Closed StephenGordan closed 9 months ago
The same problem in gameformer Interaction Prediction,but training suscessully in fameformer Open-loop Planning .
Something wrong with optimizer.step() or data ?
I am not sure what's the problem. My guess is that a deadlock occurs in the data loading phase, so I would like to suggest reducing the number of worker threads in the DataLoader.
Although the above issue hasn't been resolved, I managed to complete the model training on another device using a Docker environment. One issue is that Docker can't visualize in real-time. Is it possible to save the visualization data as videos or images?
See #10.
Why does it get stuck during training without giving any error messages?