Skip to content

Conversation

@srinadhu
Copy link

@srinadhu srinadhu commented Jul 10, 2019

@emited Please check and let me know.

  1. Added clipping before the optimizer step.
  2. Updated with .item() rather than .data[0]
  3. Input is normalized but the reconstruction is done for the unnormalized image so this is removed
  4. Loss get's NAN sometimes and so added 1e-5 as epsilon in the std2 denominator

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant