-
Notifications
You must be signed in to change notification settings - Fork 667
[ET-VK] Add optional blocklist and allowlist to vulkan partitioner to aid debugging #13326
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
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/13326
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New Failure, 1 PendingAs of commit fb3083d with merge base 71c9a4f ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This PR needs a
|
… aid debugging (pytorch#13326) Summary: ## Changes * Add `operator_allowlist` and `operator_blocklist` optional arguments to `VulkanPartitioner` * `operator_blocklist` will prevent operators in the block list to be lowered to Vulkan * `operator_allowlist` will only allow operators in the allow list to be lowered to Vulkan * `operator_allowlist` takes precedence over `operator_blocklist` ## Context When debugging models, it is useful to be able to prevent certain operators from being lowered to Vulkan, or to only allow certain operators from being lowered to Vulkan. This can help isolate which ops are causing model output to be incorrect. Test Plan: ## Test Plan Tested this feature locally while debugging example models. Co-authored-by: ssjia <[email protected]>
Summary:
Changes
operator_allowlist
andoperator_blocklist
optional arguments toVulkanPartitioner
operator_blocklist
will prevent operators in the block list to be lowered to Vulkanoperator_allowlist
will only allow operators in the allow list to be lowered to Vulkanoperator_allowlist
takes precedence overoperator_blocklist
Context
When debugging models, it is useful to be able to prevent certain operators from being lowered to Vulkan, or to only allow certain operators from being lowered to Vulkan. This can help isolate which ops are causing model output to be incorrect.
Test Plan:
Test Plan
Tested this feature locally while debugging example models.