synandi

Hi @shijy16, thank you for reporting the issue! I was able to replicate the issue using tf-nightly(2.13.0-dev20230406). Please find the screenshot below.  Thank you!

Hi @trickiwoo Thank you for reporting the issue! I was able to replicate the issue using the latest tf-nightly(2.13.0.dev20230327). Please find the screenshot below.  We are investigating this issue...

@tilakrayal I was able to replicate the issue in Ubuntu and colab. Please find the gist [here](https://colab.sandbox.google.com/gist/synandi/c7b9d6765441e78b1a374f481b1c7333/59727.ipynb) for TF v2.10 and [here](https://colab.sandbox.google.com/gist/synandi/4612a2921b20e8e081c594a0706d6a0c/59727-nightly.ipynb) for tf-nightly. Thank you!

Hi @dongyaoli10x, kindly let us know the OS Platform and Distribution used to exceute the above code. I tried executing it in Ubuntu 22.04. Please find the output below. ...

Thank you for sharing the details. I replicated the issue in CentOS. Please find the complete error below. ``` Running tests under Python 3.9.16: /home/ynandi/miniconda3/envs/tf/bin/python [ RUN ] TestDummy.test_dummy_confusing_bug 2023-02-27...

To fix the issue, as a workaround you can modify the assertion to compare the numpy values of the tensors instead of the tensors themselves. Here's the modified code: ```...

@mazeltovlee I was able to replicate the issue on Google Colab with TF v2.11 on both CPU(output is 0.0) and GPU(output is nan). Please find the gist for the same...

Hi @mazeltovlee, Apologies for the delay. On a GPU, floating point operations are performed using a special hardware that is optimized for parallel computation. This hardware may use different algorithms...

Closing as stale. Please reopen if you'd like to work on this further.

@anna-charlotte, It's generally not recommended to extend the `tf.Tensor` class, as it can lead to unexpected behavior and make your code harder to maintain. If you need to add custom...