onnxruntime
6abe9b5c - make softmax op in Attention CUDA kernel numerically stable

Commit

5 years ago

make softmax op in Attention CUDA kernel numerically stable

Author

hariharans29

hariharans29

Parents

Loading