Load correct libtpu for pytorch & jax. #1166
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
It turns out jax & pytorch are incompatible (require different libtpu versions). In order to support importing EITHER of them (but not both) we will swap in the correct libtpu during import (by monkey-patching the import code for both). https://b/213335159
rosbo
approved these changes
May 26, 2022
| sed -i "1s/^/from jax._src.cloud_tpu_init import cloud_tpu_init\ncloud_tpu_init()\n/" /opt/conda/lib/python3.7/site-packages/tensorflow/__init__.py | ||
|
|
||
| # Set these env vars so that they don't produce errs calling the metadata server to load them: | ||
| ENV TPU_ACCELERATOR_TYPE=v3-8 |
There was a problem hiding this comment.
Should the session manager set these instead?
This image could eventually be used on TPU v4 or TPU v3 pods (e.g. v3-XYZ).
There was a problem hiding this comment.
I'll stick them in the session-manager as well (which will override the image defaults), but having the defaults helped me release the image quicker.
I doubt we'll be jumping TPU arch's anytime soon though, but I agree with the idea.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
It turns out jax & pytorch are incompatible (require different libtpu
versions). In order to support importing EITHER of them (but not both)
we will swap in the correct libtpu during import (by monkey-patching the
import code for both).
https://b/213335159