| defaults: | |
| - common | |
| - config | |
| data: | |
| use_lang_emb: true | |
| num_langs: 11 | |
| lang_dim: 192 # same size as hidden_channels to facilitate the addition | |
| lang2id: | |
| chinese: 0 | |
| dutch: 1 | |
| english: 2 | |
| french: 3 | |
| german: 4 | |
| italian: 5 | |
| japanese: 6 | |
| other: 7 | |
| polish: 8 | |
| portuguese: 9 | |
| spanish: 10 | |
| use_spk_emb: false | |
| spk_embeddings_dir: null # compute on forward (model) | |
| spk_encoder_type: null # compute on forward (model) | ECAPA2SpeakerEncoder16k | |
| content_encoder_type: null # compute on forward (model) | hubert | |
| content_encoder_ckpt: null # compute on forward (model) | lengyue233/content-vec-best | |
| model: | |
| use_spk_emb: true | |
| spk_encoder_type: ECAPA2SpeakerEncoder16k | |
| spk_encoder_ckpt: null # Not used for ECAPA2SpeakerEncoder16k | |
| content_encoder_type: spin # hubert | wavlm | spin | |
| content_encoder_config: models/spin/spin.yaml # path to the config file for the content encoder | |
| content_encoder_ckpt: models/spin/spin.ckpt # or models/wavlm/WavLM-Large.pt |