bf16: true cutoff_len: 2048 dataset: train_hal_detection_1125 dataset_dir: data ddp_timeout: 180000000 deepspeed: cache/ds_z2_offload_config.json do_train: true finetuning_type: full flash_attn: auto gradient_accumulation_steps: 1 include_num_input_tokens_seen: true learning_rate: 5.0e-06 logging_steps: 1 lr_scheduler_type: cosine max_grad_norm: 1.0 max_samples: 100000 model_name_or_path: Qwen/Qwen2.5-7B-Instruct num_train_epochs: 3.0 optim: adamw_torch output_dir: saves/Qwen2.5-7B-Instruct/full/train_2025-02-07-00-42-22_qwen packing: false per_device_train_batch_size: 1 plot_loss: true preprocessing_num_workers: 16 report_to: none save_steps: 5000 stage: sft template: qwen trust_remote_code: true warmup_steps: 10