config.json

{
  "model_type": "fusion-embedding-connector",
  "adapter_rank": 384,
  "d_resampler": 384,
  "n_query": 64,
  "d_llm": 2048,
  "mrl_dims": [
    2048,
    1536,
    1024,
    512,
    256,
    128,
    64
  ],
  "trained_params": 60552832,
  "protocol": {
    "text_query_template": "<|im_start|>system\n{instruction}<|im_end|>\n<|im_start|>user\n{text}<|im_end|>\n<|im_start|>assistant\n",
    "image_doc_template": "<|im_start|>system\nRepresent the user's input.<|im_end|>\n<|im_start|>user\n<|vision_start|><|image_pad|><|vision_end|><|im_end|>\n<|im_start|>assistant\n",
    "default_query_instruction": "Retrieve images or text relevant to the user's query.",
    "pooling": "last non-pad token",
    "cross_modal_readout": "per-modality mean-centering recommended (see inference.center)"
  },
  "benchmarks": {
    "audiocaps_883_minrank5": {
      "a2t_R@1": 0.333,
      "a2t_R@5": 0.646,
      "a2t_R@10": 0.785,
      "a2t_mAP@10": 0.255,
      "t2a_R@1": 0.297,
      "t2a_R@5": 0.653,
      "t2a_R@10": 0.782,
      "t2a_mAP@10": 0.445
    },
    "maeb_audio_only_19_mean": 0.4536,
    "maeb_rank": "8 of 14 (full-coverage)",
    "speechcommands_zeroshot": 0.894,
    "ravdess_zeroshot": 0.211
  },
  "d_audio": 3584,
  "resampler_depth": 6,
  "resampler_heads": 8,
  "resampler_ffn_mult": 4,
  "resampler_dropout": 0.0,
  "adapter_act": "silu",
  "mrl_default": 1024,
  "audio_pad_id": 151654,
  "eos_id": 151645,
  "pad_id": 151643,
  "audio_pad_token": "<|audio_pad|>",
  "base_model": "Qwen/Qwen3-VL-Embedding-2B",
  "audio_model": "Qwen/Qwen2.5-Omni-7B",
  "max_text_tokens": 512,
  "n_decoder_layers": 28,
  "architectures": [
    "FusionEmbeddingModel"
  ],
  "auto_map": {
    "AutoConfig": "configuration_fusion_embedding.FusionEmbeddingConfig",
    "AutoModel": "modeling_fusion_embedding.FusionEmbeddingModel"
  },
  "base_4bit": false,
  "version": "v0.3-preview"
}