| acoustic_encoder_configs: | |
| avg_prob: 0.0 | |
| emb_size: 256 | |
| ft_size: 512 | |
| is_24k: true | |
| mix_prob: 0.0 | |
| output_ln: true | |
| speech_upstream: utter-project/mHuBERT-147 | |
| transformer_configs: | |
| attn_dropout: 0.05 | |
| depth: 4 | |
| dim: 512 | |
| dim_head: 64 | |
| ff_dropout: 0.05 | |
| heads: 8 | |
| input_dim: 800 | |
| num_register_tokens: 0 | |
| skip_posemb: false | |
| acoustic_quantizer_configs: | |
| codebook_dim: 64 | |
| codebook_size: 1024 | |
| input_dim: 256 | |
| n_codebooks: 24 | |
| possibly_no_quantizer: true | |
| quantizer_dropout: 1 | |
| additive: true | |
| apply_ln: false | |
| avg_spk_prob: 0.1 | |
| do_auggan: false | |
| drop_spk_prob: 0.0 | |
| generator_configs: | |
| cond_dim: 128 | |
| dim: 3072 | |
| hop_length: 480 | |
| input_channels: 288 | |
| intermediate_dim: 3072 | |
| n_fft: 1920 | |
| num_layers: 6 | |
| padding: same | |
| use_film: false | |
| input_layer_configs: | |
| dropout: 0.05 | |
| hidden_dims: | |
| - 1024 | |
| - 1024 | |
| - 1024 | |
| input_dim: 768 | |
| output_dim: 256 | |
| output_ln: true | |
| use_linear: false | |
| on_the_fly: false | |
| perceptual_loss_configs: | |
| layers: | |
| - 0 | |
| - 3 | |
| - 6 | |
| - 9 | |
| - 12 | |
| speech_upstream: microsoft/wavlm-large | |
| progress_encoding_configs: | |
| emb_dim: 32 | |
| emb_n: 11 | |
| quantizer_configs: | |
| dim: 256 | |
| levels: | |
| - 8 | |
| - 8 | |
| - 8 | |
| - 8 | |
| - 8 | |
| num_codebooks: 1 | |
| recon_content: true | |
| sample_rate: 24000 | |
| segauc_ratio: 0.0 | |
| use_perceptual_loss: false | |
| use_rvq_acoustic: true | |
| use_rvq_content: false | |
| use_segauc: false | |