See axolotl config

axolotl version: 0.4.1

yaml

1adapter: qlora 2auto_resume_from_checkpoints: true 3base_model: heegyu/WizardVicuna2-13b-hf 4bf16: auto 5chat_template: llama3 6dataloader_num_workers: 12 7dataset_prepared_path: null 8datasets: 9- data_files: 10 - afb1fdb32bdf02c6_train_data.json 11 ds_type: json 12 format: custom 13 path: /workspace/input_data/afb1fdb32bdf02c6_train_data.json 14 type: 15 field_instruction: premise 16 field_output: hypothesis 17 format: '{instruction}' 18 no_input_format: '{instruction}' 19 system_format: '{system}' 20 system_prompt: '' 21debug: null 22deepspeed: null 23early_stopping_patience: 3 24eval_max_new_tokens: 128 25eval_steps: 100 26eval_table_size: null 27evals_per_epoch: null 28flash_attention: false 29fp16: false 30fsdp: null 31fsdp_config: null 32gradient_accumulation_steps: 4 33gradient_checkpointing: true 34group_by_length: true 35hub_model_id: error577/d1526c12-8a5c-45c2-8a7c-e005e4428b34 36hub_repo: null 37hub_strategy: checkpoint 38hub_token: null 39learning_rate: 0.0002 40load_in_4bit: true 41load_in_8bit: false 42local_rank: null 43logging_steps: 10 44lora_alpha: 128 45lora_dropout: 0.05 46lora_fan_in_fan_out: null 47lora_model_dir: null 48lora_r: 32 49lora_target_linear: true 50lr_scheduler: cosine 51max_grad_norm: 1.0 52max_steps: null 53micro_batch_size: 1 54mlflow_experiment_name: /tmp/afb1fdb32bdf02c6_train_data.json 55model_type: AutoModelForCausalLM 56num_epochs: 3 57optimizer: adamw_torch_4bit 58output_dir: miner_id_24 59pad_to_sequence_len: true 60resume_from_checkpoint: null 61s2_attention: null 62sample_packing: false 63save_steps: 100 64sequence_len: 512 65special_tokens: 66 pad_token: </s> 67strict: false 68tf32: false 69tokenizer_type: AutoTokenizer 70train_on_inputs: false 71trust_remote_code: true 72val_set_size: 0.005 73wandb_entity: null 74wandb_mode: online 75wandb_name: d8eda0ee-8aeb-47d4-bebc-dd3dba382021 76wandb_project: Gradients-On-Demand 77wandb_run: your_name 78wandb_runid: d8eda0ee-8aeb-47d4-bebc-dd3dba382021 79warmup_steps: 30 80weight_decay: 0.0 81xformers_attention: null 82

d1526c12-8a5c-45c2-8a7c-e005e4428b34

• Loss: 2.5516

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

• learning_rate: 0.0002
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 4
• optimizer: Use OptimizerNames.ADAMW_TORCH_4BIT 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: 30
• num_epochs: 3

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