| { |
| "best_global_step": null, |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 0.75, |
| "eval_steps": 500, |
| "global_step": 75, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 659.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 821.0, |
| "completions/max_terminated_length": 821.0, |
| "completions/mean_length": 659.0, |
| "completions/mean_terminated_length": 659.0, |
| "completions/min_length": 516.0, |
| "completions/min_terminated_length": 516.0, |
| "epoch": 0.01, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.13492469489574432, |
| "kl": 6.197717993927654e-06, |
| "learning_rate": 0.0, |
| "loss": 2.8312206268310547e-07, |
| "num_tokens": 5736.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 1 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 543.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 693.0, |
| "completions/max_terminated_length": 693.0, |
| "completions/mean_length": 543.75, |
| "completions/mean_terminated_length": 543.75, |
| "completions/min_length": 426.0, |
| "completions/min_terminated_length": 426.0, |
| "epoch": 0.02, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1295754313468933, |
| "kl": 7.463787596861948e-06, |
| "learning_rate": 5.000000000000001e-07, |
| "loss": 3.725290298461914e-07, |
| "num_tokens": 11019.0, |
| "reward": 0.15250001847743988, |
| "reward_std": 0.7683477997779846, |
| "rewards/reward_environment_execution/mean": -0.09749999642372131, |
| "rewards/reward_environment_execution/std": 0.2683747410774231, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 2 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 572.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 677.0, |
| "completions/max_terminated_length": 677.0, |
| "completions/mean_length": 572.75, |
| "completions/mean_terminated_length": 572.75, |
| "completions/min_length": 490.0, |
| "completions/min_terminated_length": 490.0, |
| "epoch": 0.03, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1402619630098343, |
| "kl": 5.991519174131099e-06, |
| "learning_rate": 1.0000000000000002e-06, |
| "loss": 2.2351741790771484e-07, |
| "num_tokens": 16418.0, |
| "reward": 0.5375000238418579, |
| "reward_std": 0.004999995231628418, |
| "rewards/reward_environment_execution/mean": 0.03750000149011612, |
| "rewards/reward_environment_execution/std": 0.004999999888241291, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.30000001192092896, |
| "rewards/reward_investigation_quality/std": 0.0, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 3 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 705.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 783.0, |
| "completions/max_terminated_length": 783.0, |
| "completions/mean_length": 705.0, |
| "completions/mean_terminated_length": 705.0, |
| "completions/min_length": 634.0, |
| "completions/min_terminated_length": 634.0, |
| "epoch": 0.04, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.13879841566085815, |
| "kl": 6.360480483635911e-06, |
| "learning_rate": 1.5e-06, |
| "loss": 2.980232238769531e-07, |
| "num_tokens": 22514.0, |
| "reward": 0.24392502009868622, |
| "reward_std": 0.8475051522254944, |
| "rewards/reward_environment_execution/mean": 0.04392500966787338, |
| "rewards/reward_environment_execution/std": 0.4515362083911896, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.17500001192092896, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 4 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 677.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 785.0, |
| "completions/max_terminated_length": 785.0, |
| "completions/mean_length": 677.0, |
| "completions/mean_terminated_length": 677.0, |
| "completions/min_length": 623.0, |
| "completions/min_terminated_length": 623.0, |
| "epoch": 0.05, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10656019300222397, |
| "kl": 4.833805178350303e-06, |
| "learning_rate": 2.0000000000000003e-06, |
| "loss": 2.2351741790771484e-07, |
| "num_tokens": 28326.0, |
| "reward": 0.15250001847743988, |
| "reward_std": 0.7683477997779846, |
| "rewards/reward_environment_execution/mean": -0.09750000387430191, |
| "rewards/reward_environment_execution/std": 0.2683747410774231, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 5 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 701.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 913.0, |
| "completions/max_terminated_length": 913.0, |
| "completions/mean_length": 701.0, |
| "completions/mean_terminated_length": 701.0, |
| "completions/min_length": 553.0, |
| "completions/min_terminated_length": 553.0, |
| "epoch": 0.06, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 6.133589340606704e-05, |
| "kl": 6.346002919599414e-06, |
| "learning_rate": 2.5e-06, |
| "loss": 2.5384014179508085e-07, |
| "num_tokens": 34230.0, |
| "reward": -1.0, |
| "reward_std": 0.0, |
| "rewards/reward_environment_execution/mean": -0.5, |
| "rewards/reward_environment_execution/std": 0.0, |
| "rewards/reward_format_compliance/mean": -0.5, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.0, |
| "rewards/reward_investigation_quality/std": 0.0, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 6 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 524.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 640.0, |
| "completions/max_terminated_length": 640.0, |
| "completions/mean_length": 524.0, |
| "completions/mean_terminated_length": 524.0, |
| "completions/min_length": 450.0, |
| "completions/min_terminated_length": 450.0, |
| "epoch": 0.07, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.13970911502838135, |
| "kl": 6.830848860772676e-06, |
| "learning_rate": 3e-06, |
| "loss": 2.8312206268310547e-07, |
| "num_tokens": 39426.0, |
| "reward": -0.22999998927116394, |
| "reward_std": 0.8891193866729736, |
| "rewards/reward_environment_execution/mean": -0.23000000417232513, |
| "rewards/reward_environment_execution/std": 0.31176915764808655, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 7 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 626.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 725.0, |
| "completions/max_terminated_length": 725.0, |
| "completions/mean_length": 626.0, |
| "completions/mean_terminated_length": 626.0, |
| "completions/min_length": 537.0, |
| "completions/min_terminated_length": 537.0, |
| "epoch": 0.08, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12764990329742432, |
| "kl": 5.7269965054729255e-06, |
| "learning_rate": 3.5e-06, |
| "loss": 2.1606683731079102e-07, |
| "num_tokens": 45026.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 8 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 669.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 752.0, |
| "completions/max_terminated_length": 752.0, |
| "completions/mean_length": 669.5, |
| "completions/mean_terminated_length": 669.5, |
| "completions/min_length": 555.0, |
| "completions/min_terminated_length": 555.0, |
| "epoch": 0.09, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1321861445903778, |
| "kl": 5.624260097647493e-06, |
| "learning_rate": 4.000000000000001e-06, |
| "loss": 2.086162567138672e-07, |
| "num_tokens": 50812.0, |
| "reward": -0.22999998927116394, |
| "reward_std": 0.8891193866729736, |
| "rewards/reward_environment_execution/mean": -0.23000000417232513, |
| "rewards/reward_environment_execution/std": 0.31176915764808655, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 9 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 640.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 790.0, |
| "completions/max_terminated_length": 790.0, |
| "completions/mean_length": 640.25, |
| "completions/mean_terminated_length": 640.25, |
| "completions/min_length": 471.0, |
| "completions/min_terminated_length": 471.0, |
| "epoch": 0.1, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12317467480897903, |
| "kl": 6.860862526991696e-06, |
| "learning_rate": 4.5e-06, |
| "loss": 3.0547380447387695e-07, |
| "num_tokens": 56485.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 10 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 664.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 855.0, |
| "completions/max_terminated_length": 855.0, |
| "completions/mean_length": 664.75, |
| "completions/mean_terminated_length": 664.75, |
| "completions/min_length": 559.0, |
| "completions/min_terminated_length": 559.0, |
| "epoch": 0.11, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1155354306101799, |
| "kl": 6.481316631834488e-06, |
| "learning_rate": 5e-06, |
| "loss": 2.682209014892578e-07, |
| "num_tokens": 62252.0, |
| "reward": -0.6150000095367432, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.36500000953674316, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": -0.32499998807907104, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.07500000298023224, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 11 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 699.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 880.0, |
| "completions/max_terminated_length": 880.0, |
| "completions/mean_length": 699.5, |
| "completions/mean_terminated_length": 699.5, |
| "completions/min_length": 539.0, |
| "completions/min_terminated_length": 539.0, |
| "epoch": 0.12, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11765824258327484, |
| "kl": 5.136041295372706e-06, |
| "learning_rate": 4.99847706754774e-06, |
| "loss": 1.862645149230957e-07, |
| "num_tokens": 68150.0, |
| "reward": -0.6150000095367432, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.36500000953674316, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": -0.32499998807907104, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.07500000298023224, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 12 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 753.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 913.0, |
| "completions/max_terminated_length": 913.0, |
| "completions/mean_length": 753.5, |
| "completions/mean_terminated_length": 753.5, |
| "completions/min_length": 543.0, |
| "completions/min_terminated_length": 543.0, |
| "epoch": 0.13, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1163109764456749, |
| "kl": 6.4769044456625124e-06, |
| "learning_rate": 4.993910125649561e-06, |
| "loss": 2.682209014892578e-07, |
| "num_tokens": 74436.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 13 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 691.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 870.0, |
| "completions/max_terminated_length": 870.0, |
| "completions/mean_length": 691.25, |
| "completions/mean_terminated_length": 691.25, |
| "completions/min_length": 596.0, |
| "completions/min_terminated_length": 596.0, |
| "epoch": 0.14, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 5.699428584193811e-05, |
| "kl": 4.6748977524657676e-06, |
| "learning_rate": 4.986304738420684e-06, |
| "loss": 1.869958907718683e-07, |
| "num_tokens": 80465.0, |
| "reward": 0.5400000214576721, |
| "reward_std": 0.0, |
| "rewards/reward_environment_execution/mean": 0.03999999910593033, |
| "rewards/reward_environment_execution/std": 0.0, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.30000001192092896, |
| "rewards/reward_investigation_quality/std": 0.0, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 14 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 588.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 832.0, |
| "completions/max_terminated_length": 832.0, |
| "completions/mean_length": 588.0, |
| "completions/mean_terminated_length": 588.0, |
| "completions/min_length": 432.0, |
| "completions/min_terminated_length": 432.0, |
| "epoch": 0.15, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.00010395531717222184, |
| "kl": 9.099214707930514e-06, |
| "learning_rate": 4.975670171853926e-06, |
| "loss": 3.639685814960103e-07, |
| "num_tokens": 86089.0, |
| "reward": -1.0, |
| "reward_std": 0.0, |
| "rewards/reward_environment_execution/mean": -0.5, |
| "rewards/reward_environment_execution/std": 0.0, |
| "rewards/reward_format_compliance/mean": -0.5, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.0, |
| "rewards/reward_investigation_quality/std": 0.0, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 15 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 607.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 673.0, |
| "completions/max_terminated_length": 673.0, |
| "completions/mean_length": 607.75, |
| "completions/mean_terminated_length": 607.75, |
| "completions/min_length": 517.0, |
| "completions/min_terminated_length": 517.0, |
| "epoch": 0.16, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1489061564207077, |
| "kl": 9.158140301224194e-06, |
| "learning_rate": 4.962019382530521e-06, |
| "loss": 3.2782554626464844e-07, |
| "num_tokens": 91620.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 16 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 657.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 737.0, |
| "completions/max_terminated_length": 737.0, |
| "completions/mean_length": 657.25, |
| "completions/mean_terminated_length": 657.25, |
| "completions/min_length": 528.0, |
| "completions/min_terminated_length": 528.0, |
| "epoch": 0.17, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.14814995229244232, |
| "kl": 1.14353706521797e-05, |
| "learning_rate": 4.9453690018345144e-06, |
| "loss": 4.470348358154297e-07, |
| "num_tokens": 97513.0, |
| "reward": -0.22999998927116394, |
| "reward_std": 0.8891193866729736, |
| "rewards/reward_environment_execution/mean": -0.23000000417232513, |
| "rewards/reward_environment_execution/std": 0.31176915764808655, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 17 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 549.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 621.0, |
| "completions/max_terminated_length": 621.0, |
| "completions/mean_length": 549.25, |
| "completions/mean_terminated_length": 549.25, |
| "completions/min_length": 510.0, |
| "completions/min_terminated_length": 510.0, |
| "epoch": 0.18, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.13028942048549652, |
| "kl": 1.1230136237827537e-05, |
| "learning_rate": 4.925739315689991e-06, |
| "loss": 4.6193599700927734e-07, |
| "num_tokens": 102810.0, |
| "reward": -0.22999998927116394, |
| "reward_std": 0.8891193866729736, |
| "rewards/reward_environment_execution/mean": -0.23000000417232513, |
| "rewards/reward_environment_execution/std": 0.31176915764808655, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 18 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 565.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 686.0, |
| "completions/max_terminated_length": 686.0, |
| "completions/mean_length": 565.5, |
| "completions/mean_terminated_length": 565.5, |
| "completions/min_length": 288.0, |
| "completions/min_terminated_length": 288.0, |
| "epoch": 0.19, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.2010871171951294, |
| "kl": 1.3259304523671744e-05, |
| "learning_rate": 4.903154239845798e-06, |
| "loss": 5.438923835754395e-07, |
| "num_tokens": 108344.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 19 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 692.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 777.0, |
| "completions/max_terminated_length": 777.0, |
| "completions/mean_length": 692.25, |
| "completions/mean_terminated_length": 692.25, |
| "completions/min_length": 633.0, |
| "completions/min_terminated_length": 633.0, |
| "epoch": 0.2, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12407222390174866, |
| "kl": 1.1413530501158675e-05, |
| "learning_rate": 4.8776412907378845e-06, |
| "loss": 4.6938657760620117e-07, |
| "num_tokens": 114393.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 20 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 633.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 702.0, |
| "completions/max_terminated_length": 702.0, |
| "completions/mean_length": 633.0, |
| "completions/mean_terminated_length": 633.0, |
| "completions/min_length": 568.0, |
| "completions/min_terminated_length": 568.0, |
| "epoch": 0.21, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.14354389905929565, |
| "kl": 1.5171975746852695e-05, |
| "learning_rate": 4.849231551964771e-06, |
| "loss": 6.407499313354492e-07, |
| "num_tokens": 120177.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 21 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 669.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 978.0, |
| "completions/max_terminated_length": 978.0, |
| "completions/mean_length": 669.0, |
| "completions/mean_terminated_length": 669.0, |
| "completions/min_length": 444.0, |
| "completions/min_terminated_length": 444.0, |
| "epoch": 0.22, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12197419255971909, |
| "kl": 1.5716058214820805e-05, |
| "learning_rate": 4.817959636416969e-06, |
| "loss": 6.183981895446777e-07, |
| "num_tokens": 126125.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 22 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 774.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 1024.0, |
| "completions/mean_length": 774.25, |
| "completions/mean_terminated_length": 774.25, |
| "completions/min_length": 484.0, |
| "completions/min_terminated_length": 484.0, |
| "epoch": 0.23, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1378677934408188, |
| "kl": 2.0351369357740623e-05, |
| "learning_rate": 4.783863644106502e-06, |
| "loss": 8.493661880493164e-07, |
| "num_tokens": 132326.0, |
| "reward": -0.21249999105930328, |
| "reward_std": 0.9097756743431091, |
| "rewards/reward_environment_execution/mean": -0.23750001192092896, |
| "rewards/reward_environment_execution/std": 0.30335623025894165, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.125, |
| "rewards/reward_investigation_quality/std": 0.15000000596046448, |
| "rewards/reward_os_mechanics/mean": 0.05000000074505806, |
| "rewards/reward_os_mechanics/std": 0.10000000149011612, |
| "step": 23 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 565.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 732.0, |
| "completions/max_terminated_length": 732.0, |
| "completions/mean_length": 565.75, |
| "completions/mean_terminated_length": 565.75, |
| "completions/min_length": 314.0, |
| "completions/min_terminated_length": 314.0, |
| "epoch": 0.24, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1561768352985382, |
| "kl": 2.7805408535641618e-05, |
| "learning_rate": 4.746985115747918e-06, |
| "loss": 1.1026859283447266e-06, |
| "num_tokens": 137685.0, |
| "reward": -0.6150000095367432, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.36500000953674316, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": -0.32499998807907104, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.07500000298023224, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 24 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 668.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 909.0, |
| "completions/max_terminated_length": 909.0, |
| "completions/mean_length": 668.75, |
| "completions/mean_terminated_length": 668.75, |
| "completions/min_length": 528.0, |
| "completions/min_terminated_length": 528.0, |
| "epoch": 0.25, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.13285094499588013, |
| "kl": 2.5312373509223107e-05, |
| "learning_rate": 4.707368982147318e-06, |
| "loss": 1.0132789611816406e-06, |
| "num_tokens": 143456.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 25 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 487.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 559.0, |
| "completions/max_terminated_length": 559.0, |
| "completions/mean_length": 487.25, |
| "completions/mean_terminated_length": 487.25, |
| "completions/min_length": 420.0, |
| "completions/min_terminated_length": 420.0, |
| "epoch": 0.26, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.22987650334835052, |
| "kl": 2.307988370375824e-05, |
| "learning_rate": 4.665063509461098e-06, |
| "loss": 9.5367431640625e-07, |
| "num_tokens": 148505.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 26 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 667.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 782.0, |
| "completions/max_terminated_length": 782.0, |
| "completions/mean_length": 667.25, |
| "completions/mean_terminated_length": 667.25, |
| "completions/min_length": 533.0, |
| "completions/min_terminated_length": 533.0, |
| "epoch": 0.27, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12192773073911667, |
| "kl": 2.0764900455105817e-05, |
| "learning_rate": 4.620120240391065e-06, |
| "loss": 8.642673492431641e-07, |
| "num_tokens": 154454.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 27 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 722.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 896.0, |
| "completions/max_terminated_length": 896.0, |
| "completions/mean_length": 722.0, |
| "completions/mean_terminated_length": 722.0, |
| "completions/min_length": 637.0, |
| "completions/min_terminated_length": 637.0, |
| "epoch": 0.28, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11616191267967224, |
| "kl": 2.81208617707307e-05, |
| "learning_rate": 4.572593931387604e-06, |
| "loss": 1.1026859283447266e-06, |
| "num_tokens": 160446.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 28 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 686.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 743.0, |
| "completions/max_terminated_length": 743.0, |
| "completions/mean_length": 686.0, |
| "completions/mean_terminated_length": 686.0, |
| "completions/min_length": 620.0, |
| "completions/min_terminated_length": 620.0, |
| "epoch": 0.29, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.13982541859149933, |
| "kl": 2.6462233336133067e-05, |
| "learning_rate": 4.522542485937369e-06, |
| "loss": 1.043081283569336e-06, |
| "num_tokens": 166286.0, |
| "reward": -0.22999998927116394, |
| "reward_std": 0.8891193866729736, |
| "rewards/reward_environment_execution/mean": -0.23000000417232513, |
| "rewards/reward_environment_execution/std": 0.31176915764808655, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 29 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 673.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 847.0, |
| "completions/max_terminated_length": 847.0, |
| "completions/mean_length": 673.75, |
| "completions/mean_terminated_length": 673.75, |
| "completions/min_length": 536.0, |
| "completions/min_terminated_length": 536.0, |
| "epoch": 0.3, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.17358987033367157, |
| "kl": 4.3880153043573955e-05, |
| "learning_rate": 4.470026884016805e-06, |
| "loss": 1.9818544387817383e-06, |
| "num_tokens": 172089.0, |
| "reward": 0.6325000524520874, |
| "reward_std": 0.14221462607383728, |
| "rewards/reward_environment_execution/mean": 0.057500001043081284, |
| "rewards/reward_environment_execution/std": 0.056789085268974304, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.25, |
| "rewards/reward_investigation_quality/std": 0.057735033333301544, |
| "rewards/reward_os_mechanics/mean": 0.125, |
| "rewards/reward_os_mechanics/std": 0.15000000596046448, |
| "step": 30 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 648.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 858.0, |
| "completions/max_terminated_length": 858.0, |
| "completions/mean_length": 648.0, |
| "completions/mean_terminated_length": 648.0, |
| "completions/min_length": 461.0, |
| "completions/min_terminated_length": 461.0, |
| "epoch": 0.31, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1460961103439331, |
| "kl": 3.440224281803239e-05, |
| "learning_rate": 4.415111107797445e-06, |
| "loss": 1.3560056686401367e-06, |
| "num_tokens": 177773.0, |
| "reward": -0.2800000011920929, |
| "reward_std": 0.8353841304779053, |
| "rewards/reward_environment_execution/mean": -0.23000000417232513, |
| "rewards/reward_environment_execution/std": 0.31176915764808655, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.125, |
| "rewards/reward_investigation_quality/std": 0.15000000596046448, |
| "rewards/reward_os_mechanics/mean": -0.02500000037252903, |
| "rewards/reward_os_mechanics/std": 0.05000000074505806, |
| "step": 31 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 832.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 1227.0, |
| "completions/max_terminated_length": 1227.0, |
| "completions/mean_length": 832.5, |
| "completions/mean_terminated_length": 832.5, |
| "completions/min_length": 628.0, |
| "completions/min_terminated_length": 628.0, |
| "epoch": 0.32, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.0001798434095690027, |
| "kl": 3.118666154477978e-05, |
| "learning_rate": 4.357862063693486e-06, |
| "loss": 1.2474664572437177e-06, |
| "num_tokens": 184203.0, |
| "reward": 0.5400000214576721, |
| "reward_std": 0.0, |
| "rewards/reward_environment_execution/mean": 0.03999999910593033, |
| "rewards/reward_environment_execution/std": 0.0, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.30000001192092896, |
| "rewards/reward_investigation_quality/std": 0.0, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 32 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 519.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 580.0, |
| "completions/max_terminated_length": 580.0, |
| "completions/mean_length": 519.0, |
| "completions/mean_terminated_length": 519.0, |
| "completions/min_length": 435.0, |
| "completions/min_terminated_length": 435.0, |
| "epoch": 0.33, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.0003187077818438411, |
| "kl": 5.177023012947757e-05, |
| "learning_rate": 4.2983495008466285e-06, |
| "loss": 2.0708091597043676e-06, |
| "num_tokens": 189387.0, |
| "reward": 0.5400000214576721, |
| "reward_std": 0.0, |
| "rewards/reward_environment_execution/mean": 0.03999999910593033, |
| "rewards/reward_environment_execution/std": 0.0, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.30000001192092896, |
| "rewards/reward_investigation_quality/std": 0.0, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 33 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 670.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 767.0, |
| "completions/max_terminated_length": 767.0, |
| "completions/mean_length": 670.5, |
| "completions/mean_terminated_length": 670.5, |
| "completions/min_length": 585.0, |
| "completions/min_terminated_length": 585.0, |
| "epoch": 0.34, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12342657893896103, |
| "kl": 3.6297568385634804e-05, |
| "learning_rate": 4.236645926147493e-06, |
| "loss": 1.4156103134155273e-06, |
| "num_tokens": 195333.0, |
| "reward": -0.22999998927116394, |
| "reward_std": 0.8891193866729736, |
| "rewards/reward_environment_execution/mean": -0.23000000417232513, |
| "rewards/reward_environment_execution/std": 0.31176915764808655, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 34 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 464.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 645.0, |
| "completions/max_terminated_length": 645.0, |
| "completions/mean_length": 464.5, |
| "completions/mean_terminated_length": 464.5, |
| "completions/min_length": 195.0, |
| "completions/min_terminated_length": 195.0, |
| "epoch": 0.35, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.14799511432647705, |
| "kl": 9.590507397660986e-05, |
| "learning_rate": 4.172826515897146e-06, |
| "loss": 3.822147846221924e-06, |
| "num_tokens": 200295.0, |
| "reward": -0.6150000095367432, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.36500000953674316, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": -0.32499998807907104, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.07500000298023224, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 35 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 539.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 667.0, |
| "completions/max_terminated_length": 667.0, |
| "completions/mean_length": 539.75, |
| "completions/mean_terminated_length": 539.75, |
| "completions/min_length": 403.0, |
| "completions/min_terminated_length": 403.0, |
| "epoch": 0.36, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.15325292944908142, |
| "kl": 9.70982619037386e-05, |
| "learning_rate": 4.106969024216348e-06, |
| "loss": 3.919005393981934e-06, |
| "num_tokens": 205554.0, |
| "reward": -0.23249998688697815, |
| "reward_std": 0.8862420320510864, |
| "rewards/reward_environment_execution/mean": -0.23250000178813934, |
| "rewards/reward_environment_execution/std": 0.3089093565940857, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 36 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 544.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 787.0, |
| "completions/max_terminated_length": 787.0, |
| "completions/mean_length": 544.5, |
| "completions/mean_terminated_length": 544.5, |
| "completions/min_length": 333.0, |
| "completions/min_terminated_length": 333.0, |
| "epoch": 0.37, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.15045017004013062, |
| "kl": 7.959679260238772e-05, |
| "learning_rate": 4.039153688314146e-06, |
| "loss": 3.159046173095703e-06, |
| "num_tokens": 210828.0, |
| "reward": -0.6150000095367432, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.36500000953674316, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": -0.32499998807907104, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.07500000298023224, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 37 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 638.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 695.0, |
| "completions/max_terminated_length": 695.0, |
| "completions/mean_length": 638.75, |
| "completions/mean_terminated_length": 638.75, |
| "completions/min_length": 568.0, |
| "completions/min_terminated_length": 568.0, |
| "epoch": 0.38, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.000358834135113284, |
| "kl": 6.0015446706529474e-05, |
| "learning_rate": 3.969463130731183e-06, |
| "loss": 2.4006176317925565e-06, |
| "num_tokens": 216647.0, |
| "reward": 0.5400000214576721, |
| "reward_std": 0.0, |
| "rewards/reward_environment_execution/mean": 0.03999999910593033, |
| "rewards/reward_environment_execution/std": 0.0, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.30000001192092896, |
| "rewards/reward_investigation_quality/std": 0.0, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 38 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 555.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 891.0, |
| "completions/max_terminated_length": 891.0, |
| "completions/mean_length": 555.75, |
| "completions/mean_terminated_length": 555.75, |
| "completions/min_length": 389.0, |
| "completions/min_terminated_length": 389.0, |
| "epoch": 0.39, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.20931142568588257, |
| "kl": 6.502814630948706e-05, |
| "learning_rate": 3.897982258676867e-06, |
| "loss": 2.6226043701171875e-06, |
| "num_tokens": 221986.0, |
| "reward": 0.5375000238418579, |
| "reward_std": 0.004999995231628418, |
| "rewards/reward_environment_execution/mean": 0.03750000149011612, |
| "rewards/reward_environment_execution/std": 0.004999999888241291, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.30000001192092896, |
| "rewards/reward_investigation_quality/std": 0.0, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 39 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 638.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 786.0, |
| "completions/max_terminated_length": 786.0, |
| "completions/mean_length": 638.5, |
| "completions/mean_terminated_length": 638.5, |
| "completions/min_length": 469.0, |
| "completions/min_terminated_length": 469.0, |
| "epoch": 0.4, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1422736942768097, |
| "kl": 0.0001064265761669958, |
| "learning_rate": 3.824798160583012e-06, |
| "loss": 4.246830940246582e-06, |
| "num_tokens": 227812.0, |
| "reward": -0.22999998927116394, |
| "reward_std": 0.8891193866729736, |
| "rewards/reward_environment_execution/mean": -0.23000000417232513, |
| "rewards/reward_environment_execution/std": 0.31176915764808655, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 40 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 729.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 839.0, |
| "completions/max_terminated_length": 839.0, |
| "completions/mean_length": 729.5, |
| "completions/mean_terminated_length": 729.5, |
| "completions/min_length": 608.0, |
| "completions/min_terminated_length": 608.0, |
| "epoch": 0.41, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11686036735773087, |
| "kl": 8.182951478374889e-05, |
| "learning_rate": 3.7500000000000005e-06, |
| "loss": 3.3080577850341797e-06, |
| "num_tokens": 233834.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 41 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 582.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 772.0, |
| "completions/max_terminated_length": 772.0, |
| "completions/mean_length": 582.0, |
| "completions/mean_terminated_length": 582.0, |
| "completions/min_length": 430.0, |
| "completions/min_terminated_length": 430.0, |
| "epoch": 0.42, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.15176120400428772, |
| "kl": 0.00013200526700529736, |
| "learning_rate": 3.6736789069647273e-06, |
| "loss": 5.21540641784668e-06, |
| "num_tokens": 239438.0, |
| "reward": -0.22999998927116394, |
| "reward_std": 0.8891193866729736, |
| "rewards/reward_environment_execution/mean": -0.23000000417232513, |
| "rewards/reward_environment_execution/std": 0.31176915764808655, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 42 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 797.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 949.0, |
| "completions/max_terminated_length": 949.0, |
| "completions/mean_length": 797.75, |
| "completions/mean_terminated_length": 797.75, |
| "completions/min_length": 620.0, |
| "completions/min_terminated_length": 620.0, |
| "epoch": 0.43, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10845938324928284, |
| "kl": 5.39204447704833e-05, |
| "learning_rate": 3.595927866972694e-06, |
| "loss": 2.115964889526367e-06, |
| "num_tokens": 245737.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 43 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 650.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 885.0, |
| "completions/max_terminated_length": 885.0, |
| "completions/mean_length": 650.25, |
| "completions/mean_terminated_length": 650.25, |
| "completions/min_length": 508.0, |
| "completions/min_terminated_length": 508.0, |
| "epoch": 0.44, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.16251257061958313, |
| "kl": 0.0001057987265085103, |
| "learning_rate": 3.516841607689501e-06, |
| "loss": 4.276633262634277e-06, |
| "num_tokens": 251618.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 44 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 692.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 719.0, |
| "completions/max_terminated_length": 719.0, |
| "completions/mean_length": 692.75, |
| "completions/mean_terminated_length": 692.75, |
| "completions/min_length": 666.0, |
| "completions/min_terminated_length": 666.0, |
| "epoch": 0.45, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.15349064767360687, |
| "kl": 7.098338755895384e-05, |
| "learning_rate": 3.436516483539781e-06, |
| "loss": 2.853572368621826e-06, |
| "num_tokens": 257493.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 45 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 576.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 678.0, |
| "completions/max_terminated_length": 678.0, |
| "completions/mean_length": 576.5, |
| "completions/mean_terminated_length": 576.5, |
| "completions/min_length": 451.0, |
| "completions/min_terminated_length": 451.0, |
| "epoch": 0.46, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.16658417880535126, |
| "kl": 0.00024796422985673416, |
| "learning_rate": 3.3550503583141726e-06, |
| "loss": 9.894371032714844e-06, |
| "num_tokens": 262907.0, |
| "reward": -0.6150000095367432, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.36500000953674316, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": -0.32499998807907104, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.07500000298023224, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 46 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 679.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 1068.0, |
| "completions/max_terminated_length": 1068.0, |
| "completions/mean_length": 679.75, |
| "completions/mean_terminated_length": 679.75, |
| "completions/min_length": 452.0, |
| "completions/min_terminated_length": 452.0, |
| "epoch": 0.47, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.17304089665412903, |
| "kl": 0.00017170603496197145, |
| "learning_rate": 3.272542485937369e-06, |
| "loss": 6.839632987976074e-06, |
| "num_tokens": 268906.0, |
| "reward": -0.6150000095367432, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.36500000953674316, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": -0.32499998807907104, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.07500000298023224, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 47 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 637.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 698.0, |
| "completions/max_terminated_length": 698.0, |
| "completions/mean_length": 637.25, |
| "completions/mean_terminated_length": 637.25, |
| "completions/min_length": 598.0, |
| "completions/min_terminated_length": 598.0, |
| "epoch": 0.48, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.00041183308348990977, |
| "kl": 0.00011487708798085805, |
| "learning_rate": 3.189093389542498e-06, |
| "loss": 4.5950837375130504e-06, |
| "num_tokens": 274563.0, |
| "reward": 0.5400000214576721, |
| "reward_std": 0.0, |
| "rewards/reward_environment_execution/mean": 0.03999999910593033, |
| "rewards/reward_environment_execution/std": 0.0, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.30000001192092896, |
| "rewards/reward_investigation_quality/std": 0.0, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 48 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 802.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 1189.0, |
| "completions/max_terminated_length": 1189.0, |
| "completions/mean_length": 802.25, |
| "completions/mean_terminated_length": 802.25, |
| "completions/min_length": 501.0, |
| "completions/min_terminated_length": 501.0, |
| "epoch": 0.49, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.17298302054405212, |
| "kl": 0.00021534036750381347, |
| "learning_rate": 3.1048047389991693e-06, |
| "loss": 8.627772331237793e-06, |
| "num_tokens": 281048.0, |
| "reward": -0.22999998927116394, |
| "reward_std": 0.8891193866729736, |
| "rewards/reward_environment_execution/mean": -0.23000000417232513, |
| "rewards/reward_environment_execution/std": 0.31176915764808655, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 49 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 676.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 725.0, |
| "completions/max_terminated_length": 725.0, |
| "completions/mean_length": 676.0, |
| "completions/mean_terminated_length": 676.0, |
| "completions/min_length": 631.0, |
| "completions/min_terminated_length": 631.0, |
| "epoch": 0.5, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.13108237087726593, |
| "kl": 0.00016818426684039878, |
| "learning_rate": 3.019779227044398e-06, |
| "loss": 6.7427754402160645e-06, |
| "num_tokens": 287032.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 50 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 636.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 682.0, |
| "completions/max_terminated_length": 682.0, |
| "completions/mean_length": 636.25, |
| "completions/mean_terminated_length": 636.25, |
| "completions/min_length": 609.0, |
| "completions/min_terminated_length": 609.0, |
| "epoch": 0.51, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.0004512854793574661, |
| "kl": 0.00010897797619691119, |
| "learning_rate": 2.9341204441673267e-06, |
| "loss": 4.359118975116871e-06, |
| "num_tokens": 292677.0, |
| "reward": 0.5400000214576721, |
| "reward_std": 0.0, |
| "rewards/reward_environment_execution/mean": 0.03999999910593033, |
| "rewards/reward_environment_execution/std": 0.0, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.30000001192092896, |
| "rewards/reward_investigation_quality/std": 0.0, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 51 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 701.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 836.0, |
| "completions/max_terminated_length": 836.0, |
| "completions/mean_length": 701.75, |
| "completions/mean_terminated_length": 701.75, |
| "completions/min_length": 576.0, |
| "completions/min_terminated_length": 576.0, |
| "epoch": 0.52, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12227238714694977, |
| "kl": 0.00010721037870098371, |
| "learning_rate": 2.847932752400164e-06, |
| "loss": 4.291534423828125e-06, |
| "num_tokens": 298760.0, |
| "reward": 0.15250001847743988, |
| "reward_std": 0.7683477997779846, |
| "rewards/reward_environment_execution/mean": -0.09750000387430191, |
| "rewards/reward_environment_execution/std": 0.2683747410774231, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 52 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 628.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 686.0, |
| "completions/max_terminated_length": 686.0, |
| "completions/mean_length": 628.25, |
| "completions/mean_terminated_length": 628.25, |
| "completions/min_length": 586.0, |
| "completions/min_terminated_length": 586.0, |
| "epoch": 0.53, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1427244395017624, |
| "kl": 0.00010186010331381112, |
| "learning_rate": 2.761321158169134e-06, |
| "loss": 4.112720489501953e-06, |
| "num_tokens": 304381.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 53 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 759.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 891.0, |
| "completions/max_terminated_length": 891.0, |
| "completions/mean_length": 759.25, |
| "completions/mean_terminated_length": 759.25, |
| "completions/min_length": 688.0, |
| "completions/min_terminated_length": 688.0, |
| "epoch": 0.54, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11026006191968918, |
| "kl": 0.00010889488567045191, |
| "learning_rate": 2.6743911843603134e-06, |
| "loss": 4.366040229797363e-06, |
| "num_tokens": 310514.0, |
| "reward": -0.22999998927116394, |
| "reward_std": 0.8891193866729736, |
| "rewards/reward_environment_execution/mean": -0.23000000417232513, |
| "rewards/reward_environment_execution/std": 0.31176915764808655, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 54 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 503.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 611.0, |
| "completions/max_terminated_length": 611.0, |
| "completions/mean_length": 503.5, |
| "completions/mean_terminated_length": 503.5, |
| "completions/min_length": 390.0, |
| "completions/min_terminated_length": 390.0, |
| "epoch": 0.55, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.14209793508052826, |
| "kl": 0.000171166546351742, |
| "learning_rate": 2.587248741756253e-06, |
| "loss": 6.854534149169922e-06, |
| "num_tokens": 315808.0, |
| "reward": -0.6150000095367432, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.36500000953674316, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": -0.32499998807907104, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.07500000298023224, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 55 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 627.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 742.0, |
| "completions/max_terminated_length": 742.0, |
| "completions/mean_length": 627.0, |
| "completions/mean_terminated_length": 627.0, |
| "completions/min_length": 511.0, |
| "completions/min_terminated_length": 511.0, |
| "epoch": 0.56, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.14504601061344147, |
| "kl": 0.0003341738956805784, |
| "learning_rate": 2.5e-06, |
| "loss": 1.3366341590881348e-05, |
| "num_tokens": 321424.0, |
| "reward": -0.23249998688697815, |
| "reward_std": 0.886242151260376, |
| "rewards/reward_environment_execution/mean": -0.23250000178813934, |
| "rewards/reward_environment_execution/std": 0.3089093565940857, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 56 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 615.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 628.0, |
| "completions/max_terminated_length": 628.0, |
| "completions/mean_length": 615.5, |
| "completions/mean_terminated_length": 615.5, |
| "completions/min_length": 602.0, |
| "completions/min_terminated_length": 602.0, |
| "epoch": 0.57, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.14204947650432587, |
| "kl": 0.00015341720063588582, |
| "learning_rate": 2.4127512582437486e-06, |
| "loss": 6.288290023803711e-06, |
| "num_tokens": 327142.0, |
| "reward": 0.6147500276565552, |
| "reward_std": 0.14949996769428253, |
| "rewards/reward_environment_execution/mean": 0.16474999487400055, |
| "rewards/reward_environment_execution/std": 0.24950000643730164, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.25, |
| "rewards/reward_investigation_quality/std": 0.10000000894069672, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 57 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 660.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 896.0, |
| "completions/max_terminated_length": 896.0, |
| "completions/mean_length": 660.0, |
| "completions/mean_terminated_length": 660.0, |
| "completions/min_length": 447.0, |
| "completions/min_terminated_length": 447.0, |
| "epoch": 0.58, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.0004325012268964201, |
| "kl": 0.0001194007618323667, |
| "learning_rate": 2.325608815639687e-06, |
| "loss": 4.77603043691488e-06, |
| "num_tokens": 333058.0, |
| "reward": 0.5400000214576721, |
| "reward_std": 0.0, |
| "rewards/reward_environment_execution/mean": 0.03999999910593033, |
| "rewards/reward_environment_execution/std": 0.0, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.30000001192092896, |
| "rewards/reward_investigation_quality/std": 0.0, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 58 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 708.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 816.0, |
| "completions/max_terminated_length": 816.0, |
| "completions/mean_length": 708.5, |
| "completions/mean_terminated_length": 708.5, |
| "completions/min_length": 532.0, |
| "completions/min_terminated_length": 532.0, |
| "epoch": 0.59, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11153003573417664, |
| "kl": 0.0001295358106290223, |
| "learning_rate": 2.238678841830867e-06, |
| "loss": 5.230307579040527e-06, |
| "num_tokens": 339156.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 59 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 677.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 804.0, |
| "completions/max_terminated_length": 804.0, |
| "completions/mean_length": 677.5, |
| "completions/mean_terminated_length": 677.5, |
| "completions/min_length": 603.0, |
| "completions/min_terminated_length": 603.0, |
| "epoch": 0.6, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11006420850753784, |
| "kl": 0.00012863795018347446, |
| "learning_rate": 2.1520672475998374e-06, |
| "loss": 5.163252353668213e-06, |
| "num_tokens": 344974.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 60 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 663.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 794.0, |
| "completions/max_terminated_length": 794.0, |
| "completions/mean_length": 663.0, |
| "completions/mean_terminated_length": 663.0, |
| "completions/min_length": 600.0, |
| "completions/min_terminated_length": 600.0, |
| "epoch": 0.61, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.14184613525867462, |
| "kl": 0.00038980014687695075, |
| "learning_rate": 2.0658795558326745e-06, |
| "loss": 1.55717134475708e-05, |
| "num_tokens": 350734.0, |
| "reward": -0.22999998927116394, |
| "reward_std": 0.8891193866729736, |
| "rewards/reward_environment_execution/mean": -0.23000000417232513, |
| "rewards/reward_environment_execution/std": 0.31176915764808655, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 61 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 660.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 827.0, |
| "completions/max_terminated_length": 827.0, |
| "completions/mean_length": 660.75, |
| "completions/mean_terminated_length": 660.75, |
| "completions/min_length": 528.0, |
| "completions/min_terminated_length": 528.0, |
| "epoch": 0.62, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.00037498201709240675, |
| "kl": 0.000125649748952128, |
| "learning_rate": 1.9802207729556023e-06, |
| "loss": 5.025989594287239e-06, |
| "num_tokens": 356657.0, |
| "reward": 0.5400000214576721, |
| "reward_std": 0.0, |
| "rewards/reward_environment_execution/mean": 0.03999999910593033, |
| "rewards/reward_environment_execution/std": 0.0, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.30000001192092896, |
| "rewards/reward_investigation_quality/std": 0.0, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 62 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 476.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 645.0, |
| "completions/max_terminated_length": 645.0, |
| "completions/mean_length": 476.0, |
| "completions/mean_terminated_length": 476.0, |
| "completions/min_length": 255.0, |
| "completions/min_terminated_length": 255.0, |
| "epoch": 0.63, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.20083816349506378, |
| "kl": 0.0002776872024696786, |
| "learning_rate": 1.895195261000831e-06, |
| "loss": 1.1064112186431885e-05, |
| "num_tokens": 361677.0, |
| "reward": 0.14749999344348907, |
| "reward_std": 0.7651306986808777, |
| "rewards/reward_environment_execution/mean": -0.10250000655651093, |
| "rewards/reward_environment_execution/std": 0.26537710428237915, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 63 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 634.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 855.0, |
| "completions/max_terminated_length": 855.0, |
| "completions/mean_length": 634.25, |
| "completions/mean_terminated_length": 634.25, |
| "completions/min_length": 506.0, |
| "completions/min_terminated_length": 506.0, |
| "epoch": 0.64, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12447395920753479, |
| "kl": 0.00040143656588043086, |
| "learning_rate": 1.8109066104575023e-06, |
| "loss": 1.6085803508758545e-05, |
| "num_tokens": 367322.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 64 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 641.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 835.0, |
| "completions/max_terminated_length": 835.0, |
| "completions/mean_length": 641.75, |
| "completions/mean_terminated_length": 641.75, |
| "completions/min_length": 520.0, |
| "completions/min_terminated_length": 520.0, |
| "epoch": 0.65, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.19145435094833374, |
| "kl": 0.00024168194067897275, |
| "learning_rate": 1.7274575140626318e-06, |
| "loss": 9.670853614807129e-06, |
| "num_tokens": 372989.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 65 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 433.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 478.0, |
| "completions/max_terminated_length": 478.0, |
| "completions/mean_length": 433.5, |
| "completions/mean_terminated_length": 433.5, |
| "completions/min_length": 400.0, |
| "completions/min_terminated_length": 400.0, |
| "epoch": 0.66, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1740998774766922, |
| "kl": 0.0004632533964468166, |
| "learning_rate": 1.6449496416858285e-06, |
| "loss": 1.8537044525146484e-05, |
| "num_tokens": 377831.0, |
| "reward": -0.22999998927116394, |
| "reward_std": 0.8891193866729736, |
| "rewards/reward_environment_execution/mean": -0.23000000417232513, |
| "rewards/reward_environment_execution/std": 0.31176915764808655, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 66 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 429.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 696.0, |
| "completions/max_terminated_length": 696.0, |
| "completions/mean_length": 429.5, |
| "completions/mean_terminated_length": 429.5, |
| "completions/min_length": 228.0, |
| "completions/min_terminated_length": 228.0, |
| "epoch": 0.67, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.229574516415596, |
| "kl": 0.0006340235704556108, |
| "learning_rate": 1.56348351646022e-06, |
| "loss": 2.5369226932525635e-05, |
| "num_tokens": 382649.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 67 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 655.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 870.0, |
| "completions/max_terminated_length": 870.0, |
| "completions/mean_length": 655.25, |
| "completions/mean_terminated_length": 655.25, |
| "completions/min_length": 544.0, |
| "completions/min_terminated_length": 544.0, |
| "epoch": 0.68, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.15165656805038452, |
| "kl": 0.00021119948905834462, |
| "learning_rate": 1.4831583923105e-06, |
| "loss": 8.471310138702393e-06, |
| "num_tokens": 388546.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 68 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 611.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 764.0, |
| "completions/max_terminated_length": 764.0, |
| "completions/mean_length": 611.75, |
| "completions/mean_terminated_length": 611.75, |
| "completions/min_length": 539.0, |
| "completions/min_terminated_length": 539.0, |
| "epoch": 0.69, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1276257336139679, |
| "kl": 0.00015681719924032222, |
| "learning_rate": 1.4040721330273063e-06, |
| "loss": 6.2659382820129395e-06, |
| "num_tokens": 394097.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 69 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 742.25, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 948.0, |
| "completions/max_terminated_length": 948.0, |
| "completions/mean_length": 742.25, |
| "completions/mean_terminated_length": 742.25, |
| "completions/min_length": 557.0, |
| "completions/min_terminated_length": 557.0, |
| "epoch": 0.7, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10219409316778183, |
| "kl": 0.00011395674209779827, |
| "learning_rate": 1.3263210930352737e-06, |
| "loss": 4.589557647705078e-06, |
| "num_tokens": 400162.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 70 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 575.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 610.0, |
| "completions/max_terminated_length": 610.0, |
| "completions/mean_length": 575.75, |
| "completions/mean_terminated_length": 575.75, |
| "completions/min_length": 538.0, |
| "completions/min_terminated_length": 538.0, |
| "epoch": 0.71, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.14183653891086578, |
| "kl": 0.00029851996441720985, |
| "learning_rate": 1.2500000000000007e-06, |
| "loss": 1.1943280696868896e-05, |
| "num_tokens": 405581.0, |
| "reward": 0.1550000160932541, |
| "reward_std": 0.7699999809265137, |
| "rewards/reward_environment_execution/mean": -0.0949999988079071, |
| "rewards/reward_environment_execution/std": 0.26999998092651367, |
| "rewards/reward_format_compliance/mean": 0.02499999850988388, |
| "rewards/reward_format_compliance/std": 0.3499999940395355, |
| "rewards/reward_investigation_quality/mean": 0.22500000894069672, |
| "rewards/reward_investigation_quality/std": 0.15000002086162567, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 71 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 647.75, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 970.0, |
| "completions/max_terminated_length": 970.0, |
| "completions/mean_length": 647.75, |
| "completions/mean_terminated_length": 647.75, |
| "completions/min_length": 429.0, |
| "completions/min_terminated_length": 429.0, |
| "epoch": 0.72, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.15262170135974884, |
| "kl": 0.0001618979367776774, |
| "learning_rate": 1.1752018394169882e-06, |
| "loss": 6.4820051193237305e-06, |
| "num_tokens": 411464.0, |
| "reward": -0.22999998927116394, |
| "reward_std": 0.8891193866729736, |
| "rewards/reward_environment_execution/mean": -0.23000000417232513, |
| "rewards/reward_environment_execution/std": 0.31176915764808655, |
| "rewards/reward_format_compliance/mean": -0.15000000596046448, |
| "rewards/reward_format_compliance/std": 0.404145210981369, |
| "rewards/reward_investigation_quality/mean": 0.15000000596046448, |
| "rewards/reward_investigation_quality/std": 0.17320507764816284, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 72 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 533.5, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 845.0, |
| "completions/max_terminated_length": 845.0, |
| "completions/mean_length": 533.5, |
| "completions/mean_terminated_length": 533.5, |
| "completions/min_length": 328.0, |
| "completions/min_terminated_length": 328.0, |
| "epoch": 0.73, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.00044858516776002944, |
| "kl": 0.00018171430565416813, |
| "learning_rate": 1.1020177413231334e-06, |
| "loss": 7.26857251720503e-06, |
| "num_tokens": 416698.0, |
| "reward": 0.5400000214576721, |
| "reward_std": 0.0, |
| "rewards/reward_environment_execution/mean": 0.03999999910593033, |
| "rewards/reward_environment_execution/std": 0.0, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.30000001192092896, |
| "rewards/reward_investigation_quality/std": 0.0, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 73 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 492.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 612.0, |
| "completions/max_terminated_length": 612.0, |
| "completions/mean_length": 492.0, |
| "completions/mean_terminated_length": 492.0, |
| "completions/min_length": 387.0, |
| "completions/min_terminated_length": 387.0, |
| "epoch": 0.74, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.0007429426186718047, |
| "kl": 0.00022592291861656122, |
| "learning_rate": 1.0305368692688175e-06, |
| "loss": 9.036917617777362e-06, |
| "num_tokens": 421782.0, |
| "reward": 0.5400000214576721, |
| "reward_std": 0.0, |
| "rewards/reward_environment_execution/mean": 0.03999999910593033, |
| "rewards/reward_environment_execution/std": 0.0, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.30000001192092896, |
| "rewards/reward_investigation_quality/std": 0.0, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 74 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completion_length": 735.0, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 813.0, |
| "completions/max_terminated_length": 813.0, |
| "completions/mean_length": 735.0, |
| "completions/mean_terminated_length": 735.0, |
| "completions/min_length": 637.0, |
| "completions/min_terminated_length": 637.0, |
| "epoch": 0.75, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12998387217521667, |
| "kl": 0.0001630033293622546, |
| "learning_rate": 9.608463116858544e-07, |
| "loss": 6.616115570068359e-06, |
| "num_tokens": 427826.0, |
| "reward": 0.6147500276565552, |
| "reward_std": 0.14949996769428253, |
| "rewards/reward_environment_execution/mean": 0.16474999487400055, |
| "rewards/reward_environment_execution/std": 0.24950000643730164, |
| "rewards/reward_format_compliance/mean": 0.20000000298023224, |
| "rewards/reward_format_compliance/std": 0.0, |
| "rewards/reward_investigation_quality/mean": 0.25, |
| "rewards/reward_investigation_quality/std": 0.10000000894069672, |
| "rewards/reward_os_mechanics/mean": 0.0, |
| "rewards/reward_os_mechanics/std": 0.0, |
| "step": 75 |
| } |
| ], |
| "logging_steps": 1, |
| "max_steps": 100, |
| "num_input_tokens_seen": 427826, |
| "num_train_epochs": 1, |
| "save_steps": 25, |
| "stateful_callbacks": { |
| "TrainerControl": { |
| "args": { |
| "should_epoch_stop": false, |
| "should_evaluate": false, |
| "should_log": false, |
| "should_save": true, |
| "should_training_stop": false |
| }, |
| "attributes": {} |
| } |
| }, |
| "total_flos": 0.0, |
| "train_batch_size": 1, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
