search

sccn / sound2meg

2 stars 1 forks source link

Higher learning rate than paper's #3

Open dungscout96 opened 1 year ago

dungscout96 commented 1 year ago

We are using lr=3*10^-5 https://github.com/sccn/sound2meg/blob/db17dbf5055a45fb80001c52250ade6919b84494/reimplementation_training.py#LL149C57-L149C62

@makhalid1999 Can you reduce it to 3*10^-4 before running the experiment again?

makhalid1999 commented 1 year ago

Have updated that, and submitted to NSG, to see how it goes...