Learning with the LR of 1e-10 at @weights_biases