Multiclassexactmatch

multiclassexactmatch

Metric MulticlassExactMatch from torchmetrics (torchmetrics.classification.MulticlassExactMatch)

When to invoke this skill

The user has predictions + ground truth and asks to evaluate with MulticlassExactMatch, or mentions torchmetrics.classification.MulticlassExactMatch directly, or wants the standard torchmetrics implementation.

Reference signature

from torchmetrics.classification import MulticlassExactMatch

# MulticlassExactMatch(num_classes: int, multidim_average: Literal['global', 'samplewise'] = 'global', ignore_index: Optional[int] = None, validate_args: bool = True, **kwargs: Any) -> None

Library docstring

Compute Exact match (also known as subset accuracy) for multiclass tasks.

Exact Match is a stricter version of accuracy where all labels have to match exactly for the sample to be
correctly classified.

As input to ``forward`` and ``update`` the metric accepts the following input:

- ``preds`` (:class:`~torch.Tensor`): An int tensor of shape ``(N, ...)`` or float tensor of shape ``(N, C, ..)``.
  If preds is a floating point we apply ``torch.argmax`` along the ``C`` dimension to automatically convert
  probabilities/logits into an int tensor.
- ``target`` (:class:`~torch.Tensor`): An int tensor of shape ``(N, ...)``.

As output to ``forward`` and ``compute`` the metric returns the following output:

- ``mcem`` (:class:`~torch.Tensor`): A tensor whose returned shape depends on the ``multidim_average`` argument:

    - If ``multidim_average`` is set to ``global`` the output will be a scalar tensor
    - If ``multidim_average`` is set to ``samplewise`` the output will be a tensor of shape ``(N,)``

If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
which the reduction will then be applied over instead of the sample dimension ``N``.

Args:
    num_classes: Integer specifying the number of labels
    multidim_average:
        Defines how additionally dimensions ``...`` should be handled. Should be one of the following:

        - ``global``: Additional dimensions are flatted along the batch dimension
        - ``samplewise``: Statistic will be calculated independently for each sample on the ``N`` axis.
          The statistics in this case are calculated over the additional dimensions.

    ignore_index:
        Specifies a target value that is ignored and does not contribute to the metric calculation
    validate_args: bool indicating if input arguments and tensors should be validated for correctness.
        Set to ``False`` for faster computations.

Example (multidim tensors):
    >>> from torch import tensor
    >>> from torchmetrics.classification import MulticlassExactMatch
    >>> target = tensor([[[0, 1], [2, 1], [0, 2]], [[1, 1], [2, 0], [1, 2]]])
    >>> preds = tensor([[[0, 1], [2, 1], [0, 2

Quick recipe

import torchmetrics.classification as _m
score = _m.MulticlassExactMatch(y_true, y_pred)

Don'ts

• Don't reimplement when the library version handles edge cases (NaN, ties, empty inputs) better than a hand-rolled formula.
• Always check the library version's argument order — sklearn is (y_true, y_pred) while torchmetrics is (preds, target).