Open HaynesLi opened 1 year ago
There is no training data. When I mentioned path to the training folder I meant the directory where you want your generated models to be saved. The training data(input to the agent) is given by the unity simulation itself.
Thus it should be a self-learning type?
Yes, you just reward the agent based on its output and the agent learns by trying to maximize the reward.
But after I reviewed all codes in this project. I did not find where you define the network structure.
You only need to define the environment and the reward system. The network structure is predefined in ML-Agents.
In which part the code using the ML-Agents api? I found the Addreward() function in your code.
I followed the instruction of the project. However I met the problem during the colab part in which I can not find the training data folder. Could you offer it? Thank!