Skip to content

Start using rbe workers#401

Merged
alekstheod merged 22 commits intorocm-jaxlib-v0.7.1from
ci_start_using_rbe_workers
Oct 16, 2025
Merged

Start using rbe workers#401
alekstheod merged 22 commits intorocm-jaxlib-v0.7.1from
ci_start_using_rbe_workers

Conversation

@alekstheod
Copy link
Collaborator

Start using rbe workers (run tests and build)

@alekstheod alekstheod force-pushed the ci_start_using_rbe_workers branch 12 times, most recently from f9a5e7a to 769e105 Compare October 7, 2025 14:33
if [[ $1 == "asan" ]]; then
SANITIZER_ARGS+=("--test_env=ASAN_OPTIONS=suppressions=${SCRIPT_DIR}/asan_ignore_list.txt:use_sigaltstack=0")
SANITIZER_ARGS+=("--test_env=LSAN_OPTIONS=suppressions=${SCRIPT_DIR}/lsan_ignore_list.txt:use_sigaltstack=0")
SANITIZER_ARGS+=("--test_env=ASAN_OPTIONS=suppressions=build_tools/rocm/asan_ignore_list.txt:use_sigaltstack=0")
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why did you change this? We would still need script_dir for the tag filters?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ignore lists shall be set as a data dependency to a target so to land to rbe worker. Otherwise bazel will not bring them as they are not in the tree. The path is relative to the runfiles directory in the sandbox!

@alekstheod alekstheod force-pushed the ci_start_using_rbe_workers branch from 3fd9b68 to 538c33e Compare October 10, 2025 10:05
@alekstheod alekstheod force-pushed the ci_start_using_rbe_workers branch from 087c86b to d0efa67 Compare October 13, 2025 08:41
@alekstheod alekstheod force-pushed the ci_start_using_rbe_workers branch 2 times, most recently from eaaf8ee to ea63caa Compare October 13, 2025 11:37
@alekstheod alekstheod force-pushed the ci_start_using_rbe_workers branch from ea63caa to 156f558 Compare October 13, 2025 11:53
@alekstheod alekstheod force-pushed the ci_start_using_rbe_workers branch from 0fa0ec8 to 7fd7256 Compare October 15, 2025 12:08
@alekstheod alekstheod force-pushed the ci_start_using_rbe_workers branch 4 times, most recently from 9f4aa47 to 3008a28 Compare October 15, 2025 15:03
@alekstheod alekstheod force-pushed the ci_start_using_rbe_workers branch from 3008a28 to 2f2360a Compare October 15, 2025 15:51
],
exec_properties = {
"container-image": "docker://rocm/tensorflow-build:2.19-jammy-python3.9-rocm6.4.0@sha256:7b42dafbe5401fd84cf951a63f16dd44a84c64a6ba2eeb7675fd0919f03b2bdf",
"container-image": "docker://rocm/tensorflow-build:latest-jammy-python3.10-rocm6.4.0@sha256:9c2fbc861570735fc86dbf39dd3cefef6a13fcf6ae08358271af52f1aebd4248",

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you update sha to rocm 7.0 image?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ok

"@bazel_tools//tools/cpp:clang",
],
exec_properties = {
"container-image": "docker://rocm/tensorflow-build:latest-jammy-python3.10-rocm6.4.0@sha256:9c2fbc861570735fc86dbf39dd3cefef6a13fcf6ae08358271af52f1aebd4248",

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

same here

SANITIZER_ARGS+=("--test_env=TSAN_OPTIONS=suppressions=${SCRIPT_DIR}/tsan_ignore_list.txt::history_size=7:ignore_noninstrumented_modules=1")
SANITIZER_ARGS+=("--config=tsan")
TAG_FILTERS=$TAG_FILTERS,-notsan
EXCLUDED_TESTS+=(

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can we put all of them in one place? Any special reason for doing += here?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes they are defined above these are additional excluded tests in case of tsan config. Tsan adds delays and also produces warnings solving which is not a scope of this PR.

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please add a comment excluded for tsan

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done

--test_filter=-$(IFS=: ; echo "${EXCLUDED_TESTS[*]}") \
"${SANITIZER_ARGS[@]}" \
"$@"
"$@" \

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not sure if this works for every option being passed. Better to exchange 125 and 126 lines

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think it is better this way. The last option wins. Meaning this way we give to CI the opportunity to override whatever is defined in this script, if required.

@alekstheod alekstheod force-pushed the ci_start_using_rbe_workers branch from 2a3af07 to 32032d1 Compare October 16, 2025 13:34
@hsharsha
Copy link

LGTM

@alekstheod alekstheod merged commit 061033b into rocm-jaxlib-v0.7.1 Oct 16, 2025
8 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants