You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
mypy is buggy with ParamSpecs which forces us to insert type-ignore pragmas in certain scenarios. pyright is faster, less buggy, and gives more helpful error messages. However it doesn't support custom plugings, so some of the mazepa decorators would have to be retyped. pyright design doesn't allow us to add custom decorators on top of attrs(init=True), so all of the flow schema classes and taskable operation classes would have to have custom __init__'s. pyright doesn't support bounding a TypeVar by a Union, so some @overload's would have to be added, specifically in the tensor_ops module. Lastly, some type stub modifications would need to be done for torch and pytorch_lightning, and possibly numpy.
Uh oh!
There was an error while loading. Please reload this page.
mypy
is buggy with ParamSpecs which forces us to insert type-ignore pragmas in certain scenarios.pyright
is faster, less buggy, and gives more helpful error messages. However it doesn't support custom plugings, so some of themazepa
decorators would have to be retyped.pyright
design doesn't allow us to add custom decorators on top ofattrs(init=True)
, so all of the flow schema classes and taskable operation classes would have to have custom__init__
's.pyright
doesn't support bounding a TypeVar by a Union, so some@overload
's would have to be added, specifically in thetensor_ops
module. Lastly, some type stub modifications would need to be done fortorch
andpytorch_lightning
, and possiblynumpy
.mypy
bug trackers:Concatenate
false positive (no problem withpyright
) python/mypy#14168Concatenate
false negative (caught bypyright
, but notmypy
) python/mypy#14169The text was updated successfully, but these errors were encountered: