hyperparams.yaml
1.6 KB · 60 lines · yaml Raw
1 # ############################################################################
2 # Model: WAV2VEC base for Emotion Recognition
3 # ############################################################################
4
5
6 # Hparams NEEDED
7 HPARAMS_NEEDED: ["encoder_dim", "out_n_neurons", "label_encoder", "softmax"]
8 # Modules Needed
9 MODULES_NEEDED: ["wav2vec2", "avg_pool", "output_mlp"]
10
11 # Feature parameters
12 wav2vec2_hub: facebook/wav2vec2-base
13
14 # Pretrain folder (HuggingFace)
15 pretrained_path: speechbrain/emotion-recognition-wav2vec2-IEMOCAP
16
17 # parameters
18 encoder_dim: 768
19 out_n_neurons: 4
20
21 wav2vec2: !new:speechbrain.lobes.models.huggingface_transformers.wav2vec2.Wav2Vec2
22 source: !ref <wav2vec2_hub>
23 output_norm: True
24 freeze: True
25 save_path: wav2vec2_checkpoints
26
27 avg_pool: !new:speechbrain.nnet.pooling.StatisticsPooling
28 return_std: False
29
30 output_mlp: !new:speechbrain.nnet.linear.Linear
31 input_size: !ref <encoder_dim>
32 n_neurons: !ref <out_n_neurons>
33 bias: False
34
35 model: !new:torch.nn.ModuleList
36 - [!ref <output_mlp>]
37
38 modules:
39 wav2vec2: !ref <wav2vec2>
40 output_mlp: !ref <output_mlp>
41 avg_pool: !ref <avg_pool>
42
43 softmax: !new:speechbrain.nnet.activations.Softmax
44
45
46 label_encoder: !new:speechbrain.dataio.encoder.CategoricalEncoder
47
48
49 pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
50 loadables:
51 wav2vec2: !ref <wav2vec2>
52 model: !ref <model>
53 label_encoder: !ref <label_encoder>
54 paths:
55 wav2vec2: !ref <pretrained_path>/wav2vec2.ckpt
56 model: !ref <pretrained_path>/model.ckpt
57 label_encoder: !ref <pretrained_path>/label_encoder.txt
58
59
60