[PyTorch] Fix quantized Conv1d module parameters #62356
Closed
Conversation
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
🔗 Helpful links
💊 CI failures summary and remediationsAs of commit f860322 (more details on the Dr. CI page):
1 job timed out:
🚧 1 fixed upstream failure:These were probably caused by upstream breakages that were already fixed.
Please rebase on the
|
|
This pull request was exported from Phabricator. Differential Revision: D29957556 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D29957556 |
larryliu0820
force-pushed
the
export-D29957556
branch
from
08540c6 to
16196bc
Compare
jbschlosser
requested review from
vkuzo and
z-a-f
and removed request for
jbschlosser
|
This pull request was exported from Phabricator. Differential Revision: D29957556 |
larryliu0820
force-pushed
the
export-D29957556
branch
from
16196bc to
045873d
Compare
|
This pull request was exported from Phabricator. Differential Revision: D29957556 |
larryliu0820
force-pushed
the
export-D29957556
branch
from
045873d to
756a069
Compare
Summary: Pull Request resolved: pytorch#62356 In `torch/nn/quantized/module/conv.py`, Conv1d is making a scaler `kernel_size` into a tuple with size 2 by repeating `kernel_size` value. This logic is breaking `Conv1d` because internally it's unsqueezing the input with shape N, C, L to N, C, 1, L in [`qconv.cpp`](https://github.com/pytorch/pytorch/blob/06dfaadfc6357ed909ed15c7ef79d503c49d9475/aten/src/ATen/native/quantized/cpu/qconv.cpp#L841). Applying aforementioned kernel to this input shape will produce negative output shape in [`ConvUtils.h`](https://github.com/pytorch/FBGEMM/blob/203f7ff6e07d62b042e7d755fd1f4789d978e4d1/include/fbgemm/ConvUtils.h#L118-L119), if kernel_size > 1. Here I'm modifying the processing logic for `kernel_size` and a few other parameters, to follow the pattern of [`torch/nn/module/conv.py`](https://github.com/pytorch/pytorch/blob/aae2a3c95ee6d62e834a5e6890a12f7ecf0dd17f/torch/nn/modules/conv.py#L284-L287). Test Plan: Rely on unit test Reviewed By: kimishpatel Differential Revision: D29957556 fbshipit-source-id: be33a3ce516788125f750274b527d27a57ff8a8e
|
This pull request was exported from Phabricator. Differential Revision: D29957556 |
larryliu0820
force-pushed
the
export-D29957556
branch
from
756a069 to
f860322
Compare
|
This pull request has been merged in 70f57bc. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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:
In
torch/nn/quantized/module/conv.py, Conv1d is making a scalerkernel_sizeinto a tuple with size 2 by repeatingkernel_sizevalue. This logic is breakingConv1dbecause internally it's unsqueezing the input with shape N, C, L to N, C, 1, L inqconv.cpp. Applying aforementioned kernel to this input shape will produce negative output shape inConvUtils.h, if kernel_size > 1.Here I'm modifying the processing logic for
kernel_sizeand a few other parameters, to follow the pattern oftorch/nn/module/conv.py.Test Plan: Rely on unit test
Differential Revision: D29957556