>>> mask = torch.zeros(size=(10, ), dtype=torch.bool)
>>> y = torch.ones(size=(10, 10), dtype=torch.cdouble, requires_grad=True)
>>> y.clone().index_put_((mask, ), y[mask])
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
RuntimeError: index does not support automatic differentiation for outputs with complex dtype.
It also looks like autograd for
torch.Tensor.index_put_is not even being tested for floating point tensors correctly.cc @ezyang @anjali411 @dylanbespalko @mruberry @aocsa