Skip to content
New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

Use NNlib.conv_bias_act for Conv #1832

Draft
wants to merge 2 commits into
base: master
Choose a base branch
from

Conversation

darsnack
Copy link
Member

This closes #1302 (supersedes) and fixes #558.

PR Checklist

  • Tests are added
  • Entry in NEWS.md
  • Documentation, if applicable
  • API changes require approval from a committer (different from the author, if applicable)

@mcabbott
Copy link
Member

Xref FluxML/NNlib.jl#346, which adds gradient definitions for conv_bias_act (among other things).

@mcabbott
Copy link
Member

mcabbott commented Feb 5, 2022

Maybe worth trying to do for v0.13?

@darsnack
Copy link
Member Author

darsnack commented Feb 6, 2022

Yeah I can take a look tomorrow. What's unclear to me is whether we need some part of the NNlib PR that you linked. Currently, I don't know what code path the gradient takes for CuArrays.

@mcabbott
Copy link
Member

mcabbott commented Feb 6, 2022

I think there is no gradient defined at all without that PR. It may be close to done though.

# for free to join this conversation on GitHub. Already have an account? # to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Bottleneck in mapreducedim for convolutional layers
2 participants