Closed smorad closed 1 year ago
On it!
I see where that comes from When we use the functional RNN, we pass a set of target params that are tensors (not nn.Parameters) but the RNN isn't happy to receive a reguar tensor and not a parameter. The solution will be to use nn.Parameters even for detached params I guess, though it's goona be a bit painful
An alternative might be to use an LSTMCell
but it would be slower.
Describe the bug
The tutorial at https://github.com/pytorch/rl/blob/main/tutorials/sphinx-tutorials/dqn_with_rnn.py produces an error when run.
To Reproduce
Steps to reproduce the behavior.
Please try to provide a minimal example to reproduce the bug. Error messages and stack traces are also helpful.
Please use the markdown code blocks for both code and stack traces.
System info
Describe the characteristic of your environment:
Reason and Possible fixes
It might be due to how the torch RNN classes do strange things with their weights for efficiency.
Checklist