config.json

{
  "architectures": [
    "YoutuVLForConditionalGeneration"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "auto_map": {
    "AutoConfig": "configuration_youtu_vl.YoutuVLConfig",
    "AutoModelForCausalLM": "modeling_youtu_vl.YoutuVLForConditionalGeneration",
    "AutoProcessor": "processing_youtu_vl.YoutuVLProcessor",
    "AutoImageProcessor": "image_processing_siglip2_fast.Siglip2ImageProcessorFast"
  },
  "bos_token_id": 128000,
  "embedding_initializer_range": 0.025,
  "eos_token_id": 128001,
  "head_dim": 64,
  "hidden_act": "silu",
  "hidden_size": 2560,
  "image_token_id": 128264,
  "initializer_range": 0.0125,
  "intermediate_size": 9728,
  "kv_lora_rank": 512,
  "max_position_embeddings": 32768,
  "mlp_bias": false,
  "model_type": "youtu_vl",
  "num_attention_heads": 32,
  "num_hidden_layers": 40,
  "num_key_value_heads": 32,
  "pad_token_id": 128001,
  "q_lora_rank": 1536,
  "qk_head_dim": 192,
  "qk_nope_head_dim": 128,
  "qk_rope_head_dim": 64,
  "rms_norm_eps": 1e-06,
  "rope_interleave": true,
  "rope_theta": 500000,
  "tie_word_embeddings": true,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.56.0",
  "use_cache": false,
  "v_head_dim": 128,
  "video_token_id": 128265,
  "vision_config": {
    "attention_dropout": 0.0,
    "hidden_act": "gelu_pytorch_tanh",
    "hidden_size": 1152,
    "intermediate_size": 4304,
    "layer_norm_eps": 1e-06,
    "model_type": "siglip2_vision_model",
    "num_attention_heads": 16,
    "num_channels": 3,
    "num_hidden_layers": 27,
    "num_patches": 4096,
    "out_hidden_size": 2560,
    "patch_size": 16,
    "tokens_per_second": 2,
    "torch_dtype": "bfloat16",
    "vision_use_head": false,
    "window_size": 256,
    "fullatt_block_indexes": [
        7,
        15,
        23,
        26
    ]
  },
  "vision_end_token_id": 128263,
  "vision_start_token_id": 128262,
  "custom_tokens": {
    "<custom_1>": [282363],
    "<|image_pad|>": [128264],
    "<ins>": [283365],
    "<ref>": [283371],
    "</ref>": [283372],
    "<mask>": [27, 16499, 29],
    "</mask>": [713, 16499, 29],
    "<depth>": [
      [440, 29064, 24661],
      [440, 29064, 12672]
    ],
    "comma": [11],
    "digit_start": [15],
    "<OTHERS>": [283375],
    "<x_0>": [278267],
    "<y_2047>": [282362],
    "<mask_rle>": [7],
    "</mask_rle>": [8]
  },
  "vocab_size": 283386
}