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

Jump to bottom

Fixed failing NCCL Group test on internal CI #24429

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Fixed failing NCCL Group test on internal CI #24429

wants to merge 2 commits into from
+3 −3

Conversation

chaserileyroberts
Copy link
Contributor

Previously, this test was failing with

INTERNAL: NCCL operation ncclRecv( recv_buffer.opaque(), ToNcclCount(dtype, count), nccl_dtype, source_rank->value(), comm_, se::gpu::AsGpuStreamValue(stream)) failed: invalid argument (run with NCCL_DEBUG=WARN for details). Last NCCL warning(error) log entry (may be unrelated) 'Recv : invalid root 3 (root should be in the 0..1 range)'.

This was because of a mix of the bad channel ids and this missing replica_count attribute.

frgossen
frgossen suggested changes Apr 1, 2025
View reviewed changes
xla/tests/nccl_group_execution_test.cc Outdated
@@ -127,13 +127,13 @@ XLA_TEST_F(NcclGroupExecutionTest, NcclGroupSendRecvNoWhileLoop) {

XLA_TEST_F(NcclGroupExecutionTest, BidirectionalCommunication) {
const absl::string_view kModuleStr = R"(
HloModule module_main, entry_computation_layout={()->(u32[], u32[])}
HloModule module_main, entry_computation_layout={()->(u32[], u32[])}, replica_count=4
Copy link
Member

Choose a reason for hiding this comment

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

Shouldnt this be numn_partitions?
I would exoect this to be with channel IDs in replica mode

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I changed it to num_parititions, but I still needed to remove the channel ids to make the test pass.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@frgossen frgossen frgossen requested changes

@rosiezou rosiezou rosiezou approved these changes

@appujee appujee appujee approved these changes

At least 1 approving review is required to merge this pull request.

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@chaserileyroberts @frgossen @rosiezou @appujee