Closed mshearer0 closed 2 years ago
Building and deploying my own custom image with tensorflow-hub following instructions on pg 317,318 allows kubeflow pipeline to complete
Updating pipeline_kubeflow_gcp_buckets.py:
input_bucket and output_bucket to point to my cloud storage bucket (and copying over data and components folders to that location).
tfx_image = os.environ.get( "KUBEFLOW_TFX_IMAGE", "gcr.io/insert your project_id/ml-pipelines-tfx-custom:0.22.0",
Thank you @mshearer0 You raise a good point. I'll make the gcr url dynamic.
@mshearer0 If you have a moment, please check out https://github.com/Building-ML-Pipelines/building-machine-learning-pipelines/pull/31 and let me know if the implementation would work for you.
Thanks, should pipeline.py also be updated
And
tfx_image = os.environ.get( "KUBEFLOW_TFX_IMAGE", "gcr.io/oreilly-book/ml-pipelines-tfx-custom:0.21.4", )
In pipeline_gcp_cloud_ai.py?
@hanneshapke #31 does not work for me, could you tell me a proper image path?
@jazzsir - did you build and deploy your own image into gcr.io?
@mshearer0 I finally applied my custom image to my own registry server, thanks
@hanneshapke did you run kubeflow example code without any errors at all? my case #36
Hi @jazzsir,
Yes, it did. Could it be a permission issue to the Google Container Registry or GCS?
PS: Check out the latest updates to the example code: https://github.com/Building-ML-Pipelines/building-machine-learning-pipelines/releases/tag/examples_based_on_tfx_1.4
Error pulling image /oreilly-book/ml-pipelines-tfx-custom:0.22.0 from container registry gcr.io. Is it globally visible?