{ "config": { "base_model": "Qwen/Qwen2.5-VL-7B-Instruct", "batch_size": 1, "bf16": true, "clip_window_s": 30.0, "eval_fraction": 0.0, "fps": 2.0, "grad_accum": 8, "gt_path": "annotations/", "learning_rate": 1e-05, "logging_steps": 10, "lora_alpha": 32, "lora_dropout": 0.05, "lora_r": 16, "max_frames": 16, "max_pixels": 151200, "min_step_s": 12.0, "neg_margin_s": 8.0, "neg_per_step": 4, "num_epochs": 1.0, "output_dir": "/opt/dlami/nvme/m4-adapters/pooled-r0", "seed": 0, "short_yes": true, "train_ids": [ "cvd1-1-ethan", "cvd1-1-gabriella", "cvd1-2-ethan", "cvd1-2-gabriella", "cvd2-1-ethan", "cvd2-2-ethan", "cvd2-2-gabriella", "cvd3-1-gabriella", "cvd3-2-ethan", "cvd3-2-gabriella", "cvd4-1-ethan", "cvd4-1-gabriella", "cvd4-2-ethan" ], "video_dir": "/opt/dlami/nvme/astertech/videos" }, "dataset": { "no": 224, "total": 672, "yes": 448 }, "per_video": { "cvd1-1-ethan": { "no": 28, "total": 52, "yes": 24 }, "cvd1-1-gabriella": { "no": 20, "total": 45, "yes": 25 }, "cvd1-2-ethan": { "no": 16, "total": 54, "yes": 38 }, "cvd1-2-gabriella": { "no": 20, "total": 58, "yes": 38 }, "cvd2-1-ethan": { "no": 16, "total": 38, "yes": 22 }, "cvd2-2-ethan": { "no": 20, "total": 63, "yes": 43 }, "cvd2-2-gabriella": { "no": 20, "total": 66, "yes": 46 }, "cvd3-1-gabriella": { "no": 8, "total": 29, "yes": 21 }, "cvd3-2-ethan": { "no": 16, "total": 65, "yes": 49 }, "cvd3-2-gabriella": { "no": 16, "total": 64, "yes": 48 }, "cvd4-1-ethan": { "no": 20, "total": 48, "yes": 28 }, "cvd4-1-gabriella": { "no": 12, "total": 40, "yes": 28 }, "cvd4-2-ethan": { "no": 12, "total": 50, "yes": 38 } }, "train_ids": [ "cvd1-1-ethan", "cvd1-1-gabriella", "cvd1-2-ethan", "cvd1-2-gabriella", "cvd2-1-ethan", "cvd2-2-ethan", "cvd2-2-gabriella", "cvd3-1-gabriella", "cvd3-2-ethan", "cvd3-2-gabriella", "cvd4-1-ethan", "cvd4-1-gabriella", "cvd4-2-ethan" ] }