python – How can I customize the loss functions in the scikit-learn library?

How can I use the custom loss functions in scikit-learn? For example, instead of using the value of the mean squared, do I want to use the value of the mean squared multiplied by the true value of the sample?
I used the following code snippet:
def my_custom_loss_func (y_true, y_pred):
diff3 = (abs (y_true-y_pred)) * y_true
back diff3

clf = RandomForestRegressor (criterion = my_custom_loss_func)

knn = clf.fit (feam, labm)

I got the following error:
KeyError: