Support TensorFlow pluggable devices #2144
Open
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
|
When can this PR be merged into master? any update? |
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Outdated suggestions cannot be applied.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.
PluggableDevice architecture offers a plugin mechanism for registering devices with TensorFlow without the need to make changes in TensorFlow code. It relies on C APIs to communicate with the TensorFlow binary in a stable manner. #2038 also asks for support of pluggable devices.
This PR adds the support for pluggable device. It will try to load all .so files under the given directory from argument --tensorflow_plugins by TF_LoadPluggableDeviceLibrary. With current exported symbol "TF_", the package size is increased 4M and total ratio is increased to 101%.
We will need to add alwaylink=1 in TensorFlow kernels_experimental library (tensorflow/tensorflow#60786).