| base_model: stabilityai/stablelm-base-alpha-3b | |
| load_in_8bit: true | |
| datasets: | |
| - path: vicgalle/alpaca-gpt4 | |
| type: alpaca | |
| dataset_prepared_path: last_run_prepared | |
| val_set_size: 0.04 | |
| adapter: | |
| lora_model_dir: | |
| sequence_len: 4096 | |
| lora_r: 8 | |
| lora_alpha: 16 | |
| lora_dropout: 0.05 | |
| lora_target_modules: | |
| - q_proj | |
| - v_proj | |
| lora_fan_in_fan_out: false | |
| wandb_project: stable-llama-3b | |
| wandb_watch: | |
| wandb_run_id: | |
| wandb_log_model: checkpoint | |
| output_dir: ./stable-llama-3b | |
| batch_size: 128 | |
| micro_batch_size: 16 | |
| num_epochs: 1 | |
| learning_rate: 0.00003 | |
| train_on_inputs: false | |
| group_by_length: false | |
| bf16: true | |
| tf32: true | |
| early_stopping_patience: 3 | |
| resume_from_checkpoint: | |
| local_rank: | |