from torch import randn
from torchmetrics.regression import CosineSimilarity
metric = CosineSimilarity()
values = []
for _ in range(10):
    values.append(metric(randn(10,), randn(10,)))
fig, ax = metric.plot(values)