--n_epochs = 50 --patience = -1 --mbatch_size = 64 --lr = 0.01 --lr_scheduler = "MultiStepLR" --lr_scheduler_kwargs = {"milestones": [20, 30, 40, 50], "gamma": 0.1} --param_lr_map = {} --grad_clip = "norm" --grad_clip_kwargs = {"max_norm": 5} --regularization = 0.0 --l1_reg = 0.0 --l2_reg = 0.0 --optimizer = "Adam" --loss = "MSELoss" --initializer = null --initialization_seed = 0 --batch_shuffle_seed = 0