File size: 2,797 Bytes
01a6e6b 801e44e 01a6e6b 801e44e 01a6e6b 801e44e 01a6e6b 801e44e 01a6e6b 801e44e 01a6e6b 801e44e 01a6e6b 801e44e 01a6e6b 801e44e 01a6e6b 801e44e 01a6e6b 801e44e 01a6e6b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 |
pretrained_path: sinarashidi/s2st_fa-en_cvss
sample_rate: 16000
# URL for the HuggingFace model we want to load as encoder
wav2vec2_hub: m3hrdadfi/wav2vec2-large-xlsr-persian-v3
# Outputs
vocab_size: 103
blank_index: 102
bos_index: 100
eos_index: 101
pad_index: 102
label_smoothing: 0.0
# Encoder
features_dim: 1024
# Length Regulator
enc_kernel_size: 3
enc_stride: 2
# Transformer decoder
embedding_size: 512
d_model: 512
nhead: 8
num_encoder_layers: 0
num_decoder_layers: 6
d_ffn: 2048
transformer_dropout: 0.1
activation: !name:torch.nn.GELU
output_neurons: !ref <vocab_size>
attention_type: "RelPosMHAXL"
# Decoding parameters
min_decode_ratio: 0.0
max_decode_ratio: 1.0
wav2vec2: !new:speechbrain.lobes.models.huggingface_transformers.wav2vec2.Wav2Vec2
source: !ref <wav2vec2_hub>
output_norm: True
freeze: True
freeze_feature_extractor: True
apply_spec_augment : True
save_path: wav2vec2_checkpoints
length_regulator: !new:speechbrain.nnet.CNN.Conv1d
input_shape: [null, null, !ref <features_dim>]
out_channels: !ref <embedding_size>
kernel_size: !ref <enc_kernel_size>
stride: !ref <enc_stride>
transformer_decoder: !new:speechbrain.lobes.models.transformer.TransformerST.TransformerST # yamllint disable-line rule:line-length
input_size: !ref <embedding_size>
tgt_vocab: !ref <output_neurons>
d_model: !ref <d_model>
nhead: !ref <nhead>
num_encoder_layers: !ref <num_encoder_layers>
num_decoder_layers: !ref <num_decoder_layers>
d_ffn: !ref <d_ffn>
dropout: !ref <transformer_dropout>
activation: !ref <activation>
attention_type: !ref <attention_type>
normalize_before: True
causal: False
log_softmax: !new:speechbrain.nnet.activations.Softmax
apply_log: True
seq_lin: !new:speechbrain.nnet.linear.Linear
input_size: !ref <d_model>
n_neurons: !ref <output_neurons>
model: !new:torch.nn.ModuleList
- [!ref <length_regulator>, !ref <transformer_decoder>, !ref <seq_lin>]
encoder: !new:speechbrain.nnet.containers.LengthsCapableSequential
wav2vec2: !ref <wav2vec2>
length_regulator: !ref <length_regulator>
decoder_beamsearch: !new:speechbrain.decoders.seq2seq.S2STransformerBeamSearcher
modules: [!ref <transformer_decoder>, !ref <seq_lin>]
bos_index: !ref <bos_index>
eos_index: !ref <eos_index>
min_decode_ratio: !ref <min_decode_ratio>
max_decode_ratio: !ref <max_decode_ratio>
beam_size: 10
temperature: 1.0
modules:
encoder: !ref <encoder>
decoder: !ref <decoder_beamsearch>
pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
loadables:
model: !ref <model>
wav2vec2: !ref <wav2vec2>
paths:
wav2vec2: !ref <pretrained_path>/wav2vec2.ckpt
model: !ref <pretrained_path>/model.ckpt
|