I was wondering if this site was a good place to ask about design patterns or ways to design code that works well for data science specifically. I had in mind to post a specific question on good coding practice using TensorFlow. What stack exchange site is a good place to post my question? Is this one good?
In particular, I wanted to know what good practices for how to design TensorFlow code when using GPUs on a single machine (not distributed machines) but potentially multiple GPUs. For example, what are the differences I should take into consideration when scaling up to use GPUs? Should I for example, stop evaluating how my model is doing every iteration using the whole data set? Are there other considerations I might not even be aware when using GPUs? I've had some memory complaints from TensorFlow when I use GPUs, so maybe I might not be allocating enough memory? How does the memory and the GPU interact with each other?