config.json

{
  "architectures": [
    "InternVLOVDForDetection"
  ],
  "head_type": "small",
  "cost_bbox": 5.0,
  "cost_class": 0.0,
  "cost_giou": 2.0,
  "device_map": "cuda",
  "dim_feedforward": 1024,
  "dropout": 0.0,
  "dtype": "bfloat16",
  "eos_coef": 0.1,
  "focal_alpha": 0.75,
  "focal_gamma": 2.0,
  "freeze_backbone": true,
  "hidden_size": 1024,
  "loss_bbox": 5.0,
  "loss_cls": 0.0,
  "loss_giou": 2.0,
  "loss_mode": "bbox_only",
  "model_type": "internvl_ovd",
  "nhead": 8,
  "num_decoder_layers": 2,
  "num_queries": 1,
  "token_fpn_include_text": true,
  "token_fpn_levels": [
    16,
    8,
    4,
    2
  ],
  "transformers_version": "4.57.3",
  "use_focal_loss": false,
  "use_token_fpn": false,
  "vlm_model_name": "OpenGVLab/InternVL3_5-1B",
  "auto_map": {
    "AutoConfig": "configuration_internvl_ovd.InternVLOVDConfig",
    "AutoModel": "modeling_internvl_ovd.InternVLOVDForDetection"
  }
}