Skip to content
New issue

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

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

Already on GitHub? Sign in to your account

Implement BatchNormalization functions from cuDNN in juice #146

Draft
wants to merge 1 commit into
base: master
Choose a base branch
from

Conversation

colepoirier
Copy link

What does this PR accomplish?

  • 🦚 Feature

Closes #145.

Changes proposed by this PR:

  • Adding the necessary wrapper functions of the ffi::* cuDNN BatchNormaliztion functions to rcudnn/cudnn/src/api/.
  • Adding the corresponding functions to coaster-nn.
  • Adding the corresponding layers to juice.

Notes to reviewer:

I think that at least one additional new function https://docs.nvidia.com/deeplearning/cudnn/api/index.html#cudnnDeriveBNTensorDescriptor needs to be added in order to add the BatchNormalization functions to juice as: "This function derives a secondary tensor descriptor for the batch normalization scale, invVariance, bnBias, and bnScale subtensors from the layer's x data descriptor."

I am not sure how to proceed with this in terms of how to properly add this/implement it in rcudnn.

📜 Checklist

  • Test coverage is excellent
  • All unit tests pass
  • The juice-examples run just fine
  • Documentation is thorough, extensive and explicit

…i_batch_normalization_forward_training(), but now with a Result return type
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.

Implement BatchNormalization2d Layer
1 participant