|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
pretrained_path: TalTechNLP/voxlingua107-xls-r-300m-wav2vec |
|
|
|
|
|
sample_rate: 16000 |
|
device: 'cpu' |
|
|
|
|
|
compute_features: !new:speechbrain.lobes.features.Fbank |
|
n_mels: 60 |
|
|
|
|
|
|
|
wav2vec2_hub: facebook/wav2vec2-xls-r-300m |
|
freeze_wav2vec: True |
|
save_folder: ./save |
|
|
|
wav2vec2: !new:speechbrain.lobes.models.huggingface_wav2vec.HuggingFaceWav2Vec2 |
|
source: !ref <wav2vec2_hub> |
|
output_norm: True |
|
freeze: !ref <freeze_wav2vec> |
|
save_path: !ref <save_folder>/wav2vec2_checkpoint |
|
|
|
out_neurons: 107 |
|
|
|
classifier: !new:speechbrain.lobes.models.Xvector.Classifier |
|
input_shape: [null, null, 2048] |
|
activation: !name:torch.nn.LeakyReLU |
|
lin_blocks: 1 |
|
lin_neurons: 512 |
|
out_neurons: !ref <out_neurons> |
|
|
|
label_encoder: !new:speechbrain.dataio.encoder.CategoricalEncoder |
|
|
|
attentive: !new:speechbrain.lobes.models.ECAPA_TDNN.AttentiveStatisticsPooling |
|
channels: 1024 |
|
attention_channels: 64 |
|
|
|
modules: |
|
wav2vec2: !ref <wav2vec2> |
|
compute_features: !ref <compute_features> |
|
classifier: !ref <classifier> |
|
attentive: !ref <attentive> |
|
softmax: !ref <softmax> |
|
|
|
model: !new:torch.nn.ModuleList |
|
- [!ref <attentive>, !ref <classifier>] |
|
|
|
pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer |
|
loadables: |
|
wav2vec2: !ref <wav2vec2> |
|
classifier: !ref <classifier> |
|
label_encoder: !ref <label_encoder> |
|
model: !ref <model> |
|
paths: |
|
wav2vec2: !ref <pretrained_path>/wav2vec2.ckpt |
|
classifier: !ref <pretrained_path>/classifier.ckpt |
|
label_encoder: !ref <pretrained_path>/label_encoder.txt |
|
model: !ref <pretrained_path>/model.ckpt |
|
|
|
|
|
|
|
softmax: !new:speechbrain.nnet.activations.Softmax |
|
apply_log: True |
