Open avacaondata opened 2 years ago
It is sadly a known issue. The GPU improvements would be marginal thou, since the GPU would only be used in computing the SGD (at least this is what I understood up until now).
If anyone wants to tackle this issue and try and figure out what's breaking, it would be nice for sure!
I have the same problrm as well.Hopefully this can be solved in the future....
System information
Describe the current behavior
When trying to run this example: https://www.tensortrade.org/en/latest/examples/train_and_evaluate_using_ray.html (the official tutorial on how to use tensorboard), I cannot do it with GPU enabled. However, if I run the ray part but using Cartpole as the environment, with the same libraries environment and everything, it can use GPUs. Therefore this has something to do with tensorboard explicitly, not with ray or pytorch. The error is the following
Describe the expected behavior The expected behavior is that tensortrade envs return the data in the same format as other envs such as cartpole, so that tensors can be sent to the GPU.
Code to reproduce the issue
Other info / logs
@carlogrisetti @notadamking