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

Jax input tensor test variations #209

Open
nsmithtt opened this issue Jan 29, 2025 · 2 comments
Open

Jax input tensor test variations #209

nsmithtt opened this issue Jan 29, 2025 · 2 comments
Assignees
Labels
multichip Multichip issues

Comments

@nsmithtt
Copy link

Integrate the tests from Tap's repo into tt-xla. Feel free to update this issue as it makes sense.

As I understand it there are currently 2 flavors of inputs:

  • Manual sharding (annotated graph) + no input sharding
  • Manual sharding (annotated graph) + input sharding via device put & mesh attribute
@nsmithtt
Copy link
Author

We should mark the unsupported flavors as just unsupported for now. I've filed a follow on issue for supporting the second bullet:

@tapspatel
Copy link

I added @ajakovljevicTT to my repo and shared all the test cases. Please let me know if you need anything else from my end. As well, I iterate through many different meshes, but we will only be supporting the following in runtime

N300: 1x2 not supported: 2x1​
T3K: 1x8, 2x4 not supported: 4x2, 8x1​
TG: 1x32, 8x4 not supported: 2x16, 4x8, 6x2, 32x1

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
multichip Multichip issues
Projects
None yet
Development

No branches or pull requests

3 participants