Lukas Geiger

@mdanatg Thanks a lot for taking a look.

> Is this still an issue @Saduf2019 Yes

@mdanatg Thanks for the info. It's a bit cumbersome for library authors who want to support multiple versions of TensorFlow, but the workaround works fine. > Unfortunately the branch cut...

@Saduf2019 yes this is still an issue. The current temporary workaround relies on private TensorFlow function which isn't a longterm solution. See https://github.com/tensorflow/tensorflow/issues/44613#issuecomment-732215759

Is there already a decision made on whether a fix for this will make it into the 2.5 release?

With TF 2.6.0 this issue is now is even more subtle since `tensorflow/python/keras` was replaced with `keras`, however the old import still work. This leads to the following weirdness in...

@gadagashwini @jvishnuvardhan I think this issue should be reopend, as the stack overflow post you linked only worksaround the fact that `model.save('gs://bucket/model.h5')` isn't natively supported, whereas the TF file format...

@k-w-w This issue has been tagged as 2.1, has there been any progress on it so far?

@sayakpaul That is correct, the `SavedModel` format correctly supports GCS, however when saving a `.h5` file GCS is still not supported.

I am now seeing the following warnings on main when running some test e.g.: ```shell pytest tests/v1/engine/test_async_llm.py::test_load -s ``` ``` ERROR 05-31 01:04:14 [prometheus.py:77] Error during metrics cleanup: expected str,...