Skip to content

[CI] Skip MegaMoE v2 multi-GPU tests #5003

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
gyohuangxin merged 1 commit into from
Aug 26, 2026
Merged

[CI] Skip MegaMoE v2 multi-GPU tests #5003

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
41 changes: 2 additions & 39 deletions .github/scripts/aiter_test.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,10 +35,12 @@ else
fi

skip_tests=(
"op_tests/multigpu_tests/bench_mega_moe_v2.py"
"op_tests/multigpu_tests/test_dispatch_combine.py"
"op_tests/multigpu_tests/test_communication.py"
"op_tests/multigpu_tests/test_mori_all2all.py"
"op_tests/multigpu_tests/test_fused_ar_rms.py"
"op_tests/multigpu_tests/test_mega_moe_v2.py"
"op_tests/multigpu_tests/triton_test/test_reduce_scatter_all_gather.py"
"op_tests/multigpu_tests/triton_test/test_fused_rs_rmsnorm_quant_ag.py"
)
Expand Down Expand Up @@ -82,30 +84,6 @@ for file in "${sharded_files[@]}"; do
# batch gate so they exercise the persistent kernel at every batch size.
test_cmd=(timeout 60m python3 "$file")
case "$file" in
op_tests/multigpu_tests/bench_mega_moe_v2.py)
{
echo "Running MegaMoEV2 versus Mori EP performance guards on 8 GPUs"
} | tee -a latest_test.log
test_cmd=(
env MORI_SOCKET_IFNAME=lo MORI_SHMEM_HEAP_SIZE=40G
timeout 60m
bash -c '
set -euo pipefail
bench=$1
for spec in \
"512 uniform 0.6" \
"512 rank-mixed-skew 1.0" \
"8192 uniform 0.6" \
"8192 rank-mixed-skew 1.0"; do
read -r tokens route bias <<< "$spec"
torchrun --standalone --nproc_per_node=8 "$bench" \
--tokens "$tokens" --mtpr 8192 --route "$route" \
--hot-bias "$bias" --iters 20 --perf-guard
done
'
_ "$file"
)
;;
op_tests/multigpu_tests/test_mega_moe_gfx1250.py)
{
echo "Running gfx1250 MegaMoE fused-scatter accuracy on 8 GPUs when supported"
Expand All @@ -128,21 +106,6 @@ for file in "${sharded_files[@]}"; do
_ "$file"
)
;;
op_tests/multigpu_tests/test_mega_moe_v2.py)
{
echo "Running MegaMoEV2 v4_pro fixed-slot and compact coverage on 8 GPUs"
} | tee -a latest_test.log
test_cmd=(
env MORI_SHMEM_HEAP_SIZE=40G
timeout 60m
torchrun --standalone --nproc_per_node=8 "$file"
--network v4_pro
--bs-list 128,512
--iters 10
--accuracy-max-bs 512
--rtol 0.10
)
;;
op_tests/test_mla_persistent.py|op_tests/test_mla_persistent_round_robin.py)
{
echo "Using AITER_MLA_DECODE_PERSISTENT_MAX_BATCH=0 for $file"
Expand Down
4 changes: 0 additions & 4 deletions .github/workflows/aiter-test.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1004,15 +1004,11 @@ jobs:
bash -lc "
set -euo pipefail
pip uninstall -y amd-aiter aiter || true
apt-get update
DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends libgrpc++1.51t64
pip install -r requirements.txt
pip install --upgrade pandas pyzmq einops numpy==1.26.2
pip install --upgrade 'pybind11>=3.0.1'
pip install --upgrade 'ninja>=1.11.1'
pip install tabulate
Comment on lines 1007 to 1011
pip install 'amd-mori==1.2.2'
python3 -c 'import mori, mori.shmem'
pip install '${AITER_WHEEL_PATH}'
pip show amd-aiter
"
Expand Down
Loading