Cloud Training EfficientDet (d4)

I've tried to train a efficientdet-d4 model, based on a config. Because configs for efficientdets and tf2 are just for TPUs I tried this option, right starting ckpt (D4) was used from the model zoo.

# From the tensorflow/models/research/ directory cp object_detection/packages/tf2/setup.py . gcloud ai-platform jobs submit training whoami_object_detection_date +%m_%d_%Y_%H_%M_%S \ --job-dir=gs://${MODEL_DIR} \ --package-path ./object_detection \ --module-name object_detection.model_main_tf2 \ --runtime-version 2.1 \ --python-version 3.6 \ --scale-tier BASIC_TPU \ --region us-central1 \ -- \ --use_tpu true \ --model_dir=gs://${MODEL_DIR} \ --pipeline_config_path=gs://${PIPELINE_CONFIG_PATH}

This example from the docs .md throws several error before the last one:

• python version should be 3.7
• problem with accessing a restricted bucket (it seams as an open issue regarding tpu cloud train...)

Current obstacle is:

Does anybody experienced something similar?

Thanks,
Alex