{ "model_name": "Qwen/Qwen3-VL-8B-Thinking", "data_path": "workspace/experiments/11-11-2025", "output_dir": "workspace/experiments/11-11-2025/anonymous-512", "max_length": 2048, "batch_size": 4, "gradient_accumulation_steps": 8, "learning_rate": 0.0002, "num_epochs": 10, "warmup_ratio": 0.1, "save_steps": 2, "eval_steps": 2, "logging_steps": 1, "use_lora": true, "lora_r": 16, "lora_alpha": 32, "lora_dropout": 0.1, "early_stopping_patience": 3, "early_stopping_threshold": 0.001 }