How can I define a new data type of tensor or set grad to bool tensor?

[Geremia]

@ptrblck

Could you explain a bit what a floating point gradient on a bool value would mean and how a parameter update would work?

@fangwei123456 Yes, how would derivatives/gradients be performed on boolean tensors?
This seems related to my question here “What’s the use of torch.bool tensors?” (or my related AI.StackExchange question).