Ray.get_gpu_ids()

RLlib
1

Hi, I am using RLlib and training a dqn agent with default config on a custom env on my Windows 10 and colab. I got the following warning which is so annoying as it gives this warning over 1000 times while training the agent so I lose my print information. I wonder if anyone knows how to fix it?

I already added the following lines to my code, but it did not work.

import warnings
warnings.filterwarnings('ignore')

2021-09-10 21:35:45,807 WARNING worker.py:499 – ray.get_gpu_ids() will always return the empty list when called from the driver. This is because Ray does not manage GPU allocations to the driver process.

image
image1434×282 8.89 KB

2

Hi @deepgravity,
Thanks for the report. Yes, this is a known issue.
The next official ray release 1.7.0 is coming out that includes the fix. Stay tuned!
In the meantime, latest nightly wheel of Ray should also do.
Hope this helps.

1 Like
3

Thanks. I updated my Ray to the new version, as you recommended. But, it did not solve the issue precisely. Actually, now it does not print that line many times, but prints an empty (new) line many times :slight_smile: