{ "device": "cuda", "mechanism_checks": { "stability": [ { "lambda": -26, "q": -1.08, "predicted": "diverge", "observed": "diverge", "abs_s30": 10.06265688907345 }, { "lambda": -25, "q": -1.0, "predicted": "neutral", "observed": "neutral", "abs_s30": 1.0 }, { "lambda": -20, "q": -0.6000000000000001, "predicted": "contract", "observed": "contract", "abs_s30": 2.2107391972073433e-07 }, { "lambda": -13, "q": -0.040000000000000036, "predicted": "contract", "observed": "contract", "abs_s30": 1.1529215046068777e-42 }, { "lambda": -12.5, "q": 0.0, "predicted": "contract", "observed": "contract", "abs_s30": 0.0 }, { "lambda": -10, "q": 0.19999999999999996, "predicted": "contract", "observed": "contract", "abs_s30": 1.0737418239999928e-21 }, { "lambda": 0, "q": 1.0, "predicted": "neutral", "observed": "neutral", "abs_s30": 1.0 }, { "lambda": 5, "q": 1.4, "predicted": "diverge", "observed": "diverge", "abs_s30": 24201.43235548455 } ], "strict_stability_interval": [ -25.0, 0.0 ], "halting_constant_h": [ { "h": 0.05, "formula_T": 20, "observed_formula_T": 20, "act_clipped_T": 20 }, { "h": 0.1, "formula_T": 10, "observed_formula_T": 10, "act_clipped_T": 10 }, { "h": 0.2, "formula_T": 5, "observed_formula_T": 5, "act_clipped_T": 5 }, { "h": 0.4, "formula_T": 3, "observed_formula_T": 3, "act_clipped_T": 3 }, { "h": 0.7, "formula_T": 2, "observed_formula_T": 2, "act_clipped_T": 2 } ], "halting_bias_sweep": [ { "b": -3, "h": 0.04742587317756678, "formula_T": 22, "observed_T": 22 }, { "b": -1, "h": 0.2689414213699951, "formula_T": 4, "observed_T": 4 }, { "b": 0, "h": 0.5, "formula_T": 2, "observed_T": 2 }, { "b": 1, "h": 0.7310585786300049, "formula_T": 2, "observed_T": 2 }, { "b": 3, "h": 0.9525741268224334, "formula_T": 2, "observed_T": 2 } ] }, "mini_experiment": { "baseline": { "accuracy": 0.990234375, "avg_microsteps": 6.0, "train_seconds": 5.6903587950073415, "parameters": 13218 }, "idea": { "accuracy": 0.990234375, "avg_microsteps": 4.107421875, "train_seconds": 14.9282974020025, "parameters": 3411 } } }