Reward problem in Crawler？

[lyy0095]

I have download the training metrics of Crawler scene from https://huggingface.co/infinitejoy/MLAgents-Crawler/tensorboard .

The Cumulative Reward nearly to 2500 when Episode Length come to 1000, but in the script of Project/Assets/ML-Agents/Examples/Crawler/Scripts/CrawlerAgent.cs，the only AddReward() function was limited to [0, 1] (TouchedTarget() is never used in the whole script), so the biggest reward will be 1000, how can we got 2500?
Is there any other secret AddReward() function I do not recognized?

[lyy0095]

For any researchers, the reward problem answer is HernandezEduin/Trained-ML-Agents-Actors#1

If you are interested in examples of ML-Agent, I strongly recommend you to read Trained-ML-Agents-Actors.