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

ONNX MatMulInteger support #2071

Closed
GregoryMorse opened this issue Oct 29, 2022 · 1 comment
Closed

ONNX MatMulInteger support #2071

GregoryMorse opened this issue Oct 29, 2022 · 1 comment
Labels
enhancement New feature or request unsupported ops Issues related to unsupported operators

Comments

@GregoryMorse
Copy link
Contributor

Describe the feature request

The ONNX specification since opset 10 has specified a MatMulInteger operation which allows uint8/int8 types among other things. MatMul specifically does not allow this size.

Currently only MatMul is support, and uint8/int8 will fail to compile per the ONNX spec on that operation.

Describe scenario use case

Any places where small integer arithmetic is used, specifically when computing matrix multiplication with uint8/int8 matrix x uint8/int8 matrix -> int32 matrix.
@GregoryMorse GregoryMorse added the enhancement New feature or request label Oct 29, 2022
GregoryMorse added a commit to GregoryMorse/tensorflow-onnx that referenced this issue Oct 29, 2022
@GregoryMorse
Add Support for MatMulInteger
fdcc79b 
MatMulInteger was supported in ONNX opset v10 (not checked in proposed change, the error can be addressed on save), this specific type combination is support in TensorFlow, but the node type not identified and handled properly here.

Handles onnx#2071
@GregoryMorse GregoryMorse mentioned this issue Oct 29, 2022
Merged
@fatcat-z fatcat-z added the unsupported ops Issues related to unsupported operators label Oct 29, 2022
GregoryMorse added a commit to GregoryMorse/tensorflow-onnx that referenced this issue Nov 4, 2022
@GregoryMorse
Add Support for MatMulInteger
3e577a6 
MatMulInteger was supported in ONNX opset v10 (not checked in proposed change, the error can be addressed on save), this specific type combination is support in TensorFlow, but the node type not identified and handled properly here.

Handles onnx#2071

Signed-off-by: Gregory Morse <[email protected]>
fatcat-z added a commit that referenced this issue Nov 4, 2022
@GregoryMorse @fatcat-z
Add Support for MatMulInteger (#2072)
2c1db54 
* Add Support for MatMulInteger

MatMulInteger was supported in ONNX opset v10 (not checked in proposed change, the error can be addressed on save), this specific type combination is support in TensorFlow, but the node type not identified and handled properly here.

Handles #2071

Signed-off-by: Gregory Morse <[email protected]>

* Update math.py

Signed-off-by: Gregory Morse <[email protected]>

* Update support_status.md

Signed-off-by: Gregory Morse <[email protected]>

* Update test_backend.py

Signed-off-by: Gregory Morse <[email protected]>

Signed-off-by: Gregory Morse <[email protected]>
Co-authored-by: Jay Zhang <[email protected]>
@fatcat-z
Copy link
Collaborator

fatcat-z commented Nov 4, 2022

Close this issue since the PR has been merged. Feel free to reopen it if necessary.

@fatcat-z fatcat-z closed this as completed Nov 4, 2022
fsalv pushed a commit to fsalv/tensorflow-onnx that referenced this issue Dec 6, 2022
@GregoryMorse @fatcat-z @f-salvetti
Add Support for MatMulInteger (onnx#2072)
d8b1009 
* Add Support for MatMulInteger

MatMulInteger was supported in ONNX opset v10 (not checked in proposed change, the error can be addressed on save), this specific type combination is support in TensorFlow, but the node type not identified and handled properly here.

Handles onnx#2071

Signed-off-by: Gregory Morse <[email protected]>

* Update math.py

Signed-off-by: Gregory Morse <[email protected]>

* Update support_status.md

Signed-off-by: Gregory Morse <[email protected]>

* Update test_backend.py

Signed-off-by: Gregory Morse <[email protected]>

Signed-off-by: Gregory Morse <[email protected]>
Co-authored-by: Jay Zhang <[email protected]>
Signed-off-by: Salvetti, Francesco <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
enhancement New feature or request unsupported ops Issues related to unsupported operators
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@GregoryMorse @fatcat-z