{"atol": 0.05, "lr": 0.01, "use_single_loss": false, "iit_weight": 1.0, "behavior_weight": 1.0, "strict_weight": 0.5, "epochs": 2000, "act_fn": "gelu", "wandb_suffix": "", "device": "cuda", "clip_grad_norm": 0.1, "lr_scheduler": "", "model_pair": "strict", "same_size": false, "seed": 1234, "batch_size": 256, "include_mlp": false, "next_token": false, "detach_while_caching": true}