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

Preprocess Conv2D weights on Device #18272

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

Conversation

sankarmanoj-tt
Copy link
Contributor

@sankarmanoj-tt sankarmanoj-tt commented Feb 25, 2025

Ticket

Link to Github Issue

Problem description

Provide context for the problem.

What's changed

Describe the approach used to solve the problem.
Summarize the changes made and its impact.

Checklist

@sankarmanoj-tt sankarmanoj-tt force-pushed the smanoj/conv_device_weights branch from 08ff52d to 4d5afb1 Compare February 25, 2025 06:50
#0: WIP Conv device weights

#0: WIP Conv device weights

#0: Conv device weights

#0: 80% pass for loading weights on device

#0: Shallow conv support

#0: rebase fix

#0: Fix pad by using multicore

#0: Fix pad by using multicore

#0: Fix OOM for pad

#0: Fix device weights

#0: Re-enable tests

#0: Re-enable tests

#0: Re-enable tests

#0: Fix OOM for pad

#0: Build fix

#0: Build fix

#0: Re-enable transpose shards for Conv2D Unit Tests

#0: Tests fix

#0: Tests fix

#0: Rebase fi

#0: Tests fix

#0: Skip weights bfloat8 on grayskull

#0: Reverted types

#0: Add flag for always preprocessing weights

#0: Preprocess bias on device

#0: Fix conv bias

#0: Rebase fix

#0: Rebase fix

#0: Bug fix

#0: Skip test on N300

#18185: Change order of pad & permute

#0: Fix sweep

#0: Changed default for preprocess weights on device to false
@sankarmanoj-tt sankarmanoj-tt force-pushed the smanoj/conv_device_weights branch from 4d5afb1 to 28c6b78 Compare March 2, 2025 13:08
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