Failed : work_element_count > 0 (0 vs. 0)

[kbratsy]

Hello everyone,
After working a little on the first epoch during the training phase, I get the error I show below. When I researched the problem it says tensorflow cuda incompatibility but it didn't work at all. But after running a bit in the first era I'm getting this error. What do you think could cause this?

I did the TensorFlow installation steps in the link and installed it without any errors. But after the first epoch, I got the error I mentioned below. If there was an error about the version mismatch, it would not work at all. But as you can see in the error output, a step is running during the training phase and then it is aborted.
https://www.pugetsystems.com/labs/hpc/how-to-install-tensorflow-1-15-for-nvidia-rtx30-gpus-without-docker-or-cuda-install-2005/

Python 3.6.13
My environment is below,

ERROR

Thank you for your comments.