search

rlworkgroup / garage

A toolkit for reproducible reinforcement learning research.
MIT License
1.84k stars 309 forks source link

Reproducibility issue #2332

Open YY-GX opened 1 year ago

YY-GX commented 1 year ago

Hi all, I've used the deterministic.set_seed but still get different results. It seems that the samples that start from itr2 become different (itr0 & itr1 are the same). Do you have ideas about this issue?

sadegh16 commented 1 year ago

I think you need to set environment seed too as follows:

env._env.seed(seed) env.action_space.seed(seed)