It would be very useful to have high level autograd functions such as:
The questions are:
- Where should they go?
torch.autograd.??
- What kind of api we want to provide: should they be function to function mapping? Or should they be function + value to value mappings?
cc @ezyang @ssnl @albanD @zou3519 @gqchen
It would be very useful to have high level autograd functions such as:
The questions are:
torch.autograd.??cc @ezyang @ssnl @albanD @zou3519 @gqchen