tensorflow / tensorflow/java

API to read weights from ckpt file

Open
#420 5 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Java
Stars
928
Forks
227
PR merge metrics
No merged PRs in 30d

Description

System information

  • TensorFlow version (you are using): Nightly
  • Are you willing to contribute it (Yes/No): No

Describe the feature and the current behavior/state.
This API would help in reading the weights from ckpt file. Currently, there is no such feature in the Java SDK but the same can be achieved in python via CheckPointReader.

Will this change the current API? How?
This shouldn't affect the preexisting API. It would be a simple addition over it.

Who will benefit with this feature?
I was developing a Federated Learning Application and was stuck at the issue where I was not able to get the weights from the ckpt file (Reason: No API in Java SDK) and neither could I write a signature method in tflite for the same (Reason: tflite doesn't support dynamic tensor shapes). So this would help promote active development in FL on Android clients.

Contributor guide

Open the contributing guide

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.

Research direction

Start by comparing the requested Java SDK capability with Python's CheckPointReader and reviewing how TensorFlow checkpoint weights are represented for Java and Android clients. Done means adding a Java API that reads weights from ckpt files without changing the preexisting API, with behavior suitable for federated-learning use cases.

Written by the indexing model from the issue text.

Assessment

Tech stack
java, tensorflow
Domain
api, machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.