Check if models have same weights

6

I found this to be a better approach (also compares batch_norm layers for their running_mean and running_var params) -

def compare_models(model_1, model_2):
    models_differ = 0
    for key_item_1, key_item_2 in zip(model_1.state_dict().items(), model_2.state_dict().items()):
        if torch.equal(key_item_1[1], key_item_2[1]):
            pass
        else:
            models_differ += 1
            if (key_item_1[0] == key_item_2[0]):
                print('Mismtach found at', key_item_1[0])
            else:
                raise Exception
    if models_differ == 0:
        print('Models match perfectly! :)')
11 Likes