Cannot auto save model after training wiith ml-agents0.13.1

I use ml-agents 0.13.1 to train the agent according to the document “Making a New Learning Environment”.Everything was right during the training process.However,after the training was done,the information that INFO:mlagents.trainer:Exported /.nn file showed,I can’t find the model with that path.There is no model folder in ./ml-agents/.I did not interrupt the training by pressing Ctrl+C or stopping Unity Editor.
This happened several times.First,I installed python virtual environment and training in it,the autosave failed.Then I deactivated the virtual environment and download the python packages again,I trained again and it worked.I can find the .nn file in the./ml-agents/run-id/behavior name/.But,today, when I use it again with different run-id,it failed.Then I trained 3DBall and still can’t get the model.

OS:win10 64bit 1909
Unity:2018.4.14
ML-Agents:0.13.1
TensorFlow:2.0.1
Python:3.6.8

When the console prints

Exported ./models/ppo/<behavior_name>.nn file

The model should be at that path. . means your current directory. Make sure you are running the mlagents-learn command from the right ml-agents directory.

The fact that you got it to run with a specific run-id and not another makes me think that this is probably due to running the command in the wrong directory,

Oh,I find those models in C:\user\models. I didn’t change the directory when training.Thanks a lot,really appreciate.