Rémy Portelas

icon indicating an email [email protected]

icon location R&D Scientist in Deep Learning at Ubisoft Bordeaux. PhD in Auto. Curriculum for Deep RL sup. by P.Y. Oudeyer (Inria Bordeaux) & Katja Hofmann (MSR Cambridge

Results 2 issues of Rémy Portelas

EfficientZero high memory consumption / keeps increasing after replay buffer is full

7 comment

I am currently experimenting on scaling EfficientZero to learning setups with high-data regimes. As a first step, I am running experiments on Atari, with a replay buffer of 1M environment...

The first selfplay worker uses the same seed for all parallel environments

2 comment

I might have found an unexpected behavior in how parallel training environments are being seeded. I am referring to this line: https://github.com/YeWR/EfficientZero/blob/c533ebf5481be624d896c19f499ed4b2f7d7440d/core/selfplay_worker.py#L112 Because the rank of the first selfplay worker...