Hi,
I'm trying to follow this tutorial and found many errors
https://cloud.google.com/ai-platform/training/docs/algorithms/object-detection
The python script fails for many reasons, I've created an issue here tensorflow/tpu#820
Can you create a notebook for each step of the configuration? (service account for images stored on buckets from different projects, python version, docket SDK update, and mainly step by step python script execution for turning jpeg images into tfrecords, etc)