ggml-cuda.cu
llama.cpp
8ad92dc1 - ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext
Commit
1 year ago
- Files7
ggml-metal.m- ggml-metal.metal
ggml.c
ggml.h
llama.cpp- tests
- test-backend-ops.cpp