pytorch
138a0188 - Add support for logaddexp(float16) in CUDA and implement its reference (#91869)

Commit
2 years ago
Add support for logaddexp(float16) in CUDA and implement its reference (#91869) The reference is implemented so that it generates efficient and numerically stable triton code. Fixes https://github.com/pytorch/pytorch/issues/91683 Pull Request resolved: https://github.com/pytorch/pytorch/pull/91869 Approved by: https://github.com/ngimel
Author
Committer
Parents
Loading