M1 macs and disabling eager evaluation

Open
#498 8 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
25/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Stale
Tech stack
r, tensorflow

Research direction

Start by reading the installation code that selects tensorflow-deps and checking how the loaded TensorFlow version is resolved on M1 Macs. Reproduce the 2.5.0 request and compare it with the installed 2.6.0 result. Done means establishing whether 2.5.0 is supported on M1 and documenting or implementing the supported version-selection behavior.

Written by the indexing model from the issue text.

Description

Hi there!

I have managed to install TF version 2.6.0 on my M1 mac! Hooray!

However, I am really hoping to install TF version 2.5.0 for greta, as we would like to work out a way to test if we can disable "eager evaluation" in TF, and that is apparently not possible in 2.6.0 (https://stackoverflow.com/questions/53429896/how-do-i-disable-tensorflows-eager-execution)

Looking under the hood at the installation code, it seems that I can get TF version 2.5.0 to install by changing the code to be:

install("tensorflow-deps==2.5.0", channel = c("apple", "conda-forge"))

However this doesn't seem to work as I expected, and version 2.6.0 is still loaded.

As I understand it, some of the issues around installing version 2.5.0 on M1 macs are that it just isn't built on M1 architecture, is this the case?

I appreciate that this might be a rather complicated request, so no worries if this isn't in scope, just wanted to see if there was an outside change we could change the TF version for M1 macs, as if we have to use "eager mode" in greta, it will require re-writing substantial amounts of the code base.

(We've written a bit more around migration in greta, if you're interested, see https://github.com/greta-dev/greta/issues/306)

Dominant language
R
Stars
1.3k
Forks
316
PR merge metrics
No merged PRs in 30d

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from rstudio/tensorflow

All issues in rstudio/tensorflow

Similar issues

More R issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.