cosyvoice.yaml

# set random seed, so that you may reproduce your result.
__set_seed1: !apply:random.seed [1986]
__set_seed2: !apply:numpy.random.seed [1986]
__set_seed3: !apply:torch.manual_seed [1986]
__set_seed4: !apply:torch.cuda.manual_seed_all [1986]

# fixed params
sample_rate: 22050
text_encoder_input_size: 512
llm_input_size: 1024
llm_output_size: 1024
spk_embed_dim: 192

# model params
# for all class/function included in this repo, we use !<name> or !<new> for intialization, so that user may find all corresponding class/function according to one single yaml.
# for system/third_party class/function, we do not require this.
llm: !new:cosyvoice.llm.llm.TransformerLM
    text_encoder_input_size: !ref <text_encoder_input_size>
    llm_input_size: !ref <llm_input_size>
    llm_output_size: !ref <llm_output_size>
    text_token_size: 51866
    speech_token_size: 4096
    length_normalized_loss: True
    lsm_weight: 0
    spk_embed_dim: !ref <spk_embed_dim>
    text_encoder: !new:cosyvoice.transformer.encoder.ConformerEncoder
        input_size: !ref <text_encoder_input_size>
        output_size: 1024
        attention_heads: 16
        linear_units: 4096
        num_blocks: 6
        dropout_rate: 0.1
        positional_dropout_rate: 0.1
        attention_dropout_rate: 0.0
        normalize_before: True
        input_layer: 'linear'
        pos_enc_layer_type: 'rel_pos_espnet'
        selfattention_layer_type: 'rel_selfattn'
        use_cnn_module: False
        macaron_style: False
        use_dynamic_chunk: False
        use_dynamic_left_chunk: False
        static_chunk_size: 1
    llm: !new:cosyvoice.transformer.encoder.TransformerEncoder
        input_size: !ref <llm_input_size>
        output_size: !ref <llm_output_size>
        attention_heads: 16
        linear_units: 4096
        num_blocks: 14
        dropout_rate: 0.1
        positional_dropout_rate: 0.1
        attention_dropout_rate: 0.0
        input_layer: 'linear_legacy'
        pos_enc_layer_type: 'rel_pos_espnet'
        selfattention_layer_type: 'rel_selfattn'
        static_chunk_size: 1
    sampling: !name:cosyvoice.utils.common.ras_sampling
        top_p: 0.8
        top_k: 25
        win_size: 10
        tau_r: 0.1

flow: !new:cosyvoice.flow.flow.MaskedDiffWithXvec
    input_size: 512
    output_size: 80
    spk_embed_dim: !ref <spk_embed_dim>
    output_type: 'mel'
    vocab_size: 4096
    input_frame_rate: 50
    only_mask_loss: True
    encoder: !new:cosyvoice.transformer.encoder.ConformerEncoder
        output_size: 512
        attention_heads: 8
        linear_units: 2048
        num_blocks: 6
        dropout_rate: 0.1
        positional_dropout_rate: 0.1
        attention_dropout_rate: 0.1
        normalize_before: True
        input_layer: 'linear'
        pos_enc_layer_type: 'rel_pos_espnet'
        selfattention_layer_type: 'rel_selfattn'
        input_size: 512
        use_cnn_module: False
        macaron_style: False
    length_regulator: !new:cosyvoice.flow.length_regulator.InterpolateRegulator
        channels: 80
        sampling_ratios: [1, 1, 1, 1]
    decoder: !new:cosyvoice.flow.flow_matching.ConditionalCFM
        in_channels: 240
        n_spks: 1
        spk_emb_dim: 80
        cfm_params: !new:omegaconf.DictConfig
            content:
                sigma_min: 1e-06
                solver: 'euler'
                t_scheduler: 'cosine'
                training_cfg_rate: 0.2
                inference_cfg_rate: 0.7
                reg_loss_type: 'l1'
        estimator: !new:cosyvoice.flow.decoder.ConditionalDecoder
            in_channels: 320
            out_channels: 80
            channels: [256, 256]
            dropout: 0.0
            attention_head_dim: 64
            n_blocks: 4
            num_mid_blocks: 12
            num_heads: 8
            act_fn: 'gelu'

hift: !new:cosyvoice.hifigan.generator.HiFTGenerator
    in_channels: 80
    base_channels: 512
    nb_harmonics: 8
    sampling_rate: !ref <sample_rate>
    nsf_alpha: 0.1
    nsf_sigma: 0.003
    nsf_voiced_threshold: 10
    upsample_rates: [8, 8]
    upsample_kernel_sizes: [16, 16]
    istft_params:
        n_fft: 16
        hop_len: 4
    resblock_kernel_sizes: [3, 7, 11]
    resblock_dilation_sizes: [[1, 3, 5], [1, 3, 5], [1, 3, 5]]
    source_resblock_kernel_sizes: [7, 11]
    source_resblock_dilation_sizes: [[1, 3, 5], [1, 3, 5]]
    lrelu_slope: 0.1
    audio_limit: 0.99
    f0_predictor: !new:cosyvoice.hifigan.f0_predictor.ConvRNNF0Predictor
        num_class: 1
        in_channels: 80
        cond_channels: 512

# processor functions
parquet_opener: !name:cosyvoice.dataset.processor.parquet_opener
get_tokenizer: !name:whisper.tokenizer.get_tokenizer
    multilingual: True
    num_languages: 100
    language: 'en'
    task: 'transcribe'
allowed_special: 'all'
tokenize: !name:cosyvoice.dataset.processor.tokenize
    get_tokenizer: !ref <get_tokenizer>
    allowed_special: !ref <allowed_special>
filter: !name:cosyvoice.dataset.processor.filter
    max_length: 40960
    min_length: 0
    token_max_length: 200
    token_min_length: 1
resample: !name:cosyvoice.dataset.processor.resample
    resample_rate: !ref <sample_rate>
feat_extractor: !name:matcha.utils.audio.mel_spectrogram
    n_fft: 1024
    num_mels: 80
    sampling_rate: !ref <sample_rate>
    hop_size: 256
    win_size: 1024
    fmin: 0
    fmax: 8000
    center: False
compute_fbank: !name:cosyvoice.dataset.processor.compute_fbank
    feat_extractor: !ref <feat_extractor>
parse_embedding: !name:cosyvoice.dataset.processor.parse_embedding
    normalize: True
shuffle: !name:cosyvoice.dataset.processor.shuffle
    shuffle_size: 1000
sort: !name:cosyvoice.dataset.processor.sort
    sort_size: 500  # sort_size should be less than shuffle_size
batch: !name:cosyvoice.dataset.processor.batch
    batch_type: 'dynamic'
    max_frames_in_batch: 2000
padding: !name:cosyvoice.dataset.processor.padding

# dataset processor pipeline
data_pipeline: [
    !ref <parquet_opener>,
    !ref <tokenize>,
    !ref <filter>,
    !ref <resample>,
    !ref <compute_fbank>,
    !ref <parse_embedding>,
    !ref <shuffle>,
    !ref <sort>,
    !ref <batch>,
    !ref <padding>,
]

# train conf
train_conf:
    optim: adam
    optim_conf:
        lr: 0.001
    scheduler: warmuplr
    scheduler_conf:
        warmup_steps: 2500
    max_epoch: 200
    grad_clip: 5
    accum_grad: 2
    log_interval: 100
    save_per_step: -1