{
  "architectures": [
    "LlavaHerQwen3ForCausalLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "attn_implementation": "flash_attention_2",
  "bos_token_id": 151643,
  "ctc_decoder_config": "(2,896,32,11008)",
  "ctc_loss_weight": 1,
  "ctc_upsample_factor": 12,
  "bs_decoder_dimension": 512,
  "dropout": 0.1,
  "dropout_rate": 0.0,
  "eos_token_id": 151645,
  "freeze_blendshape_generator": false,
  "head_dim": 128,
  "hidden_act": "silu",
  "hidden_size": 4096,
  "identity_num": 935,
  "inference": false,
  "initializer_range": 0.02,
  "intermediate_size": 12288,
  "interpolate_pos": 0,
  "layer_norm_eps": 1e-05,
  "load_pretrain_speech_embedding_path": false,
  "load_pretrain_speech_encoder_weights": false,
  "load_pretrain_speech_generator_weights": false,
  "low_cpu_mem_usage": false,
  "max_position_embeddings": 40960,
  "max_seq_len": 5000,
  "max_window_layers": 36,
  "mlp_bias": false,
  "model_name_or_path": "/apdcephfs_qy3/share_4842051/haoyoozhang/ckpt/Ex-Omni-3D/WhisperEnc/Stage2_8B_TTSF_0502",
  "model_torch_dtype": "bfloat16",
  "model_type": "qwen3",
  "nhead": 8,
  "num_attention_heads": 32,
  "num_bs_decoder_layers": 6,
  "num_hidden_layers": 36,
  "num_key_value_heads": 8,
  "out_dim": 52,
  "period": 30,
  "pretrain_speech_embedding_path": "/apdcephfs_qy3/share_1443437/haoyoozhang/CodeProjects/flow_embedding.pt",
  "pretrain_speech_encoder_weights": "/apdcephfs_qy3/share_4842051/haoyoozhang/ckpt/whisper-large-v3",
  "pretrain_speech_generator_weights": "/apdcephfs_qy3/share_4842051/haoyoozhang/ckpt/Qwen3/Qwen3-0.6B",
  "pretraining_tp": 1,
  "rms_norm_eps": 1e-06,
  "rope_scaling": {
    "factor": 8.0,
    "high_freq_factor": 4.0,
    "low_freq_factor": 1.0,
    "original_max_position_embeddings": 8192,
    "rope_type": "llama3"
  },
  "rope_theta": 1000000,
  "sliding_window": null,
  "speech_embed_dim": 512,
  "speech_encoder_ds_rate": 4,
  "speech_encoder_hidden_size": 1280,
  "speech_encoder_type": "whisper",
  "speech_gen_hidden_size": 1024,
  "speech_generator_type": "ar",
  "speech_projector_type": "linear",
  "speech_vocab_size": 16384,
  "tie_word_embeddings": false,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.52.4",
  "tune_speech_generator_dpo": false,
  "unit_vocab_size": 16384,
  "use_cache": true,
  "use_layer_norm": true,
  "use_sliding_window": false,
  "vocab_size": 151936,
  "bs_tqgf_depth": 2,
  "bs_tqgf_heads": 8,
  "bs_tqgf_dim_head": 64
}