Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Unnecessary tf.distribute.MirroredStrategy() scopes in the distributed custom training tutorial #33764

Open
netw0rkf10w opened this issue Oct 27, 2019 · 0 comments

Comments

@netw0rkf10w
Copy link

@netw0rkf10w netw0rkf10w commented Oct 27, 2019

URL(s) with the issue:

https://www.tensorflow.org/tutorials/distribute/custom_training

Description of issue (what needs changing):

In the tutorial, with strategy.scope() appears almost everywhere, which gives the impression that it is required that those locations. However, when checking this example, I found that the scope is not required at all! Only strategy.experimental_run_v2 and strategy.reduce suffice. Therefore, I would propose to modify the tutorial code to keep only a minimum amount of with strategy.scope() when necessary (e.g. when defining the model).

Submit a pull request?

Yes, I can create a PR but only when you have approved that this is valid. Thanks.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
3 participants
You can’t perform that action at this time.