Ml-agents: CPU vs GPU which one is fast in "ML Agent" Learnig?
I'm going to buy a new computer.
Is it true that "ML Agent" does not use gpu performance for learning?
Does "ML Agent" have more CPU performance than GPU?
"1 Agent " = "1 Thread" true?
Hongsungchan
All 3 comments
Hi @Hongsungchan
In our current implementation of PPO, we do not make an efficient use of the GPU. This will change in future versions but for the time being, updating the model is faster on a CPU most of the time.
Each agent does not run on an individual thread. In Python, the observations of the agents are processed in batches.
vincentpierre
on 22 Apr 2019
Thank you for the discussion. We are closing this issue due to inactivity. Feel free to reopen it if you鈥檇 like to continue the discussion though.
xiaomaogy
on 21 May 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 20 May 2020
Related issues
gerardsimons
路
3Comments
jlanis
路
4Comments
DavidLining
路
3Comments
DVonk
路
3Comments
RavenLeeANU
路
4Comments
Most helpful comment
Hi @Hongsungchan
In our current implementation of PPO, we do not make an efficient use of the GPU. This will change in future versions but for the time being, updating the model is faster on a CPU most of the time.
Each agent does not run on an individual thread. In Python, the observations of the agents are processed in batches.