config.json

{
  "audio_encoder_config": {
    "model_type": "zipformer",
    "feature_dim": 80,
    "output_downsampling_factor": 2,
    "num_encoder_layers": [
      2,
      2,
      4,
      5,
      4,
      2
    ],
    "downsampling_factor": [
      1,
      2,
      4,
      8,
      4,
      2
    ],
    "encoder_dim": [
      192,
      256,
      512,
      768,
      512,
      256
    ],
    "feedforward_dim": [
      576,
      768,
      1536,
      2304,
      1536,
      768
    ],
    "warmup_batches": 4000.0,
    "dropout": null,
    "num_heads": [
      4,
      4,
      4,
      8,
      4,
      4
    ],
    "query_head_dim": [
      32
    ],
    "value_head_dim": [
      12
    ],
    "pos_head_dim": [
      4
    ],
    "pos_dim": 48,
    "encoder_unmasked_dim": [
      192,
      192,
      256,
      256,
      256,
      192
    ],
    "cnn_module_kernel": [
      31,
      31,
      15,
      15,
      15,
      31
    ],
    "causal": false,
    "chunk_size": [
      16,
      32,
      64,
      -1
    ],
    "left_context_frames": [
      64,
      128,
      256,
      -1
    ]
  },
  "llm_config": {
    "vocab_size": 152064,
    "max_position_embeddings": 32768,
    "hidden_size": 3584,
    "intermediate_size": 18944,
    "num_hidden_layers": 28,
    "num_attention_heads": 28,
    "use_sliding_window": false,
    "sliding_window": 131072,
    "max_window_layers": 28,
    "num_key_value_heads": 4,
    "hidden_act": "silu",
    "initializer_range": 0.02,
    "rms_norm_eps": 1e-06,
    "use_cache": true,
    "rope_theta": 1000000.0,
    "attention_dropout": 0.0,
    "torch_dtype": "float16",
    "tie_word_embeddings": false,
    "architectures": [
      "Qwen2ForCausalLM"
    ],
    "bos_token_id": 151643,
    "eos_token_id": 151645,
    "_name_or_path": "/projects/bejv/models/Qwen2.5-7B-Instruct",
    "transformers_version": "4.38.2",
    "model_type": "qwen2"
  },
  "use_flash_attn": false,
  "audio_encoder_projector_ds_rate": 8,
  "exclude_from_checkpoint": [
    "audio_encoder",
    "voice_encoder",
    "llm"
  ],
  "tag_audio_boundary": false,
  "audio_token": "<|AUDIO|>",
  "model_type": "tagspeech",
  "max_length": 800,
  "voice_encoder_config": {
    "model_type": "zipformer",
    "feature_dim": 80,
    "output_downsampling_factor": 2,
    "num_encoder_layers": [
      2,
      2,
      4,
      5,
      4,
      2
    ],
    "downsampling_factor": [
      1,
      2,
      4,
      8,
      4,
      2
    ],
    "encoder_dim": [
      192,
      256,
      512,
      768,
      512,
      256
    ],
    "feedforward_dim": [
      576,
      768,
      1536,
      2304,
      1536,
      768
    ],
    "warmup_batches": 4000.0,
    "dropout": null,
    "num_heads": [
      4,
      4,
      4,
      8,
      4,
      4
    ],
    "query_head_dim": [
      32
    ],
    "value_head_dim": [
      12
    ],
    "pos_head_dim": [
      4
    ],
    "pos_dim": 48,
    "encoder_unmasked_dim": [
      192,
      192,
      256,
      256,
      256,
      192
    ],
    "cnn_module_kernel": [
      31,
      31,
      15,
      15,
      15,
      31
    ],
    "causal": false,
    "chunk_size": [
      16,
      32,
      64,
      -1
    ],
    "left_context_frames": [
      64,
      128,
      256,
      -1
    ]
  },
  "semantic_projector_ds_rate": 4,
  "voice_projector_ds_rate": 4,
  "semantic_anchor_interval": 8,
  "voice_anchor_interval": 8,
  "insert_anchors_at_ends": true
}