llama.cpp
ggml : support broadcast for ggml_soft_max_ext and ggml_flash_attn_ext
#14435
Merged
Merged