llama.cpp
750f60c0 - CUDA: fix Pascal FA, deq. KV to FP16 for batch > 8 (#7681)
Commit
View On GitHub
Commit
1 year ago
CUDA: fix Pascal FA, deq. KV to FP16 for batch > 8 (#7681)
Parents
9b596417
FAQ Terms Privacy Refunds Impressum
Loading