b04901014 / FT-w2v2-ser

Official implementation for the paper Exploring Wav2vec 2.0 fine-tuning for improved speech emotion recognition

Seems to be nn.LSTM is useless

Gpwner opened this issue 3 years ago · comments

Gpwner commented 3 years ago

I don't see any usage of https://github.com/b04901014/FT-w2v2-ser/blob/main/pretrain/trainer.py#L252 ,it seems to be useless.

Li-Wei Chen commented 3 years ago

Yes, it's useless, not ran anywhere in the forward pass. We tried it during experimental stage but didn't actually use it eventually.