Vikrant Sharma

icon company Reverie Language Technologies icon location Bangalore

Results 2 comments of Vikrant Sharma

Attention doesn't work well for downsample_step=1 and outputs_per_step=1

Hello @r9y9 Any update on this? `downsample_step=1` fails to learn, but works fine with 4.

ValueError: optimizing a parameter that doesn't require gradients

Using `torch==0.4.1` seems to resolve this. Not sure yet if this would affect anything else later on.