Enable ROCm CI support #1786
Enable ROCm CI support #1786
Conversation
…g ubuntu folder for cuda Dockerfile.
…Fixed error in integration_tests.py. Fixed lint errors.
…_job_v2.yml for integration_test_8gpu.yaml.
…ily available to run the workflow.
meta-cla
bot
added
the
CLA Signed
|
No ciflow labels are configured for this repo. |
…n in linux_job_v2.yml.
… and move_aws_steps_inside_setup_rocm branch.
…tures tests inside integration_test_8gpu_features.yaml. Using linux_job_v2.yml from the main branch. Rolled back to using 8 GPU runner for ROCm.
|
|
||
| jobs: | ||
| build-test: | ||
| uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main |
There was a problem hiding this comment.
Instead of using the main branch here, could you try to use the branch 7311 from pytorch/test-infra#7329 to test this out:
| uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main | |
| uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@7311 |
We can revert it back to main if the test works
There was a problem hiding this comment.
@akashveramd We merged pytorch/pytorch#164769 so we can switch back to main branch now. @huydhn also confirmed that the AWS role was updated, and the latest retriggered job passed after that: https://github.com/pytorch/torchtitan/actions/runs/18391595425/job/52493188786?pr=1786
Please retrigger after switching back to main branch. It seems there'll be a new error related to artifact directory creation.
There was a problem hiding this comment.
@jithunnair-amd: Switched back to main. I see failures in the rocm workflow-
mkdir: cannot create directory 'artifacts-to-be-uploaded': Permission denied
https://github.com/pytorch/torchtitan/actions/runs/18424663754/job/52504224020?pr=1786
4 participants
This PR is based out of the original PR #1260.
The original PR was created in a different fork, and it was having issues setting up aws inside the workflow. Since the workflow was running from a forked PR.