Easy distributed TensorFlow on Hops Hadoop
yarntf simplifies the distributed TensorFlow submit model, for running machine learning applications on Hadoop YARN clusters.
In general it is as simple as follows.
- In your code: replace tf.train.ClusterSpec() and tf.train.Server() with yarntf.createClusterServer()
- On your cluster: submit the application with Hops-TensorFlow
Your ClusterSpec is generated automaticaly and the parameter servers stopped when all workers are completed. Specify the number of worker, ps and resources on submit.
For more details see the examples.
Work In Progress
Development is still in an early stage. Contributions are very welcome!
yarntf and Hops-TensorFlow is released under an Apache 2.0 license.
Release history Release notifications | RSS feed
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.