Normalization of Error (f) and grad param

torch / demos

Demos and tutorials around Torch7.

355 stars 301 forks source link

Closed varghesealex90 closed 7 years ago

varghesealex90 commented 7 years ago

Hello:

Sorry for the noob question, isn't it necessary to divide f and grad param by the batch size ? any reason why it is not done here?

Regards Varghese

varghesealex90 commented 7 years ago

Sorry I got the answer . Thanks