search

vita-epfl / CrowdNav

[ICRA19] Crowd-aware Robot Navigation with Attention-based Deep Reinforcement Learning
MIT License
560 stars 166 forks source link

About using the GPU #35

Closed luxiaojuntokyo closed 3 years ago

luxiaojuntokyo commented 3 years ago

When I run: python train.py --policy sarl Every time there will be: INFO: Using device: cpu

I wonder how can I use GPU to speed up the training?

My server's setting, GPU: GTX 1080Ti × 3 (VRAM: 11GB × 3)

Thanks a lot.

YuejiangLIU commented 3 years ago

Please try and add the '--gpu' argument in your command

ChanganVR commented 3 years ago

@luxiaojuntokyo the code runs faster on CPU because the speed bottleneck is on CPU-intensive operations. Feel free to explore and optimize the code such that you could use GPU to speed it up.

luxiaojuntokyo commented 3 years ago

@luxiaojuntokyo the code runs faster on CPU because the speed bottleneck is on CPU-intensive operations. Feel free to explore and optimize the code such that you could use GPU to speed it up.

Changan, thank you for your reply. If I get any result, I will comment here.

luxiaojuntokyo commented 3 years ago

Please try and add the '--gpu' argument in your command

LIU Yuejiang, thank you for your reply.