Add ignore_index option to Accuracy metric

[stas-sl]

🚀 Feature

It would be nice if Accuracy metric accepted ignore_index as it works for StatsScore or torch.nn.CrossEntropyLoss

[github-actions]

Hi! thanks for your contribution!, great first issue!

[avinashsai]

Hi,
Anyone working on this? I can try the implementation.

[SkafteNicki]

@avinashsai please feel free to do so :]
Ping me either here or on slack if you need any help.

[avinashsai]

@SkafteNicki I have a query:
This ignore_index should ignore that class label and calculate accuracy for the rest?
Ex: y_pred - [1, 0, 1, 1, 0], y_true - [1, 0, 0, 1, 0]

Accuracy(y_pred, y_true, ignore_index=0) = 100%
Ignoring class label '0', there are only two samples belonging to class '1' and hence accuracy is 100%.

[SkafteNicki]

@avinashsai that at least how I would like the argument to work, as it is similar to how torch.nn.CrossEntropyLoss works.

[SkafteNicki]

@avinashsai how is it going here?

[avinashsai]

@SkafteNicki I am working on it. Will submit PR in a couple of days.

[SkafteNicki]

Closed by #166