Skip to content
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

ci: Stabilize L0_pinned_memory flakiness #7929

Merged
merged 2 commits into from
Jan 13, 2025
Merged

ci: Stabilize L0_pinned_memory flakiness #7929

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e63d84c
Use --request-count in PA to remove PA stability criteria and make L0…
rmccorm4 Jan 13, 2025
0f4cbdc
Reduce request count from 100 -> 10 to speed up test, re-use variable
rmccorm4 Jan 13, 2025
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
11 changes: 8 additions & 3 deletions qa/L0_pinned_memory/test.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
#!/bin/bash
# Copyright (c) 2019-2021, NVIDIA CORPORATION. All rights reserved.
# Copyright (c) 2019-2025, NVIDIA CORPORATION. All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions
Expand Down Expand Up @@ -38,6 +38,9 @@ if [ ! -z "$TEST_REPO_ARCH" ]; then
REPO_VERSION=${REPO_VERSION}_${TEST_REPO_ARCH}
fi

# Use "--request-count" throughout the test to PA stability criteria and
# reduce flaky failures from PA unstable measurements.
REQUEST_COUNT=10
CLIENT=../clients/perf_client
# Only use libtorch as it accepts GPU I/O and it can handle variable shape
BACKENDS=${BACKENDS:="libtorch"}
Expand Down Expand Up @@ -91,7 +94,7 @@ for BACKEND in $BACKENDS; do

# Sanity check that the server allocates pinned memory for large size
set +e
$CLIENT -m${ENSEMBLE_NAME} --shape INPUT0:16777216
$CLIENT -m${ENSEMBLE_NAME} --shape INPUT0:16777216 --request-count ${REQUEST_COUNT}
if (( $? != 0 )); then
RET=1
fi
Expand Down Expand Up @@ -128,6 +131,7 @@ for BACKEND in $BACKENDS; do
for TENSOR_SIZE in 16384 1048576 2097152 4194304 8388608 16777216; do
$CLIENT -i grpc -u localhost:8001 -m${ENSEMBLE_NAME} \
--shape INPUT0:${TENSOR_SIZE} \
--request-count ${REQUEST_COUNT} \
>> ${BACKEND}.${TENSOR_SIZE}.pinned.log 2>&1
if (( $? != 0 )); then
RET=1
Expand All @@ -150,7 +154,7 @@ for BACKEND in $BACKENDS; do

# Sanity check that the server allocates non-pinned memory
set +e
$CLIENT -m${ENSEMBLE_NAME} --shape INPUT0:1
$CLIENT -m${ENSEMBLE_NAME} --shape INPUT0:1 --request-count ${REQUEST_COUNT}
if (( $? != 0 )); then
RET=1
fi
Expand Down Expand Up @@ -180,6 +184,7 @@ for BACKEND in $BACKENDS; do
for TENSOR_SIZE in 16384 1048576 2097152 4194304 8388608 16777216; do
$CLIENT -i grpc -u localhost:8001 -m${ENSEMBLE_NAME} \
--shape INPUT0:${TENSOR_SIZE} \
--request-count ${REQUEST_COUNT} \
>> ${BACKEND}.${TENSOR_SIZE}.nonpinned.log 2>&1
if (( $? != 0 )); then
RET=1
Expand Down
Loading