Accuracy analyis in deep learning project

[RitikaPrasai1]

I am working on a deep learning project and trying to label data to provide as training samples. I have already tried by increasing, decreasing the polygon size, number and exported those using deep learning algorithms in ArcGIS Pro.However, when I train my model using those exported samples- I always receive the accuracy as 'nan' . Could someone suggest me what is the cause. I have attached screenshot here. Thank you!

[DanPatterson]

There are many links trying to explain nan on the web.

A web search using .... deep learning accuracy nan

yields many examples, for instance....

python - Deep-Learning Nan loss reasons - Stack Overflow 

It is hard to figure out exactly your workflow from just the image.  Perhaps you can provide a textual description as well

[RitikaPrasai1]

Thank you!