Sachin Prasad

I was able to reproduce the issue in 2.10.0 version by enabling OneDNN in colab, please find the gist [here](https://colab.sandbox.google.com/gist/sachinprasadhs/25bd523cdda4800460edf313c7b27f1a/untitled73.ipynb).

Disabling the `XNNPack` will make the operations to fallback to the default implementation which is resulting slower performance. If there no particular requirement to disable the `XNNPack` then you can...

@Mollylulu , Could you please refer the above comment and close this issue. Thanks!

Thanks for reporting the bug, it's strange that it works well when the input is provided is of `int64` max value and throws error when the input is greater than...

@willi3by , You need to install tensorflow nightly version for that PR to be available. Since this PR could not make it at the time of 2.10.0 brach cut. It...

TFLite now supports 5D in selectV2, could you please try in the latest nightly version or the Tensorflow version 2.10 and confirm. More details can be found here https://github.com/tensorflow/tensorflow/blob/master/RELEASE.md#major-features-and-improvements-1

@callumm-graphcore , Could you please refer the new API [tf.random.experimental.stateless_shuffle ](https://www.tensorflow.org/api_docs/python/tf/random/experimental/stateless_shuffle) and check if this is useful in your case. Thanks!

Usually, Checks on GPU and CPU will not be same considering the performance impact. In few cases it is documented also. Check the comment [here](https://github.com/tensorflow/tensorflow/issues/57174#issuecomment-1246288003) for details. Thanks!

Apologies for the delayed response, as of now TFLite supports limited number of operations under builtin library. You can create a custom op and register under GPU kernel to allow...

Could you try compiling the model and train it with some data to check if you are facing the same behavior as saving and loading non compiled model. Now, as...