llama.cpp
ggml-cuda : perform cublas fp16 matrix multiplication as fp16
#3370
Merged
Merged
ggml-cuda : perform cublas fp16 matrix multiplication as fp16 #3370
ggerganov
merged
3 commits into
master
from
cublas-f16
ggml-cuda : perform cublas fp16 matrix multiplication as fp16
79fe5a1f
try to fix rocm build
32ada53c
cebtenzzre
commented
on 2023-09-27
restrict fp16 mat mul to volta and up
7d5674dd
ggerganov
approved these changes
on 2023-09-28
Login to write a write a comment.
Login via GitHub