How to access Google bucket files using python tensorflow?

Question

If I want to access Google bucket files using python during training job on Google Machine Learning engine (not locally). What is the correct approach from the following choices (if any)?

• use the built in open(filePath) python function where filePath is valid google bucket path e.i. 'gs://bucket/folder/file.ext'

• use the Python Client for Google Cloud Storage
  https://pypi.python.org/pypi/google-cloud-storage

the first choice is clearly much easier, if it is valid, does it requires special set up such as: setting specific bucket permissions or default storage class.

best regards.

Answer 1

Neither this or this. One should use the tensorflow Gfile module as follow:

import tensorflow as tf
with tf.gfile.GFile("gs://bucket/file", "r") as f:
        content = f.read()