from torch import rand, randint
from torchmetrics.classification import MultilabelPrecisionAtFixedRecall
metric = MultilabelPrecisionAtFixedRecall(num_labels=3, min_recall=0.5)
metric.update(rand(20, 3), randint(2, (20, 3)))
fig_, ax_ = metric.plot()  # the returned plot only shows the maximum recall value by default