Repository navigation
honor per-model packed-LHS disable in XNNPACK (#24245) - #24245
Conversation
meta-cla
Bot
added
the
CLA Signed
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/24245
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ✅ You can merge normally! (2 Unrelated Failures)As of commit 7081e1f with merge base 14b5e1d ( BROKEN TRUNK - The following jobs failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
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 D117888679. |
This PR needs a
|
meta-codesync
Bot
changed the title
Summary: Add the `disable_packed_lhs` ExecuTorch XNNPACK backend option with a false global default and per-model runtime-spec overrides. Pass the resolved settings through `XNNCompilerOptions`. When the internal XNNPACK patch is available, set `XNN_FLAG_NO_PACKED_LHS` to preserve explicit dynamic quantization; builds without that flag retain their normal runtime flags and log enabled requests as unsupported. Implementation and tests updated with Codex. Differential Revision: D117888679
34eecf0 to
dfa3003
Compare
digantdesai
left a comment
digantdesai
left a comment
There was a problem hiding this comment.
Review automatically exported from Phabricator review in Meta.
Summary: Add the `disable_packed_lhs` ExecuTorch XNNPACK backend option with a false global default and per-model runtime-spec overrides. Pass the resolved settings through `XNNCompilerOptions`. When the internal XNNPACK patch is available, set `XNN_FLAG_NO_PACKED_LHS` to preserve explicit dynamic quantization; builds without that flag retain their normal runtime flags and log enabled requests as unsupported. Reviewed By: digantdesai Differential Revision: D117888679
dfa3003 to
7081e1f
Compare
Summary:
Add the
disable_packed_lhsExecuTorch XNNPACK backend option with a false global default and per-model runtime-spec overrides. Pass the resolved settings throughXNNCompilerOptions. When the internal XNNPACK patch is available, setXNN_FLAG_NO_PACKED_LHSto preserve explicit dynamic quantization; builds without that flag retain their normal runtime flags and log enabled requests as unsupported.Reviewed By: digantdesai
Differential Revision: D117888679