optimizer.zero_grad() - Githubissues

sacmehta / ESPNet

ESPNet: Efficient Spatial Pyramid of Dilated Convolutions for Semantic Segmentation

https://sacmehta.github.io/ESPNet/

MIT License

538 stars 111 forks source link

Closed erichhhhho closed 5 years ago

erichhhhho commented 5 years ago

Hi,

I think you zero the grad two time during training. Here and Here

sacmehta commented 5 years ago

Thanks for noting it. You can remove one of them. It should not impact anything.

Also, if our work excites you, we have released a new repository. Have a look.