{
  "export_time": "2025-10-13_11-29-14",
  "trainer_cfg": {
    "batch_size": 1,
    "grad_accum_steps": 8,
    "num_workers": 2,
    "lr": 2e-05,
    "weight_decay": 0.01,
    "max_epochs": 6,
    "max_grad_norm": 1.0,
    "warmup_ratio": 0.1,
    "log_every": 5,
    "save_dir": "./checkpoints_pt",
    "resume_path": null,
    "max_seq_len": 1792,
    "left_ctx": 768,
    "max_vis_tokens": 64,
    "image_token_id": 151643,
    "PAD_ID": 0,
    "LABEL_PAD": -100,
    "label_smoothing": 0.1,
    "train_frames_min": 8,
    "train_frames_max": 14,
    "lr_lm": 5e-06,
    "lr_vision": 0.0001,
    "lr_connector": 0.0002,
    "weight_decay_lm": 0.01,
    "weight_decay_vision": 0.05,
    "weight_decay_connector": 0.05
  },
  "model_class": "InternVLChatModel",
  "device": "cuda:0",
  "notes": "Weights saved after training; if you used a monkey-patched forward at train time, weights remain standard."
}