{
    "architecture": "chatterbox_turbo",
    "dec_cond_len_seconds": 10,
    "enc_cond_len_seconds": 15,
    "gpt2": {
        "activation_function": "gelu_new",
        "n_ctx": 8196,
        "n_embd": 1024,
        "hidden_size": 1024,
        "n_head": 16,
        "n_layer": 24,
        "n_positions": 8196,
        "vocab_size": 50276,
        "layer_norm_epsilon": 1e-05,
        "attn_pdrop": 0.1,
        "embd_pdrop": 0.1,
        "resid_pdrop": 0.1
    },
    "model_type": "chatterbox_turbo",
    "quantization": {
        "group_size": 64,
        "bits": 4,
        "mode": "affine"
    },
    "quantization_config": {
        "group_size": 64,
        "bits": 4,
        "mode": "affine"
    },
    "s3gen": {
        "output_sample_rate": 24000,
        "input_sample_rate": 16000,
        "silence_token": 4299,
        "speech_vocab_size": 6561,
        "meanflow": true,
        "token_embedding_dim": 512,
        "encoder_attention_heads": 8,
        "encoder_linear_units": 2048,
        "encoder_num_blocks": 6,
        "encoder_dropout_rate": 0.1,
        "decoder_in_channels": 320,
        "decoder_out_channels": 80,
        "decoder_channels": [
            256
        ],
        "decoder_attention_head_dim": 64,
        "decoder_n_blocks": 4,
        "decoder_num_mid_blocks": 12,
        "decoder_num_heads": 8,
        "cfm_sigma_min": 1e-06,
        "cfm_t_scheduler": "cosine",
        "cfm_inference_cfg_rate": 0.7
    },
    "sample_rate": 24000,
    "t3": {
        "start_text_token": 255,
        "stop_text_token": 0,
        "text_tokens_dict_size": 50276,
        "max_text_tokens": 2048,
        "start_speech_token": 6561,
        "stop_speech_token": 6562,
        "speech_tokens_dict_size": 6563,
        "max_speech_tokens": 4096,
        "llama_config_name": "GPT2_medium",
        "input_pos_emb": null,
        "speech_cond_prompt_len": 375,
        "encoder_type": "voice_encoder",
        "speaker_embed_size": 256,
        "use_perceiver_resampler": false,
        "emotion_adv": false
    },
    "voice_encoder": {
        "num_mels": 40,
        "sample_rate": 16000,
        "speaker_embed_size": 256,
        "ve_hidden_size": 256,
        "flatten_lstm_params": false,
        "n_fft": 400,
        "hop_size": 160,
        "win_size": 400,
        "fmax": 8000,
        "fmin": 0,
        "preemphasis": 0.0,
        "mel_power": 2.0,
        "mel_type": "amp",
        "normalized_mels": false,
        "ve_partial_frames": 160,
        "ve_final_relu": true,
        "stft_magnitude_min": 0.0001
    }
}