Skip to content

Skip hanging distrib test with pytorch nightly#2816

Closed
vfdev-5 wants to merge 12 commits intopytorch:masterfrom
vfdev-5:skip-hanging-test-distrib-pth-nightly
Closed

Skip hanging distrib test with pytorch nightly#2816
vfdev-5 wants to merge 12 commits intopytorch:masterfrom
vfdev-5:skip-hanging-test-distrib-pth-nightly

Conversation

@vfdev-5
Copy link
Collaborator

@vfdev-5 vfdev-5 commented Jan 6, 2023

@vfdev-5 vfdev-5 marked this pull request as draft January 6, 2023 10:56
@vfdev-5 vfdev-5 closed this Jan 9, 2023
@vfdev-5 vfdev-5 reopened this Jan 9, 2023
@vfdev-5
Copy link
Collaborator Author

vfdev-5 commented Jan 9, 2023

Issue happens with test_idist_parallel_spawn_n_procs_native and with allinit_method as "env://", "tcp://0.0.0.0:29500" and "FILE". While debugging on GHA, process group with init_method="env://" and underlying master_addr="127.0.0.1" or localhost hangs. Same, process group creation hangs with init_method="tcp://0.0.0.0:29500".
Process group creation passes if used master_addr="0.0.0.0" or init_method="tcp://0.0.0.0:29523".

@vfdev-5
Copy link
Collaborator Author

vfdev-5 commented Jan 11, 2023

Closing in favor of #2820

@vfdev-5 vfdev-5 closed this Jan 11, 2023
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.

1 participant