You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2021/06/07 21:44:43 UTC
[GitHub] [tvm] mbrookhart opened a new pull request #8208: [Frontend] Speed up Fold Constant
mbrookhart opened a new pull request #8208:
URL: https://github.com/apache/tvm/pull/8208
@masahi
This utiliy was creating a new IRModule from the provided expression if the provided module was None. This isn't strictly necessary, the use cases are correct if we pass in an empty IRModule. The WellFormed check inside IRModule.from_expr was actually taking a lot of the time when importing large ONNX models. This change gets me a 2.6x speedup on importing the large NLP model I'm playing with today.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] mbrookhart commented on pull request #8208: [Frontend] Speed up Fold Constant
Posted by GitBox <gi...@apache.org>.
mbrookhart commented on pull request #8208:
URL: https://github.com/apache/tvm/pull/8208#issuecomment-856830311
We use it in dozens of places in the ONNX importer with mod=None, if there was a problem, ~1/3 of the onnx unit tests would have failed. I'm not sure any of the utilities in the frontend/common.py file are actually tested? https://github.com/apache/tvm/blob/main/tests/python/frontend/test_common.py
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] masahi merged pull request #8208: [Frontend] Speed up Fold Constant
Posted by GitBox <gi...@apache.org>.
masahi merged pull request #8208:
URL: https://github.com/apache/tvm/pull/8208
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org