llama.cpp
e3742272 - Revert "cuda : use CUBLAS_COMPUTE_16F for non-attention ops"

Commit

2 years ago

Revert "cuda : use CUBLAS_COMPUTE_16F for non-attention ops" This reverts commit 0f2498f25d7e278f075d060e8e77e68dacf4e90c.

Author

ggerganov

Parents