Port macOS tests from CircleCI got GHA #7175
Conversation
| @@ -0,0 +1,84 @@ | |||
| #!/usr/bin/env bash | |||
There was a problem hiding this comment.
Although it seems weird to have a standalone script for a single workflow, this script works fine for Linux as well. See #7149. I'll migrate them to this script next.
|
|
||
| set -euo pipefail | ||
|
|
||
| echo '::group::Prepare conda' |
There was a problem hiding this comment.
This line as well as the accompanying echo '::endgroup::' is used to group the output of some steps in the logs. This is only needed now, since we no longer have individual steps, but rather a single step.
| echo '::endgroup::' | ||
|
|
||
| echo '::group::Set PyTorch conda channel' | ||
| # TODO: Can we maybe have this as environment variable in the job template? For example, `IS_RELEASE`. |
There was a problem hiding this comment.
Note to self: I need to open an issue about that.
| echo '::endgroup::' | ||
|
|
||
| echo '::group::Set PyTorch GPU mutex' | ||
| case $GPU_ARCH_TYPE in |
There was a problem hiding this comment.
Not needed for macOS since we don't have GPU workflows. This is in preparation for the Linux GPU jobs. Will happen a few more times below.
| @@ -1,50 +0,0 @@ | |||
| name: Unit-tests on M1 | |||
There was a problem hiding this comment.
Since that test is already on GHA, I don't think we need the grace period where we run both workflows at the same time. LMK if you disagree.
|
There are currently three distinct failures:
|
|
Failure 1. from the comment above probably stems from an environment issue:
|
As was done with the Linux tests, this PR adds macOS tests to our GHA matrix, but leaves them on CircleCI for now. When they run smoothly for a couple of weeks, we can remove them from CircleCI and thus completing the migration for macOS.
cc @seemethere