Update to resolve CI warnings

[puhuk]

Fixes #2649

Description:
Update to resolve CI warnings
Will send soon to fix warnings in test_precision.py

Check list:

• New tests are added (if a new feature is added)
• New doc strings: description and/or example code are in RST format
• Documentation is updated (if required)

[puhuk]

For warnings from tests/ignite/metrics/test_precision.py as I think those warnings are intended to warn.
In my thought, to fix those warnings, corner cases like (torch.zeros(size=(10,)), torch.randint(0, 2, size=(10,)), 1) should be deleted.
Is that correct? Or is there other methods to fix those warnings?

[vfdev-5]

For warnings from tests/ignite/metrics/test_precision.py as I think those warnings are intended to warn.
In my thought, to fix those warnings, corner cases like (torch.zeros(size=(10,)), torch.randint(0, 2, size=(10,)), 1) should be deleted.

Can you point to the code what you are talking about ?

[puhuk]

For warnings from tests/ignite/metrics/test_precision.py as I think those warnings are intended to warn.
In my thought, to fix those warnings, corner cases like (torch.zeros(size=(10,)), torch.randint(0, 2, size=(10,)), 1) should be deleted.

Can you point to the code what you are talking about ?

For this line, because it generates data with zeros,

ignite/tests/ignite/metrics/test_precision.py

Line 154 in c506c31

(torch.zeros(size=(10,)), torch.randint(0, 2, size=(10,)), 1),

it occurs error with being set to zero in

ignite/tests/ignite/metrics/test_precision.py

Line 130 in c506c31

assert precision_score(np_y, np_y_pred, average=sk_average_parameter) == pytest.approx(pr_compute)

.

When I add zero_divion=1 like this precision_score(np_y, np_y_pred, average=sk_average_parameter, zero_division=1), it occurs error when comparing computed metric and reference.

[vfdev-5]

@puhuk here is the link on why this corner case was added : #2007
So, we need to keep this test case. Why you set zero_division=1, have you tried zero_division=0 ?

[puhuk]

No, I didn't because document says the default value is warn and it acts as 0 in https://scikit-learn.org/stable/modules/generated/sklearn.metrics.precision_score.html
But when I set with zero_division=0, those warning fixed.

[vfdev-5]

@puhuk please fix code formatting with https://github.com/pytorch/ignite/blob/master/CONTRIBUTING.md#formatting-code

bash ./tests/run_code_style.sh fmt

[vfdev-5]

lgtm, thanks

[vfdev-5]

@puhuk actually, there are still warnings:

1. about model weights : https://github.com/pytorch/ignite/runs/7903177538?check_suite_focus=true#step:15:1850 and https://github.com/pytorch/ignite/runs/7903177538?check_suite_focus=true#step:15:1856
2. "RuntimeWarning: y and y_pred should be of dtype long when entry type is binary and average!=False", https://github.com/pytorch/ignite/runs/7903177538?check_suite_focus=true#step:15:1844

Please fix them as well

warning remains

[puhuk]

Still some warings in https://github.com/pytorch/ignite/runs/7963618423?check_suite_focus=true#step:15:1826
Is it can be resolved in ignite?

[vfdev-5]

Still some warings in https://github.com/pytorch/ignite/runs/7963618423?check_suite_focus=true#step:15:1826 Is it can be resolved in ignite?

No, these warnings are in installed packages, we can't do anything with them.

[vfdev-5]

Thanks @puhuk

[vfdev-5]

This version check is incorrect

[vfdev-5]

You have to check for torchvision version here