Distributed training with TensorFlow 2
Note
The spark-tensorflow-distributor package is now deprecated and will not be pre-installed in Databricks Runtime 16.0 ML and above. Ray on Databricks is the recommended replacement for distributed training for a Tensorflow or Keras example.
The spark-tensorflow-distributor is an open-source native package in TensorFlow that helps users do distributed training with TensorFlow on their Spark clusters. It is built on top of tensorflow.distribute.Strategy, which is one of the major features in TensorFlow 2. For detailed API documentation, see docstrings. For general documentation about distributed TensorFlow, see Distributed training with TensorFlow.