Skip to content

Commit 1abbbb6

Browse files
author
Jatin Gangani
committed
set dp attention env vars
1 parent b1fbf48 commit 1abbbb6

1 file changed

Lines changed: 3 additions & 3 deletions

File tree

benchmarks/gptoss_fp4_b200_trt_slurm.sh

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -31,9 +31,6 @@ echo "MOE_BACKEND set to '$MOE_BACKEND'"
3131

3232
EXTRA_CONFIG_FILE="gptoss-fp4.yml"
3333
export TRTLLM_ENABLE_PDL=1
34-
export TRTLLM_MOE_ALLTOALL_BACKEND="mnnvlthroughput"
35-
export TRTLLM_FORCE_ALLTOALL_METHOD="MNNVL"
36-
export TRTLLM_MOE_A2A_WORKSPACE_MB="2048"
3734

3835
cat > $EXTRA_CONFIG_FILE << EOF
3936
cuda_graph_config:
@@ -52,6 +49,9 @@ moe_config:
5249
EOF
5350

5451
if [[ "$DP_ATTENTION" == "true" ]]; then
52+
export TRTLLM_MOE_ALLTOALL_BACKEND="mnnvlthroughput"
53+
export TRTLLM_FORCE_ALLTOALL_METHOD="MNNVL"
54+
export TRTLLM_MOE_A2A_WORKSPACE_MB="2048"
5555
cat << EOF >> $EXTRA_CONFIG_FILE
5656
attention_dp_config:
5757
enable_balance: true

0 commit comments

Comments
 (0)