Recoded _max_value method using a dictionary

[vfdev-5]

Description:

• Removed _max_value method and added a dictionary

Related to #5502

cc @datumbox

[facebook-github-bot]

💊 CI failures summary and remediations

As of commit 93a7905 (more details on the Dr. CI page):

---

💚 💚 Looks good so far! There are no failures yet. 💚 💚

---

This comment was automatically generated by Dr. CI (expand for details).

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

[datumbox]

Thanks @vfdev-5. A few comments + there are related failing tests.

[datumbox]

LGTM, thanks! One last nit for your consideration:

[pmeier]

We could probably also add an xfail test like

import torch
import pytest


@pytest.mark.xfail(
    reason="torch.iinfo() is not supported by torchscript. See https://github.com/pytorch/pytorch/issues/41492."
)
def test_max_value_iinfo():
    @torch.jit.script
    def max_value(image: torch.Tensor) -> int:
        return 1 if image.is_floating_point() else torch.iinfo(image.dtype).max

to nudge us to switch as soon as support is added.

[datumbox]

@pmeier Good call. If they add support our test will fail and we will certainly know about it :)

@vfdev-5 Feel free to merge on green CI.

[datumbox]

JIT on internal tests REALLY doesn't like this and fails with:

runtime error: 9.22337e+18 is outside the range of representable values of type 'long'

Replacing this with the following works:

def _max_value(dtype: torch.dtype) -> int:
    if dtype == torch.uint8:
        return 255
    elif dtype == torch.int8:
        return 127
    elif dtype == torch.int16:
        return 32767
    elif dtype == torch.int32:
        return 2147483647
    elif dtype == torch.int64:
        return 9223372036854775807
    else:
        return 1

I'll cherrypick it after it's merged internally.