Giters

santi-pdp / pase

Problem Agnostic Speech Encoder

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Fine tune the pre-trained weights

MittalShruti opened this issue · comments

commented

Hey

I want to import the pre-trained weights and then fine tune on a different language speech data. The repo has steps to train the model from scratch.

I want to fine-tune the already trained encoder-workers model on my speech data, so that I can get better features. Importing just the encoder as nn.Module, won't let me use the workers, and I won't be able to extract features the (self/)unsupervised learning way.

How should I do it?

[EDIT] Also, e.g., I trained the PASE+ model on dataset1. Now after training, get more speech data for training (dataset2). In such a case, I would want to 'fine-tune' the model trained on dataset1, using this new dataset2. And not want to train from scratch on dataset2 + dataset1.