Skip to content

Conversation

@mgehre-amd
Copy link
Collaborator

No description provided.

mgehre-amd and others added 3 commits December 2, 2024 09:14
Fixes `onnx.Pow(float,int)` and `Pow(int,float)` accuracy. Torch uses
`double` internally to compute pow if one argument is integer and the
other one is floating point (due to C++ promotion rules).

This PR keeps `onnx.Pow(int,int)` as is, which still produces numeric
mismatches for values that overflow. torch uses a pure-integer
implementation, where torch-mlir currently maps it to `Pow(float,float)`
Base automatically changed from bump_to_46a5772d to bump_to_0913b967 February 5, 2025 08:58
@mgehre-amd mgehre-amd merged commit 2091491 into bump_to_0913b967 Feb 5, 2025
4 checks passed
@mgehre-amd mgehre-amd deleted the bump_to_ee08942c branch February 5, 2025 20:49
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants