llama.cpp
2540a290 - Make CUDA compile with QK_K = 64

Commit
1 year ago
Make CUDA compile with QK_K = 64 Tests don't pass, plus we get misaligned access
Author
Iwan Kawrakow
Parents
Loading