[ONNX] Add converter for QAttention from Microsoft onnxruntime contrib opset #13654
Conversation
This file contains 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
|
Thanks for contributing to TVM! Please refer to the contributing guidelines https://tvm.apache.org/docs/contribute/ for useful information and tips. Please request code reviews from Reviewers by @-ing them in a comment. Generated by tvm-bot |
…atmul_tensorcore_cuda
|
Hello @vvchernov, @echuraev, @AndrewZhaoLuo! |
echuraev
reviewed
Dec 28, 2022
There was a problem hiding this comment.
LGTM. But I don't have a lot of knowledge in this codebase. @jwfromm, @AndrewZhaoLuo could you please take a look at this PR?
| # Currently only (batch, past_seq_length + seq_length) shape is supported. | ||
| mask_index = inputs[5] | ||
|
|
||
| # Scalar, which means a per-tensor/layer quantization |
There was a problem hiding this comment.
nit: You have absolutely the same comment for input[3] and input[7]
python/tvm/relay/frontend/onnx.py
Outdated
| result, | ||
| _op.multiply(lhs_scale, rhs_scale), | ||
| zero_point_zero, | ||
| axis=-1, # TODO(agladyshev): what is 'axis' parameter for? |
There was a problem hiding this comment.
Do you still need this todo comment?
|
Apologies, I've been quite sick. I'll try to look at this Thursday. |
AndrewZhaoLuo
approved these changes
Jan 3, 2023
This was referenced Jan 17, 2023
fzi-peccia
pushed a commit
to fzi-peccia/tvm
that referenced
this pull request
Mar 27, 2023
…b opset (apache#13654) * init QAttention converter * add type and shape checking * add test for QAttention * add tests for optional parameters * change mask_index shape * add support for 'past' input * add support for 'unidirectional' attribute * expand test coverage * fix lint * fix pylint * fix batch dimension for topi/cuda/batch_matmul_tensorcore.py::batch_matmul_tensorcore_cuda * code review fix
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.
This PR adds support for QAttention - quantized version of Attention from Microsoft onnxruntime contrib opset.
An explanation and illustration of how this layer works can be found, for example, in @lena-voita NLP course.