{
  "architectures": [
    "Qwen3VLForConditionalGeneration"
  ],
  "bio_token_ids": {
    "_bio_seq_output_pad": {
      "pad": 151687
    },
    "contact_map": {
      "end": 151682,
      "pad": 151683,
      "start": 151681
    },
    "dna": {
      "end": 151673,
      "pad": 151674,
      "start": 151672
    },
    "mol": {
      "end": 151679,
      "pad": 151680,
      "start": 151678
    },
    "protein": {
      "end": 151676,
      "pad": 151677,
      "start": 151675
    },
    "rna": {
      "end": 151670,
      "pad": 151671,
      "start": 151669
    },
    "weather": {
      "end": 151685,
      "pad": 151686,
      "start": 151684
    }
  },
  "bos_token_id": null,
  "dna_config": {
    "_name_or_path": "",
    "architectures": null,
    "chunk_size_feed_forward": 0,
    "conv_kernel_size": 7,
    "dna_encoder_hidden_size": 512,
    "dna_encoder_path": "dna",
    "dna_max_seq_length": 2048,
    "dna_projector_hidden_size": 2048,
    "dna_vocab_size": 8,
    "dropout": 0.1,
    "dtype": null,
    "id2label": {
      "0": "LABEL_0",
      "1": "LABEL_1"
    },
    "is_encoder_decoder": false,
    "label2id": {
      "LABEL_0": 0,
      "LABEL_1": 1
    },
    "model_type": "qwen3_vl_dna",
    "num_attention_heads": 8,
    "num_encoder_layers": 8,
    "num_latent_tokens": 64,
    "output_attentions": false,
    "output_hidden_states": false,
    "problem_type": null,
    "return_dict": true
  },
  "dtype": "bfloat16",
  "eos_token_id": 151645,
  "epi_ohem_enable": false,
  "epi_ohem_hard_fraction": 0.5,
  "epi_ohem_weight": 2.0,
  "image_token_id": 151655,
  "med_seg_config": {
    "cost_bbox": 5.0,
    "cost_class": 2.0,
    "cost_dice": 5.0,
    "cost_giou": 2.0,
    "cost_mask": 5.0,
    "focal_alpha": 0.25,
    "focal_gamma": 2.0,
    "freeze_sam3": true,
    "freeze_sam3_vision": false,
    "loss_w_bbox_giou": 0.5,
    "loss_w_bbox_l1": 0.5,
    "loss_w_cls": 1.0,
    "loss_w_mask_dice": 5.0,
    "loss_w_mask_dice_high": 0.0,
    "loss_w_mask_focal": 2.0,
    "loss_w_mask_semantic": 0.5,
    "loss_w_meta_ce": 0.2,
    "mask_cost_res": 128,
    "mask_h": 576,
    "mask_loss_target_res": null,
    "mask_w": 576,
    "med_query_end_id": null,
    "med_query_start_id": null,
    "meta_num_classes": 15,
    "proj_hidden_mult": 2,
    "proj_mlp": true,
    "sam3_model_path": null,
    "sam3_text_dim": 256,
    "system_prompt": "You are a medical imaging assistant specialized in image understanding for segmentation.\nYou will be provided with (1) a medical image and (2) a text query that specifies a target region to segment.\nYour goal is to produce a detailed, clinically plausible description that helps delineate the target region precisely.\nPlease describe:\n    1) The imaging modality/type if it is evident (e.g., CT, MRI, ultrasound, X-ray, endoscopy, dermoscopy, histopathology). If uncertain, explicitly say it is unknown.\n    2) The overall field of view and orientation (e.g., axial/sagittal/coronal for cross-sectional images, or a general view if not applicable).\n    3) The main anatomical structures/organs visible, and for each one provide approximate location using spatial terms. (left/right, superior/inferior, anterior/posterior, central/peripheral) and relative relationships between organs.\n    4) Most importantly, focus on the structure mentioned in the text query: describe its approximate position, boundaries, shape, size, and its appearance (intensity/color, texture, homogeneity, and contrast to surrounding tissues).\n    Be concise but detailed, use precise spatial language, and emphasize features helpful for pixel-accurate segmentation.",
    "upsample_mask_loss": false,
    "use_query_special_tokens": false
  },
  "med_seg_freeze_qwen": true,
  "med_seg_visual_fp32": false,
  "modality_configs": {},
  "model_type": "qwen3_vl",
  "mol_config": {
    "_name_or_path": "",
    "architectures": null,
    "chunk_size_feed_forward": 0,
    "dropout": 0.1,
    "dtype": null,
    "freeze_gnn": true,
    "freeze_mol_gnn": true,
    "gnn_drop_ratio": 0.0,
    "id2label": {
      "0": "LABEL_0",
      "1": "LABEL_1"
    },
    "is_encoder_decoder": false,
    "label2id": {
      "LABEL_0": 0,
      "LABEL_1": 1
    },
    "model_type": "qwen3_vl_mol",
    "mol_decoder_hidden_size": 768,
    "mol_decoder_max_seq_length": 512,
    "mol_decoder_num_heads": 12,
    "mol_decoder_num_layers": 6,
    "mol_encoder_hidden_size": 256,
    "mol_encoder_path": "mol",
    "mol_projector_hidden_size": 1024,
    "mol_vocab_size": 226,
    "num_gnn_layers": 12,
    "num_latent_tokens": 64,
    "num_resampler_heads": 8,
    "output_attentions": false,
    "output_hidden_states": false,
    "problem_type": null,
    "return_dict": true
  },
  "pad_token_id": 151643,
  "protein_config": {
    "_name_or_path": "",
    "architectures": null,
    "chunk_size_feed_forward": 0,
    "conv_kernel_size": 7,
    "dropout": 0.1,
    "dtype": null,
    "esmc_model_name": "esm2_t30_150M_UR50D",
    "freeze_backbone": true,
    "freeze_esmc": true,
    "freeze_protein_backbone": true,
    "id2label": {
      "0": "LABEL_0",
      "1": "LABEL_1"
    },
    "is_encoder_decoder": false,
    "label2id": {
      "LABEL_0": 0,
      "LABEL_1": 1
    },
    "model_type": "qwen3_vl_protein",
    "num_attention_heads": 8,
    "num_encoder_layers": 12,
    "num_latent_tokens": 64,
    "num_resampler_layers": 2,
    "output_attentions": false,
    "output_hidden_states": false,
    "problem_type": null,
    "protein_backbone_name": "esm2_t30_150M_UR50D",
    "protein_convformer_vocab_size": 33,
    "protein_encoder_hidden_size": 640,
    "protein_encoder_path": "protein",
    "protein_max_seq_length": 8192,
    "protein_projector_hidden_size": 2560,
    "protein_vocab_size": 31,
    "return_dict": true,
    "use_physicochemical_embed": true
  },
  "rna_config": {
    "_name_or_path": "",
    "architectures": null,
    "chunk_size_feed_forward": 0,
    "conv_kernel_size": 7,
    "dropout": 0.1,
    "dtype": null,
    "id2label": {
      "0": "LABEL_0",
      "1": "LABEL_1"
    },
    "is_encoder_decoder": false,
    "label2id": {
      "LABEL_0": 0,
      "LABEL_1": 1
    },
    "model_type": "qwen3_vl_rna",
    "num_attention_heads": 8,
    "num_encoder_layers": 8,
    "num_latent_tokens": 64,
    "output_attentions": false,
    "output_hidden_states": false,
    "problem_type": null,
    "return_dict": true,
    "rna_encoder_hidden_size": 512,
    "rna_encoder_path": "rna",
    "rna_max_seq_length": 4096,
    "rna_projector_hidden_size": 2048,
    "rna_vocab_size": 8
  },
  "text_config": {
    "attention_bias": false,
    "attention_dropout": 0.0,
    "bos_token_id": 151643,
    "dtype": "bfloat16",
    "eos_token_id": 151645,
    "head_dim": 128,
    "hidden_act": "silu",
    "hidden_size": 4096,
    "initializer_range": 0.02,
    "intermediate_size": 12288,
    "max_position_embeddings": 262144,
    "model_type": "qwen3_vl_text",
    "num_attention_heads": 32,
    "num_hidden_layers": 36,
    "num_key_value_heads": 8,
    "rms_norm_eps": 1e-06,
    "rope_parameters": {
      "mrope_interleaved": true,
      "mrope_section": [
        24,
        20,
        20
      ],
      "rope_theta": 5000000,
      "rope_type": "default"
    },
    "rope_theta": 5000000,
    "tie_word_embeddings": false,
    "use_cache": true,
    "vocab_size": 151936
  },
  "text_loss_weight": 1.0,
  "tie_word_embeddings": false,
  "transformers_version": "5.0.0",
  "use_cache": false,
  "video_token_id": 151656,
  "vision_config": {
    "deepstack_visual_indexes": [
      8,
      16,
      24
    ],
    "depth": 27,
    "dtype": "bfloat16",
    "hidden_act": "gelu_pytorch_tanh",
    "hidden_size": 1152,
    "in_channels": 3,
    "initializer_range": 0.02,
    "intermediate_size": 4304,
    "model_type": "qwen3_vl",
    "num_heads": 16,
    "num_position_embeddings": 2304,
    "out_hidden_size": 4096,
    "patch_size": 16,
    "spatial_merge_size": 2,
    "temporal_patch_size": 2
  },
  "vision_end_token_id": 151653,
  "vision_start_token_id": 151652,
  "weather_config": {
    "_name_or_path": "",
    "architectures": null,
    "attn_type": "flash",
    "chunk_size_feed_forward": 0,
    "const_chans": 6,
    "decoder_depth": 12,
    "dtype": null,
    "embed_freq": 256,
    "embed_mode": "add",
    "embed_types": [
      "step",
      "hour",
      "doy",
      "lead_hour"
    ],
    "encoder_depth": 12,
    "ffn_type": "geglu_ffn",
    "frame_interval": "6h",
    "hidden_size": 2048,
    "id2label": {
      "0": "LABEL_0",
      "1": "LABEL_1"
    },
    "image_size": [
      721,
      1440
    ],
    "in_chans": 70,
    "in_frames": 1,
    "is_encoder_decoder": false,
    "label2id": {
      "LABEL_0": 0,
      "LABEL_1": 1
    },
    "lead_hour_scaling": true,
    "lower_chans": 0,
    "mask_type": "h",
    "max_rollout_steps": 200,
    "meteo_data_path": null,
    "meteo_loss_coef": 1.0,
    "meteo_loss_type": "l1_channel",
    "meteo_loss_weight": 1.0,
    "meteo_out_channels": 70,
    "meteo_output_size": [
      721,
      1440
    ],
    "meteo_token_id": 151655,
    "mlp_ratio": 4.0,
    "model_type": "qwen3_vl_weather",
    "n_kv_heads": null,
    "norm_type": "adarms",
    "num_heads": 32,
    "output_attentions": false,
    "output_hidden_states": false,
    "patch_size": 6,
    "pl_chans": null,
    "problem_type": null,
    "qwenvl_dim": 4096,
    "return_dict": true,
    "time_embed_dropout": 0.0,
    "upper_chans": 70,
    "window_size": 20
  }
}