{
  "allow_neg_eigval": false,
  "architectures": [
    "PanoVLMForConditionalGeneration"
  ],
  "attn_mode": "chunk",
  "auto_map": {
    "AutoConfig": "configuration_panovlm.PanoVLMConfig",
    "AutoImageProcessor": "image_processing_panovlm.PanoVLMImageProcessor",
    "AutoModel": "modeling_panovlm.PanoVLMForConditionalGeneration",
    "AutoModelForImageTextToText": "modeling_panovlm.PanoVLMForConditionalGeneration",
    "AutoProcessor": "processing_panovlm.PanoVLMProcessor"
  },
  "bos_token_id": 1,
  "conv_bias": false,
  "conv_size": 4,
  "dtype": "bfloat16",
  "eos_token_id": 2,
  "expand_v": 1.0,
  "fuse_bitlinear": true,
  "fuse_norm": true,
  "head_dim": 128,
  "hidden_size": 1024,
  "image_seq_length": 256,
  "image_size": 1024,
  "image_token_id": 32001,
  "lower_bound": -1.0,
  "max_position_embeddings": 131072,
  "mlp_hidden_dim": 2816,
  "model_type": "panovlm",
  "num_heads": 8,
  "num_hidden_layers": 24,
  "num_reserved_token_slots": 128,
  "num_v_heads": null,
  "pad_token_id": null,
  "projector_type": "linear",
  "rms_norm_eps": 1e-05,
  "safe_gate": true,
  "tie_word_embeddings": false,
  "transformers_version": "5.8.1",
  "use_bitlinear": false,
  "use_lower_bound": false,
  "use_short_conv": true,
  "vision_config": {
    "cls_ratio": 2.0,
    "downsamples": [
      false,
      true,
      true,
      true,
      true
    ],
    "embed_dims": [
      96,
      192,
      384,
      768,
      1536
    ],
    "inference_mode": true,
    "layers": [
      2,
      12,
      24,
      4,
      2
    ],
    "lkc_use_act": true,
    "mlp_ratios": [
      4,
      4,
      4,
      4,
      4
    ],
    "norm_layer": {
      "func_args": [],
      "func_kwargs": {
        "eps": 1e-05
      },
      "func_name": "LayerNorm2d"
    },
    "num_classes": 0,
    "pos_embs": [
      null,
      null,
      null,
      {
        "func_args": [],
        "func_kwargs": {
          "spatial_shape": [
            7,
            7
          ]
        },
        "func_name": "RepConditionalPosEnc"
      },
      {
        "func_args": [],
        "func_kwargs": {
          "spatial_shape": [
            7,
            7
          ]
        },
        "func_name": "RepConditionalPosEnc"
      }
    ],
    "se_downsamples": [
      false,
      false,
      false,
      false,
      false
    ],
    "stem_use_scale_branch": false,
    "token_mixers": [
      "repmixer",
      "repmixer",
      "repmixer",
      "attention",
      "attention"
    ]
  },
  "vision_feature_dim": 3072,
  "vocab_size": 32000
}