{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.28111830789548353,
  "eval_steps": 100,
  "global_step": 32100,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "eval_Final acc": 0.0,
      "eval_R1": 0.6413,
      "eval_R2": 0.3941,
      "eval_RL": 0.5744,
      "eval_RLsum": 0.5348,
      "eval_acc": 0.0,
      "eval_loss": 1.605442762374878,
      "eval_runtime": 1471.833,
      "eval_samples_per_second": 15.916,
      "eval_steps_per_second": 7.958,
      "step": 1
    },
    {
      "epoch": 0.0,
      "eval_Final acc": 0.0,
      "eval_R1": 0.7537,
      "eval_R2": 0.5609,
      "eval_RL": 0.729,
      "eval_RLsum": 0.7202,
      "eval_acc": 0.0,
      "eval_loss": 0.9451379179954529,
      "eval_runtime": 1465.7086,
      "eval_samples_per_second": 15.983,
      "eval_steps_per_second": 7.991,
      "step": 100
    },
    {
      "epoch": 0.0,
      "eval_Final acc": 0.7623580636899172,
      "eval_R1": 0.9493,
      "eval_R2": 0.8885,
      "eval_RL": 0.9427,
      "eval_RLsum": 0.9427,
      "eval_acc": 0.1172,
      "eval_loss": 0.15940502285957336,
      "eval_runtime": 1455.1674,
      "eval_samples_per_second": 16.098,
      "eval_steps_per_second": 8.049,
      "step": 200
    },
    {
      "epoch": 0.0,
      "eval_Final acc": 0.7699564586357039,
      "eval_R1": 0.9575,
      "eval_R2": 0.9086,
      "eval_RL": 0.9519,
      "eval_RLsum": 0.9519,
      "eval_acc": 0.2093,
      "eval_loss": 0.12018325924873352,
      "eval_runtime": 1457.2408,
      "eval_samples_per_second": 16.076,
      "eval_steps_per_second": 8.038,
      "step": 300
    },
    {
      "epoch": 0.0,
      "eval_Final acc": 0.7807564244856142,
      "eval_R1": 0.9642,
      "eval_R2": 0.9253,
      "eval_RL": 0.9599,
      "eval_RLsum": 0.9599,
      "eval_acc": 0.2186,
      "eval_loss": 0.0955982431769371,
      "eval_runtime": 1457.0581,
      "eval_samples_per_second": 16.078,
      "eval_steps_per_second": 8.039,
      "step": 400
    },
    {
      "epoch": 0.0,
      "grad_norm": 4.5625,
      "learning_rate": 8.333333333333334e-06,
      "loss": 0.4187,
      "step": 500
    },
    {
      "epoch": 0.0,
      "eval_Final acc": 0.792837018697174,
      "eval_R1": 0.9681,
      "eval_R2": 0.9354,
      "eval_RL": 0.9648,
      "eval_RLsum": 0.9648,
      "eval_acc": 0.2321,
      "eval_loss": 0.09067051112651825,
      "eval_runtime": 1455.1044,
      "eval_samples_per_second": 16.099,
      "eval_steps_per_second": 8.05,
      "step": 500
    },
    {
      "epoch": 0.01,
      "eval_Final acc": 0.783744557329463,
      "eval_R1": 0.9717,
      "eval_R2": 0.9407,
      "eval_RL": 0.9684,
      "eval_RLsum": 0.9684,
      "eval_acc": 0.3027,
      "eval_loss": 0.0751085877418518,
      "eval_runtime": 1457.3186,
      "eval_samples_per_second": 16.075,
      "eval_steps_per_second": 8.037,
      "step": 600
    },
    {
      "epoch": 0.01,
      "eval_Final acc": 0.7924101425766242,
      "eval_R1": 0.9729,
      "eval_R2": 0.9436,
      "eval_RL": 0.9698,
      "eval_RLsum": 0.9698,
      "eval_acc": 0.2989,
      "eval_loss": 0.07116048038005829,
      "eval_runtime": 1458.0436,
      "eval_samples_per_second": 16.067,
      "eval_steps_per_second": 8.033,
      "step": 700
    },
    {
      "epoch": 0.01,
      "eval_Final acc": 0.7832323059848032,
      "eval_R1": 0.9722,
      "eval_R2": 0.9426,
      "eval_RL": 0.969,
      "eval_RLsum": 0.969,
      "eval_acc": 0.3228,
      "eval_loss": 0.0740852877497673,
      "eval_runtime": 1456.9463,
      "eval_samples_per_second": 16.079,
      "eval_steps_per_second": 8.039,
      "step": 800
    },
    {
      "epoch": 0.01,
      "eval_Final acc": 0.7784512934346453,
      "eval_R1": 0.9773,
      "eval_R2": 0.9525,
      "eval_RL": 0.9739,
      "eval_RLsum": 0.974,
      "eval_acc": 0.3635,
      "eval_loss": 0.059417903423309326,
      "eval_runtime": 1456.8308,
      "eval_samples_per_second": 16.08,
      "eval_steps_per_second": 8.04,
      "step": 900
    },
    {
      "epoch": 0.01,
      "grad_norm": 1.6796875,
      "learning_rate": 1.6666666666666667e-05,
      "loss": 0.0688,
      "step": 1000
    },
    {
      "epoch": 0.01,
      "eval_Final acc": 0.797618031247332,
      "eval_R1": 0.977,
      "eval_R2": 0.9548,
      "eval_RL": 0.9753,
      "eval_RLsum": 0.9754,
      "eval_acc": 0.3543,
      "eval_loss": 0.058632686734199524,
      "eval_runtime": 1454.9906,
      "eval_samples_per_second": 16.1,
      "eval_steps_per_second": 8.05,
      "step": 1000
    },
    {
      "epoch": 0.01,
      "eval_Final acc": 0.7904892000341501,
      "eval_R1": 0.9788,
      "eval_R2": 0.9563,
      "eval_RL": 0.9762,
      "eval_RLsum": 0.9762,
      "eval_acc": 0.3487,
      "eval_loss": 0.06547216325998306,
      "eval_runtime": 1455.5453,
      "eval_samples_per_second": 16.094,
      "eval_steps_per_second": 8.047,
      "step": 1100
    },
    {
      "epoch": 0.01,
      "eval_Final acc": 0.7834884316571331,
      "eval_R1": 0.9769,
      "eval_R2": 0.9535,
      "eval_RL": 0.9746,
      "eval_RLsum": 0.9746,
      "eval_acc": 0.3247,
      "eval_loss": 0.06337419152259827,
      "eval_runtime": 1455.4383,
      "eval_samples_per_second": 16.095,
      "eval_steps_per_second": 8.048,
      "step": 1200
    },
    {
      "epoch": 0.01,
      "eval_Final acc": 0.7833176812089132,
      "eval_R1": 0.9787,
      "eval_R2": 0.9561,
      "eval_RL": 0.9757,
      "eval_RLsum": 0.9757,
      "eval_acc": 0.3857,
      "eval_loss": 0.055473439395427704,
      "eval_runtime": 1455.1315,
      "eval_samples_per_second": 16.099,
      "eval_steps_per_second": 8.049,
      "step": 1300
    },
    {
      "epoch": 0.01,
      "eval_Final acc": 0.7908307009305899,
      "eval_R1": 0.9768,
      "eval_R2": 0.9542,
      "eval_RL": 0.9746,
      "eval_RLsum": 0.9746,
      "eval_acc": 0.3544,
      "eval_loss": 0.057770680636167526,
      "eval_runtime": 1453.8142,
      "eval_samples_per_second": 16.113,
      "eval_steps_per_second": 8.057,
      "step": 1400
    },
    {
      "epoch": 0.01,
      "grad_norm": 1.7578125,
      "learning_rate": 2.5e-05,
      "loss": 0.0573,
      "step": 1500
    },
    {
      "epoch": 0.01,
      "eval_Final acc": 0.7988132843848715,
      "eval_R1": 0.9785,
      "eval_R2": 0.9576,
      "eval_RL": 0.9762,
      "eval_RLsum": 0.9763,
      "eval_acc": 0.385,
      "eval_loss": 0.054414428770542145,
      "eval_runtime": 1454.4943,
      "eval_samples_per_second": 16.106,
      "eval_steps_per_second": 8.053,
      "step": 1500
    },
    {
      "epoch": 0.01,
      "eval_Final acc": 0.7952702125843081,
      "eval_R1": 0.9778,
      "eval_R2": 0.9535,
      "eval_RL": 0.9746,
      "eval_RLsum": 0.9746,
      "eval_acc": 0.3673,
      "eval_loss": 0.061701055616140366,
      "eval_runtime": 1453.2555,
      "eval_samples_per_second": 16.12,
      "eval_steps_per_second": 8.06,
      "step": 1600
    },
    {
      "epoch": 0.01,
      "eval_Final acc": 0.8124306326304107,
      "eval_R1": 0.9777,
      "eval_R2": 0.9563,
      "eval_RL": 0.9758,
      "eval_RLsum": 0.9758,
      "eval_acc": 0.3971,
      "eval_loss": 0.056059580296278,
      "eval_runtime": 1453.5256,
      "eval_samples_per_second": 16.117,
      "eval_steps_per_second": 8.058,
      "step": 1700
    },
    {
      "epoch": 0.02,
      "eval_Final acc": 0.8063689917186032,
      "eval_R1": 0.9781,
      "eval_R2": 0.9549,
      "eval_RL": 0.9759,
      "eval_RLsum": 0.9759,
      "eval_acc": 0.366,
      "eval_loss": 0.06471270322799683,
      "eval_runtime": 1452.9948,
      "eval_samples_per_second": 16.123,
      "eval_steps_per_second": 8.061,
      "step": 1800
    },
    {
      "epoch": 0.02,
      "eval_Final acc": 0.7845983095705626,
      "eval_R1": 0.9811,
      "eval_R2": 0.9621,
      "eval_RL": 0.9788,
      "eval_RLsum": 0.9788,
      "eval_acc": 0.4019,
      "eval_loss": 0.04981780797243118,
      "eval_runtime": 1454.5264,
      "eval_samples_per_second": 16.106,
      "eval_steps_per_second": 8.053,
      "step": 1900
    },
    {
      "epoch": 0.02,
      "grad_norm": 2.859375,
      "learning_rate": 3.3333333333333335e-05,
      "loss": 0.0525,
      "step": 2000
    },
    {
      "epoch": 0.02,
      "eval_Final acc": 0.7960812772133526,
      "eval_R1": 0.9807,
      "eval_R2": 0.9611,
      "eval_RL": 0.9783,
      "eval_RLsum": 0.9783,
      "eval_acc": 0.3725,
      "eval_loss": 0.05053915083408356,
      "eval_runtime": 1454.7495,
      "eval_samples_per_second": 16.103,
      "eval_steps_per_second": 8.052,
      "step": 2000
    },
    {
      "epoch": 0.02,
      "eval_Final acc": 0.7927516434730641,
      "eval_R1": 0.9802,
      "eval_R2": 0.96,
      "eval_RL": 0.9777,
      "eval_RLsum": 0.9778,
      "eval_acc": 0.3841,
      "eval_loss": 0.05365655571222305,
      "eval_runtime": 1452.1804,
      "eval_samples_per_second": 16.132,
      "eval_steps_per_second": 8.066,
      "step": 2100
    },
    {
      "epoch": 0.02,
      "eval_Final acc": 0.8036369845470844,
      "eval_R1": 0.9804,
      "eval_R2": 0.9612,
      "eval_RL": 0.9783,
      "eval_RLsum": 0.9783,
      "eval_acc": 0.4144,
      "eval_loss": 0.05007287114858627,
      "eval_runtime": 1455.5749,
      "eval_samples_per_second": 16.094,
      "eval_steps_per_second": 8.047,
      "step": 2200
    },
    {
      "epoch": 0.02,
      "eval_Final acc": 0.8013745411081704,
      "eval_R1": 0.9794,
      "eval_R2": 0.9596,
      "eval_RL": 0.9775,
      "eval_RLsum": 0.9775,
      "eval_acc": 0.3875,
      "eval_loss": 0.05094440281391144,
      "eval_runtime": 1453.0962,
      "eval_samples_per_second": 16.121,
      "eval_steps_per_second": 8.061,
      "step": 2300
    },
    {
      "epoch": 0.02,
      "eval_Final acc": 0.7896354477930505,
      "eval_R1": 0.9805,
      "eval_R2": 0.9617,
      "eval_RL": 0.9783,
      "eval_RLsum": 0.9783,
      "eval_acc": 0.4016,
      "eval_loss": 0.05133907496929169,
      "eval_runtime": 1454.3241,
      "eval_samples_per_second": 16.108,
      "eval_steps_per_second": 8.054,
      "step": 2400
    },
    {
      "epoch": 0.02,
      "grad_norm": 0.81640625,
      "learning_rate": 4.166666666666667e-05,
      "loss": 0.0483,
      "step": 2500
    },
    {
      "epoch": 0.02,
      "eval_Final acc": 0.802441731409545,
      "eval_R1": 0.9803,
      "eval_R2": 0.9602,
      "eval_RL": 0.9782,
      "eval_RLsum": 0.9782,
      "eval_acc": 0.3766,
      "eval_loss": 0.0638379454612732,
      "eval_runtime": 1456.4588,
      "eval_samples_per_second": 16.084,
      "eval_steps_per_second": 8.042,
      "step": 2500
    },
    {
      "epoch": 0.02,
      "eval_Final acc": 0.7980449073678818,
      "eval_R1": 0.9814,
      "eval_R2": 0.9632,
      "eval_RL": 0.9796,
      "eval_RLsum": 0.9796,
      "eval_acc": 0.4159,
      "eval_loss": 0.048462945967912674,
      "eval_runtime": 1454.6004,
      "eval_samples_per_second": 16.105,
      "eval_steps_per_second": 8.052,
      "step": 2600
    },
    {
      "epoch": 0.02,
      "eval_Final acc": 0.8023990437974899,
      "eval_R1": 0.9816,
      "eval_R2": 0.9629,
      "eval_RL": 0.9793,
      "eval_RLsum": 0.9793,
      "eval_acc": 0.4099,
      "eval_loss": 0.050914596766233444,
      "eval_runtime": 1453.4126,
      "eval_samples_per_second": 16.118,
      "eval_steps_per_second": 8.059,
      "step": 2700
    },
    {
      "epoch": 0.02,
      "eval_Final acc": 0.7963374028856826,
      "eval_R1": 0.9818,
      "eval_R2": 0.9629,
      "eval_RL": 0.9794,
      "eval_RLsum": 0.9794,
      "eval_acc": 0.4046,
      "eval_loss": 0.05119272321462631,
      "eval_runtime": 1452.8409,
      "eval_samples_per_second": 16.124,
      "eval_steps_per_second": 8.062,
      "step": 2800
    },
    {
      "epoch": 0.03,
      "eval_Final acc": 0.8194314010074276,
      "eval_R1": 0.9824,
      "eval_R2": 0.965,
      "eval_RL": 0.9807,
      "eval_RLsum": 0.9807,
      "eval_acc": 0.4058,
      "eval_loss": 0.04595864564180374,
      "eval_runtime": 1453.233,
      "eval_samples_per_second": 16.12,
      "eval_steps_per_second": 8.06,
      "step": 2900
    },
    {
      "epoch": 0.03,
      "grad_norm": 2.359375,
      "learning_rate": 5e-05,
      "loss": 0.0483,
      "step": 3000
    },
    {
      "epoch": 0.03,
      "eval_Final acc": 0.8083753094851874,
      "eval_R1": 0.981,
      "eval_R2": 0.9618,
      "eval_RL": 0.9791,
      "eval_RLsum": 0.9791,
      "eval_acc": 0.4029,
      "eval_loss": 0.050352487713098526,
      "eval_runtime": 1453.1272,
      "eval_samples_per_second": 16.121,
      "eval_steps_per_second": 8.061,
      "step": 3000
    },
    {
      "epoch": 0.03,
      "eval_Final acc": 0.7971057799026723,
      "eval_R1": 0.9808,
      "eval_R2": 0.9626,
      "eval_RL": 0.9791,
      "eval_RLsum": 0.9792,
      "eval_acc": 0.4255,
      "eval_loss": 0.04512492194771767,
      "eval_runtime": 1453.2885,
      "eval_samples_per_second": 16.119,
      "eval_steps_per_second": 8.06,
      "step": 3100
    },
    {
      "epoch": 0.03,
      "eval_Final acc": 0.8262187313241697,
      "eval_R1": 0.9824,
      "eval_R2": 0.9663,
      "eval_RL": 0.9813,
      "eval_RLsum": 0.9814,
      "eval_acc": 0.4177,
      "eval_loss": 0.04345691576600075,
      "eval_runtime": 1450.9307,
      "eval_samples_per_second": 16.145,
      "eval_steps_per_second": 8.073,
      "step": 3200
    },
    {
      "epoch": 0.03,
      "eval_Final acc": 0.8077776829164176,
      "eval_R1": 0.9828,
      "eval_R2": 0.9662,
      "eval_RL": 0.9812,
      "eval_RLsum": 0.9812,
      "eval_acc": 0.4478,
      "eval_loss": 0.04140510410070419,
      "eval_runtime": 1455.0171,
      "eval_samples_per_second": 16.1,
      "eval_steps_per_second": 8.05,
      "step": 3300
    },
    {
      "epoch": 0.03,
      "eval_Final acc": 0.8141808247246649,
      "eval_R1": 0.9834,
      "eval_R2": 0.9677,
      "eval_RL": 0.9821,
      "eval_RLsum": 0.9821,
      "eval_acc": 0.4464,
      "eval_loss": 0.04060707613825798,
      "eval_runtime": 1476.0047,
      "eval_samples_per_second": 15.871,
      "eval_steps_per_second": 7.936,
      "step": 3400
    },
    {
      "epoch": 0.03,
      "grad_norm": 1.5078125,
      "learning_rate": 4.997804822366226e-05,
      "loss": 0.0413,
      "step": 3500
    },
    {
      "epoch": 0.03,
      "eval_Final acc": 0.790702638094425,
      "eval_R1": 0.9824,
      "eval_R2": 0.965,
      "eval_RL": 0.9803,
      "eval_RLsum": 0.9803,
      "eval_acc": 0.432,
      "eval_loss": 0.04334761202335358,
      "eval_runtime": 1456.2692,
      "eval_samples_per_second": 16.086,
      "eval_steps_per_second": 8.043,
      "step": 3500
    },
    {
      "epoch": 0.03,
      "eval_Final acc": 0.8168701442841287,
      "eval_R1": 0.9835,
      "eval_R2": 0.9677,
      "eval_RL": 0.9821,
      "eval_RLsum": 0.9821,
      "eval_acc": 0.4265,
      "eval_loss": 0.040950775146484375,
      "eval_runtime": 1463.6764,
      "eval_samples_per_second": 16.005,
      "eval_steps_per_second": 8.002,
      "step": 3600
    },
    {
      "epoch": 0.03,
      "eval_Final acc": 0.8131990096474003,
      "eval_R1": 0.9819,
      "eval_R2": 0.9648,
      "eval_RL": 0.9804,
      "eval_RLsum": 0.9804,
      "eval_acc": 0.3961,
      "eval_loss": 0.042560186237096786,
      "eval_runtime": 1457.9476,
      "eval_samples_per_second": 16.068,
      "eval_steps_per_second": 8.034,
      "step": 3700
    },
    {
      "epoch": 0.03,
      "eval_Final acc": 0.8021856057372151,
      "eval_R1": 0.9835,
      "eval_R2": 0.9665,
      "eval_RL": 0.9816,
      "eval_RLsum": 0.9816,
      "eval_acc": 0.4433,
      "eval_loss": 0.040598947554826736,
      "eval_runtime": 1455.5227,
      "eval_samples_per_second": 16.095,
      "eval_steps_per_second": 8.047,
      "step": 3800
    },
    {
      "epoch": 0.03,
      "eval_Final acc": 0.8159310168189191,
      "eval_R1": 0.9849,
      "eval_R2": 0.9689,
      "eval_RL": 0.9828,
      "eval_RLsum": 0.9828,
      "eval_acc": 0.4842,
      "eval_loss": 0.039128851145505905,
      "eval_runtime": 1454.1229,
      "eval_samples_per_second": 16.11,
      "eval_steps_per_second": 8.055,
      "step": 3900
    },
    {
      "epoch": 0.04,
      "grad_norm": 1.484375,
      "learning_rate": 4.995609644732452e-05,
      "loss": 0.0344,
      "step": 4000
    },
    {
      "epoch": 0.04,
      "eval_Final acc": 0.8003500384188509,
      "eval_R1": 0.9828,
      "eval_R2": 0.9657,
      "eval_RL": 0.9808,
      "eval_RLsum": 0.9808,
      "eval_acc": 0.4485,
      "eval_loss": 0.04084988310933113,
      "eval_runtime": 1458.2574,
      "eval_samples_per_second": 16.064,
      "eval_steps_per_second": 8.032,
      "step": 4000
    },
    {
      "epoch": 0.04,
      "eval_Final acc": 0.8207974045931871,
      "eval_R1": 0.9848,
      "eval_R2": 0.9696,
      "eval_RL": 0.9832,
      "eval_RLsum": 0.9832,
      "eval_acc": 0.46,
      "eval_loss": 0.038339026272296906,
      "eval_runtime": 1460.3625,
      "eval_samples_per_second": 16.041,
      "eval_steps_per_second": 8.021,
      "step": 4100
    },
    {
      "epoch": 0.04,
      "eval_Final acc": 0.8277981729702041,
      "eval_R1": 0.9852,
      "eval_R2": 0.9708,
      "eval_RL": 0.9837,
      "eval_RLsum": 0.9837,
      "eval_acc": 0.4786,
      "eval_loss": 0.036994535475969315,
      "eval_runtime": 1455.9411,
      "eval_samples_per_second": 16.09,
      "eval_steps_per_second": 8.045,
      "step": 4200
    },
    {
      "epoch": 0.04,
      "eval_Final acc": 0.8134978229317852,
      "eval_R1": 0.9841,
      "eval_R2": 0.9693,
      "eval_RL": 0.983,
      "eval_RLsum": 0.983,
      "eval_acc": 0.456,
      "eval_loss": 0.03735179081559181,
      "eval_runtime": 1455.6736,
      "eval_samples_per_second": 16.093,
      "eval_steps_per_second": 8.046,
      "step": 4300
    },
    {
      "epoch": 0.04,
      "eval_Final acc": 0.825663792367455,
      "eval_R1": 0.9847,
      "eval_R2": 0.97,
      "eval_RL": 0.9831,
      "eval_RLsum": 0.9832,
      "eval_acc": 0.48,
      "eval_loss": 0.03692484274506569,
      "eval_runtime": 1455.1544,
      "eval_samples_per_second": 16.099,
      "eval_steps_per_second": 8.049,
      "step": 4400
    },
    {
      "epoch": 0.04,
      "grad_norm": 0.62890625,
      "learning_rate": 4.993414467098678e-05,
      "loss": 0.0332,
      "step": 4500
    },
    {
      "epoch": 0.04,
      "eval_Final acc": 0.8109365662084863,
      "eval_R1": 0.9855,
      "eval_R2": 0.9714,
      "eval_RL": 0.984,
      "eval_RLsum": 0.984,
      "eval_acc": 0.465,
      "eval_loss": 0.03752029687166214,
      "eval_runtime": 1452.4959,
      "eval_samples_per_second": 16.128,
      "eval_steps_per_second": 8.064,
      "step": 4500
    },
    {
      "epoch": 0.04,
      "eval_Final acc": 0.8263467941603346,
      "eval_R1": 0.9865,
      "eval_R2": 0.9721,
      "eval_RL": 0.9845,
      "eval_RLsum": 0.9845,
      "eval_acc": 0.4893,
      "eval_loss": 0.03496832400560379,
      "eval_runtime": 1452.9712,
      "eval_samples_per_second": 16.123,
      "eval_steps_per_second": 8.061,
      "step": 4600
    },
    {
      "epoch": 0.04,
      "eval_Final acc": 0.8236147869888159,
      "eval_R1": 0.9859,
      "eval_R2": 0.971,
      "eval_RL": 0.984,
      "eval_RLsum": 0.984,
      "eval_acc": 0.4757,
      "eval_loss": 0.03735872730612755,
      "eval_runtime": 1452.9684,
      "eval_samples_per_second": 16.123,
      "eval_steps_per_second": 8.061,
      "step": 4700
    },
    {
      "epoch": 0.04,
      "eval_Final acc": 0.827883548194314,
      "eval_R1": 0.9853,
      "eval_R2": 0.9711,
      "eval_RL": 0.9839,
      "eval_RLsum": 0.9839,
      "eval_acc": 0.4599,
      "eval_loss": 0.037006475031375885,
      "eval_runtime": 1453.0228,
      "eval_samples_per_second": 16.122,
      "eval_steps_per_second": 8.061,
      "step": 4800
    },
    {
      "epoch": 0.04,
      "eval_Final acc": 0.8364637582173653,
      "eval_R1": 0.9859,
      "eval_R2": 0.9724,
      "eval_RL": 0.9846,
      "eval_RLsum": 0.9846,
      "eval_acc": 0.4934,
      "eval_loss": 0.03393573313951492,
      "eval_runtime": 1452.694,
      "eval_samples_per_second": 16.126,
      "eval_steps_per_second": 8.063,
      "step": 4900
    },
    {
      "epoch": 0.04,
      "grad_norm": 0.98046875,
      "learning_rate": 4.991219289464904e-05,
      "loss": 0.0305,
      "step": 5000
    },
    {
      "epoch": 0.04,
      "eval_Final acc": 0.8354392555280458,
      "eval_R1": 0.9858,
      "eval_R2": 0.9718,
      "eval_RL": 0.9844,
      "eval_RLsum": 0.9844,
      "eval_acc": 0.4975,
      "eval_loss": 0.03460527956485748,
      "eval_runtime": 1453.5753,
      "eval_samples_per_second": 16.116,
      "eval_steps_per_second": 8.058,
      "step": 5000
    },
    {
      "epoch": 0.04,
      "eval_Final acc": 0.841799709724238,
      "eval_R1": 0.9861,
      "eval_R2": 0.9725,
      "eval_RL": 0.9847,
      "eval_RLsum": 0.9847,
      "eval_acc": 0.4892,
      "eval_loss": 0.03439662605524063,
      "eval_runtime": 1456.3191,
      "eval_samples_per_second": 16.086,
      "eval_steps_per_second": 8.043,
      "step": 5100
    },
    {
      "epoch": 0.05,
      "eval_Final acc": 0.8333048749252967,
      "eval_R1": 0.986,
      "eval_R2": 0.9722,
      "eval_RL": 0.9847,
      "eval_RLsum": 0.9847,
      "eval_acc": 0.4991,
      "eval_loss": 0.03394049033522606,
      "eval_runtime": 1452.9392,
      "eval_samples_per_second": 16.123,
      "eval_steps_per_second": 8.062,
      "step": 5200
    },
    {
      "epoch": 0.05,
      "eval_Final acc": 0.8340305643302314,
      "eval_R1": 0.9855,
      "eval_R2": 0.971,
      "eval_RL": 0.9839,
      "eval_RLsum": 0.984,
      "eval_acc": 0.4957,
      "eval_loss": 0.034873489290475845,
      "eval_runtime": 1454.048,
      "eval_samples_per_second": 16.111,
      "eval_steps_per_second": 8.055,
      "step": 5300
    },
    {
      "epoch": 0.05,
      "eval_Final acc": 0.8334329377614617,
      "eval_R1": 0.985,
      "eval_R2": 0.9701,
      "eval_RL": 0.9836,
      "eval_RLsum": 0.9837,
      "eval_acc": 0.4827,
      "eval_loss": 0.0355926938354969,
      "eval_runtime": 1451.844,
      "eval_samples_per_second": 16.135,
      "eval_steps_per_second": 8.068,
      "step": 5400
    },
    {
      "epoch": 0.05,
      "grad_norm": 0.263671875,
      "learning_rate": 4.9890241118311296e-05,
      "loss": 0.0294,
      "step": 5500
    },
    {
      "epoch": 0.05,
      "eval_Final acc": 0.8330060616409118,
      "eval_R1": 0.9846,
      "eval_R2": 0.9697,
      "eval_RL": 0.9832,
      "eval_RLsum": 0.9832,
      "eval_acc": 0.4824,
      "eval_loss": 0.0374641939997673,
      "eval_runtime": 1451.8914,
      "eval_samples_per_second": 16.135,
      "eval_steps_per_second": 8.067,
      "step": 5500
    },
    {
      "epoch": 0.05,
      "eval_Final acc": 0.8362076325450354,
      "eval_R1": 0.9861,
      "eval_R2": 0.9722,
      "eval_RL": 0.9847,
      "eval_RLsum": 0.9847,
      "eval_acc": 0.4621,
      "eval_loss": 0.03642026707530022,
      "eval_runtime": 1452.5171,
      "eval_samples_per_second": 16.128,
      "eval_steps_per_second": 8.064,
      "step": 5600
    },
    {
      "epoch": 0.05,
      "eval_Final acc": 0.8437206522667122,
      "eval_R1": 0.9864,
      "eval_R2": 0.973,
      "eval_RL": 0.985,
      "eval_RLsum": 0.985,
      "eval_acc": 0.4869,
      "eval_loss": 0.03417905047535896,
      "eval_runtime": 1453.5418,
      "eval_samples_per_second": 16.116,
      "eval_steps_per_second": 8.058,
      "step": 5700
    },
    {
      "epoch": 0.05,
      "eval_Final acc": 0.8427388371894476,
      "eval_R1": 0.9855,
      "eval_R2": 0.9713,
      "eval_RL": 0.9842,
      "eval_RLsum": 0.9842,
      "eval_acc": 0.4967,
      "eval_loss": 0.033821187913417816,
      "eval_runtime": 1453.0231,
      "eval_samples_per_second": 16.122,
      "eval_steps_per_second": 8.061,
      "step": 5800
    },
    {
      "epoch": 0.05,
      "eval_Final acc": 0.8383847007598395,
      "eval_R1": 0.9865,
      "eval_R2": 0.9733,
      "eval_RL": 0.9851,
      "eval_RLsum": 0.9852,
      "eval_acc": 0.5177,
      "eval_loss": 0.03298434987664223,
      "eval_runtime": 1453.3307,
      "eval_samples_per_second": 16.119,
      "eval_steps_per_second": 8.059,
      "step": 5900
    },
    {
      "epoch": 0.05,
      "grad_norm": 0.65234375,
      "learning_rate": 4.9868289341973555e-05,
      "loss": 0.0269,
      "step": 6000
    },
    {
      "epoch": 0.05,
      "eval_Final acc": 0.8430803380858874,
      "eval_R1": 0.9865,
      "eval_R2": 0.9732,
      "eval_RL": 0.9853,
      "eval_RLsum": 0.9853,
      "eval_acc": 0.5183,
      "eval_loss": 0.03350769728422165,
      "eval_runtime": 1452.7517,
      "eval_samples_per_second": 16.125,
      "eval_steps_per_second": 8.063,
      "step": 6000
    },
    {
      "epoch": 0.05,
      "eval_Final acc": 0.8421412106206779,
      "eval_R1": 0.9867,
      "eval_R2": 0.9733,
      "eval_RL": 0.9854,
      "eval_RLsum": 0.9854,
      "eval_acc": 0.5083,
      "eval_loss": 0.03241712972521782,
      "eval_runtime": 1452.2064,
      "eval_samples_per_second": 16.131,
      "eval_steps_per_second": 8.066,
      "step": 6100
    },
    {
      "epoch": 0.05,
      "eval_Final acc": 0.8307009305899428,
      "eval_R1": 0.9875,
      "eval_R2": 0.9742,
      "eval_RL": 0.9857,
      "eval_RLsum": 0.9857,
      "eval_acc": 0.5239,
      "eval_loss": 0.03130584582686424,
      "eval_runtime": 1452.2195,
      "eval_samples_per_second": 16.131,
      "eval_steps_per_second": 8.066,
      "step": 6200
    },
    {
      "epoch": 0.06,
      "eval_Final acc": 0.8461111585417912,
      "eval_R1": 0.9866,
      "eval_R2": 0.9733,
      "eval_RL": 0.9853,
      "eval_RLsum": 0.9854,
      "eval_acc": 0.5294,
      "eval_loss": 0.031051861122250557,
      "eval_runtime": 1451.5689,
      "eval_samples_per_second": 16.138,
      "eval_steps_per_second": 8.069,
      "step": 6300
    },
    {
      "epoch": 0.06,
      "eval_Final acc": 0.8422692734568428,
      "eval_R1": 0.9882,
      "eval_R2": 0.9758,
      "eval_RL": 0.9865,
      "eval_RLsum": 0.9866,
      "eval_acc": 0.5361,
      "eval_loss": 0.03030543029308319,
      "eval_runtime": 1452.4067,
      "eval_samples_per_second": 16.129,
      "eval_steps_per_second": 8.065,
      "step": 6400
    },
    {
      "epoch": 0.06,
      "grad_norm": 0.1962890625,
      "learning_rate": 4.9846337565635814e-05,
      "loss": 0.0259,
      "step": 6500
    },
    {
      "epoch": 0.06,
      "eval_Final acc": 0.8405617689746435,
      "eval_R1": 0.9874,
      "eval_R2": 0.975,
      "eval_RL": 0.9862,
      "eval_RLsum": 0.9862,
      "eval_acc": 0.5432,
      "eval_loss": 0.03065246157348156,
      "eval_runtime": 1452.6122,
      "eval_samples_per_second": 16.127,
      "eval_steps_per_second": 8.063,
      "step": 6500
    },
    {
      "epoch": 0.06,
      "eval_Final acc": 0.8482028515324853,
      "eval_R1": 0.9875,
      "eval_R2": 0.9752,
      "eval_RL": 0.9863,
      "eval_RLsum": 0.9863,
      "eval_acc": 0.5112,
      "eval_loss": 0.03139213100075722,
      "eval_runtime": 1450.679,
      "eval_samples_per_second": 16.148,
      "eval_steps_per_second": 8.074,
      "step": 6600
    },
    {
      "epoch": 0.06,
      "eval_Final acc": 0.8590881926065056,
      "eval_R1": 0.9873,
      "eval_R2": 0.9751,
      "eval_RL": 0.9863,
      "eval_RLsum": 0.9863,
      "eval_acc": 0.5372,
      "eval_loss": 0.031337395310401917,
      "eval_runtime": 1451.3656,
      "eval_samples_per_second": 16.141,
      "eval_steps_per_second": 8.07,
      "step": 6700
    },
    {
      "epoch": 0.06,
      "eval_Final acc": 0.8578075642448562,
      "eval_R1": 0.9875,
      "eval_R2": 0.9756,
      "eval_RL": 0.9865,
      "eval_RLsum": 0.9865,
      "eval_acc": 0.5365,
      "eval_loss": 0.029987240210175514,
      "eval_runtime": 1452.5988,
      "eval_samples_per_second": 16.127,
      "eval_steps_per_second": 8.063,
      "step": 6800
    },
    {
      "epoch": 0.06,
      "eval_Final acc": 0.8523435499018185,
      "eval_R1": 0.9868,
      "eval_R2": 0.974,
      "eval_RL": 0.9856,
      "eval_RLsum": 0.9856,
      "eval_acc": 0.5236,
      "eval_loss": 0.029863450676202774,
      "eval_runtime": 1463.7558,
      "eval_samples_per_second": 16.004,
      "eval_steps_per_second": 8.002,
      "step": 6900
    },
    {
      "epoch": 0.06,
      "grad_norm": 0.51953125,
      "learning_rate": 4.982438578929807e-05,
      "loss": 0.0251,
      "step": 7000
    },
    {
      "epoch": 0.06,
      "eval_Final acc": 0.8515324852727738,
      "eval_R1": 0.988,
      "eval_R2": 0.9757,
      "eval_RL": 0.9865,
      "eval_RLsum": 0.9865,
      "eval_acc": 0.5462,
      "eval_loss": 0.0297914557158947,
      "eval_runtime": 1453.0463,
      "eval_samples_per_second": 16.122,
      "eval_steps_per_second": 8.061,
      "step": 7000
    },
    {
      "epoch": 0.06,
      "eval_Final acc": 0.8449159054042517,
      "eval_R1": 0.9876,
      "eval_R2": 0.9754,
      "eval_RL": 0.9865,
      "eval_RLsum": 0.9865,
      "eval_acc": 0.5391,
      "eval_loss": 0.028796358034014702,
      "eval_runtime": 1463.6264,
      "eval_samples_per_second": 16.005,
      "eval_steps_per_second": 8.003,
      "step": 7100
    },
    {
      "epoch": 0.06,
      "eval_Final acc": 0.8518312985571587,
      "eval_R1": 0.988,
      "eval_R2": 0.9759,
      "eval_RL": 0.9867,
      "eval_RLsum": 0.9867,
      "eval_acc": 0.5419,
      "eval_loss": 0.029616739600896835,
      "eval_runtime": 1449.9364,
      "eval_samples_per_second": 16.157,
      "eval_steps_per_second": 8.078,
      "step": 7200
    },
    {
      "epoch": 0.06,
      "eval_Final acc": 0.8484162895927602,
      "eval_R1": 0.9882,
      "eval_R2": 0.9759,
      "eval_RL": 0.9866,
      "eval_RLsum": 0.9866,
      "eval_acc": 0.5557,
      "eval_loss": 0.027878882363438606,
      "eval_runtime": 1451.89,
      "eval_samples_per_second": 16.135,
      "eval_steps_per_second": 8.067,
      "step": 7300
    },
    {
      "epoch": 0.06,
      "eval_Final acc": 0.8571672500640314,
      "eval_R1": 0.988,
      "eval_R2": 0.9757,
      "eval_RL": 0.9866,
      "eval_RLsum": 0.9866,
      "eval_acc": 0.5347,
      "eval_loss": 0.029337283223867416,
      "eval_runtime": 1451.1447,
      "eval_samples_per_second": 16.143,
      "eval_steps_per_second": 8.072,
      "step": 7400
    },
    {
      "epoch": 0.07,
      "grad_norm": 0.470703125,
      "learning_rate": 4.980243401296033e-05,
      "loss": 0.0227,
      "step": 7500
    },
    {
      "epoch": 0.07,
      "eval_Final acc": 0.8491846666097499,
      "eval_R1": 0.9871,
      "eval_R2": 0.9743,
      "eval_RL": 0.9859,
      "eval_RLsum": 0.9859,
      "eval_acc": 0.5351,
      "eval_loss": 0.029166363179683685,
      "eval_runtime": 1463.7668,
      "eval_samples_per_second": 16.004,
      "eval_steps_per_second": 8.002,
      "step": 7500
    },
    {
      "epoch": 0.07,
      "eval_Final acc": 0.8567403739434816,
      "eval_R1": 0.9881,
      "eval_R2": 0.9769,
      "eval_RL": 0.9872,
      "eval_RLsum": 0.9872,
      "eval_acc": 0.5605,
      "eval_loss": 0.027573315426707268,
      "eval_runtime": 1451.2217,
      "eval_samples_per_second": 16.142,
      "eval_steps_per_second": 8.071,
      "step": 7600
    },
    {
      "epoch": 0.07,
      "eval_Final acc": 0.8582771279774609,
      "eval_R1": 0.9877,
      "eval_R2": 0.9757,
      "eval_RL": 0.9865,
      "eval_RLsum": 0.9865,
      "eval_acc": 0.5501,
      "eval_loss": 0.029309125617146492,
      "eval_runtime": 1461.7473,
      "eval_samples_per_second": 16.026,
      "eval_steps_per_second": 8.013,
      "step": 7700
    },
    {
      "epoch": 0.07,
      "eval_Final acc": 0.8508067958678391,
      "eval_R1": 0.9878,
      "eval_R2": 0.9759,
      "eval_RL": 0.9866,
      "eval_RLsum": 0.9866,
      "eval_acc": 0.5448,
      "eval_loss": 0.028771745041012764,
      "eval_runtime": 1452.0859,
      "eval_samples_per_second": 16.133,
      "eval_steps_per_second": 8.066,
      "step": 7800
    },
    {
      "epoch": 0.07,
      "eval_Final acc": 0.8591308802185605,
      "eval_R1": 0.9885,
      "eval_R2": 0.9771,
      "eval_RL": 0.9875,
      "eval_RLsum": 0.9875,
      "eval_acc": 0.5522,
      "eval_loss": 0.029200535267591476,
      "eval_runtime": 1452.9102,
      "eval_samples_per_second": 16.124,
      "eval_steps_per_second": 8.062,
      "step": 7900
    },
    {
      "epoch": 0.07,
      "grad_norm": 0.6171875,
      "learning_rate": 4.978048223662259e-05,
      "loss": 0.0197,
      "step": 8000
    },
    {
      "epoch": 0.07,
      "eval_Final acc": 0.8661316485955776,
      "eval_R1": 0.9884,
      "eval_R2": 0.9765,
      "eval_RL": 0.9872,
      "eval_RLsum": 0.9872,
      "eval_acc": 0.5499,
      "eval_loss": 0.0286564901471138,
      "eval_runtime": 1462.7564,
      "eval_samples_per_second": 16.015,
      "eval_steps_per_second": 8.007,
      "step": 8000
    },
    {
      "epoch": 0.07,
      "eval_Final acc": 0.855801246478272,
      "eval_R1": 0.988,
      "eval_R2": 0.9756,
      "eval_RL": 0.9865,
      "eval_RLsum": 0.9866,
      "eval_acc": 0.5429,
      "eval_loss": 0.029118115082383156,
      "eval_runtime": 1452.209,
      "eval_samples_per_second": 16.131,
      "eval_steps_per_second": 8.066,
      "step": 8100
    },
    {
      "epoch": 0.07,
      "eval_Final acc": 0.8709553487577905,
      "eval_R1": 0.9888,
      "eval_R2": 0.9777,
      "eval_RL": 0.9877,
      "eval_RLsum": 0.9878,
      "eval_acc": 0.5563,
      "eval_loss": 0.02814970724284649,
      "eval_runtime": 1454.2282,
      "eval_samples_per_second": 16.109,
      "eval_steps_per_second": 8.054,
      "step": 8200
    },
    {
      "epoch": 0.07,
      "eval_Final acc": 0.8565269358832067,
      "eval_R1": 0.9882,
      "eval_R2": 0.9769,
      "eval_RL": 0.9872,
      "eval_RLsum": 0.9872,
      "eval_acc": 0.548,
      "eval_loss": 0.029164429754018784,
      "eval_runtime": 1451.5073,
      "eval_samples_per_second": 16.139,
      "eval_steps_per_second": 8.07,
      "step": 8300
    },
    {
      "epoch": 0.07,
      "eval_Final acc": 0.8612225732092547,
      "eval_R1": 0.9887,
      "eval_R2": 0.9774,
      "eval_RL": 0.9875,
      "eval_RLsum": 0.9875,
      "eval_acc": 0.5511,
      "eval_loss": 0.02785548008978367,
      "eval_runtime": 1451.9513,
      "eval_samples_per_second": 16.134,
      "eval_steps_per_second": 8.067,
      "step": 8400
    },
    {
      "epoch": 0.07,
      "grad_norm": 0.8984375,
      "learning_rate": 4.975853046028485e-05,
      "loss": 0.0208,
      "step": 8500
    },
    {
      "epoch": 0.07,
      "eval_Final acc": 0.8587893793221207,
      "eval_R1": 0.9882,
      "eval_R2": 0.9766,
      "eval_RL": 0.987,
      "eval_RLsum": 0.9871,
      "eval_acc": 0.5564,
      "eval_loss": 0.028132641687989235,
      "eval_runtime": 1449.8904,
      "eval_samples_per_second": 16.157,
      "eval_steps_per_second": 8.079,
      "step": 8500
    },
    {
      "epoch": 0.08,
      "eval_Final acc": 0.8593016306667804,
      "eval_R1": 0.9879,
      "eval_R2": 0.9759,
      "eval_RL": 0.9866,
      "eval_RLsum": 0.9867,
      "eval_acc": 0.5491,
      "eval_loss": 0.028305571526288986,
      "eval_runtime": 1451.1721,
      "eval_samples_per_second": 16.143,
      "eval_steps_per_second": 8.071,
      "step": 8600
    },
    {
      "epoch": 0.08,
      "eval_Final acc": 0.8723213523435499,
      "eval_R1": 0.9887,
      "eval_R2": 0.9773,
      "eval_RL": 0.9876,
      "eval_RLsum": 0.9876,
      "eval_acc": 0.5382,
      "eval_loss": 0.029259830713272095,
      "eval_runtime": 1450.5589,
      "eval_samples_per_second": 16.15,
      "eval_steps_per_second": 8.075,
      "step": 8700
    },
    {
      "epoch": 0.08,
      "eval_Final acc": 0.8727482284640997,
      "eval_R1": 0.9891,
      "eval_R2": 0.9779,
      "eval_RL": 0.9879,
      "eval_RLsum": 0.9879,
      "eval_acc": 0.5636,
      "eval_loss": 0.02689511515200138,
      "eval_runtime": 1453.2034,
      "eval_samples_per_second": 16.12,
      "eval_steps_per_second": 8.06,
      "step": 8800
    },
    {
      "epoch": 0.08,
      "eval_Final acc": 0.8654913344147528,
      "eval_R1": 0.9886,
      "eval_R2": 0.9773,
      "eval_RL": 0.9875,
      "eval_RLsum": 0.9875,
      "eval_acc": 0.5658,
      "eval_loss": 0.02876410074532032,
      "eval_runtime": 1449.9472,
      "eval_samples_per_second": 16.156,
      "eval_steps_per_second": 8.078,
      "step": 8900
    },
    {
      "epoch": 0.08,
      "grad_norm": 0.439453125,
      "learning_rate": 4.973657868394711e-05,
      "loss": 0.0189,
      "step": 9000
    },
    {
      "epoch": 0.08,
      "eval_Final acc": 0.8724921027917698,
      "eval_R1": 0.9888,
      "eval_R2": 0.9778,
      "eval_RL": 0.9878,
      "eval_RLsum": 0.9878,
      "eval_acc": 0.5667,
      "eval_loss": 0.027498234063386917,
      "eval_runtime": 1450.1339,
      "eval_samples_per_second": 16.154,
      "eval_steps_per_second": 8.077,
      "step": 9000
    },
    {
      "epoch": 0.08,
      "eval_Final acc": 0.8678391530777768,
      "eval_R1": 0.9886,
      "eval_R2": 0.9777,
      "eval_RL": 0.9877,
      "eval_RLsum": 0.9877,
      "eval_acc": 0.5776,
      "eval_loss": 0.025310663506388664,
      "eval_runtime": 1451.0223,
      "eval_samples_per_second": 16.144,
      "eval_steps_per_second": 8.072,
      "step": 9100
    },
    {
      "epoch": 0.08,
      "eval_Final acc": 0.8777426790745325,
      "eval_R1": 0.9893,
      "eval_R2": 0.9788,
      "eval_RL": 0.9884,
      "eval_RLsum": 0.9884,
      "eval_acc": 0.5735,
      "eval_loss": 0.026753507554531097,
      "eval_runtime": 1452.8789,
      "eval_samples_per_second": 16.124,
      "eval_steps_per_second": 8.062,
      "step": 9200
    },
    {
      "epoch": 0.08,
      "eval_Final acc": 0.8771450525057628,
      "eval_R1": 0.9893,
      "eval_R2": 0.9788,
      "eval_RL": 0.9883,
      "eval_RLsum": 0.9883,
      "eval_acc": 0.5752,
      "eval_loss": 0.02651958540081978,
      "eval_runtime": 1461.0082,
      "eval_samples_per_second": 16.034,
      "eval_steps_per_second": 8.017,
      "step": 9300
    },
    {
      "epoch": 0.08,
      "eval_Final acc": 0.8802185605737215,
      "eval_R1": 0.9889,
      "eval_R2": 0.9778,
      "eval_RL": 0.9879,
      "eval_RLsum": 0.9879,
      "eval_acc": 0.5708,
      "eval_loss": 0.025540869683027267,
      "eval_runtime": 1462.5119,
      "eval_samples_per_second": 16.018,
      "eval_steps_per_second": 8.009,
      "step": 9400
    },
    {
      "epoch": 0.08,
      "grad_norm": 0.74609375,
      "learning_rate": 4.971462690760937e-05,
      "loss": 0.0185,
      "step": 9500
    },
    {
      "epoch": 0.08,
      "eval_Final acc": 0.8808588747545463,
      "eval_R1": 0.9892,
      "eval_R2": 0.9787,
      "eval_RL": 0.9883,
      "eval_RLsum": 0.9883,
      "eval_acc": 0.5713,
      "eval_loss": 0.02617293782532215,
      "eval_runtime": 1451.0746,
      "eval_samples_per_second": 16.144,
      "eval_steps_per_second": 8.072,
      "step": 9500
    },
    {
      "epoch": 0.08,
      "eval_Final acc": 0.8802185605737215,
      "eval_R1": 0.9896,
      "eval_R2": 0.9793,
      "eval_RL": 0.9886,
      "eval_RLsum": 0.9887,
      "eval_acc": 0.5765,
      "eval_loss": 0.025828788056969643,
      "eval_runtime": 1452.1408,
      "eval_samples_per_second": 16.132,
      "eval_steps_per_second": 8.066,
      "step": 9600
    },
    {
      "epoch": 0.08,
      "eval_Final acc": 0.8950311619568001,
      "eval_R1": 0.9895,
      "eval_R2": 0.9793,
      "eval_RL": 0.9888,
      "eval_RLsum": 0.9888,
      "eval_acc": 0.5819,
      "eval_loss": 0.02496754750609398,
      "eval_runtime": 1453.5941,
      "eval_samples_per_second": 16.116,
      "eval_steps_per_second": 8.058,
      "step": 9700
    },
    {
      "epoch": 0.09,
      "eval_Final acc": 0.8919149662767865,
      "eval_R1": 0.99,
      "eval_R2": 0.9798,
      "eval_RL": 0.989,
      "eval_RLsum": 0.989,
      "eval_acc": 0.5924,
      "eval_loss": 0.02453671582043171,
      "eval_runtime": 1450.8477,
      "eval_samples_per_second": 16.146,
      "eval_steps_per_second": 8.073,
      "step": 9800
    },
    {
      "epoch": 0.09,
      "eval_Final acc": 0.8859387005890891,
      "eval_R1": 0.9898,
      "eval_R2": 0.9793,
      "eval_RL": 0.9887,
      "eval_RLsum": 0.9887,
      "eval_acc": 0.5839,
      "eval_loss": 0.025181995704770088,
      "eval_runtime": 1451.3926,
      "eval_samples_per_second": 16.14,
      "eval_steps_per_second": 8.07,
      "step": 9900
    },
    {
      "epoch": 0.09,
      "grad_norm": 0.58203125,
      "learning_rate": 4.9692675131271626e-05,
      "loss": 0.017,
      "step": 10000
    },
    {
      "epoch": 0.09,
      "eval_Final acc": 0.8914880901562366,
      "eval_R1": 0.9903,
      "eval_R2": 0.9805,
      "eval_RL": 0.9895,
      "eval_RLsum": 0.9895,
      "eval_acc": 0.5802,
      "eval_loss": 0.025952670723199844,
      "eval_runtime": 1452.6528,
      "eval_samples_per_second": 16.126,
      "eval_steps_per_second": 8.063,
      "step": 10000
    },
    {
      "epoch": 0.09,
      "eval_Final acc": 0.8896525228378724,
      "eval_R1": 0.9896,
      "eval_R2": 0.9795,
      "eval_RL": 0.9888,
      "eval_RLsum": 0.9888,
      "eval_acc": 0.5995,
      "eval_loss": 0.024223260581493378,
      "eval_runtime": 1452.818,
      "eval_samples_per_second": 16.125,
      "eval_steps_per_second": 8.062,
      "step": 10100
    },
    {
      "epoch": 0.09,
      "eval_Final acc": 0.8901647741825323,
      "eval_R1": 0.9899,
      "eval_R2": 0.9802,
      "eval_RL": 0.9892,
      "eval_RLsum": 0.9892,
      "eval_acc": 0.5976,
      "eval_loss": 0.023528151214122772,
      "eval_runtime": 1450.7267,
      "eval_samples_per_second": 16.148,
      "eval_steps_per_second": 8.074,
      "step": 10200
    },
    {
      "epoch": 0.09,
      "eval_Final acc": 0.8992999231622983,
      "eval_R1": 0.9896,
      "eval_R2": 0.9797,
      "eval_RL": 0.989,
      "eval_RLsum": 0.989,
      "eval_acc": 0.5917,
      "eval_loss": 0.024023424834012985,
      "eval_runtime": 1452.1557,
      "eval_samples_per_second": 16.132,
      "eval_steps_per_second": 8.066,
      "step": 10300
    },
    {
      "epoch": 0.09,
      "eval_Final acc": 0.8960129770340647,
      "eval_R1": 0.99,
      "eval_R2": 0.98,
      "eval_RL": 0.9891,
      "eval_RLsum": 0.9891,
      "eval_acc": 0.6019,
      "eval_loss": 0.023717278614640236,
      "eval_runtime": 1450.9718,
      "eval_samples_per_second": 16.145,
      "eval_steps_per_second": 8.073,
      "step": 10400
    },
    {
      "epoch": 0.09,
      "grad_norm": 0.466796875,
      "learning_rate": 4.9670723354933884e-05,
      "loss": 0.0163,
      "step": 10500
    },
    {
      "epoch": 0.09,
      "eval_Final acc": 0.8959276018099548,
      "eval_R1": 0.9896,
      "eval_R2": 0.979,
      "eval_RL": 0.9886,
      "eval_RLsum": 0.9886,
      "eval_acc": 0.5796,
      "eval_loss": 0.025336040183901787,
      "eval_runtime": 1448.595,
      "eval_samples_per_second": 16.172,
      "eval_steps_per_second": 8.086,
      "step": 10500
    },
    {
      "epoch": 0.09,
      "eval_Final acc": 0.8957995389737898,
      "eval_R1": 0.9901,
      "eval_R2": 0.9803,
      "eval_RL": 0.9893,
      "eval_RLsum": 0.9893,
      "eval_acc": 0.5992,
      "eval_loss": 0.024455588310956955,
      "eval_runtime": 1452.4703,
      "eval_samples_per_second": 16.128,
      "eval_steps_per_second": 8.064,
      "step": 10600
    },
    {
      "epoch": 0.09,
      "eval_Final acc": 0.9006659267480577,
      "eval_R1": 0.9898,
      "eval_R2": 0.9799,
      "eval_RL": 0.9891,
      "eval_RLsum": 0.9891,
      "eval_acc": 0.5939,
      "eval_loss": 0.025212198495864868,
      "eval_runtime": 1453.3691,
      "eval_samples_per_second": 16.118,
      "eval_steps_per_second": 8.059,
      "step": 10700
    },
    {
      "epoch": 0.09,
      "eval_Final acc": 0.9012635533168275,
      "eval_R1": 0.9903,
      "eval_R2": 0.9805,
      "eval_RL": 0.9894,
      "eval_RLsum": 0.9894,
      "eval_acc": 0.6025,
      "eval_loss": 0.023420142009854317,
      "eval_runtime": 1451.8761,
      "eval_samples_per_second": 16.135,
      "eval_steps_per_second": 8.067,
      "step": 10800
    },
    {
      "epoch": 0.1,
      "eval_Final acc": 0.9087338854264493,
      "eval_R1": 0.9902,
      "eval_R2": 0.9804,
      "eval_RL": 0.9894,
      "eval_RLsum": 0.9894,
      "eval_acc": 0.6133,
      "eval_loss": 0.023686418309807777,
      "eval_runtime": 1451.2152,
      "eval_samples_per_second": 16.142,
      "eval_steps_per_second": 8.071,
      "step": 10900
    },
    {
      "epoch": 0.1,
      "grad_norm": 0.53515625,
      "learning_rate": 4.964877157859614e-05,
      "loss": 0.015,
      "step": 11000
    },
    {
      "epoch": 0.1,
      "eval_Final acc": 0.8992999231622983,
      "eval_R1": 0.99,
      "eval_R2": 0.9799,
      "eval_RL": 0.9891,
      "eval_RLsum": 0.9891,
      "eval_acc": 0.604,
      "eval_loss": 0.024508416652679443,
      "eval_runtime": 1450.3538,
      "eval_samples_per_second": 16.152,
      "eval_steps_per_second": 8.076,
      "step": 11000
    },
    {
      "epoch": 0.1,
      "eval_Final acc": 0.8949457867326902,
      "eval_R1": 0.9905,
      "eval_R2": 0.9807,
      "eval_RL": 0.9895,
      "eval_RLsum": 0.9895,
      "eval_acc": 0.6036,
      "eval_loss": 0.023102611303329468,
      "eval_runtime": 1452.4964,
      "eval_samples_per_second": 16.128,
      "eval_steps_per_second": 8.064,
      "step": 11100
    },
    {
      "epoch": 0.1,
      "eval_Final acc": 0.9003671134636728,
      "eval_R1": 0.9906,
      "eval_R2": 0.9813,
      "eval_RL": 0.9898,
      "eval_RLsum": 0.9898,
      "eval_acc": 0.6195,
      "eval_loss": 0.025366883724927902,
      "eval_runtime": 1450.5915,
      "eval_samples_per_second": 16.149,
      "eval_steps_per_second": 8.075,
      "step": 11200
    },
    {
      "epoch": 0.1,
      "eval_Final acc": 0.8933236574746009,
      "eval_R1": 0.9899,
      "eval_R2": 0.9799,
      "eval_RL": 0.9889,
      "eval_RLsum": 0.989,
      "eval_acc": 0.6119,
      "eval_loss": 0.024762660264968872,
      "eval_runtime": 1450.5609,
      "eval_samples_per_second": 16.15,
      "eval_steps_per_second": 8.075,
      "step": 11300
    },
    {
      "epoch": 0.1,
      "eval_Final acc": 0.9096730128916588,
      "eval_R1": 0.9905,
      "eval_R2": 0.9811,
      "eval_RL": 0.9897,
      "eval_RLsum": 0.9897,
      "eval_acc": 0.6061,
      "eval_loss": 0.023372193798422813,
      "eval_runtime": 1451.9756,
      "eval_samples_per_second": 16.134,
      "eval_steps_per_second": 8.067,
      "step": 11400
    },
    {
      "epoch": 0.1,
      "grad_norm": 0.98046875,
      "learning_rate": 4.96268198022584e-05,
      "loss": 0.015,
      "step": 11500
    },
    {
      "epoch": 0.1,
      "eval_Final acc": 0.9110390164774183,
      "eval_R1": 0.9904,
      "eval_R2": 0.9808,
      "eval_RL": 0.9896,
      "eval_RLsum": 0.9896,
      "eval_acc": 0.6084,
      "eval_loss": 0.022345341742038727,
      "eval_runtime": 1451.9936,
      "eval_samples_per_second": 16.134,
      "eval_steps_per_second": 8.067,
      "step": 11500
    },
    {
      "epoch": 0.1,
      "eval_Final acc": 0.9126184581234525,
      "eval_R1": 0.9907,
      "eval_R2": 0.981,
      "eval_RL": 0.9898,
      "eval_RLsum": 0.9899,
      "eval_acc": 0.6209,
      "eval_loss": 0.021934418007731438,
      "eval_runtime": 1449.8533,
      "eval_samples_per_second": 16.157,
      "eval_steps_per_second": 8.079,
      "step": 11600
    },
    {
      "epoch": 0.1,
      "eval_Final acc": 0.9121915820029027,
      "eval_R1": 0.991,
      "eval_R2": 0.9818,
      "eval_RL": 0.9901,
      "eval_RLsum": 0.9902,
      "eval_acc": 0.6197,
      "eval_loss": 0.021505460143089294,
      "eval_runtime": 1450.5832,
      "eval_samples_per_second": 16.149,
      "eval_steps_per_second": 8.075,
      "step": 11700
    },
    {
      "epoch": 0.1,
      "eval_Final acc": 0.9144113378297618,
      "eval_R1": 0.9907,
      "eval_R2": 0.9814,
      "eval_RL": 0.9899,
      "eval_RLsum": 0.9899,
      "eval_acc": 0.6187,
      "eval_loss": 0.021871628239750862,
      "eval_runtime": 1451.8257,
      "eval_samples_per_second": 16.136,
      "eval_steps_per_second": 8.068,
      "step": 11800
    },
    {
      "epoch": 0.1,
      "eval_Final acc": 0.9141978997694868,
      "eval_R1": 0.991,
      "eval_R2": 0.9819,
      "eval_RL": 0.9902,
      "eval_RLsum": 0.9902,
      "eval_acc": 0.6257,
      "eval_loss": 0.02094077877700329,
      "eval_runtime": 1463.4648,
      "eval_samples_per_second": 16.007,
      "eval_steps_per_second": 8.004,
      "step": 11900
    },
    {
      "epoch": 0.11,
      "grad_norm": 0.23828125,
      "learning_rate": 4.960486802592066e-05,
      "loss": 0.0134,
      "step": 12000
    },
    {
      "epoch": 0.11,
      "eval_Final acc": 0.9180824724664902,
      "eval_R1": 0.9906,
      "eval_R2": 0.9813,
      "eval_RL": 0.9899,
      "eval_RLsum": 0.9899,
      "eval_acc": 0.6147,
      "eval_loss": 0.021829580888152122,
      "eval_runtime": 1463.5331,
      "eval_samples_per_second": 16.006,
      "eval_steps_per_second": 8.003,
      "step": 12000
    },
    {
      "epoch": 0.11,
      "eval_Final acc": 0.9104413899086485,
      "eval_R1": 0.9909,
      "eval_R2": 0.9817,
      "eval_RL": 0.9901,
      "eval_RLsum": 0.9901,
      "eval_acc": 0.6231,
      "eval_loss": 0.021953416988253593,
      "eval_runtime": 1459.3648,
      "eval_samples_per_second": 16.052,
      "eval_steps_per_second": 8.026,
      "step": 12100
    },
    {
      "epoch": 0.11,
      "eval_Final acc": 0.911508580210023,
      "eval_R1": 0.9907,
      "eval_R2": 0.9814,
      "eval_RL": 0.9899,
      "eval_RLsum": 0.9899,
      "eval_acc": 0.6235,
      "eval_loss": 0.021894315257668495,
      "eval_runtime": 1463.2725,
      "eval_samples_per_second": 16.009,
      "eval_steps_per_second": 8.005,
      "step": 12200
    },
    {
      "epoch": 0.11,
      "eval_Final acc": 0.9195765388884146,
      "eval_R1": 0.9911,
      "eval_R2": 0.9821,
      "eval_RL": 0.9904,
      "eval_RLsum": 0.9904,
      "eval_acc": 0.6261,
      "eval_loss": 0.0205834973603487,
      "eval_runtime": 1450.6918,
      "eval_samples_per_second": 16.148,
      "eval_steps_per_second": 8.074,
      "step": 12300
    },
    {
      "epoch": 0.11,
      "eval_Final acc": 0.9150516520105866,
      "eval_R1": 0.9906,
      "eval_R2": 0.9815,
      "eval_RL": 0.99,
      "eval_RLsum": 0.99,
      "eval_acc": 0.6164,
      "eval_loss": 0.020932447165250778,
      "eval_runtime": 1450.9827,
      "eval_samples_per_second": 16.145,
      "eval_steps_per_second": 8.072,
      "step": 12400
    },
    {
      "epoch": 0.11,
      "grad_norm": 1.828125,
      "learning_rate": 4.958291624958292e-05,
      "loss": 0.0129,
      "step": 12500
    },
    {
      "epoch": 0.11,
      "eval_Final acc": 0.9190215999316999,
      "eval_R1": 0.9909,
      "eval_R2": 0.9821,
      "eval_RL": 0.9902,
      "eval_RLsum": 0.9902,
      "eval_acc": 0.6214,
      "eval_loss": 0.020663928240537643,
      "eval_runtime": 1450.2698,
      "eval_samples_per_second": 16.153,
      "eval_steps_per_second": 8.076,
      "step": 12500
    },
    {
      "epoch": 0.11,
      "eval_Final acc": 0.9282848117476309,
      "eval_R1": 0.9912,
      "eval_R2": 0.9822,
      "eval_RL": 0.9905,
      "eval_RLsum": 0.9905,
      "eval_acc": 0.634,
      "eval_loss": 0.020196454599499702,
      "eval_runtime": 1449.7517,
      "eval_samples_per_second": 16.159,
      "eval_steps_per_second": 8.079,
      "step": 12600
    },
    {
      "epoch": 0.11,
      "eval_Final acc": 0.9266199948774866,
      "eval_R1": 0.9912,
      "eval_R2": 0.9823,
      "eval_RL": 0.9903,
      "eval_RLsum": 0.9904,
      "eval_acc": 0.6292,
      "eval_loss": 0.020464591681957245,
      "eval_runtime": 1450.6096,
      "eval_samples_per_second": 16.149,
      "eval_steps_per_second": 8.075,
      "step": 12700
    },
    {
      "epoch": 0.11,
      "eval_Final acc": 0.9174421582856656,
      "eval_R1": 0.9913,
      "eval_R2": 0.982,
      "eval_RL": 0.9903,
      "eval_RLsum": 0.9903,
      "eval_acc": 0.6388,
      "eval_loss": 0.019640225917100906,
      "eval_runtime": 1449.7946,
      "eval_samples_per_second": 16.158,
      "eval_steps_per_second": 8.079,
      "step": 12800
    },
    {
      "epoch": 0.11,
      "eval_Final acc": 0.9144967130538718,
      "eval_R1": 0.9905,
      "eval_R2": 0.981,
      "eval_RL": 0.9898,
      "eval_RLsum": 0.9898,
      "eval_acc": 0.6319,
      "eval_loss": 0.020924171432852745,
      "eval_runtime": 1450.2548,
      "eval_samples_per_second": 16.153,
      "eval_steps_per_second": 8.077,
      "step": 12900
    },
    {
      "epoch": 0.11,
      "grad_norm": 0.59765625,
      "learning_rate": 4.956096447324517e-05,
      "loss": 0.0122,
      "step": 13000
    },
    {
      "epoch": 0.11,
      "eval_Final acc": 0.9319986339964142,
      "eval_R1": 0.9909,
      "eval_R2": 0.9819,
      "eval_RL": 0.9903,
      "eval_RLsum": 0.9903,
      "eval_acc": 0.639,
      "eval_loss": 0.02068125642836094,
      "eval_runtime": 1450.1468,
      "eval_samples_per_second": 16.154,
      "eval_steps_per_second": 8.077,
      "step": 13000
    },
    {
      "epoch": 0.11,
      "eval_Final acc": 0.9329377614616239,
      "eval_R1": 0.9909,
      "eval_R2": 0.9818,
      "eval_RL": 0.9903,
      "eval_RLsum": 0.9903,
      "eval_acc": 0.6378,
      "eval_loss": 0.02071656845510006,
      "eval_runtime": 1450.5848,
      "eval_samples_per_second": 16.149,
      "eval_steps_per_second": 8.075,
      "step": 13100
    },
    {
      "epoch": 0.12,
      "eval_Final acc": 0.9319132587723042,
      "eval_R1": 0.9912,
      "eval_R2": 0.9826,
      "eval_RL": 0.9907,
      "eval_RLsum": 0.9907,
      "eval_acc": 0.6419,
      "eval_loss": 0.020106444135308266,
      "eval_runtime": 1451.3754,
      "eval_samples_per_second": 16.141,
      "eval_steps_per_second": 8.07,
      "step": 13200
    },
    {
      "epoch": 0.12,
      "eval_Final acc": 0.9173567830615555,
      "eval_R1": 0.991,
      "eval_R2": 0.982,
      "eval_RL": 0.9904,
      "eval_RLsum": 0.9904,
      "eval_acc": 0.6347,
      "eval_loss": 0.0205816850066185,
      "eval_runtime": 1449.0746,
      "eval_samples_per_second": 16.166,
      "eval_steps_per_second": 8.083,
      "step": 13300
    },
    {
      "epoch": 0.12,
      "eval_Final acc": 0.9335780756424485,
      "eval_R1": 0.991,
      "eval_R2": 0.9822,
      "eval_RL": 0.9905,
      "eval_RLsum": 0.9906,
      "eval_acc": 0.6372,
      "eval_loss": 0.020543359220027924,
      "eval_runtime": 1449.6548,
      "eval_samples_per_second": 16.16,
      "eval_steps_per_second": 8.08,
      "step": 13400
    },
    {
      "epoch": 0.12,
      "grad_norm": 0.27734375,
      "learning_rate": 4.953901269690744e-05,
      "loss": 0.0111,
      "step": 13500
    },
    {
      "epoch": 0.12,
      "eval_Final acc": 0.930205754290105,
      "eval_R1": 0.9912,
      "eval_R2": 0.9825,
      "eval_RL": 0.9907,
      "eval_RLsum": 0.9907,
      "eval_acc": 0.6448,
      "eval_loss": 0.019825078547000885,
      "eval_runtime": 1450.8026,
      "eval_samples_per_second": 16.147,
      "eval_steps_per_second": 8.073,
      "step": 13500
    },
    {
      "epoch": 0.12,
      "eval_Final acc": 0.9357978314693076,
      "eval_R1": 0.9911,
      "eval_R2": 0.9821,
      "eval_RL": 0.9905,
      "eval_RLsum": 0.9905,
      "eval_acc": 0.645,
      "eval_loss": 0.020004283636808395,
      "eval_runtime": 1461.8559,
      "eval_samples_per_second": 16.025,
      "eval_steps_per_second": 8.012,
      "step": 13600
    },
    {
      "epoch": 0.12,
      "eval_Final acc": 0.9339195765388885,
      "eval_R1": 0.9912,
      "eval_R2": 0.9824,
      "eval_RL": 0.9905,
      "eval_RLsum": 0.9905,
      "eval_acc": 0.6342,
      "eval_loss": 0.020793380215764046,
      "eval_runtime": 1463.9178,
      "eval_samples_per_second": 16.002,
      "eval_steps_per_second": 8.001,
      "step": 13700
    },
    {
      "epoch": 0.12,
      "eval_Final acc": 0.9270468709980364,
      "eval_R1": 0.9909,
      "eval_R2": 0.9822,
      "eval_RL": 0.9904,
      "eval_RLsum": 0.9904,
      "eval_acc": 0.6307,
      "eval_loss": 0.02046465128660202,
      "eval_runtime": 1450.1003,
      "eval_samples_per_second": 16.155,
      "eval_steps_per_second": 8.077,
      "step": 13800
    },
    {
      "epoch": 0.12,
      "eval_Final acc": 0.9397677793904209,
      "eval_R1": 0.9912,
      "eval_R2": 0.9825,
      "eval_RL": 0.9907,
      "eval_RLsum": 0.9907,
      "eval_acc": 0.6438,
      "eval_loss": 0.01898295432329178,
      "eval_runtime": 1451.5667,
      "eval_samples_per_second": 16.138,
      "eval_steps_per_second": 8.069,
      "step": 13900
    },
    {
      "epoch": 0.12,
      "grad_norm": 0.40234375,
      "learning_rate": 4.9517060920569696e-05,
      "loss": 0.011,
      "step": 14000
    },
    {
      "epoch": 0.12,
      "eval_Final acc": 0.923162298301033,
      "eval_R1": 0.9908,
      "eval_R2": 0.9817,
      "eval_RL": 0.9902,
      "eval_RLsum": 0.9902,
      "eval_acc": 0.6253,
      "eval_loss": 0.0198197178542614,
      "eval_runtime": 1450.2616,
      "eval_samples_per_second": 16.153,
      "eval_steps_per_second": 8.076,
      "step": 14000
    },
    {
      "epoch": 0.12,
      "eval_Final acc": 0.9402800307350807,
      "eval_R1": 0.9913,
      "eval_R2": 0.9826,
      "eval_RL": 0.9907,
      "eval_RLsum": 0.9907,
      "eval_acc": 0.6412,
      "eval_loss": 0.019281338900327682,
      "eval_runtime": 1450.3797,
      "eval_samples_per_second": 16.152,
      "eval_steps_per_second": 8.076,
      "step": 14100
    },
    {
      "epoch": 0.12,
      "eval_Final acc": 0.9471100486638777,
      "eval_R1": 0.9914,
      "eval_R2": 0.9829,
      "eval_RL": 0.9909,
      "eval_RLsum": 0.9909,
      "eval_acc": 0.6502,
      "eval_loss": 0.017984973266720772,
      "eval_runtime": 1452.0929,
      "eval_samples_per_second": 16.133,
      "eval_steps_per_second": 8.066,
      "step": 14200
    },
    {
      "epoch": 0.13,
      "eval_Final acc": 0.944249978656194,
      "eval_R1": 0.9913,
      "eval_R2": 0.9828,
      "eval_RL": 0.9908,
      "eval_RLsum": 0.9908,
      "eval_acc": 0.6404,
      "eval_loss": 0.019001377746462822,
      "eval_runtime": 1451.6302,
      "eval_samples_per_second": 16.138,
      "eval_steps_per_second": 8.069,
      "step": 14300
    },
    {
      "epoch": 0.13,
      "eval_Final acc": 0.9420729104413899,
      "eval_R1": 0.9916,
      "eval_R2": 0.9832,
      "eval_RL": 0.991,
      "eval_RLsum": 0.991,
      "eval_acc": 0.6516,
      "eval_loss": 0.018046874552965164,
      "eval_runtime": 1453.4282,
      "eval_samples_per_second": 16.118,
      "eval_steps_per_second": 8.059,
      "step": 14400
    },
    {
      "epoch": 0.13,
      "grad_norm": 0.35546875,
      "learning_rate": 4.9495109144231955e-05,
      "loss": 0.0107,
      "step": 14500
    },
    {
      "epoch": 0.13,
      "eval_Final acc": 0.94190215999317,
      "eval_R1": 0.9912,
      "eval_R2": 0.9827,
      "eval_RL": 0.9907,
      "eval_RLsum": 0.9907,
      "eval_acc": 0.6478,
      "eval_loss": 0.018823372200131416,
      "eval_runtime": 1451.0015,
      "eval_samples_per_second": 16.145,
      "eval_steps_per_second": 8.072,
      "step": 14500
    },
    {
      "epoch": 0.13,
      "eval_Final acc": 0.9454452317937334,
      "eval_R1": 0.9913,
      "eval_R2": 0.9827,
      "eval_RL": 0.9908,
      "eval_RLsum": 0.9908,
      "eval_acc": 0.6468,
      "eval_loss": 0.0185859352350235,
      "eval_runtime": 1463.0333,
      "eval_samples_per_second": 16.012,
      "eval_steps_per_second": 8.006,
      "step": 14600
    },
    {
      "epoch": 0.13,
      "eval_Final acc": 0.9363100828139674,
      "eval_R1": 0.9915,
      "eval_R2": 0.983,
      "eval_RL": 0.9909,
      "eval_RLsum": 0.9909,
      "eval_acc": 0.6489,
      "eval_loss": 0.019154921174049377,
      "eval_runtime": 1451.3524,
      "eval_samples_per_second": 16.141,
      "eval_steps_per_second": 8.07,
      "step": 14700
    },
    {
      "epoch": 0.13,
      "eval_Final acc": 0.94190215999317,
      "eval_R1": 0.9915,
      "eval_R2": 0.9831,
      "eval_RL": 0.9909,
      "eval_RLsum": 0.9909,
      "eval_acc": 0.6486,
      "eval_loss": 0.018625011667609215,
      "eval_runtime": 1450.9013,
      "eval_samples_per_second": 16.146,
      "eval_steps_per_second": 8.073,
      "step": 14800
    },
    {
      "epoch": 0.13,
      "eval_Final acc": 0.9416033467087851,
      "eval_R1": 0.9919,
      "eval_R2": 0.9839,
      "eval_RL": 0.9914,
      "eval_RLsum": 0.9914,
      "eval_acc": 0.6491,
      "eval_loss": 0.019322440028190613,
      "eval_runtime": 1450.5226,
      "eval_samples_per_second": 16.15,
      "eval_steps_per_second": 8.075,
      "step": 14900
    },
    {
      "epoch": 0.13,
      "grad_norm": 0.193359375,
      "learning_rate": 4.9473157367894214e-05,
      "loss": 0.0104,
      "step": 15000
    },
    {
      "epoch": 0.13,
      "eval_Final acc": 0.941944847605225,
      "eval_R1": 0.9917,
      "eval_R2": 0.9836,
      "eval_RL": 0.9912,
      "eval_RLsum": 0.9912,
      "eval_acc": 0.6556,
      "eval_loss": 0.017888953909277916,
      "eval_runtime": 1450.2096,
      "eval_samples_per_second": 16.154,
      "eval_steps_per_second": 8.077,
      "step": 15000
    },
    {
      "epoch": 0.13,
      "eval_Final acc": 0.9384017758046614,
      "eval_R1": 0.9915,
      "eval_R2": 0.9832,
      "eval_RL": 0.991,
      "eval_RLsum": 0.991,
      "eval_acc": 0.6525,
      "eval_loss": 0.017996517941355705,
      "eval_runtime": 1457.0489,
      "eval_samples_per_second": 16.078,
      "eval_steps_per_second": 8.039,
      "step": 15100
    },
    {
      "epoch": 0.13,
      "eval_Final acc": 0.9480064885170324,
      "eval_R1": 0.9917,
      "eval_R2": 0.9835,
      "eval_RL": 0.9912,
      "eval_RLsum": 0.9912,
      "eval_acc": 0.6528,
      "eval_loss": 0.01767478697001934,
      "eval_runtime": 1462.9178,
      "eval_samples_per_second": 16.013,
      "eval_steps_per_second": 8.007,
      "step": 15200
    },
    {
      "epoch": 0.13,
      "eval_Final acc": 0.9430547255186544,
      "eval_R1": 0.9918,
      "eval_R2": 0.9835,
      "eval_RL": 0.9911,
      "eval_RLsum": 0.9911,
      "eval_acc": 0.6535,
      "eval_loss": 0.017685310915112495,
      "eval_runtime": 1450.3083,
      "eval_samples_per_second": 16.152,
      "eval_steps_per_second": 8.076,
      "step": 15300
    },
    {
      "epoch": 0.13,
      "eval_Final acc": 0.9410910953641254,
      "eval_R1": 0.9915,
      "eval_R2": 0.9831,
      "eval_RL": 0.991,
      "eval_RLsum": 0.991,
      "eval_acc": 0.6514,
      "eval_loss": 0.017296113073825836,
      "eval_runtime": 1451.3492,
      "eval_samples_per_second": 16.141,
      "eval_steps_per_second": 8.07,
      "step": 15400
    },
    {
      "epoch": 0.14,
      "grad_norm": 0.019775390625,
      "learning_rate": 4.945120559155647e-05,
      "loss": 0.0098,
      "step": 15500
    },
    {
      "epoch": 0.14,
      "eval_Final acc": 0.9433108511909843,
      "eval_R1": 0.9914,
      "eval_R2": 0.9828,
      "eval_RL": 0.9908,
      "eval_RLsum": 0.9908,
      "eval_acc": 0.6579,
      "eval_loss": 0.017405711114406586,
      "eval_runtime": 1462.1068,
      "eval_samples_per_second": 16.022,
      "eval_steps_per_second": 8.011,
      "step": 15500
    },
    {
      "epoch": 0.14,
      "eval_Final acc": 0.944335353880304,
      "eval_R1": 0.9918,
      "eval_R2": 0.9835,
      "eval_RL": 0.9912,
      "eval_RLsum": 0.9912,
      "eval_acc": 0.6587,
      "eval_loss": 0.017218494787812233,
      "eval_runtime": 1449.9491,
      "eval_samples_per_second": 16.156,
      "eval_steps_per_second": 8.078,
      "step": 15600
    },
    {
      "epoch": 0.14,
      "eval_Final acc": 0.948774865534022,
      "eval_R1": 0.992,
      "eval_R2": 0.9838,
      "eval_RL": 0.9914,
      "eval_RLsum": 0.9914,
      "eval_acc": 0.6619,
      "eval_loss": 0.016830172389745712,
      "eval_runtime": 1450.7441,
      "eval_samples_per_second": 16.148,
      "eval_steps_per_second": 8.074,
      "step": 15700
    },
    {
      "epoch": 0.14,
      "eval_Final acc": 0.9457440450781184,
      "eval_R1": 0.992,
      "eval_R2": 0.9837,
      "eval_RL": 0.9914,
      "eval_RLsum": 0.9914,
      "eval_acc": 0.6616,
      "eval_loss": 0.018399380147457123,
      "eval_runtime": 1462.926,
      "eval_samples_per_second": 16.013,
      "eval_steps_per_second": 8.007,
      "step": 15800
    },
    {
      "epoch": 0.14,
      "eval_Final acc": 0.9450610432852387,
      "eval_R1": 0.9916,
      "eval_R2": 0.9832,
      "eval_RL": 0.991,
      "eval_RLsum": 0.9911,
      "eval_acc": 0.6537,
      "eval_loss": 0.017210980877280235,
      "eval_runtime": 1450.764,
      "eval_samples_per_second": 16.147,
      "eval_steps_per_second": 8.074,
      "step": 15900
    },
    {
      "epoch": 0.14,
      "grad_norm": 0.2490234375,
      "learning_rate": 4.942925381521873e-05,
      "loss": 0.0092,
      "step": 16000
    },
    {
      "epoch": 0.14,
      "eval_Final acc": 0.9471954238879877,
      "eval_R1": 0.9917,
      "eval_R2": 0.9834,
      "eval_RL": 0.9911,
      "eval_RLsum": 0.9911,
      "eval_acc": 0.6601,
      "eval_loss": 0.016908491030335426,
      "eval_runtime": 1451.7265,
      "eval_samples_per_second": 16.137,
      "eval_steps_per_second": 8.068,
      "step": 16000
    },
    {
      "epoch": 0.14,
      "eval_Final acc": 0.946427046870998,
      "eval_R1": 0.992,
      "eval_R2": 0.984,
      "eval_RL": 0.9915,
      "eval_RLsum": 0.9915,
      "eval_acc": 0.6547,
      "eval_loss": 0.017686331644654274,
      "eval_runtime": 1451.0736,
      "eval_samples_per_second": 16.144,
      "eval_steps_per_second": 8.072,
      "step": 16100
    },
    {
      "epoch": 0.14,
      "eval_Final acc": 0.9481772389652523,
      "eval_R1": 0.992,
      "eval_R2": 0.9839,
      "eval_RL": 0.9914,
      "eval_RLsum": 0.9914,
      "eval_acc": 0.6649,
      "eval_loss": 0.016972754150629044,
      "eval_runtime": 1462.9166,
      "eval_samples_per_second": 16.013,
      "eval_steps_per_second": 8.007,
      "step": 16200
    },
    {
      "epoch": 0.14,
      "eval_Final acc": 0.9416033467087851,
      "eval_R1": 0.9921,
      "eval_R2": 0.984,
      "eval_RL": 0.9914,
      "eval_RLsum": 0.9915,
      "eval_acc": 0.6693,
      "eval_loss": 0.017500566318631172,
      "eval_runtime": 1452.1501,
      "eval_samples_per_second": 16.132,
      "eval_steps_per_second": 8.066,
      "step": 16300
    },
    {
      "epoch": 0.14,
      "eval_Final acc": 0.948860240758132,
      "eval_R1": 0.992,
      "eval_R2": 0.9838,
      "eval_RL": 0.9913,
      "eval_RLsum": 0.9913,
      "eval_acc": 0.6632,
      "eval_loss": 0.017114121466875076,
      "eval_runtime": 1460.939,
      "eval_samples_per_second": 16.035,
      "eval_steps_per_second": 8.017,
      "step": 16400
    },
    {
      "epoch": 0.14,
      "grad_norm": 0.0016326904296875,
      "learning_rate": 4.940730203888099e-05,
      "loss": 0.0082,
      "step": 16500
    },
    {
      "epoch": 0.14,
      "eval_Final acc": 0.9450610432852387,
      "eval_R1": 0.992,
      "eval_R2": 0.9839,
      "eval_RL": 0.9914,
      "eval_RLsum": 0.9914,
      "eval_acc": 0.6675,
      "eval_loss": 0.016872666776180267,
      "eval_runtime": 1452.1284,
      "eval_samples_per_second": 16.132,
      "eval_steps_per_second": 8.066,
      "step": 16500
    },
    {
      "epoch": 0.15,
      "eval_Final acc": 0.9508665585247161,
      "eval_R1": 0.9919,
      "eval_R2": 0.9838,
      "eval_RL": 0.9914,
      "eval_RLsum": 0.9914,
      "eval_acc": 0.6604,
      "eval_loss": 0.017369089648127556,
      "eval_runtime": 1463.9971,
      "eval_samples_per_second": 16.001,
      "eval_steps_per_second": 8.001,
      "step": 16600
    },
    {
      "epoch": 0.15,
      "eval_Final acc": 0.9541535046529497,
      "eval_R1": 0.992,
      "eval_R2": 0.9841,
      "eval_RL": 0.9915,
      "eval_RLsum": 0.9915,
      "eval_acc": 0.661,
      "eval_loss": 0.016900382936000824,
      "eval_runtime": 1462.7224,
      "eval_samples_per_second": 16.015,
      "eval_steps_per_second": 8.008,
      "step": 16700
    },
    {
      "epoch": 0.15,
      "eval_Final acc": 0.9528301886792453,
      "eval_R1": 0.992,
      "eval_R2": 0.9841,
      "eval_RL": 0.9915,
      "eval_RLsum": 0.9916,
      "eval_acc": 0.6713,
      "eval_loss": 0.017047865316271782,
      "eval_runtime": 1463.7473,
      "eval_samples_per_second": 16.004,
      "eval_steps_per_second": 8.002,
      "step": 16800
    },
    {
      "epoch": 0.15,
      "eval_Final acc": 0.9502689319559464,
      "eval_R1": 0.9921,
      "eval_R2": 0.9841,
      "eval_RL": 0.9916,
      "eval_RLsum": 0.9916,
      "eval_acc": 0.6742,
      "eval_loss": 0.016844719648361206,
      "eval_runtime": 1462.0385,
      "eval_samples_per_second": 16.023,
      "eval_steps_per_second": 8.011,
      "step": 16900
    },
    {
      "epoch": 0.15,
      "grad_norm": 0.19921875,
      "learning_rate": 4.938535026254324e-05,
      "loss": 0.0084,
      "step": 17000
    },
    {
      "epoch": 0.15,
      "eval_Final acc": 0.956031759583369,
      "eval_R1": 0.9921,
      "eval_R2": 0.9842,
      "eval_RL": 0.9916,
      "eval_RLsum": 0.9916,
      "eval_acc": 0.6706,
      "eval_loss": 0.016474982723593712,
      "eval_runtime": 1451.4334,
      "eval_samples_per_second": 16.14,
      "eval_steps_per_second": 8.07,
      "step": 17000
    },
    {
      "epoch": 0.15,
      "eval_Final acc": 0.9571416374967985,
      "eval_R1": 0.9919,
      "eval_R2": 0.9837,
      "eval_RL": 0.9915,
      "eval_RLsum": 0.9915,
      "eval_acc": 0.6663,
      "eval_loss": 0.016770988702774048,
      "eval_runtime": 1451.201,
      "eval_samples_per_second": 16.142,
      "eval_steps_per_second": 8.071,
      "step": 17100
    },
    {
      "epoch": 0.15,
      "eval_Final acc": 0.9605993340732519,
      "eval_R1": 0.9922,
      "eval_R2": 0.9845,
      "eval_RL": 0.9918,
      "eval_RLsum": 0.9918,
      "eval_acc": 0.667,
      "eval_loss": 0.016466723755002022,
      "eval_runtime": 1451.1773,
      "eval_samples_per_second": 16.143,
      "eval_steps_per_second": 8.071,
      "step": 17200
    },
    {
      "epoch": 0.15,
      "eval_Final acc": 0.9587210791428328,
      "eval_R1": 0.9923,
      "eval_R2": 0.9845,
      "eval_RL": 0.9918,
      "eval_RLsum": 0.9918,
      "eval_acc": 0.6677,
      "eval_loss": 0.01687614992260933,
      "eval_runtime": 1464.9901,
      "eval_samples_per_second": 15.991,
      "eval_steps_per_second": 7.995,
      "step": 17300
    },
    {
      "epoch": 0.15,
      "eval_Final acc": 0.9598309570562623,
      "eval_R1": 0.9923,
      "eval_R2": 0.9845,
      "eval_RL": 0.9918,
      "eval_RLsum": 0.9918,
      "eval_acc": 0.6693,
      "eval_loss": 0.016489630565047264,
      "eval_runtime": 1459.8108,
      "eval_samples_per_second": 16.047,
      "eval_steps_per_second": 8.024,
      "step": 17400
    },
    {
      "epoch": 0.15,
      "grad_norm": 0.271484375,
      "learning_rate": 4.936339848620551e-05,
      "loss": 0.0086,
      "step": 17500
    },
    {
      "epoch": 0.15,
      "eval_Final acc": 0.9530009391274652,
      "eval_R1": 0.9921,
      "eval_R2": 0.9842,
      "eval_RL": 0.9916,
      "eval_RLsum": 0.9916,
      "eval_acc": 0.6529,
      "eval_loss": 0.017549073323607445,
      "eval_runtime": 1462.3921,
      "eval_samples_per_second": 16.019,
      "eval_steps_per_second": 8.009,
      "step": 17500
    },
    {
      "epoch": 0.15,
      "eval_Final acc": 0.9570989498847434,
      "eval_R1": 0.9922,
      "eval_R2": 0.9845,
      "eval_RL": 0.9917,
      "eval_RLsum": 0.9917,
      "eval_acc": 0.6663,
      "eval_loss": 0.01672600954771042,
      "eval_runtime": 1451.0259,
      "eval_samples_per_second": 16.144,
      "eval_steps_per_second": 8.072,
      "step": 17600
    },
    {
      "epoch": 0.16,
      "eval_Final acc": 0.9632032784086059,
      "eval_R1": 0.9924,
      "eval_R2": 0.9849,
      "eval_RL": 0.992,
      "eval_RLsum": 0.992,
      "eval_acc": 0.6742,
      "eval_loss": 0.016187025234103203,
      "eval_runtime": 1452.7428,
      "eval_samples_per_second": 16.125,
      "eval_steps_per_second": 8.063,
      "step": 17700
    },
    {
      "epoch": 0.16,
      "eval_Final acc": 0.9606420216853069,
      "eval_R1": 0.9925,
      "eval_R2": 0.9851,
      "eval_RL": 0.9921,
      "eval_RLsum": 0.9921,
      "eval_acc": 0.6739,
      "eval_loss": 0.016911525279283524,
      "eval_runtime": 1462.6471,
      "eval_samples_per_second": 16.016,
      "eval_steps_per_second": 8.008,
      "step": 17800
    },
    {
      "epoch": 0.16,
      "eval_Final acc": 0.9603858960129771,
      "eval_R1": 0.9924,
      "eval_R2": 0.9848,
      "eval_RL": 0.992,
      "eval_RLsum": 0.992,
      "eval_acc": 0.6718,
      "eval_loss": 0.015750443562865257,
      "eval_runtime": 1461.3035,
      "eval_samples_per_second": 16.031,
      "eval_steps_per_second": 8.015,
      "step": 17900
    },
    {
      "epoch": 0.16,
      "grad_norm": 0.373046875,
      "learning_rate": 4.9341446709867766e-05,
      "loss": 0.0083,
      "step": 18000
    },
    {
      "epoch": 0.16,
      "eval_Final acc": 0.958166140186118,
      "eval_R1": 0.9923,
      "eval_R2": 0.9847,
      "eval_RL": 0.9918,
      "eval_RLsum": 0.9919,
      "eval_acc": 0.6733,
      "eval_loss": 0.015923548489809036,
      "eval_runtime": 1462.3367,
      "eval_samples_per_second": 16.02,
      "eval_steps_per_second": 8.01,
      "step": 18000
    },
    {
      "epoch": 0.16,
      "eval_Final acc": 0.9608981473576368,
      "eval_R1": 0.9923,
      "eval_R2": 0.9846,
      "eval_RL": 0.9918,
      "eval_RLsum": 0.9918,
      "eval_acc": 0.6748,
      "eval_loss": 0.015872864052653313,
      "eval_runtime": 1462.417,
      "eval_samples_per_second": 16.019,
      "eval_steps_per_second": 8.009,
      "step": 18100
    },
    {
      "epoch": 0.16,
      "eval_Final acc": 0.9552633825663792,
      "eval_R1": 0.9924,
      "eval_R2": 0.9847,
      "eval_RL": 0.9919,
      "eval_RLsum": 0.9919,
      "eval_acc": 0.6726,
      "eval_loss": 0.01545502059161663,
      "eval_runtime": 1465.9123,
      "eval_samples_per_second": 15.98,
      "eval_steps_per_second": 7.99,
      "step": 18200
    },
    {
      "epoch": 0.16,
      "eval_Final acc": 0.9608981473576368,
      "eval_R1": 0.9925,
      "eval_R2": 0.985,
      "eval_RL": 0.9921,
      "eval_RLsum": 0.9921,
      "eval_acc": 0.675,
      "eval_loss": 0.015620005317032337,
      "eval_runtime": 1463.6929,
      "eval_samples_per_second": 16.005,
      "eval_steps_per_second": 8.002,
      "step": 18300
    },
    {
      "epoch": 0.16,
      "eval_Final acc": 0.960428583625032,
      "eval_R1": 0.9925,
      "eval_R2": 0.985,
      "eval_RL": 0.9921,
      "eval_RLsum": 0.9921,
      "eval_acc": 0.6786,
      "eval_loss": 0.015367608517408371,
      "eval_runtime": 1462.0373,
      "eval_samples_per_second": 16.023,
      "eval_steps_per_second": 8.011,
      "step": 18400
    },
    {
      "epoch": 0.16,
      "grad_norm": 0.400390625,
      "learning_rate": 4.9319494933530025e-05,
      "loss": 0.0083,
      "step": 18500
    },
    {
      "epoch": 0.16,
      "eval_Final acc": 0.9613677110902417,
      "eval_R1": 0.9926,
      "eval_R2": 0.9853,
      "eval_RL": 0.9922,
      "eval_RLsum": 0.9922,
      "eval_acc": 0.6787,
      "eval_loss": 0.01588550955057144,
      "eval_runtime": 1464.8106,
      "eval_samples_per_second": 15.993,
      "eval_steps_per_second": 7.996,
      "step": 18500
    },
    {
      "epoch": 0.16,
      "eval_Final acc": 0.9643558439340904,
      "eval_R1": 0.9927,
      "eval_R2": 0.9854,
      "eval_RL": 0.9923,
      "eval_RLsum": 0.9923,
      "eval_acc": 0.6848,
      "eval_loss": 0.01595507375895977,
      "eval_runtime": 1451.6879,
      "eval_samples_per_second": 16.137,
      "eval_steps_per_second": 8.069,
      "step": 18600
    },
    {
      "epoch": 0.16,
      "eval_Final acc": 0.9656791599077947,
      "eval_R1": 0.9928,
      "eval_R2": 0.9856,
      "eval_RL": 0.9924,
      "eval_RLsum": 0.9924,
      "eval_acc": 0.6849,
      "eval_loss": 0.01566305197775364,
      "eval_runtime": 1465.7513,
      "eval_samples_per_second": 15.982,
      "eval_steps_per_second": 7.991,
      "step": 18700
    },
    {
      "epoch": 0.16,
      "eval_Final acc": 0.9619653376590114,
      "eval_R1": 0.9926,
      "eval_R2": 0.9853,
      "eval_RL": 0.9923,
      "eval_RLsum": 0.9923,
      "eval_acc": 0.6763,
      "eval_loss": 0.015803711488842964,
      "eval_runtime": 1462.0691,
      "eval_samples_per_second": 16.022,
      "eval_steps_per_second": 8.011,
      "step": 18800
    },
    {
      "epoch": 0.17,
      "eval_Final acc": 0.9595748313839324,
      "eval_R1": 0.9926,
      "eval_R2": 0.9854,
      "eval_RL": 0.9922,
      "eval_RLsum": 0.9922,
      "eval_acc": 0.6855,
      "eval_loss": 0.01540306955575943,
      "eval_runtime": 1468.5007,
      "eval_samples_per_second": 15.952,
      "eval_steps_per_second": 7.976,
      "step": 18900
    },
    {
      "epoch": 0.17,
      "grad_norm": 0.1865234375,
      "learning_rate": 4.9297543157192284e-05,
      "loss": 0.0074,
      "step": 19000
    },
    {
      "epoch": 0.17,
      "eval_Final acc": 0.9589345172031076,
      "eval_R1": 0.9927,
      "eval_R2": 0.9855,
      "eval_RL": 0.9923,
      "eval_RLsum": 0.9924,
      "eval_acc": 0.6815,
      "eval_loss": 0.015605304390192032,
      "eval_runtime": 1461.1003,
      "eval_samples_per_second": 16.033,
      "eval_steps_per_second": 8.017,
      "step": 19000
    },
    {
      "epoch": 0.17,
      "eval_Final acc": 0.9629898403483309,
      "eval_R1": 0.9927,
      "eval_R2": 0.9856,
      "eval_RL": 0.9923,
      "eval_RLsum": 0.9923,
      "eval_acc": 0.6774,
      "eval_loss": 0.015724962577223778,
      "eval_runtime": 1451.6079,
      "eval_samples_per_second": 16.138,
      "eval_steps_per_second": 8.069,
      "step": 19100
    },
    {
      "epoch": 0.17,
      "eval_Final acc": 0.9646546572184752,
      "eval_R1": 0.9924,
      "eval_R2": 0.985,
      "eval_RL": 0.9921,
      "eval_RLsum": 0.9921,
      "eval_acc": 0.6785,
      "eval_loss": 0.01578380912542343,
      "eval_runtime": 1463.3246,
      "eval_samples_per_second": 16.009,
      "eval_steps_per_second": 8.004,
      "step": 19200
    },
    {
      "epoch": 0.17,
      "eval_Final acc": 0.9619226500469563,
      "eval_R1": 0.9927,
      "eval_R2": 0.9854,
      "eval_RL": 0.9923,
      "eval_RLsum": 0.9923,
      "eval_acc": 0.6812,
      "eval_loss": 0.01560690812766552,
      "eval_runtime": 1451.0867,
      "eval_samples_per_second": 16.144,
      "eval_steps_per_second": 8.072,
      "step": 19300
    },
    {
      "epoch": 0.17,
      "eval_Final acc": 0.9591479552633826,
      "eval_R1": 0.9924,
      "eval_R2": 0.985,
      "eval_RL": 0.992,
      "eval_RLsum": 0.992,
      "eval_acc": 0.6785,
      "eval_loss": 0.015781858935952187,
      "eval_runtime": 1461.9851,
      "eval_samples_per_second": 16.023,
      "eval_steps_per_second": 8.012,
      "step": 19400
    },
    {
      "epoch": 0.17,
      "grad_norm": 1.3515625,
      "learning_rate": 4.927559138085454e-05,
      "loss": 0.0075,
      "step": 19500
    },
    {
      "epoch": 0.17,
      "eval_Final acc": 0.9643131563220353,
      "eval_R1": 0.9926,
      "eval_R2": 0.9854,
      "eval_RL": 0.9923,
      "eval_RLsum": 0.9923,
      "eval_acc": 0.6828,
      "eval_loss": 0.015539612621068954,
      "eval_runtime": 1450.2843,
      "eval_samples_per_second": 16.153,
      "eval_steps_per_second": 8.076,
      "step": 19500
    },
    {
      "epoch": 0.17,
      "eval_Final acc": 0.9615384615384616,
      "eval_R1": 0.9926,
      "eval_R2": 0.9853,
      "eval_RL": 0.9922,
      "eval_RLsum": 0.9922,
      "eval_acc": 0.6776,
      "eval_loss": 0.01577220857143402,
      "eval_runtime": 1450.3034,
      "eval_samples_per_second": 16.152,
      "eval_steps_per_second": 8.076,
      "step": 19600
    },
    {
      "epoch": 0.17,
      "eval_Final acc": 0.967600102450269,
      "eval_R1": 0.9925,
      "eval_R2": 0.9851,
      "eval_RL": 0.9922,
      "eval_RLsum": 0.9922,
      "eval_acc": 0.6807,
      "eval_loss": 0.01566699519753456,
      "eval_runtime": 1462.0717,
      "eval_samples_per_second": 16.022,
      "eval_steps_per_second": 8.011,
      "step": 19700
    },
    {
      "epoch": 0.17,
      "eval_Final acc": 0.9692222317083582,
      "eval_R1": 0.9924,
      "eval_R2": 0.985,
      "eval_RL": 0.9921,
      "eval_RLsum": 0.9921,
      "eval_acc": 0.6781,
      "eval_loss": 0.01571837067604065,
      "eval_runtime": 1460.0425,
      "eval_samples_per_second": 16.045,
      "eval_steps_per_second": 8.022,
      "step": 19800
    },
    {
      "epoch": 0.17,
      "eval_Final acc": 0.9638009049773756,
      "eval_R1": 0.9925,
      "eval_R2": 0.985,
      "eval_RL": 0.9921,
      "eval_RLsum": 0.9921,
      "eval_acc": 0.6783,
      "eval_loss": 0.015627918764948845,
      "eval_runtime": 1461.8926,
      "eval_samples_per_second": 16.024,
      "eval_steps_per_second": 8.012,
      "step": 19900
    },
    {
      "epoch": 0.18,
      "grad_norm": 0.2470703125,
      "learning_rate": 4.92536396045168e-05,
      "loss": 0.0076,
      "step": 20000
    },
    {
      "epoch": 0.18,
      "eval_Final acc": 0.9644412191582002,
      "eval_R1": 0.9922,
      "eval_R2": 0.9846,
      "eval_RL": 0.9918,
      "eval_RLsum": 0.9918,
      "eval_acc": 0.6752,
      "eval_loss": 0.015452449209988117,
      "eval_runtime": 1466.0438,
      "eval_samples_per_second": 15.979,
      "eval_steps_per_second": 7.99,
      "step": 20000
    },
    {
      "epoch": 0.18,
      "eval_Final acc": 0.958294203022283,
      "eval_R1": 0.992,
      "eval_R2": 0.9844,
      "eval_RL": 0.9917,
      "eval_RLsum": 0.9917,
      "eval_acc": 0.6706,
      "eval_loss": 0.01574222557246685,
      "eval_runtime": 1455.4379,
      "eval_samples_per_second": 16.095,
      "eval_steps_per_second": 8.048,
      "step": 20100
    },
    {
      "epoch": 0.18,
      "eval_Final acc": 0.9609408349696918,
      "eval_R1": 0.9923,
      "eval_R2": 0.9849,
      "eval_RL": 0.992,
      "eval_RLsum": 0.992,
      "eval_acc": 0.6742,
      "eval_loss": 0.014955781400203705,
      "eval_runtime": 1464.5448,
      "eval_samples_per_second": 15.995,
      "eval_steps_per_second": 7.998,
      "step": 20200
    },
    {
      "epoch": 0.18,
      "eval_Final acc": 0.9608554597455818,
      "eval_R1": 0.9926,
      "eval_R2": 0.9852,
      "eval_RL": 0.9922,
      "eval_RLsum": 0.9922,
      "eval_acc": 0.6778,
      "eval_loss": 0.015123295597732067,
      "eval_runtime": 1470.0029,
      "eval_samples_per_second": 15.936,
      "eval_steps_per_second": 7.968,
      "step": 20300
    },
    {
      "epoch": 0.18,
      "eval_Final acc": 0.9660633484162896,
      "eval_R1": 0.9926,
      "eval_R2": 0.9853,
      "eval_RL": 0.9922,
      "eval_RLsum": 0.9922,
      "eval_acc": 0.6779,
      "eval_loss": 0.014950262382626534,
      "eval_runtime": 1453.4102,
      "eval_samples_per_second": 16.118,
      "eval_steps_per_second": 8.059,
      "step": 20400
    },
    {
      "epoch": 0.18,
      "grad_norm": 0.341796875,
      "learning_rate": 4.923168782817906e-05,
      "loss": 0.0072,
      "step": 20500
    },
    {
      "epoch": 0.18,
      "eval_Final acc": 0.9644412191582002,
      "eval_R1": 0.9926,
      "eval_R2": 0.9852,
      "eval_RL": 0.9922,
      "eval_RLsum": 0.9922,
      "eval_acc": 0.6775,
      "eval_loss": 0.015031445771455765,
      "eval_runtime": 1453.6013,
      "eval_samples_per_second": 16.116,
      "eval_steps_per_second": 8.058,
      "step": 20500
    },
    {
      "epoch": 0.18,
      "eval_Final acc": 0.9700759839494578,
      "eval_R1": 0.9927,
      "eval_R2": 0.9855,
      "eval_RL": 0.9924,
      "eval_RLsum": 0.9924,
      "eval_acc": 0.683,
      "eval_loss": 0.01463630422949791,
      "eval_runtime": 1467.7451,
      "eval_samples_per_second": 15.961,
      "eval_steps_per_second": 7.98,
      "step": 20600
    },
    {
      "epoch": 0.18,
      "eval_Final acc": 0.9696917954409631,
      "eval_R1": 0.9926,
      "eval_R2": 0.9852,
      "eval_RL": 0.9922,
      "eval_RLsum": 0.9922,
      "eval_acc": 0.6795,
      "eval_loss": 0.014612710103392601,
      "eval_runtime": 1456.4539,
      "eval_samples_per_second": 16.084,
      "eval_steps_per_second": 8.042,
      "step": 20700
    },
    {
      "epoch": 0.18,
      "eval_Final acc": 0.9676427900623239,
      "eval_R1": 0.9927,
      "eval_R2": 0.9853,
      "eval_RL": 0.9923,
      "eval_RLsum": 0.9923,
      "eval_acc": 0.6751,
      "eval_loss": 0.014858444221317768,
      "eval_runtime": 1474.1963,
      "eval_samples_per_second": 15.891,
      "eval_steps_per_second": 7.945,
      "step": 20800
    },
    {
      "epoch": 0.18,
      "eval_Final acc": 0.9708016733543926,
      "eval_R1": 0.9927,
      "eval_R2": 0.9854,
      "eval_RL": 0.9924,
      "eval_RLsum": 0.9924,
      "eval_acc": 0.6729,
      "eval_loss": 0.014847170561552048,
      "eval_runtime": 1467.518,
      "eval_samples_per_second": 15.963,
      "eval_steps_per_second": 7.982,
      "step": 20900
    },
    {
      "epoch": 0.18,
      "grad_norm": 0.2353515625,
      "learning_rate": 4.920973605184131e-05,
      "loss": 0.0076,
      "step": 21000
    },
    {
      "epoch": 0.18,
      "eval_Final acc": 0.9703747972338428,
      "eval_R1": 0.9927,
      "eval_R2": 0.9856,
      "eval_RL": 0.9924,
      "eval_RLsum": 0.9924,
      "eval_acc": 0.6745,
      "eval_loss": 0.01479652151465416,
      "eval_runtime": 1472.683,
      "eval_samples_per_second": 15.907,
      "eval_steps_per_second": 7.954,
      "step": 21000
    },
    {
      "epoch": 0.18,
      "eval_Final acc": 0.9728079911209767,
      "eval_R1": 0.9927,
      "eval_R2": 0.9855,
      "eval_RL": 0.9924,
      "eval_RLsum": 0.9924,
      "eval_acc": 0.6805,
      "eval_loss": 0.014527148567140102,
      "eval_runtime": 1467.6556,
      "eval_samples_per_second": 15.962,
      "eval_steps_per_second": 7.981,
      "step": 21100
    },
    {
      "epoch": 0.19,
      "eval_Final acc": 0.9738324938102962,
      "eval_R1": 0.9929,
      "eval_R2": 0.9859,
      "eval_RL": 0.9926,
      "eval_RLsum": 0.9926,
      "eval_acc": 0.6839,
      "eval_loss": 0.014385683462023735,
      "eval_runtime": 1456.2262,
      "eval_samples_per_second": 16.087,
      "eval_steps_per_second": 8.043,
      "step": 21200
    },
    {
      "epoch": 0.19,
      "eval_Final acc": 0.9712285494749424,
      "eval_R1": 0.9927,
      "eval_R2": 0.9856,
      "eval_RL": 0.9924,
      "eval_RLsum": 0.9924,
      "eval_acc": 0.6847,
      "eval_loss": 0.014538643881678581,
      "eval_runtime": 1455.2082,
      "eval_samples_per_second": 16.098,
      "eval_steps_per_second": 8.049,
      "step": 21300
    },
    {
      "epoch": 0.19,
      "eval_Final acc": 0.9768633142662,
      "eval_R1": 0.9927,
      "eval_R2": 0.9856,
      "eval_RL": 0.9925,
      "eval_RLsum": 0.9925,
      "eval_acc": 0.6839,
      "eval_loss": 0.014480886980891228,
      "eval_runtime": 1464.5231,
      "eval_samples_per_second": 15.996,
      "eval_steps_per_second": 7.998,
      "step": 21400
    },
    {
      "epoch": 0.19,
      "grad_norm": 0.26171875,
      "learning_rate": 4.918778427550357e-05,
      "loss": 0.0074,
      "step": 21500
    },
    {
      "epoch": 0.19,
      "eval_Final acc": 0.974387432767011,
      "eval_R1": 0.9927,
      "eval_R2": 0.9855,
      "eval_RL": 0.9925,
      "eval_RLsum": 0.9925,
      "eval_acc": 0.6786,
      "eval_loss": 0.014647131785750389,
      "eval_runtime": 1454.7879,
      "eval_samples_per_second": 16.103,
      "eval_steps_per_second": 8.051,
      "step": 21500
    },
    {
      "epoch": 0.19,
      "eval_Final acc": 0.9725091778365919,
      "eval_R1": 0.9928,
      "eval_R2": 0.9858,
      "eval_RL": 0.9926,
      "eval_RLsum": 0.9926,
      "eval_acc": 0.6848,
      "eval_loss": 0.01433196198195219,
      "eval_runtime": 1455.1142,
      "eval_samples_per_second": 16.099,
      "eval_steps_per_second": 8.05,
      "step": 21600
    },
    {
      "epoch": 0.19,
      "eval_Final acc": 0.9748569964996158,
      "eval_R1": 0.9929,
      "eval_R2": 0.9859,
      "eval_RL": 0.9927,
      "eval_RLsum": 0.9927,
      "eval_acc": 0.6846,
      "eval_loss": 0.014561287127435207,
      "eval_runtime": 1454.6349,
      "eval_samples_per_second": 16.104,
      "eval_steps_per_second": 8.052,
      "step": 21700
    },
    {
      "epoch": 0.19,
      "eval_Final acc": 0.9721676769401519,
      "eval_R1": 0.9927,
      "eval_R2": 0.9856,
      "eval_RL": 0.9925,
      "eval_RLsum": 0.9925,
      "eval_acc": 0.6847,
      "eval_loss": 0.014359143562614918,
      "eval_runtime": 1468.7397,
      "eval_samples_per_second": 15.95,
      "eval_steps_per_second": 7.975,
      "step": 21800
    },
    {
      "epoch": 0.19,
      "eval_Final acc": 0.974472807991121,
      "eval_R1": 0.9928,
      "eval_R2": 0.9859,
      "eval_RL": 0.9926,
      "eval_RLsum": 0.9926,
      "eval_acc": 0.6841,
      "eval_loss": 0.01475589256733656,
      "eval_runtime": 1473.6914,
      "eval_samples_per_second": 15.896,
      "eval_steps_per_second": 7.948,
      "step": 21900
    },
    {
      "epoch": 0.19,
      "grad_norm": 0.11962890625,
      "learning_rate": 4.916583249916584e-05,
      "loss": 0.0067,
      "step": 22000
    },
    {
      "epoch": 0.19,
      "eval_Final acc": 0.9737471185861862,
      "eval_R1": 0.9927,
      "eval_R2": 0.9857,
      "eval_RL": 0.9925,
      "eval_RLsum": 0.9925,
      "eval_acc": 0.6866,
      "eval_loss": 0.01425244938582182,
      "eval_runtime": 1469.1152,
      "eval_samples_per_second": 15.946,
      "eval_steps_per_second": 7.973,
      "step": 22000
    },
    {
      "epoch": 0.19,
      "eval_Final acc": 0.9729787415691966,
      "eval_R1": 0.9927,
      "eval_R2": 0.9856,
      "eval_RL": 0.9925,
      "eval_RLsum": 0.9925,
      "eval_acc": 0.6846,
      "eval_loss": 0.014375461265444756,
      "eval_runtime": 1465.8673,
      "eval_samples_per_second": 15.981,
      "eval_steps_per_second": 7.99,
      "step": 22100
    },
    {
      "epoch": 0.19,
      "eval_Final acc": 0.9739605566464612,
      "eval_R1": 0.9927,
      "eval_R2": 0.9856,
      "eval_RL": 0.9925,
      "eval_RLsum": 0.9925,
      "eval_acc": 0.6869,
      "eval_loss": 0.014116480015218258,
      "eval_runtime": 1467.7642,
      "eval_samples_per_second": 15.96,
      "eval_steps_per_second": 7.98,
      "step": 22200
    },
    {
      "epoch": 0.2,
      "eval_Final acc": 0.9754973106804405,
      "eval_R1": 0.9927,
      "eval_R2": 0.9857,
      "eval_RL": 0.9925,
      "eval_RLsum": 0.9925,
      "eval_acc": 0.6877,
      "eval_loss": 0.014198748394846916,
      "eval_runtime": 1467.0445,
      "eval_samples_per_second": 15.968,
      "eval_steps_per_second": 7.984,
      "step": 22300
    },
    {
      "epoch": 0.2,
      "eval_Final acc": 0.9761376248612652,
      "eval_R1": 0.9928,
      "eval_R2": 0.9858,
      "eval_RL": 0.9925,
      "eval_RLsum": 0.9925,
      "eval_acc": 0.688,
      "eval_loss": 0.014363715425133705,
      "eval_runtime": 1467.5905,
      "eval_samples_per_second": 15.962,
      "eval_steps_per_second": 7.981,
      "step": 22400
    },
    {
      "epoch": 0.2,
      "grad_norm": 0.263671875,
      "learning_rate": 4.9143880722828096e-05,
      "loss": 0.0068,
      "step": 22500
    },
    {
      "epoch": 0.2,
      "eval_Final acc": 0.9769913771023649,
      "eval_R1": 0.9928,
      "eval_R2": 0.986,
      "eval_RL": 0.9926,
      "eval_RLsum": 0.9926,
      "eval_acc": 0.6934,
      "eval_loss": 0.014523769728839397,
      "eval_runtime": 1466.4197,
      "eval_samples_per_second": 15.975,
      "eval_steps_per_second": 7.987,
      "step": 22500
    },
    {
      "epoch": 0.2,
      "eval_Final acc": 0.9765645009818151,
      "eval_R1": 0.9928,
      "eval_R2": 0.9858,
      "eval_RL": 0.9926,
      "eval_RLsum": 0.9926,
      "eval_acc": 0.6859,
      "eval_loss": 0.013950865715742111,
      "eval_runtime": 1467.1214,
      "eval_samples_per_second": 15.967,
      "eval_steps_per_second": 7.984,
      "step": 22600
    },
    {
      "epoch": 0.2,
      "eval_Final acc": 0.9774609408349697,
      "eval_R1": 0.993,
      "eval_R2": 0.986,
      "eval_RL": 0.9927,
      "eval_RLsum": 0.9927,
      "eval_acc": 0.6888,
      "eval_loss": 0.014014782384037971,
      "eval_runtime": 1465.5297,
      "eval_samples_per_second": 15.985,
      "eval_steps_per_second": 7.992,
      "step": 22700
    },
    {
      "epoch": 0.2,
      "eval_Final acc": 0.9723811150004269,
      "eval_R1": 0.9928,
      "eval_R2": 0.9857,
      "eval_RL": 0.9925,
      "eval_RLsum": 0.9926,
      "eval_acc": 0.6854,
      "eval_loss": 0.013970132917165756,
      "eval_runtime": 1455.0288,
      "eval_samples_per_second": 16.1,
      "eval_steps_per_second": 8.05,
      "step": 22800
    },
    {
      "epoch": 0.2,
      "eval_Final acc": 0.9756253735166055,
      "eval_R1": 0.9929,
      "eval_R2": 0.986,
      "eval_RL": 0.9927,
      "eval_RLsum": 0.9927,
      "eval_acc": 0.6906,
      "eval_loss": 0.013664424419403076,
      "eval_runtime": 1465.5051,
      "eval_samples_per_second": 15.985,
      "eval_steps_per_second": 7.992,
      "step": 22900
    },
    {
      "epoch": 0.2,
      "grad_norm": 0.08447265625,
      "learning_rate": 4.9121928946490354e-05,
      "loss": 0.0071,
      "step": 23000
    },
    {
      "epoch": 0.2,
      "eval_Final acc": 0.9722530521642619,
      "eval_R1": 0.9928,
      "eval_R2": 0.9857,
      "eval_RL": 0.9925,
      "eval_RLsum": 0.9925,
      "eval_acc": 0.6858,
      "eval_loss": 0.013724473305046558,
      "eval_runtime": 1466.5841,
      "eval_samples_per_second": 15.973,
      "eval_steps_per_second": 7.987,
      "step": 23000
    },
    {
      "epoch": 0.2,
      "eval_Final acc": 0.9757107487407154,
      "eval_R1": 0.9929,
      "eval_R2": 0.9859,
      "eval_RL": 0.9926,
      "eval_RLsum": 0.9926,
      "eval_acc": 0.6891,
      "eval_loss": 0.013421189039945602,
      "eval_runtime": 1464.274,
      "eval_samples_per_second": 15.998,
      "eval_steps_per_second": 7.999,
      "step": 23100
    },
    {
      "epoch": 0.2,
      "eval_Final acc": 0.9771621275505848,
      "eval_R1": 0.9929,
      "eval_R2": 0.986,
      "eval_RL": 0.9927,
      "eval_RLsum": 0.9927,
      "eval_acc": 0.6886,
      "eval_loss": 0.01364926714450121,
      "eval_runtime": 1456.3856,
      "eval_samples_per_second": 16.085,
      "eval_steps_per_second": 8.043,
      "step": 23200
    },
    {
      "epoch": 0.2,
      "eval_Final acc": 0.9768633142662,
      "eval_R1": 0.9931,
      "eval_R2": 0.9862,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6912,
      "eval_loss": 0.01379765011370182,
      "eval_runtime": 1454.6457,
      "eval_samples_per_second": 16.104,
      "eval_steps_per_second": 8.052,
      "step": 23300
    },
    {
      "epoch": 0.2,
      "eval_Final acc": 0.9754546230683856,
      "eval_R1": 0.9929,
      "eval_R2": 0.9858,
      "eval_RL": 0.9926,
      "eval_RLsum": 0.9926,
      "eval_acc": 0.6875,
      "eval_loss": 0.013711774721741676,
      "eval_runtime": 1466.1836,
      "eval_samples_per_second": 15.978,
      "eval_steps_per_second": 7.989,
      "step": 23400
    },
    {
      "epoch": 0.21,
      "grad_norm": 0.1943359375,
      "learning_rate": 4.909997717015261e-05,
      "loss": 0.0061,
      "step": 23500
    },
    {
      "epoch": 0.21,
      "eval_Final acc": 0.9769913771023649,
      "eval_R1": 0.9929,
      "eval_R2": 0.9858,
      "eval_RL": 0.9926,
      "eval_RLsum": 0.9926,
      "eval_acc": 0.6913,
      "eval_loss": 0.01385662704706192,
      "eval_runtime": 1501.1497,
      "eval_samples_per_second": 15.605,
      "eval_steps_per_second": 7.803,
      "step": 23500
    },
    {
      "epoch": 0.21,
      "eval_Final acc": 0.978997694868949,
      "eval_R1": 0.993,
      "eval_R2": 0.9861,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6932,
      "eval_loss": 0.01365965511649847,
      "eval_runtime": 1503.0888,
      "eval_samples_per_second": 15.585,
      "eval_steps_per_second": 7.793,
      "step": 23600
    },
    {
      "epoch": 0.21,
      "eval_Final acc": 0.9801075727823786,
      "eval_R1": 0.9931,
      "eval_R2": 0.9861,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6933,
      "eval_loss": 0.013601411134004593,
      "eval_runtime": 1492.6715,
      "eval_samples_per_second": 15.694,
      "eval_steps_per_second": 7.847,
      "step": 23700
    },
    {
      "epoch": 0.21,
      "eval_Final acc": 0.9786561939725091,
      "eval_R1": 0.9931,
      "eval_R2": 0.9862,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6903,
      "eval_loss": 0.013755897991359234,
      "eval_runtime": 1463.4758,
      "eval_samples_per_second": 16.007,
      "eval_steps_per_second": 8.004,
      "step": 23800
    },
    {
      "epoch": 0.21,
      "eval_Final acc": 0.9795953214377188,
      "eval_R1": 0.993,
      "eval_R2": 0.986,
      "eval_RL": 0.9927,
      "eval_RLsum": 0.9927,
      "eval_acc": 0.6916,
      "eval_loss": 0.013785951770842075,
      "eval_runtime": 1460.93,
      "eval_samples_per_second": 16.035,
      "eval_steps_per_second": 8.017,
      "step": 23900
    },
    {
      "epoch": 0.21,
      "grad_norm": 0.0849609375,
      "learning_rate": 4.907802539381487e-05,
      "loss": 0.0066,
      "step": 24000
    },
    {
      "epoch": 0.21,
      "eval_Final acc": 0.979253820541279,
      "eval_R1": 0.993,
      "eval_R2": 0.9861,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6934,
      "eval_loss": 0.013886190950870514,
      "eval_runtime": 1462.5153,
      "eval_samples_per_second": 16.018,
      "eval_steps_per_second": 8.009,
      "step": 24000
    },
    {
      "epoch": 0.21,
      "eval_Final acc": 0.9777597541193546,
      "eval_R1": 0.9931,
      "eval_R2": 0.9861,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6924,
      "eval_loss": 0.013507825322449207,
      "eval_runtime": 1453.2744,
      "eval_samples_per_second": 16.119,
      "eval_steps_per_second": 8.06,
      "step": 24100
    },
    {
      "epoch": 0.21,
      "eval_Final acc": 0.9795099462136088,
      "eval_R1": 0.993,
      "eval_R2": 0.9861,
      "eval_RL": 0.9927,
      "eval_RLsum": 0.9927,
      "eval_acc": 0.6921,
      "eval_loss": 0.013588870875537395,
      "eval_runtime": 1464.1469,
      "eval_samples_per_second": 16.0,
      "eval_steps_per_second": 8.0,
      "step": 24200
    },
    {
      "epoch": 0.21,
      "eval_Final acc": 0.9776316912831896,
      "eval_R1": 0.9929,
      "eval_R2": 0.986,
      "eval_RL": 0.9927,
      "eval_RLsum": 0.9927,
      "eval_acc": 0.6898,
      "eval_loss": 0.013585035689175129,
      "eval_runtime": 1461.8917,
      "eval_samples_per_second": 16.024,
      "eval_steps_per_second": 8.012,
      "step": 24300
    },
    {
      "epoch": 0.21,
      "eval_Final acc": 0.9805771365149833,
      "eval_R1": 0.9931,
      "eval_R2": 0.9863,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6916,
      "eval_loss": 0.013763057999312878,
      "eval_runtime": 1463.0093,
      "eval_samples_per_second": 16.012,
      "eval_steps_per_second": 8.006,
      "step": 24400
    },
    {
      "epoch": 0.21,
      "grad_norm": 0.20703125,
      "learning_rate": 4.9056073617477124e-05,
      "loss": 0.0063,
      "step": 24500
    },
    {
      "epoch": 0.21,
      "eval_Final acc": 0.9798087594979937,
      "eval_R1": 0.993,
      "eval_R2": 0.9861,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6887,
      "eval_loss": 0.01329925749450922,
      "eval_runtime": 1465.4449,
      "eval_samples_per_second": 15.986,
      "eval_steps_per_second": 7.993,
      "step": 24500
    },
    {
      "epoch": 0.22,
      "eval_Final acc": 0.9804490736788184,
      "eval_R1": 0.9931,
      "eval_R2": 0.9862,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6927,
      "eval_loss": 0.013285613618791103,
      "eval_runtime": 1462.5507,
      "eval_samples_per_second": 16.017,
      "eval_steps_per_second": 8.009,
      "step": 24600
    },
    {
      "epoch": 0.22,
      "eval_Final acc": 0.9818150772645778,
      "eval_R1": 0.9931,
      "eval_R2": 0.9862,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6915,
      "eval_loss": 0.013250759802758694,
      "eval_runtime": 1461.3039,
      "eval_samples_per_second": 16.031,
      "eval_steps_per_second": 8.015,
      "step": 24700
    },
    {
      "epoch": 0.22,
      "eval_Final acc": 0.9810467002475881,
      "eval_R1": 0.993,
      "eval_R2": 0.9861,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6916,
      "eval_loss": 0.013293116353452206,
      "eval_runtime": 1452.0946,
      "eval_samples_per_second": 16.133,
      "eval_steps_per_second": 8.066,
      "step": 24800
    },
    {
      "epoch": 0.22,
      "eval_Final acc": 0.9806625117390934,
      "eval_R1": 0.993,
      "eval_R2": 0.9862,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6921,
      "eval_loss": 0.013502970337867737,
      "eval_runtime": 1462.0832,
      "eval_samples_per_second": 16.022,
      "eval_steps_per_second": 8.011,
      "step": 24900
    },
    {
      "epoch": 0.22,
      "grad_norm": 0.1953125,
      "learning_rate": 4.903412184113938e-05,
      "loss": 0.0059,
      "step": 25000
    },
    {
      "epoch": 0.22,
      "eval_Final acc": 0.9803210108426534,
      "eval_R1": 0.993,
      "eval_R2": 0.9863,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6932,
      "eval_loss": 0.01366698369383812,
      "eval_runtime": 1460.4792,
      "eval_samples_per_second": 16.04,
      "eval_steps_per_second": 8.02,
      "step": 25000
    },
    {
      "epoch": 0.22,
      "eval_Final acc": 0.9796380090497737,
      "eval_R1": 0.993,
      "eval_R2": 0.9861,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6957,
      "eval_loss": 0.013300583697855473,
      "eval_runtime": 1449.798,
      "eval_samples_per_second": 16.158,
      "eval_steps_per_second": 8.079,
      "step": 25100
    },
    {
      "epoch": 0.22,
      "eval_Final acc": 0.9799795099462136,
      "eval_R1": 0.993,
      "eval_R2": 0.9862,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.697,
      "eval_loss": 0.013188117183744907,
      "eval_runtime": 1452.5688,
      "eval_samples_per_second": 16.127,
      "eval_steps_per_second": 8.064,
      "step": 25200
    },
    {
      "epoch": 0.22,
      "eval_Final acc": 0.9822846409971826,
      "eval_R1": 0.9931,
      "eval_R2": 0.9864,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6916,
      "eval_loss": 0.013676137663424015,
      "eval_runtime": 1466.4976,
      "eval_samples_per_second": 15.974,
      "eval_steps_per_second": 7.987,
      "step": 25300
    },
    {
      "epoch": 0.22,
      "eval_Final acc": 0.9802783232305985,
      "eval_R1": 0.9931,
      "eval_R2": 0.9862,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6937,
      "eval_loss": 0.013502080924808979,
      "eval_runtime": 1468.8367,
      "eval_samples_per_second": 15.949,
      "eval_steps_per_second": 7.974,
      "step": 25400
    },
    {
      "epoch": 0.22,
      "grad_norm": 0.27734375,
      "learning_rate": 4.901217006480164e-05,
      "loss": 0.0065,
      "step": 25500
    },
    {
      "epoch": 0.22,
      "eval_Final acc": 0.9810467002475881,
      "eval_R1": 0.993,
      "eval_R2": 0.9862,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6961,
      "eval_loss": 0.013518081977963448,
      "eval_runtime": 1463.973,
      "eval_samples_per_second": 16.002,
      "eval_steps_per_second": 8.001,
      "step": 25500
    },
    {
      "epoch": 0.22,
      "eval_Final acc": 0.9814735763681379,
      "eval_R1": 0.9932,
      "eval_R2": 0.9864,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6941,
      "eval_loss": 0.013471427373588085,
      "eval_runtime": 1454.3619,
      "eval_samples_per_second": 16.107,
      "eval_steps_per_second": 8.054,
      "step": 25600
    },
    {
      "epoch": 0.23,
      "eval_Final acc": 0.9799795099462136,
      "eval_R1": 0.993,
      "eval_R2": 0.9861,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6939,
      "eval_loss": 0.013599563390016556,
      "eval_runtime": 1450.6791,
      "eval_samples_per_second": 16.148,
      "eval_steps_per_second": 8.074,
      "step": 25700
    },
    {
      "epoch": 0.23,
      "eval_Final acc": 0.9822846409971826,
      "eval_R1": 0.9931,
      "eval_R2": 0.9863,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6949,
      "eval_loss": 0.013590376824140549,
      "eval_runtime": 1452.6426,
      "eval_samples_per_second": 16.126,
      "eval_steps_per_second": 8.063,
      "step": 25800
    },
    {
      "epoch": 0.23,
      "eval_Final acc": 0.9819858277127977,
      "eval_R1": 0.9931,
      "eval_R2": 0.9863,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6904,
      "eval_loss": 0.01379045657813549,
      "eval_runtime": 1461.8335,
      "eval_samples_per_second": 16.025,
      "eval_steps_per_second": 8.013,
      "step": 25900
    },
    {
      "epoch": 0.23,
      "grad_norm": 0.2275390625,
      "learning_rate": 4.899021828846391e-05,
      "loss": 0.0063,
      "step": 26000
    },
    {
      "epoch": 0.23,
      "eval_Final acc": 0.9830530180141723,
      "eval_R1": 0.9931,
      "eval_R2": 0.9863,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6927,
      "eval_loss": 0.013356298208236694,
      "eval_runtime": 1452.8591,
      "eval_samples_per_second": 16.124,
      "eval_steps_per_second": 8.062,
      "step": 26000
    },
    {
      "epoch": 0.23,
      "eval_Final acc": 0.9810040126355332,
      "eval_R1": 0.9931,
      "eval_R2": 0.9862,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6941,
      "eval_loss": 0.013375646434724331,
      "eval_runtime": 1451.5645,
      "eval_samples_per_second": 16.138,
      "eval_steps_per_second": 8.069,
      "step": 26100
    },
    {
      "epoch": 0.23,
      "eval_Final acc": 0.9822419533851277,
      "eval_R1": 0.9933,
      "eval_R2": 0.9866,
      "eval_RL": 0.9931,
      "eval_RLsum": 0.9931,
      "eval_acc": 0.6951,
      "eval_loss": 0.013354193419218063,
      "eval_runtime": 1450.6751,
      "eval_samples_per_second": 16.148,
      "eval_steps_per_second": 8.074,
      "step": 26200
    },
    {
      "epoch": 0.23,
      "eval_Final acc": 0.9815589515922479,
      "eval_R1": 0.9932,
      "eval_R2": 0.9863,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6925,
      "eval_loss": 0.013482308015227318,
      "eval_runtime": 1452.6613,
      "eval_samples_per_second": 16.126,
      "eval_steps_per_second": 8.063,
      "step": 26300
    },
    {
      "epoch": 0.23,
      "eval_Final acc": 0.979168445317169,
      "eval_R1": 0.993,
      "eval_R2": 0.9862,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6951,
      "eval_loss": 0.013421417213976383,
      "eval_runtime": 1461.2158,
      "eval_samples_per_second": 16.032,
      "eval_steps_per_second": 8.016,
      "step": 26400
    },
    {
      "epoch": 0.23,
      "grad_norm": 0.2490234375,
      "learning_rate": 4.8968266512126166e-05,
      "loss": 0.0061,
      "step": 26500
    },
    {
      "epoch": 0.23,
      "eval_Final acc": 0.9811747630837531,
      "eval_R1": 0.9931,
      "eval_R2": 0.9862,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6926,
      "eval_loss": 0.013355622068047523,
      "eval_runtime": 1463.7882,
      "eval_samples_per_second": 16.004,
      "eval_steps_per_second": 8.002,
      "step": 26500
    },
    {
      "epoch": 0.23,
      "eval_Final acc": 0.9818150772645778,
      "eval_R1": 0.9932,
      "eval_R2": 0.9864,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.691,
      "eval_loss": 0.013367006555199623,
      "eval_runtime": 1452.3069,
      "eval_samples_per_second": 16.13,
      "eval_steps_per_second": 8.065,
      "step": 26600
    },
    {
      "epoch": 0.23,
      "eval_Final acc": 0.981345513531973,
      "eval_R1": 0.9932,
      "eval_R2": 0.9864,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6916,
      "eval_loss": 0.013578394427895546,
      "eval_runtime": 1451.8922,
      "eval_samples_per_second": 16.135,
      "eval_steps_per_second": 8.067,
      "step": 26700
    },
    {
      "epoch": 0.23,
      "eval_Final acc": 0.9811320754716981,
      "eval_R1": 0.9932,
      "eval_R2": 0.9864,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6907,
      "eval_loss": 0.013182532973587513,
      "eval_runtime": 1451.294,
      "eval_samples_per_second": 16.141,
      "eval_steps_per_second": 8.071,
      "step": 26800
    },
    {
      "epoch": 0.24,
      "eval_Final acc": 0.979253820541279,
      "eval_R1": 0.993,
      "eval_R2": 0.9863,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6953,
      "eval_loss": 0.013427864760160446,
      "eval_runtime": 1452.6694,
      "eval_samples_per_second": 16.126,
      "eval_steps_per_second": 8.063,
      "step": 26900
    },
    {
      "epoch": 0.24,
      "grad_norm": 0.1435546875,
      "learning_rate": 4.8946314735788425e-05,
      "loss": 0.0065,
      "step": 27000
    },
    {
      "epoch": 0.24,
      "eval_Final acc": 0.9783573806881243,
      "eval_R1": 0.993,
      "eval_R2": 0.9861,
      "eval_RL": 0.9927,
      "eval_RLsum": 0.9927,
      "eval_acc": 0.6921,
      "eval_loss": 0.012859231792390347,
      "eval_runtime": 1452.2213,
      "eval_samples_per_second": 16.131,
      "eval_steps_per_second": 8.066,
      "step": 27000
    },
    {
      "epoch": 0.24,
      "eval_Final acc": 0.9807051993511483,
      "eval_R1": 0.9932,
      "eval_R2": 0.9865,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6953,
      "eval_loss": 0.01281033270061016,
      "eval_runtime": 1452.6557,
      "eval_samples_per_second": 16.126,
      "eval_steps_per_second": 8.063,
      "step": 27100
    },
    {
      "epoch": 0.24,
      "eval_Final acc": 0.978912319644839,
      "eval_R1": 0.9931,
      "eval_R2": 0.9863,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6935,
      "eval_loss": 0.013128886930644512,
      "eval_runtime": 1463.4582,
      "eval_samples_per_second": 16.007,
      "eval_steps_per_second": 8.004,
      "step": 27200
    },
    {
      "epoch": 0.24,
      "eval_Final acc": 0.9796806966618288,
      "eval_R1": 0.993,
      "eval_R2": 0.9863,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6964,
      "eval_loss": 0.013543019071221352,
      "eval_runtime": 1452.6081,
      "eval_samples_per_second": 16.127,
      "eval_steps_per_second": 8.063,
      "step": 27300
    },
    {
      "epoch": 0.24,
      "eval_Final acc": 0.9782720054640144,
      "eval_R1": 0.9931,
      "eval_R2": 0.9863,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6947,
      "eval_loss": 0.01297728531062603,
      "eval_runtime": 1464.2327,
      "eval_samples_per_second": 15.999,
      "eval_steps_per_second": 7.999,
      "step": 27400
    },
    {
      "epoch": 0.24,
      "grad_norm": 0.0008544921875,
      "learning_rate": 4.8924362959450684e-05,
      "loss": 0.006,
      "step": 27500
    },
    {
      "epoch": 0.24,
      "eval_Final acc": 0.978912319644839,
      "eval_R1": 0.9932,
      "eval_R2": 0.9864,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6931,
      "eval_loss": 0.012972823344171047,
      "eval_runtime": 1464.3665,
      "eval_samples_per_second": 15.997,
      "eval_steps_per_second": 7.999,
      "step": 27500
    },
    {
      "epoch": 0.24,
      "eval_Final acc": 0.979083070093059,
      "eval_R1": 0.9932,
      "eval_R2": 0.9865,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6966,
      "eval_loss": 0.012808305211365223,
      "eval_runtime": 1452.1504,
      "eval_samples_per_second": 16.132,
      "eval_steps_per_second": 8.066,
      "step": 27600
    },
    {
      "epoch": 0.24,
      "eval_Final acc": 0.9788269444207292,
      "eval_R1": 0.993,
      "eval_R2": 0.9863,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6973,
      "eval_loss": 0.012933765538036823,
      "eval_runtime": 1450.4379,
      "eval_samples_per_second": 16.151,
      "eval_steps_per_second": 8.075,
      "step": 27700
    },
    {
      "epoch": 0.24,
      "eval_Final acc": 0.9800221975582686,
      "eval_R1": 0.993,
      "eval_R2": 0.9863,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.695,
      "eval_loss": 0.013324824161827564,
      "eval_runtime": 1464.3177,
      "eval_samples_per_second": 15.998,
      "eval_steps_per_second": 7.999,
      "step": 27800
    },
    {
      "epoch": 0.24,
      "eval_Final acc": 0.9780585674037394,
      "eval_R1": 0.993,
      "eval_R2": 0.9862,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6966,
      "eval_loss": 0.012795492075383663,
      "eval_runtime": 1451.11,
      "eval_samples_per_second": 16.144,
      "eval_steps_per_second": 8.072,
      "step": 27900
    },
    {
      "epoch": 0.25,
      "grad_norm": 0.0011749267578125,
      "learning_rate": 4.890241118311294e-05,
      "loss": 0.006,
      "step": 28000
    },
    {
      "epoch": 0.25,
      "eval_Final acc": 0.9779305045675745,
      "eval_R1": 0.9932,
      "eval_R2": 0.9864,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6965,
      "eval_loss": 0.012530622072517872,
      "eval_runtime": 1451.3978,
      "eval_samples_per_second": 16.14,
      "eval_steps_per_second": 8.07,
      "step": 28000
    },
    {
      "epoch": 0.25,
      "eval_Final acc": 0.9779731921796295,
      "eval_R1": 0.9931,
      "eval_R2": 0.9863,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.696,
      "eval_loss": 0.012704298831522465,
      "eval_runtime": 1469.1509,
      "eval_samples_per_second": 15.945,
      "eval_steps_per_second": 7.973,
      "step": 28100
    },
    {
      "epoch": 0.25,
      "eval_Final acc": 0.9779305045675745,
      "eval_R1": 0.9931,
      "eval_R2": 0.9865,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6971,
      "eval_loss": 0.012863822281360626,
      "eval_runtime": 1463.7862,
      "eval_samples_per_second": 16.004,
      "eval_steps_per_second": 8.002,
      "step": 28200
    },
    {
      "epoch": 0.25,
      "eval_Final acc": 0.979253820541279,
      "eval_R1": 0.9932,
      "eval_R2": 0.9865,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6965,
      "eval_loss": 0.013012239709496498,
      "eval_runtime": 1463.1868,
      "eval_samples_per_second": 16.01,
      "eval_steps_per_second": 8.005,
      "step": 28300
    },
    {
      "epoch": 0.25,
      "eval_Final acc": 0.9778878169555195,
      "eval_R1": 0.9931,
      "eval_R2": 0.9864,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6995,
      "eval_loss": 0.01300707645714283,
      "eval_runtime": 1450.9666,
      "eval_samples_per_second": 16.145,
      "eval_steps_per_second": 8.073,
      "step": 28400
    },
    {
      "epoch": 0.25,
      "grad_norm": 0.138671875,
      "learning_rate": 4.8880459406775195e-05,
      "loss": 0.0064,
      "step": 28500
    },
    {
      "epoch": 0.25,
      "eval_Final acc": 0.9784000683001793,
      "eval_R1": 0.9932,
      "eval_R2": 0.9864,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6961,
      "eval_loss": 0.012779178097844124,
      "eval_runtime": 1464.5898,
      "eval_samples_per_second": 15.995,
      "eval_steps_per_second": 7.997,
      "step": 28500
    },
    {
      "epoch": 0.25,
      "eval_Final acc": 0.979083070093059,
      "eval_R1": 0.9932,
      "eval_R2": 0.9865,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6965,
      "eval_loss": 0.01298822183161974,
      "eval_runtime": 1450.7516,
      "eval_samples_per_second": 16.147,
      "eval_steps_per_second": 8.074,
      "step": 28600
    },
    {
      "epoch": 0.25,
      "eval_Final acc": 0.9778878169555195,
      "eval_R1": 0.9932,
      "eval_R2": 0.9865,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6986,
      "eval_loss": 0.01306883804500103,
      "eval_runtime": 1461.7899,
      "eval_samples_per_second": 16.026,
      "eval_steps_per_second": 8.013,
      "step": 28700
    },
    {
      "epoch": 0.25,
      "eval_Final acc": 0.979083070093059,
      "eval_R1": 0.9931,
      "eval_R2": 0.9864,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6998,
      "eval_loss": 0.012865162454545498,
      "eval_runtime": 1463.6383,
      "eval_samples_per_second": 16.005,
      "eval_steps_per_second": 8.003,
      "step": 28800
    },
    {
      "epoch": 0.25,
      "eval_Final acc": 0.979253820541279,
      "eval_R1": 0.9932,
      "eval_R2": 0.9865,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.7023,
      "eval_loss": 0.012838166207075119,
      "eval_runtime": 1464.7455,
      "eval_samples_per_second": 15.993,
      "eval_steps_per_second": 7.997,
      "step": 28900
    },
    {
      "epoch": 0.25,
      "grad_norm": 0.2001953125,
      "learning_rate": 4.885850763043745e-05,
      "loss": 0.0063,
      "step": 29000
    },
    {
      "epoch": 0.25,
      "eval_Final acc": 0.9783573806881243,
      "eval_R1": 0.9933,
      "eval_R2": 0.9867,
      "eval_RL": 0.9931,
      "eval_RLsum": 0.9931,
      "eval_acc": 0.7012,
      "eval_loss": 0.012496843002736568,
      "eval_runtime": 1462.9544,
      "eval_samples_per_second": 16.013,
      "eval_steps_per_second": 8.006,
      "step": 29000
    },
    {
      "epoch": 0.25,
      "eval_Final acc": 0.9774609408349697,
      "eval_R1": 0.9932,
      "eval_R2": 0.9864,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6969,
      "eval_loss": 0.012720301747322083,
      "eval_runtime": 1463.6145,
      "eval_samples_per_second": 16.006,
      "eval_steps_per_second": 8.003,
      "step": 29100
    },
    {
      "epoch": 0.26,
      "eval_Final acc": 0.9784427559122343,
      "eval_R1": 0.9933,
      "eval_R2": 0.9866,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6981,
      "eval_loss": 0.012680739164352417,
      "eval_runtime": 1449.976,
      "eval_samples_per_second": 16.156,
      "eval_steps_per_second": 8.078,
      "step": 29200
    },
    {
      "epoch": 0.26,
      "eval_Final acc": 0.978912319644839,
      "eval_R1": 0.9933,
      "eval_R2": 0.9866,
      "eval_RL": 0.9931,
      "eval_RLsum": 0.9931,
      "eval_acc": 0.7006,
      "eval_loss": 0.01270331908017397,
      "eval_runtime": 1464.1018,
      "eval_samples_per_second": 16.0,
      "eval_steps_per_second": 8.0,
      "step": 29300
    },
    {
      "epoch": 0.26,
      "eval_Final acc": 0.9787842568086741,
      "eval_R1": 0.9931,
      "eval_R2": 0.9863,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9928,
      "eval_acc": 0.6978,
      "eval_loss": 0.012856976129114628,
      "eval_runtime": 1463.3902,
      "eval_samples_per_second": 16.008,
      "eval_steps_per_second": 8.004,
      "step": 29400
    },
    {
      "epoch": 0.26,
      "grad_norm": 0.353515625,
      "learning_rate": 4.883655585409971e-05,
      "loss": 0.0057,
      "step": 29500
    },
    {
      "epoch": 0.26,
      "eval_Final acc": 0.9802783232305985,
      "eval_R1": 0.993,
      "eval_R2": 0.9863,
      "eval_RL": 0.9928,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6993,
      "eval_loss": 0.012988944537937641,
      "eval_runtime": 1450.9054,
      "eval_samples_per_second": 16.146,
      "eval_steps_per_second": 8.073,
      "step": 29500
    },
    {
      "epoch": 0.26,
      "eval_Final acc": 0.9792111329292239,
      "eval_R1": 0.9931,
      "eval_R2": 0.9864,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.7013,
      "eval_loss": 0.012645288370549679,
      "eval_runtime": 1450.3308,
      "eval_samples_per_second": 16.152,
      "eval_steps_per_second": 8.076,
      "step": 29600
    },
    {
      "epoch": 0.26,
      "eval_Final acc": 0.9778878169555195,
      "eval_R1": 0.9932,
      "eval_R2": 0.9865,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6983,
      "eval_loss": 0.012678520753979683,
      "eval_runtime": 1463.3974,
      "eval_samples_per_second": 16.008,
      "eval_steps_per_second": 8.004,
      "step": 29700
    },
    {
      "epoch": 0.26,
      "eval_Final acc": 0.9792965081533339,
      "eval_R1": 0.9932,
      "eval_R2": 0.9865,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6974,
      "eval_loss": 0.012662285938858986,
      "eval_runtime": 1463.9507,
      "eval_samples_per_second": 16.002,
      "eval_steps_per_second": 8.001,
      "step": 29800
    },
    {
      "epoch": 0.26,
      "eval_Final acc": 0.9781866302399044,
      "eval_R1": 0.9933,
      "eval_R2": 0.9867,
      "eval_RL": 0.9931,
      "eval_RLsum": 0.9931,
      "eval_acc": 0.7025,
      "eval_loss": 0.012851212173700333,
      "eval_runtime": 1451.8736,
      "eval_samples_per_second": 16.135,
      "eval_steps_per_second": 8.068,
      "step": 29900
    },
    {
      "epoch": 0.26,
      "grad_norm": 0.287109375,
      "learning_rate": 4.881460407776198e-05,
      "loss": 0.0061,
      "step": 30000
    },
    {
      "epoch": 0.26,
      "eval_Final acc": 0.979168445317169,
      "eval_R1": 0.9933,
      "eval_R2": 0.9867,
      "eval_RL": 0.9931,
      "eval_RLsum": 0.9931,
      "eval_acc": 0.7001,
      "eval_loss": 0.012811372056603432,
      "eval_runtime": 1454.3297,
      "eval_samples_per_second": 16.108,
      "eval_steps_per_second": 8.054,
      "step": 30000
    },
    {
      "epoch": 0.26,
      "eval_Final acc": 0.9784000683001793,
      "eval_R1": 0.9932,
      "eval_R2": 0.9865,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6973,
      "eval_loss": 0.013009893707931042,
      "eval_runtime": 1462.9574,
      "eval_samples_per_second": 16.013,
      "eval_steps_per_second": 8.006,
      "step": 30100
    },
    {
      "epoch": 0.26,
      "eval_Final acc": 0.9794245709894989,
      "eval_R1": 0.9933,
      "eval_R2": 0.9867,
      "eval_RL": 0.993,
      "eval_RLsum": 0.9931,
      "eval_acc": 0.6999,
      "eval_loss": 0.013131324201822281,
      "eval_runtime": 1463.0724,
      "eval_samples_per_second": 16.012,
      "eval_steps_per_second": 8.006,
      "step": 30200
    },
    {
      "epoch": 0.27,
      "eval_Final acc": 0.9785281311363442,
      "eval_R1": 0.9933,
      "eval_R2": 0.9865,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6985,
      "eval_loss": 0.01299890037626028,
      "eval_runtime": 1451.5078,
      "eval_samples_per_second": 16.139,
      "eval_steps_per_second": 8.07,
      "step": 30300
    },
    {
      "epoch": 0.27,
      "eval_Final acc": 0.9811320754716981,
      "eval_R1": 0.9933,
      "eval_R2": 0.9867,
      "eval_RL": 0.9931,
      "eval_RLsum": 0.9931,
      "eval_acc": 0.6985,
      "eval_loss": 0.013059684075415134,
      "eval_runtime": 1451.6354,
      "eval_samples_per_second": 16.138,
      "eval_steps_per_second": 8.069,
      "step": 30400
    },
    {
      "epoch": 0.27,
      "grad_norm": 0.1787109375,
      "learning_rate": 4.8792652301424237e-05,
      "loss": 0.0058,
      "step": 30500
    },
    {
      "epoch": 0.27,
      "eval_Final acc": 0.9817723896525229,
      "eval_R1": 0.9933,
      "eval_R2": 0.9868,
      "eval_RL": 0.9932,
      "eval_RLsum": 0.9932,
      "eval_acc": 0.7025,
      "eval_loss": 0.013005061075091362,
      "eval_runtime": 1452.6582,
      "eval_samples_per_second": 16.126,
      "eval_steps_per_second": 8.063,
      "step": 30500
    },
    {
      "epoch": 0.27,
      "eval_Final acc": 0.9793818833774439,
      "eval_R1": 0.9933,
      "eval_R2": 0.9866,
      "eval_RL": 0.9931,
      "eval_RLsum": 0.9931,
      "eval_acc": 0.7,
      "eval_loss": 0.013058282434940338,
      "eval_runtime": 1466.8192,
      "eval_samples_per_second": 15.971,
      "eval_steps_per_second": 7.985,
      "step": 30600
    },
    {
      "epoch": 0.27,
      "eval_Final acc": 0.9805771365149833,
      "eval_R1": 0.9931,
      "eval_R2": 0.9863,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6975,
      "eval_loss": 0.01274099387228489,
      "eval_runtime": 1462.5234,
      "eval_samples_per_second": 16.018,
      "eval_steps_per_second": 8.009,
      "step": 30700
    },
    {
      "epoch": 0.27,
      "eval_Final acc": 0.983821395031162,
      "eval_R1": 0.9931,
      "eval_R2": 0.9864,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.6988,
      "eval_loss": 0.012721989303827286,
      "eval_runtime": 1466.802,
      "eval_samples_per_second": 15.971,
      "eval_steps_per_second": 7.985,
      "step": 30800
    },
    {
      "epoch": 0.27,
      "eval_Final acc": 0.9820285153248527,
      "eval_R1": 0.9931,
      "eval_R2": 0.9864,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6978,
      "eval_loss": 0.012523257173597813,
      "eval_runtime": 1450.487,
      "eval_samples_per_second": 16.15,
      "eval_steps_per_second": 8.075,
      "step": 30900
    },
    {
      "epoch": 0.27,
      "grad_norm": 0.474609375,
      "learning_rate": 4.8770700525086495e-05,
      "loss": 0.0062,
      "step": 31000
    },
    {
      "epoch": 0.27,
      "eval_Final acc": 0.9830530180141723,
      "eval_R1": 0.9933,
      "eval_R2": 0.9867,
      "eval_RL": 0.9931,
      "eval_RLsum": 0.9931,
      "eval_acc": 0.7006,
      "eval_loss": 0.012635809369385242,
      "eval_runtime": 1461.9373,
      "eval_samples_per_second": 16.024,
      "eval_steps_per_second": 8.012,
      "step": 31000
    },
    {
      "epoch": 0.27,
      "eval_Final acc": 0.9840775207034919,
      "eval_R1": 0.9932,
      "eval_R2": 0.9865,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.6982,
      "eval_loss": 0.012647534720599651,
      "eval_runtime": 1453.2223,
      "eval_samples_per_second": 16.12,
      "eval_steps_per_second": 8.06,
      "step": 31100
    },
    {
      "epoch": 0.27,
      "eval_Final acc": 0.9824980790574576,
      "eval_R1": 0.9932,
      "eval_R2": 0.9866,
      "eval_RL": 0.9931,
      "eval_RLsum": 0.9931,
      "eval_acc": 0.7008,
      "eval_loss": 0.012633519247174263,
      "eval_runtime": 1469.198,
      "eval_samples_per_second": 15.945,
      "eval_steps_per_second": 7.972,
      "step": 31200
    },
    {
      "epoch": 0.27,
      "eval_Final acc": 0.9849739605566464,
      "eval_R1": 0.9932,
      "eval_R2": 0.9867,
      "eval_RL": 0.9931,
      "eval_RLsum": 0.9931,
      "eval_acc": 0.702,
      "eval_loss": 0.012578523717820644,
      "eval_runtime": 1462.7782,
      "eval_samples_per_second": 16.015,
      "eval_steps_per_second": 8.007,
      "step": 31300
    },
    {
      "epoch": 0.27,
      "eval_Final acc": 0.9835225817467771,
      "eval_R1": 0.9934,
      "eval_R2": 0.9869,
      "eval_RL": 0.9932,
      "eval_RLsum": 0.9933,
      "eval_acc": 0.7041,
      "eval_loss": 0.012371215969324112,
      "eval_runtime": 1469.2878,
      "eval_samples_per_second": 15.944,
      "eval_steps_per_second": 7.972,
      "step": 31400
    },
    {
      "epoch": 0.28,
      "grad_norm": 0.287109375,
      "learning_rate": 4.8748748748748754e-05,
      "loss": 0.0059,
      "step": 31500
    },
    {
      "epoch": 0.28,
      "eval_Final acc": 0.9830957056262273,
      "eval_R1": 0.9934,
      "eval_R2": 0.9867,
      "eval_RL": 0.9931,
      "eval_RLsum": 0.9932,
      "eval_acc": 0.699,
      "eval_loss": 0.01243219617754221,
      "eval_runtime": 1463.0593,
      "eval_samples_per_second": 16.012,
      "eval_steps_per_second": 8.006,
      "step": 31500
    },
    {
      "epoch": 0.28,
      "eval_Final acc": 0.9827968923418424,
      "eval_R1": 0.9933,
      "eval_R2": 0.9867,
      "eval_RL": 0.9931,
      "eval_RLsum": 0.9931,
      "eval_acc": 0.7031,
      "eval_loss": 0.012519530951976776,
      "eval_runtime": 1462.521,
      "eval_samples_per_second": 16.018,
      "eval_steps_per_second": 8.009,
      "step": 31600
    },
    {
      "epoch": 0.28,
      "eval_Final acc": 0.9831383932382822,
      "eval_R1": 0.9932,
      "eval_R2": 0.9864,
      "eval_RL": 0.9929,
      "eval_RLsum": 0.9929,
      "eval_acc": 0.7005,
      "eval_loss": 0.012659478932619095,
      "eval_runtime": 1460.6217,
      "eval_samples_per_second": 16.038,
      "eval_steps_per_second": 8.019,
      "step": 31700
    },
    {
      "epoch": 0.28,
      "eval_Final acc": 0.9834798941347221,
      "eval_R1": 0.9932,
      "eval_R2": 0.9865,
      "eval_RL": 0.993,
      "eval_RLsum": 0.993,
      "eval_acc": 0.7009,
      "eval_loss": 0.012534371577203274,
      "eval_runtime": 1461.887,
      "eval_samples_per_second": 16.024,
      "eval_steps_per_second": 8.012,
      "step": 31800
    },
    {
      "epoch": 0.28,
      "eval_Final acc": 0.983650644582942,
      "eval_R1": 0.9932,
      "eval_R2": 0.9866,
      "eval_RL": 0.9931,
      "eval_RLsum": 0.9931,
      "eval_acc": 0.6986,
      "eval_loss": 0.012843219563364983,
      "eval_runtime": 1463.7915,
      "eval_samples_per_second": 16.004,
      "eval_steps_per_second": 8.002,
      "step": 31900
    },
    {
      "epoch": 0.28,
      "grad_norm": 0.361328125,
      "learning_rate": 4.872679697241101e-05,
      "loss": 0.0058,
      "step": 32000
    },
    {
      "epoch": 0.28,
      "eval_Final acc": 0.9846324596602066,
      "eval_R1": 0.9934,
      "eval_R2": 0.9867,
      "eval_RL": 0.9932,
      "eval_RLsum": 0.9932,
      "eval_acc": 0.6988,
      "eval_loss": 0.0125523516908288,
      "eval_runtime": 1461.815,
      "eval_samples_per_second": 16.025,
      "eval_steps_per_second": 8.013,
      "step": 32000
    },
    {
      "epoch": 0.28,
      "eval_Final acc": 0.9843336463758218,
      "eval_R1": 0.9934,
      "eval_R2": 0.9868,
      "eval_RL": 0.9932,
      "eval_RLsum": 0.9932,
      "eval_acc": 0.7014,
      "eval_loss": 0.012504671700298786,
      "eval_runtime": 1467.3939,
      "eval_samples_per_second": 15.964,
      "eval_steps_per_second": 7.982,
      "step": 32100
    }
  ],
  "logging_steps": 500,
  "max_steps": 1141860,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 300,
  "total_flos": 2.663546147578921e+18,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}