The "Training and Deploying a TensorFlow Model in Vertex AI" lab's (lab_exercise_long.ipynb) CustomContainerTrainingJob errors out with CPU quota violation about training-data-analyst

[yumemio]

Cloud Console shows that the assigned quota for aiplatform.googleapis.com/custom_model_training_c2_cpus is zero for all regions.

Switching the machine type to N1 or E2 could help, since aiplatform.googleapis.com/custom_model_training_cpus has a more generous limit (20 or 42 CPUs depending on the region).

I've tested with e2-standard-4 and the pipeline is running as expected. I'm not sure if this workaround conflicts with the Qwiklabs policy of no deviation from instructions!:roll_eyes:

[MrCsabaToth]

@yumemio Thanks for debugging. This makes sense, the e boxes are efficient and so less resource consuming. I may give it a go when I have time.