| { |
| "best_global_step": null, |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 0.75, |
| "eval_steps": 500, |
| "global_step": 150, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.005, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1034354716539383, |
| "kl": 0.0, |
| "learning_rate": 0.0, |
| "loss": 0.0, |
| "num_tokens": 8624.0, |
| "reward": 0.8125, |
| "reward_std": 0.3720118999481201, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.3720119297504425, |
| "step": 1 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.01, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.0, |
| "kl": 0.0, |
| "learning_rate": 1.5000000000000002e-07, |
| "loss": 0.0, |
| "num_tokens": 17248.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 2 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.015, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1181466355919838, |
| "kl": 0.0010664464207366109, |
| "learning_rate": 3.0000000000000004e-07, |
| "loss": 0.0, |
| "num_tokens": 25872.0, |
| "reward": 0.7291666269302368, |
| "reward_std": 0.3666396141052246, |
| "rewards/compute_privacy_reward/mean": 0.7291666269302368, |
| "rewards/compute_privacy_reward/std": 0.3666396141052246, |
| "step": 3 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.02, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12828628718852997, |
| "kl": 0.000932232360355556, |
| "learning_rate": 4.5e-07, |
| "loss": 0.0, |
| "num_tokens": 34496.0, |
| "reward": 0.875, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 4 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.025, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.2313728630542755, |
| "kl": 0.001018319046124816, |
| "learning_rate": 6.000000000000001e-07, |
| "loss": 0.0, |
| "num_tokens": 43120.0, |
| "reward": 0.875, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 5 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.03, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1034809872508049, |
| "kl": 0.0008933571516536176, |
| "learning_rate": 7.5e-07, |
| "loss": 0.0, |
| "num_tokens": 51744.0, |
| "reward": 0.625, |
| "reward_std": 0.4432026147842407, |
| "rewards/compute_privacy_reward/mean": 0.625, |
| "rewards/compute_privacy_reward/std": 0.4432026445865631, |
| "step": 6 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.035, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10621684044599533, |
| "kl": 0.0009254220058210194, |
| "learning_rate": 9e-07, |
| "loss": 0.0, |
| "num_tokens": 60368.0, |
| "reward": 0.9791666269302368, |
| "reward_std": 0.058925580233335495, |
| "rewards/compute_privacy_reward/mean": 0.9791666269302368, |
| "rewards/compute_privacy_reward/std": 0.0589255727827549, |
| "step": 7 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.04, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.9167926438967697e-05, |
| "kl": 0.0009798445971682668, |
| "learning_rate": 1.05e-06, |
| "loss": 0.0, |
| "num_tokens": 68992.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 8 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.045, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.8048785932478495e-05, |
| "kl": 0.0009403201402164996, |
| "learning_rate": 1.2000000000000002e-06, |
| "loss": 0.0, |
| "num_tokens": 77616.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 9 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.05, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.09035244584083557, |
| "kl": 0.0009671517182141542, |
| "learning_rate": 1.35e-06, |
| "loss": 0.0, |
| "num_tokens": 86240.0, |
| "reward": 0.96875, |
| "reward_std": 0.0883883461356163, |
| "rewards/compute_privacy_reward/mean": 0.96875, |
| "rewards/compute_privacy_reward/std": 0.0883883461356163, |
| "step": 10 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.055, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.13590653240680695, |
| "kl": 0.001038418267853558, |
| "learning_rate": 1.5e-06, |
| "loss": 0.0, |
| "num_tokens": 94864.0, |
| "reward": 0.875, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 11 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.06, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1292438954114914, |
| "kl": 0.001032966305501759, |
| "learning_rate": 1.65e-06, |
| "loss": 0.0, |
| "num_tokens": 103488.0, |
| "reward": 0.84375, |
| "reward_std": 0.22903135418891907, |
| "rewards/compute_privacy_reward/mean": 0.84375, |
| "rewards/compute_privacy_reward/std": 0.22903135418891907, |
| "step": 12 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.065, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11210907250642776, |
| "kl": 0.001149487216025591, |
| "learning_rate": 1.8e-06, |
| "loss": 0.0, |
| "num_tokens": 112112.0, |
| "reward": 0.8125, |
| "reward_std": 0.2587745785713196, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.25877460837364197, |
| "step": 13 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.07, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.977427746169269e-05, |
| "kl": 0.0010270538041368127, |
| "learning_rate": 1.95e-06, |
| "loss": 0.0, |
| "num_tokens": 120736.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 14 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.075, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 4.218954563839361e-05, |
| "kl": 0.0009941478492692113, |
| "learning_rate": 2.1e-06, |
| "loss": 0.0, |
| "num_tokens": 129360.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 15 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.08, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12708093225955963, |
| "kl": 0.0014116306556388736, |
| "learning_rate": 2.25e-06, |
| "loss": 0.0, |
| "num_tokens": 137984.0, |
| "reward": 0.8125, |
| "reward_std": 0.3720118999481201, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.3720119297504425, |
| "step": 16 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.085, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10569458454847336, |
| "kl": 0.0008829380385577679, |
| "learning_rate": 2.4000000000000003e-06, |
| "loss": 0.0, |
| "num_tokens": 146608.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 17 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.09, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12139219045639038, |
| "kl": 0.0010356823913753033, |
| "learning_rate": 2.55e-06, |
| "loss": 0.0, |
| "num_tokens": 155232.0, |
| "reward": 0.9236111044883728, |
| "reward_std": 0.17552515864372253, |
| "rewards/compute_privacy_reward/mean": 0.9236111044883728, |
| "rewards/compute_privacy_reward/std": 0.17552517354488373, |
| "step": 18 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.095, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11737910658121109, |
| "kl": 0.0010564889525994658, |
| "learning_rate": 2.7e-06, |
| "loss": 0.0, |
| "num_tokens": 163856.0, |
| "reward": 0.8413461446762085, |
| "reward_std": 0.21408309042453766, |
| "rewards/compute_privacy_reward/mean": 0.8413461446762085, |
| "rewards/compute_privacy_reward/std": 0.21408309042453766, |
| "step": 19 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.1, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11085058748722076, |
| "kl": 0.0010880294721573591, |
| "learning_rate": 2.85e-06, |
| "loss": 0.0, |
| "num_tokens": 172480.0, |
| "reward": 0.8125, |
| "reward_std": 0.3720118999481201, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.3720119297504425, |
| "step": 20 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.105, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.16246397793293, |
| "kl": 0.0009519192390143871, |
| "learning_rate": 3e-06, |
| "loss": 0.0, |
| "num_tokens": 181104.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 21 |
| }, |
| { |
| "completion_length": 937.625, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 333.0, |
| "completions/mean_length": 937.625, |
| "completions/mean_terminated_length": 333.0, |
| "completions/min_length": 333.0, |
| "completions/min_terminated_length": 333.0, |
| "epoch": 0.11, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.20288348197937012, |
| "kl": 0.0010024518705904484, |
| "learning_rate": 2.999771542734587e-06, |
| "loss": 0.0, |
| "num_tokens": 189037.0, |
| "reward": 0.9895833730697632, |
| "reward_std": 0.029462767764925957, |
| "rewards/compute_privacy_reward/mean": 0.9895833730697632, |
| "rewards/compute_privacy_reward/std": 0.029462775215506554, |
| "step": 22 |
| }, |
| { |
| "completion_length": 1007.875, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 895.0, |
| "completions/mean_length": 1007.875, |
| "completions/mean_terminated_length": 895.0, |
| "completions/min_length": 895.0, |
| "completions/min_terminated_length": 895.0, |
| "epoch": 0.115, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.31133848428726196, |
| "kl": 0.001384518574923277, |
| "learning_rate": 2.9990862405286437e-06, |
| "loss": 0.0, |
| "num_tokens": 197532.0, |
| "reward": 0.8958333730697632, |
| "reward_std": 0.19795581698417664, |
| "rewards/compute_privacy_reward/mean": 0.8958333730697632, |
| "rewards/compute_privacy_reward/std": 0.19795581698417664, |
| "step": 23 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.12, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11477329581975937, |
| "kl": 0.0010207787854596972, |
| "learning_rate": 2.9979443021318607e-06, |
| "loss": 0.0, |
| "num_tokens": 206156.0, |
| "reward": 0.8125, |
| "reward_std": 0.2587745785713196, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.25877460837364197, |
| "step": 24 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.125, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.13319747149944305, |
| "kl": 0.0012949397787451744, |
| "learning_rate": 2.9963460753897363e-06, |
| "loss": 0.0, |
| "num_tokens": 214780.0, |
| "reward": 0.8125, |
| "reward_std": 0.3720118999481201, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.3720119297504425, |
| "step": 25 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.13, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1052631065249443, |
| "kl": 0.001011731568723917, |
| "learning_rate": 2.994292047137618e-06, |
| "loss": 0.0, |
| "num_tokens": 223404.0, |
| "reward": 0.8125, |
| "reward_std": 0.2587745785713196, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.25877460837364197, |
| "step": 26 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.135, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1082906424999237, |
| "kl": 0.0008408679277636111, |
| "learning_rate": 2.99178284305241e-06, |
| "loss": 0.0, |
| "num_tokens": 232028.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 27 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.14, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11048393696546555, |
| "kl": 0.0008934384095482528, |
| "learning_rate": 2.9888192274619833e-06, |
| "loss": 0.0, |
| "num_tokens": 240652.0, |
| "reward": 0.625, |
| "reward_std": 0.3535533845424652, |
| "rewards/compute_privacy_reward/mean": 0.625, |
| "rewards/compute_privacy_reward/std": 0.3535533845424652, |
| "step": 28 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.145, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11940355598926544, |
| "kl": 0.0011573235969990492, |
| "learning_rate": 2.9854021031123555e-06, |
| "loss": 0.0, |
| "num_tokens": 249276.0, |
| "reward": 0.6875, |
| "reward_std": 0.3720118999481201, |
| "rewards/compute_privacy_reward/mean": 0.6875, |
| "rewards/compute_privacy_reward/std": 0.3720119297504425, |
| "step": 29 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.15, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.0002610312949400395, |
| "kl": 0.0011443651746958494, |
| "learning_rate": 2.981532510892707e-06, |
| "loss": 0.0, |
| "num_tokens": 257900.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 30 |
| }, |
| { |
| "completion_length": 934.625, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 309.0, |
| "completions/mean_length": 934.625, |
| "completions/mean_terminated_length": 309.0, |
| "completions/min_length": 309.0, |
| "completions/min_terminated_length": 309.0, |
| "epoch": 0.155, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.13400329649448395, |
| "kl": 0.0009456340922042727, |
| "learning_rate": 2.9772116295183124e-06, |
| "loss": 0.0, |
| "num_tokens": 265809.0, |
| "reward": 0.7916666269302368, |
| "reward_std": 0.38575834035873413, |
| "rewards/compute_privacy_reward/mean": 0.7916666269302368, |
| "rewards/compute_privacy_reward/std": 0.3857583701610565, |
| "step": 31 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.16, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.15540945529937744, |
| "kl": 0.0010692230425775051, |
| "learning_rate": 2.972440775171496e-06, |
| "loss": 0.0, |
| "num_tokens": 274433.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 32 |
| }, |
| { |
| "completion_length": 1003.75, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 862.0, |
| "completions/mean_length": 1003.75, |
| "completions/mean_terminated_length": 862.0, |
| "completions/min_length": 862.0, |
| "completions/min_terminated_length": 862.0, |
| "epoch": 0.165, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.13121868669986725, |
| "kl": 0.0009638073388487101, |
| "learning_rate": 2.9672214011007086e-06, |
| "loss": 0.0, |
| "num_tokens": 282895.0, |
| "reward": 0.8125, |
| "reward_std": 0.3720118999481201, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.3720119297504425, |
| "step": 33 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.17, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.736206831992604e-05, |
| "kl": 0.0009622175712138414, |
| "learning_rate": 2.961555097177853e-06, |
| "loss": 0.0, |
| "num_tokens": 291519.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 34 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.175, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10691909492015839, |
| "kl": 0.0008742180070839822, |
| "learning_rate": 2.9554435894139947e-06, |
| "loss": 0.0, |
| "num_tokens": 300143.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 35 |
| }, |
| { |
| "completion_length": 996.875, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 807.0, |
| "completions/mean_length": 996.875, |
| "completions/mean_terminated_length": 807.0, |
| "completions/min_length": 807.0, |
| "completions/min_terminated_length": 807.0, |
| "epoch": 0.18, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10549346357584, |
| "kl": 0.0010430242400616407, |
| "learning_rate": 2.9488887394336023e-06, |
| "loss": 0.0, |
| "num_tokens": 308550.0, |
| "reward": 0.75, |
| "reward_std": 0.37796446681022644, |
| "rewards/compute_privacy_reward/mean": 0.75, |
| "rewards/compute_privacy_reward/std": 0.37796446681022644, |
| "step": 36 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.185, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.3038460910320282, |
| "kl": 0.0008555782260373235, |
| "learning_rate": 2.9418925439074784e-06, |
| "loss": 0.0, |
| "num_tokens": 317174.0, |
| "reward": 0.6875, |
| "reward_std": 0.3720118999481201, |
| "rewards/compute_privacy_reward/mean": 0.6875, |
| "rewards/compute_privacy_reward/std": 0.3720119297504425, |
| "step": 37 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.19, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.516529118816834e-05, |
| "kl": 0.0008900034008547664, |
| "learning_rate": 2.9344571339445534e-06, |
| "loss": 0.0, |
| "num_tokens": 325798.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 38 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.195, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 5.1143713790224865e-05, |
| "kl": 0.0010267372708767653, |
| "learning_rate": 2.9265847744427307e-06, |
| "loss": 0.0, |
| "num_tokens": 334422.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 39 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.2, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.13654102385044098, |
| "kl": 0.0011425098637118936, |
| "learning_rate": 2.9182778633989753e-06, |
| "loss": 0.0, |
| "num_tokens": 343046.0, |
| "reward": 0.875, |
| "reward_std": 0.3535533845424652, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.3535533845424652, |
| "step": 40 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.205, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 3.4272001357749104e-05, |
| "kl": 0.0010022853966802359, |
| "learning_rate": 2.9095389311788626e-06, |
| "loss": 0.0, |
| "num_tokens": 351670.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 41 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.21, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12202563881874084, |
| "kl": 0.001005432684905827, |
| "learning_rate": 2.9003706397458025e-06, |
| "loss": 0.0, |
| "num_tokens": 360294.0, |
| "reward": 0.7708333730697632, |
| "reward_std": 0.33258843421936035, |
| "rewards/compute_privacy_reward/mean": 0.7708333730697632, |
| "rewards/compute_privacy_reward/std": 0.33258846402168274, |
| "step": 42 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.215, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.5954943592078052e-05, |
| "kl": 0.0010314591927453876, |
| "learning_rate": 2.8907757818501814e-06, |
| "loss": 0.0, |
| "num_tokens": 368918.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 43 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.22, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.00014445575652644038, |
| "kl": 0.0012217889307066798, |
| "learning_rate": 2.880757280178661e-06, |
| "loss": 0.0, |
| "num_tokens": 377542.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 44 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.225, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10734281688928604, |
| "kl": 0.0009893818059936166, |
| "learning_rate": 2.8703181864639013e-06, |
| "loss": 0.0, |
| "num_tokens": 386166.0, |
| "reward": 0.84375, |
| "reward_std": 0.22903135418891907, |
| "rewards/compute_privacy_reward/mean": 0.84375, |
| "rewards/compute_privacy_reward/std": 0.22903135418891907, |
| "step": 45 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.23, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10895420610904694, |
| "kl": 0.0009800479747354984, |
| "learning_rate": 2.859461680554975e-06, |
| "loss": 0.0, |
| "num_tokens": 394790.0, |
| "reward": 0.6875, |
| "reward_std": 0.45806270837783813, |
| "rewards/compute_privacy_reward/mean": 0.6875, |
| "rewards/compute_privacy_reward/std": 0.45806270837783813, |
| "step": 46 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.235, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.0007260672282427549, |
| "kl": 0.00257990974932909, |
| "learning_rate": 2.8481910694487506e-06, |
| "loss": 0.0, |
| "num_tokens": 403414.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 47 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.24, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11148799955844879, |
| "kl": 0.0011035795323550701, |
| "learning_rate": 2.8365097862825516e-06, |
| "loss": 0.0, |
| "num_tokens": 412038.0, |
| "reward": 0.84375, |
| "reward_std": 0.29693374037742615, |
| "rewards/compute_privacy_reward/mean": 0.84375, |
| "rewards/compute_privacy_reward/std": 0.29693374037742615, |
| "step": 48 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.245, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.09957773983478546, |
| "kl": 0.0008715721196494997, |
| "learning_rate": 2.8244213892883906e-06, |
| "loss": 0.0, |
| "num_tokens": 420662.0, |
| "reward": 0.5625, |
| "reward_std": 0.3204349875450134, |
| "rewards/compute_privacy_reward/mean": 0.5625, |
| "rewards/compute_privacy_reward/std": 0.3204349875450134, |
| "step": 49 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.25, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 3.079672751482576e-05, |
| "kl": 0.0011631341185420752, |
| "learning_rate": 2.811929560709094e-06, |
| "loss": 0.0, |
| "num_tokens": 429286.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 50 |
| }, |
| { |
| "completion_length": 1014.375, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 947.0, |
| "completions/mean_length": 1014.375, |
| "completions/mean_terminated_length": 947.0, |
| "completions/min_length": 947.0, |
| "completions/min_terminated_length": 947.0, |
| "epoch": 0.255, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.08795179426670074, |
| "kl": 0.0008526367018930614, |
| "learning_rate": 2.7990381056766585e-06, |
| "loss": 0.0, |
| "num_tokens": 437833.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 51 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.26, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12447066605091095, |
| "kl": 0.0010829587699845433, |
| "learning_rate": 2.7857509510531684e-06, |
| "loss": 0.0, |
| "num_tokens": 446457.0, |
| "reward": 0.875, |
| "reward_std": 0.3535533845424652, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.3535533845424652, |
| "step": 52 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.265, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.13497750461101532, |
| "kl": 0.0009853476658463478, |
| "learning_rate": 2.772072144234639e-06, |
| "loss": 0.0, |
| "num_tokens": 455081.0, |
| "reward": 0.8125, |
| "reward_std": 0.2587745785713196, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.25877460837364197, |
| "step": 53 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.27, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1315947026014328, |
| "kl": 0.0009250645525753498, |
| "learning_rate": 2.758005851918136e-06, |
| "loss": 0.0, |
| "num_tokens": 463705.0, |
| "reward": 0.75, |
| "reward_std": 0.37796446681022644, |
| "rewards/compute_privacy_reward/mean": 0.75, |
| "rewards/compute_privacy_reward/std": 0.37796446681022644, |
| "step": 54 |
| }, |
| { |
| "completion_length": 994.75, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 790.0, |
| "completions/mean_length": 994.75, |
| "completions/mean_terminated_length": 790.0, |
| "completions/min_length": 790.0, |
| "completions/min_terminated_length": 790.0, |
| "epoch": 0.275, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.7504169591120444e-05, |
| "kl": 0.0009734573541209102, |
| "learning_rate": 2.7435563588325624e-06, |
| "loss": 0.0, |
| "num_tokens": 472095.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 55 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.28, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.6432204574812204e-05, |
| "kl": 0.000817618565633893, |
| "learning_rate": 2.728728066433488e-06, |
| "loss": 0.0, |
| "num_tokens": 480719.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 56 |
| }, |
| { |
| "completion_length": 906.5, |
| "completions/clipped_ratio": 0.75, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 922.0, |
| "completions/mean_length": 906.5, |
| "completions/mean_terminated_length": 554.0, |
| "completions/min_length": 186.0, |
| "completions/min_terminated_length": 186.0, |
| "epoch": 0.285, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.2514869272708893, |
| "kl": 0.015548394061625004, |
| "learning_rate": 2.713525491562421e-06, |
| "loss": 0.0, |
| "num_tokens": 488403.0, |
| "reward": 0.8125, |
| "reward_std": 0.2587745785713196, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.25877460837364197, |
| "step": 57 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.29, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.9795588741544634e-05, |
| "kl": 0.0008735397132113576, |
| "learning_rate": 2.6979532650709395e-06, |
| "loss": 0.0, |
| "num_tokens": 497027.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 58 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.295, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 2.2923424694454297e-05, |
| "kl": 0.0009889747016131878, |
| "learning_rate": 2.6820161304100827e-06, |
| "loss": 0.0, |
| "num_tokens": 505651.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 59 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.3, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11676385253667831, |
| "kl": 0.0010271839564666152, |
| "learning_rate": 2.6657189421854562e-06, |
| "loss": 0.0, |
| "num_tokens": 514275.0, |
| "reward": 0.8125, |
| "reward_std": 0.2587745785713196, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.25877460837364197, |
| "step": 60 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.305, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.0959247425198555, |
| "kl": 0.0009773718193173409, |
| "learning_rate": 2.649066664678467e-06, |
| "loss": 0.0, |
| "num_tokens": 522899.0, |
| "reward": 0.875, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 61 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.31, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.14202216267585754, |
| "kl": 0.0009774481877684593, |
| "learning_rate": 2.632064370334158e-06, |
| "loss": 0.0, |
| "num_tokens": 531523.0, |
| "reward": 0.9166666269302368, |
| "reward_std": 0.17817416787147522, |
| "rewards/compute_privacy_reward/mean": 0.9166666269302368, |
| "rewards/compute_privacy_reward/std": 0.17817416787147522, |
| "step": 62 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.315, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.9618792066466995e-05, |
| "kl": 0.0009177947649732232, |
| "learning_rate": 2.6147172382160914e-06, |
| "loss": 0.0, |
| "num_tokens": 540147.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 63 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.32, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.811989386624191e-05, |
| "kl": 0.0009169476688839495, |
| "learning_rate": 2.597030552428756e-06, |
| "loss": 0.0, |
| "num_tokens": 548771.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 64 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.325, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11086320877075195, |
| "kl": 0.001360249938443303, |
| "learning_rate": 2.5790097005079765e-06, |
| "loss": 0.0, |
| "num_tokens": 557395.0, |
| "reward": 0.625, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.625, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 65 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.33, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.7990078049479052e-05, |
| "kl": 0.0007709726342000067, |
| "learning_rate": 2.5606601717798212e-06, |
| "loss": 0.0, |
| "num_tokens": 566019.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 66 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.335, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 5.119280831422657e-05, |
| "kl": 0.001086847041733563, |
| "learning_rate": 2.5419875556884957e-06, |
| "loss": 0.0, |
| "num_tokens": 574643.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 67 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.34, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.09366317093372345, |
| "kl": 0.0008181668817996979, |
| "learning_rate": 2.522997540093748e-06, |
| "loss": 0.0, |
| "num_tokens": 583267.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 68 |
| }, |
| { |
| "completion_length": 950.125, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 433.0, |
| "completions/mean_length": 950.125, |
| "completions/mean_terminated_length": 433.0, |
| "completions/min_length": 433.0, |
| "completions/min_terminated_length": 433.0, |
| "epoch": 0.345, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.13151833415031433, |
| "kl": 0.0010982896201312542, |
| "learning_rate": 2.5036959095382875e-06, |
| "loss": 0.0, |
| "num_tokens": 591300.0, |
| "reward": 0.8125, |
| "reward_std": 0.2587745785713196, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.25877460837364197, |
| "step": 69 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.35, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.09004732221364975, |
| "kl": 0.0011112384963780642, |
| "learning_rate": 2.484088543485761e-06, |
| "loss": 0.0, |
| "num_tokens": 599924.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 70 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.355, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 98.65550994873047, |
| "kl": 17.318056106567383, |
| "learning_rate": 2.464181414529809e-06, |
| "loss": 0.0173, |
| "num_tokens": 608548.0, |
| "reward": 0.8125, |
| "reward_std": 0.2587745785713196, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.25877460837364197, |
| "step": 71 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.36, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1246163472533226, |
| "kl": 0.0010302275186404586, |
| "learning_rate": 2.4439805865747562e-06, |
| "loss": 0.0, |
| "num_tokens": 617172.0, |
| "reward": 0.875, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 72 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.365, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12120372802019119, |
| "kl": 0.001782815670594573, |
| "learning_rate": 2.4234922129884873e-06, |
| "loss": 0.0, |
| "num_tokens": 625796.0, |
| "reward": 0.75, |
| "reward_std": 0.37796446681022644, |
| "rewards/compute_privacy_reward/mean": 0.75, |
| "rewards/compute_privacy_reward/std": 0.37796446681022644, |
| "step": 73 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.37, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 3.0413952117669396e-05, |
| "kl": 0.000998888281174004, |
| "learning_rate": 2.4027225347280728e-06, |
| "loss": 0.0, |
| "num_tokens": 634420.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 74 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.375, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10706434398889542, |
| "kl": 0.0008194810943678021, |
| "learning_rate": 2.3816778784387097e-06, |
| "loss": 0.0, |
| "num_tokens": 643044.0, |
| "reward": 0.8046875, |
| "reward_std": 0.2702460289001465, |
| "rewards/compute_privacy_reward/mean": 0.8046875, |
| "rewards/compute_privacy_reward/std": 0.27024605870246887, |
| "step": 75 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.38, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1093151718378067, |
| "kl": 0.0007514241151511669, |
| "learning_rate": 2.3603646545265692e-06, |
| "loss": 0.0, |
| "num_tokens": 651668.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 76 |
| }, |
| { |
| "completion_length": 997.5, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 812.0, |
| "completions/mean_length": 997.5, |
| "completions/mean_terminated_length": 812.0, |
| "completions/min_length": 812.0, |
| "completions/min_terminated_length": 812.0, |
| "epoch": 0.385, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12932655215263367, |
| "kl": 0.0010467175161466002, |
| "learning_rate": 2.3387893552061204e-06, |
| "loss": 0.0, |
| "num_tokens": 660080.0, |
| "reward": 0.875, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 77 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.39, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 3.121865665889345e-05, |
| "kl": 0.000994647154584527, |
| "learning_rate": 2.316958552522541e-06, |
| "loss": 0.0, |
| "num_tokens": 668704.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 78 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.395, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11221697181463242, |
| "kl": 0.0014477845979854465, |
| "learning_rate": 2.2948788963498076e-06, |
| "loss": 0.0, |
| "num_tokens": 677328.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 79 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.4, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11178349703550339, |
| "kl": 0.0011074937647208571, |
| "learning_rate": 2.2725571123650813e-06, |
| "loss": 0.0, |
| "num_tokens": 685952.0, |
| "reward": 0.875, |
| "reward_std": 0.3535533845424652, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.3535533845424652, |
| "step": 80 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.405, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1577489674091339, |
| "kl": 0.0012230502907186747, |
| "learning_rate": 2.25e-06, |
| "loss": 0.0, |
| "num_tokens": 694576.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 81 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.41, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 2.0535368093987927e-05, |
| "kl": 0.0010358804138377309, |
| "learning_rate": 2.227214430369506e-06, |
| "loss": 0.0, |
| "num_tokens": 703200.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 82 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.415, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 2.3772499844199046e-05, |
| "kl": 0.0010962329106405377, |
| "learning_rate": 2.204207344178836e-06, |
| "loss": 0.0, |
| "num_tokens": 711824.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 83 |
| }, |
| { |
| "completion_length": 948.625, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 421.0, |
| "completions/mean_length": 948.625, |
| "completions/mean_terminated_length": 421.0, |
| "completions/min_length": 421.0, |
| "completions/min_terminated_length": 421.0, |
| "epoch": 0.42, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11506448686122894, |
| "kl": 0.0010746553307399154, |
| "learning_rate": 2.18098574960932e-06, |
| "loss": 0.0, |
| "num_tokens": 719845.0, |
| "reward": 0.8671875, |
| "reward_std": 0.19604772329330444, |
| "rewards/compute_privacy_reward/mean": 0.8671875, |
| "rewards/compute_privacy_reward/std": 0.19604772329330444, |
| "step": 84 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.425, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11664219200611115, |
| "kl": 0.0009337576339021325, |
| "learning_rate": 2.157556720183616e-06, |
| "loss": 0.0, |
| "num_tokens": 728469.0, |
| "reward": 0.875, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 85 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.43, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10617554187774658, |
| "kl": 0.0008689322858117521, |
| "learning_rate": 2.1339273926110494e-06, |
| "loss": 0.0, |
| "num_tokens": 737093.0, |
| "reward": 0.8333333730697632, |
| "reward_std": 0.35634833574295044, |
| "rewards/compute_privacy_reward/mean": 0.8333333730697632, |
| "rewards/compute_privacy_reward/std": 0.3563483655452728, |
| "step": 86 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.435, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.733417229843326e-05, |
| "kl": 0.0008813230087980628, |
| "learning_rate": 2.1101049646137005e-06, |
| "loss": 0.0, |
| "num_tokens": 745717.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 87 |
| }, |
| { |
| "completion_length": 928.875, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 263.0, |
| "completions/mean_length": 928.875, |
| "completions/mean_terminated_length": 263.0, |
| "completions/min_length": 263.0, |
| "completions/min_terminated_length": 263.0, |
| "epoch": 0.44, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.614055871963501, |
| "kl": 0.10425230860710144, |
| "learning_rate": 2.0860966927339105e-06, |
| "loss": 0.0001, |
| "num_tokens": 753580.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 88 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.445, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 3.0654395231977105e-05, |
| "kl": 0.0012105577625334263, |
| "learning_rate": 2.061909890123868e-06, |
| "loss": 0.0, |
| "num_tokens": 762204.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 89 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.45, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.00014825160906184465, |
| "kl": 0.0009850619826465845, |
| "learning_rate": 2.03755192431795e-06, |
| "loss": 0.0, |
| "num_tokens": 770828.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 90 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.455, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1180417612195015, |
| "kl": 0.000987752340734005, |
| "learning_rate": 2.0130302149885033e-06, |
| "loss": 0.0, |
| "num_tokens": 779452.0, |
| "reward": 0.6875, |
| "reward_std": 0.3720118999481201, |
| "rewards/compute_privacy_reward/mean": 0.6875, |
| "rewards/compute_privacy_reward/std": 0.3720119297504425, |
| "step": 91 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.46, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11703640222549438, |
| "kl": 0.0011730747064575553, |
| "learning_rate": 1.988352231685735e-06, |
| "loss": 0.0, |
| "num_tokens": 788076.0, |
| "reward": 0.90625, |
| "reward_std": 0.18600594997406006, |
| "rewards/compute_privacy_reward/mean": 0.90625, |
| "rewards/compute_privacy_reward/std": 0.18600596487522125, |
| "step": 92 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.465, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 2.518199835321866e-05, |
| "kl": 0.0010467178653925657, |
| "learning_rate": 1.963525491562421e-06, |
| "loss": 0.0, |
| "num_tokens": 796700.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 93 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.47, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.19600749015808105, |
| "kl": 0.0008916480001062155, |
| "learning_rate": 1.9385575570841053e-06, |
| "loss": 0.0, |
| "num_tokens": 805324.0, |
| "reward": 0.875, |
| "reward_std": 0.3535533845424652, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.3535533845424652, |
| "step": 94 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.475, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.09082144498825073, |
| "kl": 0.0008060715044848621, |
| "learning_rate": 1.9134560337254986e-06, |
| "loss": 0.0, |
| "num_tokens": 813948.0, |
| "reward": 0.5625, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.5625, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 95 |
| }, |
| { |
| "completion_length": 950.875, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 439.0, |
| "completions/mean_length": 950.875, |
| "completions/mean_terminated_length": 439.0, |
| "completions/min_length": 439.0, |
| "completions/min_terminated_length": 439.0, |
| "epoch": 0.48, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1263413280248642, |
| "kl": 0.0013786425115540624, |
| "learning_rate": 1.888228567653781e-06, |
| "loss": 0.0, |
| "num_tokens": 821987.0, |
| "reward": 0.6041666269302368, |
| "reward_std": 0.19795581698417664, |
| "rewards/compute_privacy_reward/mean": 0.6041666269302368, |
| "rewards/compute_privacy_reward/std": 0.19795581698417664, |
| "step": 96 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.485, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10073529183864594, |
| "kl": 0.0008537114481441677, |
| "learning_rate": 1.8628828433995015e-06, |
| "loss": 0.0, |
| "num_tokens": 830611.0, |
| "reward": 0.8125, |
| "reward_std": 0.2587745785713196, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.25877460837364197, |
| "step": 97 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.49, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11044912040233612, |
| "kl": 0.00106201006565243, |
| "learning_rate": 1.8374265815157977e-06, |
| "loss": 0.0, |
| "num_tokens": 839235.0, |
| "reward": 0.875, |
| "reward_std": 0.2314550131559372, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.2314550131559372, |
| "step": 98 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.495, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.096685990691185, |
| "kl": 0.0009302418329752982, |
| "learning_rate": 1.8118675362266389e-06, |
| "loss": 0.0, |
| "num_tokens": 847859.0, |
| "reward": 0.75, |
| "reward_std": 0.26726123690605164, |
| "rewards/compute_privacy_reward/mean": 0.75, |
| "rewards/compute_privacy_reward/std": 0.26726123690605164, |
| "step": 99 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.5, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10617872327566147, |
| "kl": 0.0009280540398322046, |
| "learning_rate": 1.7862134930648174e-06, |
| "loss": 0.0, |
| "num_tokens": 856483.0, |
| "reward": 0.875, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 100 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.505, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1047428548336029, |
| "kl": 0.0010330090299248695, |
| "learning_rate": 1.7604722665003958e-06, |
| "loss": 0.0, |
| "num_tokens": 865107.0, |
| "reward": 0.875, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 101 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.51, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11728937923908234, |
| "kl": 0.0014484968269243836, |
| "learning_rate": 1.7346516975603465e-06, |
| "loss": 0.0, |
| "num_tokens": 873731.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 102 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.515, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11015930026769638, |
| "kl": 0.0009305118583142757, |
| "learning_rate": 1.7087596514400981e-06, |
| "loss": 0.0, |
| "num_tokens": 882355.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 103 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.52, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1054692417383194, |
| "kl": 0.0008813097956590354, |
| "learning_rate": 1.6828040151077213e-06, |
| "loss": 0.0, |
| "num_tokens": 890979.0, |
| "reward": 0.75, |
| "reward_std": 0.37796446681022644, |
| "rewards/compute_privacy_reward/mean": 0.75, |
| "rewards/compute_privacy_reward/std": 0.37796446681022644, |
| "step": 104 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.525, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1155865490436554, |
| "kl": 0.0010020399931818247, |
| "learning_rate": 1.6567926949014804e-06, |
| "loss": 0.0, |
| "num_tokens": 899603.0, |
| "reward": 0.875, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 105 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.53, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12112700194120407, |
| "kl": 0.0015760967507958412, |
| "learning_rate": 1.6307336141214877e-06, |
| "loss": 0.0, |
| "num_tokens": 908227.0, |
| "reward": 0.8125, |
| "reward_std": 0.2587745785713196, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.25877460837364197, |
| "step": 106 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.535, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 2.4447414034511894e-05, |
| "kl": 0.0010565246921032667, |
| "learning_rate": 1.6046347106161879e-06, |
| "loss": 0.0, |
| "num_tokens": 916851.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 107 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.54, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1220158115029335, |
| "kl": 0.001237599179148674, |
| "learning_rate": 1.578503934364416e-06, |
| "loss": 0.0, |
| "num_tokens": 925475.0, |
| "reward": 0.90625, |
| "reward_std": 0.1293872892856598, |
| "rewards/compute_privacy_reward/mean": 0.90625, |
| "rewards/compute_privacy_reward/std": 0.12938730418682098, |
| "step": 108 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.545, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 2.833184589690063e-05, |
| "kl": 0.0011500224936753511, |
| "learning_rate": 1.5523492450537518e-06, |
| "loss": 0.0, |
| "num_tokens": 934099.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 109 |
| }, |
| { |
| "completion_length": 986.125, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 721.0, |
| "completions/mean_length": 986.125, |
| "completions/mean_terminated_length": 721.0, |
| "completions/min_length": 721.0, |
| "completions/min_terminated_length": 721.0, |
| "epoch": 0.55, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11168019473552704, |
| "kl": 0.0009486325434409082, |
| "learning_rate": 1.5261786096559255e-06, |
| "loss": 0.0, |
| "num_tokens": 942420.0, |
| "reward": 0.625, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.625, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 110 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.555, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1225506141781807, |
| "kl": 0.0009468000498600304, |
| "learning_rate": 1.5e-06, |
| "loss": 0.0, |
| "num_tokens": 951044.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 111 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.56, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11512339860200882, |
| "kl": 0.0008956408128142357, |
| "learning_rate": 1.4738213903440748e-06, |
| "loss": 0.0, |
| "num_tokens": 959668.0, |
| "reward": 0.8125, |
| "reward_std": 0.2587745785713196, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.25877460837364197, |
| "step": 112 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.565, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10893312841653824, |
| "kl": 0.0008773457375355065, |
| "learning_rate": 1.4476507549462489e-06, |
| "loss": 0.0, |
| "num_tokens": 968292.0, |
| "reward": 0.3125, |
| "reward_std": 0.45806270837783813, |
| "rewards/compute_privacy_reward/mean": 0.3125, |
| "rewards/compute_privacy_reward/std": 0.45806270837783813, |
| "step": 113 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.57, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.533588147140108e-05, |
| "kl": 0.0009089583763852715, |
| "learning_rate": 1.4214960656355842e-06, |
| "loss": 0.0, |
| "num_tokens": 976916.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 114 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.575, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.7000613297568634e-05, |
| "kl": 0.0008393891621381044, |
| "learning_rate": 1.395365289383812e-06, |
| "loss": 0.0, |
| "num_tokens": 985540.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 115 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.58, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.00031007660436443985, |
| "kl": 0.001272182329557836, |
| "learning_rate": 1.3692663858785126e-06, |
| "loss": 0.0, |
| "num_tokens": 994164.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 116 |
| }, |
| { |
| "completion_length": 947.0, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 408.0, |
| "completions/mean_length": 947.0, |
| "completions/mean_terminated_length": 408.0, |
| "completions/min_length": 408.0, |
| "completions/min_terminated_length": 408.0, |
| "epoch": 0.585, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10772430896759033, |
| "kl": 0.0011644808109849691, |
| "learning_rate": 1.3432073050985201e-06, |
| "loss": 0.0, |
| "num_tokens": 1002172.0, |
| "reward": 0.9895833730697632, |
| "reward_std": 0.029462767764925957, |
| "rewards/compute_privacy_reward/mean": 0.9895833730697632, |
| "rewards/compute_privacy_reward/std": 0.029462775215506554, |
| "step": 117 |
| }, |
| { |
| "completion_length": 1006.0, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 880.0, |
| "completions/mean_length": 1006.0, |
| "completions/mean_terminated_length": 880.0, |
| "completions/min_length": 880.0, |
| "completions/min_terminated_length": 880.0, |
| "epoch": 0.59, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10412620007991791, |
| "kl": 0.0007835485157556832, |
| "learning_rate": 1.317195984892279e-06, |
| "loss": 0.0, |
| "num_tokens": 1010652.0, |
| "reward": 0.8125, |
| "reward_std": 0.3720118999481201, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.3720119297504425, |
| "step": 118 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.595, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.20948749780654907, |
| "kl": 0.001019683200865984, |
| "learning_rate": 1.2912403485599022e-06, |
| "loss": 0.0, |
| "num_tokens": 1019276.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 119 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.6, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10544629395008087, |
| "kl": 0.0008191296947188675, |
| "learning_rate": 1.2653483024396534e-06, |
| "loss": 0.0, |
| "num_tokens": 1027900.0, |
| "reward": 0.9583333730697632, |
| "reward_std": 0.11785111576318741, |
| "rewards/compute_privacy_reward/mean": 0.9583333730697632, |
| "rewards/compute_privacy_reward/std": 0.117851123213768, |
| "step": 120 |
| }, |
| { |
| "completion_length": 956.0, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 480.0, |
| "completions/mean_length": 956.0, |
| "completions/mean_terminated_length": 480.0, |
| "completions/min_length": 480.0, |
| "completions/min_terminated_length": 480.0, |
| "epoch": 0.605, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12782183289527893, |
| "kl": 0.0013312390074133873, |
| "learning_rate": 1.2395277334996047e-06, |
| "loss": 0.0, |
| "num_tokens": 1035980.0, |
| "reward": 0.8125, |
| "reward_std": 0.2587745785713196, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.25877460837364197, |
| "step": 121 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.61, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10687465220689774, |
| "kl": 0.0010700662387534976, |
| "learning_rate": 1.2137865069351827e-06, |
| "loss": 0.0, |
| "num_tokens": 1044604.0, |
| "reward": 0.9583333730697632, |
| "reward_std": 0.11785111576318741, |
| "rewards/compute_privacy_reward/mean": 0.9583333730697632, |
| "rewards/compute_privacy_reward/std": 0.117851123213768, |
| "step": 122 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.615, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 2.7855163352796808e-05, |
| "kl": 0.0009706842829473317, |
| "learning_rate": 1.1881324637733612e-06, |
| "loss": 0.0, |
| "num_tokens": 1053228.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 123 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.62, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10759395360946655, |
| "kl": 0.0010831992840394378, |
| "learning_rate": 1.1625734184842028e-06, |
| "loss": 0.0, |
| "num_tokens": 1061852.0, |
| "reward": 0.8541666865348816, |
| "reward_std": 0.27368006110191345, |
| "rewards/compute_privacy_reward/mean": 0.8541666865348816, |
| "rewards/compute_privacy_reward/std": 0.27368009090423584, |
| "step": 124 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.625, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10416800528764725, |
| "kl": 0.0010248625185340643, |
| "learning_rate": 1.1371171566004986e-06, |
| "loss": 0.0, |
| "num_tokens": 1070476.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 125 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.63, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10788594186306, |
| "kl": 0.0013441079063341022, |
| "learning_rate": 1.1117714323462188e-06, |
| "loss": 0.0, |
| "num_tokens": 1079100.0, |
| "reward": 0.8125, |
| "reward_std": 0.2587745785713196, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.25877460837364197, |
| "step": 126 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.635, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.0995379239320755, |
| "kl": 0.0009179638000205159, |
| "learning_rate": 1.0865439662745013e-06, |
| "loss": 0.0, |
| "num_tokens": 1087724.0, |
| "reward": 0.875, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 127 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.64, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12063811719417572, |
| "kl": 0.0010087619302794337, |
| "learning_rate": 1.061442442915895e-06, |
| "loss": 0.0, |
| "num_tokens": 1096348.0, |
| "reward": 0.8125, |
| "reward_std": 0.3720118999481201, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.3720119297504425, |
| "step": 128 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.645, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.14395400881767273, |
| "kl": 0.0011259987950325012, |
| "learning_rate": 1.036474508437579e-06, |
| "loss": 0.0, |
| "num_tokens": 1104972.0, |
| "reward": 0.9479166269302368, |
| "reward_std": 0.1473139226436615, |
| "rewards/compute_privacy_reward/mean": 0.9479166269302368, |
| "rewards/compute_privacy_reward/std": 0.1473139226436615, |
| "step": 129 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.65, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11925813555717468, |
| "kl": 0.0011099607218056917, |
| "learning_rate": 1.0116477683142654e-06, |
| "loss": 0.0, |
| "num_tokens": 1113596.0, |
| "reward": 0.828125, |
| "reward_std": 0.35942351818084717, |
| "rewards/compute_privacy_reward/mean": 0.828125, |
| "rewards/compute_privacy_reward/std": 0.35942351818084717, |
| "step": 130 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.655, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11185282468795776, |
| "kl": 0.0010316207772120833, |
| "learning_rate": 9.86969785011497e-07, |
| "loss": 0.0, |
| "num_tokens": 1122220.0, |
| "reward": 0.75, |
| "reward_std": 0.26726123690605164, |
| "rewards/compute_privacy_reward/mean": 0.75, |
| "rewards/compute_privacy_reward/std": 0.26726123690605164, |
| "step": 131 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.66, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.2678123414516449, |
| "kl": 0.0014533624053001404, |
| "learning_rate": 9.624480756820497e-07, |
| "loss": 0.0, |
| "num_tokens": 1130844.0, |
| "reward": 0.9375, |
| "reward_std": 0.1767766922712326, |
| "rewards/compute_privacy_reward/mean": 0.9375, |
| "rewards/compute_privacy_reward/std": 0.1767766922712326, |
| "step": 132 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.665, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11791135370731354, |
| "kl": 0.0009261510567739606, |
| "learning_rate": 9.380901098761319e-07, |
| "loss": 0.0, |
| "num_tokens": 1139468.0, |
| "reward": 0.75, |
| "reward_std": 0.37796446681022644, |
| "rewards/compute_privacy_reward/mean": 0.75, |
| "rewards/compute_privacy_reward/std": 0.37796446681022644, |
| "step": 133 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.67, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.9650735339382663e-05, |
| "kl": 0.0009384769364260137, |
| "learning_rate": 9.139033072660897e-07, |
| "loss": 0.0, |
| "num_tokens": 1148092.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 134 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.675, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10626892745494843, |
| "kl": 0.0010822066105902195, |
| "learning_rate": 8.898950353863e-07, |
| "loss": 0.0, |
| "num_tokens": 1156716.0, |
| "reward": 0.875, |
| "reward_std": 0.3535533845424652, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.3535533845424652, |
| "step": 135 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.68, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.00028548898990266025, |
| "kl": 0.0012995110591873527, |
| "learning_rate": 8.660726073889511e-07, |
| "loss": 0.0, |
| "num_tokens": 1165340.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 136 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.685, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.12054099887609482, |
| "kl": 0.0010441445047035813, |
| "learning_rate": 8.424432798163837e-07, |
| "loss": 0.0, |
| "num_tokens": 1173964.0, |
| "reward": 0.6875, |
| "reward_std": 0.45806270837783813, |
| "rewards/compute_privacy_reward/mean": 0.6875, |
| "rewards/compute_privacy_reward/std": 0.45806270837783813, |
| "step": 137 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.69, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11257300525903702, |
| "kl": 0.0010643939021974802, |
| "learning_rate": 8.190142503906799e-07, |
| "loss": 0.0, |
| "num_tokens": 1182588.0, |
| "reward": 0.625, |
| "reward_std": 0.3535533845424652, |
| "rewards/compute_privacy_reward/mean": 0.625, |
| "rewards/compute_privacy_reward/std": 0.3535533845424652, |
| "step": 138 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.695, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10603896528482437, |
| "kl": 0.000935895717702806, |
| "learning_rate": 7.957926558211642e-07, |
| "loss": 0.0, |
| "num_tokens": 1191212.0, |
| "reward": 0.8958333134651184, |
| "reward_std": 0.294627845287323, |
| "rewards/compute_privacy_reward/mean": 0.8958333134651184, |
| "rewards/compute_privacy_reward/std": 0.294627845287323, |
| "step": 139 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.7, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11416691541671753, |
| "kl": 0.0010176518699154258, |
| "learning_rate": 7.727855696304945e-07, |
| "loss": 0.0, |
| "num_tokens": 1199836.0, |
| "reward": 0.8125, |
| "reward_std": 0.3720118999481201, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.3720119297504425, |
| "step": 140 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.705, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11740383505821228, |
| "kl": 0.00109928532037884, |
| "learning_rate": 7.500000000000003e-07, |
| "loss": 0.0, |
| "num_tokens": 1208460.0, |
| "reward": 0.625, |
| "reward_std": 0.3535533845424652, |
| "rewards/compute_privacy_reward/mean": 0.625, |
| "rewards/compute_privacy_reward/std": 0.3535533845424652, |
| "step": 141 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.71, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 0.001488029956817627, |
| "kl": 0.0010139566147699952, |
| "learning_rate": 7.274428876349186e-07, |
| "loss": 0.0, |
| "num_tokens": 1217084.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 142 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.715, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.6253690123558044, |
| "kl": 0.0013599335215985775, |
| "learning_rate": 7.051211036501928e-07, |
| "loss": 0.0, |
| "num_tokens": 1225708.0, |
| "reward": 0.5, |
| "reward_std": 0.37796446681022644, |
| "rewards/compute_privacy_reward/mean": 0.5, |
| "rewards/compute_privacy_reward/std": 0.37796446681022644, |
| "step": 143 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.72, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11948256939649582, |
| "kl": 0.0010317479027435184, |
| "learning_rate": 6.830414474774594e-07, |
| "loss": 0.0, |
| "num_tokens": 1234332.0, |
| "reward": 0.875, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 144 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.725, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10708877444267273, |
| "kl": 0.0009449583012610674, |
| "learning_rate": 6.6121064479388e-07, |
| "loss": 0.0, |
| "num_tokens": 1242956.0, |
| "reward": 0.9107142686843872, |
| "reward_std": 0.16967642307281494, |
| "rewards/compute_privacy_reward/mean": 0.9107142686843872, |
| "rewards/compute_privacy_reward/std": 0.16967640817165375, |
| "step": 145 |
| }, |
| { |
| "completion_length": 994.5, |
| "completions/clipped_ratio": 0.875, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 788.0, |
| "completions/mean_length": 994.5, |
| "completions/mean_terminated_length": 788.0, |
| "completions/min_length": 788.0, |
| "completions/min_terminated_length": 788.0, |
| "epoch": 0.73, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.1336914449930191, |
| "kl": 0.0012226947583258152, |
| "learning_rate": 6.396353454734313e-07, |
| "loss": 0.0, |
| "num_tokens": 1251344.0, |
| "reward": 0.875, |
| "reward_std": 0.2314550280570984, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.2314550280570984, |
| "step": 146 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.735, |
| "frac_reward_zero_std": 1.0, |
| "grad_norm": 1.4293345884652808e-05, |
| "kl": 0.0009187616524286568, |
| "learning_rate": 6.183221215612905e-07, |
| "loss": 0.0, |
| "num_tokens": 1259968.0, |
| "reward": 1.0, |
| "reward_std": 0.0, |
| "rewards/compute_privacy_reward/mean": 1.0, |
| "rewards/compute_privacy_reward/std": 0.0, |
| "step": 147 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.74, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.4137856066226959, |
| "kl": 0.0009601294295862317, |
| "learning_rate": 5.972774652719274e-07, |
| "loss": 0.0, |
| "num_tokens": 1268592.0, |
| "reward": 0.90625, |
| "reward_std": 0.18600594997406006, |
| "rewards/compute_privacy_reward/mean": 0.90625, |
| "rewards/compute_privacy_reward/std": 0.18600596487522125, |
| "step": 148 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.745, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.11015897244215012, |
| "kl": 0.0011308238608762622, |
| "learning_rate": 5.765077870115125e-07, |
| "loss": 0.0, |
| "num_tokens": 1277216.0, |
| "reward": 0.875, |
| "reward_std": 0.3535533845424652, |
| "rewards/compute_privacy_reward/mean": 0.875, |
| "rewards/compute_privacy_reward/std": 0.3535533845424652, |
| "step": 149 |
| }, |
| { |
| "completion_length": 1024.0, |
| "completions/clipped_ratio": 1.0, |
| "completions/max_length": 1024.0, |
| "completions/max_terminated_length": 0.0, |
| "completions/mean_length": 1024.0, |
| "completions/mean_terminated_length": 0.0, |
| "completions/min_length": 1024.0, |
| "completions/min_terminated_length": 0.0, |
| "epoch": 0.75, |
| "frac_reward_zero_std": 0.0, |
| "grad_norm": 0.10181056708097458, |
| "kl": 0.0010204908903688192, |
| "learning_rate": 5.560194134252441e-07, |
| "loss": 0.0, |
| "num_tokens": 1285840.0, |
| "reward": 0.8125, |
| "reward_std": 0.3720118999481201, |
| "rewards/compute_privacy_reward/mean": 0.8125, |
| "rewards/compute_privacy_reward/std": 0.3720119297504425, |
| "step": 150 |
| } |
| ], |
| "logging_steps": 1, |
| "max_steps": 200, |
| "num_input_tokens_seen": 1285840, |
| "num_train_epochs": 1, |
| "save_steps": 50, |
| "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": 8, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
