What is the usage of Caffe2:TypeMeta
?
I found this description:
TypeMeta is a thin class that allows us to store the type of a container such as a blob, or the data type of a tensor, with a unique run-time id. It also stores some additional data such as the item size and the name of the type for run-time inspection. Definition at line 324 of file typeid.h.
But, really I am not sure how does it work.
In the following example, it gets z.dtype()
which is at:Tensor
and then it saves it in dtype
as caffe2::TypeMeta
.
caffe2::TypeMeta dtype;
z = torch::rand({1,4});
dtype = z.dtype();
Not sure what is the point of that and how it is possible to get the data back from dtype
as it is done in pytorch/adam.cpp at master · pytorch/pytorch · GitHub