0

I can by now run my TensorFlow model's trainer locally with Python 3.6 because Cloud ML now supports Python 3. However, if I try to run the same model in the cloud with trainingInput.pythonVersion: "3.6" in config.yaml I receive the following error:

INVALID_ARGUMENT: Field: python_version Error: The specified Python version '3.6' is not supported.
- '@type': type.googleapis.com/google.rpc.BadRequest
  fieldViolations:
  - description: The specified Python version '3.6' is not supported.
    field: python_version

What can be done about this?

Prabhakar
  • 1,138
  • 2
  • 14
  • 30
Drux
  • 11,992
  • 13
  • 66
  • 116

1 Answers1

1

trainingInput.pythonVersion: "3.5" (instead of "3.6") apparently works.

I hope the fact that I am now runing Python 3.6 (python -V) locally, but Python 3.5 in the cloud won't cause subtle issues with Cloud ML Engine down the road.

Drux
  • 11,992
  • 13
  • 66
  • 116
  • 1
    You can see the full list of supported packages and versions at https://cloud.google.com/ml-engine/docs/runtime-version-list. – MatthewScarpino Dec 24 '17 at 16:28
  • FYI I have by now installed Python 3.5 (next to Python 3.6) on my local development machine and am now using this (i.e. a single) version for running trainers both locally and in the cloud. – Drux Dec 24 '17 at 16:53
  • 2
    bummer 3.6 isn't supported. now i have to rewrite all my literal string interpolations – Lem Ko Jun 21 '18 at 15:45