score:2

Accepted answer

There is a little bit about the number of threads to use at

https://www.tensorflow.org/versions/r0.9/how_tos/reading_data/index.html#batching

Unfortunately, I don't think there is a simple answer for batch sizes. The efficient batch size for a network depends on a lot of details about the network. In practice, if you care about optimal performance you're going to need to do a bunch of trial and error (maybe starting from the values used by a similar network).