Understanding precision and reacall in torchmetric

Documentation. Precision — PyTorch-Metrics 0.10.2 documentation

image

from torchmetrics import Precision
from torchmetrics.classification import BinaryPrecision

preds  = torch.tensor([0, 0, 0, 0])
target = torch.tensor([0, 0, 0, 0])

general_precision = Precision(num_classes=2)
bin_precision = BinaryPrecision()

# both should be same?
general_precision(preds, target)
# (tensor(1.)

bin_precision(preds, target) 
# tensor(0.))

From equation, and from preds and target, as there are now True Positive, the numerator becomes 0 and so the answer should be 0. But in general precision, how does it calcuate this?

[ Same goes to (F1Score - BinaryF1Score, BinaryRecall, Recall). ]

ok, -_-

precision = Precision(num_classes=1, multiclass=False)