Ricky Chen

It seems the issue was updated. Right now, the error is func outputs a tuple but the initial state is a tensor. Their shapes need to match. (Error message could...

Our implementation is written purely in PyTorch, so it just conforms to however PyTorch is executed. That is, if you run it normally it only allocates one process. But if...

Yeah, just define it anywhere. In order to use `odeint_adjoint`, it's good practice to define them as part of the module. ``` global_params = ... class ODEfunc(nn.Module): def __init__(self): self.parameters...

I've just enabled complex types on the latest commit. Give it a try and if there are issues, let me know. To install the latest commit: ``` pip install git+https://github.com/rtqichen/torchdiffeq...

The current API isn't designed with this mind unfortunately. The only way to implement it is to use the internal functions. The equivalent to the `step` method is the `_advance`...

No sorry, zero progress has been made since 2019. If anyone wants to submit a PR for this, I can approve it.

Do you have a minimal working example, and your PyTorch version?

Not entirely sure I follow the issue. If the issue is that the ODE needs to be updated once you solve past t_{i+1}, then you can also set the time...

There is a min_step argument if you use any RK adaptive-step solver like the default `dopri5`. https://github.com/rtqichen/torchdiffeq/blob/master/torchdiffeq/_impl/rk_common.py#L146 ``` odeint(..., method="dopri5", options={"min_step": 0.01}) ``` Depending on your application, it might be...

Thanks for letting me know. Latest version has been updated on pip.