A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 10, 06:30 UTC
[ROCm][CI] Handle pytest status code 5 when a shard isn't allocated any tests (#32040)
Signed-off-by: Divakar Verma <divakar.verma@amd.com>
Divakar Verma committed
Jan 12, 2026 at 16:35 UTC
a28d9f4470550835a8b36b9e558d4d47b30c7127
2 files changed
+11
-2
.buildkite/scripts/hardware_ci/run-amd-test.sh
+10
-1
@@ -209,12 +209,21 @@ if [[ $commands == *"--shard-id="* ]]; then
209
wait "${pid}"
210
STATUS+=($?)
211
done
212
+ at_least_one_shard_with_tests=0
213
for st in "${STATUS[@]}"; do
213
- if [[ ${st} -ne 0 ]]; then
214
+ if [[ ${st} -ne 0 ]] && [[ ${st} -ne 5 ]]; then
215
echo "One of the processes failed with $st"
216
exit "${st}"
217
+ elif [[ ${st} -eq 5 ]]; then
218
+ echo "Shard exited with status 5 (no tests collected) - treating as success"
219
+ else # This means st is 0
220
+ at_least_one_shard_with_tests=1
221
fi
222
done
223
+ if [[ ${#STATUS[@]} -gt 0 && ${at_least_one_shard_with_tests} -eq 0 ]]; then
224
+ echo "All shards reported no tests collected. Failing the build."
225
+ exit 1
226
+ fi
227
else
228
echo "Render devices: $BUILDKITE_AGENT_META_DATA_RENDER_DEVICES"
229
docker run \
.buildkite/test-amd.yaml
+1
-1
@@ -870,7 +870,7 @@ steps:
870
- label: Language Models Tests (Extra Standard) %N
871
timeout_in_minutes: 45
872
mirror_hardwares: [amdexperimental]
873
- agent_pool: mi325_2
873
+ agent_pool: mi325_8
874
# grade: Blocking
875
torch_nightly: true
876
source_file_dependencies: