{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 0.25749346208006435, "eval_steps": 500, "global_step": 20, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.1611328125, "completions/max_length": 10637.0, "completions/mean_length": 1210.8232421875, "completions/min_length": 21.0, "depth_nums": 0.1611328125, "entropy": 6.012878332287073, "epoch": 0.012874673104003219, "grad_norm": 20.75629234313965, "learning_rate": 3e-06, "loss": -0.0, "num_tokens": 4379765.0, "reward": 1.2593750953674316, "reward_std": 0.6255239248275757, "rewards/accuracy_reward_with_llm/mean": 0.556640625, "rewards/accuracy_reward_with_llm/std": 0.4970242381095886, "rewards/format_reward/mean": 0.7021484375, "rewards/format_reward/std": 0.45753735303878784, "rewards/visual_gain_reward/mean": 0.0005859375232830644, "rewards/visual_gain_reward/std": 0.010814730077981949, "segmentation_nums": 0.15234375, "step": 1, "total_gen_nums": 0.3134765625 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.1845703125, "completions/max_length": 11014.0, "completions/mean_length": 1261.671875, "completions/min_length": 18.0, "depth_nums": 0.1767578125, "entropy": 5.868425816297531, "epoch": 0.025749346208006438, "grad_norm": 28.14173126220703, "learning_rate": 2.9987834972573546e-06, "loss": 0.0001, "num_tokens": 8647497.0, "reward": 1.1925780773162842, "reward_std": 0.6145570278167725, "rewards/accuracy_reward_with_llm/mean": 0.5205078125, "rewards/accuracy_reward_with_llm/std": 0.49982336163520813, "rewards/format_reward/mean": 0.671875, "rewards/format_reward/std": 0.4697600305080414, "rewards/visual_gain_reward/mean": 0.00019531250291038305, "rewards/visual_gain_reward/std": 0.0062500000931322575, "segmentation_nums": 0.142578125, "step": 2, "total_gen_nums": 0.3193359375 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.16796875, "completions/max_length": 8707.0, "completions/mean_length": 1188.65625, "completions/min_length": 18.0, "depth_nums": 0.2060546875, "entropy": 5.8414958491921425, "epoch": 0.03862401931200966, "grad_norm": 33.402503967285156, "learning_rate": 2.9951359622013153e-06, "loss": 0.0, "num_tokens": 12897429.0, "reward": 1.220117211341858, "reward_std": 0.6342688798904419, "rewards/accuracy_reward_with_llm/mean": 0.5419921875, "rewards/accuracy_reward_with_llm/std": 0.4984769821166992, "rewards/format_reward/mean": 0.677734375, "rewards/format_reward/std": 0.4675724506378174, "rewards/visual_gain_reward/mean": 0.0003906250058207661, "rewards/visual_gain_reward/std": 0.008834514766931534, "segmentation_nums": 0.119140625, "step": 3, "total_gen_nums": 0.3251953125 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.1435546875, "completions/max_length": 8885.0, "completions/mean_length": 1253.751953125, "completions/min_length": 24.0, "depth_nums": 0.21484375, "entropy": 5.834823831915855, "epoch": 0.051498692416012876, "grad_norm": 13.457406997680664, "learning_rate": 2.989063311147081e-06, "loss": -0.0, "num_tokens": 17253257.0, "reward": 1.2726562023162842, "reward_std": 0.6112420558929443, "rewards/accuracy_reward_with_llm/mean": 0.5498046875, "rewards/accuracy_reward_with_llm/std": 0.4977564215660095, "rewards/format_reward/mean": 0.72265625, "rewards/format_reward/std": 0.44790637493133545, "rewards/visual_gain_reward/mean": 0.00019531250291038305, "rewards/visual_gain_reward/std": 0.0062500000931322575, "segmentation_nums": 0.1337890625, "step": 4, "total_gen_nums": 0.3486328125 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.140625, "completions/max_length": 8690.0, "completions/mean_length": 1261.009765625, "completions/min_length": 16.0, "depth_nums": 0.2158203125, "entropy": 5.8213246911764145, "epoch": 0.06437336552001609, "grad_norm": 12.64630126953125, "learning_rate": 2.9805753939568693e-06, "loss": 0.0001, "num_tokens": 21637837.0, "reward": 1.2509765625, "reward_std": 0.6066232323646545, "rewards/accuracy_reward_with_llm/mean": 0.51953125, "rewards/accuracy_reward_with_llm/std": 0.49986252188682556, "rewards/format_reward/mean": 0.7314453125, "rewards/format_reward/std": 0.4434242844581604, "rewards/visual_gain_reward/mean": 0.0, "rewards/visual_gain_reward/std": 0.0, "segmentation_nums": 0.1279296875, "step": 5, "total_gen_nums": 0.34375 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.1494140625, "completions/max_length": 8820.0, "completions/mean_length": 1227.6279296875, "completions/min_length": 20.0, "depth_nums": 0.1904296875, "entropy": 5.848804991692305, "epoch": 0.07724803862401931, "grad_norm": 10.392542839050293, "learning_rate": 2.9696859780634016e-06, "loss": 0.0, "num_tokens": 25954420.0, "reward": 1.3287110328674316, "reward_std": 0.5927157998085022, "rewards/accuracy_reward_with_llm/mean": 0.5888671875, "rewards/accuracy_reward_with_llm/std": 0.49227967858314514, "rewards/format_reward/mean": 0.7392578125, "rewards/format_reward/std": 0.4392540454864502, "rewards/visual_gain_reward/mean": 0.0005859375232830644, "rewards/visual_gain_reward/std": 0.010814730077981949, "segmentation_nums": 0.1474609375, "step": 6, "total_gen_nums": 0.337890625 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.076171875, "completions/max_length": 8984.0, "completions/mean_length": 1092.6708984375, "completions/min_length": 22.0, "depth_nums": 0.162109375, "entropy": 5.889556594192982, "epoch": 0.09012271172802253, "grad_norm": 62.187469482421875, "learning_rate": 2.956412726139078e-06, "loss": 0.0, "num_tokens": 30148439.0, "reward": 1.4267578125, "reward_std": 0.5138623118400574, "rewards/accuracy_reward_with_llm/mean": 0.5888671875, "rewards/accuracy_reward_with_llm/std": 0.49227967858314514, "rewards/format_reward/mean": 0.837890625, "rewards/format_reward/std": 0.36873120069503784, "rewards/visual_gain_reward/mean": 0.0, "rewards/visual_gain_reward/std": 0.0, "segmentation_nums": 0.146484375, "step": 7, "total_gen_nums": 0.30859375 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.09375, "completions/max_length": 9044.0, "completions/mean_length": 1225.798828125, "completions/min_length": 26.0, "depth_nums": 0.21875, "entropy": 5.749105926603079, "epoch": 0.10299738483202575, "grad_norm": 4.391378402709961, "learning_rate": 2.9407771674470586e-06, "loss": 0.0, "num_tokens": 34471603.0, "reward": 1.3777344226837158, "reward_std": 0.5527065992355347, "rewards/accuracy_reward_with_llm/mean": 0.568359375, "rewards/accuracy_reward_with_llm/std": 0.49554696679115295, "rewards/format_reward/mean": 0.80859375, "rewards/format_reward/std": 0.3936002850532532, "rewards/visual_gain_reward/mean": 0.0007812500116415322, "rewards/visual_gain_reward/std": 0.012481658719480038, "segmentation_nums": 0.1279296875, "step": 8, "total_gen_nums": 0.3466796875 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.080078125, "completions/max_length": 8989.0, "completions/mean_length": 1115.748046875, "completions/min_length": 9.0, "depth_nums": 0.185546875, "entropy": 5.770555350929499, "epoch": 0.11587205793602896, "grad_norm": 5.756628036499023, "learning_rate": 2.922804662920718e-06, "loss": -0.0001, "num_tokens": 38669155.0, "reward": 1.4249999523162842, "reward_std": 0.5238240361213684, "rewards/accuracy_reward_with_llm/mean": 0.5947265625, "rewards/accuracy_reward_with_llm/std": 0.4911847710609436, "rewards/format_reward/mean": 0.830078125, "rewards/format_reward/std": 0.3757476806640625, "rewards/visual_gain_reward/mean": 0.00019531250291038305, "rewards/visual_gain_reward/std": 0.006250000558793545, "segmentation_nums": 0.134765625, "step": 9, "total_gen_nums": 0.3203125 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0810546875, "completions/max_length": 8784.0, "completions/mean_length": 1067.3955078125, "completions/min_length": 20.0, "depth_nums": 0.19921875, "entropy": 5.8339343555271626, "epoch": 0.12874673104003218, "grad_norm": 3.002086877822876, "learning_rate": 2.9025243640281224e-06, "loss": -0.0, "num_tokens": 42779380.0, "reward": 1.4367187023162842, "reward_std": 0.4874913692474365, "rewards/accuracy_reward_with_llm/mean": 0.5869140625, "rewards/accuracy_reward_with_llm/std": 0.4926286041736603, "rewards/format_reward/mean": 0.849609375, "rewards/format_reward/std": 0.35762855410575867, "rewards/visual_gain_reward/mean": 0.00019531250291038305, "rewards/visual_gain_reward/std": 0.0062500000931322575, "segmentation_nums": 0.109375, "step": 10, "total_gen_nums": 0.30859375 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0888671875, "completions/max_length": 5823.0, "completions/mean_length": 1022.0263671875, "completions/min_length": 3.0, "depth_nums": 0.169921875, "entropy": 5.852696340531111, "epoch": 0.1416214041440354, "grad_norm": 2.620856523513794, "learning_rate": 2.8799691654882364e-06, "loss": -0.0001, "num_tokens": 46905439.0, "reward": 1.451562523841858, "reward_std": 0.5207765102386475, "rewards/accuracy_reward_with_llm/mean": 0.611328125, "rewards/accuracy_reward_with_llm/std": 0.4876866936683655, "rewards/format_reward/mean": 0.83984375, "rewards/format_reward/std": 0.36693012714385986, "rewards/visual_gain_reward/mean": 0.0003906250058207661, "rewards/visual_gain_reward/std": 0.008834514766931534, "segmentation_nums": 0.1298828125, "step": 11, "total_gen_nums": 0.2998046875 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0556640625, "completions/max_length": 9018.0, "completions/mean_length": 1051.802734375, "completions/min_length": 33.0, "depth_nums": 0.18359375, "entropy": 5.979498442262411, "epoch": 0.15449607724803863, "grad_norm": 4.12182092666626, "learning_rate": 2.8551756519155732e-06, "loss": 0.0, "num_tokens": 51027613.0, "reward": 1.531640648841858, "reward_std": 0.4414806663990021, "rewards/accuracy_reward_with_llm/mean": 0.6396484375, "rewards/accuracy_reward_with_llm/std": 0.480336993932724, "rewards/format_reward/mean": 0.8916015625, "rewards/format_reward/std": 0.3110348582267761, "rewards/visual_gain_reward/mean": 0.0003906250058207661, "rewards/visual_gain_reward/std": 0.008834514766931534, "segmentation_nums": 0.1142578125, "step": 12, "total_gen_nums": 0.2978515625 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0302734375, "completions/max_length": 6003.0, "completions/mean_length": 804.115234375, "completions/min_length": 30.0, "depth_nums": 0.140625, "entropy": 6.25120921805501, "epoch": 0.16737075035204185, "grad_norm": 3.7815661430358887, "learning_rate": 2.8281840384798147e-06, "loss": -0.0, "num_tokens": 54883153.0, "reward": 1.576562523841858, "reward_std": 0.410561740398407, "rewards/accuracy_reward_with_llm/mean": 0.6474609375, "rewards/accuracy_reward_with_llm/std": 0.4779941439628601, "rewards/format_reward/mean": 0.9287109375, "rewards/format_reward/std": 0.2574327886104584, "rewards/visual_gain_reward/mean": 0.0003906250058207661, "rewards/visual_gain_reward/std": 0.008834514766931534, "segmentation_nums": 0.05859375, "step": 13, "total_gen_nums": 0.19921875 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0400390625, "completions/max_length": 8630.0, "completions/mean_length": 996.5263671875, "completions/min_length": 41.0, "depth_nums": 0.1728515625, "entropy": 6.108686868101358, "epoch": 0.18024542345604505, "grad_norm": 6.139822483062744, "learning_rate": 2.7990381056766585e-06, "loss": 0.0, "num_tokens": 58956744.0, "reward": 1.508398413658142, "reward_std": 0.45154306292533875, "rewards/accuracy_reward_with_llm/mean": 0.6005859375, "rewards/accuracy_reward_with_llm/std": 0.490017294883728, "rewards/format_reward/mean": 0.9072265625, "rewards/format_reward/std": 0.29025644063949585, "rewards/visual_gain_reward/mean": 0.0005859375232830644, "rewards/visual_gain_reward/std": 0.010814730077981949, "segmentation_nums": 0.0927734375, "step": 14, "total_gen_nums": 0.265625 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.029296875, "completions/max_length": 8819.0, "completions/mean_length": 1031.1953125, "completions/min_length": 29.0, "depth_nums": 0.1884765625, "entropy": 6.060374926775694, "epoch": 0.19312009656004828, "grad_norm": 7.209692001342773, "learning_rate": 2.7677851283156924e-06, "loss": -0.0, "num_tokens": 63076588.0, "reward": 1.534570336341858, "reward_std": 0.45015260577201843, "rewards/accuracy_reward_with_llm/mean": 0.6083984375, "rewards/accuracy_reward_with_llm/std": 0.4883468747138977, "rewards/format_reward/mean": 0.92578125, "rewards/format_reward/std": 0.26225462555885315, "rewards/visual_gain_reward/mean": 0.0003906250058207661, "rewards/visual_gain_reward/std": 0.008834514766931534, "segmentation_nums": 0.08203125, "step": 15, "total_gen_nums": 0.2705078125 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0380859375, "completions/max_length": 8755.0, "completions/mean_length": 924.4208984375, "completions/min_length": 36.0, "depth_nums": 0.1435546875, "entropy": 6.233527477830648, "epoch": 0.2059947696640515, "grad_norm": 4.891477108001709, "learning_rate": 2.7344757988404844e-06, "loss": 0.0, "num_tokens": 67082447.0, "reward": 1.54296875, "reward_std": 0.4315545856952667, "rewards/accuracy_reward_with_llm/mean": 0.62109375, "rewards/accuracy_reward_with_llm/std": 0.4853517711162567, "rewards/format_reward/mean": 0.921875, "rewards/format_reward/std": 0.26849931478500366, "rewards/visual_gain_reward/mean": 0.0, "rewards/visual_gain_reward/std": 0.0, "segmentation_nums": 0.0732421875, "step": 16, "total_gen_nums": 0.216796875 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0380859375, "completions/max_length": 6920.0, "completions/mean_length": 941.1455078125, "completions/min_length": 9.0, "depth_nums": 0.1611328125, "entropy": 6.030625708401203, "epoch": 0.21886944276805473, "grad_norm": 2.913377285003662, "learning_rate": 2.699164145105252e-06, "loss": -0.0, "num_tokens": 71067000.0, "reward": 1.5294921398162842, "reward_std": 0.3852996826171875, "rewards/accuracy_reward_with_llm/mean": 0.6005859375, "rewards/accuracy_reward_with_llm/std": 0.490017294883728, "rewards/format_reward/mean": 0.9287109375, "rewards/format_reward/std": 0.2574327886104584, "rewards/visual_gain_reward/mean": 0.00019531250291038305, "rewards/visual_gain_reward/std": 0.006250000558793545, "segmentation_nums": 0.0751953125, "step": 17, "total_gen_nums": 0.236328125 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0302734375, "completions/max_length": 8720.0, "completions/mean_length": 976.4716796875, "completions/min_length": 48.0, "depth_nums": 0.1728515625, "entropy": 6.0241236835718155, "epoch": 0.23174411587205793, "grad_norm": 1.8226603269577026, "learning_rate": 2.6619074427414816e-06, "loss": -0.0, "num_tokens": 75093701.0, "reward": 1.563867211341858, "reward_std": 0.3692563772201538, "rewards/accuracy_reward_with_llm/mean": 0.6259765625, "rewards/accuracy_reward_with_llm/std": 0.4841061532497406, "rewards/format_reward/mean": 0.9375, "rewards/format_reward/std": 0.242179736495018, "rewards/visual_gain_reward/mean": 0.0003906250058207661, "rewards/visual_gain_reward/std": 0.008834514766931534, "segmentation_nums": 0.08203125, "step": 18, "total_gen_nums": 0.2548828125 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.03125, "completions/max_length": 9009.0, "completions/mean_length": 965.1865234375, "completions/min_length": 41.0, "depth_nums": 0.1689453125, "entropy": 6.073840655386448, "epoch": 0.24461878897606115, "grad_norm": 2.7203147411346436, "learning_rate": 2.6227661222566517e-06, "loss": 0.0, "num_tokens": 79134800.0, "reward": 1.5759766101837158, "reward_std": 0.3747516870498657, "rewards/accuracy_reward_with_llm/mean": 0.6435546875, "rewards/accuracy_reward_with_llm/std": 0.4791829288005829, "rewards/format_reward/mean": 0.931640625, "rewards/format_reward/std": 0.25248491764068604, "rewards/visual_gain_reward/mean": 0.0007812500116415322, "rewards/visual_gain_reward/std": 0.012481658719480038, "segmentation_nums": 0.0693359375, "step": 19, "total_gen_nums": 0.23828125 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0244140625, "completions/max_length": 8861.0, "completions/mean_length": 1025.93359375, "completions/min_length": 42.0, "depth_nums": 0.1904296875, "entropy": 5.9863284938037395, "epoch": 0.25749346208006435, "grad_norm": 2.001307249069214, "learning_rate": 2.581803671015722e-06, "loss": 0.0001, "num_tokens": 83173602.0, "reward": 1.569726586341858, "reward_std": 0.38102471828460693, "rewards/accuracy_reward_with_llm/mean": 0.640625, "rewards/accuracy_reward_with_llm/std": 0.480051726102829, "rewards/format_reward/mean": 0.9287109375, "rewards/format_reward/std": 0.2574327886104584, "rewards/visual_gain_reward/mean": 0.0003906250058207661, "rewards/visual_gain_reward/std": 0.008834514766931534, "segmentation_nums": 0.0634765625, "step": 20, "total_gen_nums": 0.25390625 } ], "logging_steps": 1.0, "max_steps": 78, "num_input_tokens_seen": 83173602, "num_train_epochs": 1, "save_steps": 10, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": false }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 1, "trial_name": null, "trial_params": null }