{ "schema_version": 1, "status": "complete", "config_fingerprint": "0667d4ea7a9eeb08101909c8f4b52860faadd491ce6d4ceb8094405e04a5f75e", "dataset_fingerprint": "6cf9eb3d777a338eb007dcc5606b2aac6605bf899a191ea399ccd77d5527acfb", "cursor": { "global_step": 4510, "micro_step": 36198, "dataset_epoch": 2, "sample_offset": 9760, "accumulation_step": 0 }, "extra": { "last_metrics": { "global_step": 4510, "behavior_loss": 0.012179180164821446, "behavior_microbatches": 8, "behavior_singleton_microbatches": 0, "oom_replayed_as_singletons": false, "gradient_norm": 0.01002525258809328, "learning_rate_used": 1e-06, "step_seconds": 12.761, "peak_reserved_gib": 24.904, "optimizer_state_device": "cpu_between_updates", "weighted_behavior_objective": 0.012179180164821446, "validation_behavior_loss": 0.016795611083507536, "lr_window_phase": "continue", "lr_window_start_step": 4440, "lr_window_end_step": 4510, "lr_window_point_count": 8, "lr_window_log_slope_per_step": 3.681781263404904e-05, "lr_window_fitted_log_descent": -0.002577246884383433, "lr_window_relative_descent": -0.0025805708400730154, "lr_window_residual_mad_scale": 0.0009766390549257966, "lr_window_descent_to_noise": -2.638893940790897, "lr_window_accepted": false, "lr_window_elbow_step": 4480, "lr_training_window_start_step": 4450, "lr_training_window_end_step": 4510, "lr_training_window_point_count": 7, "lr_training_window_log_slope_per_step": -0.0011663531877474339, "lr_training_window_fitted_log_descent": 0.06998119126484603, "lr_training_window_relative_descent": 0.06758864278070709, "lr_training_window_residual_mad_scale": 0.06787014756190643, "lr_training_window_descent_to_noise": 1.031104156669382, "lr_training_window_accepted": true, "lr_training_window_elbow_step": 4510, "lr_window_any_signal_accepted": true, "validation_interval": 10, "next_validation_step": 4520, "learning_rate": 1e-06 }, "reason": "periodic", "parent": null, "lr_control": { "version": 3, "config": { "factor": 0.5, "minimum_learning_rate": 1e-06, "window_steps": 50, "probe_every_steps": 10, "confirmation_steps": 20, "minimum_descent_to_noise": 1.0, "minimum_relative_descent": 0.0 }, "state": { "learning_rate": 1e-06, "points": [ { "step": 4510, "loss": 0.016795611083507536 } ], "training_points": [], "training_loss_sum": 0.0, "training_loss_count": 0, "last_training_step": 4510, "window_start_step": 4510, "next_validation_step": 4520, "last_analysis": { "start_step": 4440, "end_step": 4510, "point_count": 8, "log_slope_per_step": 3.681781263404904e-05, "fitted_log_descent": -0.002577246884383433, "relative_descent": -0.0025805708400730154, "residual_mad_scale": 0.0009766390549257966, "descent_to_noise": -2.638893940790897, "accepted": false, "elbow_step": 4480 }, "last_training_analysis": { "start_step": 4450, "end_step": 4510, "point_count": 7, "log_slope_per_step": -0.0011663531877474339, "fitted_log_descent": 0.06998119126484603, "relative_descent": 0.06758864278070709, "residual_mad_scale": 0.06787014756190643, "descent_to_noise": 1.031104156669382, "accepted": true, "elbow_step": 4510 }, "confirming": false, "pending_rollback": null, "reductions": 9 } }, "rollback_replay": null }, "files": { "adapters.safetensors": "1f2277f2e045b144719b952e572b4f73b1df0ceb70df39881499cb4f8001d328", "training_state.pt": "6862b5f3a9a6be15900b5d6b910825df3ba4cd325ba2c53ee5f7a2121d3eaeb5" } }