SGNAS
SGNAS copied to clipboard
The compatibility for the later version pytorch
In the original implementation, an error will occur "RuntimeError: A view was created in no_grad mode and is being modified inplace with grad mode enabled" using PyTorch 1.8.
Thanks for the pull request! To make sure the gradient flow and update the architecture generator, I think that the architecture parameter may not be detached in line 222 of lookup_table_builder.py.