config.json

{
  "architectures": [
    "LlavaForConditionalGeneration"
  ],
  "image_seq_length": 256,
  "image_token_index": 128260,
  "model_type": "llava",
  "multimodal_projector_bias": true,
  "projector_hidden_act": "gelu",
  "text_config": {
    "_attn_implementation_autoset": true,
    "_name_or_path": "hongyuw/bitnet-b1.58-2B-4T-bf16",
    "architectures": [
      "BitNetForCausalLM"
    ],
    "attention_bias": false,
    "attention_dropout": 0.0,
    "auto_map": {
      "AutoConfig": "hongyuw/bitnet-b1.58-2B-4T-bf16--configuration_bitnet.BitNetConfig",
      "AutoModelForCausalLM": "hongyuw/bitnet-b1.58-2B-4T-bf16--modeling_bitnet.BitNetForCausalLM"
    },
    "hidden_act": "relu2",
    "hidden_size": 2560,
    "initializer_range": 0.02,
    "intermediate_size": 6912,
    "max_position_embeddings": 4096,
    "model_path": "hongyuw/bitnet-b1.58-2B-4T-bf16",
    "model_type": "BitNet",
    "num_attention_heads": 20,
    "num_hidden_layers": 30,
    "num_key_value_heads": 5,
    "quantization_config": {
      "linear_class": "autobitlinear",
      "quant_method": "bitnet",
      "quantization_mode": "online"
    },
    "rms_norm_eps": 1e-05,
    "rope_theta": 500000.0,
    "torch_dtype": "bfloat16",
    "use_cache": true,
    "vocab_size": 128264
  },
  "torch_dtype": "bfloat16",
  "transformers_version": "4.51.0.dev0",
  "vision_config": {
    "_attn_implementation_autoset": true,
    "attention_dropout": 0.0,
    "hidden_act": "gelu_pytorch_tanh",
    "hidden_size": 1152,
    "image_size": 224,
    "intermediate_size": 4304,
    "layer_norm_eps": 1e-06,
    "model_type": "siglip_vision_model",
    "num_attention_heads": 16,
    "num_channels": 3,
    "num_hidden_layers": 26,
    "patch_size": 14,
    "vision_use_head": false,
    "vit_act_bits": 8,
    "vit_weight_bits": 1
  },
  "vision_feature_layer": -1,
  "vision_feature_select_strategy": "full"
}