Vertex AI: Custom Training Job and Prediction Using Managed Datasets Reviews

I believe this qwiklab needs a refresh. There are some minor changes in how notebooks are being created (button names, etc.), but the major issue is that the provided code doesn't work when run manually (Task 3). I received following errors: 1) AttributeError: module 'pandas' has no attribute 'Int64Index' - this was easy to fix by removing GCS import functionality not used in this lab 2) Bigger issue are errors related to the fact that when run manually the code tries to substitute treat name and sex as NUMERIC fields and fails to convert them to float. I didn't have enough time to fix this. Regardless to the code issues while running manually the code works correctly in following tasks, so I was able to successful train, deploy and predict. Nevertheless I was almost about to stop after failing the Task 3.

Remigiusz S. · Reviewed about 2 years ago

Rita A. · Reviewed about 2 years ago

Joseph M. · Reviewed about 2 years ago

Wahaj S. · Reviewed about 2 years ago

Oshadha S. · Reviewed about 2 years ago

Jordi B. · Reviewed about 2 years ago

Edwin Isaac S. · Reviewed about 2 years ago

Joe M. · Reviewed about 2 years ago

Muhammad Saif Z. · Reviewed about 2 years ago

Oliwer B. · Reviewed about 2 years ago

Dzmitry S. · Reviewed about 2 years ago

It worked well on my own gcp account, but on the lab account the training job failed with a permission to bigquery error. Also in both cases there was an error related to pandas version, solved by installing pandas 1.4.0 or below.

Sergio Adolfo D. · Reviewed about 2 years ago

Vishaal S. · Reviewed about 2 years ago

mist. S. · Reviewed about 2 years ago

Vijay Y. · Reviewed about 2 years ago

Tot N. · Reviewed about 2 years ago

Palak S. · Reviewed about 2 years ago

Hitesh K. · Reviewed about 2 years ago

Rohan D. · Reviewed about 2 years ago

Ankit R. · Reviewed about 2 years ago

Sergio Adolfo D. · Reviewed about 2 years ago

Omar Y. · Reviewed about 2 years ago

Out of date and doesn't work from Task 4 onwards.

Sam A. · Reviewed about 2 years ago

Omar Y. · Reviewed about 2 years ago

1. Notebook creation is poorly explained and does not match UI (what tab?, what kind of instance?) this results in issues when trying to train in the Jupyter environment 2. Most of the lab description is either out of date with GCP UI or badly described. The jupyter screenshot does not match the real instance. etc 3. Deployment of trained model fails with: Failed to create endpoint "titanic_endpoint" due to the error: Model artifact directory is expected to contain exactly one of: [saved_model.pb, saved_model.pbtxt]. Please re-upload the Model with correct artifact directory..

Topoleanu T. · Reviewed about 2 years ago