import torch
from torchmetrics.wrappers import BootStrapper
from torchmetrics.regression import MeanSquaredError
metric = BootStrapper(MeanSquaredError(), num_bootstraps=20)
values = [ ]
for _ in range(3):
    values.append(metric(torch.randn(100,), torch.randn(100,)))
fig_, ax_ = metric.plot(values)