| bf16: true | |
| cutoff_len: 2048 | |
| dataset: demo_data | |
| dataset_dir: data | |
| ddp_timeout: 180000000 | |
| do_train: true | |
| double_quantization: true | |
| eval_steps: 20 | |
| eval_strategy: steps | |
| finetuning_type: lora | |
| flash_attn: auto | |
| gradient_accumulation_steps: 8 | |
| include_num_input_tokens_seen: true | |
| learning_rate: 5.0e-05 | |
| logging_steps: 5 | |
| lora_alpha: 16 | |
| lora_dropout: 0.05 | |
| lora_rank: 8 | |
| lora_target: all | |
| lr_scheduler_type: cosine | |
| max_grad_norm: 1.0 | |
| max_samples: 100000 | |
| model_name_or_path: microsoft/Phi-3-medium-4k-instruct | |
| num_train_epochs: 10.0 | |
| optim: adamw_torch | |
| output_dir: saves/Phi-3-14B-8k-Instruct/lora/train_2025-04-09-16-08 | |
| packing: false | |
| per_device_eval_batch_size: 2 | |
| per_device_train_batch_size: 2 | |
| plot_loss: true | |
| preprocessing_num_workers: 16 | |
| quantization_bit: 8 | |
| quantization_method: bnb | |
| report_to: | |
| - tensorboard | |
| save_steps: 20 | |
| stage: sft | |
| template: phi | |
| trust_remote_code: true | |
| use_rslora: true | |
| val_size: 0.115 | |
| warmup_steps: 0 | |