llama.cpp
opencl: Q6_K GEMM/GEMV fix for ne01 of weights that are not multiples of 128.
#25464

Merged

Commits

opencl: fix garbled output for Q6_K weights with ne01 % 128 != 0 on Adreno

lhez committed 55 days ago
opencl: reserve alignment slack for the SOA subbuffer carve in alloc size

lhez committed 55 days ago
opencl: use lm based q6_k mm when ne1 is not multiple of 128

lhez committed 55 days ago

Loading