Yong Tang

@LinGeLin That will depends on how you want to use unstructured data in TF. If the unstructured data is merely image which you can assume it is an image (jpg/bmp/etc)...

We could automate though a manual trigger (vs. PR push) might still be needed. The GitHub Actions is not super reliable, and PyPI.org has one issue of you cannot override...

@rllin As was discussed in the meeting, the extra compression support likely will be landed in tensorflow-io package. I will transfer the issue to tensorflow/io repo and we can continue...

@rllin Adding additional ZSTANDARD compressions (outside of zlib/gzip) likely falls under tensorflow-io as tensorflow normally only support a very minimal compression types (only zlib/gzip, not just for TFRecordDataset but also...

@AlexFuster if you have black installed, can you just run `black audio_ops.py` and see if the file is formatted automatically?

@AlexFuster Can you rebase the PR with the latest master branch? We recently made some enhancement on testing which I think will resolve some issues shown in GitHub Actions CI.

> Oh yeah, I forgot to mention It does just support floating point audios. Is this a problem? @AlexFuster The existing implementation support both int16 and float32. Though I think...

@AlexFuster The test failures are legitimate. The reason being 1. `Tensor` in tensorflow may not be treated as python native value (e.g, `1`, `1.1`, etc). Things like `int(x)` may not...

With `int16` tests may pass, though in the original implementation the resample actually support passing `rate_in`, `rate_out` as tensor and input with unknown shape. I think with new implementation this...

> Why would you want a tensor with a shape other than () being passed as rate_in? @AlexFuster Sorry it should be `a tensor without constant value`. In graph mode...