Calling model.predict() from multiple processes

[Rampagy]

Maybe this is an issue with my implementation, tflearn or tensorflow, I guess I am not sure. The code hangs when I call model.predict() in multiple processes, but I know the exact same code works when evaluated in a single process. See my reproduction steps here.

I know I have seen lots of issues with tensorflow hanging when trying to predict within multiple processes, so I am not sure if this is a tflearn issue or a tensorflow issue.

[ckyleda]

Can confirm that this is a significant issue that's existed for literally as long as Tensorflow has - I don't understand how this still hasn't been fixed yet.

There seems to be no way to multiprocess inference in Tensorflow. (At least as far as I have found.)

[Rampagy]

Dang. Thanks for looking into it.

[Lancerchiang]

@Rampagy Hi please refer to the solution I wrote for this issue. Multiprocessing works fine with tensorflow.
#tensorflow/tensorflow#8220