Old warnings from F.interpolate docs should be cleaned up

[jbschlosser]

Issue description

The F.interpolate() docs include a couple of old warnings regarding the behavior of align_corners and recompute_scale_factor referencing PyTorch v0.3.1 and v1.6.0, respectively. These should be removed as no longer relevant. Also, somewhat related: the docstring for recompute_scale_factor does not specify what the default value for that arg is and should be updated as well.

cc @brianjo @mruberry @albanD @jbschlosser @walterddr @kshitij12345

[pejato]

Hi @jbschlosser , I see the code that should be changed. Could I take this on?

I also see this block of code in interpolate that references the 1.6.0 change, should this be removed as well?

def interpolate( ... ):

    ...

    if recompute_scale_factor is None:
        # only warn when the scales have floating values since
        # the result for ints is the same with/without recompute_scale_factor
        if scale_factors is not None:
            for scale in scale_factors:
                if math.floor(scale) != scale:
                    warnings.warn(
                        "The default behavior for interpolate/upsample with float scale_factor changed "
                        "in 1.6.0 to align with other frameworks/libraries, and now uses scale_factor directly, "
                        "instead of relying on the computed output size. "
                        "If you wish to restore the old behavior, please set recompute_scale_factor=True. "
                        "See the documentation of nn.Upsample for details. "
                    )
                    break

[jbschlosser]

Hey @pejato! We would certainly welcome a PR if you'd like to contribute :) And yes you're correct that the warning in the code should be removed now as well.

[pejato]

@jbschlosser Great! I can aim to have a PR up by tomorrow. Does this need a unit test?

[albanD]

Hi,

There are a couple tests in test/test_nn.py that handle these warnings for example in test_upsamplingNearest2d. You can update one of the test to properly assert that no warning is raised anymore (self.assertEqual(len(w), 0)) and remove the warning catchers for the rest.