Skip to content

Conversation

Kacper-Pietkun
Copy link
Contributor

I added tests for custom ops defined in vllm_gaudi/ops:

  • For the tests of ops that are not using cuda kernels - native ops and hpu ops are triggered for the same input and their outputs are compared
  • For others tests that are using cuda kernels (so cannot be called with vllm-gaudi plugin) I created separate directory to store some predefined small tensors (to avoid hardcoding them in tests) - weights, inputs and outputs. Such tests run hpu ops with loaded inputs and weights and compare their outputs with the loaded outputs.

Signed-off-by: Kacper Pietkun <[email protected]>
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