Megatron-DeepSpeed
c9afebc6 - Revert "Checking we use fused kernels to compute scaled masked softmax on prefix lm (#209)"

Commit

4 years ago

Revert "Checking we use fused kernels to compute scaled masked softmax on prefix lm (#209)" This reverts commit b227590f3840b6bbede1ad56813e9dfdc7c971ce.

Author

thomasw21

Parents

b227590f

Megatron-DeepSpeed c9afebc6 - Revert "Checking we use fused kernels to compute scaled masked softmax on prefix lm (#209)"

Megatron-DeepSpeed
c9afebc6 - Revert "Checking we use fused kernels to compute scaled masked softmax on prefix lm (#209)"