{ "schema_version": 1, "status": "complete", "config_fingerprint": "0667d4ea7a9eeb08101909c8f4b52860faadd491ce6d4ceb8094405e04a5f75e", "dataset_fingerprint": "6cf9eb3d777a338eb007dcc5606b2aac6605bf899a191ea399ccd77d5527acfb", "cursor": { "global_step": 3510, "micro_step": 28165, "dataset_epoch": 1, "sample_offset": 24960, "accumulation_step": 0 }, "extra": { "last_metrics": { "global_step": 3510, "behavior_loss": 0.014460635284194723, "behavior_microbatches": 8, "behavior_singleton_microbatches": 0, "oom_replayed_as_singletons": false, "gradient_norm": 0.022714726626873016, "learning_rate_used": 1e-06, "step_seconds": 12.606, "peak_reserved_gib": 24.904, "optimizer_state_device": "cpu_between_updates", "weighted_behavior_objective": 0.014460635284194723, "validation_behavior_loss": 0.01682164315134287, "lr_window_phase": "minimum_lr_plateau", "lr_window_start_step": 3440, "lr_window_end_step": 3510, "lr_window_point_count": 8, "lr_window_log_slope_per_step": 3.468046811199166e-05, "lr_window_fitted_log_descent": -0.002427632767839416, "lr_window_relative_descent": -0.002430581854217184, "lr_window_residual_mad_scale": 0.0030586733246124, "lr_window_descent_to_noise": -0.7936881484874001, "lr_window_accepted": false, "lr_window_elbow_step": 3450, "lr_training_window_start_step": 3450, "lr_training_window_end_step": 3510, "lr_training_window_point_count": 7, "lr_training_window_log_slope_per_step": 0.002190928189992114, "lr_training_window_fitted_log_descent": -0.13145569139952684, "lr_training_window_relative_descent": -0.1404873732034586, "lr_training_window_residual_mad_scale": 0.012542356750465129, "lr_training_window_descent_to_noise": -10.48094022637746, "lr_training_window_accepted": false, "lr_training_window_elbow_step": 3460, "lr_window_any_signal_accepted": false, "validation_interval": 10, "next_validation_step": 3520, "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": 3510, "loss": 0.01682164315134287 } ], "training_points": [], "training_loss_sum": 0.0, "training_loss_count": 0, "last_training_step": 3510, "window_start_step": 3510, "next_validation_step": 3520, "last_analysis": { "start_step": 3440, "end_step": 3510, "point_count": 8, "log_slope_per_step": 3.468046811199166e-05, "fitted_log_descent": -0.002427632767839416, "relative_descent": -0.002430581854217184, "residual_mad_scale": 0.0030586733246124, "descent_to_noise": -0.7936881484874001, "accepted": false, "elbow_step": 3450 }, "last_training_analysis": { "start_step": 3450, "end_step": 3510, "point_count": 7, "log_slope_per_step": 0.002190928189992114, "fitted_log_descent": -0.13145569139952684, "relative_descent": -0.1404873732034586, "residual_mad_scale": 0.012542356750465129, "descent_to_noise": -10.48094022637746, "accepted": false, "elbow_step": 3460 }, "confirming": false, "pending_rollback": null, "reductions": 9 } }, "rollback_replay": null }, "files": { "adapters.safetensors": "3f78038fb2a496d336c5437a1129ce105ca66448e2de752e8f1abe8650c74700", "training_state.pt": "62aa73c036a6c23d5b51eba117eeaf32582b72c87c27309a2e0a424a52730921" } }