Closed javirk closed 3 years ago
Ok, I created a new environment, reinstalled torchdyn & Co. and now there are no problems with NeuralDE
.
pytorch 1.9+cuda10.2
torchdyn 0.2.2.1
torchdyn 0.2.2.1
torchsde 0.2.5i
But then the tutorials are not updated. I would do a PR but I don't know if they're kept like that for some reason.
Hi @javirk. The docs are outdated in several places and will be massively improved for the next release planned for early July. NeuralDE
is the legacy version kept for backward compatibility, which has now been replaced by NeuralODE
.
Out of curiosity, did you install torchdyn
via pip
? 0.2.2.1
is outdated and 0.3.2
should be on PyPI.
Yes, I installed it via pip
. I just tried in a blank new environment and it retrieves 0.2.2.1
for some reason. I have pip 21.1.3
(last version).
pip install torchdyn
with pip 21.1.3
retrieves torchdyn==0.4.*
for me at the moment, check if this works on your end as well.
Yes, it works now. Thank you very much.
I think this can be closed now.
I am seeing in the docs (classification or Lagrangian neural nets tutorials) that the ODE is instantiated as
NeuralDE
. When I try it in my environment, with torchdyn 0.2.0, I getNameError: name 'NeuralDE' is not defined
. However, the tutorials in github say it'sNeuralODE
, and that works for me.Are the docs wrong or
NeuralDE
andNeuralODE
are two different things and I'm doing something wrong?Thanks!