consolidate Linux workflows on CPU and GPU #7189
Conversation
… into consolidate-linux-ci
| eval "$(${CONDA_PATH} shell.bash hook)" | ||
|
|
||
| echo '::group::Set PyTorch conda channel and wheel index' | ||
| # TODO: Can we maybe have this as environment variable in the job template? For example, `IS_RELEASE`. |
There was a problem hiding this comment.
Will open an issue in test-infra for that
| # Prepare conda | ||
| CONDA_PATH=$(which conda) | ||
| eval "$(${CONDA_PATH} shell.bash hook)" |
There was a problem hiding this comment.
IIUC, all the generic jobs come with conda pre-installed and use bash, right? If so, could we maybe run conda init before we enter the script? Otherwise stuff like conda activate doesn't work without running the highlighted commands before.
There was a problem hiding this comment.
At least the conda part is not true for Windows. @osalpekar is this intentional?
There was a problem hiding this comment.
It is not intentional - since this is the first application of the windows job, we'll need to identify and patch these gaps as we go :) This line would need to be added to the windows_job to ensure conda is installed: https://github.com/pytorch/test-infra/blob/main/.github/workflows/macos_job.yml#L84
| # Prepare conda | ||
| CONDA_PATH=$(which conda) | ||
| eval "$(${CONDA_PATH} shell.bash hook)" |
There was a problem hiding this comment.
It is not intentional - since this is the first application of the windows job, we'll need to identify and patch these gaps as we go :) This line would need to be added to the windows_job to ensure conda is installed: https://github.com/pytorch/test-infra/blob/main/.github/workflows/macos_job.yml#L84
|
Hey @pmeier! You merged this PR, but no labels were added. The list of valid labels is available at https://github.com/pytorch/vision/blob/main/.github/process_commit.py |
Reviewed By: vmoens Differential Revision: D44416540 fbshipit-source-id: 13d9ced6c5723b1cf99d8a45a38683c09f51c533
Spin off from #7149. Instead of having CPU and GPU test separately, we can have them in a single workflow as well. We should prefer that to avoid copy-pasting to much stuff where it is not necessary.
cc @seemethere