llama.cpp
cuda: fix grid.y overflow in non-contiguous dequantize/convert kernels
#19999
Merged
Merged
cuda: fix grid.y overflow in non-contiguous dequantize/convert kernels #19999
am17an
merged
1 commit into
ggml-org:master
from
oobabooga:fix-cuda-grid-y-overflow
cuda: cap grid.y at 65535 in non-contiguous dequantize/convert kernels
5bdfdc5d
am17an
approved these changes
on 2026-03-01
Login to write a write a comment.
Login via GitHub