v0.6.0

• Support Spark 2.4.0 and TensorFlow 1.12.0.
• Simplify docker/travis setup.

Release 0.5.0

• Support Spark 2.3.1 and Tensorflow 1.10.0

Release 0.4.0

• Manually append meta data via appendShape()
• Skip operations that have no input and no output

Release 0.2.9

This is the final release for 0.2.9.

Notable changes since 0.2.8:

• Upgrades tensorflow dependency from version 1.1.0 to 1.3.0
• map_blocks, map_row APIs now accept Pandas DataFrames as input
• Adds support for tensorflow variables. Note that these variables cannot be shared between the worker nodes.

Release 0.2.8

This is the final release for 0.2.8.

Notable changes since 0.2.5:

• uses the official java API for tensorflow
• support for image ingest (see inception example)
• support for multiple hardware platforms (CPU, GPU) and operating systems (linux, macos). Windows should also work but it has not been tested.
• support for Spark 2.1.x and Spark 2.2.x
• some usability and performance fixes, which should give a better experience for users
• more flexible input names for mapRows.

Release 0.2.8 RC0

This is the first release candidate for 0.2.8.

Notable changes:

• uses the official java API for tensorflow
• support for image ingest (see inception example)
• support for Spark 2.1.x
• the same release should support both CPU and GPU clusters
• some usability and performance fixes, which should give a better experience for users