llama.cpp icon indicating copy to clipboard operation
llama.cpp copied to clipboard
Published 5 months ago • verified publisher icon ggml-org

opencl: fix rms_norm_mul

Open lhez opened this issue 5 months ago • 0 comments

The rms_norm_mul kernel produces incorrect result when ne00 = 768. This PR changes how the kernel does reduction to get the sum. This seems to fix the issue.

lhez avatar Nov 13 '25 19:11 lhez