Python Codec

Important Note: You must have RL-Glue Core installed to use this codec.

Important Links

Frequently Asked Questions

We're waiting for your questions!


Python Codec Speed Tests

These are ad-hoc tests done without being careful.  They are meant to give a flavor of the speed of this codec.

Without NumPy (old)


Type 1 (Big Observations)     :: 498 milliseconds per step (2 steps per second)
Type 2 (Regular Observations) :: .60 milliseconds per step (1667 steps per second)

With NumPy (thanks to James Bergstra)

Type 1 (Big Observations)     :: 41 milliseconds per step (24.4 steps per second)
Type 2 (Regular Observations) :: .60 milliseconds per step (1667 steps per second)

That's a 10x speedup on large observations!

Comments