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

Add ImageNet ResNet Workload Variant #594

Merged
merged 6 commits into from
Dec 8, 2023

Conversation

pomonam
Copy link
Contributor

@pomonam pomonam commented Dec 6, 2023

The PR adds 3 workload variants for ImageNet:

  1. Change all ReLU activations to SiLU.
  2. Change all ReLU activations to GELU.
  3. Change the scale of the last BN initialization (for residual block) to 8.

In addition, I added modeldiff tests for the first two variants, and they passed.

@pomonam pomonam requested a review from a team as a code owner December 6, 2023 21:48
Copy link

github-actions bot commented Dec 6, 2023

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅

@pomonam pomonam requested a review from priyakasimbeg December 6, 2023 21:49
@priyakasimbeg
Copy link
Contributor

priyakasimbeg commented Dec 7, 2023

Thanks for getting this in! Just one request; could you add the variant workload names to this list in docker/scripts/startup.sh please?

@pomonam
Copy link
Contributor Author

pomonam commented Dec 8, 2023

Thanks for getting this in! Just one request; could you add the variant workload names to this list in docker/scripts/startup.sh please?

Sure thing! I just added the names in the docker setup.

@pomonam pomonam requested a review from runame December 8, 2023 16:42
@priyakasimbeg priyakasimbeg merged commit 64d1a85 into dev Dec 8, 2023
31 checks passed
@github-actions github-actions bot locked and limited conversation to collaborators Dec 8, 2023
@pomonam pomonam deleted the juhan/imagenet_resnet_variant branch December 9, 2023 04:25
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants