Handle Byte scalar type in quantized conv2d NHWC (#18427)#18427
Open
khazaei wants to merge 2 commits intopytorch:mainfrom
Open
Handle Byte scalar type in quantized conv2d NHWC (#18427)#18427khazaei wants to merge 2 commits intopytorch:mainfrom
khazaei wants to merge 2 commits intopytorch:mainfrom
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/18427
Note: Links to docs will display an error until the docs builds have been completed. ❌ 2 New Failures, 2 Unrelated FailuresAs of commit 3757d54 with merge base 45a9717 ( NEW FAILURES - The following jobs have 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
This PR needs a
|
meta-codesync
bot
changed the title
khazaei
force-pushed
the
export-D97819850
branch
from
478d713 to
6940ed5
Compare
khazaei
pushed a commit
to khazaei/executorch
that referenced
this pull request
Mar 24, 2026
Summary: Extend the quantized conv2d NHWC operator to accept `ScalarType::Byte` (uint8) inputs in addition to `ScalarType::Char` (int8), enabling unsigned 8-bit quantized inference through the same HiFi code path. Differential Revision: D97819850
khazaei
force-pushed
the
export-D97819850
branch
from
6940ed5 to
d77c9f0
Compare
khazaei
pushed a commit
to khazaei/executorch
that referenced
this pull request
Mar 24, 2026
Summary: Extend the quantized conv2d NHWC operator to accept `ScalarType::Byte` (uint8) inputs in addition to `ScalarType::Char` (int8), enabling unsigned 8-bit quantized inference through the same HiFi code path. Differential Revision: D97819850
khazaei
force-pushed
the
export-D97819850
branch
from
d77c9f0 to
00878ec
Compare
khazaei
pushed a commit
to khazaei/executorch
that referenced
this pull request
Mar 24, 2026
Summary: Pull Request resolved: pytorch#18427 Extend the quantized conv2d NHWC operator to accept `ScalarType::Byte` (uint8) inputs in addition to `ScalarType::Char` (int8), enabling unsigned 8-bit quantized inference through the same HiFi code path. Differential Revision: D97819850
khazaei
force-pushed
the
export-D97819850
branch
from
00878ec to
a9b575d
Compare
…ouped convolutions Summary: Rearrange weights from [OC, KH, KW] to [KH, KW, OC] for depthwise conv2d when `groups == input_channels` . The NNLib depthwise function (`xa_nn_conv2d_depthwise_per_chan_sym8sxasym8s`) expects weights in [KH, KW, OC] layout, but when the weight tensor is 4D [OC, KH, KW, IC/G], it arrives in [OC, KH, KW] order. A temporary buffer is allocated via `allocate_temp_memory` to hold the rearranged weights, which are then passed to the depthwise kernel instead of the original `p_kernel`. Also updates `test_custom_ops.py` to target the `Artemis_HiFi4_UT_v3` backend. Differential Revision: D97833420
khazaei
pushed a commit
to khazaei/executorch
that referenced
this pull request
Mar 24, 2026
Summary: Extend the quantized conv2d NHWC operator to accept `ScalarType::Byte` (uint8) inputs in addition to `ScalarType::Char` (int8), enabling unsigned 8-bit quantized inference through the same HiFi code path. Differential Revision: D97819850
khazaei
force-pushed
the
export-D97819850
branch
from
a9b575d to
157ad59
Compare
Summary: Pull Request resolved: pytorch#18427 Extend the quantized conv2d NHWC operator to accept `ScalarType::Byte` (uint8) inputs in addition to `ScalarType::Char` (int8), enabling unsigned 8-bit quantized inference through the same HiFi code path. Differential Revision: D97819850
khazaei
force-pushed
the
export-D97819850
branch
from
157ad59 to
3757d54
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
1 participant
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:
Extend the quantized conv2d NHWC operator to accept
ScalarType::Byte(uint8) inputs in addition toScalarType::Char(int8), enabling unsigned 8-bit quantized inference through the same HiFi code path.Differential Revision: D97819850