Hi, you may have already covered this but I was wondering what sets Ray apart from TensorFlow’s own parallel training methods? We have had problems with the built-in parallel training and wanted to know if we could run into the same with Ray.
Hi Brian, welcome to the Ray community! 
Can you elaborate more on what issues you’re having with the built-in parallel training with TensorFlow?
The built-in methods just never seemed to work well. We think it had something to do with the recombination algorithm but the models accuracy always seemed to suffer after training in parallel with tensorflow. Just wondering if its the same with Ray?