onnxruntime
6d681ee2 - past-present bnsh, non-cache past-present.

Commit

3 years ago

past-present bnsh, non-cache past-present.

References

aciddelgado/gqa_bnsh

#17674 - [CUDA] GroupQueryAttention operator using FlashAttention

Author

aciddelgado

aciddelgado

Parents

Loading