frozenleaves/vllm
1
0
Fork 0
mirror of https://github.com/vllm-project/vllm.git synced 2025-10-20 14:53:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Kernel] Fix awq error when n is not divisable by 128 (#13227)

Browse Source
This commit is contained in:
Jinzhen Lin
2025-02-14 12:07:05 +08:00
committed by GitHub
parent 410886950a
commit 8c32b08a86
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
csrc/quantization/awq/gemm_kernels.cu
View File

@ -334,7 +334,7 @@ __global__ void __launch_bounds__(64)
}
// TODO: Shang: Hoist loop invariance.
for (int ax1_0_1 = 0; ax1_0_1 < 4; ++ax1_0_1) {
for (int ax1_0_1 = 0; ax1_0_1 < (N / 32); ++ax1_0_1) {
for (int local_id = 0; local_id < 8; ++local_id) {
int row_offset = (((int)blockIdx_y) / j_factors1) * 16 +
((int)threadIdx.x) / 4 + (local_id % 4) / 2 * 8;