Ml-agents: How to check GPU works properly.
Hi guys,
I installed cuda and cudnn for tensorflow gpu version. But I do not know how to check if it works or just work as CPU version. Can someone give me some hints for that? thanks..
RavenLeeANU
All 4 comments
According to this thread you can run this command to see if there is a GPU to use.
tf.test.is_gpu_available()
vincentpierre
on 17 Dec 2018
I used this command with result: TRUE, however when I train my GFX card load is not going up at all and training is quite slow. Cpu is like 50-100% busy.
my tf version is 1.12.0
oddEventHorizon
on 29 Dec 2018
ML-Agents is not optimized for GPU usage yet.
In general, Reinforcement Learning is hard to be optimized for GPUs, because especially on-policy algorithms do not sample enough data to efficiently use a GPU. So most of the time, you are better of with a CPU for now. Highly scaled up systems like IMPALA or OpenAI Five take advantage of a GPU.
MarcoMeter
on 2 Jan 2019
This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.
![lock[bot] picture](https://avatars1.githubusercontent.com/in/6672?v=4&s=40)
lock[bot]
on 2 Jan 2020
Related issues
tensorgpu
路
3Comments
gerardsimons
路
3Comments
GuntherFox
路
3Comments
MarcPilgaard
路
3Comments
dlindmark
路
3Comments
Most helpful comment
ML-Agents is not optimized for GPU usage yet.
In general, Reinforcement Learning is hard to be optimized for GPUs, because especially on-policy algorithms do not sample enough data to efficiently use a GPU. So most of the time, you are better of with a CPU for now. Highly scaled up systems like IMPALA or OpenAI Five take advantage of a GPU.