{
  "architectures": [
    "RepConTURModel"
  ],
  "auto_map": {
    "AutoConfig": "configuration_repcontur.RepConTURConfig",
    "AutoModel": "modeling_repcontur.RepConTURModel"
  },
  "ct_encoder": {
    "codebook_size": 8192,
    "dim": 512,
    "dim_head": 32,
    "encoder_type": "ctvit",
    "heads": 8,
    "image_size": 240,
    "patch_size": 20,
    "spatial_depth": 4,
    "temporal_depth": 4,
    "temporal_patch_size": 10
  },
  "ct_pooler": {
    "mlp_depth": 2,
    "n_heads": 8,
    "n_queries": 256
  },
  "decoder_layer": 4,
  "dtype": "bfloat16",
  "generation_kwargs": {},
  "img_context_token_id": 151667,
  "internvl_model_path": "OpenGVLab/InternVL2_5-1B",
  "llm_hidden_size": 896,
  "model_type": "repcontur",
  "num_heads": 32,
  "num_maskformer_stages": 5,
  "query_dim": 2048,
  "seg_token_idx": 151674,
  "special_tokens": [
    "[SEG]",
    "<p>",
    "</p>",
    "<vp>",
    "</vp>"
  ],
  "template": {
    "INSTRUCTION": "<|user|>\n{input}<|end|>\n<|assistant|>\n",
    "SEP": "\n",
    "STOP_WORDS": [
      "<|end|>"
    ],
    "SUFFIX": "<|end|>",
    "SUFFIX_AS_EOS": true,
    "SYSTEM": "<|system|>\n{system}<|end|>\n"
  },
  "text_embedding_dim": 2560,
  "transformers_version": "4.57.1",
  "vocab_size": 151679,
  "voxtell_arch_kwargs": {
    "conv_bias": true,
    "conv_op": "torch.nn.modules.conv.Conv3d",
    "dropout_op": null,
    "dropout_op_kwargs": null,
    "features_per_stage": [
      32,
      64,
      128,
      256,
      320,
      320
    ],
    "kernel_sizes": [
      [
        3,
        3,
        3
      ],
      [
        3,
        3,
        3
      ],
      [
        3,
        3,
        3
      ],
      [
        3,
        3,
        3
      ],
      [
        3,
        3,
        3
      ],
      [
        3,
        3,
        3
      ]
    ],
    "n_blocks_per_stage": [
      1,
      3,
      4,
      6,
      6,
      6
    ],
    "n_conv_per_stage_decoder": [
      1,
      1,
      1,
      1,
      1
    ],
    "n_stages": 6,
    "nonlin": "torch.nn.LeakyReLU",
    "nonlin_kwargs": {
      "inplace": true
    },
    "norm_op": "torch.nn.modules.instancenorm.InstanceNorm3d",
    "norm_op_kwargs": {
      "affine": true,
      "eps": 1e-05
    },
    "strides": [
      [
        1,
        1,
        1
      ],
      [
        2,
        2,
        2
      ],
      [
        2,
        2,
        2
      ],
      [
        2,
        2,
        2
      ],
      [
        2,
        2,
        2
      ],
      [
        2,
        2,
        2
      ]
    ]
  },
  "voxtell_decoder_layer": 4,
  "voxtell_kw_requires_import": [
    "conv_op",
    "norm_op",
    "dropout_op",
    "nonlin"
  ],
  "voxtell_num_heads": 32,
  "voxtell_num_maskformer_stages": 5,
  "voxtell_query_dim": 2048,
  "voxtell_text_embedding_dim": 2560
}