config.json

{
  "_name_or_path": "VideoRefer-VideoLLaMA3-7B",
  "architectures": [
    "Videollama3Qwen2ForCausalLM"
  ],
  "attention_dropout": 0.0,
  "auto_map": {
    "AutoConfig": "configuration_videollama3.Videollama3Qwen2Config",
    "AutoModelForCausalLM": "modeling_videollama3.Videollama3Qwen2ForCausalLM"
  },
  "bos_token_id": 151643,
  "eos_token_id": 151645,
  "hidden_act": "silu",
  "hidden_size": 3584,
  "image_aspect_ratio": "qwen2vl",
  "image_size": 28,
  "image_token_index": 151665,
  "image_token_length": 1,
  "initializer_range": 0.02,
  "intermediate_size": 18944,
  "is_alignment": false,
  "llm_lr": 1e-05,
  "max_frames": 180,
  "max_position_embeddings": 32768,
  "max_window_layers": 28,
  "mm_hidden_size": 1152,
  "mm_max_length": 8192,
  "mm_projector_lr": 1e-05,
  "mm_projector_type": "mlp2x_gelu",
  "mm_vision_encoder": "DAMO-NLP-SG/VL3-SigLIP-NaViT",
  "mm_vision_select_feature": "patch",
  "mm_vision_select_layer": -2,
  "model_type": "videollama3_qwen2",
  "num_attention_heads": 28,
  "num_hidden_layers": 28,
  "num_key_value_heads": 4,
  "region_encoder_lr": 1e-05,
  "region_token_index": 151668,
  "rms_norm_eps": 1e-06,
  "rope_scaling": null,
  "rope_theta": 1000000.0,
  "sliding_window": null,
  "spatial_merge_size": 2,
  "tie_word_embeddings": false,
  "tokenizer_model_max_length": 16384,
  "tokenizer_padding_side": "right",
  "torch_dtype": "bfloat16",
  "transformers_version": "4.46.3",
  "use_cache": true,
  "use_mm_proj": true,
  "use_sliding_window": false,
  "use_token_compression": true,
  "vision_encoder": null,
  "vision_encoder_config": {
    "hidden_size": 1152,
    "intermediate_size": 4304,
    "model_type": "videollama3_vision_encoder",
    "num_attention_heads": 16,
    "num_hidden_layers": 27,
    "patch_size": 14
  },
  "vision_encoder_lr": 0.0,
  "vocab_size": 151669
}