llama.cpp
CUDA & CPU: support F32 kernel type for `CONV_TRANSPOSE_2D`
#17094
Merged
Merged