search

ryanjulian / rllab

rllab is a framework for developing and evaluating reinforcement learning algorithms, fully compatible with OpenAI Gym.
Other
1 stars 21 forks source link

Asynchronous plotting for Theano #111

Closed jonashen closed 6 years ago

jonashen commented 6 years ago

Issues are equivalent to TensorFlow. See #1.

jonashen commented 6 years ago

@ryanjulian I discovered quite an interesting bug. When I implement async plotting using multiprocessing, Ubuntu works with no issue but Mac OSX will throw a glfw-related segmentation fault. However, when I switch to multithreading, Mac OSX works with no issue but Ubuntu takes forever to load the window, if at all (the likely culprit being Python's GIL). I'll implement both by first checking which OS the current machine running the code is, and then either fire up a thread or process accordingly.

jonashen commented 6 years ago

PR

ryanjulian commented 6 years ago

Great debugging. SGTM