See axolotl config

axolotl version: 0.4.1

yaml

1adapter: lora 2auto_find_batch_size: false 3base_model: TinyLlama/TinyLlama-1.1B-Chat-v0.6 4bf16: auto 5chat_template: llama3 6dataloader_num_workers: 12 7dataset_prepared_path: null 8datasets: 9- data_files: 10 - 5098d357e8e641a8_train_data.json 11 ds_type: json 12 format: custom 13 path: /workspace/input_data/5098d357e8e641a8_train_data.json 14 type: 15 field_instruction: prompt 16 field_output: chosen 17 format: '{instruction}' 18 no_input_format: '{instruction}' 19 system_format: '{system}' 20 system_prompt: '' 21debug: null 22deepspeed: null 23early_stopping_patience: 2 24early_stopping_threshold: 1.0e-05 25eval_max_new_tokens: 128 26eval_steps: 40 27eval_strategy: null 28flash_attention: true 29fp16: null 30fsdp: null 31fsdp_config: null 32gradient_accumulation_steps: 8 33gradient_checkpointing: false 34group_by_length: false 35hub_model_id: mrferr3t/fab7c3df-bd15-4208-9d66-373061379161 36hub_repo: null 37hub_strategy: checkpoint 38hub_token: null 39learning_rate: 0.0004 40load_in_4bit: false 41load_in_8bit: false 42local_rank: null 43logging_steps: 40 44lora_alpha: 16 45lora_dropout: 0.05 46lora_fan_in_fan_out: null 47lora_model_dir: null 48lora_r: 8 49lora_target_linear: true 50lr_scheduler: cosine 51max_steps: null 52micro_batch_size: 16 53mlflow_experiment_name: /tmp/5098d357e8e641a8_train_data.json 54model_type: AutoModelForCausalLM 55num_epochs: 100 56optimizer: adamw_bnb_8bit 57output_dir: miner_id_24 58pad_to_sequence_len: true 59resume_from_checkpoint: null 60s2_attention: null 61sample_packing: false 62save_steps: 40 63saves_per_epoch: 0 64sequence_len: 512 65strict: false 66tf32: false 67tokenizer_type: AutoTokenizer 68train_on_inputs: false 69trust_remote_code: true 70val_set_size: .05000000 71wandb_entity: null 72wandb_mode: disabled 73wandb_name: 358492dc-bec6-4426-a767-ecbc452dc239 74wandb_project: Gradients-On-Demand 75wandb_run: your_name 76wandb_runid: 358492dc-bec6-4426-a767-ecbc452dc239 77warmup_steps: 100 78weight_decay: 0.0 79xformers_attention: null 80

fab7c3df-bd15-4208-9d66-373061379161

• Loss: 1.1762

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

• learning_rate: 0.0004
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• num_epochs: 100

Training results

Framework versions

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