ML Agents don't move while training

cdrakep · January 16, 2021, 6:00pm

Version information:
ml-agents: 0.21.0,
ml-agents-envs: 0.21.0,
Communicator API: 1.2.0,
TensorFlow: 2.3.1

StreamExecutor device (0): GeForce RTX 2080 Ti, Compute Capability 7.5

Hi, I’m not sure what the issue is here. Check out this video -

The agents move when Unity isn’t connected to mlagents-learn, but when I set it up to train they just do the first step and freeze. This is consistent across all the examples and mlagents-learn yaml files.

I’ve tried ml-agents releases 10, 11, and 12 with Unity 2018.4.30f1, 2019.4,18f1, and 2020.1.13f1. Same behavior across all versions, works fine with pre-trained inference, does not work when I try to us mlagents-learn. I’ve let it sit overnight and produce a 50,000,000 iteration file (toward end of video) The agents appear to move, but the training is bad so they fail.

I’ve been googling and watching videos of people doing the exact steps I’m doing, and the agents actually move and these values increase.

cdrakep · January 16, 2021, 7:28pm

Followed the steps on our OSX build machine, works as expected.

cdrakep · January 16, 2021, 8:04pm

Got it working -

I noticed on OSX it was saying “framework: pytorch” and on windows it said “framework: tensorflow”

I installed Anaconda and made a clean environment with these instructions →

github.com

miyamotok0105/unity-ml-agents/blob/master/docs/Installation-Windows.md

# Installing ML-Agents for Windows

ML-Agents supports Windows 10. While it might be possible to run ML-Agents using other versions of Windows, it has not been tested on other versions. Furthermore, ML-Agents has not been tested on a Windows VM such as Bootcamp or Parallels.

To use ML-Agents, you install Python and the required Python packages as outlined below. This guide also covers how set up GPU-based training (for advanced users). GPU-based training is not required for the v0.3 release of ML-Agents. However, training on a GPU might be required by future versions and features.

## Step 1: Install Python via Anaconda

[Download](https://www.anaconda.com/download/#windows) and install Anaconda for Windows. By using Anaconda, you can manage separate environments for different distributions of Python. Python 3 is required as we no longer support Python 2. In this guide, we are using Python version 3.6 and Anaconda version 5.1 ([64-bit](https://repo.continuum.io/archive/Anaconda3-5.1.0-Windows-x86_64.exe) or [32-bit](https://repo.continuum.io/archive/Anaconda3-5.1.0-Windows-x86.exe) direct links).

<p align="center">
    <img src="images/anaconda_install.PNG" 
        alt="Anaconda Install" 
        width="500" border="10" />
</p>

We recommend the default _advanced installation options_. However, select the options appropriate for your specific situation.

<p align="center">
    <img src="images/anaconda_default.PNG"

This file has been truncated. show original

Except when it throws the error about numpy I reinstalled mlagents with “pip install mlagents --use-feature=2020-resolver” and now the trainer works as expected.

Topic		Replies	Views
Agent Not Moving When Training Unity Engine ML-Agents , com_unity_ml-agents	1	943	August 28, 2021
Setting up mlAgents for actual trainig is not working for me Unity Engine ML-Agents , com_unity_ml-agents	7	2514	March 2, 2021
The ml-agent doesn't move, the arrow controls don't even work. Unity Engine ML-Agents , Question , com_unity_ml-agents	7	483	January 5, 2025
Trying Build And Run Project with already trained Ml Agent Unity Engine ML-Agents , Question , com_unity_ml-agents	8	1963	October 7, 2020
ML-agents command line environment doesn't sync with Unity environment. Unity Engine ML-Agents , Performance , Unity-Documentation , com_unity_ml-agents	3	650	February 2, 2022

ML Agents don't move while training

Related topics