams = data_args.num_beams if data_args.num_beams is not None else training_args.generation_num_beams if training_args.do_eval: logger.info("*** Evaluate ***") metrics = trainer.evaluate(max_length=max_length, num_beams=num_beams, metric_key_prefix="eval") max_eval_samples = data_args.max_eval_samples if data_args.max_eval_samples is not None else len(eval_dataset) metrics["eval_samples"] = min(max_eval_samples, len(eval_dataset)) trainer.log_metrics("eval", metrics) trainer.save_metrics("eval", metrics) # Prediction if training_args.do_predict: logger.info("*** Predict ***") predict_results = trainer.predict( predict_dataset, metric_key_prefix="predict", max_length=max_length, num_beams=num_beams ) metrics = predict_results.metrics max_predict_samples = ( data_args.max_predict_samples if data_args.max_predict_samples is not None else len(predict_dataset) ) metrics["predict_samples"] = min(max_predict_samples, len(predict_dataset)) trainer.log_metrics("predict", metrics) trainer.save_metrics("predict", metrics) if trainer.is_world_process_zero(): if training_args.predict_with_generate: outputs = predict_results.predictions predictions = tokenizer.batch_decode( outputs.reshape(outputs.shape[0] * outputs.shape[1], outputs.shape[-1]), skip_special_tokens=True, clean_up_tokenization_spaces=True ) predictions = [pred.strip() for pred in predictions] # predictions = tokenizer.batch_decode( # predict_results.predictions, skip_special_tokens=True, clean_up_tokenization_spaces=True # ) # predictions = [pred.strip() for pred in predictions] output_prediction_file=None if data_args.prediction_mode == "gen": output_prediction_file = os.path.join(data_args.data_dir, unlabeled_gen_train_iter_file) ##### inputs = predict_dataset_copy["input"] outputs = predict_dataset_copy["conclusions"] is_golds = predict_dataset_copy["is_gold"] with open(output_prediction_file, "w", encoding="utf-8") as f: for index,(i,o,igs) in enumerate(zip(inputs, outputs ,is_golds)): gold_idx = igs.index(1) # print('-----------------------') # print(len(predictions),index) # print(predictions[index*num_return_seq:(index+1)*num_return_seq]) example1 = {"input": i, "conclusions": [o[gold_idx]]+predictions[index*num_return_seq:(index+1)*num_return_seq], "is_gold": [1]+[0]*num_return_seq, "ver_prob": [-1]*(num_return_seq+1)} # example2 = {"input": i, "conclusion": p, "is_gold": 0, "ver_prob": -1} json.dump(example1, f); f.write("\n") # json.dump(example2, f); f.write("\n") elif data_args.prediction_mode == "ver": output_prediction_file = os.path.join(data_args.data_dir, unlabeled_ver_train_iter_file) ##### inputs = predict_dataset_copy["input"] outputs = predict_dataset_copy["conclusions"] is_golds = predict_dataset_copy["is_gold"] with open(output_prediction_file, "w", encoding="utf-8") as f: for index,(i,o,igs) in enumerate(zip(inputs, outputs ,is_golds)): gold_idx = igs.index(1) p = predictions[index * num_return_seq:(index + 1) * num_return_seq] gt = o[gold_idx] all_ids = list(range(len(igs))) all_ids.remove(gold_idx) es_negs = [o[nid] for nid in list(np.random.choice(list(all_ids),2,replace=False))] example = {"input": i, "conclusions": [gt,p[0]]+es_negs, "is_gold":[1,0]+[0]*len(es_negs)} json.dump(example, f) f.write("\n") ## Default write to file # with open(output_prediction_file, "w", encoding="utf-8") as writer: # writer.write("\n".join(predictions)) kwargs = {"finetuned_from": model_args.model_name_or_path, "tasks": "translation"} if data_args.dataset_name is not None: kwargs["dataset_tags"] = data_args.dataset_name if data_args.dataset_config_name is not None: kwargs["dataset_args"] = data_args.dataset_config_name kwargs["dataset"] = f"{data_args.dataset_name} {data_args.dataset_config_name}" else: kwargs["dataset"] = data_args.dataset_name # languages = [l for l in [data_args.source_lang, data_args.target_lang] if l is not None] # if len(languages) > 0: # kwargs["language"] = languages # if training_args.push_to_hub: # trainer.push_to_hub(**kwargs) # else: # trainer.create_model_card(**kwargs) return results def _mp_fn(index): # For xla_spawn (TPUs) main() if __name__ == "__main__": main()

ContextualSP/logigan/pre-training/hf_generation_multi_es.py/0

import numpy as np from collections import defaultdict import re from nltk.corpus import stopwords from enum import Enum from itertools import permutations import re import json import random # words = stopwords.words('english') from collections import defaultdict from functools import reduce class ResType(Enum): ENTITY = 1, RELATION = 2 class Sample: def __init__(self, query, sparql, tag): self.query = query self.sparql = sparql self.tag = tag def string(): return '\t'.join([self.query, self.sparql, self.tag]) class Helper: def __init__(self): self.stop_words = ["Did", "and", ",", "'s", "M0", "M1", "M2", "M3", "M4", "M5","M6", "whose", "Whose", \ "What", "did", "Was", "was", "Which", "Were", "were", "that", "M", "a"] self.stop_words += stopwords.words('english') def load_phrase_table(self): self.dict = defaultdict(list) data = open("./coor_file-0520.filter_wto_no") for i in data: i = eval(i.strip())[0] self.dict[i[0]].append(i[1]) # maxlen = max(maxlen, len(i[0].split())) # if len(i[0].split()) > 3: # print(i) # print(self.dict) # self.indexTree = Trie() # for key in self.dict: # self.indexTree.add(key) def count_var(self, lf): value = {'?x0':2.5, '?x1':2, '?x2':1.5, '?x3':1, '?x4':0.5, '?x5':0} a1, r, a2 = lf.split() cn1, cn2, cn3 = 0, 0, 0 if a1.startswith("?x"): cn1 = 5 cn1 += value[a1] if a2.startswith("?x"): cn2 = 3 cn2 += value[a2] if r == 'a': cn3 = -1 return cn1 + cn2 + cn3 def add_index(self, file): data = np.array(open(file).readlines()).reshape(-1, 3) res = [] for item in data: _, src, trg = item # print(_, src, trg) src_new = [] for idx, token in enumerate(src.split(), 1): src_new.append(f'{token} ({idx})') res.append('\n'.join([_, ' '.join(src_new), trg])) return "\n".join(res) def update_output_format(self, version): for type in ["formula2query", "query2formula"]: file = f"/home/v-yinguo/DKI/GIZA_alignment/{version}/{type}/{type}.A3.final" open(f'{file}.update', "w").write(self.add_index(file)) def statistic_coor(self, q2f, f2q): ## type = 1是query2formula的映射 ## type = 0是formula2query的映射 def split_up(sent): return [token for token in sent.split() if not token.startswith("(")] def split_bottom(sent): tokens = re.split(r' *\({[0-9 ]*}\) *', sent)[1:-1] return tokens def split_alignment(sent): alignment = [re.findall(r'[0-9].', i) for i in re.findall(r'\({[0-9 ]*}\)', sent)] ##把开始的NULL去掉 # print(alignment) alignment = [[int(ii) for ii in i] if len(i) else [0] for i in alignment[1:]] ## 将alignment 处理成连续的span alignment_combine = [] # print(alignment) for sub_align in alignment: sub_align_span = [] start, end = 0, 0 for idx in range(len(sub_align) - 1): if sub_align[idx + 1] == sub_align[idx] + 1: end += 1 else: sub_align_span.append((sub_align[start], sub_align[end])) start, end = idx + 1, idx + 1 sub_align_span.append((sub_align[start], sub_align[end])) alignment_combine.append(sub_align_span) return alignment_combine coor_dict = defaultdict(int) for type, file in enumerate([f2q, q2f]): # print(f"file name:{file}") for index, line in enumerate(open(file),1): line = line.strip() # print(line) if (not type and index % 5 == 3) or (type and index % 5 == 4) : query = line elif (not type and index % 5 == 4) or (type and index % 5 == 3) : formula = line elif index %5 == 0: query_tokens = split_bottom(query) if type else split_up(query) formula_tokens= split_up(formula) if type else split_bottom(formula) alignment = split_alignment(query) if type else split_alignment(formula) # print(f"len formula:{len(formula_tokens)}, len(query):{len(query_tokens)},len alignment:{len(alignment)}") # print(f"query:{query_tokens}\nformula:{formula_tokens}\nalignment:{alignment}") assert(len(formula_tokens) == len(alignment) or len(query_tokens) == len(alignment)) if not type: for spans, formula_token in zip(alignment, formula_tokens): for sub_span in spans: s_pos, e_pos = sub_span[0], sub_span[1] query_token = ' '.join(query_tokens[s_pos-1:e_pos]) if s_pos == e_pos: if s_pos and not re.match(r'M[0-9]', query_token) and query_token not in self.stop_words: # print(f"1-1 matching:{s_pos}-{e_pos} {query_token}-{formula_token}") coor_dict[(query_token, formula_token)] += 1 else: # print(f"multi matching:{s_pos}-{e_pos} {query_token}-{formula_token}") coor_dict[(query_token, formula_token)] += 1 else: # print(f"\n\nori query:{query}\nori formula:{formula}") # print(f"query:{query_tokens}\nformula:{formula_tokens}\nalignment:{alignment}") for spans, query_token in zip(alignment, query_tokens): if query_token in self.stop_words or re.match(r'M[0-9]', query_token): continue for sub_span in spans: s_pos, e_pos = sub_span[0], sub_span[1] formula_token = ' '.join(formula_tokens[s_pos-1:e_pos]) if s_pos == e_pos: if s_pos: # print(f"1-1 matching:{s_pos}-{e_pos} {query_token}-{formula_token}") coor_dict[(query_token, formula_token)] += 1 else: # print(f"multi matching:{s_pos}-{e_pos} {query_token}-{formula_token}") coor_dict[(query_token, formula_token)] += 1 pass # print(coor_dict) sorted_coor_dict = sorted(coor_dict.items(), key=lambda s:s[1], reverse = True) # print(sorted_coor_dict) return sorted_coor_dict def filter_str(self, line): qf, cnt = eval(line.strip()) key, v = qf if key in self.stop_words: return if v.count("|||") > 1: return key_entities = re.findall(r'M[0-9]', key) v_entities = re.findall(r'M[0-9]', v) key_entities.sort() v_entities.sort() if len(key_entities) == 0: v = re.sub(r'M[0-9]', 'M', v) # coor_dict[(key, v)] += cnt elif len(v_entities) == 0: key = re.sub(r'M[0-9]', 'M', key) elif len(key_entities) == len(v_entities): key = re.sub(r'M[0-9]', 'M', key) v = re.sub(r'M[0-9]', 'M', v) else: return while(len(key)): if key.split()[0] in self.stop_words and '#is#M' not in v: key = ' '.join(key.split()[1:]) elif key.split()[-1] in self.stop_words and '#is#M' not in v: key = ' '.join(key.split()[:-1]) else: break if len(key) == 0: return if len(key.split()) == 1 and len(v.split(" ")) > 2: return if len(key.split()) == 1 and (("FILTER" not in v and len(v.split("|||")) > 1) or ("FILTER" in v and len(v.split("|||")) > 2)): return if len(key.split()) > 1 and len(set(key.split()) - set(self.stop_words)) == 0 and not '#is#M' in v: return if v.startswith("FILTER"): return v = list(set(v.split())) if len(v) > 1: return v.sort() v = ' '.join(v) v = re.sub(r'\?x[0-9]', '?x', v) return (key, v), cnt def filter_result(self, src1): coor_dict = defaultdict(int) ## src1是 debug_opt for file in [open(src1)]: for line in file: # print("after key:", key) result = self.filter_str(line) if result: coor_dict[result[0]] += result[1] # elif len(key_entities) == len(v_entities): # key = re.sub(r'M[0-9]', 'M', key) # v = re.sub(r'M[0-9]', 'M', v) # coor_dict[(key, v)] += cnt # coor_dict[(key, v)] += cnt # coor_dict[(key, v)] += cnt sorted_coor_dict = sorted(coor_dict.items(), key=lambda s:s[1], reverse = True) return sorted_coor_dict def filter_result_pred(self, src1): coor_dict = defaultdict(int) ## src1是 debug_opt for file in [open(src1)]: for line in file: # print("after key:", key) result = self.filter_str(line) if result: key, v = result[0] # v = "?x#ns:people.person.nationality#ns:m.0f8l9c" a1, r, a2 = v.split('#') if re.match(r'\?x[0-9]*|M[0-9]*', a1) and re.match(r'\?x[0-9]*|M[0-9]*', a2) and r!='is': # print("1111111") v = r[3:] if r.startswith('ns:') else r else: # print("222222222") r = r[3:] if r.startswith('ns:') else r a2 = a2[3:] if a2.startswith('ns:') else a2 a2 = 'm_'+a2[2:] if a2.startswith('m.') else a2 v = f"{r} {a2}" # print(key, v) coor_dict[(key, v)] += result[1] # elif len(key_entities) == len(v_entities): # key = re.sub(r'M[0-9]', 'M', key) # v = re.sub(r'M[0-9]', 'M', v) # coor_dict[(key, v)] += cnt # coor_dict[(key, v)] += cnt # coor_dict[(key, v)] += cnt sorted_coor_dict = sorted(coor_dict.items(), key=lambda s:s[1], reverse = True) return sorted_coor_dict def term_extract(self, query): terms = [] entities = [] query = query.split() idx = 0 ####三元组 while idx < len(query): if re.match(r'M[0-9]', query[idx]): entities.append(( query[idx:idx+1],query[idx:idx+1] ,(idx, idx))) idx += 1 # elif idx +3 <= len(query) and ' '.join(query[idx:idx+3]) in self.dict: # terms.append((' '.join(query[idx:idx+3]), self.dict.get(' '.join(query[idx:idx+3])),(idx, idx+2))) # idx += 3 elif idx +1 <= len(query) and ' '.join(query[idx:idx+1]) in self.dict: terms.append((' '.join(query[idx:idx+1]), self.dict.get(' '.join(query[idx:idx+1])), (idx, idx))) idx += 1 else: idx +=1 ## 二元组 idx = 0 while idx < len(query) - 3: if idx +3 <= len(query) and ' '.join(query[idx:idx+3]) in self.dict: terms.append((' '.join(query[idx:idx+3]), self.dict.get(' '.join(query[idx:idx+3])),(idx, idx+2))) idx += 1 idx = 0 while idx < len(query) - 2: if idx +2 <= len(query) and' '.join(query[idx:idx+2]) in self.dict: terms.append(( ' '.join(query[idx:idx+2]), self.dict.get(' '.join(query[idx:idx+2])), (idx, idx+1))) idx += 1 terms = sorted(terms, key = lambda s:s[2][0]) return entities, terms pass def term_extract_v2(self, query, type): ## 0520 ##改了新版的兼容 识别Did M terms = [] entities = [] if query.startswith("Did M") or query.startswith("Was M") or query.startswith("Were M") or query.startswith("Was a"): if type in ['mcd2', 'mcd3']: nl_pattern = query.split()[0] +" " + query.split()[1] terms.append((nl_pattern, [f'?x0#is#{query.split()[1]}'], (0, 1))) else: nl_pattern = query.split()[0] +" M" terms.append((nl_pattern, ['?x0#is#M'], (0, 1))) # print("terms:", terms) query = query.split() idx = 0 ####三元组 while idx < len(query): if re.match(r'M[0-9]', query[idx]): entities.append(( query[idx:idx+1],query[idx:idx+1] ,(idx, idx))) idx += 1 # elif idx +3 <= len(query) and ' '.join(query[idx:idx+3]) in self.dict: # terms.append((' '.join(query[idx:idx+3]), self.dict.get(' '.join(query[idx:idx+3])),(idx, idx+2))) # idx += 3 elif idx +1 <= len(query) and ' '.join(query[idx:idx+1]) in self.dict: terms.append((' '.join(query[idx:idx+1]), self.dict.get(' '.join(query[idx:idx+1])), (idx, idx))) idx += 1 else: idx +=1 ## 二元组 idx = 0 while idx < len(query) - 3: if idx +3 <= len(query) and ' '.join(query[idx:idx+3]) in self.dict: terms.append((' '.join(query[idx:idx+3]), self.dict.get(' '.join(query[idx:idx+3])),(idx, idx+2))) idx += 1 idx = 0 while idx < len(query) - 2: if idx +2 <= len(query) and' '.join(query[idx:idx+2]) in self.dict: terms.append(( ' '.join(query[idx:idx+2]), self.dict.get(' '.join(query[idx:idx+2])), (idx, idx+1))) idx += 1 terms = sorted(terms, key = lambda s:s[2][0]) # print(query, entities, terms) return entities, terms pass def fill_skeleton(self, query, skeleton): ## fill skeleton 是之前的细粒度版本 ## 就是?x a M, ?x nationality 以及 gender的都做区分 ## v2的版本把他们都做成?x P M def preprocess_sparql(query): tokens = [] for token in query: # Replace 'ns:' prefixes. if token.startswith('ns:'): token = token[3:] # Replace mid prefixes. if token.startswith('m.'): token = 'm_' + token[2:] tokens.append(token) return ' '.join(tokens) def transform_term_to_pattern(term): # print("term here:", term) term_split = [] for i in term.split(): term_split += i.split("|||") # print("term split:", term_split) skeleton_list = [] term_list = [] for i in term_split: if i.startswith("FILTER"): continue # print(i) i = preprocess_sparql(i.split("#")) a1, r, a2 = i.split() if a1.startswith("?x") and a2.startswith("?x"): skeleton_list.append(f"{a1} P {a2}") elif a1.startswith("?x") and a2.startswith("M"): skeleton_list.append(f"{a1} P M") elif a2.startswith("?x") and a1.startswith("M"): skeleton_list.append(f"M P {a2}") elif a1.startswith("M") and a2.startswith("M"): skeleton_list.append(f"M P M") elif r == "a": skeleton_list.append(f"{a1} a M") else: skeleton_list.append(f"{a1} V S") term_list.append(i) return skeleton_list, ' . '.join(term_list) entities, terms = self.term_extract(query) # print(f"\nquery:{query}\nskeleton:{skeleton}\nsparql:{sparql}") candidate_terms = defaultdict(set) for term in terms: for sub_term in term[1]: sub_pattern , sub_term = transform_term_to_pattern(sub_term) # print(sub_pattern) if " ".join(sub_pattern) in skeleton: candidate_terms[" ".join(sub_pattern)].add(sub_term) candidate_triplets = defaultdict(list) # print("candidate_term:", candidate_terms) for candidate_skeleton, candidate_terms in candidate_terms.items(): # a1, r, a2 = candidate_term.split("#") for candidate_term in candidate_terms: candidate_term = candidate_term.replace("#", " ") if candidate_term.count("M") == 1: candidate_triplets[candidate_skeleton] += [''.join(candidate_term.replace("M", entity[0][0])) for entity in entities] elif candidate_term.count("M") == 2: candidate_term = list(candidate_term) index_m = candidate_term.index('M') candidate_term[index_m] = 'W' index_m = candidate_term.index('M') candidate_term[index_m] = 'Y' candidate_term = ''.join(candidate_term) for i in permutations(entities, 2): a1, a2 = i[0][0][0], i[1][0][0] # print(a1, a2, candidate_term) candidate_term_ = candidate_term.replace("W", a1) candidate_term_ = candidate_term_.replace("Y", a2) candidate_triplets[candidate_skeleton].append(candidate_term_) else: candidate_triplets[candidate_skeleton].append(candidate_term) # print(entities) # for i in terms: # print(i) # print(terms) # print("candidate_terms:", candidate_terms) # print("candidate_triplets:", candidate_triplets) return candidate_triplets # print(terms) # for term in terms: # for candidate_term in term: # if candiidate_term # for pattern in skeleton: # pass def fill_skeleton_v2(self, query, skeleton): ## 通过query + 对齐的双语词典align的结果得到候选的candidate triples ## 候选的cndidate_triples通过给定的skeleton来过滤 ## 就是?x a M, ?x nationality 以及 gender的都做区分 ## v2的版本把他们都做成?x P M def preprocess_sparql(query): tokens = [] for token in query: # Replace 'ns:' prefixes. if token.startswith('ns:'): token = token[3:] # Replace mid prefixes. if token.startswith('m.'): token = 'm_' + token[2:] tokens.append(token) return ' '.join(tokens) def transform_term_to_pattern(term): term_split = [] for i in term.split(): term_split += i.split("|||") # print("term split:", term_split) skeleton_list = [] term_list = [] for i in term_split: if i.startswith("FILTER"): continue # print(i) i = preprocess_sparql(i.split("#")) a1, r, a2 = i.split() if a1.startswith("?x") and a2.startswith("?x"): ## ?x P ?x skeleton_list.append(f"{a1} P {a2}") elif a1.startswith("?x") and a2.startswith("M"): ## ?x P M skeleton_list.append(f"{a1} P M") elif a2.startswith("?x") and a1.startswith("M"): ## M P ?x skeleton_list.append(f"M P {a2}") elif a1.startswith("M") and a2.startswith("M"): ## M P M skeleton_list.append(f"M P M") elif a1.startswith("?x") and r == "a": ## ?x a M => ?x P M skeleton_list.append(f"{a1} P M") else: ## ?x nationality/gender => ?x P M skeleton_list.append(f"{a1} P M") term_list.append(i) return skeleton_list, ' . '.join(term_list) entities, terms, Mflag = self.term_extract(query) # print(f"\nquery:{query}\nskeleton:{skeleton}\nsparql:{sparql}") candidate_terms = defaultdict(set) for term in terms: for sub_term in term[1]: # print("sub_term:", sub_term) sub_pattern , sub_term = transform_term_to_pattern(sub_term) # print("sub_pattern:", sub_pattern) if " ".join(sub_pattern) in skeleton: candidate_terms[" ".join(sub_pattern)].add(sub_term) candidate_triplets = defaultdict(list) # print("candidate_term:", candidate_terms) for candidate_skeleton, candidate_terms in candidate_terms.items(): # a1, r, a2 = candidate_term.split("#") for candidate_term in candidate_terms: candidate_term = candidate_term.replace("#", " ") if candidate_term.count("M") == 1: candidate_triplets[candidate_skeleton] += [''.join(candidate_term.replace("M", entity[0][0])) for entity in entities] elif candidate_term.count("M") == 2: candidate_term = list(candidate_term) index_m = candidate_term.index('M') candidate_term[index_m] = 'W' index_m = candidate_term.index('M') candidate_term[index_m] = 'Y' candidate_term = ''.join(candidate_term) for i in permutations(entities, 2): a1, a2 = i[0][0][0], i[1][0][0] # print(a1, a2, candidate_term) candidate_term_ = candidate_term.replace("W", a1) candidate_term_ = candidate_term_.replace("Y", a2) candidate_triplets[candidate_skeleton].append(candidate_term_) else: candidate_triplets[candidate_skeleton].append(candidate_term) # print(entities) # for i in terms: # print(i) # print(terms) # print("candidate_terms:", candidate_terms) # print("candidate_triplets:", candidate_triplets) return candidate_triplets def fill_skeleton_v3(self, query, skeleton, split): ## 通过query + 对齐的双语词典align的结果得到候选的candidate triples ## 候选的cndidate_triples通过给定的skeleton来过滤 ## 就是?x a M, ?x nationality 以及 gender的都做区分 ## v3的版本是把原始的M P ?x 换成了?x版本 无M开头的sparql # Mflag = False # for triple in skeleton.split(" . "): # if triple.strip().startswith("M"): # Mflag def preprocess_sparql(query): tokens = [] for token in query: # Replace 'ns:' prefixes. if token.startswith('ns:'): token = token[3:] # Replace mid prefixes. if token.startswith('m.'): token = 'm_' + token[2:] tokens.append(token) return ' '.join(tokens) def check_valid(skeleton_list, skeleton_pattern): skeleton_pattern = re.sub(r'\?x[0-9]', "?x", skeleton_pattern) # skeleton = re.sub(r'\?x[0-9]', "?x", skeleton) for skeleton in skeleton_list: if re.sub(r'\?x[0-9]', "?x", skeleton) not in skeleton_pattern: return False return True def transform_term_to_pattern(term): term_split = [] for i in term.split(): term_split += i.split("|||") skeleton_list = [] term_list = [] for i in term_split: if i.startswith("FILTER"): continue i = preprocess_sparql(i.split("#")) a1, r, a2 = i.split() if a1.startswith("?x") and a2.startswith("?x"): ## ?x P ?x skeleton_list.append(f"{a1} P {a2}") elif a1.startswith("?x") and a2.startswith("M"): ## ?x P M skeleton_list.append(f"{a1} P M") elif a1.startswith("?x") and r == "a": ## ?x a M => ?x P M skeleton_list.append(f"{a1} a M") else: skeleton_list.append(f"{a1} V S") term_list.append(i) skeleton_str = [] return skeleton_list, ' . '.join(term_list) entities, terms = self.term_extract_v2(query, split) # print("terms:", terms) # print(f"\nquery:{query}\nskeleton:{skeleton}\nsparql:{sparql}") candidate_terms = defaultdict(set) for term in terms: for sub_term in term[1]: sub_pattern , sub_term = transform_term_to_pattern(sub_term) if check_valid(sub_pattern, skeleton): candidate_terms[" ".join(sub_pattern)].add(sub_term) candidate_triplets = defaultdict(list) # print("candidate_term:", candidate_terms) for candidate_skeleton, candidate_terms in candidate_terms.items(): # a1, r, a2 = candidate_term.split("#") for candidate_term in candidate_terms: candidate_term = candidate_term.replace("#", " ") if candidate_term.count("M") == 1: if candidate_term.startswith("?x0 is M") and split in ['mcd2', 'mcd3']: candidate_triplets[candidate_skeleton] += [candidate_term] else: candidate_triplets[candidate_skeleton] += [''.join(candidate_term.replace("M", entity[0][0])) for entity in entities] elif candidate_term.count("M") == 2: candidate_term = list(candidate_term) index_m = candidate_term.index('M') candidate_term[index_m] = 'W' index_m = candidate_term.index('M') candidate_term[index_m] = 'Y' candidate_term = ''.join(candidate_term) for i in permutations(entities, 2): a1, a2 = i[0][0][0], i[1][0][0] # print(a1, a2, candidate_term) candidate_term_ = candidate_term.replace("W", a1) candidate_term_ = candidate_term_.replace("Y", a2) candidate_triplets[candidate_skeleton].append(candidate_term_) else: candidate_triplets[candidate_skeleton].append(candidate_term) # print(entities) # for i in terms: # print(i) # print(terms) # print("candidate_terms:", candidate_terms) # print("candidate_triplets:", candidate_triplets) return candidate_triplets def modify_skeleton(self,sparql): sparql = sparql.replace("SELECT count(*) WHERE { ", " ") sparql = sparql.replace("SELECT DISTINCT ?x0 WHERE { ", " ") sparql_list = sparql.strip().split(" . ") skeleton_list = [] # print(sparql) for item in sparql_list: if item.startswith("FILTER"): continue a1, r, a2 = item.strip().split() # print(a1, r, a2) if a1.startswith("?x") and a2.startswith("?x"): skeleton_list.append(f"{a1} P {a2}") elif a1.startswith("?x") and a2.startswith("M"): skeleton_list.append(f"{a1} P M") elif a2.startswith("?x") and a1.startswith("M"): skeleton_list.append(f"M P {a2}") elif a1.startswith("M") and a2.startswith("M"): skeleton_list.append(f"M P M") elif a1.startswith("?x") and r == "a": skeleton_list.append(f"{a1} a M") elif re.match(r'M[0-9]', a1): skeleton_list.append(f"M V S") else: skeleton_list.append(f"{a1} V S") skeleton_set = list(set(skeleton_list)) skeleton_set.sort(key=skeleton_list.index) return sparql, " . ".join(skeleton_set) def clear_sparql(self, sparql): return re.findall(r'[{](.*?)[}]', sparql.replace('\n', ' '))[0] def clear_skeleton(self, skeleton): # print(skeleton) # print(re.findall(r'[{](.*?)[}]', skeleton.strip())[0]) skeleton = [i for i in re.findall(r'[{](.*?)[}]', skeleton)[0].strip().split(" . ") if not i.startswith("FILTER")] for idx, item in enumerate(skeleton): a1, r, a2 = item.split() if re.match(r'M[0-9]', a1): a1 = 'M' if re.match(r'M[0-9]', a2): a2 = 'M' if re.match(r'P[0-9]', r): r = 'P' skeleton[idx] = " ".join([a1, r, a2]) return " . ".join(skeleton) def split_skeleton(self, skeleton, flag): # print("skeleton:", skeleton) if isinstance(skeleton, list): skeleton = " . ".join(skeleton) sparql_groups_part= self.split_sub_skeleton(skeleton, flag, 0) sparql_groups = [] for i in sparql_groups_part: xidx = 0 isplit = i.split(" . ") for kkidx, kk in enumerate(isplit): if kk.startswith('?x'): xidx =kkidx ##记录当前分割开的sparql是用那个中间变量！ variable_idx = int(kk[2]) break if xidx > 0: sub_sparql_groups_p1 = self.split_sub_skeleton(' . '.join(isplit[xidx:]), False, variable_idx) sub_sparql_groups_p2 = self.split_sub_skeleton(' . '.join(isplit[:xidx]), False, variable_idx) for i in sub_sparql_groups_p2: for j in sub_sparql_groups_p1: sparql_groups.append(i+' . ' + j) else: sparql_groups.append(i) return sparql_groups def split_sub_skeleton(self, sparql, Mflag, count): # split_lf_results = [] # split_lf_combine_results = [] def trans_tuple_str(tuple_list): t_all = tuple_list[0] for i in range(1, len(tuple_list)): if isinstance(tuple_list[i], tuple): t_all += tuple_list[i] else: return False return ' '.join(t_all) results, triples, FILTER_triples = [], [], [] for clf in sparql.split(" . "): if clf.startswith("FILTER"): continue elif len(clf.split()) != 3: continue a1, r, a2 = clf.split() var_cnt = self.count_var(clf) triples.append((a1, r, a2, var_cnt)) split_dict = defaultdict(list) sorted_triples = sorted(triples, key=lambda k: k[-1]) # print("sorted triples:", sparql, "\n", sorted_triples) ##划分方法 for triple in sorted_triples: if isinstance(triple, tuple) and len(triple) == 4: arg1, rel, arg2, _ = triple triple = (arg1, rel, arg2) ## 对于两个变量的三元组 ## 把他们尽可能的插入之前已有的三元组中 ## [?x0 ?x1] [?x1, ?x2] [?x2, ?x3] if arg1.startswith('?x') and arg2.startswith('?x'): ##对于链式 的特定修正！！！ ## 每次需要更新他们匹配的组 arg_max = arg1 if arg1 > arg2 else arg2 arg_min = arg2 if arg1 > arg2 else arg1 if len(split_dict[arg_max]) > 0: for cur_list in split_dict[arg_max]: cur_list_ = cur_list[:] cur_list_.insert(0, triple) if Mflag and len(split_dict[arg_min]) > 0: for j in split_dict[arg_min]: j+=cur_list_ else: split_dict[arg_min].append(cur_list_) else: split_dict[arg_max].append([triple]) ## 如果只有一个变量 ## 看能不能为之前添加的做补充 ## 形如(?x, r, M)为之前（M, r, ?x)的做补充 elif (not Mflag and arg1.startswith('?x') and not arg1.startswith("?x0")) \ or (Mflag and arg1.startswith('?x')): flag = True for t in split_dict[arg1]: if t[0][0] != arg1: t.append(triple) flag = False if flag: split_dict[arg1].append([triple]) # print("h:",split_dict) ##都没有 为该变量的第一个三元组关系 elif (not Mflag and arg1.startswith("M") and arg2.startswith("?x") and not arg2.startswith("?x0")) \ or (Mflag and arg1.startswith("M") and arg2.startswith("?x")): flag =True for t in split_dict[arg2]: t.append(triple) flag = False if flag: split_dict[arg2].append([triple]) else: variable = arg2 if arg2.startswith("?x") else arg1 split_dict[variable].append([triple]) else: split_dict[triple] = [triple] final_split = [] for v in split_dict.values(): for vv in v: vv_len = len(vv) xidx, xflag = 0, False for idx in range(vv_len): # if SPOUSE_PRED in vv[idx] or SIBLING_PRED in vv[idx]: # # print("vvidx:", vv[idx]) # a1, r, a2 = vv[idx] # vv.append(f"FILTER ( {a1} != {a2} )") vv[idx] = ' '.join(vv[idx]) if not xflag and vv[idx].startswith("?x"): xidx, xflag = idx, True vv = ' . '.join(vv[:xidx] + sorted(list(set(vv[xidx:vv_len]))) + vv[vv_len:]) # print(vv,vv[2], count, vv.startswith('?x'), vv[2]!=count) if not (vv.startswith('?x') and int(vv[2])!=count): ##去掉不合法的?x final_split.append(vv) return final_split def distribute_triples_to_skeleton(self, skeleton_groups, candidate_triplets): print("skeleton_groups:", skeleton_groups) print("candidate_triplets:", candidate_triplets) fn = lambda x, code=',': reduce(lambda x, y: [str(i)+code+str(j) for i in x for j in y], x) ans = [] def replace_variable(pattern, candidates): # print("pattern:", pattern) # print("candidates:", candidates) a1, _, a2 = pattern.split() modify_candidates = [] for idx, candidate in enumerate(candidates): a1_c, r_c, a2_c = candidate.split() a1_c = a1 if a1_c == "?x" else a1_c a2_c = a2 if a2_c == "?x" else a2_c modify_candidates.append(' '.join([a1_c, r_c, a2_c])) # print("modify:", modify_candidates) return modify_candidates for skeleton_group in skeleton_groups: skeleton_group = skeleton_group.split(" . ") if len(skeleton_group) == 1: if skeleton_group[0] in candidate_triplets: ans += candidate_triplets.get(skeleton_group[0]) temp_candidates = candidate_triplets.get(re.sub(r'\?x[0-9]', '?x', skeleton_group[0]), []) ans += replace_variable(skeleton_group[0], temp_candidates) else: # print("11",re.sub(r'\?x[0-9]', '?x', skeleton_group[1])) # print("22",candidate_triplets.get(re.sub(r'\?x[0-9]', '?x', skeleton_group[1]))) triples_groups = [replace_variable(skeleton_item, candidate_triplets.get(re.sub(r'\?x[0-9]', '?x', skeleton_item), [])) for skeleton_item in skeleton_group] # print(triples_groups) ans += fn(triples_groups, ' . ') return ans def generate_samples(self, query, sparql, triples, type): # triples = triples.split(" . ") pos_ans = [] neg_ans = [] valid_cnt = len([i for i in sparql.split(" . ") if not i.startswith("FILTER")]) # print(f"query:{query}\nsparql:{sparql}") coverage_sparql = set() for triple_group in triples: flag = True for triple in triple_group.split(" . "): # print(triple) if triple not in sparql.split(" . "): flag = False continue else: coverage_sparql.add(triple) # print(flag, triple_group) if flag: pos_ans.append(Sample(query, triple_group, flag).__dict__) # print(triple_group) else: neg_ans.append(Sample(query, triple_group, flag).__dict__) # coverage_sparql = " . ".join(coverage_sparql).split(" . ") coverage = True if len(coverage_sparql) == valid_cnt else False # if not coverage: # print(coverage_sparql) # print(sparql) # print(f"query:{query}\nsparql:{sparql}") # print(pos_ans, triples) if type == "train": # print(len(pos_ans), len(neg_ans)) return coverage, pos_ans + random.sample(neg_ans, min(len(neg_ans), len(pos_ans))) else: return coverage, pos_ans + neg_ans def generate_samples_v2(self, query, trie_sparql, triples, type): ##目的是为了统计primitive prediction中 的P/R/F1 pos_ans = [] neg_ans = [] # valid_cnt = len([i for i in sparql.split(" . ") if not i.startswith("FILTER")]) valid_paths = [path.strip() for path in trie_sparql.split("###")] # print(f"query:{query}\nsparql:{sparql}") # print("triples:", triples) coverage_sparql = set() for triple_group in triples: if triple_group in valid_paths: pos_ans.append(Sample(query, triple_group, True).__dict__) # print(triple_group) else: # print("triple_group:", triple_group) # print("valid_paths:", valid_paths) neg_ans.append(Sample(query, triple_group, False).__dict__) p = float(len(set(triples) & set(valid_paths))) / float(len(triples)) r = float(len(set(triples) & set(valid_paths))) / float(len(valid_paths)) # f = 2 * p*r / (p+r) # print(p, r) if type == "train": # print(len(pos_ans), len(neg_ans)) return (p, r), pos_ans + random.sample(neg_ans, min(len(neg_ans), len(pos_ans))) else: return (p, r), len(pos_ans), len(neg_ans), pos_ans + neg_ans if __name__ == '__main__': helper = Helper() #####一些GIZA++的指令 #参加博客：http://codepothunter.github.io/2016/07/11/How-to-use-GIZA-for-alignment/ ############################################生成.update文件############################################ # helper.update_output_format("v4-0520") ############################################生成统计共现文件############################################# # qf_file统计结果在debug_opt中 # fq_file的统计结果在debug_opt2中 # qf_file = f"/home/v-yinguo/DKI/GIZA_alignment/v4-0520/query2formula/query2formula.A3.final.update" # fq_file = f"/home/v-yinguo/DKI/GIZA_alignment/v4-0520/formula2query/formula2query.A3.final.update" # res_file = open(f"/home/v-yinguo/DKI/GIZA_alignment/coor_file-0520","w") # sorted_res = helper.statistic_coor(qf_file, fq_file) # for k in sorted_res: # # print("here") # res_file.write(str(k) + '\n') # print(k) ############################################只过滤共献词表3################################################ # helper = Helper() # src = "/home/v-yinguo/DKI/GIZA_alignment/coor_file-0520" # sorted_res = helper.filter_result(src) # f = open("./coor_file-0520.filter_wto_no", "w") # for k in sorted_res: # # print("k") # if k[-1] > 200: # f.write(str(k) + '\n') ###########################################只保留词和predicate的对应################################################ ##生成phrase_table.pred文件 # helper = Helper() # src = "/home/v-yinguo/DKI/GIZA_alignment/coor_file-0520" # sorted_res = helper.filter_result_pred(src) # f = open("./coor_file-0520.filter.pred", "w") # for k in sorted_res: # # print("k") # if k[-1] > 30: # f.write(str(k) + '\n') ############################################过滤共现词表################################################ # src1, src2 = "/home/v-yinguo/DKI/GIZA_alignment/debug_opt", "/home/v-yinguo/DKI/GIZA_alignment/debug_opt" # sorted_res = helper.filter_result(src1, src2) # f = open("./debug_opt4", "w") # for k in sorted_res: # # print("here") # if k[-1] > 50: # f.write(str(k) + '\n') # f.close()

ContextualSP/poset_decoding/data/generate_phrase_table.py/0

coverage: status: project: default: # basic target: auto threshold: 3% base: auto # advanced branches: null if_no_uploads: error if_not_found: success if_ci_failed: error only_pulls: false flags: null paths: null patch: default: threshold: 1%

ContextualSP/poset_decoding/traversal_path_prediction/MatchZoo-py/.codecov.yml/0

# -*- coding: utf-8 -*- # # Configuration file for the Sphinx documentation builder. # # This file does only contain a selection of the most common options. For a # full list see the documentation: # http://www.sphinx-doc.org/en/master/config # -- Path setup -------------------------------------------------------------- # If extensions (or modules to document with autodoc) are in another directory, # add these directories to sys.path here. If the directory is relative to the # documentation root, use os.path.abspath to make it absolute, like shown here. # import os import sys sys.path.insert(0, os.path.abspath('../..')) sys.path.insert(0, os.path.abspath('../../matchzoo')) sys.path.insert(0, os.path.abspath('../../matchzoo/auto')) sys.path.insert(0, os.path.abspath('../../matchzoo/data_pack')) sys.path.insert(0, os.path.abspath('../../matchzoo/dataloader')) sys.path.insert(0, os.path.abspath('../../matchzoo/datasets')) sys.path.insert(0, os.path.abspath('../../matchzoo/engine')) sys.path.insert(0, os.path.abspath('../../matchzoo/embedding')) sys.path.insert(0, os.path.abspath('../../matchzoo/losses')) sys.path.insert(0, os.path.abspath('../../matchzoo/models')) sys.path.insert(0, os.path.abspath('../../matchzoo/modules')) sys.path.insert(0, os.path.abspath('../../matchzoo/metrics')) sys.path.insert(0, os.path.abspath('../../matchzoo/preprocessors')) sys.path.insert(0, os.path.abspath('../../matchzoo/utils')) sys.path.insert(0, os.path.abspath('../../matchzoo/tasks')) sys.path.insert(0, os.path.abspath('../../matchzoo/trainers')) # -- Project information ----------------------------------------------------- project = 'MatchZoo-py' author = 'MatchZoo' # The short X.Y version version = '' # The full version, including alpha/beta/rc tags release = '1.0' # -- General configuration --------------------------------------------------- # If your documentation needs a minimal Sphinx version, state it here. # # needs_sphinx = '1.0' # Add any Sphinx extension module names here, as strings. They can be # extensions coming with Sphinx (named 'sphinx.ext.*') or your custom # ones. extensions = [ 'autoapi.extension', 'sphinx.ext.autodoc', 'sphinx.ext.mathjax', 'sphinx.ext.napoleon', 'sphinx_autodoc_typehints', ] autoapi_dirs = ['../../matchzoo'] # Add any paths that contain templates here, relative to this directory. templates_path = ['_templates'] # The suffix(es) of source filenames. # You can specify multiple suffix as a list of string: # # source_suffix = ['.rst', '.md'] source_suffix = '.rst' # from recommonmark.parser import CommonMarkParser # source_parsers = { # '.md':CommonMarkParser # } # The master toctree document. master_doc = 'index' # The language for content autogenerated by Sphinx. Refer to documentation # for a list of supported languages. # # This is also used if you do content translation via gettext catalogs. # Usually you set "language" from the command line for these cases. language = None # List of patterns, relative to source directory, that match files and # directories to ignore when looking for source files. # This pattern also affects html_static_path and html_extra_path . exclude_patterns = ['_build', 'Thumbs.db', '.DS_Store'] # The name of the Pygments (syntax highlighting) style to use. pygments_style = 'sphinx' # -- Options for HTML output ------------------------------------------------- # The theme to use for HTML and HTML Help pages. See the documentation for # a list of builtin themes. # html_theme = 'sphinx_rtd_theme' # Theme options are theme-specific and customize the look and feel of a theme # further. For a list of options available for each theme, see the # documentation. # # html_theme_options = {} # Add any paths that contain custom static files (such as style sheets) here, # relative to this directory. They are copied after the builtin static files, # so a file named "default.css" will overwrite the builtin "default.css". html_static_path = ['_static'] # Custom sidebar templates, must be a dictionary that maps document names # to template names. # # The default sidebars (for documents that don't match any pattern) are # defined by theme itself. Builtin themes are using these templates by # default: ``['localtoc.html', 'relations.html', 'sourcelink.html', # 'searchbox.html']``. # # html_sidebars = {} # -- Options for HTMLHelp output --------------------------------------------- # Output file base name for HTML help builder. htmlhelp_basename = 'MatchZoodoc' # -- Options for LaTeX output ------------------------------------------------ latex_elements = { # The paper size ('letterpaper' or 'a4paper'). # # 'papersize': 'letterpaper', # The font size ('10pt', '11pt' or '12pt'). # # 'pointsize': '10pt', # Additional stuff for the LaTeX preamble. # # 'preamble': '', # Latex figure (float) alignment # # 'figure_align': 'htbp', } # Grouping the document tree into LaTeX files. List of tuples # (source start file, target name, title, # author, documentclass [howto, manual, or own class]). latex_documents = [ (master_doc, 'MatchZoo.tex', 'MatchZoo Documentation', 'MatchZoo', 'manual'), ] # -- Options for manual page output ------------------------------------------ # One entry per manual page. List of tuples # (source start file, name, description, authors, manual section). man_pages = [ (master_doc, 'matchzoo', 'MatchZoo Documentation', [author], 1) ] # -- Options for Texinfo output ---------------------------------------------- # Grouping the document tree into Texinfo files. List of tuples # (source start file, target name, title, author, # dir menu entry, description, category) texinfo_documents = [ (master_doc, 'MatchZoo', 'MatchZoo Documentation', author, 'MatchZoo', 'One line description of project.', 'Miscellaneous'), ] # -- Extension configuration -------------------------------------------------

ContextualSP/poset_decoding/traversal_path_prediction/MatchZoo-py/docs/source/conf.py/0

from . import callbacks from .dataset import Dataset from .dataloader import DataLoader from .dataloader_builder import DataLoaderBuilder from .dataset_builder import DatasetBuilder

ContextualSP/poset_decoding/traversal_path_prediction/MatchZoo-py/matchzoo/dataloader/__init__.py/0

from .embedding import Embedding from .embedding import load_from_file

ContextualSP/poset_decoding/traversal_path_prediction/MatchZoo-py/matchzoo/embedding/__init__.py/0

"""Average precision metric for ranking.""" import numpy as np from matchzoo.engine.base_metric import RankingMetric from . import Precision class AveragePrecision(RankingMetric): """Average precision metric.""" ALIAS = ['average_precision', 'ap'] def __init__(self, threshold: float = 0.): """ :class:`AveragePrecision` constructor. :param threshold: The label threshold of relevance degree. """ self._threshold = threshold def __repr__(self) -> str: """:return: Formated string representation of the metric.""" return f"{self.ALIAS[0]}({self._threshold})" def __call__(self, y_true: np.array, y_pred: np.array) -> float: """ Calculate average precision (area under PR curve). Example: >>> y_true = [0, 1] >>> y_pred = [0.1, 0.6] >>> round(AveragePrecision()(y_true, y_pred), 2) 0.75 >>> round(AveragePrecision()([], []), 2) 0.0 :param y_true: The ground true label of each document. :param y_pred: The predicted scores of each document. :return: Average precision. """ precision_metrics = [Precision(k + 1) for k in range(len(y_pred))] out = [metric(y_true, y_pred) for metric in precision_metrics] if not out: return 0. return np.mean(out).item()

ContextualSP/poset_decoding/traversal_path_prediction/MatchZoo-py/matchzoo/metrics/average_precision.py/0

"""A simple densely connected baseline model.""" import typing import torch from matchzoo.engine.base_model import BaseModel from matchzoo.engine.param_table import ParamTable from matchzoo.engine import hyper_spaces class DenseBaseline(BaseModel): """ A simple densely connected baseline model. Examples: >>> model = DenseBaseline() >>> model.params['mlp_num_layers'] = 2 >>> model.params['mlp_num_units'] = 300 >>> model.params['mlp_num_fan_out'] = 128 >>> model.params['mlp_activation_func'] = 'relu' >>> model.guess_and_fill_missing_params(verbose=0) >>> model.build() """ @classmethod def get_default_params(cls) -> ParamTable: """:return: model default parameters.""" params = super().get_default_params( with_embedding=True, with_multi_layer_perceptron=True ) params['mlp_num_units'] = 256 params.get('mlp_num_units').hyper_space = \ hyper_spaces.quniform(16, 512) params.get('mlp_num_layers').hyper_space = \ hyper_spaces.quniform(1, 5) return params def build(self): """Build.""" self.embeddinng = self._make_default_embedding_layer() self.mlp = self._make_multi_layer_perceptron_layer( 2 * self._params['embedding_output_dim'] ) self.out = self._make_output_layer( self._params['mlp_num_fan_out'] ) def forward(self, inputs): """Forward.""" input_left, input_right = inputs['text_left'], inputs['text_right'] input_left = self.embeddinng(input_left.long()).sum(1) input_right = self.embeddinng(input_right.long()).sum(1) x = torch.cat((input_left, input_right), dim=1) return self.out(self.mlp(x))

ContextualSP/poset_decoding/traversal_path_prediction/MatchZoo-py/matchzoo/models/dense_baseline.py/0

"""Attention module.""" import typing import torch import torch.nn as nn import torch.nn.functional as F class Attention(nn.Module): """ Attention module. :param input_size: Size of input. :param mask: An integer to mask the invalid values. Defaults to 0. Examples: >>> import torch >>> attention = Attention(input_size=10) >>> x = torch.randn(4, 5, 10) >>> x.shape torch.Size([4, 5, 10]) >>> x_mask = torch.BoolTensor(4, 5) >>> attention(x, x_mask).shape torch.Size([4, 5]) """ def __init__(self, input_size: int = 100): """Attention constructor.""" super().__init__() self.linear = nn.Linear(input_size, 1, bias=False) def forward(self, x, x_mask): """Perform attention on the input.""" x = self.linear(x).squeeze(dim=-1) x = x.masked_fill(x_mask, -float('inf')) return F.softmax(x, dim=-1) class BidirectionalAttention(nn.Module): """Computing the soft attention between two sequence.""" def __init__(self): """Init.""" super().__init__() def forward(self, v1, v1_mask, v2, v2_mask): """Forward.""" similarity_matrix = v1.bmm(v2.transpose(2, 1).contiguous()) v2_v1_attn = F.softmax( similarity_matrix.masked_fill( v1_mask.unsqueeze(2), -1e-7), dim=1) v1_v2_attn = F.softmax( similarity_matrix.masked_fill( v2_mask.unsqueeze(1), -1e-7), dim=2) attended_v1 = v1_v2_attn.bmm(v2) attended_v2 = v2_v1_attn.transpose(1, 2).bmm(v1) attended_v1.masked_fill_(v1_mask.unsqueeze(2), 0) attended_v2.masked_fill_(v2_mask.unsqueeze(2), 0) return attended_v1, attended_v2 class MatchModule(nn.Module): """ Computing the match representation for Match LSTM. :param hidden_size: Size of hidden vectors. :param dropout_rate: Dropout rate of the projection layer. Defaults to 0. Examples: >>> import torch >>> attention = MatchModule(hidden_size=10) >>> v1 = torch.randn(4, 5, 10) >>> v1.shape torch.Size([4, 5, 10]) >>> v2 = torch.randn(4, 5, 10) >>> v2_mask = torch.ones(4, 5).to(dtype=torch.uint8) >>> attention(v1, v2, v2_mask).shape torch.Size([4, 5, 20]) """ def __init__(self, hidden_size, dropout_rate=0): """Init.""" super().__init__() self.v2_proj = nn.Linear(hidden_size, hidden_size) self.proj = nn.Linear(hidden_size * 4, hidden_size * 2) self.dropout = nn.Dropout(p=dropout_rate) def forward(self, v1, v2, v2_mask): """Computing attention vectors and projection vectors.""" proj_v2 = self.v2_proj(v2) similarity_matrix = v1.bmm(proj_v2.transpose(2, 1).contiguous()) v1_v2_attn = F.softmax( similarity_matrix.masked_fill( v2_mask.unsqueeze(1).bool(), -1e-7), dim=2) v2_wsum = v1_v2_attn.bmm(v2) fusion = torch.cat([v1, v2_wsum, v1 - v2_wsum, v1 * v2_wsum], dim=2) match = self.dropout(F.relu(self.proj(fusion))) return match

ContextualSP/poset_decoding/traversal_path_prediction/MatchZoo-py/matchzoo/modules/attention.py/0

from matchzoo.data_pack import DataPack from .units import Vocabulary from .build_unit_from_data_pack import build_unit_from_data_pack def build_vocab_unit( data_pack: DataPack, mode: str = 'both', verbose: int = 1 ) -> Vocabulary: """ Build a :class:`preprocessor.units.Vocabulary` given `data_pack`. The `data_pack` should be preprocessed forehand, and each item in `text_left` and `text_right` columns of the `data_pack` should be a list of tokens. :param data_pack: The :class:`DataPack` to build vocabulary upon. :param mode: One of 'left', 'right', and 'both', to determine the source data for building the :class:`VocabularyUnit`. :param verbose: Verbosity. :return: A built vocabulary unit. """ return build_unit_from_data_pack( unit=Vocabulary(), data_pack=data_pack, mode=mode, flatten=True, verbose=verbose )

ContextualSP/poset_decoding/traversal_path_prediction/MatchZoo-py/matchzoo/preprocessors/build_vocab_unit.py/0

import typing import numpy as np from .unit import Unit class TruncatedLength(Unit): """ TruncatedLengthUnit Class. Process unit to truncate the text that exceeds the set length. Examples: >>> from matchzoo.preprocessors.units import TruncatedLength >>> truncatedlen = TruncatedLength(3) >>> truncatedlen.transform(list(range(1, 6))) == [3, 4, 5] True >>> truncatedlen.transform(list(range(2))) == [0, 1] True """ def __init__( self, text_length: int, truncate_mode: str = 'pre' ): """ Class initialization. :param text_length: the specified maximum length of text. :param truncate_mode: String, `pre` or `post`: remove values from sequences larger than :attr:`text_length`, either at the beginning or at the end of the sequences. """ self._text_length = text_length self._truncate_mode = truncate_mode def transform(self, input_: list) -> list: """ Truncate the text that exceeds the specified maximum length. :param input_: list of tokenized tokens. :return tokens: list of tokenized tokens in fixed length if its origin length larger than :attr:`text_length`. """ if len(input_) <= self._text_length: truncated_tokens = input_ else: if self._truncate_mode == 'pre': truncated_tokens = input_[-self._text_length:] elif self._truncate_mode == 'post': truncated_tokens = input_[:self._text_length] else: raise ValueError('{} is not a vaild ' 'truncate mode.'.format(self._truncate_mode)) return truncated_tokens

ContextualSP/poset_decoding/traversal_path_prediction/MatchZoo-py/matchzoo/preprocessors/units/truncated_length.py/0

import typing import torch from torch import nn from torch import optim import matchzoo from matchzoo.engine.base_metric import ( BaseMetric, RankingMetric, ClassificationMetric ) activation = nn.ModuleDict([ ['relu', nn.ReLU()], ['hardtanh', nn.Hardtanh()], ['relu6', nn.ReLU6()], ['sigmoid', nn.Sigmoid()], ['tanh', nn.Tanh()], ['softmax', nn.Softmax()], ['softmax2d', nn.Softmax2d()], ['logsoftmax', nn.LogSoftmax()], ['elu', nn.ELU()], ['selu', nn.SELU()], ['celu', nn.CELU()], ['hardshrink', nn.Hardshrink()], ['leakyrelu', nn.LeakyReLU()], ['logsigmoid', nn.LogSigmoid()], ['softplus', nn.Softplus()], ['softshrink', nn.Softshrink()], ['prelu', nn.PReLU()], ['softsign', nn.Softsign()], ['softmin', nn.Softmin()], ['tanhshrink', nn.Tanhshrink()], ['rrelu', nn.RReLU()], ['glu', nn.GLU()], ]) loss = nn.ModuleDict([ ['l1', nn.L1Loss()], ['nll', nn.NLLLoss()], ['kldiv', nn.KLDivLoss()], ['mse', nn.MSELoss()], ['bce', nn.BCELoss()], ['bce_with_logits', nn.BCEWithLogitsLoss()], ['cosine_embedding', nn.CosineEmbeddingLoss()], ['ctc', nn.CTCLoss()], ['hinge_embedding', nn.HingeEmbeddingLoss()], ['margin_ranking', nn.MarginRankingLoss()], ['multi_label_margin', nn.MultiLabelMarginLoss()], ['multi_label_soft_margin', nn.MultiLabelSoftMarginLoss()], ['multi_margin', nn.MultiMarginLoss()], ['smooth_l1', nn.SmoothL1Loss()], ['soft_margin', nn.SoftMarginLoss()], ['cross_entropy', nn.CrossEntropyLoss()], ['triplet_margin', nn.TripletMarginLoss()], ['poisson_nll', nn.PoissonNLLLoss()] ]) optimizer = dict({ 'adadelta': optim.Adadelta, 'adagrad': optim.Adagrad, 'adam': optim.Adam, 'sparse_adam': optim.SparseAdam, 'adamax': optim.Adamax, 'asgd': optim.ASGD, 'lbfgs': optim.LBFGS, 'rmsprop': optim.RMSprop, 'rprop': optim.Rprop, 'sgd': optim.SGD }) def _parse( identifier: typing.Union[str, typing.Type[nn.Module], nn.Module], dictionary: nn.ModuleDict, target: str ) -> nn.Module: """ Parse loss and activation. :param identifier: activation identifier, one of - String: name of a activation - Torch Modele subclass - Torch Module instance (it will be returned unchanged). :param dictionary: nn.ModuleDict instance. Map string identifier to nn.Module instance. :return: A :class:`nn.Module` instance """ if isinstance(identifier, str): if identifier in dictionary: return dictionary[identifier] else: raise ValueError( f'Could not interpret {target} identifier: ' + str(identifier) ) elif isinstance(identifier, nn.Module): return identifier elif issubclass(identifier, nn.Module): return identifier() else: raise ValueError( f'Could not interpret {target} identifier: ' + str(identifier) ) def parse_activation( identifier: typing.Union[str, typing.Type[nn.Module], nn.Module] ) -> nn.Module: """ Retrieves a torch Module instance. :param identifier: activation identifier, one of - String: name of a activation - Torch Modele subclass - Torch Module instance (it will be returned unchanged). :return: A :class:`nn.Module` instance Examples:: >>> from torch import nn >>> from matchzoo.utils import parse_activation Use `str` as activation: >>> activation = parse_activation('relu') >>> type(activation) <class 'torch.nn.modules.activation.ReLU'> Use :class:`torch.nn.Module` subclasses as activation: >>> type(parse_activation(nn.ReLU)) <class 'torch.nn.modules.activation.ReLU'> Use :class:`torch.nn.Module` instances as activation: >>> type(parse_activation(nn.ReLU())) <class 'torch.nn.modules.activation.ReLU'> """ return _parse(identifier, activation, 'activation') def parse_loss( identifier: typing.Union[str, typing.Type[nn.Module], nn.Module], task: typing.Optional[str] = None ) -> nn.Module: """ Retrieves a torch Module instance. :param identifier: loss identifier, one of - String: name of a loss - Torch Module subclass - Torch Module instance (it will be returned unchanged). :param task: Task type for determining specific loss. :return: A :class:`nn.Module` instance Examples:: >>> from torch import nn >>> from matchzoo.utils import parse_loss Use `str` as loss: >>> loss = parse_loss('mse') >>> type(loss) <class 'torch.nn.modules.loss.MSELoss'> Use :class:`torch.nn.Module` subclasses as loss: >>> type(parse_loss(nn.MSELoss)) <class 'torch.nn.modules.loss.MSELoss'> Use :class:`torch.nn.Module` instances as loss: >>> type(parse_loss(nn.MSELoss())) <class 'torch.nn.modules.loss.MSELoss'> """ return _parse(identifier, loss, 'loss') def _parse_metric( metric: typing.Union[str, typing.Type[BaseMetric], BaseMetric], Metrix: typing.Type[BaseMetric] ) -> BaseMetric: """ Parse metric. :param metrc: Input metric in any form. :param Metrix: Base Metric class. Either :class:`matchzoo.engine.base_metric.RankingMetric` or :class:`matchzoo.engine.base_metric.ClassificationMetric`. :return: A :class:`BaseMetric` instance """ if isinstance(metric, str): metric = metric.lower() # ignore case for subclass in Metrix.__subclasses__(): if metric == subclass.ALIAS or metric in subclass.ALIAS: return subclass() elif isinstance(metric, Metrix): return metric elif issubclass(metric, Metrix): return metric() raise ValueError(f'`{metric}` can not be used in current task.') def parse_metric( metric: typing.Union[str, typing.Type[BaseMetric], BaseMetric], task: str ) -> BaseMetric: """ Parse input metric in any form into a :class:`BaseMetric` instance. :param metric: Input metric in any form. :param task: Task type for determining specific metric. :return: A :class:`BaseMetric` instance Examples:: >>> from matchzoo import metrics >>> from matchzoo.utils import parse_metric Use `str` as MatchZoo metrics: >>> mz_metric = parse_metric('map', 'ranking') >>> type(mz_metric) <class 'matchzoo.metrics.mean_average_precision.MeanAveragePrecision'> Use :class:`matchzoo.engine.BaseMetric` subclasses as MatchZoo metrics: >>> type(parse_metric(metrics.AveragePrecision, 'ranking')) <class 'matchzoo.metrics.average_precision.AveragePrecision'> Use :class:`matchzoo.engine.BaseMetric` instances as MatchZoo metrics: >>> type(parse_metric(metrics.AveragePrecision(), 'ranking')) <class 'matchzoo.metrics.average_precision.AveragePrecision'> """ if task is None: raise ValueError( 'Should specify one `BaseTask`.' ) if task == 'ranking': return _parse_metric(metric, RankingMetric) if task == 'classification': return _parse_metric(metric, ClassificationMetric) else: raise ValueError( 'Should be a Ranking or Classification task.' ) def parse_optimizer( identifier: typing.Union[str, typing.Type[optim.Optimizer]], ) -> optim.Optimizer: """ Parse input metric in any form into a :class:`Optimizer` class. :param optimizer: Input optimizer in any form. :return: A :class:`Optimizer` class Examples:: >>> from torch import optim >>> from matchzoo.utils import parse_optimizer Use `str` as optimizer: >>> parse_optimizer('adam') <class 'torch.optim.adam.Adam'> Use :class:`torch.optim.Optimizer` subclasses as optimizer: >>> parse_optimizer(optim.Adam) <class 'torch.optim.adam.Adam'> """ if isinstance(identifier, str): identifier = identifier.lower() # ignore case if identifier in optimizer: return optimizer[identifier] else: raise ValueError( f'Could not interpret optimizer identifier: ' + str(identifier) ) elif issubclass(identifier, optim.Optimizer): return identifier else: raise ValueError( f'Could not interpret optimizer identifier: ' + str(identifier) )

ContextualSP/poset_decoding/traversal_path_prediction/MatchZoo-py/matchzoo/utils/parse.py/0

""" These tests are simplied because the original verion takes too much time to run, making CI fails as it reaches the time limit. """ import torch import pytest from pathlib import Path import shutil import matchzoo as mz @pytest.fixture(scope='module', params=[ mz.tasks.Ranking(losses=mz.losses.RankCrossEntropyLoss(num_neg=2)), mz.tasks.Classification(num_classes=2), ]) def task(request): return request.param @pytest.fixture(scope='module') def train_raw(task): return mz.datasets.toy.load_data('train', task)[:10] @pytest.fixture(scope='module', params=mz.models.list_available()) def model_class(request): return request.param @pytest.fixture(scope='module') def embedding(): return mz.datasets.toy.load_embedding() @pytest.fixture(scope='module') def setup(task, model_class, train_raw, embedding): return mz.auto.prepare( task=task, model_class=model_class, data_pack=train_raw, embedding=embedding ) @pytest.fixture(scope='module') def model(setup): return setup[0] @pytest.fixture(scope='module') def preprocessor(setup): return setup[1] @pytest.fixture(scope='module') def dataset_builder(setup): return setup[2] @pytest.fixture(scope='module') def dataloader_builder(setup): return setup[3] @pytest.fixture(scope='module') def dataloader(train_raw, preprocessor, dataset_builder, dataloader_builder): return dataloader_builder.build( dataset_builder.build(preprocessor.transform(train_raw))) @pytest.fixture(scope='module') def optimizer(model): return torch.optim.Adam(model.parameters()) @pytest.fixture(scope='module') def save_dir(): return Path('.matchzoo_test_save_load_tmpdir') @pytest.mark.slow def test_model_fit_eval_predict(model, optimizer, dataloader, save_dir): trainer = mz.trainers.Trainer( model=model, optimizer=optimizer, trainloader=dataloader, validloader=dataloader, epochs=2, save_dir=save_dir, verbose=0 ) trainer.run() if save_dir.exists(): shutil.rmtree(save_dir)

ContextualSP/poset_decoding/traversal_path_prediction/MatchZoo-py/tests/models/test_models.py/0

<jupyter_start><jupyter_code>%run init.ipynb preprocessor = mz.models.Bert.get_default_preprocessor() train_pack_processed = preprocessor.transform(train_pack_raw) dev_pack_processed = preprocessor.transform(dev_pack_raw) test_pack_processed = preprocessor.transform(test_pack_raw) trainset = mz.dataloader.Dataset( data_pack=train_pack_processed, mode='pair', num_dup=2, num_neg=1 ) testset = mz.dataloader.Dataset( data_pack=test_pack_processed ) padding_callback = mz.models.Bert.get_default_padding_callback() trainloader = mz.dataloader.DataLoader( dataset=trainset, batch_size=20, stage='train', resample=True, sort=False, callback=padding_callback ) testloader = mz.dataloader.DataLoader( dataset=testset, batch_size=20, stage='dev', callback=padding_callback ) model = mz.models.Bert() model.params['task'] = ranking_task model.params['mode'] = 'bert-base-uncased' model.params['dropout_rate'] = 0.2 model.build() print(model) print('Trainable params: ', sum(p.numel() for p in model.parameters() if p.requires_grad)) no_decay = ['bias', 'LayerNorm.weight'] optimizer_grouped_parameters = [ {'params': [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)], 'weight_decay': 5e-5}, {'params': [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)], 'weight_decay': 0.0} ] from pytorch_transformers import AdamW, WarmupLinearSchedule optimizer = AdamW(optimizer_grouped_parameters, lr=5e-5, betas=(0.9, 0.98), eps=1e-8) scheduler = WarmupLinearSchedule(optimizer, warmup_steps=6, t_total=-1) trainer = mz.trainers.Trainer( model=model, optimizer=optimizer, scheduler=scheduler, trainloader=trainloader, validloader=testloader, validate_interval=None, epochs=10 ) trainer.run()<jupyter_output><empty_output>

ContextualSP/poset_decoding/traversal_path_prediction/MatchZoo-py/tutorials/ranking/bert.ipynb/0

<jupyter_start><jupyter_code>import pandas as pd from transformers import AutoTokenizer, AutoModelForSequenceClassification import numpy as np import torch from torch.nn.functional import softmax from copy import deepcopy import enchant from sklearn.metrics.pairwise import cosine_similarity from scipy import sparse from tqdm import trange, tqdm import json import nltk import spacy import string import re import pickle import inflect nlp = spacy.load('en_core_web_sm')<jupyter_output><empty_output><jupyter_text>Global Variables<jupyter_code>bigram_reservered_words = ['against', 'and', 'area', 'average', 'since', 'away', 'section', 'by', 'class', 'club', 'code', 'cup', 'current', 'date', 'data', 'district', 'elected', 'engine', 'episode', 'event', 'final', 'finish', 'first', 'for', 'from', 'game', 'games', 'goals', 'gold', 'grid', 'height', 'high', 'home', 'id', 'in', 'incumbent', 'international', 'laps', 'league', 'list', 'log', 'loss', 'losses', 'lost', 'method', 'age', 'name', 'nation', 'no', 'notes', 'number', 'of', 'one', 'two', 'three', 'four', 'yes', 'no', 'yards', 'five' 'other', 'outcome', 'overall', 'par', 'party', 'per', 'pick', 'played', 'player', 'points', 'pos', 'rank', 'record', 'region', 'release', 'report', 'res', 'result', 'results','round', 'score', 'season', 'second', 'series', 'singles', 'start', 'end', 'state', 'status', 'table', 'team', 'types', 'the', 'first', 'second', 'third', 'time', 'to', 'total', 'type', 'up', 'week', 'weeks', 'year', 'unit', 'version', 'years', 'ends', 'ended', 'min', 'max', 'make', 'statistics', 'stats', 'in', 'on', 'to', 'see', 'feet', 'subject'] preps = ["aboard","about","above","across","after","against","along","amid","among","as","at","before","behind","below","beneath","beside","besides","between","beyond","but","by","concerning","considering","despite","down","during","except","excepting","excluding","following","for","from","in","inside","into","like","minus","near","of","off","on","onto","opposite","outside","over","past","per","plus","regarding","round","save","since","than","through","to","toward","towards","under","underneath","unlike","until","up","upon","versus","via","with","within","without"] bigram_reservered_words = list(set(bigram_reservered_words + preps)) ### Templates to be used for checking # template1 = lambda table_name, col: f"We are told of the {table_name}'s {trim_col_name(table_name, col)}." # template2 = lambda table_name, col: f"We are informed of the {trim_col_name(table_name, col)} of the {table_name}." # template3 = lambda table_name, col: f"We know {trim_col_name(table_name, col)} of the {table_name}." # template4 = lambda table_name, col: f"We collect {table_name}'s {trim_col_name(table_name, col)}." # TEMPLATES = [template1, template2, template3, template4] if STRICT_MODE else [template1, template2] tablename_black_list = ["statistics", "data", "table", "summary", "sketch", "list"] date_marks = ["date", "dates", "year", "years", "month", "months", "day", "days", "daytime", "minute", "minutes", "second", "seconds", "time"] num_marks = ["num", "number", "sum", "amount", "count", "total", "#", "No.", "no.", "scores", "rating", "rank", "height", "weight", "age", "time", "times", "temperature", "year", "years", "month", "months", "day", "days", "minute", "minutes", "second", "seconds", "average", "sum", "grade", "fee", "cost", "value", "rate"] # words explicitly has numeric implications def template1(table_name, col, col_type): """Template 1 placeholder filling for give table name, column name, and column type""" table_name = get_singular_word(table_name) capital_tname = table_name[0].capitalize() + table_name[1:] trimmed_col_name = trim_col_naive(table_name, col) type_prompt = trim_type(col_type) if col_type == "date" and any([dm in trimmed_col_name for dm in date_marks]): type_prompt = "" if col_type == "number" and any([nm in trimmed_col_name for nm in num_marks]): type_prompt = "" return f"{capital_tname} {trimmed_col_name}{trim_type(col_type)}." TEMPLATES = [template1] device = "cuda" nli_tokenizer = AutoTokenizer.from_pretrained("roberta-large-mnli") nli_model = AutoModelForSequenceClassification.from_pretrained("roberta-large-mnli").to(device)<jupyter_output><empty_output><jupyter_text>Preprocessing Functions<jupyter_code>inflect_engine = inflect.engine() def get_singular_word(word): """ Reduce a given word (string) to singular form """ ans = inflect_engine.singular_noun(word) return ans if ans else word def read_dense_table_vectors(path, delim="\t"): """ Read backend dense table vectors """ with open(path, "r") as f: tid2vals = {} for line in f.readlines(): if len(line) == 0: continue units = line.split(delim) table_id, vals = units[0], units[1:] tid2vals[table_id] = np.array(vals).astype(float) return tid2vals def trim_col(tname, col): """ Normalize table name. :tname: table name :col: column name to be trained """ if tname == '' or col == '': return col if tname == ' ' or col == ' ': return col tname_tokens = nlp(tname) col_tokens = nlp(col) if tname_tokens[-1].lemma_ == col_tokens[0].lemma_: return col_tokens[1:] return col # template1 = lambda table_name, col: f"{table_name[0].capitalize() + table_name[1:]} {trim_col(table_name, col)}." def trim_col_naive(tname, col): if tname == '' or col == '': return col if tname.lower() == col.lower(): return 'name' return col.lower() def trim_type(col_type): if col_type in ["text", "bool"]: return "" return " " + "time" if col_type == "date" else "number"<jupyter_output><empty_output><jupyter_text>Load datasets<jupyter_code>with open("./data/tid2tables.pkl","rb") as f: tid2tables = pickle.load(f) idx2tid = {i:tid for i,tid in enumerate(tid2tables.keys())} tid2idx = {tid:i for i,tid in enumerate(tid2tables.keys())} spiders = [] with open("./data/spider/spider-tables.jsonl", "r") as f: for line in f.readlines(): table = json.loads(line) spiders.append(table) wtqs = [] with open("./data/WTQ/wtq-tables.jsonl", "r") as f: # no table name, no domain name for line in f.readlines(): table = json.loads(line) # dict_keys(['file_name', 'table_name', 'column_types', 'column_names', 'column_values']) wtqs.append(table) wsqls_train = [] with open("./data/wikisql/wikisql-tables.jsonl", "r") as f: for line in f.readlines(): table = json.loads(line) # dict_keys(['refer_cols_index', 'domain', 'table_id', 'table_name', 'column_names', 'column_types', 'column_values']) wsqls_train.append(table) with open("./data/wikisql/wikisql_train.tables2question.json", "r") as f: wsql_train_table2qs = json.load(f) with open("./data/spider/spider-table2questions.json", "r") as f: spider_table2qs = json.load(f) with open("./data/WTQ/wtq-table2questions.json", "r") as f: wtq_table2qs = json.load(f) idx2word, word2idx = {}, {} word2vec = {} with open("./data/numberbatch/nb_emb.txt", "r") as f: cnt = -2 for line in tqdm(f.readlines(), desc="Building word2vec...", leave=True): cnt += 1 if cnt == -1: continue units = line.split(" ") word, emb = units[0], np.array(units[1:]).astype(float) word2vec[word] = emb idx2word[cnt] = word word2idx[word] = cnt EMB_DIM=300 with open("./data/syndict_pipeline.json") as f: synonym_dic = json.load(f)<jupyter_output><empty_output><jupyter_text>Dense Retrieval Setup<jupyter_code>import os import torch import torch.nn as nn import numpy as np from transformers import TapasModel, TapasConfig, TapasTokenizer, BertModel, BertTokenizer def build_projection_layer(weight_path: str): with open(weight_path, 'rb') as f: weights = torch.from_numpy(np.load(f)) linear = nn.Linear(weights.size(0), weights.size(1), bias=False) linear.weight.data = weights return linear MAX_LEN = 1024 DUMMY_TABLE = pd.DataFrame({}) basepath = os.path.join("tapas-torch", "tapas_retrieval") table_model_path = os.path.join(basepath, "tapas_nq_hn_retriever_large_table", "checkpoint") table_model = TapasModel.from_pretrained(table_model_path).to(device) tapas_tokenizer = TapasTokenizer.from_pretrained(table_model_path) table_model_config = TapasConfig.from_pretrained(table_model_path) query_model_path = os.path.join(basepath, "tapas_nq_hn_retriever_large_query", "checkpoint") query_model = TapasModel.from_pretrained(query_model_path).to(device) text_projection_layer = build_projection_layer(os.path.join(basepath, "projection_layer", "text_projection.npy")).to(device) table_projection_layer = build_projection_layer(os.path.join(basepath, "projection_layer", "table_projection.npy")).to(device) def form_table(dic_table, col_name_key="column_names", max_row_limit=10, max_cell_val_len=50): # output a dataframe """ Build source table text for dense vector computation. Done via resampling strategy. :dic_table: table as a dictionary :col_name_key: the key name in passed dic_table storing column names (as a list). :max_row_limit: maximum number of rows for cosntructed table :max_cell_val_len: cell values will be truncated to this length. """ col_names = dic_table["column_names"] col_vals = {k : list(set(v)) for k,v in dic_table["column_values"].items()} try: longest_unique = min(max([len(v) for v in col_vals.values()]), max_row_limit) except: if len(dic_table[col_name_key]) == 0: return DUMMY_TABLE else: return pd.DataFrame({k:[] for k in dic_table[col_name_key]}) col2vals = {n : [str(elem)[:max_cell_val_len] for elem in np.random.choice(v, longest_unique, replace=True)] for n,v in col_vals.items()} return pd.DataFrame(col2vals) # takes a while to load 615144 * 2 vectors... wdc_dense_a = read_dense_table_vectors(path="./wdc/wdc_dense_A.txt") wdc_dense_b = read_dense_table_vectors(path="./wdc/wdc_dense_B.txt") idx2tid = {i:k for i,k in enumerate(wdc_dense_a.keys())} tid2idx = {k:i for i,k in idx2tid.items()} bm_mat_A = torch.stack([torch.Tensor(vs) for vs in wdc_dense_a.values()], dim = 0).to(device) bm_mat_B = torch.stack([torch.Tensor(vs) for vs in wdc_dense_b.values()], dim = 0).to(device) bm_mat_A = bm_mat_A / torch.norm(bm_mat_A, dim=-1).unsqueeze(-1) bm_mat_B = bm_mat_B / torch.norm(bm_mat_B, dim=-1).unsqueeze(-1)<jupyter_output><empty_output><jupyter_text>Core NLI algorithms<jupyter_code>def trim_col_name(table_name, col_name): if table_name == col_name: return table_name + " name" return col_name def check_spell(col_name): """ Check whether a column name (multiwords allowed) is valid english word. """ return all([checker.check(w) for w in col_name.split(" ") if w != ""]) def batchify(pair_dict): """ form batch of a pair of ori-rpl. Two directions. """ split_idx = [] batch_ori = [] batch_rpl = [] prev_end_idx = 0 for dic in pair_dict: key_map, pairs, _, _ = dic.values() split_idx.append((prev_end_idx, prev_end_idx + 2 * len(pairs),)) prev_end_idx = prev_end_idx + 2 * len(pairs) # 2 * because of reverse for (ori, rpl) in pairs: batch_ori.append(ori) batch_ori.append(rpl) # reverse batch_rpl.append(rpl) batch_rpl.append(ori) # reverse return split_idx, batch_ori, batch_rpl def aggregate(split_idx, scores, strict=True): contras, neus, ents = [],[],[] for s,e in split_idx: one_rpl_scores = scores[s:e,:] if strict: """ For REPLACE cols Prefer high PRECISION of repalceablility! (If we REPLACE with a UNreplaceable col, we run into trouble) Reject as many LOW confidence candidate as possible. If NLI give HIGH ent-score, then two columns should almost always be mutally replaceable! """ contra, neu, ent = torch.min(one_rpl_scores, dim=0)[0].squeeze() else: """ For ADD cols Prefer high RECALL of repalceablility! (If we ADD a replaceable col, we run into trouble) Accept as many LOW confidence candidate as possible. If NLI still suggests LOW ent-score, then two columns should almost always be mutally UNreplaceable! """ contra, neu, ent = torch.max(one_rpl_scores, dim=0)[0].squeeze() # Prefer high recall contras.append(float(contra.item())) neus.append(float(neu.item())) ents.append(float(ent.item())) return contras, neus, ents def construct_pairs_for_nli_test(tables, table_id_key="table_id", table_name_key="table_name", col_type_key="column_types", col_name_key="column_names", pending_rpls_key="column_names_syn"): """ Given a list of dictionary-represneted tables, and pending replacements cols, construct pairs of ori-rpl :table_id_key: key name in table dict for table name :col_type_key: key name in table dict for types of columns :col_name_key: key name in table dict for names of columns :pending_rpls_key: key name in table dict for pending keys to be replaced """ assert isinstance(tables, list), "Please pass a list of tables." assert table_id_key in tables[0], "Each Table must have an id." assert table_name_key in tables[0], "Table name is required, but the key is missing." assert col_name_key in tables[0], "column name key is required but missing" assert col_type_key in tables[0], "column type key is required but missing" assert pending_rpls_key in tables[0], "Pending replacement columns is required, but the key is missing." constructed_pairs = [] for i in trange(len(tables)): tab = tables[i] tname = tab[table_name_key] if tab[table_name_key] != "s" else tab[table_name_key][:-1] pending_rpls = tab[pending_rpls_key] col2type = {col: tp for col,tp in zip(tab[col_name_key], tab[col_type_key])} for ori_col, rpl_col_list in pending_rpls.items(): for rpl_col in rpl_col_list: if not check_spell(rpl_col): continue rpl_dic = {"key_map": None, "pairs": [], "table_id": tab[table_id_key], "table_name": tab[table_name_key]} for template in TEMPLATES: sent_ori = template(tname, ori_col, col2type[ori_col]) sent_rpl = template(tname, rpl_col, col2type[ori_col]) rpl_dic["key_map"] = (ori_col, rpl_col,) rpl_dic["pairs"] += ((sent_ori, sent_rpl,),) constructed_pairs.append(rpl_dic) return constructed_pairs def nli_test_across_tables(constructed_pairs, batch_size=256): """ The major interface for NLI verification. Given constructed pairs (results from construct_pairs_for_nli_test), use batch computation to speed up the verfication process. """ assert batch_size % 8 == 0, "Batch size must be a multiple of 8." results = [] completed_pairs = 0 total_batches = len(constructed_pairs) // batch_size + 1 pbar = tqdm(total = total_batches) with torch.no_grad(): while completed_pairs < len(constructed_pairs): batch_contras, batch_neus, batch_ents = [],[],[] prev_completed = completed_pairs completed_pairs = min(completed_pairs + batch_size, len(constructed_pairs)) batch = constructed_pairs[prev_completed:completed_pairs] split_idx, batch_ori, batch_rpl = batchify(batch) inputs = nli_tokenizer(batch_ori, batch_rpl, padding="longest", return_tensors="pt").to(device) logits = nli_model(**inputs).logits scores = softmax(logits, dim=1) # [batch, 3] batch_contras, batch_neus, batch_ents = aggregate(split_idx, scores) del inputs; del logits; del scores; torch.cuda.empty_cache() batch_contras, batch_neus, batch_ents = np.array(batch_contras), np.array(batch_neus), np.array(batch_ents) for b, c, n, e in zip(batch, batch_contras, batch_neus, batch_ents): results.append({"key_map": b["key_map"], "scores": (c, n, e,)}) pbar.update(1) pbar.close() return results def trim_name(text): for ch in ['\\','`','*','{','}','[',']','(',')','>', '<', '#','+','\'', '"']: if ch in text: text = text.replace(ch, "") text.replace("-", " ") text.replace(".", " ") return text def extract_emb(list_of_names): """ Extract numberbatch word embeddings for a given list of strings """ assert isinstance(list_of_names, list), "Expected list as input" output_matrix = np.zeros([len(list_of_names), EMB_DIM]) for i,name in enumerate(list_of_names): name = trim_name(name) units = name.split() # notice "_" is covered by our nb_emb! name_emb = np.zeros(EMB_DIM) for word in units: if "_" in word and word2vec.get(word, None) is None: sub_words = word.split("_") local_emb_mat = extract_emb(sub_words) emb = np.mean(local_emb_mat, axis=0) else: emb = word2vec.get(word, np.zeros(EMB_DIM)) name_emb += emb name_emb /= len(units) output_matrix[i,:] = name_emb return output_matrix def reranker(tgt_names, cand_names, topk=10): """ Do reranking (usually among few hundreds of candidates) and return topk per numberbatch word2vec similarity """ if len(cand_names) == 0: return {} tgt_mat = extract_emb(tgt_names) cand_mat = extract_emb(cand_names) sim_mat = tgt_mat @ cand_mat.T topk = min(len(cand_names), topk) top_scores, top_idx = [v.squeeze().numpy() for v in torch.topk(torch.Tensor(sim_mat), topk, dim=-1)] rec_dic = {} for i, tgt in enumerate(tgt_names): if len(top_idx.shape) == 0: top_idx = np.array([top_idx]) if len(top_idx.shape) == 1: top_idx = top_idx[None, :] rec_dic[tgt] = [cand_names[idx] for idx in top_idx[i]] return rec_dic def retriver(query_table, queries=None, retrieve_strategy="query_dense", topk_tables=50, col_name_key="column_names", target_expand_keys=100): """ Tapas based dense retrieval for finding topk most similar tabels from table base. :query_table: The table whose topk similar will be found :quries: the user NL queries attached with the query_table :retrieve_strategy: Choose from ["query_dense", "table_dense"], qd uses NL query as retrieval query vector, and td uses table as retreival query vector. :topk_tables: Return k most similar tables """ ori_cols = set(query_table[col_name_key]) top_tables_tid = [] if retrieve_strategy == "query_dense": top_tables_tid = retrieve_tables_query_dense(queries,k=topk_tables) elif retrieve_strategy == "table_dense": pass elif retrieve_strategy == "tfidf": pass else: raise NotImplementedError top_tables = [tid2tables[tid] for tid in top_tables_tid] expanded_cols = set() for t in top_tables: if len(expanded_cols) >= target_expand_keys: break expanded_cols = expanded_cols.union(t[col_name_key]) expanded_cols = expanded_cols.difference(ori_cols) return list(ori_cols), list(expanded_cols) def retrieve_tables_tfidf(query_table, tfidf_mat, col_name_key="column_names", table_doc_key="doc"): """ TF-IDF based retrieval for finding most similar tables from DB. """ assert table_doc_key in query_table and col_name_key in query_table query_tfidf = vectorizer.transform(query_table) scores = cosine_similarity(query_tfidf, tfidf_mat)[0] top_scores, indices = [t.squeeze().numpy() for t in torch.topk(torch.Tensor(scores), 1000)] return [idx2tid[i] for i in indices] def retrieve_tables_query_dense(queries, k=50): """' Interface for finding most similar table via dense retrieval. call goes from here. """ assert isinstance(queries, list), "input queries must be a list of strings" torch.cuda.empty_cache() with torch.no_grad(): q_inputs = tapas_tokenizer(table=DUMMY_TABLE, queries=queries, padding=True, truncation=True, return_tensors="pt").to(device) qb = query_model(**q_inputs).pooler_output qb = text_projection_layer(qb) qb = qb / torch.norm(qb, dim=-1).unsqueeze(-1) cos = torch.matmul(qb, bm_mat_B.transpose(0, 1)) cos = torch.mean(cos, dim=0) top_score, top_idx = [v.data.cpu().numpy() for v in torch.topk(cos, k=k)] # top_idx = top_idx.data.cpu().numpy() #### ab means table encoded with encoder A, query encoded with encoder B. return [idx2tid[i] for i in top_idx] checker = enchant.Dict("en_US") def _ends_with_id(string): if len(string) < 2: return False return string[-2:].lower() == "id" def _fill_type_info(string, col_type, delim): """ Add type description for a given column """ if col_type == "date" and not any([dm in string for dm in date_marks]): return delim.join([string, "time"]) if col_type == "number" and not any([nm in string for nm in num_marks]): return delim.join([string, "number"]) return string def contains_number(text): """ Judege whether the passed string contain number """ return len(re.findall("[-+]?[.]?[\d]+(?:,\d\d\d)*[\.]?\d*(?:[eE][-+]?\d+)?", text)) > 0 def _get_replacement(tok1, tok2, tok1_is_reserved, tok2_is_reserved): """ Given a bi-gram, replce the word whose IDF is higher with its synonym. """ if tok1_is_reserved and tok2_is_reserved: return (None, None) if tok1_is_reserved and (not check_spell(tok2) or contains_number(tok2)): return (None, None) if tok2_is_reserved and (not check_spell(tok1) or contains_number(tok1)): return (None, None) if tok1_is_reserved: syn_dic = synonym_dic.get(tok2.lower(), None) return (tok2, syn_dic) if syn_dic is not None else (None, None) if tok2_is_reserved: syn_dic = synonym_dic.get(tok1.lower(), None) return (tok1, syn_dic) if syn_dic is not None else (None, None) # both are not reserved, pick one with higher tfidf val def extract_idf(vocab): vocab_idx = vectorizer.vocabulary_.get(vocab, None) idf = 0 if vocab_idx is None else vectorizer.idf_[vocab_idx] return idf first_tgt = tok1 if extract_idf(tok1) <= extract_idf(tok2) else tok2 # rare is better second_tgt = tok2 if first_tgt == tok1 else tok1 syn_dic_first = synonym_dic.get(first_tgt.lower(), None) if syn_dic_first is not None: return (first_tgt, syn_dic_first) syn_dic_second = synonym_dic.get(second_tgt.lower(), None) return (second_tgt, syn_dic_second) if syn_dic_second is not None else (None, None)<jupyter_output><empty_output><jupyter_text>REPLACE & ADD Interface<jupyter_code>def normalize_token(token): """ Do strict noramlization for a given token. All punctuations will be removed. """ def remove_articles(text): return re.sub(r'\b(a|an|the)\b', ' ', text) def white_space_fix(text): return ' '.join(text.split()) def remove_punc(text): exclude = set(string.punctuation) return ''.join(ch for ch in text if ch not in exclude) def lower(text): return text.lower() return white_space_fix(remove_articles(remove_punc(lower(token)))) def trim_retrieval_results(replacement_dict): """ All overly short (char len < 4) / misspelled / contains numbers tokens are not replaceable. Filter them out. """ out_dic = deepcopy(replacement_dict) for col in replacement_dict.keys(): replacements = replacement_dict[col] filter_rpls = [r for r in replacements if len(r) > 4] out_dic[col] = filter_rpls return out_dic def consider_rpl(token): """ Judge whether a token is suitable for replacement. All overly short (char len < 4) / misspelled / contains numbers col are not replaceable. """ norm_token = normalize_token(token) if len(norm_token) < 4: return False, token if contains_number(norm_token): return False, token if not check_spell(norm_token): return False, token return True, norm_token<jupyter_output><empty_output><jupyter_text>core functions<jupyter_code>def replace_and_add_for_give_tables(path, table2qs, batch_size=512, replace_threshold=0.75, add_threshold=0.4, output_prefix="", output_dir="./processed_data", delim=" ", topk_tables=50, max_cands_per_col=10, table_name_key="table_name", col_type_key="column_types", col_name_key="column_names"): """ The highest-level interface for replacement and addition across all tables stored in a given path. One call, handle all. :path: Target tables path :table2qs: The queries corresponding to the each of the tables. :batch_size: bsz for NLI checking. 512 recommended. :replace_threshold: If NLI entailment score is higher than this threshold under STRICT mode, then the rpl pair is accepted. :add_treshold: If NLI entailment score is lower than this threshold under LOOSE mode, then the add pair is accepted. :output_prefix: File name prefix for output file. :output_dir: Output file directory. :delim: Delimiator for column names. Single white space by default. :topk_tables: How many most similar tables to consider from dense retrieval. :max_cands_per_col: Max nubmer of pairs to be considered for each column (both add and rpl). This directly influences the final amount to be checked by NLI. """ if not os.path.exists(output_dir): os.makedirs(output_dir) # STEP 0: Prepare tables print("STEP 0 : Prepare tables...\n") tables = [] tables_template = {} cnt = 0 with open(path, "r") as f: for line in f.readlines(): table = json.loads(line) table_copy = deepcopy(table) table_copy["rpls_retrieval"] = {} table_copy["rpls_syndict"] = {} tables.append(table_copy) table["REPLACE"] = {tname:[] for tname in table[col_name_key]} table["ADD"] = {tname:[] for tname in table[col_name_key]} tables_template[table["table_id"]] = table # if cnt == 5: break cnt += 1 # STEP 1: "retrieval" for add / replacement print('STEP 1: Dense retrieval for add / replacement...\n') for tab in tqdm(tables, position=0): tid = tab["table_id"] queries = table2qs.get(tid, None) if queries is not None: queries = queries[:10] if len(queries) > 10 else queries ori_cols, expanded_cols = retriver(query_table=tab, queries=queries, topk_tables=topk_tables) rec_dic = reranker(ori_cols, expanded_cols, topk=max_cands_per_col) # We will find analog & synonyms in this list tab["rpls_retrieval"] = trim_retrieval_results(rec_dic) # STEP 2 : synonym dict for replacement print('STEP 2 : synonym dict for replacement...\n') for tab in tqdm(tables, position=1): if isinstance(tab["column_types"], list): col2type = {c:t for c,t in zip(tab["column_names"], tab["column_types"])} else: col2type = tab["column_types"] for col in tab["column_names"]: col_type = col2type.get(col, "text") if _ends_with_id(col): continue tokens = [w.lower() for w in col.split(delim)] keep_original, normalized_tokens = [], [] tok2syn = {} for tok in tokens: can_rpl, tok = consider_rpl(tok) normalized_tokens.append(tok) syn_dic = synonym_dic.get(tok, None) if syn_dic is None: syn_dic = synonym_dic.get(get_singular_word(tok), None) keep_ori = (not can_rpl) or (syn_dic is None) # skip eihter because not replaceable or not in dic keep_original.append(keep_ori) if keep_ori == True: continue rec_dic = reranker(tgt_names=[tok], cand_names=list(set(syn_dic["synonyms"])), topk=10) tok2syn.update(rec_dic) syn_rpl_candidates = set() # Genereate syn-replaced candidates patience = 5 # if 5 in steps there is no new candidate added, break the loop. while True: if len(syn_rpl_candidates) >= max_cands_per_col or patience == 0: syn_rpl_candidates = list(syn_rpl_candidates.difference(set([" ".join(normalized_tokens)]))) break rpl_threshold = 1 if len(tokens) == 1 else (0.75 if len(tokens) == 2 else 0.5) do_rpl_coins = np.random.rand(len(tokens)) <= rpl_threshold # only keep original for 20% of time new_cand = [] for i, tok in enumerate(normalized_tokens): if not keep_original[i] and do_rpl_coins[i]: all_syns = tok2syn.get(tok, [tok]) syn = np.random.choice(all_syns) new_cand.append(syn) else: new_cand.append(tok) new_cand = delim.join(new_cand) len_before = len(syn_rpl_candidates) syn_rpl_candidates.add(new_cand) if len(syn_rpl_candidates) > len_before: patience = 5 else: patience -= 1 tab["rpls_syndict"].update({col : syn_rpl_candidates}) # #STEP 3: filter syn dict replacement with NLI print('STEP 3: filter syn dict replacement with NLI...\n') STRICT_MODE = True constructed_pairs_rpl_syndict = construct_pairs_for_nli_test(tables, pending_rpls_key="rpls_syndict", table_name_key=table_name_key, col_type_key=col_type_key, col_name_key=col_name_key) results_rpl_syndict = nli_test_across_tables(constructed_pairs_rpl_syndict, batch_size=batch_size) for i, dic in enumerate(results_rpl_syndict): table_id = constructed_pairs_rpl_syndict[i]["table_id"] table = tables_template[table_id] ent = dic["scores"][2] if ent >= replace_threshold: ori, rpl = dic["key_map"] # print(f"{ori} -> {rpl}") table["REPLACE"][ori].append(rpl) # update REPLACE key # print(results_rpl_syndict) # STEP 4: filter retrieval replacement with NLI print('STEP 4: filter retrieval replacement with NLI...\n') STRICT_MODE = True constructed_pairs_rpl_retrieval = construct_pairs_for_nli_test(tables, pending_rpls_key="rpls_retrieval", table_name_key=table_name_key, col_type_key=col_type_key, col_name_key=col_name_key) results_rpl_retrieval = nli_test_across_tables(constructed_pairs_rpl_retrieval, batch_size=batch_size) for i, dic in enumerate(results_rpl_retrieval): table_id = constructed_pairs_rpl_retrieval[i]["table_id"] table = tables_template[table_id] ent = dic["scores"][2] if ent >= replace_threshold: ori, rpl = dic["key_map"] table["REPLACE"][ori].append(rpl) # update REPLACE key # STEP 5: prune all replaceable from retrieval results & filter substring overlap from original col print('STEP 5: prune all replaceable from retrieval results & filter substring overlap from original col...\n') for table in tables: tid = table["table_id"] all_columns = table[col_name_key] rpl_dict = tables_template[tid]["REPLACE"] for rpl_col in table["rpls_retrieval"].keys(): rpl_candidates = table["rpls_retrieval"][rpl_col] add_candidates = [] # ADD operation candidates comes from here for rpl in rpl_candidates: if any([rpl in c for c in all_columns]) or any([rpl in c for c in rpl_dict[rpl_col]]): continue add_candidates.append(rpl) table["rpls_retrieval"][rpl_col] = add_candidates # print(tables[8]["rpls_retrieval"]) # STEP 6 : filter leftover retrieval ADD candidates with NLI print('STEP 6 : filter leftover retrieval ADD candidates with NLI\n') STRICT_MODE = False constructed_pairs_add_retrieval = construct_pairs_for_nli_test(tables, pending_rpls_key="rpls_retrieval", table_name_key=table_name_key, col_type_key=col_type_key, col_name_key=col_name_key) results_add_retrieval = nli_test_across_tables(constructed_pairs_add_retrieval, batch_size=batch_size) for i, dic in enumerate(results_add_retrieval): table_id = constructed_pairs_add_retrieval[i]["table_id"] table = tables_template[table_id] ent = dic["scores"][2] if ent <= add_threshold: ori, rpl = dic["key_map"] table["ADD"][ori].append(rpl) # update ADD key # print(tables_template["SPIDER_8"]["ADD"]) # STEP 7: Write replace + add results to new file print('STEP 7: Write REPLACE & ADD results to new file \n') with open(f"{output_dir}/{output_prefix}-pipeline-output.jsonl", "w") as f: for table in tables_template.values(): json.dump(table, f) f.write("\n")<jupyter_output><empty_output><jupyter_text>Leave For Running Spider<jupyter_code>replace_and_add_for_give_tables("./data/spider/spider-tables.jsonl", table2qs=spider_table2qs, output_prefix="spider", table_name_key="table_name", replace_threshold=0.50, batch_size=512)<jupyter_output><empty_output><jupyter_text>WTQ<jupyter_code>replace_and_add_for_give_tables("./data/WTQ/wtq-tables.jsonl", table2qs=wtq_table2qs, output_prefix="wtq", table_name_key="pred_table_name", replace_threshold=0.50, batch_size=512)<jupyter_output><empty_output><jupyter_text>WikiSQL<jupyter_code>replace_and_add_for_give_tables("./data/wikisql/wikisql-tables.jsonl", table2qs=wsql_train_table2qs, output_prefix="wsql-train", table_name_key="pred_table_name", replace_threshold=0.70, batch_size=128)<jupyter_output><empty_output>

ContextualSP/robustness_of_text_to_sql/CTA/pipeline.ipynb/0

# Robustness of Text-to-SQL Models This repository contains the data and code in the following paper: > [**Towards Robustness of Text-to-SQL Models Against Natural and Realistic Adversarial Table Perturbation**](https://aclanthology.org/2022.acl-long.142.pdf) <br/> > Xinyu Pi*, Bing Wang*, Yan Gao, Jiaqi Guo, Zhoujun Li, Jian-Guang Lou<br/> > ACL 2022 Long Papers ## Introduction This repository is the official implementation of our paper *Towards Robustness of Text-to-SQL Models Against Natural and Realistic Adversarial Table Perturbation*. In this paper, we curate **ADVETA**, the first robustness evaluation benchmark featuring natural and realistic adversarial table perturbation. To defend against this perturbation, we build a systematic adversarial training example generation framework **CTA**, tailored for better contextualization of tabular data. ## ADVETA <img src="misc/ATP.png" height=450> We manually curate the **ADVE**rsarial **T**able perturb**A**tion (ADVETA) benchmark based on three mainstream Text-to-SQL datasets, Spider, WikiSQL and WTQ. For each table from the original development set, we conduct RPL/ADD annotation separately, perturbing only table columns. We release our data in `adveta_1.0.zip` file. ## CTA <img src="misc/CTA.png" height=400> ### Requirement - python: 3.8 - cuda: 10.1 - torch: 1.7.1 install dependencies: ```bash conda create -n cta python=3.8 -y conda activate cta conda install pytorch==1.7.1 cudatoolkit=10.1 -c pytorch -y python -m spacy download en_core_web_sm pip install -r requirements.txt ``` ### Introduction Contextualized Table Augmentation (CTA) framework as an adversarial training example generation approach tailored for tabular data. Before you run `pipeline.ipynb`, you should download data files and checkpoints from [Google Drive](https://drive.google.com/file/d/1HqP1P5QqytGZTM_Kx8Bbq0EuyKaD9raV/view?usp=sharing). notes: - We download number-batch word embedding from [here](https://conceptnet.s3.amazonaws.com/downloads/2019/numberbatch/numberbatch-en-19.08.txt.gz) as `./data/nb_emb.txt`. - We pre-compute processed-WDC tables using Tapas dense retrieval models. Store output to `./wdc/wdc_dense_A.txt` and `./wdc/wdc_dense_B.txt` (Tapas have two encoders). ### Run Just run the `pipeline.ipynb` and have fun. ## Cite ``` @inproceedings{pi-etal-2022-towards, title = "Towards Robustness of Text-to-{SQL} Models Against Natural and Realistic Adversarial Table Perturbation", author = "Pi, Xinyu and Wang, Bing and Gao, Yan and Guo, Jiaqi and Li, Zhoujun and Lou, Jian-Guang", month = may, year = "2022", address = "Dublin, Ireland", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2022.acl-long.142", pages = "2007--2022" } ```

ContextualSP/robustness_of_text_to_sql/README.md/0

set seed=1 set config_file=train_configs/concat.none.jsonnet set model_file=checkpoints_sparc/sparc_concat_none_model set tables_file=dataset_sparc/tables.json set database_path=dataset_sparc/database set dataset_path=dataset_sparc set train_data_path=dataset_sparc/train.json set validation_data_path=dataset_sparc/dev.json set pretrained_file=glove/glove.twitter.27B.100d.txt allennlp train -s %model_file% %config_file% ^ --include-package dataset_reader.sparc_reader ^ --include-package models.sparc_parser ^ -o {"""model.serialization_dir""":"""%model_file%""","""random_seed""":"""%seed%""","""numpy_seed""":"""%seed%""","""pytorch_seed""":"""%seed%""","""dataset_reader.tables_file""":"""%tables_file%""","""dataset_reader.database_path""":"""%database_path%""","""train_data_path""":"""%train_data_path%""","""validation_data_path""":"""%validation_data_path%""","""model.text_embedder.tokens.pretrained_file""":"""%pretrained_file%""","""model.dataset_path""":"""%dataset_path%"""}

ContextualSP/semantic_parsing_in_context/bash_files/windows/train_sparc.bat/0

from allennlp.models.archival import load_archive from allennlp.predictors.predictor import Predictor # WARNING: Do not exclude these imports from predictor.sparc_predictor import SparcPredictor from dataset_reader.sparc_reader import SparcDatasetReader from models.sparc_parser import SparcParser class PredictManager: def __init__(self, archive_file, tables_file, database_path): overrides = "{\"dataset_reader.tables_file\":\"" + tables_file + "\",\"dataset_reader.database_path\":" +\ "\"" + database_path + "\"}" archive = load_archive(archive_file, overrides=overrides) self.predictor = Predictor.from_archive( archive, predictor_name="sparc") def predict_result(self, ques_inter: str, ques_database: str): param = { "database_id": ques_database, "question": ques_inter } restate = self.predictor.predict_json(param)["best_predict_sql"] return restate if __name__ == '__main__': manager = PredictManager(archive_file="model.tar.gz", tables_file="dataset_sparc/tables.json", database_path="dataset_sparc/database") # the input dialogue is separate by `;`, and the second argument is database_id result = manager.predict_result("What are all the airlines;Of these, which is Jetblue Airways", "flight_2") print(result)

ContextualSP/semantic_parsing_in_context/predict.py/0

# pylint: disable=anomalous-backslash-in-string """ A ``Text2SqlTableContext`` represents the SQL context in which an utterance appears for the any of the text2sql datasets, with the grammar and the valid actions. """ from typing import List, Dict from dataset_readers.dataset_util.spider_utils import Table GRAMMAR_DICTIONARY = {} GRAMMAR_DICTIONARY["statement"] = ['(query ws iue ws query)', '(query ws)'] GRAMMAR_DICTIONARY["iue"] = ['"intersect"', '"except"', '"union"'] GRAMMAR_DICTIONARY["query"] = ['(ws select_core ws groupby_clause ws orderby_clause ws limit)', '(ws select_core ws groupby_clause ws orderby_clause)', '(ws select_core ws groupby_clause ws limit)', '(ws select_core ws orderby_clause ws limit)', '(ws select_core ws groupby_clause)', '(ws select_core ws orderby_clause)', '(ws select_core)'] GRAMMAR_DICTIONARY["select_core"] = ['(select_with_distinct ws select_results ws from_clause ws where_clause)', '(select_with_distinct ws select_results ws from_clause)', '(select_with_distinct ws select_results ws where_clause)', '(select_with_distinct ws select_results)'] GRAMMAR_DICTIONARY["select_with_distinct"] = ['(ws "select" ws "distinct")', '(ws "select")'] GRAMMAR_DICTIONARY["select_results"] = ['(ws select_result ws "," ws select_results)', '(ws select_result)'] GRAMMAR_DICTIONARY["select_result"] = ['"*"', '(table_source ws ".*")', 'expr', 'col_ref'] GRAMMAR_DICTIONARY["from_clause"] = ['(ws "from" ws table_source ws join_clauses)', '(ws "from" ws source)'] GRAMMAR_DICTIONARY["join_clauses"] = ['(join_clause ws join_clauses)', 'join_clause'] GRAMMAR_DICTIONARY["join_clause"] = ['"join" ws table_source ws "on" ws join_condition_clause'] GRAMMAR_DICTIONARY["join_condition_clause"] = ['(join_condition ws "and" ws join_condition_clause)', 'join_condition'] GRAMMAR_DICTIONARY["join_condition"] = ['ws col_ref ws "=" ws col_ref'] GRAMMAR_DICTIONARY["source"] = ['(ws single_source ws "," ws source)', '(ws single_source)'] GRAMMAR_DICTIONARY["single_source"] = ['table_source', 'source_subq'] GRAMMAR_DICTIONARY["source_subq"] = ['("(" ws query ws ")")'] # GRAMMAR_DICTIONARY["source_subq"] = ['("(" ws query ws ")" ws "as" ws name)', '("(" ws query ws ")")'] GRAMMAR_DICTIONARY["limit"] = ['("limit" ws non_literal_number)'] GRAMMAR_DICTIONARY["where_clause"] = ['(ws "where" wsp expr ws where_conj)', '(ws "where" wsp expr)'] GRAMMAR_DICTIONARY["where_conj"] = ['(ws "and" wsp expr ws where_conj)', '(ws "and" wsp expr)'] GRAMMAR_DICTIONARY["groupby_clause"] = ['(ws "group" ws "by" ws group_clause ws "having" ws expr)', '(ws "group" ws "by" ws group_clause)'] GRAMMAR_DICTIONARY["group_clause"] = ['(ws expr ws "," ws group_clause)', '(ws expr)'] GRAMMAR_DICTIONARY["orderby_clause"] = ['ws "order" ws "by" ws order_clause'] GRAMMAR_DICTIONARY["order_clause"] = ['(ordering_term ws "," ws order_clause)', 'ordering_term'] GRAMMAR_DICTIONARY["ordering_term"] = ['(ws expr ws ordering)', '(ws expr)'] GRAMMAR_DICTIONARY["ordering"] = ['(ws "asc")', '(ws "desc")'] GRAMMAR_DICTIONARY["col_ref"] = ['(table_name ws "." ws column_name)', 'column_name'] GRAMMAR_DICTIONARY["table_source"] = ['(table_name ws "as" ws table_alias)', 'table_name'] GRAMMAR_DICTIONARY["table_name"] = ["table_alias"] GRAMMAR_DICTIONARY["table_alias"] = ['"t1"', '"t2"', '"t3"', '"t4"'] GRAMMAR_DICTIONARY["column_name"] = [] GRAMMAR_DICTIONARY["ws"] = ['~"\s*"i'] GRAMMAR_DICTIONARY['wsp'] = ['~"\s+"i'] GRAMMAR_DICTIONARY["expr"] = ['in_expr', # Like expressions. '(value wsp "like" wsp string)', # Between expressions. '(value ws "between" wsp value ws "and" wsp value)', # Binary expressions. '(value ws binaryop wsp expr)', # Unary expressions. '(unaryop ws expr)', 'source_subq', 'value'] GRAMMAR_DICTIONARY["in_expr"] = ['(value wsp "not" wsp "in" wsp string_set)', '(value wsp "in" wsp string_set)', '(value wsp "not" wsp "in" wsp expr)', '(value wsp "in" wsp expr)'] GRAMMAR_DICTIONARY["value"] = ['parenval', '"YEAR(CURDATE())"', 'number', 'boolean', 'function', 'col_ref', 'string'] GRAMMAR_DICTIONARY["parenval"] = ['"(" ws expr ws ")"'] GRAMMAR_DICTIONARY["function"] = ['(fname ws "(" ws "distinct" ws arg_list_or_star ws ")")', '(fname ws "(" ws arg_list_or_star ws ")")'] GRAMMAR_DICTIONARY["arg_list_or_star"] = ['arg_list', '"*"'] GRAMMAR_DICTIONARY["arg_list"] = ['(expr ws "," ws arg_list)', 'expr'] # TODO(MARK): Massive hack, remove and modify the grammar accordingly # GRAMMAR_DICTIONARY["number"] = ['~"\d*\.?\d+"i', "'3'", "'4'"] GRAMMAR_DICTIONARY["non_literal_number"] = ['"1"', '"2"', '"3"', '"4"'] GRAMMAR_DICTIONARY["number"] = ['ws "value" ws'] GRAMMAR_DICTIONARY["string_set"] = ['ws "(" ws string_set_vals ws ")"'] GRAMMAR_DICTIONARY["string_set_vals"] = ['(string ws "," ws string_set_vals)', 'string'] # GRAMMAR_DICTIONARY["string"] = ['~"\'.*?\'"i'] GRAMMAR_DICTIONARY["string"] = ['"\'" ws "value" ws "\'"'] GRAMMAR_DICTIONARY["fname"] = ['"count"', '"sum"', '"max"', '"min"', '"avg"', '"all"'] GRAMMAR_DICTIONARY["boolean"] = ['"true"', '"false"'] # TODO(MARK): This is not tight enough. AND/OR are strictly boolean value operators. GRAMMAR_DICTIONARY["binaryop"] = ['"+"', '"-"', '"*"', '"/"', '"="', '"!="', '"<>"', '">="', '"<="', '">"', '"<"', '"and"', '"or"', '"like"'] GRAMMAR_DICTIONARY["unaryop"] = ['"+"', '"-"', '"not"', '"not"'] def update_grammar_with_tables(grammar_dictionary: Dict[str, List[str]], schema: Dict[str, Table]) -> None: table_names = sorted([f'"{table.lower()}"' for table in list(schema.keys())], reverse=True) grammar_dictionary['table_name'] += table_names all_columns = set() for table in schema.values(): all_columns.update([f'"{table.name.lower()}@{column.name.lower()}"' for column in table.columns if column.name != '*']) sorted_columns = sorted([column for column in all_columns], reverse=True) grammar_dictionary['column_name'] += sorted_columns def update_grammar_to_be_table_names_free(grammar_dictionary: Dict[str, List[str]]): """ Remove table names from column names, remove aliases """ grammar_dictionary["column_name"] = [] grammar_dictionary["table_name"] = [] grammar_dictionary["col_ref"] = ['column_name'] grammar_dictionary["table_source"] = ['table_name'] del grammar_dictionary["table_alias"] def update_grammar_flip_joins(grammar_dictionary: Dict[str, List[str]]): """ Remove table names from column names, remove aliases """ # using a simple rule such as join_clauses-> [(join_clauses ws join_clause), join_clause] # resulted in a max recursion error, so for now just using a predefined max # number of joins grammar_dictionary["join_clauses"] = ['(join_clauses_1 ws join_clause)', 'join_clause'] grammar_dictionary["join_clauses_1"] = ['(join_clauses_2 ws join_clause)', 'join_clause'] grammar_dictionary["join_clauses_2"] = ['(join_clause ws join_clause)', 'join_clause']

ContextualSP/unified_parser_text_to_sql/semparse/contexts/spider_db_grammar.py/0

import os, sys import json import sqlite3 import traceback import argparse import tqdm from ..process_sql import get_sql from .schema import Schema, get_schemas_from_json if __name__ == "__main__": parser = argparse.ArgumentParser() parser.add_argument("--input", required=True) parser.add_argument("--tables", required=True) parser.add_argument("--output", required=True) args = parser.parse_args() sql_path = args.input output_file = args.output table_file = args.tables schemas, db_names, tables = get_schemas_from_json(table_file) with open(sql_path) as inf: sql_data = json.load(inf) sql_data_new = [] for data in tqdm.tqdm(sql_data): try: db_id = data["db_id"] schema = schemas[db_id] table = tables[db_id] schema = Schema(schema, table) sql = data["query"] sql_label = get_sql(schema, sql) data["sql"] = sql_label sql_data_new.append(data) except: print("db_id: ", db_id) print("sql: ", sql) raise with open(output_file, "wt") as out: json.dump(sql_data_new, out, sort_keys=True, indent=4, separators=(",", ": "))

ContextualSP/unified_parser_text_to_sql/third_party/spider/preprocess/parse_raw_json.py/0

from PIL import Image import io import torch from torch.utils.data import Dataset import torchvision.transforms as transforms class ImageNet_Withhold(Dataset): def __init__(self, data_root, ann_file='', transform=None, train=True, task ='train'): super(ImageNet_Withhold, self).__init__() ann_file = ann_file + '/' + 'val_true.txt' train_split = (task == 'train' or task == 'val') self.data_root = data_root + '/'+ ('train' if train_split else 'val') self.data = [] self.nb_classes = 0 folders = {} cnt = 0 self.z = ZipReader() # if train: # for member in self.tarfile.getmembers(): # print(member) # self.tarfile = tarfile.open(self.data_root) f = open(ann_file) prefix = 'data/sdb/imagenet'+'/'+ ('train' if train_split else 'val') + '/' for line in f: tmp = line.strip().split('\t')[0] class_pic = tmp.split('/') class_tmp = class_pic[0] pic = class_pic[1] if class_tmp in folders: # print(self.tarfile.getmember(('train/' if train else 'val/') + tmp[0] + '.JPEG')) self.data.append((class_tmp + '.zip', prefix + tmp + '.JPEG', folders[class_tmp])) else: folders[class_tmp] = cnt cnt += 1 self.data.append((class_tmp + '.zip', prefix + tmp + '.JPEG',folders[class_tmp])) normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]) if transform is not None: self.transforms = transform else: if train: self.transforms = transforms.Compose([ transforms.RandomSizedCrop(224), transforms.RandomHorizontalFlip(), transforms.ToTensor(), normalize, ]) else: self.transforms = transforms.Compose([ transforms.Scale(256), transforms.CenterCrop(224), transforms.ToTensor(), normalize, ]) self.nb_classes = cnt def __len__(self): return len(self.data) def __getitem__(self, idx): # print('extract_file', time.time()-start_time) iob = self.z.read(self.data_root + '/' + self.data[idx][0], self.data[idx][1]) iob = io.BytesIO(iob) img = Image.open(iob).convert('RGB') target = self.data[idx][2] if self.transforms is not None: img = self.transforms(img) # print('open', time.time()-start_time) return img, target

Cream/AutoFormer/lib/imagenet_withhold.py/0

import logging import torch.nn as nn import torch.utils.checkpoint as cp from ..runner import load_checkpoint from .weight_init import constant_init, kaiming_init def conv3x3(in_planes, out_planes, stride=1, dilation=1): "3x3 convolution with padding" return nn.Conv2d( in_planes, out_planes, kernel_size=3, stride=stride, padding=dilation, dilation=dilation, bias=False) class BasicBlock(nn.Module): expansion = 1 def __init__(self, inplanes, planes, stride=1, dilation=1, downsample=None, style='pytorch', with_cp=False): super(BasicBlock, self).__init__() self.conv1 = conv3x3(inplanes, planes, stride, dilation) self.bn1 = nn.BatchNorm2d(planes) self.relu = nn.ReLU(inplace=True) self.conv2 = conv3x3(planes, planes) self.bn2 = nn.BatchNorm2d(planes) self.downsample = downsample self.stride = stride self.dilation = dilation assert not with_cp def forward(self, x): residual = x out = self.conv1(x) out = self.bn1(out) out = self.relu(out) out = self.conv2(out) out = self.bn2(out) if self.downsample is not None: residual = self.downsample(x) out += residual out = self.relu(out) return out class Bottleneck(nn.Module): expansion = 4 def __init__(self, inplanes, planes, stride=1, dilation=1, downsample=None, style='pytorch', with_cp=False): """Bottleneck block. If style is "pytorch", the stride-two layer is the 3x3 conv layer, if it is "caffe", the stride-two layer is the first 1x1 conv layer. """ super(Bottleneck, self).__init__() assert style in ['pytorch', 'caffe'] if style == 'pytorch': conv1_stride = 1 conv2_stride = stride else: conv1_stride = stride conv2_stride = 1 self.conv1 = nn.Conv2d( inplanes, planes, kernel_size=1, stride=conv1_stride, bias=False) self.conv2 = nn.Conv2d( planes, planes, kernel_size=3, stride=conv2_stride, padding=dilation, dilation=dilation, bias=False) self.bn1 = nn.BatchNorm2d(planes) self.bn2 = nn.BatchNorm2d(planes) self.conv3 = nn.Conv2d( planes, planes * self.expansion, kernel_size=1, bias=False) self.bn3 = nn.BatchNorm2d(planes * self.expansion) self.relu = nn.ReLU(inplace=True) self.downsample = downsample self.stride = stride self.dilation = dilation self.with_cp = with_cp def forward(self, x): def _inner_forward(x): residual = x out = self.conv1(x) out = self.bn1(out) out = self.relu(out) out = self.conv2(out) out = self.bn2(out) out = self.relu(out) out = self.conv3(out) out = self.bn3(out) if self.downsample is not None: residual = self.downsample(x) out += residual return out if self.with_cp and x.requires_grad: out = cp.checkpoint(_inner_forward, x) else: out = _inner_forward(x) out = self.relu(out) return out def make_res_layer(block, inplanes, planes, blocks, stride=1, dilation=1, style='pytorch', with_cp=False): downsample = None if stride != 1 or inplanes != planes * block.expansion: downsample = nn.Sequential( nn.Conv2d( inplanes, planes * block.expansion, kernel_size=1, stride=stride, bias=False), nn.BatchNorm2d(planes * block.expansion), ) layers = [] layers.append( block( inplanes, planes, stride, dilation, downsample, style=style, with_cp=with_cp)) inplanes = planes * block.expansion for i in range(1, blocks): layers.append( block(inplanes, planes, 1, dilation, style=style, with_cp=with_cp)) return nn.Sequential(*layers) class ResNet(nn.Module): """ResNet backbone. Args: depth (int): Depth of resnet, from {18, 34, 50, 101, 152}. num_stages (int): Resnet stages, normally 4. strides (Sequence[int]): Strides of the first block of each stage. dilations (Sequence[int]): Dilation of each stage. out_indices (Sequence[int]): Output from which stages. style (str): `pytorch` or `caffe`. If set to "pytorch", the stride-two layer is the 3x3 conv layer, otherwise the stride-two layer is the first 1x1 conv layer. frozen_stages (int): Stages to be frozen (all param fixed). -1 means not freezing any parameters. bn_eval (bool): Whether to set BN layers as eval mode, namely, freeze running stats (mean and var). bn_frozen (bool): Whether to freeze weight and bias of BN layers. with_cp (bool): Use checkpoint or not. Using checkpoint will save some memory while slowing down the training speed. """ arch_settings = { 18: (BasicBlock, (2, 2, 2, 2)), 34: (BasicBlock, (3, 4, 6, 3)), 50: (Bottleneck, (3, 4, 6, 3)), 101: (Bottleneck, (3, 4, 23, 3)), 152: (Bottleneck, (3, 8, 36, 3)) } def __init__(self, depth, num_stages=4, strides=(1, 2, 2, 2), dilations=(1, 1, 1, 1), out_indices=(0, 1, 2, 3), style='pytorch', frozen_stages=-1, bn_eval=True, bn_frozen=False, with_cp=False): super(ResNet, self).__init__() if depth not in self.arch_settings: raise KeyError('invalid depth {} for resnet'.format(depth)) assert num_stages >= 1 and num_stages <= 4 block, stage_blocks = self.arch_settings[depth] stage_blocks = stage_blocks[:num_stages] assert len(strides) == len(dilations) == num_stages assert max(out_indices) < num_stages self.out_indices = out_indices self.style = style self.frozen_stages = frozen_stages self.bn_eval = bn_eval self.bn_frozen = bn_frozen self.with_cp = with_cp self.inplanes = 64 self.conv1 = nn.Conv2d( 3, 64, kernel_size=7, stride=2, padding=3, bias=False) self.bn1 = nn.BatchNorm2d(64) self.relu = nn.ReLU(inplace=True) self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) self.res_layers = [] for i, num_blocks in enumerate(stage_blocks): stride = strides[i] dilation = dilations[i] planes = 64 * 2**i res_layer = make_res_layer( block, self.inplanes, planes, num_blocks, stride=stride, dilation=dilation, style=self.style, with_cp=with_cp) self.inplanes = planes * block.expansion layer_name = 'layer{}'.format(i + 1) self.add_module(layer_name, res_layer) self.res_layers.append(layer_name) self.feat_dim = block.expansion * 64 * 2**(len(stage_blocks) - 1) def init_weights(self, pretrained=None): if isinstance(pretrained, str): logger = logging.getLogger() load_checkpoint(self, pretrained, strict=False, logger=logger) elif pretrained is None: for m in self.modules(): if isinstance(m, nn.Conv2d): kaiming_init(m) elif isinstance(m, nn.BatchNorm2d): constant_init(m, 1) else: raise TypeError('pretrained must be a str or None') def forward(self, x): x = self.conv1(x) x = self.bn1(x) x = self.relu(x) x = self.maxpool(x) outs = [] for i, layer_name in enumerate(self.res_layers): res_layer = getattr(self, layer_name) x = res_layer(x) if i in self.out_indices: outs.append(x) if len(outs) == 1: return outs[0] else: return tuple(outs) def train(self, mode=True): super(ResNet, self).train(mode) if self.bn_eval: for m in self.modules(): if isinstance(m, nn.BatchNorm2d): m.eval() if self.bn_frozen: for params in m.parameters(): params.requires_grad = False if mode and self.frozen_stages >= 0: for param in self.conv1.parameters(): param.requires_grad = False for param in self.bn1.parameters(): param.requires_grad = False self.bn1.eval() self.bn1.weight.requires_grad = False self.bn1.bias.requires_grad = False for i in range(1, self.frozen_stages + 1): mod = getattr(self, 'layer{}'.format(i)) mod.eval() for param in mod.parameters(): param.requires_grad = False

Cream/CDARTS/CDARTS_detection/mmcv/cnn/resnet.py/0

import numpy as np from .colorspace import bgr2rgb, rgb2bgr def imnormalize(img, mean, std, to_rgb=True): img = img.astype(np.float32) if to_rgb: img = bgr2rgb(img) return (img - mean) / std def imdenormalize(img, mean, std, to_bgr=True): img = (img * std) + mean if to_bgr: img = rgb2bgr(img) return img

Cream/CDARTS/CDARTS_detection/mmcv/image/transforms/normalize.py/0

class Hook(object): def before_run(self, runner): pass def after_run(self, runner): pass def before_epoch(self, runner): pass def after_epoch(self, runner): pass def before_iter(self, runner): pass def after_iter(self, runner): pass def before_train_epoch(self, runner): self.before_epoch(runner) def before_val_epoch(self, runner): self.before_epoch(runner) def after_train_epoch(self, runner): self.after_epoch(runner) def after_val_epoch(self, runner): self.after_epoch(runner) def before_train_iter(self, runner): self.before_iter(runner) def before_val_iter(self, runner): self.before_iter(runner) def after_train_iter(self, runner): self.after_iter(runner) def arch_after_train_iter(self, runner): self.after_iter(runner) def after_val_iter(self, runner): self.after_iter(runner) def every_n_epochs(self, runner, n): return (runner.epoch + 1) % n == 0 if n > 0 else False def every_n_inner_iters(self, runner, n): return (runner.inner_iter + 1) % n == 0 if n > 0 else False def every_n_iters(self, runner, n): return (runner.iter + 1) % n == 0 if n > 0 else False def end_of_epoch(self, runner): return runner.inner_iter + 1 == len(runner.data_loader)

Cream/CDARTS/CDARTS_detection/mmcv/runner/hooks/hook.py/0

from .config import ConfigDict, Config from .misc import (is_str, iter_cast, list_cast, tuple_cast, is_seq_of, is_list_of, is_tuple_of, slice_list, concat_list, check_prerequisites, requires_package, requires_executable) from .path import (is_filepath, fopen, check_file_exist, mkdir_or_exist, symlink, scandir, FileNotFoundError) from .progressbar import ProgressBar, track_progress, track_parallel_progress from .timer import Timer, TimerError, check_time __all__ = [ 'ConfigDict', 'Config', 'is_str', 'iter_cast', 'list_cast', 'tuple_cast', 'is_seq_of', 'is_list_of', 'is_tuple_of', 'slice_list', 'concat_list', 'check_prerequisites', 'requires_package', 'requires_executable', 'is_filepath', 'fopen', 'check_file_exist', 'mkdir_or_exist', 'symlink', 'scandir', 'FileNotFoundError', 'ProgressBar', 'track_progress', 'track_parallel_progress', 'Timer', 'TimerError', 'check_time' ]

Cream/CDARTS/CDARTS_detection/mmcv/utils/__init__.py/0

from .color import Color, color_val from .image import imshow, imshow_bboxes, imshow_det_bboxes from .optflow import flowshow, flow2rgb, make_color_wheel __all__ = [ 'Color', 'color_val', 'imshow', 'imshow_bboxes', 'imshow_det_bboxes', 'flowshow', 'flow2rgb', 'make_color_wheel' ]

Cream/CDARTS/CDARTS_detection/mmcv/visualization/__init__.py/0

from .anchor_generator import AnchorGenerator from .anchor_target import anchor_target, anchor_inside_flags from .guided_anchor_target import ga_loc_target, ga_shape_target __all__ = [ 'AnchorGenerator', 'anchor_target', 'anchor_inside_flags', 'ga_loc_target', 'ga_shape_target' ]

Cream/CDARTS/CDARTS_detection/mmdet/core/anchor/__init__.py/0

import numpy as np import torch from .random_sampler import RandomSampler class InstanceBalancedPosSampler(RandomSampler): def _sample_pos(self, assign_result, num_expected, **kwargs): pos_inds = torch.nonzero(assign_result.gt_inds > 0) if pos_inds.numel() != 0: pos_inds = pos_inds.squeeze(1) if pos_inds.numel() <= num_expected: return pos_inds else: unique_gt_inds = assign_result.gt_inds[pos_inds].unique() num_gts = len(unique_gt_inds) num_per_gt = int(round(num_expected / float(num_gts)) + 1) sampled_inds = [] for i in unique_gt_inds: inds = torch.nonzero(assign_result.gt_inds == i.item()) if inds.numel() != 0: inds = inds.squeeze(1) else: continue if len(inds) > num_per_gt: inds = self.random_choice(inds, num_per_gt) sampled_inds.append(inds) sampled_inds = torch.cat(sampled_inds) if len(sampled_inds) < num_expected: num_extra = num_expected - len(sampled_inds) extra_inds = np.array( list(set(pos_inds.cpu()) - set(sampled_inds.cpu()))) if len(extra_inds) > num_extra: extra_inds = self.random_choice(extra_inds, num_extra) extra_inds = torch.from_numpy(extra_inds).to( assign_result.gt_inds.device).long() sampled_inds = torch.cat([sampled_inds, extra_inds]) elif len(sampled_inds) > num_expected: sampled_inds = self.random_choice(sampled_inds, num_expected) return sampled_inds

Cream/CDARTS/CDARTS_detection/mmdet/core/bbox/samplers/instance_balanced_pos_sampler.py/0

import copy import torch import torch.nn as nn from mmcv.runner import OptimizerHook from .utils import cast_tensor_type from ..utils.dist_utils import allreduce_grads class Fp16OptimizerHook(OptimizerHook): """FP16 optimizer hook. The steps of fp16 optimizer is as follows. 1. Scale the loss value. 2. BP in the fp16 model. 2. Copy gradients from fp16 model to fp32 weights. 3. Update fp32 weights. 4. Copy updated parameters from fp32 weights to fp16 model. Refer to https://arxiv.org/abs/1710.03740 for more details. Args: loss_scale (float): Scale factor multiplied with loss. """ def __init__(self, grad_clip=None, coalesce=True, bucket_size_mb=-1, loss_scale=512., distributed=True): self.grad_clip = grad_clip self.coalesce = coalesce self.bucket_size_mb = bucket_size_mb self.loss_scale = loss_scale self.distributed = distributed def before_run(self, runner): # keep a copy of fp32 weights runner.optimizer.param_groups = copy.deepcopy( runner.optimizer.param_groups) # convert model to fp16 wrap_fp16_model(runner.model) def copy_grads_to_fp32(self, fp16_net, fp32_weights): """Copy gradients from fp16 model to fp32 weight copy.""" for fp32_param, fp16_param in zip(fp32_weights, fp16_net.parameters()): if fp16_param.grad is not None: if fp32_param.grad is None: fp32_param.grad = fp32_param.data.new(fp32_param.size()) fp32_param.grad.copy_(fp16_param.grad) def copy_params_to_fp16(self, fp16_net, fp32_weights): """Copy updated params from fp32 weight copy to fp16 model.""" for fp16_param, fp32_param in zip(fp16_net.parameters(), fp32_weights): fp16_param.data.copy_(fp32_param.data) def after_train_iter(self, runner): # clear grads of last iteration runner.model.zero_grad() runner.optimizer.zero_grad() # scale the loss value scaled_loss = runner.outputs['loss'] * self.loss_scale scaled_loss.backward() # copy fp16 grads in the model to fp32 params in the optimizer fp32_weights = [] for param_group in runner.optimizer.param_groups: fp32_weights += param_group['params'] self.copy_grads_to_fp32(runner.model, fp32_weights) # allreduce grads if self.distributed: allreduce_grads(fp32_weights, self.coalesce, self.bucket_size_mb) # scale the gradients back for param in fp32_weights: if param.grad is not None: param.grad.div_(self.loss_scale) if self.grad_clip is not None: self.clip_grads(fp32_weights) # update fp32 params runner.optimizer.step() # copy fp32 params to the fp16 model self.copy_params_to_fp16(runner.model, fp32_weights) def wrap_fp16_model(model): # convert model to fp16 model.half() # patch the normalization layers to make it work in fp32 mode patch_norm_fp32(model) # set `fp16_enabled` flag for m in model.modules(): if hasattr(m, 'fp16_enabled'): m.fp16_enabled = True def patch_norm_fp32(module): if isinstance(module, (nn.modules.batchnorm._BatchNorm, nn.GroupNorm)): module.float() module.forward = patch_forward_method(module.forward, torch.half, torch.float) for child in module.children(): patch_norm_fp32(child) return module def patch_forward_method(func, src_type, dst_type, convert_output=True): """Patch the forward method of a module. Args: func (callable): The original forward method. src_type (torch.dtype): Type of input arguments to be converted from. dst_type (torch.dtype): Type of input arguments to be converted to. convert_output (bool): Whether to convert the output back to src_type. Returns: callable: The patched forward method. """ def new_forward(*args, **kwargs): output = func(*cast_tensor_type(args, src_type, dst_type), **cast_tensor_type(kwargs, src_type, dst_type)) if convert_output: output = cast_tensor_type(output, dst_type, src_type) return output return new_forward

Cream/CDARTS/CDARTS_detection/mmdet/core/fp16/hooks.py/0

import numpy as np from torch.utils.data.dataset import ConcatDataset as _ConcatDataset from .registry import DATASETS @DATASETS.register_module class ConcatDataset(_ConcatDataset): """A wrapper of concatenated dataset. Same as :obj:`torch.utils.data.dataset.ConcatDataset`, but concat the group flag for image aspect ratio. Args: datasets (list[:obj:`Dataset`]): A list of datasets. """ def __init__(self, datasets): super(ConcatDataset, self).__init__(datasets) self.CLASSES = datasets[0].CLASSES if hasattr(datasets[0], 'flag'): flags = [] for i in range(0, len(datasets)): flags.append(datasets[i].flag) self.flag = np.concatenate(flags) @DATASETS.register_module class RepeatDataset(object): """A wrapper of repeated dataset. The length of repeated dataset will be `times` larger than the original dataset. This is useful when the data loading time is long but the dataset is small. Using RepeatDataset can reduce the data loading time between epochs. Args: dataset (:obj:`Dataset`): The dataset to be repeated. times (int): Repeat times. """ def __init__(self, dataset, times): self.dataset = dataset self.times = times self.CLASSES = dataset.CLASSES if hasattr(self.dataset, 'flag'): self.flag = np.tile(self.dataset.flag, times) self._ori_len = len(self.dataset) def __getitem__(self, idx): return self.dataset[idx % self._ori_len] def __len__(self): return self.times * self._ori_len

Cream/CDARTS/CDARTS_detection/mmdet/datasets/dataset_wrappers.py/0

from .backbones import * # noqa: F401,F403 from .necks import * # noqa: F401,F403 from .roi_extractors import * # noqa: F401,F403 from .anchor_heads import * # noqa: F401,F403 from .shared_heads import * # noqa: F401,F403 from .bbox_heads import * # noqa: F401,F403 from .mask_heads import * # noqa: F401,F403 from .losses import * # noqa: F401,F403 from .detectors import * # noqa: F401,F403 from .registry import (BACKBONES, NECKS, ROI_EXTRACTORS, SHARED_HEADS, HEADS, LOSSES, DETECTORS) from .builder import (build_backbone, build_neck, build_roi_extractor, build_shared_head, build_head, build_loss, build_detector) __all__ = [ 'BACKBONES', 'NECKS', 'ROI_EXTRACTORS', 'SHARED_HEADS', 'HEADS', 'LOSSES', 'DETECTORS', 'build_backbone', 'build_neck', 'build_roi_extractor', 'build_shared_head', 'build_head', 'build_loss', 'build_detector' ]

Cream/CDARTS/CDARTS_detection/mmdet/models/__init__.py/0

import torch import torch.nn as nn import torch.nn.functional as F import math import time import numpy as np from .fbnet_blocks import * from .fbnet_arch import predefine_archs import logging from torch.nn.modules.batchnorm import _BatchNorm from mmcv.cnn import constant_init, kaiming_init from .utils import load_checkpoint from ..registry import BACKBONES @BACKBONES.register_module class FBNet(nn.Module): def __init__(self, arch='fbnet_c', out_indices=(5, 9, 17, 22), frozen_stages=-1): super(FBNet, self).__init__() print('Model is {}.'.format(arch)) self.out_indices = out_indices self.frozen_stages = frozen_stages self.arch = arch self.input_size = 800 self.build_backbone(self.arch, self.input_size) def build_backbone(self, arch, input_size): genotypes = predefine_archs[arch]['genotypes'] strides = predefine_archs[arch]['strides'] out_channels = predefine_archs[arch]['out_channels'] self.layers = nn.ModuleList() self.layers.append(ConvBNReLU(input_size, in_channels=3, out_channels=out_channels[0], kernel_size=3, stride=strides[0], padding=1, bias=True, relu_type='relu', bn_type='bn')) input_size = input_size // strides[0] _in_channels = out_channels[0] for genotype, stride, _out_channels in zip(genotypes[1:], strides[1:], out_channels[1:]): if genotype.endswith('sb'): self.layers.append(SUPER_PRIMITIVES[genotype](input_size, _in_channels, _out_channels, stride)) else: self.layers.append(PRIMITIVES[genotype](input_size, _in_channels, _out_channels, stride)) input_size = input_size // stride _in_channels = _out_channels for m in self.modules(): if isinstance(m, nn.SyncBatchNorm): m._specify_ddp_gpu_num(1) def init_weights(self, pretrained=None): if isinstance(pretrained, str): logger = logging.getLogger() load_checkpoint(self, pretrained, strict=False, logger=logger) elif pretrained is None: for m in self.modules(): if isinstance(m, nn.Conv2d): kaiming_init(m) elif isinstance(m, (_BatchNorm, nn.GroupNorm)): constant_init(m, 1) else: raise TypeError('pretrained must be a str or None') def forward(self, x, alphas=None): outs = [] cnt = 0 for i, layer in enumerate(self.layers): x = layer(x) if i in self.out_indices: outs.append(x) return outs

Cream/CDARTS/CDARTS_detection/mmdet/models/backbones/fbnet.py/0

import numpy as np import torch import torch.nn as nn import torch.nn.functional as F norm_cfg_ = { 'BN': nn.BatchNorm2d, 'SyncBN': nn.SyncBatchNorm, 'GN': nn.GroupNorm, } OPS = { 'skip': lambda input_size, in_channels, out_channels, stride, bn='BN': Identity(input_size, in_channels, out_channels, stride), 'ir_k3_e1': lambda input_size, in_channels, out_channels, stride, bn='BN': MBBlock(input_size, in_channels, out_channels, 1, stride, 3, bn=bn), 'ir_k3_e1_r2': lambda input_size, in_channels, out_channels, stride, bn='BN': MBBlock(input_size, in_channels, out_channels, 1, stride, 3, dilation=2, bn=bn), 'ir_k3_e3': lambda input_size, in_channels, out_channels, stride, bn='BN': MBBlock(input_size, in_channels, out_channels, 3, stride, 3, bn=bn), 'ir_k3_e6': lambda input_size, in_channels, out_channels, stride, bn='BN': MBBlock(input_size, in_channels, out_channels, 6, stride, 3, bn=bn), 'ir_k3_e6_r2': lambda input_size, in_channels, out_channels, stride, bn='BN': MBBlock(input_size, in_channels, out_channels, 6, stride, 3, dilation=2, bn=bn), 'ir_k3_s2': lambda input_size, in_channels, out_channels, stride, bn='BN': MBBlock(input_size, in_channels, out_channels, 1, stride, 3, 2, bn=bn), 'ir_k5_e1': lambda input_size, in_channels, out_channels, stride, bn='BN': MBBlock(input_size, in_channels, out_channels, 1, stride, 5, bn=bn), 'ir_k5_e1_r2': lambda input_size, in_channels, out_channels, stride, bn='BN': MBBlock(input_size, in_channels, out_channels, 1, stride, 5, dilation=2, bn=bn), 'ir_k5_e3': lambda input_size, in_channels, out_channels, stride, bn='BN': MBBlock(input_size, in_channels, out_channels, 3, stride, 5, bn=bn), 'ir_k5_e6': lambda input_size, in_channels, out_channels, stride, bn='BN': MBBlock(input_size, in_channels, out_channels, 6, stride, 5, bn=bn), 'ir_k5_e6_r2': lambda input_size, in_channels, out_channels, stride, bn='BN': MBBlock(input_size, in_channels, out_channels, 6, stride, 5, dilation=2, bn=bn), 'ir_k5_s2': lambda input_size, in_channels, out_channels, stride, bn='BN': MBBlock(input_size, in_channels, out_channels, 1, stride, 5, 2, bn=bn), 'ir_k7_e3': lambda input_size, in_channels, out_channels, stride, bn='BN': MBBlock(input_size, in_channels, out_channels, 3, stride, 7, bn=bn), 'ir_k7_e6': lambda input_size, in_channels, out_channels, stride, bn='BN': MBBlock(input_size, in_channels, out_channels, 6, stride, 7, bn=bn), 'sep_k3' : lambda input_size, in_channels, out_channels, stride, bn='BN': SepConv(input_size, in_channels, out_channels, 1, stride, 3), 'sep_k5' : lambda input_size, in_channels, out_channels, stride, bn='BN': SepConv(input_size, in_channels, out_channels, 1, stride, 5), 'conv1' : lambda input_size, in_channels, out_channels, stride, bn='BN': ConvBNReLU(input_size, in_channels, out_channels, 1, stride, bn_type=bn), 'conv3' : lambda input_size, in_channels, out_channels, stride, bn='BN': ConvBNReLU(input_size, in_channels, out_channels, 3, stride, bn_type=bn), 'conv5' : lambda input_size, in_channels, out_channels, stride, bn='BN': ConvBNReLU(input_size, in_channels, out_channels, 5, stride, bn_type=bn), 'avgpool': lambda input_size, in_channels, out_channels, stride, bn='BN': AvgPool(input_size, in_channels, stride), } class AvgPool(nn.Module): def __init__(self, stride): super(AvgPool, self).__init__() self.stride = stride def forward(self, x): return F.avg_pool2d(x, self.stride) class ChannelShuffle(nn.Module): def __init__(self, groups=1): super(ChannelShuffle, self).__init__() self.groups = groups def forward(self, x): if self.groups == 1: return x N, C, H, W = x.size() cpg = C // self.groups # channels per group out = x.view(N, self.groups, cpg, H, W) out = out.permute(0, 2, 1, 3, 4).contiguous() out = out.view(N, C, H, W) return out class ConvBNReLU(nn.Module): def __init__(self, input_size, in_channels, out_channels, kernel_size, stride, dilation=1, bias=False, relu_type='relu', bn_type='BN', groups=1): super(ConvBNReLU, self).__init__() assert(relu_type in ['relu', 'none']) padding = (kernel_size - 1) * dilation // 2 if bn_type == 'none': bias = True self.conv = nn.Conv2d(in_channels, out_channels, kernel_size, stride=stride, padding=padding, dilation=dilation, bias=bias, groups=groups) nn.init.kaiming_normal_(self.conv.weight, mode="fan_out", nonlinearity="relu") if self.conv.bias is not None: nn.init.constant_(self.conv.bias, 0.0) if bn_type == 'none' : self.bn = nn.Sequential() elif bn_type == 'GN': norm_layer = norm_cfg_[bn_type] self.bn = norm_layer(num_channels=out_channels, num_groups=32) else: norm_layer = norm_cfg_[bn_type] self.bn = norm_layer(out_channels) self.relu = nn.ReLU(inplace=True) if relu_type == 'relu' else nn.Sequential() def forward(self, x): out = self.conv(x) out = self.relu(self.bn(out)) return out class SE(nn.Module): def __init__(self, input_size, in_channels, se_ratio): super(SE, self).__init__() self.in_channels, self.se_ratio = in_channels, se_ratio self.pooling = nn.AdaptiveAvgPool2d((1, 1)) self.fc1 = nn.Conv2d(in_channels, max(1, int(in_channels * se_ratio)), 1, bias=False) self.fc2 = nn.Conv2d(max(1, int(in_channels * se_ratio)), in_channels, 1, bias=False) def forward(self, x): out = self.pooling(x) out = self.fc1(out) out = F.relu(out) out = self.fc2(out) out = F.sigmoid(out) return out class Identity(nn.Module): def __init__(self, input_size, in_channels, out_channels, stride): super(Identity, self).__init__() if in_channels != out_channels or stride != 1: self.conv = ConvBNReLU(input_size, in_channels, out_channels, kernel_size=1, stride=stride, padding=0, bias=False, relu_type='relu', bn_type='bn') else: self.conv = nn.Sequential() def forward(self, x): return self.conv(x) class SepConv(nn.Module): def __init__(self, input_size, in_channels, out_channels, expansion, stride, kernel_size, groups=1, bn_type='BN'): super(SepConv, self).__init__() self.conv1 = ConvBNReLU(input_size, in_channels, in_channels, kernel_size=kernel_size, stride=stride, bias=False, relu_type='relu', bn_type=bn_type, groups=in_channels) self.conv2 = ConvBNReLU(input_size//stride, in_channels, out_channels, kernel_size=1, stride=1, bias=False, relu_type='none', bn_type=bn_type, groups=groups) def forward(self, x): out = self.conv1(x) out = self.conv2(out) return out class MBBlock(nn.Module): def __init__(self, input_size, in_channels, out_channels, expansion, stride, kernel_size, dilation=1, groups=1, has_se=False, bn='BN'): super(MBBlock, self).__init__() self.in_channels = in_channels self.out_channels =out_channels self.has_se = has_se self.stride = stride self.groups = groups mid_channels = in_channels * expansion self.conv1 = ConvBNReLU(input_size, in_channels, mid_channels, kernel_size=1, stride=1, dilation=1, bias=False, relu_type='relu', bn_type=bn, groups=groups) self.conv2 = ConvBNReLU(input_size, mid_channels, mid_channels, kernel_size=kernel_size, stride=stride, dilation=dilation, bias=False, relu_type='relu', bn_type=bn, groups=mid_channels) self.conv3 = ConvBNReLU(input_size//self.stride, mid_channels, out_channels, kernel_size=1, stride=1, dilation=1, bias=False, relu_type='none', bn_type=bn, groups=groups) if has_se == True: self.se = SE(input_size, mid_channels, se_ratio=0.05) if groups != 1: self.shuffle = ChannelShuffle(input_size, in_channels, groups) def forward(self, x): out = self.conv1(x) if self.groups != 1: out = self.shuffle(out) out = self.conv2(out) if self.has_se: out = out * self.se(out) out = self.conv3(out) if self.in_channels == self.out_channels and self.stride == 1: out = out + x return out

Cream/CDARTS/CDARTS_detection/mmdet/models/bbox_heads/auto_head/mbblock_ops.py/0

from .single_stage import SingleStageDetector from ..registry import DETECTORS @DETECTORS.register_module class RetinaNet(SingleStageDetector): def __init__(self, backbone, neck, bbox_head, train_cfg=None, test_cfg=None, pretrained=None): super(RetinaNet, self).__init__(backbone, neck, bbox_head, train_cfg, test_cfg, pretrained)

Cream/CDARTS/CDARTS_detection/mmdet/models/detectors/retinanet.py/0

import mmcv import numpy as np import pycocotools.mask as mask_util import torch import torch.nn as nn from ..builder import build_loss from ..registry import HEADS from ..utils import ConvModule from mmdet.core import mask_target, force_fp32, auto_fp16 @HEADS.register_module class FCNMaskHead(nn.Module): def __init__(self, num_convs=4, roi_feat_size=14, in_channels=256, conv_kernel_size=3, conv_out_channels=256, upsample_method='deconv', upsample_ratio=2, num_classes=81, class_agnostic=False, conv_cfg=None, norm_cfg=None, loss_mask=dict( type='CrossEntropyLoss', use_mask=True, loss_weight=1.0)): super(FCNMaskHead, self).__init__() if upsample_method not in [None, 'deconv', 'nearest', 'bilinear']: raise ValueError( 'Invalid upsample method {}, accepted methods ' 'are "deconv", "nearest", "bilinear"'.format(upsample_method)) self.num_convs = num_convs self.roi_feat_size = roi_feat_size # WARN: not used and reserved self.in_channels = in_channels self.conv_kernel_size = conv_kernel_size self.conv_out_channels = conv_out_channels self.upsample_method = upsample_method self.upsample_ratio = upsample_ratio self.num_classes = num_classes self.class_agnostic = class_agnostic self.conv_cfg = conv_cfg self.norm_cfg = norm_cfg self.fp16_enabled = False self.loss_mask = build_loss(loss_mask) self.convs = nn.ModuleList() for i in range(self.num_convs): in_channels = ( self.in_channels if i == 0 else self.conv_out_channels) padding = (self.conv_kernel_size - 1) // 2 self.convs.append( ConvModule( in_channels, self.conv_out_channels, self.conv_kernel_size, padding=padding, conv_cfg=conv_cfg, norm_cfg=norm_cfg)) upsample_in_channels = ( self.conv_out_channels if self.num_convs > 0 else in_channels) if self.upsample_method is None: self.upsample = None elif self.upsample_method == 'deconv': self.upsample = nn.ConvTranspose2d( upsample_in_channels, self.conv_out_channels, self.upsample_ratio, stride=self.upsample_ratio) else: self.upsample = nn.Upsample( scale_factor=self.upsample_ratio, mode=self.upsample_method) out_channels = 1 if self.class_agnostic else self.num_classes logits_in_channel = ( self.conv_out_channels if self.upsample_method == 'deconv' else upsample_in_channels) self.conv_logits = nn.Conv2d(logits_in_channel, out_channels, 1) self.relu = nn.ReLU(inplace=True) self.debug_imgs = None def init_weights(self): for m in [self.upsample, self.conv_logits]: if m is None: continue nn.init.kaiming_normal_( m.weight, mode='fan_out', nonlinearity='relu') nn.init.constant_(m.bias, 0) @auto_fp16() def forward(self, x): for conv in self.convs: x = conv(x) if self.upsample is not None: x = self.upsample(x) if self.upsample_method == 'deconv': x = self.relu(x) mask_pred = self.conv_logits(x) return mask_pred def get_target(self, sampling_results, gt_masks, rcnn_train_cfg): pos_proposals = [res.pos_bboxes for res in sampling_results] pos_assigned_gt_inds = [ res.pos_assigned_gt_inds for res in sampling_results ] mask_targets = mask_target(pos_proposals, pos_assigned_gt_inds, gt_masks, rcnn_train_cfg) return mask_targets @force_fp32(apply_to=('mask_pred', )) def loss(self, mask_pred, mask_targets, labels): loss = dict() if self.class_agnostic: loss_mask = self.loss_mask(mask_pred, mask_targets, torch.zeros_like(labels)) else: loss_mask = self.loss_mask(mask_pred, mask_targets, labels) loss['loss_mask'] = loss_mask return loss def get_seg_masks(self, mask_pred, det_bboxes, det_labels, rcnn_test_cfg, ori_shape, scale_factor, rescale): """Get segmentation masks from mask_pred and bboxes. Args: mask_pred (Tensor or ndarray): shape (n, #class+1, h, w). For single-scale testing, mask_pred is the direct output of model, whose type is Tensor, while for multi-scale testing, it will be converted to numpy array outside of this method. det_bboxes (Tensor): shape (n, 4/5) det_labels (Tensor): shape (n, ) img_shape (Tensor): shape (3, ) rcnn_test_cfg (dict): rcnn testing config ori_shape: original image size Returns: list[list]: encoded masks """ if isinstance(mask_pred, torch.Tensor): mask_pred = mask_pred.sigmoid().cpu().numpy() assert isinstance(mask_pred, np.ndarray) # when enabling mixed precision training, mask_pred may be float16 # numpy array mask_pred = mask_pred.astype(np.float32) cls_segms = [[] for _ in range(self.num_classes - 1)] bboxes = det_bboxes.cpu().numpy()[:, :4] labels = det_labels.cpu().numpy() + 1 if rescale: img_h, img_w = ori_shape[:2] else: img_h = np.round(ori_shape[0] * scale_factor).astype(np.int32) img_w = np.round(ori_shape[1] * scale_factor).astype(np.int32) scale_factor = 1.0 for i in range(bboxes.shape[0]): bbox = (bboxes[i, :] / scale_factor).astype(np.int32) label = labels[i] w = max(bbox[2] - bbox[0] + 1, 1) h = max(bbox[3] - bbox[1] + 1, 1) if not self.class_agnostic: mask_pred_ = mask_pred[i, label, :, :] else: mask_pred_ = mask_pred[i, 0, :, :] im_mask = np.zeros((img_h, img_w), dtype=np.uint8) bbox_mask = mmcv.imresize(mask_pred_, (w, h)) bbox_mask = (bbox_mask > rcnn_test_cfg.mask_thr_binary).astype( np.uint8) im_mask[bbox[1]:bbox[1] + h, bbox[0]:bbox[0] + w] = bbox_mask rle = mask_util.encode( np.array(im_mask[:, :, np.newaxis], order='F'))[0] cls_segms[label - 1].append(rle) return cls_segms

Cream/CDARTS/CDARTS_detection/mmdet/models/mask_heads/fcn_mask_head.py/0

from .non_local import NonLocal2D from .generalized_attention import GeneralizedAttention __all__ = ['NonLocal2D', 'GeneralizedAttention']

Cream/CDARTS/CDARTS_detection/mmdet/models/plugins/__init__.py/0

from .functions.deform_conv import deform_conv, modulated_deform_conv from .functions.deform_pool import deform_roi_pooling from .modules.deform_conv import (DeformConv, ModulatedDeformConv, DeformConvPack, ModulatedDeformConvPack) from .modules.deform_pool import (DeformRoIPooling, DeformRoIPoolingPack, ModulatedDeformRoIPoolingPack) __all__ = [ 'DeformConv', 'DeformConvPack', 'ModulatedDeformConv', 'ModulatedDeformConvPack', 'DeformRoIPooling', 'DeformRoIPoolingPack', 'ModulatedDeformRoIPoolingPack', 'deform_conv', 'modulated_deform_conv', 'deform_roi_pooling' ]

Cream/CDARTS/CDARTS_detection/mmdet/ops/dcn/__init__.py/0

import math import torch from torch.autograd import Function from torch.nn.modules.utils import _pair from .. import masked_conv2d_cuda class MaskedConv2dFunction(Function): @staticmethod def forward(ctx, features, mask, weight, bias, padding=0, stride=1): assert mask.dim() == 3 and mask.size(0) == 1 assert features.dim() == 4 and features.size(0) == 1 assert features.size()[2:] == mask.size()[1:] pad_h, pad_w = _pair(padding) stride_h, stride_w = _pair(stride) if stride_h != 1 or stride_w != 1: raise ValueError( 'Stride could not only be 1 in masked_conv2d currently.') if not features.is_cuda: raise NotImplementedError out_channel, in_channel, kernel_h, kernel_w = weight.size() batch_size = features.size(0) out_h = int( math.floor((features.size(2) + 2 * pad_h - (kernel_h - 1) - 1) / stride_h + 1)) out_w = int( math.floor((features.size(3) + 2 * pad_w - (kernel_h - 1) - 1) / stride_w + 1)) mask_inds = torch.nonzero(mask[0] > 0) output = features.new_zeros(batch_size, out_channel, out_h, out_w) if mask_inds.numel() > 0: mask_h_idx = mask_inds[:, 0].contiguous() mask_w_idx = mask_inds[:, 1].contiguous() data_col = features.new_zeros(in_channel * kernel_h * kernel_w, mask_inds.size(0)) masked_conv2d_cuda.masked_im2col_forward(features, mask_h_idx, mask_w_idx, kernel_h, kernel_w, pad_h, pad_w, data_col) masked_output = torch.addmm(1, bias[:, None], 1, weight.view(out_channel, -1), data_col) masked_conv2d_cuda.masked_col2im_forward(masked_output, mask_h_idx, mask_w_idx, out_h, out_w, out_channel, output) return output @staticmethod def backward(ctx, grad_output): return (None, ) * 5 masked_conv2d = MaskedConv2dFunction.apply

Cream/CDARTS/CDARTS_detection/mmdet/ops/masked_conv/functions/masked_conv.py/0

from torch.autograd import Function from .. import roi_align_cuda class RoIAlignFunction(Function): @staticmethod def forward(ctx, features, rois, out_size, spatial_scale, sample_num=0): if isinstance(out_size, int): out_h = out_size out_w = out_size elif isinstance(out_size, tuple): assert len(out_size) == 2 assert isinstance(out_size[0], int) assert isinstance(out_size[1], int) out_h, out_w = out_size else: raise TypeError( '"out_size" must be an integer or tuple of integers') ctx.spatial_scale = spatial_scale ctx.sample_num = sample_num ctx.save_for_backward(rois) ctx.feature_size = features.size() batch_size, num_channels, data_height, data_width = features.size() num_rois = rois.size(0) output = features.new_zeros(num_rois, num_channels, out_h, out_w) if features.is_cuda: roi_align_cuda.forward(features, rois, out_h, out_w, spatial_scale, sample_num, output) else: raise NotImplementedError return output @staticmethod def backward(ctx, grad_output): feature_size = ctx.feature_size spatial_scale = ctx.spatial_scale sample_num = ctx.sample_num rois = ctx.saved_tensors[0] assert (feature_size is not None and grad_output.is_cuda) batch_size, num_channels, data_height, data_width = feature_size out_w = grad_output.size(3) out_h = grad_output.size(2) grad_input = grad_rois = None if ctx.needs_input_grad[0]: grad_input = rois.new_zeros(batch_size, num_channels, data_height, data_width) roi_align_cuda.backward(grad_output.contiguous(), rois, out_h, out_w, spatial_scale, sample_num, grad_input) return grad_input, grad_rois, None, None, None roi_align = RoIAlignFunction.apply

Cream/CDARTS/CDARTS_detection/mmdet/ops/roi_align/functions/roi_align.py/0