{
  "model_type": "qwen3_asr",
  "params": {
    "mel_param": "qwen3_asr_mel.ncnn.param",
    "mel_bin": "qwen3_asr_mel.ncnn.bin",
    "audio_conv_param": "qwen3_asr_audio_conv.ncnn.param",
    "audio_conv_bin": "qwen3_asr_audio_conv.ncnn.bin",
    "audio_encoder_param": "qwen3_asr_audio_encoder.ncnn.param",
    "audio_encoder_bin": "qwen3_asr_audio_encoder.ncnn.bin",
    "text_embed_param": "qwen3_asr_text_embed.ncnn.param",
    "text_embed_bin": "qwen3_asr_text_embed.ncnn.bin",
    "text_decoder_param": "qwen3_asr_decoder.ncnn.param",
    "text_decoder_bin": "qwen3_asr_decoder.ncnn.bin",
    "lm_head_param": "qwen3_asr_lm_head.ncnn.param",
    "lm_head_bin": "qwen3_asr_text_embed.ncnn.bin"
  },
  "tokenizer": {
    "type": "bbpe",
    "vocab_file": "vocab.txt",
    "merges_file": "merges.txt",
    "bos": "",
    "eos": "<|im_end|>",
    "pad": "<|endoftext|>",
    "unk": "",
    "eos_ids": [
      151645,
      151643
    ],
    "additional_special_tokens": [
      "<|endoftext|>",
      "<|im_start|>",
      "<|im_end|>",
      "<|object_ref_start|>",
      "<|object_ref_end|>",
      "<|box_start|>",
      "<|box_end|>",
      "<|quad_start|>",
      "<|quad_end|>",
      "<|vision_start|>",
      "<|vision_end|>",
      "<|vision_pad|>",
      "<|image_pad|>",
      "<|video_pad|>",
      "<tool_call>",
      "</tool_call>",
      "<|fim_prefix|>",
      "<|fim_middle|>",
      "<|fim_suffix|>",
      "<|fim_pad|>",
      "<|repo_name|>",
      "<|file_sep|>",
      "<tool_response>",
      "</tool_response>",
      "<think>",
      "</think>",
      "<|audio_start|>",
      "<|audio_end|>",
      "<tts_pad>",
      "<tts_text_bos>",
      "<tts_text_eod>",
      "<tts_text_bos_single>",
      "<non_speech>",
      "<|audio_pad|>",
      "<blank1>",
      "<blank2>",
      "<blank3>",
      "<blank4>",
      "<blank5>",
      "<blank6>",
      "<blank7>",
      "<blank8>",
      "<blank9>",
      "<blank10>",
      "<blank11>",
      "<blank12>",
      "<blank13>",
      "<blank14>",
      "<blank15>",
      "<blank16>",
      "<blank17>",
      "<blank18>",
      "<blank19>",
      "<blank20>",
      "<blank21>",
      "<blank22>",
      "<blank23>",
      "<blank24>",
      "<blank25>",
      "<blank26>",
      "<blank27>",
      "<asr_text>"
    ]
  },
  "setting": {
    "attn_cnt": 28,
    "hidden_size": 1024,
    "head_dim": 128,
    "num_attention_heads": 16,
    "num_key_value_heads": 8,
    "vocab_size": 151936,
    "kv_cache": true,
    "rope": {
      "type": "RoPE",
      "rope_head_dim": 128,
      "rope_theta": 1000000.0
    },
    "audio_pad_token_id": 151676,
    "audio_start_token_id": 151669,
    "audio_end_token_id": 151670,
    "asr_text_token_id": 151704,
    "special_token_id_begin": 151643,
    "audio": {
      "sampling_rate": 16000,
      "n_fft": 400,
      "hop_length": 160,
      "n_mels": 128,
      "chunk_frames": 100,
      "tokens_per_chunk": 13
    }
  }
}