config.json

{
  "architectures": [
    "VLMQwenForCausalLM"
  ],
  "attention_dropout": 0.0,
  "bos_token_id": 151643,
  "depth": 12,
  "dim": 768,
  "eos_token_id": 151645,
  "hidden_act": "silu",
  "hidden_size": 3584,
  "high_input_size": [
    32,
    768
  ],
  "high_output_size": [
    64,
    128
  ],
  "initializer_range": 0.02,
  "input_size": [
    256,
    256,
    128
  ],
  "intermediate_size": 18944,
  "layer_types": [
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention",
    "full_attention"
  ],
  "low_input_size": [
    256,
    384
  ],
  "low_output_size": [
    192,
    128
  ],
  "max_position_embeddings": 32768,
  "max_window_layers": 28,
  "mm_hidden_size": 768,
  "mm_mlp_depth": 2,
  "mm_projector_type": "mixer",
  "model_type": "vlm_qwen",
  "num_attention_heads": 28,
  "num_hidden_layers": 28,
  "num_key_value_heads": 4,
  "patch_size": [
    16,
    16,
    16
  ],
  "proj_out_num": 256,
  "rms_norm_eps": 1e-06,
  "rope_scaling": null,
  "rope_theta": 1000000.0,
  "sliding_window": null,
  "tie_word_embeddings": false,
  "torch_dtype": "float32",
  "transformers_version": "4.55.0.dev0",
  "use_cache": true,
  "use_sliding_window": false,
  "vision_select_feature": "cls_patch",
  "vision_select_layer": -2,
  "vision_tower": "dcformer",
  "vocab_size": 151666
}