about train the network in notebook two-layer-network

[zibo-zibo]

I'm work on cs231n's 2021 spring assignment, and I got stuck in the two-layer-network. The loss of the train process are all inf, not some decreasing scalars.
The new assignment is more modular, the computation of loss and derivative are in different files, but the code is almost same of yours. I can't find where is the error which result in the inf loss. Have you ever encountered this situation.
It like this
input_size = 32 * 32 * 3 hidden_size = 50 num_classes = 10 model = TwoLayerNet(input_size, hidden_size, num_classes) solver = None solver = Solver(model, data, update_rule='sgd') solver.train()