This function should only be used to load models saved in python.
For it to work correctly you need to use torch.save
with the flag:
_use_new_zipfile_serialization=True
and also remove all nn.Parameter
classes from the tensors in the dict.
load_state_dict(path, ..., legacy_stream = FALSE)
a named list of tensors.
to the state dict file
additional arguments that are currently not used.
if TRUE
then the state dict is loaded using a
a legacy way of handling streams.
The above might change with development of this in pytorch's C++ api.