Enable 16-bit activations in Cadence Quantizer For fully_connected and linear #15010
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
meta-cla
bot
added
the
CLA Signed
This PR needs a
|
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/15010
Note: Links to docs will display an error until the docs builds have been completed. ❌ 2 New FailuresAs of commit 388201c with merge base a12219d ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
RahulC7
force-pushed
the
export-D84284794
branch
from
ee7a0f7 to
408cbcd
Compare
RahulC7
added a commit
to RahulC7/executorch
that referenced
this pull request
Oct 13, 2025
…d linear (pytorch#15010) Summary: # Context We currently only support 8-bit for most operators. We would like to add generic ops for 16-bit activations, for the following ops: - quantized_fully_connected - quantized_linear - quantized_conv (all flavors) - quantized_matmul # This Diff Here, we add support for `quantized_linear` and `quantized_fully_connected`. We need to do the following: 1. Allow 16-bit activations in `quantized_fully_connected_out.cpp` and `quantized_linear_out.cpp`. 2. Allow 16-bit activations in `ref_implementations.py`, so tests can run with 16-bit activations to validate the quantization is correct. 3. Add a quantizer(`CadenceWith16BitLinearActivationsQuantizer`) for checking this works and create a unit test. Reviewed By: hsharma35 Differential Revision: D84284794
DrJessop
approved these changes
Oct 14, 2025
…d linear (pytorch#15010) Summary: # Context We currently only support 8-bit for most operators. We would like to add generic ops for 16-bit activations, for the following ops: - quantized_fully_connected - quantized_linear - quantized_conv (all flavors) - quantized_matmul # This Diff Here, we add support for `quantized_linear` and `quantized_fully_connected`. We need to do the following: 1. Allow 16-bit activations in `quantized_fully_connected_out.cpp` and `quantized_linear_out.cpp`. 2. Allow 16-bit activations in `ref_implementations.py`, so tests can run with 16-bit activations to validate the quantization is correct. 3. Add a quantizer(`CadenceWith16BitLinearActivationsQuantizer`) for checking this works and create a unit test. Reviewed By: DrJessop, hsharma35 Differential Revision: D84284794
RahulC7
force-pushed
the
export-D84284794
branch
from
408cbcd to
388201c
Compare
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:
Context
We currently only support 8-bit for most operators. We would like to add generic ops for 16-bit activations, for the following ops:
This Diff
Here, we add support for
quantized_linearandquantized_fully_connected. We need to do the following:quantized_fully_connected_out.cppandquantized_linear_out.cpp.ref_implementations.py, so tests can run with 16-bit activations to validate the quantization is correct.CadenceWith16BitLinearActivationsQuantizer) for checking this works and create a unit test.Differential Revision: D84284794