| { |
| "best_global_step": null, |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 0.08202099737532809, |
| "eval_steps": 500, |
| "global_step": 500, |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 6.05, |
| "completions/max_terminated_length": 6.05, |
| "completions/mean_length": 4.61875, |
| "completions/mean_terminated_length": 4.61875, |
| "completions/min_length": 3.5, |
| "completions/min_terminated_length": 3.5, |
| "entropy": 0.8203600860819279, |
| "epoch": 0.0032808398950131233, |
| "frac_reward_zero_std": 0.3, |
| "grad_norm": 5.8125, |
| "learning_rate": 4.990500000000001e-06, |
| "loss": 0.06519373655319213, |
| "num_tokens": 5635.0, |
| "reward": 1.1270193845033645, |
| "reward_std": 0.04764379555272171, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.050634765625, |
| "rewards/reward_length/std": 0.015013676788657903, |
| "rewards/reward_semantic/mean": 0.43183712139725683, |
| "rewards/reward_semantic/std": 0.15772892236709596, |
| "step": 20, |
| "step_time": 1.6850911805999202 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 7.15, |
| "completions/max_terminated_length": 7.15, |
| "completions/mean_length": 5.00625, |
| "completions/mean_terminated_length": 5.00625, |
| "completions/min_length": 3.65, |
| "completions/min_terminated_length": 3.65, |
| "entropy": 0.929626553109847, |
| "epoch": 0.006561679790026247, |
| "frac_reward_zero_std": 0.2, |
| "grad_norm": 0.0, |
| "learning_rate": 4.9805e-06, |
| "loss": 0.07899624109268188, |
| "num_tokens": 11660.0, |
| "reward": 1.0556338340044022, |
| "reward_std": 0.017533331926097163, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.0642578125, |
| "rewards/reward_length/std": 0.013992739858804271, |
| "rewards/reward_semantic/mean": 0.19615575280040504, |
| "rewards/reward_semantic/std": 0.05699479430913925, |
| "step": 40, |
| "step_time": 1.79874156090018 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 5.5, |
| "completions/max_terminated_length": 5.5, |
| "completions/mean_length": 4.325, |
| "completions/mean_terminated_length": 4.325, |
| "completions/min_length": 3.5, |
| "completions/min_terminated_length": 3.5, |
| "entropy": 0.7027173407775991, |
| "epoch": 0.00984251968503937, |
| "frac_reward_zero_std": 0.3, |
| "grad_norm": 0.0, |
| "learning_rate": 4.9705e-06, |
| "loss": 0.023067036271095277, |
| "num_tokens": 17440.0, |
| "reward": 1.1290264517068862, |
| "reward_std": 0.03574381650541909, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.052685546875, |
| "rewards/reward_length/std": 0.012657346948981286, |
| "rewards/reward_semantic/mean": 0.4388690486550331, |
| "rewards/reward_semantic/std": 0.11865786015987397, |
| "step": 60, |
| "step_time": 1.6641022313501708 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 7.05, |
| "completions/max_terminated_length": 7.05, |
| "completions/mean_length": 4.89375, |
| "completions/mean_terminated_length": 4.89375, |
| "completions/min_length": 3.45, |
| "completions/min_terminated_length": 3.45, |
| "entropy": 0.8099277728608285, |
| "epoch": 0.013123359580052493, |
| "frac_reward_zero_std": 0.2, |
| "grad_norm": 2.578125, |
| "learning_rate": 4.9605e-06, |
| "loss": 0.12280056476593018, |
| "num_tokens": 23631.0, |
| "reward": 1.08654942214489, |
| "reward_std": 0.040316034654097166, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.06103515625, |
| "rewards/reward_length/std": 0.022364351549185812, |
| "rewards/reward_semantic/mean": 0.2986706353724003, |
| "rewards/reward_semantic/std": 0.1318330906331539, |
| "step": 80, |
| "step_time": 1.7635547111498453 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 6.75, |
| "completions/max_terminated_length": 6.75, |
| "completions/mean_length": 4.75, |
| "completions/mean_terminated_length": 4.75, |
| "completions/min_length": 3.4, |
| "completions/min_terminated_length": 3.4, |
| "entropy": 0.8052310052706162, |
| "epoch": 0.016404199475065617, |
| "frac_reward_zero_std": 0.2, |
| "grad_norm": 3.171875, |
| "learning_rate": 4.9505e-06, |
| "loss": 0.007156229019165039, |
| "num_tokens": 29599.0, |
| "reward": 1.0940562933683395, |
| "reward_std": 0.030761140065442304, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.06103515625, |
| "rewards/reward_length/std": 0.017728388286195697, |
| "rewards/reward_semantic/mean": 0.3236935347318649, |
| "rewards/reward_semantic/std": 0.1025692880153656, |
| "step": 100, |
| "step_time": 1.6173622274000992 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 5.6, |
| "completions/max_terminated_length": 5.6, |
| "completions/mean_length": 4.125, |
| "completions/mean_terminated_length": 4.125, |
| "completions/min_length": 3.15, |
| "completions/min_terminated_length": 3.15, |
| "entropy": 0.7014230243818019, |
| "epoch": 0.01968503937007874, |
| "frac_reward_zero_std": 0.3, |
| "grad_norm": 0.0, |
| "learning_rate": 4.9405e-06, |
| "loss": 0.07218956351280212, |
| "num_tokens": 35107.0, |
| "reward": 1.1007445603609085, |
| "reward_std": 0.028740251205454113, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.04814453125, |
| "rewards/reward_length/std": 0.010129839461296797, |
| "rewards/reward_semantic/mean": 0.34383928664028646, |
| "rewards/reward_semantic/std": 0.09484268203377724, |
| "step": 120, |
| "step_time": 1.5441937706006683 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 7.25, |
| "completions/max_terminated_length": 7.25, |
| "completions/mean_length": 4.9625, |
| "completions/mean_terminated_length": 4.9625, |
| "completions/min_length": 3.5, |
| "completions/min_terminated_length": 3.5, |
| "entropy": 0.7183661407629188, |
| "epoch": 0.022965879265091863, |
| "frac_reward_zero_std": 0.15, |
| "grad_norm": 2.609375, |
| "learning_rate": 4.9305000000000005e-06, |
| "loss": 0.05585476756095886, |
| "num_tokens": 40989.0, |
| "reward": 1.1004581451416016, |
| "reward_std": 0.03431669266719837, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.059765625, |
| "rewards/reward_length/std": 0.017409664555452765, |
| "rewards/reward_semantic/mean": 0.34482143111526964, |
| "rewards/reward_semantic/std": 0.11257617846131325, |
| "step": 140, |
| "step_time": 1.7615427238992196 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 7.05, |
| "completions/max_terminated_length": 7.05, |
| "completions/mean_length": 4.89375, |
| "completions/mean_terminated_length": 4.89375, |
| "completions/min_length": 3.5, |
| "completions/min_terminated_length": 3.5, |
| "entropy": 0.6915914016732131, |
| "epoch": 0.026246719160104987, |
| "frac_reward_zero_std": 0.1, |
| "grad_norm": 2.859375, |
| "learning_rate": 4.920500000000001e-06, |
| "loss": 0.0472392201423645, |
| "num_tokens": 47220.0, |
| "reward": 1.0850708901882171, |
| "reward_std": 0.06178898175858194, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.065576171875, |
| "rewards/reward_length/std": 0.01732458183541894, |
| "rewards/reward_semantic/mean": 0.2944990089163184, |
| "rewards/reward_semantic/std": 0.2051414344459772, |
| "step": 160, |
| "step_time": 1.7252537605003453 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 6.25, |
| "completions/max_terminated_length": 6.25, |
| "completions/mean_length": 4.55625, |
| "completions/mean_terminated_length": 4.55625, |
| "completions/min_length": 3.35, |
| "completions/min_terminated_length": 3.35, |
| "entropy": 0.8429647823562846, |
| "epoch": 0.02952755905511811, |
| "frac_reward_zero_std": 0.05, |
| "grad_norm": 3.078125, |
| "learning_rate": 4.910500000000001e-06, |
| "loss": 0.08106310367584228, |
| "num_tokens": 53053.0, |
| "reward": 1.0679614007472993, |
| "reward_std": 0.03796071938340902, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.0566650390625, |
| "rewards/reward_length/std": 0.014682336029363797, |
| "rewards/reward_semantic/mean": 0.23598214369267226, |
| "rewards/reward_semantic/std": 0.12518299631774427, |
| "step": 180, |
| "step_time": 1.7508945095991293 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 6.2, |
| "completions/max_terminated_length": 6.2, |
| "completions/mean_length": 4.5375, |
| "completions/mean_terminated_length": 4.5375, |
| "completions/min_length": 3.75, |
| "completions/min_terminated_length": 3.75, |
| "entropy": 0.5915892604841246, |
| "epoch": 0.03280839895013123, |
| "frac_reward_zero_std": 0.35, |
| "grad_norm": 3.71875, |
| "learning_rate": 4.9005e-06, |
| "loss": 0.0888818621635437, |
| "num_tokens": 58931.0, |
| "reward": 1.120530515909195, |
| "reward_std": 0.0405463597693597, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.0493896484375, |
| "rewards/reward_length/std": 0.008753356069792061, |
| "rewards/reward_semantic/mean": 0.4100000007078052, |
| "rewards/reward_semantic/std": 0.13480054624378682, |
| "step": 200, |
| "step_time": 1.713693901949955 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 6.45, |
| "completions/max_terminated_length": 6.45, |
| "completions/mean_length": 5.15, |
| "completions/mean_terminated_length": 5.15, |
| "completions/min_length": 4.2, |
| "completions/min_terminated_length": 4.2, |
| "entropy": 0.6395838386631112, |
| "epoch": 0.03608923884514436, |
| "frac_reward_zero_std": 0.35, |
| "grad_norm": 2.625, |
| "learning_rate": 4.8905e-06, |
| "loss": 0.06417204141616821, |
| "num_tokens": 65139.0, |
| "reward": 1.1249804645776749, |
| "reward_std": 0.027039075889479135, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.056640625, |
| "rewards/reward_length/std": 0.011077208817005158, |
| "rewards/reward_semantic/mean": 0.4260416656732559, |
| "rewards/reward_semantic/std": 0.08868714347481728, |
| "step": 220, |
| "step_time": 1.7278446534002796 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 5.4, |
| "completions/max_terminated_length": 5.4, |
| "completions/mean_length": 4.2625, |
| "completions/mean_terminated_length": 4.2625, |
| "completions/min_length": 3.0, |
| "completions/min_terminated_length": 3.0, |
| "entropy": 0.5652526923571713, |
| "epoch": 0.03937007874015748, |
| "frac_reward_zero_std": 0.35, |
| "grad_norm": 8.125, |
| "learning_rate": 4.8805e-06, |
| "loss": 0.07150443196296692, |
| "num_tokens": 71221.0, |
| "reward": 1.1466667413711549, |
| "reward_std": 0.03061878081643954, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.063330078125, |
| "rewards/reward_length/std": 0.015704452409408987, |
| "rewards/reward_semantic/mean": 0.4994440995156765, |
| "rewards/reward_semantic/std": 0.10041465926915408, |
| "step": 240, |
| "step_time": 1.6952891380005894 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 6.3, |
| "completions/max_terminated_length": 6.3, |
| "completions/mean_length": 4.46875, |
| "completions/mean_terminated_length": 4.46875, |
| "completions/min_length": 3.35, |
| "completions/min_terminated_length": 3.35, |
| "entropy": 0.672205779588694, |
| "epoch": 0.04265091863517061, |
| "frac_reward_zero_std": 0.2, |
| "grad_norm": 6.4375, |
| "learning_rate": 4.8705e-06, |
| "loss": 0.09235309958457946, |
| "num_tokens": 77416.0, |
| "reward": 1.0690006405115127, |
| "reward_std": 0.019543477255501783, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.0533203125, |
| "rewards/reward_length/std": 0.017377534485422074, |
| "rewards/reward_semantic/mean": 0.23888888889923693, |
| "rewards/reward_semantic/std": 0.0630335971713066, |
| "step": 260, |
| "step_time": 1.7545264535998286 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 5.3, |
| "completions/max_terminated_length": 5.3, |
| "completions/mean_length": 4.525, |
| "completions/mean_terminated_length": 4.525, |
| "completions/min_length": 3.65, |
| "completions/min_terminated_length": 3.65, |
| "entropy": 0.5580058551760885, |
| "epoch": 0.045931758530183726, |
| "frac_reward_zero_std": 0.45, |
| "grad_norm": 4.0625, |
| "learning_rate": 4.8605e-06, |
| "loss": 0.02879692018032074, |
| "num_tokens": 83332.0, |
| "reward": 1.1027960389852525, |
| "reward_std": 0.011529068512027152, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.0558349609375, |
| "rewards/reward_length/std": 0.008870826044585556, |
| "rewards/reward_semantic/mean": 0.3519593259319663, |
| "rewards/reward_semantic/std": 0.03782847560942173, |
| "step": 280, |
| "step_time": 1.5900321766996968 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 6.35, |
| "completions/max_terminated_length": 6.35, |
| "completions/mean_length": 4.56875, |
| "completions/mean_terminated_length": 4.56875, |
| "completions/min_length": 3.2, |
| "completions/min_terminated_length": 3.2, |
| "entropy": 0.8288409212742408, |
| "epoch": 0.04921259842519685, |
| "frac_reward_zero_std": 0.2, |
| "grad_norm": 0.0, |
| "learning_rate": 4.8505e-06, |
| "loss": 0.044028478860855105, |
| "num_tokens": 89567.0, |
| "reward": 1.1026858359575271, |
| "reward_std": 0.03983475278655533, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.062744140625, |
| "rewards/reward_length/std": 0.018717235408257694, |
| "rewards/reward_semantic/mean": 0.35274351127445697, |
| "rewards/reward_semantic/std": 0.13158041909337043, |
| "step": 300, |
| "step_time": 1.7004065297001034 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 5.65, |
| "completions/max_terminated_length": 5.65, |
| "completions/mean_length": 4.5875, |
| "completions/mean_terminated_length": 4.5875, |
| "completions/min_length": 3.5, |
| "completions/min_terminated_length": 3.5, |
| "entropy": 0.7240128130373705, |
| "epoch": 0.05249343832020997, |
| "frac_reward_zero_std": 0.2, |
| "grad_norm": 5.0625, |
| "learning_rate": 4.8405000000000005e-06, |
| "loss": 0.03284294605255127, |
| "num_tokens": 95157.0, |
| "reward": 1.0648824661970138, |
| "reward_std": 0.014824012622557347, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.055029296875, |
| "rewards/reward_length/std": 0.010819258564151824, |
| "rewards/reward_semantic/mean": 0.2254464277997613, |
| "rewards/reward_semantic/std": 0.04843873754143715, |
| "step": 320, |
| "step_time": 1.6396856073492927 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 6.25, |
| "completions/max_terminated_length": 6.25, |
| "completions/mean_length": 4.625, |
| "completions/mean_terminated_length": 4.625, |
| "completions/min_length": 3.35, |
| "completions/min_terminated_length": 3.35, |
| "entropy": 0.8246265778172528, |
| "epoch": 0.0557742782152231, |
| "frac_reward_zero_std": 0.35, |
| "grad_norm": 0.0, |
| "learning_rate": 4.830500000000001e-06, |
| "loss": 0.05039528012275696, |
| "num_tokens": 100857.0, |
| "reward": 1.0999272018671036, |
| "reward_std": 0.012020479088823777, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.0510986328125, |
| "rewards/reward_length/std": 0.013131194305606187, |
| "rewards/reward_semantic/mean": 0.3416071431711316, |
| "rewards/reward_semantic/std": 0.03859963417053223, |
| "step": 340, |
| "step_time": 1.6233859046002181 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 5.75, |
| "completions/max_terminated_length": 5.75, |
| "completions/mean_length": 4.49375, |
| "completions/mean_terminated_length": 4.49375, |
| "completions/min_length": 3.7, |
| "completions/min_terminated_length": 3.7, |
| "entropy": 0.49938802123069764, |
| "epoch": 0.05905511811023622, |
| "frac_reward_zero_std": 0.3, |
| "grad_norm": 0.0, |
| "learning_rate": 4.8205e-06, |
| "loss": 0.01368042677640915, |
| "num_tokens": 106664.0, |
| "reward": 1.1010554283857346, |
| "reward_std": 0.045361624303768625, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.0576416015625, |
| "rewards/reward_length/std": 0.010658931022044271, |
| "rewards/reward_semantic/mean": 0.3464583337306976, |
| "rewards/reward_semantic/std": 0.15053959861397742, |
| "step": 360, |
| "step_time": 1.6211198363504082 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 5.8, |
| "completions/max_terminated_length": 5.8, |
| "completions/mean_length": 4.74375, |
| "completions/mean_terminated_length": 4.74375, |
| "completions/min_length": 3.6, |
| "completions/min_terminated_length": 3.6, |
| "entropy": 0.6371999262647023, |
| "epoch": 0.062335958005249346, |
| "frac_reward_zero_std": 0.3, |
| "grad_norm": 9.1875, |
| "learning_rate": 4.8105e-06, |
| "loss": 0.03233610987663269, |
| "num_tokens": 112511.0, |
| "reward": 1.1021533191204071, |
| "reward_std": 0.02930783515548683, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.05693359375, |
| "rewards/reward_length/std": 0.012558965157950297, |
| "rewards/reward_semantic/mean": 0.3500000026077032, |
| "rewards/reward_semantic/std": 0.09658735543489456, |
| "step": 380, |
| "step_time": 1.6564469950499188 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 6.05, |
| "completions/max_terminated_length": 6.05, |
| "completions/mean_length": 4.33125, |
| "completions/mean_terminated_length": 4.33125, |
| "completions/min_length": 3.35, |
| "completions/min_terminated_length": 3.35, |
| "entropy": 0.7747498400121913, |
| "epoch": 0.06561679790026247, |
| "frac_reward_zero_std": 0.15, |
| "grad_norm": 16.0, |
| "learning_rate": 4.8005e-06, |
| "loss": 0.02885372340679169, |
| "num_tokens": 118188.0, |
| "reward": 1.1014230579137803, |
| "reward_std": 0.054096709901932626, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.054931640625, |
| "rewards/reward_length/std": 0.01498162968782708, |
| "rewards/reward_semantic/mean": 0.3472321443259716, |
| "rewards/reward_semantic/std": 0.17982315719127656, |
| "step": 400, |
| "step_time": 1.6521750739995695 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 5.55, |
| "completions/max_terminated_length": 5.55, |
| "completions/mean_length": 4.25, |
| "completions/mean_terminated_length": 4.25, |
| "completions/min_length": 3.25, |
| "completions/min_terminated_length": 3.25, |
| "entropy": 0.6093234239519916, |
| "epoch": 0.0688976377952756, |
| "frac_reward_zero_std": 0.2, |
| "grad_norm": 8.1875, |
| "learning_rate": 4.7905e-06, |
| "loss": 0.07366046905517579, |
| "num_tokens": 123980.0, |
| "reward": 1.088520696759224, |
| "reward_std": 0.02127447940329148, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.0520263671875, |
| "rewards/reward_length/std": 0.010646102635655552, |
| "rewards/reward_semantic/mean": 0.30374008174985645, |
| "rewards/reward_semantic/std": 0.06995791383087635, |
| "step": 420, |
| "step_time": 1.598766735350273 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 5.6, |
| "completions/max_terminated_length": 5.6, |
| "completions/mean_length": 4.4625, |
| "completions/mean_terminated_length": 4.4625, |
| "completions/min_length": 3.4, |
| "completions/min_terminated_length": 3.4, |
| "entropy": 0.6000309210250634, |
| "epoch": 0.07217847769028872, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 6.25, |
| "learning_rate": 4.7805e-06, |
| "loss": 0.02538236975669861, |
| "num_tokens": 130310.0, |
| "reward": 1.1284031838178634, |
| "reward_std": 0.030537298787385224, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.0548828125, |
| "rewards/reward_length/std": 0.012618192343506963, |
| "rewards/reward_semantic/mean": 0.43715773820877074, |
| "rewards/reward_semantic/std": 0.10091984458267689, |
| "step": 440, |
| "step_time": 1.6581264014497719 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 6.75, |
| "completions/max_terminated_length": 6.75, |
| "completions/mean_length": 4.9, |
| "completions/mean_terminated_length": 4.9, |
| "completions/min_length": 3.45, |
| "completions/min_terminated_length": 3.45, |
| "entropy": 0.7827204793691636, |
| "epoch": 0.07545931758530183, |
| "frac_reward_zero_std": 0.15, |
| "grad_norm": 2.140625, |
| "learning_rate": 4.7705e-06, |
| "loss": 0.09432601332664489, |
| "num_tokens": 136206.0, |
| "reward": 1.061219373345375, |
| "reward_std": 0.04166477765902528, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.067041015625, |
| "rewards/reward_length/std": 0.016452802077401428, |
| "rewards/reward_semantic/mean": 0.21523809637874364, |
| "rewards/reward_semantic/std": 0.13769038021564484, |
| "step": 460, |
| "step_time": 1.763348830850282 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 5.95, |
| "completions/max_terminated_length": 5.95, |
| "completions/mean_length": 4.5125, |
| "completions/mean_terminated_length": 4.5125, |
| "completions/min_length": 3.3, |
| "completions/min_terminated_length": 3.3, |
| "entropy": 0.6544731822835275, |
| "epoch": 0.07874015748031496, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.0, |
| "learning_rate": 4.7605e-06, |
| "loss": 0.095479816198349, |
| "num_tokens": 142168.0, |
| "reward": 1.110293486714363, |
| "reward_std": 0.03730325770447962, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.0526123046875, |
| "rewards/reward_length/std": 0.013677800679579378, |
| "rewards/reward_semantic/mean": 0.37641369104385375, |
| "rewards/reward_semantic/std": 0.12245006412267685, |
| "step": 480, |
| "step_time": 1.693691986000158 |
| }, |
| { |
| "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, |
| "completions/clipped_ratio": 0.0, |
| "completions/max_length": 6.2, |
| "completions/max_terminated_length": 6.2, |
| "completions/mean_length": 4.5375, |
| "completions/mean_terminated_length": 4.5375, |
| "completions/min_length": 3.3, |
| "completions/min_terminated_length": 3.3, |
| "entropy": 0.7134705329648569, |
| "epoch": 0.08202099737532809, |
| "frac_reward_zero_std": 0.2, |
| "grad_norm": 0.0, |
| "learning_rate": 4.7505000000000005e-06, |
| "loss": 0.04971726834774017, |
| "num_tokens": 148190.0, |
| "reward": 1.059982505440712, |
| "reward_std": 0.029997232874302428, |
| "rewards/reward_grounded/mean": 1.0, |
| "rewards/reward_grounded/std": 0.0, |
| "rewards/reward_length/mean": -0.053564453125, |
| "rewards/reward_length/std": 0.012155310017988086, |
| "rewards/reward_semantic/mean": 0.20886905267834663, |
| "rewards/reward_semantic/std": 0.0987570121884346, |
| "step": 500, |
| "step_time": 1.7915035136496953 |
| } |
| ], |
| "logging_steps": 20, |
| "max_steps": 10000, |
| "num_input_tokens_seen": 148190, |
| "num_train_epochs": 2, |
| "save_steps": 500, |
| "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 |
| } |
|
|
