torch metrics but for tinygrad that all there is to it. I found the interface really convienents for logging metrics and creating balanced loss functions whihc i found convienient for rapid expirementation and eeking out some extra performance with the same models
python3 -m pip install git+https://github.com/torn8to/tinymetrics.git
or
pip install git+https://github.com/torn8to/tinymetrics.git
from tinymetrics.losses import regression
from tinygrad import Tensor
from numpy as np
p,gt = Tensor(np.array([1,5,7])), Tensor(np.array([1,5,3]))
rmse = regression.MeanSquaredError(rooted=True)
value = rmse(p,gt)