{"next_token": false, "non_ioi_thresh": 0.65, "use_per_token_check": false, "batch_size": 512, "num_workers": 0, "early_stop": true, "lr_scheduler": null, "scheduler_val_metric": ["val/accuracy", "val/IIA"], "scheduler_mode": "max", "scheduler_kwargs": {}, "clip_grad_norm": 1.0, "seed": 0, "lr": 0.001, "detach_while_caching": true, "optimizer_kwargs": {"betas": [0.9, 0.9]}, "atol": 0.05, "use_single_loss": false, "iit_weight": 1.0, "behavior_weight": 1.0, "val_IIA_sampling": "random", "use_all_tokens_for_behavior": false, "strict_weight": 0.4, "siit_sampling": "individual"}