See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: NousResearch/Meta-Llama-3-8B-Alternate-Tokenizer
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 950bcce6cc9a1d4d_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/950bcce6cc9a1d4d_train_data.json
  type:
    field_instruction: darija
    field_output: eng
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: null
eval_max_new_tokens: 256
eval_table_size: null
evals_per_epoch: 4
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: mamung/6e88ab70-6a8d-4b8c-b0b3-77184d17a20d
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.00015
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 5
lora_alpha: 128
lora_dropout: 0.1
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 64
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
- o_proj
- gate_proj
- down_proj
- up_proj
lr_scheduler: cosine
max_steps: 100
micro_batch_size: 8
mlflow_experiment_name: /tmp/950bcce6cc9a1d4d_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optim_args:
  adam_beta1: 0.9
  adam_beta2: 0.95
  adam_epsilon: 2.0e-05
optimizer: adamw_torch
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
saves_per_epoch: 4
sequence_len: 1024
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.1
wandb_entity: eddysang
wandb_mode: online
wandb_name: d4e86d7e-2b2e-45b9-bf98-c0da59e845ed
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: d4e86d7e-2b2e-45b9-bf98-c0da59e845ed
warmup_steps: 20
weight_decay: 0.02
xformers_attention: false

6e88ab70-6a8d-4b8c-b0b3-77184d17a20d

This model is a fine-tuned version of NousResearch/Meta-Llama-3-8B-Alternate-Tokenizer on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.3061

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.00015
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=adam_beta1=0.9,adam_beta2=0.95,adam_epsilon=2e-05
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 20
• training_steps: 100

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0008	1	8.7248
8.547	0.0071	9	6.6192
5.2814	0.0141	18	4.4651
4.2927	0.0212	27	4.0005
3.8531	0.0283	36	3.8258
3.6351	0.0353	45	3.6262
3.6745	0.0424	54	3.5406
3.2906	0.0495	63	3.4501
3.4344	0.0565	72	3.3794
3.4602	0.0636	81	3.3322
3.0172	0.0707	90	3.3157
3.4561	0.0778	99	3.3061

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.5.0+cu124
• Datasets 3.0.1
• Tokenizers 0.20.1