litellm/llms/petals.py

import os, types
import json
from enum import Enum
import requests
import time
from typing import Callable, Optional
import litellm
from litellm.utils import ModelResponse, Usage
from .prompt_templates.factory import prompt_factory, custom_prompt

class PetalsError(Exception):
    def __init__(self, status_code, message):
        self.status_code = status_code
        self.message = message
        super().__init__(
            self.message
        )  # Call the base class constructor with the parameters it needs

class PetalsConfig():
    """
    Reference: https://github.com/petals-infra/chat.petals.dev#post-apiv1generate
    The `PetalsConfig` class encapsulates the configuration for the Petals API. The properties of this class are described below:

    - `max_length` (integer): This represents the maximum length of the generated text (including the prefix) in tokens.

    - `max_new_tokens` (integer): This represents the maximum number of newly generated tokens (excluding the prefix).

    The generation parameters are compatible with `.generate()` from Hugging Face's Transformers library:

    - `do_sample` (boolean, optional): If set to 0 (default), the API runs greedy generation. If set to 1, the API performs sampling using the parameters below:

    - `temperature` (float, optional): This value sets the temperature for sampling.
    
    - `top_k` (integer, optional): This value sets the limit for top-k sampling.
    
    - `top_p` (float, optional): This value sets the limit for top-p (nucleus) sampling.
    
    - `repetition_penalty` (float, optional): This helps apply the repetition penalty during text generation, as discussed in this paper.
    """
    max_length: Optional[int]=None
    max_new_tokens: Optional[int]=litellm.max_tokens # petals requires max tokens to be set
    do_sample: Optional[bool]=None
    temperature: Optional[float]=None
    top_k: Optional[int]=None
    top_p: Optional[float]=None
    repetition_penalty: Optional[float]=None

    def __init__(self,
                 max_length: Optional[int]=None,
                 max_new_tokens: Optional[int]=litellm.max_tokens, # petals requires max tokens to be set
                 do_sample: Optional[bool]=None,
                 temperature: Optional[float]=None,
                 top_k: Optional[int]=None,
                 top_p: Optional[float]=None,
                 repetition_penalty: Optional[float]=None) -> None:
        locals_ = locals()
        for key, value in locals_.items():
            if key != 'self' and value is not None:
                setattr(self.__class__, key, value)
    
    @classmethod
    def get_config(cls):
        return {k: v for k, v in cls.__dict__.items() 
                if not k.startswith('__') 
                and not isinstance(v, (types.FunctionType, types.BuiltinFunctionType, classmethod, staticmethod)) 
                and v is not None}
    
def completion(
    model: str,
    messages: list,
    api_base: Optional[str], 
    model_response: ModelResponse,
    print_verbose: Callable,
    encoding,
    logging_obj,
    optional_params=None,
    stream=False,
    litellm_params=None,
    logger_fn=None,
):
    ## Load Config
    config = litellm.PetalsConfig.get_config()
    for k, v in config.items(): 
        if k not in optional_params: # completion(top_k=3) > petals_config(top_k=3) <- allows for dynamic variables to be passed in
            optional_params[k] = v

    if model in litellm.custom_prompt_dict:
        # check if the model has a registered custom prompt
        model_prompt_details = litellm.custom_prompt_dict[model]
        prompt = custom_prompt(
            role_dict=model_prompt_details["roles"], 
            initial_prompt_value=model_prompt_details["initial_prompt_value"],  
            final_prompt_value=model_prompt_details["final_prompt_value"], 
            messages=messages
        )
    else:
        prompt = prompt_factory(model=model, messages=messages)

    if api_base: 
        ## LOGGING
        logging_obj.pre_call(
                input=prompt,
                api_key="",
                additional_args={"complete_input_dict": optional_params, "api_base": api_base},
            )
        data = {
            "model": model,
            "inputs": prompt,
            **optional_params
        }
        
        ## COMPLETION CALL
        response = requests.post(api_base, data=data)
        
        ## LOGGING
        logging_obj.post_call(
                input=prompt,
                api_key="",
                original_response=response.text,
                additional_args={"complete_input_dict": optional_params},
            )
        
        ## RESPONSE OBJECT
        try:
            output_text = response.json()["outputs"]
        except Exception as e:
            PetalsError(status_code=response.status_code, message=str(e))

    else: 
        try:
            import torch
            from transformers import AutoTokenizer
            from petals import AutoDistributedModelForCausalLM # type: ignore
        except:
            raise Exception(
                "Importing torch, transformers, petals failed\nTry pip installing petals \npip install git+https://github.com/bigscience-workshop/petals"
            )
        
        model = model

        tokenizer = AutoTokenizer.from_pretrained(model, use_fast=False, add_bos_token=False)
        model_obj = AutoDistributedModelForCausalLM.from_pretrained(model)

        ## LOGGING
        logging_obj.pre_call(
                input=prompt,
                api_key="",
                additional_args={"complete_input_dict": optional_params},
            )
        
        ## COMPLETION CALL
        inputs = tokenizer(prompt, return_tensors="pt")["input_ids"]
        
        # optional params: max_new_tokens=1,temperature=0.9, top_p=0.6
        outputs = model_obj.generate(inputs, **optional_params)

        ## LOGGING
        logging_obj.post_call(
                input=prompt,
                api_key="",
                original_response=outputs,
                additional_args={"complete_input_dict": optional_params},
            )
        ## RESPONSE OBJECT
        output_text = tokenizer.decode(outputs[0])
    
    if len(output_text) > 0:
        model_response["choices"][0]["message"]["content"] = output_text

    prompt_tokens = len(
        encoding.encode(prompt)
    ) 
    completion_tokens = len(
        encoding.encode(model_response["choices"][0]["message"].get("content"))
    )

    model_response["created"] = int(time.time())
    model_response["model"] = model
    usage = Usage(
            prompt_tokens=prompt_tokens,
            completion_tokens=completion_tokens,
            total_tokens=prompt_tokens + completion_tokens
        )
    model_response.usage = usage
    return model_response

def embedding():
    # logic for parsing in - calling - parsing out model embedding calls
    pass