Repository navigation
Prepack stored inputs of clones and clamps (#24436) - #24436
JakeStevens wants to merge 1 commit into
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/24436
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New Failure, 3 Pending, 2 Unclassified FailuresAs of commit 406d2e9 with merge base cb3f24e ( NEW FAILURE - The following job has failed:
UNCLASSIFIED FAILURES - DrCI could not classify the following jobs because the workflow did not run on the merge base. The failures may be pre-existing on trunk or introduced by this PR:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
@JakeStevens has exported this pull request. If you are a Meta employee, you can view the originating Diff in D129088169. |
This PR needs a
|
Summary: ## Ulterior Motive Unblock Swin V2's stored query/key/value biases and attention logit scales. ## Rationale **What**: Add `aten.clone.default` and `aten.clamp.default` to the existing stored-input prepack list. **Why**: Clone expects a tensor rather than a TensorRef. Clamp also needs a bound input resource; leaving its stored input as a reference can leave the shader descriptor unwritten. ## Details - Reuse the existing explicit prepack path without changing either kernel. - Cover both operators in the prepack pass and numerical engine tests. - Preserve the existing mutation guards and all nonstored-input paths. Authored with AI using Metamate on agentcloud. Differential Revision: D129088169
9077856 to
406d2e9
Compare
Summary:
Ulterior Motive
Unblock Swin V2's stored query/key/value biases and attention logit scales.
Rationale
What: Add
aten.clone.defaultandaten.clamp.defaultto the existing stored-input prepack list.Why: Clone expects a tensor rather than a TensorRef. Clamp also needs a bound input resource; leaving its stored input as a reference can leave the shader descriptor unwritten.
Details
Authored with AI using Metamate on agentcloud.
Differential Revision: D129088169