Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Pytorch MatMul Substitution #1292

Open
wants to merge 36 commits into
base: main
Choose a base branch
from
Open

Pytorch MatMul Substitution #1292

wants to merge 36 commits into from

Conversation

itai-berman
Copy link
Collaborator

@itai-berman itai-berman commented Dec 12, 2024 •
edited
Loading

Pull Request Description:

Add substitution to support MatMul function & operator by doing Reshape->Split->2D MatMul->Stack->Reshape

Checklist before requesting a review:

  • I set the appropriate labels on the pull request.
  • I have added/updated the release note draft (if necessary).
  • I have updated the documentation to reflect my changes (if necessary).
  • All function and files are well documented.
  • All function and classes have type hints.
  • There is a licenses in all file.
  • The function and variable names are informative.
  • I have checked for code duplications.
  • I have added new unittest (if necessary).

elad-c
elad-c requested changes Dec 15, 2024
View reviewed changes
...l_compression_toolkit/core/pytorch/graph_substitutions/substitutions/matmul_decomposition.py Outdated Show resolved Hide resolved
...l_compression_toolkit/core/pytorch/graph_substitutions/substitutions/matmul_decomposition.py Outdated Show resolved Hide resolved
...l_compression_toolkit/core/pytorch/graph_substitutions/substitutions/matmul_decomposition.py Outdated Show resolved Hide resolved
...l_compression_toolkit/core/pytorch/graph_substitutions/substitutions/matmul_decomposition.py Outdated Show resolved Hide resolved
...l_compression_toolkit/core/pytorch/graph_substitutions/substitutions/matmul_decomposition.py Outdated Show resolved Hide resolved
...l_compression_toolkit/core/pytorch/graph_substitutions/substitutions/matmul_decomposition.py Outdated Show resolved Hide resolved
...l_compression_toolkit/core/pytorch/graph_substitutions/substitutions/matmul_decomposition.py Outdated Show resolved Hide resolved
...l_compression_toolkit/core/pytorch/graph_substitutions/substitutions/matmul_decomposition.py Show resolved Hide resolved
tests/pytorch_tests/model_tests/feature_models/matmul_test.py Outdated Show resolved Hide resolved
tests/pytorch_tests/model_tests/feature_models/matmul_test.py Show resolved Hide resolved
ofirgo and others added 6 commits December 19, 2024 09:59
@ofirgo
restricting matplotlib to version < 3.10.0 to fix an issue with tenso…
ef3270b 
…rboard when using this version of matplotlib (sony#1298)
@lior-dikstein @ofirgo
Refactor Target Platform Capabilities - Phase 3 (sony#1297)
b125f96 
* Refactor Target Platform Capabilities - Phase 3
Remove context manager functionality from the Target Platform Model.
Fix all tests and TP models.

---------

Co-authored-by: liord <[email protected]>
Co-authored-by: Ofir Gordon <[email protected]>
@ofirgo
TPC attach 2 framework (sony#1296)
3c536f2 
Create attach to framework class for keras and pytorch and implement 'attach' method that returns a TPC
@elad-c
Replace max tensor with max cut (sony#1295)
ceaf820 
Replace MaxTensor with MaxCut for activation mixed precision (Experimental).
@elad-c
Fix bug (sony#1303)
ae8e60c
@elad-c
Move splitting ops from default to qpreserving configs in TPCv4. (son…
dcbfc5c 
…y#1304)
@yarden-yagil-sony yarden-yagil-sony self-requested a review December 30, 2024 12:50
lior-dikstein and others added 8 commits December 31, 2024 17:57
@lior-dikstein @elad-c
Refactor Target Platform Capabilities - Phase 4 (sony#1301)
afed6e3 
* Refactor Target Platform Capabilities - Phase 4
Replace schema classes from dataclass to pydantic 'BaseModel'.
Fix tests to support pydantic schema classes.
Add test for exporting TargetPlatformModel class to json
---------

Co-authored-by: liord <[email protected]>
Co-authored-by: Elad Cohen <[email protected]>
@itai-berman
add substitution for functional linear
5b2a55d
@itai-berman
change test and substitution to support kwargs input, update output s…
14b6db7 
…hape
@itai-berman
add matmul decomposition substituion and tests
b677a5a
@itai-berman
fix license
8b42444
@itai-berman
add substitution for functional linear
8403eca
@itai-berman
change test and substitution to support kwargs input, update output s…
b198bec 
…hape
@itai-berman
fix license
9fa5710
@itai-berman
add substitution for functional linear
9a480b3
@itai-berman
change test and substitution to support kwargs input, update output s…
728cd90 
…hape
@itai-berman
fix license
f535add
@itai-berman
add substitution for functional linear
aba2cb0
@itai-berman
change test and substitution to support kwargs input, update output s…
99f7694 
…hape
@itai-berman
fix license
98d78a4
@itai-berman
add tests for different input dimensions with only float tpc
bdc9057
@itai-berman
skip unnecessary nodes, add expand to support broadcasting
6e8246e
@itai-berman
Merge remote-tracking branch 'origin/pytorch_matmul_substitution' int…
8ae4bed 
…o pytorch_matmul_substitution

# Conflicts:
#	model_compression_toolkit/core/pytorch/graph_substitutions/substitutions/matmul_decomposition.py
#	tests/pytorch_tests/model_tests/feature_models/matmul_test.py
#	tests/pytorch_tests/model_tests/test_feature_models_runner.py
@itai-berman itai-berman requested a review from elad-c January 2, 2025 12:31
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@elad-c elad-c elad-c approved these changes

@Idan-BenAmi Idan-BenAmi Awaiting requested review from Idan-BenAmi

@yarden-yagil-sony yarden-yagil-sony Awaiting requested review from yarden-yagil-sony

Assignees
No one assigned
Labels
auto:core auto:target_platform_capabilities auto:tests pr: minor feature
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@itai-berman @elad-c @ofirgo @lior-dikstein