search

francoisgermain / SpeechDenoisingWithDeepFeatureLosses

Speech Denoising with Deep Feature Losses
MIT License
183 stars 53 forks source link

How can i increase the batch size? #10

Open byuns9334 opened 5 years ago

byuns9334 commented 5 years ago

I can't find any information about batch size on the code, and the paper. How can I increase the batch size on the source code? And what is the default setting for batch size in current code? Thanks!

Duffyxx commented 4 years ago

The model input is the whole speech wav array, so it is hard to keep every speech file to be the same length, and hard to running on bigger batch size, the best way to increase batch size I think is to clip speech to frames and feed the same length array into network.