See axolotl config

axolotl version: 0.4.1

yaml

1adapter: lora 2base_model: DeepMount00/Llama-3-8b-Ita 3bf16: true 4chat_template: llama3 5dataloader_num_workers: 24 6dataset_prepared_path: null 7datasets: 8- data_files: 9 - d61c5229482d7018_train_data.json 10 ds_type: json 11 format: custom 12 path: /workspace/input_data/d61c5229482d7018_train_data.json 13 type: 14 field_input: es 15 field_instruction: de 16 field_output: en 17 format: '{instruction} {input}' 18 no_input_format: '{instruction}' 19 system_format: '{system}' 20 system_prompt: '' 21debug: null 22deepspeed: null 23device_map: auto 24do_eval: true 25early_stopping_patience: 3 26eval_batch_size: 2 27eval_max_new_tokens: 128 28eval_steps: 150 29eval_table_size: null 30evals_per_epoch: null 31flash_attention: true 32fp16: false 33fsdp: null 34fsdp_config: null 35gradient_accumulation_steps: 8 36gradient_checkpointing: false 37group_by_length: true 38hub_model_id: abaddon182/f4d7c437-2c2f-4c91-8887-1016f684962d 39hub_repo: null 40hub_strategy: checkpoint 41hub_token: null 42learning_rate: 0.0002 43load_in_4bit: false 44load_in_8bit: false 45local_rank: null 46logging_steps: 50 47lora_alpha: 64 48lora_dropout: 0.05 49lora_fan_in_fan_out: null 50lora_model_dir: null 51lora_r: 32 52lora_target_linear: true 53lr_scheduler: cosine 54max_grad_norm: 1.0 55max_steps: 3000 56micro_batch_size: 2 57mlflow_experiment_name: /tmp/d61c5229482d7018_train_data.json 58model_type: AutoModelForCausalLM 59num_epochs: 10 60optim_args: 61 adam_beta1: 0.9 62 adam_beta2: 0.999 63 adam_epsilon: 1e-8 64optimizer: adamw_torch_fused 65output_dir: miner_id_24 66pad_to_sequence_len: true 67resume_from_checkpoint: null 68s2_attention: null 69sample_packing: false 70save_steps: 150 71saves_per_epoch: null 72sequence_len: 512 73special_tokens: 74 pad_token: <|eot_id|> 75strict: false 76tf32: true 77tokenizer_type: AutoTokenizer 78train_on_inputs: false 79trust_remote_code: true 80val_set_size: 0.05 81wandb_entity: null 82wandb_mode: online 83wandb_name: 36783109-ff18-4e86-96fe-41347eabea38 84wandb_project: Gradients-On-Demand 85wandb_run: your_name 86wandb_runid: 36783109-ff18-4e86-96fe-41347eabea38 87warmup_steps: 50 88weight_decay: 0.0 89xformers_attention: null 90

f4d7c437-2c2f-4c91-8887-1016f684962d

• Loss: 0.3622

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

• learning_rate: 0.0002
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=adam_beta1=0.9,adam_beta2=0.999,adam_epsilon=1e-8
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 50
• training_steps: 3000

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