Closed dalbhanj closed 4 years ago
@dalbhanj thank you for proposing the issue, I'll take it over and fix it up.
@PatrickXYS I don't see this issue, module launch_crd
should come from container image, What's the image we are using now? Is it a regression issue after your refactor?
Yes, it should come from container image, which should be the image hosted by me. There's no launch_crd
now. It should be a regression issue after my factor, I'm working on it.
@PatrickXYS As we don't have CI now, try to make sure things work before check in.
It should be fixed by the merged PR, @dalbhanj let us know if it's good to go.
This isn't resolved yet bcoz it only works in 'us-west-2' region, yeah?
After investigation, region settings are not the root cause. Let's keep the investigation and updated.
@PatrickXYS
Can you extract region as an argument? Why do we stick to us-west-2
now?
Notebook affected: https://github.com/aws-samples/eks-kubeflow-workshop/blob/master/notebooks/05_Kubeflow_Pipeline/05_03_Pipeline_mnist.ipynb
After you submit pipeline, step 3 kubeflow-launch-tfjob fails with this error Traceback (most recent call last): File "/ml/launch_tfjob.py", line 22, in
import launch_crd
ImportError: No module named launch_crd