Adding mixed quantization support #14134
Merged
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🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/14134
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 5188058 with merge base 181ed4d ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D81519735 |
This PR needs a
|
mgiordy
force-pushed
the
export-D81519735
branch
from
1e64ab6 to
789bf47
Compare
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 10, 2025
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Differential Revision: D81519735
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D81519735 |
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 10, 2025
Summary: Pull Request resolved: pytorch#14134 # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Differential Revision: D81519735
mgiordy
force-pushed
the
export-D81519735
branch
2 times, most recently
from
2ffaef2 to
49f3938
Compare
Contributor
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 16, 2025
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Reviewed By: skrtskrtfb Differential Revision: D81519735
mgiordy
force-pushed
the
export-D81519735
branch
from
49f3938 to
569f2e6
Compare
Contributor
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 23, 2025
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Reviewed By: skrtskrtfb Differential Revision: D81519735
mgiordy
force-pushed
the
export-D81519735
branch
from
569f2e6 to
406d706
Compare
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 24, 2025
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Reviewed By: skrtskrtfb Differential Revision: D81519735
Contributor
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 25, 2025
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Reviewed By: skrtskrtfb Differential Revision: D81519735
mgiordy
force-pushed
the
export-D81519735
branch
from
406d706 to
2c0b42b
Compare
Contributor
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 25, 2025
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Reviewed By: skrtskrtfb Differential Revision: D81519735
mgiordy
force-pushed
the
export-D81519735
branch
from
2c0b42b to
61701f4
Compare
Contributor
mgiordy
force-pushed
the
export-D81519735
branch
from
61701f4 to
10689f5
Compare
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 25, 2025
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Reviewed By: skrtskrtfb Differential Revision: D81519735
Contributor
mgiordy
force-pushed
the
export-D81519735
branch
from
10689f5 to
2840bbd
Compare
Contributor
mgiordy
force-pushed
the
export-D81519735
branch
from
2840bbd to
12aae49
Compare
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 26, 2025
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Reviewed By: skrtskrtfb Differential Revision: D81519735
Contributor
mgiordy
force-pushed
the
export-D81519735
branch
from
12aae49 to
b556b8d
Compare
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 26, 2025
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Reviewed By: skrtskrtfb Differential Revision: D81519735
Contributor
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 26, 2025
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Reviewed By: skrtskrtfb Differential Revision: D81519735
mgiordy
force-pushed
the
export-D81519735
branch
from
b556b8d to
ed2bdec
Compare
Contributor
mgiordy
force-pushed
the
export-D81519735
branch
from
ed2bdec to
5b7459b
Compare
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 27, 2025
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Reviewed By: skrtskrtfb Differential Revision: D81519735
Contributor
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 27, 2025
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Reviewed By: skrtskrtfb Differential Revision: D81519735
mgiordy
force-pushed
the
export-D81519735
branch
from
5b7459b to
1f03b87
Compare
Contributor
mgiordy
force-pushed
the
export-D81519735
branch
from
1f03b87 to
a2cceb1
Compare
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 29, 2025
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Reviewed By: skrtskrtfb Differential Revision: D81519735
Contributor
mgiordy
force-pushed
the
export-D81519735
branch
from
a2cceb1 to
792d571
Compare
mgiordy
pushed a commit
to mgiordy/executorch
that referenced
this pull request
Sep 29, 2025
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Reviewed By: skrtskrtfb Differential Revision: D81519735
Contributor
Summary: # Context This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point. # In this diff 1. Op nodes are returned from each pattern matching 2. Dequantize nodes are bypassed if not needed in the final graph. Reviewed By: skrtskrtfb, mcremon-meta Differential Revision: D81519735
mgiordy
force-pushed
the
export-D81519735
branch
from
792d571 to
63ae795
Compare
Contributor
skrtskrtfb
approved these changes
Sep 29, 2025
skrtskrtfb
approved these changes
Sep 30, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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
This Diff adds support for mixed quantization operators in Executorch. Now weights and biases can be quantized, while inputs and activations are kept in floating point.
In this diff
Differential Revision: D81519735