what replaces aiplatform.CustomContainerTrainingJobRunOp in google-cloud-pipeline-components V2

[raihan1khan]

this example currently doesn't work - https://github.com/GoogleCloudPlatform/fraudfinder/blob/main/vertex_ai/06_formalization.ipynb, it's using CustomContainerTrainingJobRunOp. Any idea

[raihan1khan]

@rimolive

[rimolive]

cc @connor-mccarthy

It will be helpful if you can share the output of the block that's not working.

[raihan1khan]

so this block is not working with sdk v2

custom training job component - script

train_model_op = vertex_ai_components.CustomContainerTrainingJobRunOp(
    display_name=JOB_NAME,
    model_display_name=MODEL_NAME,
    container_uri=IMAGE_URI,
    staging_bucket=bucket_name,
    dataset=dataset_create_op.outputs["dataset"],
    base_output_dir=bucket_name,
    args=ARGS,
    replica_count=replica_count,
    machine_type=machine_type,
    training_fraction_split=train_split,
    validation_fraction_split=val_split,
    test_fraction_split=test_split,
    model_serving_container_image_uri=MODEL_SERVING_IMAGE_URI,
    project=project_id,
    location=region,
).after(dataset_create_op)

[giliolimirko]

CustomContainerTrainingJobRunOp has been removed in v2 and therefore it does not exists. You should run kfp=1.8

[raihan1khan]

@giliolimirko - that's why i was asking what replaced custom container training job run op, if we are using 2.x then what's the solution. How can i use custom container training job

[rimolive]

I'd prefer to hear from the Google team to hear more about this. cc @chensun @zijianjoy @connor-mccarthy

As far as I could see, and a quick research, says that there's no replacement for this component. Instead, you should begin using Containerized Python components.