tensorflow-io 0.16.0 release

[yongtang]

With TensorFlow 2.4 release candidate out soon (this week or next), we can consider having a v0.16.0 release for 2.3, as was discussed in our last meeting.

Here is the list of items for 0.16.0:

• Bump tensorflow-io version to 0.16.0 (PR Bump the version to 0.16.0 #1157)
• Update RELEASE.md to capture 0.16.0 changes (PR Bump the version to 0.16.0 #1157)
• Update README.md to update 0.16.0 release date (PR Add entry of 0.16.0 release date on README.md #1159)
• When GitHub Actions complete the build against the master, download the artifact with the name tensorflow-io-release. Uncompress the content and all binaries for release are inside.
• Do minimal sanity test to make sure at least the first section of getting started example code works locally.
• Push all binaries to PyPI.org.
• Address the points related to docker releases in [docker] push version specific tag during docker release #1153 (comment)
• Add an overview description to our docker releases. Currently, it is blank.

And follow up immediately after 0.16.0 release:

• Update docs/tutorials/colorspace.ipynb to use pip install tensorflow-io (instead of tensorflow-io-nightly). (PR: [tutorials] modify tfio version in notebooks #1160)
• Update docs/tutorials/kafka.ipynb to use pip install tensorflow-io (instead of tensorflow-io-nightly). (PR: [tutorials] modify tfio version in notebooks #1160)
• Update docs/tutorials/elasticsearch.ipynb to use pip install tensorflow-io (instead of tensorflow-io-nightly). (PR: [tutorials] modify tfio version in notebooks #1160)
• Push to tensorflow.org/io (with the help from docs team).

Finally, post 0.16.0 release:

• Bump TF version to 2.4.0 (rc) (PR Convert AZFS to use C modular file system API #1143)
• Resolve any issues in Convert AZFS to use C modular file system API #1143 to use modular file systems C API (as discussed in our last meeting). (PR Convert AZFS to use C modular file system API #1143)
  Note since Expose TF_RegisterFilesystemPlugin C API tensorflow#43680 has been merged, we are finally unblocked.
• Send email to SIG IO

/cc @vlasenkoalexey @terrytangyuan @kvignesh1420 @BryanCutler @dmitrievanthony

[yongtang]

@kvignesh1420 Do you have a PyPI.org account? If you do I can add you to the maintainer of tensorflow-io repo https://pypi.org/project/tensorflow-io/ if you prefer.

[kvignesh1420]

@yongtang my PyPI handle is https://pypi.org/user/kvignesh1420/.

Also, I have added a few points to the task list above. (related to docker releases and updates to notebooks)

[yongtang]

@kvignesh1420 Thanks! Invite for PyPI.org has been sent out.

[kvignesh1420]

@terrytangyuan if time permits, can this issue #1142 be addressed before the release? We can build and add the r-latest tag to this image.

[terrytangyuan]

Looks like we are already releasing so I don't think we will have enough time to fix that.

[kvignesh1420]

@yongtang here is a draft of the overview content in docker hub.

-- START --

TensorFlow-IO Docker Images

TensorFlow I/O is a collection of Dataset, streaming, and file system extensions maintained by TensorFlow SIG-IO that extend TensorFlow's built-in support for io related operations.

The dockerfiles are available on Github along with the list of tags.

Base Image Tags

• 0.x.y, latest, and nightly tags come with TensorFlow-IO pre-installed. Versioned tags contain their version, the latest tags contain the latest release and the nightly images come with the TensorFlow-IO nightly Python package. Additionally, since Tensorflow-IO is dependent on Tensorflow, the images come with the compatible Tensorflow version.

• -devel images are based on Tensorflow's custom-op images and come with Bazel. These are ideal for exploring and developing changes to TensorFlow-IO.

Running Containers

Start the container with the latest release of the package.

$ docker run -it --rm tfsigio/tfio bash

Start the development container and mount the ${PWD} to explore and build the ops.

$ docker run -it --rm --name tfio-dev --net=host -v ${PWD}:/v -w /v tfsigio/tfio:latest-devel bash

Please check the docs for development using docker.

-- END --

cc: @terrytangyuan @BryanCutler @vlasenkoalexey

[yongtang]

Thanks @kvignesh1420 👍 ! The README page of the docker hub has been updated.

[yongtang]

I tested manually with release binary candidates and sanity looks fine.

[yongtang]

Below are the release binary candidates:

0f558f6f8add7c061a7359e49fa7fd16c3b20fe19b432e0b5dcdcf1b097fad75  tensorflow_io-0.16.0-cp36-cp36m-macosx_10_13_x86_64.whl
f3878516cfa9d7eb69dd4039ef6565e6d5912782ddbcc1e027064c85d493b516  tensorflow_io-0.16.0-cp36-cp36m-manylinux2010_x86_64.whl
f3c1088ea03dd1beb4ada99aa9319801dae32e01bdfd599a14e0a47a2d04c75c  tensorflow_io-0.16.0-cp36-cp36m-win_amd64.whl
4b613a4129667b85b8cf09481e95babe15b7f597d8162c6614c359174f1aed44  tensorflow_io-0.16.0-cp37-cp37m-macosx_10_13_x86_64.whl
638174fcd38a8c18090c67be0c47b060c95cbdeabf9e9823566f8805717d6acc  tensorflow_io-0.16.0-cp37-cp37m-manylinux2010_x86_64.whl
32b77ff85ceec5c6390d500dfaae41ef55c28b6497eeaf4ee3a9e84249fa0035  tensorflow_io-0.16.0-cp37-cp37m-win_amd64.whl
ba39848f61dfdaffaf8df35814d339316bcb8272d2fa6aff8552c4e9c8c4c278  tensorflow_io-0.16.0-cp38-cp38-macosx_10_13_x86_64.whl
29a343c7921f4a9d880ae7ad5b200c4dda2e1affc857e8e05087383f18a41eb2  tensorflow_io-0.16.0-cp38-cp38-manylinux2010_x86_64.whl
427b099eb17f70872c0f7dbe81360b88a93b93d57e1c3135dbbff682f5ae2f46  tensorflow_io-0.16.0-cp38-cp38-win_amd64.whl

[yongtang]

The release of 0.16.0 is available:

https://pypi.org/project/tensorflow-io/0.16.0/#files

https://github.com/tensorflow/io/releases/tag/v0.16.0

[kvignesh1420]

@MarkDaoust @lamberta can we update the docs in the website as the PR #1160 has been merged?

[yongtang]

PR #1143 will cover most of the post 0.16.0 items. I think we haven't send out email to SIG IO list yet. Will send out shortly.