This splits off the test requirements update from #167 to reduce the number of files modified in that diff. It supersedes #166.
There are several changes due to the mypy update, which caught a few new issues with type annotations. In general strict type adherence is best, so we've created some new variables for things such as Tensor conversion (specifically, in UnetDataTransform and VarNetDataTransform).
Since we now have a standardized requirements framework and versioning we can drop support for pre-1.7.0 PyTorch FFTs and drop packaging as a requirement.
This splits off the test requirements update from #167 to reduce the number of files modified in that diff. It supersedes #166.
mypyupdate, which caught a few new issues with type annotations. In general strict type adherence is best, so we've created some new variables for things such asTensorconversion (specifically, inUnetDataTransformandVarNetDataTransform).packagingas a requirement.