All geek questions in one place

Google Cloud ML Tensorflow Version

docs for setting up Google Cloud ML offer to install the version of Tensorflow r0.11. I noticed that the TensorFlow features recently available in r0.12 throw exceptions when running on Cloud ML. Is there a timeline for Cloud ML that supports r0.12? Will switching between r0.11 and r0.12 be optional or mandatory?

+6
source share
1 answer

Yes, you can specify --runtime-version = 0.12 to get assembly 0.12. This is a new feature and is documented at https://cloud.google.com/ml/docs/concepts/runtime-version-list

Please note, however, that assembly 0.12 is not yet considered stable, and the exact design of Tensorflow may change. Once Tensorflow version 1.0 is available, it will also be supported, and Tensorflow versions prior to 1.0 will begin to become obsolete.

(see https://cloud.google.com/sdk/gcloud/reference/beta/ml/jobs/submit/training for use.)

+7
source

Source: https://habr.com/ru/post/1014847/

More articles:


All Articles