You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I wanted to ask if it would be possible to provide a mechanism to use torch.float32 as dtype for some of the adaptive solvers.
On Apple M1 (mps) torch.float64 is not supported...
I encountered this issue as well, but I think all you need to do is to specify options={'dtype': torch.float32} in your odeint call (assuming you're using odeint with an adaptive solver like I do).
Hello,
Thanks for your amazing package!
I wanted to ask if it would be possible to provide a mechanism to use
torch.float32
asdtype
for some of the adaptive solvers.On Apple M1 (mps)
torch.float64
is not supported...E.g. here
torchdiffeq/torchdiffeq/_impl/dopri5.py
Line 6 in d6ee52b
the
dtype
is fixed independent of the input.I am new to your package (and also a bit to pytorch, since I usually use Julia), therefore, sorry if I was overlooking something obvious.
The text was updated successfully, but these errors were encountered: