Skip to content

Commit 12d030a

Browse files
committed
fix(sglang): serialize B300 TP4 DP scheduling
1 parent 6c4ec1d commit 12d030a

1 file changed

Lines changed: 3 additions & 0 deletions

File tree

benchmarks/single_node/agentic/dsv4_fp4_b300_sglang.sh

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -135,6 +135,9 @@ if [ "$DP_ATTENTION" = "true" ]; then
135135
--moe-runner-backend flashinfer_mxfp4
136136
--disable-flashinfer-autotune
137137
)
138+
if [ "$TP" -eq 4 ]; then
139+
PARALLEL_ARGS+=(--disable-overlap-schedule)
140+
fi
138141
MEM_FRACTION_STATIC=0.88
139142
CHUNKED_PREFILL_SIZE=16384
140143
else

0 commit comments

Comments
 (0)