llama.cpp
opencl: fold the gpt-oss MoE per-expert bias adds into the epilogue (op/kernel fusion)
#26431
Merged
Merged
opencl: fold the gpt-oss MoE per-expert bias adds into the epilogue (op/kernel fusion) #26431
lhez
merged
2 commits into
ggml-org:master
from
qualcomm:hq/opencl-moe-bias-glu-fusion-upstream
opencl: fold the gpt-oss MoE bias adds into swiglu_oai
c5ef5073
opencl: fold the MoE down-projection bias into the combine
79b151d7
max-krasnyansky
approved these changes
on 2026-08-21
lhez
approved these changes
on 2026-08-21
Login to write a write a comment.
Login via GitHub