{"atol": 0.05, "lr": 0.001, "use_single_loss": false, "iit_weight": 1.0, "behavior_weight": 1.0, "strict_weight": 0.4, "epochs": 2000, "act_fn": "gelu", "wandb_suffix": "strict_38", "device": "cpu", "clip_grad_norm": 0.1, "lr_scheduler": ""} |
{"atol": 0.05, "lr": 0.001, "use_single_loss": false, "iit_weight": 1.0, "behavior_weight": 1.0, "strict_weight": 0.4, "epochs": 2000, "act_fn": "gelu", "wandb_suffix": "strict_38", "device": "cpu", "clip_grad_norm": 0.1, "lr_scheduler": ""} |