search

IntelLabs / coach

Reinforcement Learning Coach by Intel AI Lab enables easy experimentation with state of the art Reinforcement Learning algorithms
https://intellabs.github.io/coach/
Apache License 2.0
2.33k stars 461 forks source link

The same argsort is performed batch_size times #28

Closed Shmuma closed 6 years ago

Shmuma commented 6 years ago

In this line we have np.argsort applied to all quantilles in the loop: https://github.com/NervanaSystems/coach/blob/7bdba396d20bd54726a296af22b4e688af7f00cb/agents/qr_dqn_agent.py#L55

It would be batch_size times faster to calc it once.

itaicaspi-intel commented 6 years ago

Thanks! This improvement was added in 11faf19