Adel-Moumen
commited on
Commit
•
9035a95
1
Parent(s):
c4fc6d1
Delete hyperparams_develop.yaml
Browse files- hyperparams_develop.yaml +0 -68
hyperparams_develop.yaml
DELETED
@@ -1,68 +0,0 @@
|
|
1 |
-
# ################################
|
2 |
-
# Model: wav2vec2 + DNN + CTC
|
3 |
-
# Augmentation: SpecAugment
|
4 |
-
# Authors: Sung-Lin Yeh 2021, Titouan Parcollet 2022
|
5 |
-
# ################################
|
6 |
-
|
7 |
-
# URL for the biggest Fairseq english wav2vec2 model.
|
8 |
-
wav2vec2_hub: facebook/wav2vec2-large-960h-lv60-self
|
9 |
-
sample_rate: 16000
|
10 |
-
|
11 |
-
# Model parameters
|
12 |
-
activation: !name:torch.nn.LeakyReLU
|
13 |
-
dnn_layers: 2
|
14 |
-
dnn_neurons: 1024
|
15 |
-
freeze_wav2vec: True
|
16 |
-
|
17 |
-
# Outputs
|
18 |
-
output_neurons: 31 # BPE size, index(blank/eos/bos) = 0
|
19 |
-
|
20 |
-
# Decoding parameters
|
21 |
-
blank_index: 0
|
22 |
-
bos_index: 1
|
23 |
-
eos_index: 2
|
24 |
-
|
25 |
-
enc: !new:speechbrain.lobes.models.VanillaNN.VanillaNN
|
26 |
-
input_shape: [null, null, 1024]
|
27 |
-
activation: !ref <activation>
|
28 |
-
dnn_blocks: !ref <dnn_layers>
|
29 |
-
dnn_neurons: !ref <dnn_neurons>
|
30 |
-
|
31 |
-
wav2vec2: !new:speechbrain.lobes.models.huggingface_transformers.Wav2Vec2
|
32 |
-
source: !ref <wav2vec2_hub>
|
33 |
-
output_norm: True
|
34 |
-
freeze: True
|
35 |
-
save_path: model_checkpoints
|
36 |
-
|
37 |
-
ctc_lin: !new:speechbrain.nnet.linear.Linear
|
38 |
-
input_size: !ref <dnn_neurons>
|
39 |
-
n_neurons: !ref <output_neurons>
|
40 |
-
|
41 |
-
log_softmax: !new:speechbrain.nnet.activations.Softmax
|
42 |
-
apply_log: True
|
43 |
-
|
44 |
-
ctc_cost: !name:speechbrain.nnet.losses.ctc_loss
|
45 |
-
blank_index: !ref <blank_index>
|
46 |
-
|
47 |
-
asr_model: !new:torch.nn.ModuleList
|
48 |
-
- [!ref <enc>, !ref <ctc_lin>]
|
49 |
-
|
50 |
-
tokenizer: !new:speechbrain.dataio.encoder.CTCTextEncoder
|
51 |
-
|
52 |
-
encoder: !new:speechbrain.nnet.containers.LengthsCapableSequential
|
53 |
-
wav2vec2: !ref <wav2vec2>
|
54 |
-
enc: !ref <enc>
|
55 |
-
ctc_lin: !ref <ctc_lin>
|
56 |
-
log_softmax: !ref <log_softmax>
|
57 |
-
|
58 |
-
decoding_function: !name:speechbrain.decoders.ctc_greedy_decode
|
59 |
-
blank_id: !ref <blank_index>
|
60 |
-
|
61 |
-
modules:
|
62 |
-
encoder: !ref <encoder>
|
63 |
-
|
64 |
-
pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
|
65 |
-
loadables:
|
66 |
-
wav2vec2: !ref <wav2vec2>
|
67 |
-
asr: !ref <asr_model>
|
68 |
-
tokenizer: !ref <tokenizer>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|