Improve metrics tests

[vfdev-5]

The idea is to add available_device pytest fixture into the appropriate metric tests. For example:

@pytest.mark.parametrize("n_times", range(5))
- def test_multiclass_input(n_times, test_data):
+ def test_multiclass_input(n_times, available_device, test_data):
    y_pred, y, num_classes, batch_size = test_data
-    cm = ConfusionMatrix(num_classes=num_classes)
+    cm = ConfusionMatrix(num_classes=num_classes, device=available_device)
    ...

This will generate 2 more tests and run test_multiclass_input on cpu, cuda if available, mps if available.

No need to add available_device fixture to the tests:

1. checking if an error is raised, for example test_confusion_matrix.py::test_num_classes_wrong_input:

def test_num_classes_wrong_input():
    with pytest.raises(ValueError, match="Argument num_classes needs to be > 1"):
        ConfusionMatrix(num_classes=1)

2. running distributed tests, e.g. tests inside TestDistributed or test_distrib_*.

Files to update:

• test_average_precision.py
• test_classification_report.py
• test_average_precision.py
• test_cohen_kappa.py
• test_confusion_matrix.py
• test_cosine_similarity.py
• test_entropy.py
• test_fbeta.py
• test_frequency.py
• test_hsic.py
• test_js_divergence.py
• test_kl_divergence.py
• test_loss.py
• test_maximum_mean_discrepancy.py
• test_mean_absolute_error.py
• test_mean_pairwise_distance.py
• test_mean_squared_error.py
• test_multilabel_confusion_matrix.py
• test_mutual_information.py
• test_precision_recall_curve.py
• test_roc_auc.py
• test_roc_curve.py
• test_root_mean_squared_error.py
• test_top_k_categorical_accuracy.py

Additional lot of metrics:

• clustering / test_calinski_harabasz_score.py ... test_silhouette_score.py
• nlp / test_bleu.py ...
• regression / ...

Please, split the work into multiple PRs: one PR for updates in a single file.
This issue can't be assigned to a single person and can be tackled in collaboration.

For the questions/details, please ask here or on our discord.

[github-actions]

Hey 👋, I've just created a thread for this issue on PyTorch-Ignite Discord where you can quickly talk to the community on the topic.

^{🤖 This comment was automatically posted by Discuss on Discord}