zm_config.json

{
  "library_name": "zeromodels",
  "zeromodels_version": "1.2.2",
  "model_module": "zeromodels.models.qwen3_vl",
  "model_class": "Qwen3VLConditionalGenerate",
  "variant": "qwen3-vl-32b-instruct",
  "weights": "model.weights.json",
  "schema_version": 2,
  "weight_dtype": "bfloat16",
  "model_type": "qwen3_vl",
  "text_config": {
    "vocab_size": 151936,
    "embed_dim": 5120,
    "mlp_dim": 25600,
    "num_layers": 64,
    "num_heads": 64,
    "num_kv_heads": 8,
    "head_dim": 128,
    "norm_eps": 1e-06,
    "rope_theta": 5000000,
    "mrope_section": [
      24,
      20,
      20
    ],
    "tie_embeddings": false
  },
  "vision_config": {
    "depth": 27,
    "embed_dim": 1152,
    "mlp_dim": 4304,
    "num_heads": 16,
    "out_dim": 5120,
    "act": "gelu_pytorch_tanh",
    "num_position_embeddings": 2304,
    "deepstack_visual_indexes": [
      8,
      16,
      24
    ],
    "patch_size": 16,
    "spatial_merge_size": 2,
    "temporal_patch_size": 2,
    "in_channels": 3
  },
  "image_token_id": 151655,
  "video_token_id": 151656,
  "vision_start_token_id": 151652,
  "vision_end_token_id": 151653
}