Fix optional bias and batch handling in cadence::fully_connected (#19194)#19194
Open
hsharma35 wants to merge 2 commits intopytorch:mainfrom
Open
Fix optional bias and batch handling in cadence::fully_connected (#19194)#19194hsharma35 wants to merge 2 commits intopytorch:mainfrom
hsharma35 wants to merge 2 commits intopytorch:mainfrom
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/19194
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: ❌ 1 New Failure, 2 Unrelated FailuresAs of commit a1d8229 with merge base 5a206ab ( NEW FAILURE - The following job has failed:
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. |
meta-cla
Bot
added
the
CLA Signed
Contributor
|
@hsharma35 has exported this pull request. If you are a Meta employee, you can view the originating Diff in D102821213. |
This PR needs a
|
mcremon-meta
approved these changes
Apr 28, 2026
meta-codesync
Bot
changed the title
hsharma35
added a commit
to hsharma35/executorch
that referenced
this pull request
Apr 28, 2026
…orch#19194) Summary: Fixes two bugs in the generic and HiFi cadence::fully_connected implementations. First, the optional bias was dereferenced without a has_value() guard, causing a crash for bias-free inputs. Second, only the first input row was computed because the batch loop was missing; a loop over leading_dims (the product of all non-channel input dimensions) is now added to correctly process batched and multi-sequence inputs. Reviewed By: mcremon-meta Differential Revision: D102821213
hsharma35
force-pushed
the
export-D102821213
branch
from
8246093 to
506d3b8
Compare
…kernels (pytorch#19193) Summary: PR pytorch#19193 Fixes two correctness bugs in the HiFi kernels for cadence::quantized_conv1d_ncl.out and cadence::quantized_conv1d_nlc.out. The int8 path (xa_nn_conv2d_per_chan_sym8sxasym8s) produces incorrect results with stride > 1 on some backends (e.g., Artemis HiFi4) and is now redirected to the generic fallback for that case. The uint8 path overflowed WORD32 when computing out_multiplier32 if eff_scale >= 1.0 (i.e., output_scale > bias_scale), which is now clamped to INT32_MAX. Reviewed By: zonglinpeng Differential Revision: D102821209
…orch#19194) Summary: Fixes two bugs in the generic and HiFi cadence::fully_connected implementations. First, the optional bias was dereferenced without a has_value() guard, causing a crash for bias-free inputs. Second, only the first input row was computed because the batch loop was missing; a loop over leading_dims (the product of all non-channel input dimensions) is now added to correctly process batched and multi-sequence inputs. Reviewed By: mcremon-meta Differential Revision: D102821213
hsharma35
force-pushed
the
export-D102821213
branch
from
506d3b8 to
a1d8229
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary:
Fixes two bugs in the generic and HiFi cadence::fully_connected implementations. First, the optional bias was dereferenced without a has_value() guard, causing a crash for bias-free inputs. Second, only the first input row was computed because the batch loop was missing; a loop over leading_dims (the product of all non-channel input dimensions) is now added to correctly process batched and multi-sequence inputs.
Reviewed By: mcremon-meta
Differential Revision: D102821213