How to solve the problem of Nan value during training?

tristandeleu / pytorch-maml-rl

Reinforcement Learning with Model-Agnostic Meta-Learning in Pytorch

MIT License

827 stars 158 forks source link

How to solve the problem of Nan value during training? #65

Open outshine-J opened 2 years ago

outshine-J commented 2 years ago

Dear author: When I run the train.py file, when it runs for about 5000 iterations, the neural network has a Nan value, which causes the program to break unexpectedly. How can I solve this situation?