{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "eval_steps": 500,
  "global_step": 6332,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0015792798483891346,
      "grad_norm": 8.3125,
      "learning_rate": 2.998815540113708e-05,
      "loss": 0.597,
      "mean_token_accuracy": 0.8960636019706726,
      "step": 5,
      "train/kl_loss_qwen": 0.06855956092476845,
      "train/kl_loss_r1": 0.11469546605767392,
      "train/total_kl": 0.18325502723455428
    },
    {
      "epoch": 0.003158559696778269,
      "grad_norm": 20.125,
      "learning_rate": 2.9976310802274165e-05,
      "loss": 0.7472,
      "mean_token_accuracy": 0.8972001910209656,
      "step": 10,
      "train/kl_loss_qwen": 0.06861092746257783,
      "train/kl_loss_r1": 0.2576677232980728,
      "train/total_kl": 0.32627865076065066
    },
    {
      "epoch": 0.004737839545167404,
      "grad_norm": 9.625,
      "learning_rate": 2.9964466203411243e-05,
      "loss": 1.2359,
      "mean_token_accuracy": 0.8659369230270386,
      "step": 15,
      "train/kl_loss_qwen": 0.19941518902778627,
      "train/kl_loss_r1": 0.5403151184320449,
      "train/total_kl": 0.7397302955389022
    },
    {
      "epoch": 0.006317119393556538,
      "grad_norm": 5.53125,
      "learning_rate": 2.995262160454833e-05,
      "loss": 0.5792,
      "mean_token_accuracy": 0.8915262579917907,
      "step": 20,
      "train/kl_loss_qwen": 0.04952048547565937,
      "train/kl_loss_r1": 0.11130407452583313,
      "train/total_kl": 0.16082455813884736
    },
    {
      "epoch": 0.007896399241945672,
      "grad_norm": 7.125,
      "learning_rate": 2.9940777005685407e-05,
      "loss": 0.5741,
      "mean_token_accuracy": 0.8891668438911438,
      "step": 25,
      "train/kl_loss_qwen": 0.05235663205385208,
      "train/kl_loss_r1": 0.1069216139614582,
      "train/total_kl": 0.1592782460153103
    },
    {
      "epoch": 0.009475679090334808,
      "grad_norm": 11.8125,
      "learning_rate": 2.992893240682249e-05,
      "loss": 0.509,
      "mean_token_accuracy": 0.912922716140747,
      "step": 30,
      "train/kl_loss_qwen": 0.05729825086891651,
      "train/kl_loss_r1": 0.12653971910476686,
      "train/total_kl": 0.18383796960115434
    },
    {
      "epoch": 0.011054958938723942,
      "grad_norm": 5.0,
      "learning_rate": 2.991708780795957e-05,
      "loss": 0.4608,
      "mean_token_accuracy": 0.9034641861915589,
      "step": 35,
      "train/kl_loss_qwen": 0.040385838225483896,
      "train/kl_loss_r1": 0.09278027042746544,
      "train/total_kl": 0.1331661120057106
    },
    {
      "epoch": 0.012634238787113077,
      "grad_norm": 5.6875,
      "learning_rate": 2.9905243209096653e-05,
      "loss": 0.5804,
      "mean_token_accuracy": 0.887318778038025,
      "step": 40,
      "train/kl_loss_qwen": 0.04102672897279262,
      "train/kl_loss_r1": 0.12314091846346856,
      "train/total_kl": 0.16416764855384827
    },
    {
      "epoch": 0.014213518635502211,
      "grad_norm": 4.53125,
      "learning_rate": 2.9893398610233732e-05,
      "loss": 0.4971,
      "mean_token_accuracy": 0.8992031812667847,
      "step": 45,
      "train/kl_loss_qwen": 0.04518002718687057,
      "train/kl_loss_r1": 0.11724689826369286,
      "train/total_kl": 0.1624269276857376
    },
    {
      "epoch": 0.015792798483891344,
      "grad_norm": 6.09375,
      "learning_rate": 2.9881554011370817e-05,
      "loss": 0.5238,
      "mean_token_accuracy": 0.8840829849243164,
      "step": 50,
      "train/kl_loss_qwen": 0.03807361088693142,
      "train/kl_loss_r1": 0.09139700084924698,
      "train/total_kl": 0.12947061359882356
    },
    {
      "epoch": 0.01737207833228048,
      "grad_norm": 6.25,
      "learning_rate": 2.9869709412507896e-05,
      "loss": 0.4987,
      "mean_token_accuracy": 0.9054428577423096,
      "step": 55,
      "train/kl_loss_qwen": 0.053597450256347656,
      "train/kl_loss_r1": 0.09445101767778397,
      "train/total_kl": 0.14804846942424774
    },
    {
      "epoch": 0.018951358180669616,
      "grad_norm": 6.1875,
      "learning_rate": 2.985786481364498e-05,
      "loss": 0.5019,
      "mean_token_accuracy": 0.9012048602104187,
      "step": 60,
      "train/kl_loss_qwen": 0.04631386026740074,
      "train/kl_loss_r1": 0.10713352411985397,
      "train/total_kl": 0.15344738587737083
    },
    {
      "epoch": 0.02053063802905875,
      "grad_norm": 11.3125,
      "learning_rate": 2.984602021478206e-05,
      "loss": 0.4744,
      "mean_token_accuracy": 0.92316153049469,
      "step": 65,
      "train/kl_loss_qwen": 0.0651139885187149,
      "train/kl_loss_r1": 0.1392371878027916,
      "train/total_kl": 0.20435118079185485
    },
    {
      "epoch": 0.022109917877447885,
      "grad_norm": 5.96875,
      "learning_rate": 2.983417561591914e-05,
      "loss": 0.4672,
      "mean_token_accuracy": 0.8972129464149475,
      "step": 70,
      "train/kl_loss_qwen": 0.03818983249366283,
      "train/kl_loss_r1": 0.08857940062880516,
      "train/total_kl": 0.12676922976970673
    },
    {
      "epoch": 0.023689197725837018,
      "grad_norm": 12.9375,
      "learning_rate": 2.9822331017056224e-05,
      "loss": 0.5646,
      "mean_token_accuracy": 0.8939262151718139,
      "step": 75,
      "train/kl_loss_qwen": 0.04693145453929901,
      "train/kl_loss_r1": 0.11174595803022384,
      "train/total_kl": 0.1586774155497551
    },
    {
      "epoch": 0.025268477574226154,
      "grad_norm": 7.25,
      "learning_rate": 2.9810486418193306e-05,
      "loss": 0.4816,
      "mean_token_accuracy": 0.9227402448654175,
      "step": 80,
      "train/kl_loss_qwen": 0.06289013884961606,
      "train/kl_loss_r1": 0.16410403251647948,
      "train/total_kl": 0.22699416875839235
    },
    {
      "epoch": 0.026847757422615286,
      "grad_norm": 5.03125,
      "learning_rate": 2.9798641819330384e-05,
      "loss": 0.4457,
      "mean_token_accuracy": 0.9102887392044068,
      "step": 85,
      "train/kl_loss_qwen": 0.049068006500601766,
      "train/kl_loss_r1": 0.09716275930404664,
      "train/total_kl": 0.1462307646870613
    },
    {
      "epoch": 0.028427037271004423,
      "grad_norm": 5.375,
      "learning_rate": 2.978679722046747e-05,
      "loss": 0.485,
      "mean_token_accuracy": 0.9039726495742798,
      "step": 90,
      "train/kl_loss_qwen": 0.04189151488244534,
      "train/kl_loss_r1": 0.11585308015346527,
      "train/total_kl": 0.15774459838867189
    },
    {
      "epoch": 0.030006317119393555,
      "grad_norm": 4.34375,
      "learning_rate": 2.9774952621604548e-05,
      "loss": 0.4823,
      "mean_token_accuracy": 0.9130605220794678,
      "step": 95,
      "train/kl_loss_qwen": 0.06395440921187401,
      "train/kl_loss_r1": 0.11665031760931015,
      "train/total_kl": 0.1806047275662422
    },
    {
      "epoch": 0.03158559696778269,
      "grad_norm": 4.59375,
      "learning_rate": 2.9763108022741633e-05,
      "loss": 0.43,
      "mean_token_accuracy": 0.9186366915702819,
      "step": 100,
      "train/kl_loss_qwen": 0.05134199559688568,
      "train/kl_loss_r1": 0.10748056173324586,
      "train/total_kl": 0.15882255434989928
    },
    {
      "epoch": 0.03316487681617183,
      "grad_norm": 5.15625,
      "learning_rate": 2.9751263423878712e-05,
      "loss": 0.4282,
      "mean_token_accuracy": 0.9085503935813903,
      "step": 105,
      "train/kl_loss_qwen": 0.0480798214673996,
      "train/kl_loss_r1": 0.09819269627332687,
      "train/total_kl": 0.14627251774072647
    },
    {
      "epoch": 0.03474415666456096,
      "grad_norm": 5.9375,
      "learning_rate": 2.9739418825015794e-05,
      "loss": 0.4829,
      "mean_token_accuracy": 0.8989167809486389,
      "step": 110,
      "train/kl_loss_qwen": 0.050010932609438896,
      "train/kl_loss_r1": 0.11596636474132538,
      "train/total_kl": 0.16597729921340942
    },
    {
      "epoch": 0.03632343651295009,
      "grad_norm": 9.4375,
      "learning_rate": 2.9727574226152876e-05,
      "loss": 0.4531,
      "mean_token_accuracy": 0.9218316435813904,
      "step": 115,
      "train/kl_loss_qwen": 0.05941658318042755,
      "train/kl_loss_r1": 0.13600580245256425,
      "train/total_kl": 0.19542238712310792
    },
    {
      "epoch": 0.03790271636133923,
      "grad_norm": 5.875,
      "learning_rate": 2.9715729627289958e-05,
      "loss": 0.4459,
      "mean_token_accuracy": 0.9076899290084839,
      "step": 120,
      "train/kl_loss_qwen": 0.04677486214786768,
      "train/kl_loss_r1": 0.09317785315215588,
      "train/total_kl": 0.13995271548628807
    },
    {
      "epoch": 0.039481996209728365,
      "grad_norm": 4.65625,
      "learning_rate": 2.9703885028427037e-05,
      "loss": 0.4398,
      "mean_token_accuracy": 0.9047970175743103,
      "step": 125,
      "train/kl_loss_qwen": 0.03303897306323052,
      "train/kl_loss_r1": 0.09023606330156327,
      "train/total_kl": 0.12327503561973571
    },
    {
      "epoch": 0.0410612760581175,
      "grad_norm": 4.78125,
      "learning_rate": 2.969204042956412e-05,
      "loss": 0.4287,
      "mean_token_accuracy": 0.9076829433441163,
      "step": 130,
      "train/kl_loss_qwen": 0.03501458279788494,
      "train/kl_loss_r1": 0.08678344935178757,
      "train/total_kl": 0.12179803252220153
    },
    {
      "epoch": 0.04264055590650663,
      "grad_norm": 4.4375,
      "learning_rate": 2.96801958307012e-05,
      "loss": 0.4449,
      "mean_token_accuracy": 0.9096317172050477,
      "step": 135,
      "train/kl_loss_qwen": 0.05215136893093586,
      "train/kl_loss_r1": 0.10034886747598648,
      "train/total_kl": 0.1525002360343933
    },
    {
      "epoch": 0.04421983575489577,
      "grad_norm": 8.875,
      "learning_rate": 2.9668351231838282e-05,
      "loss": 0.4651,
      "mean_token_accuracy": 0.9113608121871948,
      "step": 140,
      "train/kl_loss_qwen": 0.04918167479336262,
      "train/kl_loss_r1": 0.1308824487030506,
      "train/total_kl": 0.18006412386894227
    },
    {
      "epoch": 0.0457991156032849,
      "grad_norm": 7.34375,
      "learning_rate": 2.9656506632975364e-05,
      "loss": 0.4539,
      "mean_token_accuracy": 0.914273452758789,
      "step": 145,
      "train/kl_loss_qwen": 0.037452077865600585,
      "train/kl_loss_r1": 0.10261543095111847,
      "train/total_kl": 0.1400675117969513
    },
    {
      "epoch": 0.047378395451674035,
      "grad_norm": 5.9375,
      "learning_rate": 2.9644662034112443e-05,
      "loss": 0.4864,
      "mean_token_accuracy": 0.9147412776947021,
      "step": 150,
      "train/kl_loss_qwen": 0.05881679952144623,
      "train/kl_loss_r1": 0.14455455988645555,
      "train/total_kl": 0.20337136089801788
    },
    {
      "epoch": 0.04895767530006317,
      "grad_norm": 4.3125,
      "learning_rate": 2.963281743524953e-05,
      "loss": 0.4574,
      "mean_token_accuracy": 0.9119827151298523,
      "step": 155,
      "train/kl_loss_qwen": 0.052772169932723045,
      "train/kl_loss_r1": 0.11696249693632126,
      "train/total_kl": 0.1697346657514572
    },
    {
      "epoch": 0.05053695514845231,
      "grad_norm": 4.46875,
      "learning_rate": 2.9620972836386607e-05,
      "loss": 0.4369,
      "mean_token_accuracy": 0.9130014657974244,
      "step": 160,
      "train/kl_loss_qwen": 0.04390869811177254,
      "train/kl_loss_r1": 0.10849741548299789,
      "train/total_kl": 0.15240611135959625
    },
    {
      "epoch": 0.05211623499684144,
      "grad_norm": 3.953125,
      "learning_rate": 2.960912823752369e-05,
      "loss": 0.4349,
      "mean_token_accuracy": 0.919090187549591,
      "step": 165,
      "train/kl_loss_qwen": 0.05664343386888504,
      "train/kl_loss_r1": 0.11979135572910309,
      "train/total_kl": 0.17643478661775588
    },
    {
      "epoch": 0.05369551484523057,
      "grad_norm": 4.71875,
      "learning_rate": 2.959728363866077e-05,
      "loss": 0.4566,
      "mean_token_accuracy": 0.9074759125709534,
      "step": 170,
      "train/kl_loss_qwen": 0.03763047084212303,
      "train/kl_loss_r1": 0.11284382492303849,
      "train/total_kl": 0.15047429502010345
    },
    {
      "epoch": 0.05527479469361971,
      "grad_norm": 3.6875,
      "learning_rate": 2.9585439039797853e-05,
      "loss": 0.4691,
      "mean_token_accuracy": 0.9052929759025574,
      "step": 175,
      "train/kl_loss_qwen": 0.04656125195324421,
      "train/kl_loss_r1": 0.10648671239614486,
      "train/total_kl": 0.15304796546697616
    },
    {
      "epoch": 0.056854074542008845,
      "grad_norm": 4.5625,
      "learning_rate": 2.9573594440934935e-05,
      "loss": 0.4519,
      "mean_token_accuracy": 0.9051965713500977,
      "step": 180,
      "train/kl_loss_qwen": 0.04833948947489262,
      "train/kl_loss_r1": 0.10998060405254365,
      "train/total_kl": 0.15832009613513948
    },
    {
      "epoch": 0.05843335439039798,
      "grad_norm": 4.5625,
      "learning_rate": 2.9561749842072017e-05,
      "loss": 0.3868,
      "mean_token_accuracy": 0.9177706241607666,
      "step": 185,
      "train/kl_loss_qwen": 0.031213388219475745,
      "train/kl_loss_r1": 0.06958115547895431,
      "train/total_kl": 0.10079454332590103
    },
    {
      "epoch": 0.06001263423878711,
      "grad_norm": 5.15625,
      "learning_rate": 2.9549905243209095e-05,
      "loss": 0.3506,
      "mean_token_accuracy": 0.9317142486572265,
      "step": 190,
      "train/kl_loss_qwen": 0.03400315977632999,
      "train/kl_loss_r1": 0.08720123767852783,
      "train/total_kl": 0.1212043970823288
    },
    {
      "epoch": 0.06159191408717625,
      "grad_norm": 6.09375,
      "learning_rate": 2.953806064434618e-05,
      "loss": 0.5022,
      "mean_token_accuracy": 0.9038637638092041,
      "step": 195,
      "train/kl_loss_qwen": 0.05283312201499939,
      "train/kl_loss_r1": 0.11915318369865417,
      "train/total_kl": 0.17198630422353745
    },
    {
      "epoch": 0.06317119393556538,
      "grad_norm": 21.625,
      "learning_rate": 2.952621604548326e-05,
      "loss": 0.5946,
      "mean_token_accuracy": 0.9095802426338195,
      "step": 200,
      "train/kl_loss_qwen": 0.10119329243898392,
      "train/kl_loss_r1": 0.19896390438079833,
      "train/total_kl": 0.30015720427036285
    },
    {
      "epoch": 0.06475047378395452,
      "grad_norm": 4.15625,
      "learning_rate": 2.951437144662034e-05,
      "loss": 0.4827,
      "mean_token_accuracy": 0.8999685764312744,
      "step": 205,
      "train/kl_loss_qwen": 0.056330349668860434,
      "train/kl_loss_r1": 0.10734975934028626,
      "train/total_kl": 0.16368010640144348
    },
    {
      "epoch": 0.06632975363234365,
      "grad_norm": 5.15625,
      "learning_rate": 2.9502526847757423e-05,
      "loss": 0.479,
      "mean_token_accuracy": 0.903225576877594,
      "step": 210,
      "train/kl_loss_qwen": 0.036679952219128606,
      "train/kl_loss_r1": 0.100700543820858,
      "train/total_kl": 0.1373804971575737
    },
    {
      "epoch": 0.06790903348073278,
      "grad_norm": 3.25,
      "learning_rate": 2.9490682248894505e-05,
      "loss": 0.3973,
      "mean_token_accuracy": 0.922663950920105,
      "step": 215,
      "train/kl_loss_qwen": 0.0410617146641016,
      "train/kl_loss_r1": 0.09844419062137603,
      "train/total_kl": 0.13950590491294862
    },
    {
      "epoch": 0.06948831332912192,
      "grad_norm": 6.125,
      "learning_rate": 2.9478837650031587e-05,
      "loss": 0.5403,
      "mean_token_accuracy": 0.891838264465332,
      "step": 220,
      "train/kl_loss_qwen": 0.0530395407229662,
      "train/kl_loss_r1": 0.13303317576646806,
      "train/total_kl": 0.18607271909713746
    },
    {
      "epoch": 0.07106759317751106,
      "grad_norm": 5.09375,
      "learning_rate": 2.946699305116867e-05,
      "loss": 0.4455,
      "mean_token_accuracy": 0.9003235220909118,
      "step": 225,
      "train/kl_loss_qwen": 0.03510046619921923,
      "train/kl_loss_r1": 0.08129789903759957,
      "train/total_kl": 0.1163983628153801
    },
    {
      "epoch": 0.07264687302590019,
      "grad_norm": 4.53125,
      "learning_rate": 2.9455148452305748e-05,
      "loss": 0.4453,
      "mean_token_accuracy": 0.9028073430061341,
      "step": 230,
      "train/kl_loss_qwen": 0.031616070121526715,
      "train/kl_loss_r1": 0.08886473998427391,
      "train/total_kl": 0.12048080861568451
    },
    {
      "epoch": 0.07422615287428933,
      "grad_norm": 6.25,
      "learning_rate": 2.9443303853442833e-05,
      "loss": 0.5085,
      "mean_token_accuracy": 0.90663161277771,
      "step": 235,
      "train/kl_loss_qwen": 0.05736028254032135,
      "train/kl_loss_r1": 0.13584401607513427,
      "train/total_kl": 0.1932042956352234
    },
    {
      "epoch": 0.07580543272267846,
      "grad_norm": 4.21875,
      "learning_rate": 2.9431459254579912e-05,
      "loss": 0.487,
      "mean_token_accuracy": 0.9022761344909668,
      "step": 240,
      "train/kl_loss_qwen": 0.05404329039156437,
      "train/kl_loss_r1": 0.11493499130010605,
      "train/total_kl": 0.16897827982902527
    },
    {
      "epoch": 0.07738471257106759,
      "grad_norm": 3.21875,
      "learning_rate": 2.9419614655716994e-05,
      "loss": 0.4273,
      "mean_token_accuracy": 0.9091532230377197,
      "step": 245,
      "train/kl_loss_qwen": 0.03419891633093357,
      "train/kl_loss_r1": 0.09884620010852814,
      "train/total_kl": 0.13304511457681656
    },
    {
      "epoch": 0.07896399241945673,
      "grad_norm": 3.09375,
      "learning_rate": 2.9407770056854076e-05,
      "loss": 0.4358,
      "mean_token_accuracy": 0.9110920786857605,
      "step": 250,
      "train/kl_loss_qwen": 0.046397218108177186,
      "train/kl_loss_r1": 0.10456390678882599,
      "train/total_kl": 0.15096112191677094
    },
    {
      "epoch": 0.08054327226784586,
      "grad_norm": 5.59375,
      "learning_rate": 2.9395925457991154e-05,
      "loss": 0.4237,
      "mean_token_accuracy": 0.9208296179771424,
      "step": 255,
      "train/kl_loss_qwen": 0.04299980103969574,
      "train/kl_loss_r1": 0.12058884203433991,
      "train/total_kl": 0.16358864456415176
    },
    {
      "epoch": 0.082122552116235,
      "grad_norm": 4.03125,
      "learning_rate": 2.938408085912824e-05,
      "loss": 0.528,
      "mean_token_accuracy": 0.9103268384933472,
      "step": 260,
      "train/kl_loss_qwen": 0.07274995595216752,
      "train/kl_loss_r1": 0.15502492040395738,
      "train/total_kl": 0.2277748703956604
    },
    {
      "epoch": 0.08370183196462413,
      "grad_norm": 3.140625,
      "learning_rate": 2.9372236260265318e-05,
      "loss": 0.4201,
      "mean_token_accuracy": 0.9096172094345093,
      "step": 265,
      "train/kl_loss_qwen": 0.03658049404621124,
      "train/kl_loss_r1": 0.08660418316721916,
      "train/total_kl": 0.12318467870354652
    },
    {
      "epoch": 0.08528111181301326,
      "grad_norm": 4.65625,
      "learning_rate": 2.93603916614024e-05,
      "loss": 0.3815,
      "mean_token_accuracy": 0.9104852080345154,
      "step": 270,
      "train/kl_loss_qwen": 0.024139636382460593,
      "train/kl_loss_r1": 0.07269874587655067,
      "train/total_kl": 0.09683838188648224
    },
    {
      "epoch": 0.0868603916614024,
      "grad_norm": 3.796875,
      "learning_rate": 2.9348547062539482e-05,
      "loss": 0.4166,
      "mean_token_accuracy": 0.9154416918754578,
      "step": 275,
      "train/kl_loss_qwen": 0.04606907367706299,
      "train/kl_loss_r1": 0.10962127298116683,
      "train/total_kl": 0.15569035112857818
    },
    {
      "epoch": 0.08843967150979154,
      "grad_norm": 4.875,
      "learning_rate": 2.9336702463676564e-05,
      "loss": 0.8508,
      "mean_token_accuracy": 0.9076127290725708,
      "step": 280,
      "train/kl_loss_qwen": 0.1508233666419983,
      "train/kl_loss_r1": 0.40986614376306535,
      "train/total_kl": 0.5606895059347152
    },
    {
      "epoch": 0.09001895135818067,
      "grad_norm": 4.125,
      "learning_rate": 2.9324857864813646e-05,
      "loss": 0.4099,
      "mean_token_accuracy": 0.9162441611289978,
      "step": 285,
      "train/kl_loss_qwen": 0.04260625131428242,
      "train/kl_loss_r1": 0.09926703125238419,
      "train/total_kl": 0.1418732836842537
    },
    {
      "epoch": 0.0915982312065698,
      "grad_norm": 5.0625,
      "learning_rate": 2.9313013265950728e-05,
      "loss": 0.4375,
      "mean_token_accuracy": 0.9190158605575561,
      "step": 290,
      "train/kl_loss_qwen": 0.049399647116661075,
      "train/kl_loss_r1": 0.12138292789459229,
      "train/total_kl": 0.17078257501125335
    },
    {
      "epoch": 0.09317751105495894,
      "grad_norm": 5.125,
      "learning_rate": 2.9301168667087807e-05,
      "loss": 0.4329,
      "mean_token_accuracy": 0.9062884211540222,
      "step": 295,
      "train/kl_loss_qwen": 0.024093149788677694,
      "train/kl_loss_r1": 0.07717221677303314,
      "train/total_kl": 0.10126536786556244
    },
    {
      "epoch": 0.09475679090334807,
      "grad_norm": 3.65625,
      "learning_rate": 2.9289324068224892e-05,
      "loss": 0.4211,
      "mean_token_accuracy": 0.9014742851257325,
      "step": 300,
      "train/kl_loss_qwen": 0.04225325286388397,
      "train/kl_loss_r1": 0.08502504676580429,
      "train/total_kl": 0.12727829813957214
    },
    {
      "epoch": 0.09633607075173721,
      "grad_norm": 8.0625,
      "learning_rate": 2.927747946936197e-05,
      "loss": 0.4498,
      "mean_token_accuracy": 0.9137883186340332,
      "step": 305,
      "train/kl_loss_qwen": 0.045832043141126634,
      "train/kl_loss_r1": 0.1273800939321518,
      "train/total_kl": 0.17321213483810424
    },
    {
      "epoch": 0.09791535060012634,
      "grad_norm": 5.9375,
      "learning_rate": 2.9265634870499053e-05,
      "loss": 0.3803,
      "mean_token_accuracy": 0.9171325325965881,
      "step": 310,
      "train/kl_loss_qwen": 0.029794520139694212,
      "train/kl_loss_r1": 0.09493155777454376,
      "train/total_kl": 0.12472607791423798
    },
    {
      "epoch": 0.09949463044851548,
      "grad_norm": 4.28125,
      "learning_rate": 2.9253790271636135e-05,
      "loss": 0.3505,
      "mean_token_accuracy": 0.9299791336059571,
      "step": 315,
      "train/kl_loss_qwen": 0.02822284698486328,
      "train/kl_loss_r1": 0.08262794315814972,
      "train/total_kl": 0.110850790143013
    },
    {
      "epoch": 0.10107391029690461,
      "grad_norm": 5.46875,
      "learning_rate": 2.9241945672773217e-05,
      "loss": 0.4432,
      "mean_token_accuracy": 0.9070752620697021,
      "step": 320,
      "train/kl_loss_qwen": 0.03599231615662575,
      "train/kl_loss_r1": 0.10620761513710023,
      "train/total_kl": 0.14219993054866792
    },
    {
      "epoch": 0.10265319014529374,
      "grad_norm": 4.34375,
      "learning_rate": 2.92301010739103e-05,
      "loss": 0.4035,
      "mean_token_accuracy": 0.9100895643234252,
      "step": 325,
      "train/kl_loss_qwen": 0.03184035867452621,
      "train/kl_loss_r1": 0.08496054112911225,
      "train/total_kl": 0.11680089980363846
    },
    {
      "epoch": 0.10423246999368288,
      "grad_norm": 6.3125,
      "learning_rate": 2.921825647504738e-05,
      "loss": 0.4781,
      "mean_token_accuracy": 0.8939521312713623,
      "step": 330,
      "train/kl_loss_qwen": 0.041493885964155194,
      "train/kl_loss_r1": 0.08955786675214768,
      "train/total_kl": 0.13105175495147706
    },
    {
      "epoch": 0.10581174984207202,
      "grad_norm": 5.21875,
      "learning_rate": 2.920641187618446e-05,
      "loss": 0.4373,
      "mean_token_accuracy": 0.922801160812378,
      "step": 335,
      "train/kl_loss_qwen": 0.061493751406669614,
      "train/kl_loss_r1": 0.1285446584224701,
      "train/total_kl": 0.19003840684890747
    },
    {
      "epoch": 0.10739102969046115,
      "grad_norm": 3.9375,
      "learning_rate": 2.9194567277321544e-05,
      "loss": 0.4023,
      "mean_token_accuracy": 0.9231971383094788,
      "step": 340,
      "train/kl_loss_qwen": 0.05469193607568741,
      "train/kl_loss_r1": 0.10607736557722092,
      "train/total_kl": 0.16076930165290831
    },
    {
      "epoch": 0.10897030953885029,
      "grad_norm": 3.765625,
      "learning_rate": 2.9182722678458623e-05,
      "loss": 0.4937,
      "mean_token_accuracy": 0.8905720472335815,
      "step": 345,
      "train/kl_loss_qwen": 0.033227978833019735,
      "train/kl_loss_r1": 0.09125299751758575,
      "train/total_kl": 0.12448097988963128
    },
    {
      "epoch": 0.11054958938723942,
      "grad_norm": 4.375,
      "learning_rate": 2.9170878079595705e-05,
      "loss": 0.4768,
      "mean_token_accuracy": 0.8965606689453125,
      "step": 350,
      "train/kl_loss_qwen": 0.03288907930254936,
      "train/kl_loss_r1": 0.0899883434176445,
      "train/total_kl": 0.1228774219751358
    },
    {
      "epoch": 0.11212886923562855,
      "grad_norm": 4.65625,
      "learning_rate": 2.9159033480732787e-05,
      "loss": 0.4279,
      "mean_token_accuracy": 0.904407286643982,
      "step": 355,
      "train/kl_loss_qwen": 0.031035296991467475,
      "train/kl_loss_r1": 0.0875077024102211,
      "train/total_kl": 0.11854299902915955
    },
    {
      "epoch": 0.11370814908401769,
      "grad_norm": 3.921875,
      "learning_rate": 2.914718888186987e-05,
      "loss": 0.4076,
      "mean_token_accuracy": 0.9209004878997803,
      "step": 360,
      "train/kl_loss_qwen": 0.04115631431341171,
      "train/kl_loss_r1": 0.10898292958736419,
      "train/total_kl": 0.15013924688100816
    },
    {
      "epoch": 0.11528742893240682,
      "grad_norm": 7.625,
      "learning_rate": 2.913534428300695e-05,
      "loss": 0.4304,
      "mean_token_accuracy": 0.9129171371459961,
      "step": 365,
      "train/kl_loss_qwen": 0.04491574205458164,
      "train/kl_loss_r1": 0.10503538995981217,
      "train/total_kl": 0.14995113313198088
    },
    {
      "epoch": 0.11686670878079596,
      "grad_norm": 5.875,
      "learning_rate": 2.9123499684144033e-05,
      "loss": 0.4378,
      "mean_token_accuracy": 0.917508578300476,
      "step": 370,
      "train/kl_loss_qwen": 0.04991894215345383,
      "train/kl_loss_r1": 0.11731350570917129,
      "train/total_kl": 0.1672324448823929
    },
    {
      "epoch": 0.1184459886291851,
      "grad_norm": 3.953125,
      "learning_rate": 2.911165508528111e-05,
      "loss": 0.4209,
      "mean_token_accuracy": 0.9044533848762513,
      "step": 375,
      "train/kl_loss_qwen": 0.03777746744453907,
      "train/kl_loss_r1": 0.07677295953035354,
      "train/total_kl": 0.11455042660236359
    },
    {
      "epoch": 0.12002526847757422,
      "grad_norm": 4.25,
      "learning_rate": 2.9099810486418193e-05,
      "loss": 0.403,
      "mean_token_accuracy": 0.9301979780197144,
      "step": 380,
      "train/kl_loss_qwen": 0.04332649745047092,
      "train/kl_loss_r1": 0.11153730005025864,
      "train/total_kl": 0.15486379861831664
    },
    {
      "epoch": 0.12160454832596336,
      "grad_norm": 5.09375,
      "learning_rate": 2.9087965887555275e-05,
      "loss": 0.4301,
      "mean_token_accuracy": 0.9139994621276856,
      "step": 385,
      "train/kl_loss_qwen": 0.04693395085632801,
      "train/kl_loss_r1": 0.1062821701169014,
      "train/total_kl": 0.1532161220908165
    },
    {
      "epoch": 0.1231838281743525,
      "grad_norm": 3.453125,
      "learning_rate": 2.9076121288692354e-05,
      "loss": 0.3351,
      "mean_token_accuracy": 0.9313080430030822,
      "step": 390,
      "train/kl_loss_qwen": 0.035632068291306496,
      "train/kl_loss_r1": 0.07446864098310471,
      "train/total_kl": 0.11010070890188217
    },
    {
      "epoch": 0.12476310802274163,
      "grad_norm": 4.34375,
      "learning_rate": 2.906427668982944e-05,
      "loss": 0.455,
      "mean_token_accuracy": 0.9174779534339905,
      "step": 395,
      "train/kl_loss_qwen": 0.05457628481090069,
      "train/kl_loss_r1": 0.15018678307533265,
      "train/total_kl": 0.20476306676864625
    },
    {
      "epoch": 0.12634238787113075,
      "grad_norm": 3.875,
      "learning_rate": 2.9052432090966518e-05,
      "loss": 0.436,
      "mean_token_accuracy": 0.909913980960846,
      "step": 400,
      "train/kl_loss_qwen": 0.03823860697448254,
      "train/kl_loss_r1": 0.09224527180194855,
      "train/total_kl": 0.130483877658844
    },
    {
      "epoch": 0.1279216677195199,
      "grad_norm": 3.640625,
      "learning_rate": 2.9040587492103603e-05,
      "loss": 0.4007,
      "mean_token_accuracy": 0.9207257151603698,
      "step": 405,
      "train/kl_loss_qwen": 0.03845406137406826,
      "train/kl_loss_r1": 0.109235218167305,
      "train/total_kl": 0.14768928289413452
    },
    {
      "epoch": 0.12950094756790903,
      "grad_norm": 7.9375,
      "learning_rate": 2.9028742893240682e-05,
      "loss": 0.4428,
      "mean_token_accuracy": 0.9095982909202576,
      "step": 410,
      "train/kl_loss_qwen": 0.06052665114402771,
      "train/kl_loss_r1": 0.11685522198677063,
      "train/total_kl": 0.1773818701505661
    },
    {
      "epoch": 0.13108022741629816,
      "grad_norm": 3.125,
      "learning_rate": 2.9016898294377764e-05,
      "loss": 0.4334,
      "mean_token_accuracy": 0.911774218082428,
      "step": 415,
      "train/kl_loss_qwen": 0.0361639253795147,
      "train/kl_loss_r1": 0.08238687217235566,
      "train/total_kl": 0.11855079829692841
    },
    {
      "epoch": 0.1326595072646873,
      "grad_norm": 4.40625,
      "learning_rate": 2.9005053695514846e-05,
      "loss": 0.4347,
      "mean_token_accuracy": 0.9105106472969056,
      "step": 420,
      "train/kl_loss_qwen": 0.048740137740969657,
      "train/kl_loss_r1": 0.08928831964731217,
      "train/total_kl": 0.13802845627069474
    },
    {
      "epoch": 0.13423878711307644,
      "grad_norm": 4.84375,
      "learning_rate": 2.8993209096651928e-05,
      "loss": 0.4469,
      "mean_token_accuracy": 0.9053473591804504,
      "step": 425,
      "train/kl_loss_qwen": 0.03426402285695076,
      "train/kl_loss_r1": 0.10157537758350373,
      "train/total_kl": 0.13583939969539643
    },
    {
      "epoch": 0.13581806696146556,
      "grad_norm": 4.5,
      "learning_rate": 2.8981364497789006e-05,
      "loss": 0.5532,
      "mean_token_accuracy": 0.9127062082290649,
      "step": 430,
      "train/kl_loss_qwen": 0.07363911382853985,
      "train/kl_loss_r1": 0.20122179612517357,
      "train/total_kl": 0.2748609110713005
    },
    {
      "epoch": 0.13739734680985471,
      "grad_norm": 5.6875,
      "learning_rate": 2.8969519898926092e-05,
      "loss": 0.4176,
      "mean_token_accuracy": 0.9096400499343872,
      "step": 435,
      "train/kl_loss_qwen": 0.04292279854416847,
      "train/kl_loss_r1": 0.08915165811777115,
      "train/total_kl": 0.13207445442676544
    },
    {
      "epoch": 0.13897662665824384,
      "grad_norm": 4.59375,
      "learning_rate": 2.895767530006317e-05,
      "loss": 0.443,
      "mean_token_accuracy": 0.905884861946106,
      "step": 440,
      "train/kl_loss_qwen": 0.04267049990594387,
      "train/kl_loss_r1": 0.0992314226925373,
      "train/total_kl": 0.14190192222595216
    },
    {
      "epoch": 0.14055590650663297,
      "grad_norm": 3.390625,
      "learning_rate": 2.8945830701200256e-05,
      "loss": 0.3724,
      "mean_token_accuracy": 0.926804494857788,
      "step": 445,
      "train/kl_loss_qwen": 0.040928006544709204,
      "train/kl_loss_r1": 0.08868414610624313,
      "train/total_kl": 0.12961215376853943
    },
    {
      "epoch": 0.14213518635502212,
      "grad_norm": 4.40625,
      "learning_rate": 2.8933986102337334e-05,
      "loss": 0.4129,
      "mean_token_accuracy": 0.9207128882408142,
      "step": 450,
      "train/kl_loss_qwen": 0.041717945225536825,
      "train/kl_loss_r1": 0.09797072559595107,
      "train/total_kl": 0.1396886721253395
    },
    {
      "epoch": 0.14371446620341125,
      "grad_norm": 6.09375,
      "learning_rate": 2.8922141503474416e-05,
      "loss": 0.4452,
      "mean_token_accuracy": 0.8937878489494324,
      "step": 455,
      "train/kl_loss_qwen": 0.028465672209858895,
      "train/kl_loss_r1": 0.08130454644560814,
      "train/total_kl": 0.10977021902799607
    },
    {
      "epoch": 0.14529374605180037,
      "grad_norm": 4.9375,
      "learning_rate": 2.8910296904611498e-05,
      "loss": 0.6694,
      "mean_token_accuracy": 0.9046720504760742,
      "step": 460,
      "train/kl_loss_qwen": 0.10907440520823002,
      "train/kl_loss_r1": 0.2476026453077793,
      "train/total_kl": 0.35667705833911895
    },
    {
      "epoch": 0.14687302590018952,
      "grad_norm": 3.984375,
      "learning_rate": 2.889845230574858e-05,
      "loss": 0.3373,
      "mean_token_accuracy": 0.9334591627120972,
      "step": 465,
      "train/kl_loss_qwen": 0.028183384239673613,
      "train/kl_loss_r1": 0.09278561770915986,
      "train/total_kl": 0.12096900343894959
    },
    {
      "epoch": 0.14845230574857865,
      "grad_norm": 10.5,
      "learning_rate": 2.888660770688566e-05,
      "loss": 0.4439,
      "mean_token_accuracy": 0.9239588499069213,
      "step": 470,
      "train/kl_loss_qwen": 0.05102039538323879,
      "train/kl_loss_r1": 0.15917339473962783,
      "train/total_kl": 0.21019378900527955
    },
    {
      "epoch": 0.15003158559696778,
      "grad_norm": 2.546875,
      "learning_rate": 2.8874763108022744e-05,
      "loss": 0.4147,
      "mean_token_accuracy": 0.9094501256942749,
      "step": 475,
      "train/kl_loss_qwen": 0.02994098700582981,
      "train/kl_loss_r1": 0.08021924495697022,
      "train/total_kl": 0.11016023010015488
    },
    {
      "epoch": 0.15161086544535693,
      "grad_norm": 4.40625,
      "learning_rate": 2.8862918509159823e-05,
      "loss": 0.4155,
      "mean_token_accuracy": 0.9191853761672973,
      "step": 480,
      "train/kl_loss_qwen": 0.0378162007778883,
      "train/kl_loss_r1": 0.1032797209918499,
      "train/total_kl": 0.14109592363238335
    },
    {
      "epoch": 0.15319014529374606,
      "grad_norm": 3.78125,
      "learning_rate": 2.8851073910296908e-05,
      "loss": 0.3896,
      "mean_token_accuracy": 0.9102899312973023,
      "step": 485,
      "train/kl_loss_qwen": 0.03835316747426987,
      "train/kl_loss_r1": 0.07900936752557755,
      "train/total_kl": 0.11736253648996353
    },
    {
      "epoch": 0.15476942514213518,
      "grad_norm": 3.578125,
      "learning_rate": 2.8839229311433987e-05,
      "loss": 0.4158,
      "mean_token_accuracy": 0.9168619632720947,
      "step": 490,
      "train/kl_loss_qwen": 0.04100581556558609,
      "train/kl_loss_r1": 0.10428861528635025,
      "train/total_kl": 0.14529443085193633
    },
    {
      "epoch": 0.15634870499052433,
      "grad_norm": 3.34375,
      "learning_rate": 2.882738471257107e-05,
      "loss": 0.3887,
      "mean_token_accuracy": 0.9177898645401001,
      "step": 495,
      "train/kl_loss_qwen": 0.03854363039135933,
      "train/kl_loss_r1": 0.08364018350839615,
      "train/total_kl": 0.12218381464481354
    },
    {
      "epoch": 0.15792798483891346,
      "grad_norm": 8.0,
      "learning_rate": 2.881554011370815e-05,
      "loss": 0.4066,
      "mean_token_accuracy": 0.9216166973114014,
      "step": 500,
      "train/kl_loss_qwen": 0.04159573204815388,
      "train/kl_loss_r1": 0.11968117952346802,
      "train/total_kl": 0.16127691119909288
    },
    {
      "epoch": 0.15950726468730259,
      "grad_norm": 3.375,
      "learning_rate": 2.880369551484523e-05,
      "loss": 0.3254,
      "mean_token_accuracy": 0.9259904026985168,
      "step": 505,
      "train/kl_loss_qwen": 0.030062363110482694,
      "train/kl_loss_r1": 0.06661505103111268,
      "train/total_kl": 0.09667741358280182
    },
    {
      "epoch": 0.1610865445356917,
      "grad_norm": 4.4375,
      "learning_rate": 2.879185091598231e-05,
      "loss": 0.401,
      "mean_token_accuracy": 0.921536922454834,
      "step": 510,
      "train/kl_loss_qwen": 0.05205292329192161,
      "train/kl_loss_r1": 0.1018543653190136,
      "train/total_kl": 0.15390728861093522
    },
    {
      "epoch": 0.16266582438408086,
      "grad_norm": 3.578125,
      "learning_rate": 2.8780006317119393e-05,
      "loss": 0.4377,
      "mean_token_accuracy": 0.9153391242027282,
      "step": 515,
      "train/kl_loss_qwen": 0.060998298972845075,
      "train/kl_loss_r1": 0.1110582858324051,
      "train/total_kl": 0.17205658555030823
    },
    {
      "epoch": 0.16424510423247,
      "grad_norm": 5.3125,
      "learning_rate": 2.8768161718256475e-05,
      "loss": 0.5099,
      "mean_token_accuracy": 0.9028707623481751,
      "step": 520,
      "train/kl_loss_qwen": 0.06397873610258102,
      "train/kl_loss_r1": 0.12318305522203446,
      "train/total_kl": 0.18716179132461547
    },
    {
      "epoch": 0.16582438408085912,
      "grad_norm": 5.625,
      "learning_rate": 2.8756317119393557e-05,
      "loss": 0.4024,
      "mean_token_accuracy": 0.9106868028640747,
      "step": 525,
      "train/kl_loss_qwen": 0.03646775484085083,
      "train/kl_loss_r1": 0.08387937545776367,
      "train/total_kl": 0.12034712731838226
    },
    {
      "epoch": 0.16740366392924827,
      "grad_norm": 4.65625,
      "learning_rate": 2.874447252053064e-05,
      "loss": 0.3981,
      "mean_token_accuracy": 0.9222624659538269,
      "step": 530,
      "train/kl_loss_qwen": 0.03656016550958156,
      "train/kl_loss_r1": 0.10247425511479377,
      "train/total_kl": 0.1390344202518463
    },
    {
      "epoch": 0.1689829437776374,
      "grad_norm": 3.5625,
      "learning_rate": 2.8732627921667718e-05,
      "loss": 0.5562,
      "mean_token_accuracy": 0.9067546010017395,
      "step": 535,
      "train/kl_loss_qwen": 0.06597321853041649,
      "train/kl_loss_r1": 0.17305902093648912,
      "train/total_kl": 0.2390322357416153
    },
    {
      "epoch": 0.17056222362602652,
      "grad_norm": 3.515625,
      "learning_rate": 2.8720783322804803e-05,
      "loss": 0.3896,
      "mean_token_accuracy": 0.919617235660553,
      "step": 540,
      "train/kl_loss_qwen": 0.04390577785670757,
      "train/kl_loss_r1": 0.08733106479048729,
      "train/total_kl": 0.13123684376478195
    },
    {
      "epoch": 0.17214150347441567,
      "grad_norm": 8.125,
      "learning_rate": 2.870893872394188e-05,
      "loss": 0.4346,
      "mean_token_accuracy": 0.9107903361320495,
      "step": 545,
      "train/kl_loss_qwen": 0.034388281404972076,
      "train/kl_loss_r1": 0.10430290699005126,
      "train/total_kl": 0.138691183924675
    },
    {
      "epoch": 0.1737207833228048,
      "grad_norm": 3.25,
      "learning_rate": 2.8697094125078967e-05,
      "loss": 0.3827,
      "mean_token_accuracy": 0.9236801505088806,
      "step": 550,
      "train/kl_loss_qwen": 0.041596776060760024,
      "train/kl_loss_r1": 0.09700752794742584,
      "train/total_kl": 0.13860430419445038
    },
    {
      "epoch": 0.17530006317119393,
      "grad_norm": 4.28125,
      "learning_rate": 2.8685249526216046e-05,
      "loss": 0.3968,
      "mean_token_accuracy": 0.9150171518325806,
      "step": 555,
      "train/kl_loss_qwen": 0.03237848430871963,
      "train/kl_loss_r1": 0.09319581836462021,
      "train/total_kl": 0.12557430267333985
    },
    {
      "epoch": 0.17687934301958308,
      "grad_norm": 3.84375,
      "learning_rate": 2.8673404927353128e-05,
      "loss": 0.4101,
      "mean_token_accuracy": 0.9032036662101746,
      "step": 560,
      "train/kl_loss_qwen": 0.03721548542380333,
      "train/kl_loss_r1": 0.07992707639932632,
      "train/total_kl": 0.11714256405830384
    },
    {
      "epoch": 0.1784586228679722,
      "grad_norm": 8.25,
      "learning_rate": 2.866156032849021e-05,
      "loss": 0.4642,
      "mean_token_accuracy": 0.908664071559906,
      "step": 565,
      "train/kl_loss_qwen": 0.05835045520216227,
      "train/kl_loss_r1": 0.11581412255764008,
      "train/total_kl": 0.17416457682847977
    },
    {
      "epoch": 0.18003790271636133,
      "grad_norm": 2.59375,
      "learning_rate": 2.864971572962729e-05,
      "loss": 0.3384,
      "mean_token_accuracy": 0.9242323517799378,
      "step": 570,
      "train/kl_loss_qwen": 0.03214266635477543,
      "train/kl_loss_r1": 0.07008601352572441,
      "train/total_kl": 0.10222867727279664
    },
    {
      "epoch": 0.18161718256475048,
      "grad_norm": 5.0625,
      "learning_rate": 2.863787113076437e-05,
      "loss": 0.399,
      "mean_token_accuracy": 0.9151182413101197,
      "step": 575,
      "train/kl_loss_qwen": 0.04417599849402905,
      "train/kl_loss_r1": 0.09617904052138329,
      "train/total_kl": 0.14035504013299943
    },
    {
      "epoch": 0.1831964624131396,
      "grad_norm": 3.984375,
      "learning_rate": 2.8626026531901455e-05,
      "loss": 0.4182,
      "mean_token_accuracy": 0.9164104700088501,
      "step": 580,
      "train/kl_loss_qwen": 0.04810438416898251,
      "train/kl_loss_r1": 0.11305789649486542,
      "train/total_kl": 0.16116228252649306
    },
    {
      "epoch": 0.18477574226152874,
      "grad_norm": 14.875,
      "learning_rate": 2.8614181933038534e-05,
      "loss": 0.4284,
      "mean_token_accuracy": 0.9225537776947021,
      "step": 585,
      "train/kl_loss_qwen": 0.04667378515005112,
      "train/kl_loss_r1": 0.11427764669060707,
      "train/total_kl": 0.1609514281153679
    },
    {
      "epoch": 0.1863550221099179,
      "grad_norm": 3.0,
      "learning_rate": 2.860233733417562e-05,
      "loss": 0.4038,
      "mean_token_accuracy": 0.9220267534255981,
      "step": 590,
      "train/kl_loss_qwen": 0.04543819911777973,
      "train/kl_loss_r1": 0.1040770024061203,
      "train/total_kl": 0.14951520562171935
    },
    {
      "epoch": 0.18793430195830702,
      "grad_norm": 5.5625,
      "learning_rate": 2.8590492735312698e-05,
      "loss": 0.3957,
      "mean_token_accuracy": 0.9152343153953553,
      "step": 595,
      "train/kl_loss_qwen": 0.02967059575021267,
      "train/kl_loss_r1": 0.08191915675997734,
      "train/total_kl": 0.11158975511789322
    },
    {
      "epoch": 0.18951358180669614,
      "grad_norm": 2.875,
      "learning_rate": 2.857864813644978e-05,
      "loss": 0.4017,
      "mean_token_accuracy": 0.9134599685668945,
      "step": 600,
      "train/kl_loss_qwen": 0.0327123086899519,
      "train/kl_loss_r1": 0.09271356016397476,
      "train/total_kl": 0.12542586922645568
    },
    {
      "epoch": 0.1910928616550853,
      "grad_norm": 3.328125,
      "learning_rate": 2.8566803537586862e-05,
      "loss": 0.4025,
      "mean_token_accuracy": 0.9242550015449524,
      "step": 605,
      "train/kl_loss_qwen": 0.0499387551099062,
      "train/kl_loss_r1": 0.11244929134845734,
      "train/total_kl": 0.16238804608583451
    },
    {
      "epoch": 0.19267214150347442,
      "grad_norm": 5.3125,
      "learning_rate": 2.8554958938723944e-05,
      "loss": 0.4036,
      "mean_token_accuracy": 0.9153047204017639,
      "step": 610,
      "train/kl_loss_qwen": 0.03762825503945351,
      "train/kl_loss_r1": 0.08287687674164772,
      "train/total_kl": 0.12050513178110123
    },
    {
      "epoch": 0.19425142135186355,
      "grad_norm": 4.625,
      "learning_rate": 2.8543114339861023e-05,
      "loss": 0.5771,
      "mean_token_accuracy": 0.8914734363555908,
      "step": 615,
      "train/kl_loss_qwen": 0.07855067662894726,
      "train/kl_loss_r1": 0.1495678573846817,
      "train/total_kl": 0.22811853438615798
    },
    {
      "epoch": 0.19583070120025267,
      "grad_norm": 4.03125,
      "learning_rate": 2.8531269740998108e-05,
      "loss": 0.4371,
      "mean_token_accuracy": 0.9092644453048706,
      "step": 620,
      "train/kl_loss_qwen": 0.050358827039599416,
      "train/kl_loss_r1": 0.09986221492290497,
      "train/total_kl": 0.15022104233503342
    },
    {
      "epoch": 0.19740998104864182,
      "grad_norm": 4.0625,
      "learning_rate": 2.8519425142135186e-05,
      "loss": 0.3977,
      "mean_token_accuracy": 0.9199515819549561,
      "step": 625,
      "train/kl_loss_qwen": 0.04157613757997751,
      "train/kl_loss_r1": 0.0886114403605461,
      "train/total_kl": 0.13018757998943328
    },
    {
      "epoch": 0.19898926089703095,
      "grad_norm": 3.921875,
      "learning_rate": 2.850758054327227e-05,
      "loss": 0.3932,
      "mean_token_accuracy": 0.910289752483368,
      "step": 630,
      "train/kl_loss_qwen": 0.028871962428092958,
      "train/kl_loss_r1": 0.0790595531463623,
      "train/total_kl": 0.10793151557445527
    },
    {
      "epoch": 0.20056854074542008,
      "grad_norm": 4.40625,
      "learning_rate": 2.849573594440935e-05,
      "loss": 0.3877,
      "mean_token_accuracy": 0.924757432937622,
      "step": 635,
      "train/kl_loss_qwen": 0.045140248537063596,
      "train/kl_loss_r1": 0.09230539053678513,
      "train/total_kl": 0.1374456375837326
    },
    {
      "epoch": 0.20214782059380923,
      "grad_norm": 6.78125,
      "learning_rate": 2.848389134554643e-05,
      "loss": 0.4605,
      "mean_token_accuracy": 0.9073841214179993,
      "step": 640,
      "train/kl_loss_qwen": 0.04638505652546883,
      "train/kl_loss_r1": 0.10768231302499771,
      "train/total_kl": 0.15406737327575684
    },
    {
      "epoch": 0.20372710044219836,
      "grad_norm": 4.6875,
      "learning_rate": 2.8472046746683514e-05,
      "loss": 0.3801,
      "mean_token_accuracy": 0.9125670075416565,
      "step": 645,
      "train/kl_loss_qwen": 0.025409231334924696,
      "train/kl_loss_r1": 0.07233659029006959,
      "train/total_kl": 0.09774582087993622
    },
    {
      "epoch": 0.20530638029058748,
      "grad_norm": 4.96875,
      "learning_rate": 2.8460202147820593e-05,
      "loss": 0.4244,
      "mean_token_accuracy": 0.9137134909629822,
      "step": 650,
      "train/kl_loss_qwen": 0.044654354080557825,
      "train/kl_loss_r1": 0.11554057672619819,
      "train/total_kl": 0.160194930434227
    },
    {
      "epoch": 0.20688566013897663,
      "grad_norm": 4.34375,
      "learning_rate": 2.8448357548957675e-05,
      "loss": 0.4776,
      "mean_token_accuracy": 0.9174468874931335,
      "step": 655,
      "train/kl_loss_qwen": 0.0691050760447979,
      "train/kl_loss_r1": 0.15570700615644456,
      "train/total_kl": 0.2248120754957199
    },
    {
      "epoch": 0.20846493998736576,
      "grad_norm": 5.9375,
      "learning_rate": 2.8436512950094757e-05,
      "loss": 0.4497,
      "mean_token_accuracy": 0.9179223418235779,
      "step": 660,
      "train/kl_loss_qwen": 0.05040717124938965,
      "train/kl_loss_r1": 0.10901356935501098,
      "train/total_kl": 0.15942074060440065
    },
    {
      "epoch": 0.2100442198357549,
      "grad_norm": 4.15625,
      "learning_rate": 2.842466835123184e-05,
      "loss": 0.3903,
      "mean_token_accuracy": 0.918436586856842,
      "step": 665,
      "train/kl_loss_qwen": 0.03432295508682728,
      "train/kl_loss_r1": 0.09101127088069916,
      "train/total_kl": 0.1253342241048813
    },
    {
      "epoch": 0.21162349968414404,
      "grad_norm": 4.21875,
      "learning_rate": 2.841282375236892e-05,
      "loss": 0.4757,
      "mean_token_accuracy": 0.9196832418441773,
      "step": 670,
      "train/kl_loss_qwen": 0.06941994354128837,
      "train/kl_loss_r1": 0.15589446872472762,
      "train/total_kl": 0.22531441301107408
    },
    {
      "epoch": 0.21320277953253317,
      "grad_norm": 4.25,
      "learning_rate": 2.8400979153506003e-05,
      "loss": 0.4044,
      "mean_token_accuracy": 0.9070070743560791,
      "step": 675,
      "train/kl_loss_qwen": 0.02508469931781292,
      "train/kl_loss_r1": 0.0729275494813919,
      "train/total_kl": 0.09801224917173386
    },
    {
      "epoch": 0.2147820593809223,
      "grad_norm": 3.984375,
      "learning_rate": 2.838913455464308e-05,
      "loss": 0.3684,
      "mean_token_accuracy": 0.920681643486023,
      "step": 680,
      "train/kl_loss_qwen": 0.038502507656812665,
      "train/kl_loss_r1": 0.09550532251596451,
      "train/total_kl": 0.13400783091783525
    },
    {
      "epoch": 0.21636133922931144,
      "grad_norm": 5.71875,
      "learning_rate": 2.8377289955780167e-05,
      "loss": 0.405,
      "mean_token_accuracy": 0.9182163357734681,
      "step": 685,
      "train/kl_loss_qwen": 0.04167989417910576,
      "train/kl_loss_r1": 0.09086664766073227,
      "train/total_kl": 0.1325465440750122
    },
    {
      "epoch": 0.21794061907770057,
      "grad_norm": 3.5,
      "learning_rate": 2.8365445356917245e-05,
      "loss": 0.3854,
      "mean_token_accuracy": 0.9178756952285767,
      "step": 690,
      "train/kl_loss_qwen": 0.02976965680718422,
      "train/kl_loss_r1": 0.07755026966333389,
      "train/total_kl": 0.10731992721557618
    },
    {
      "epoch": 0.2195198989260897,
      "grad_norm": 3.59375,
      "learning_rate": 2.8353600758054327e-05,
      "loss": 0.4816,
      "mean_token_accuracy": 0.8993646502494812,
      "step": 695,
      "train/kl_loss_qwen": 0.05840725116431713,
      "train/kl_loss_r1": 0.09280273914337159,
      "train/total_kl": 0.1512099876999855
    },
    {
      "epoch": 0.22109917877447885,
      "grad_norm": 5.78125,
      "learning_rate": 2.834175615919141e-05,
      "loss": 0.4719,
      "mean_token_accuracy": 0.9162763595581055,
      "step": 700,
      "train/kl_loss_qwen": 0.060653197765350345,
      "train/kl_loss_r1": 0.12878852635622023,
      "train/total_kl": 0.18944172114133834
    },
    {
      "epoch": 0.22267845862286798,
      "grad_norm": 5.53125,
      "learning_rate": 2.832991156032849e-05,
      "loss": 0.9092,
      "mean_token_accuracy": 0.8932477593421936,
      "step": 705,
      "train/kl_loss_qwen": 0.16318726614117623,
      "train/kl_loss_r1": 0.42447653263807295,
      "train/total_kl": 0.587663796544075
    },
    {
      "epoch": 0.2242577384712571,
      "grad_norm": 3.09375,
      "learning_rate": 2.8318066961465573e-05,
      "loss": 0.406,
      "mean_token_accuracy": 0.9242049813270569,
      "step": 710,
      "train/kl_loss_qwen": 0.04510703757405281,
      "train/kl_loss_r1": 0.11025056838989258,
      "train/total_kl": 0.15535760223865508
    },
    {
      "epoch": 0.22583701831964625,
      "grad_norm": 2.71875,
      "learning_rate": 2.8306222362602655e-05,
      "loss": 0.3393,
      "mean_token_accuracy": 0.9313870310783386,
      "step": 715,
      "train/kl_loss_qwen": 0.02871016077697277,
      "train/kl_loss_r1": 0.09356620982289314,
      "train/total_kl": 0.12227637022733688
    },
    {
      "epoch": 0.22741629816803538,
      "grad_norm": 5.0625,
      "learning_rate": 2.8294377763739734e-05,
      "loss": 0.4224,
      "mean_token_accuracy": 0.9268312335014344,
      "step": 720,
      "train/kl_loss_qwen": 0.0612876083701849,
      "train/kl_loss_r1": 0.13620872125029565,
      "train/total_kl": 0.19749632626771926
    },
    {
      "epoch": 0.2289955780164245,
      "grad_norm": 5.9375,
      "learning_rate": 2.828253316487682e-05,
      "loss": 0.4011,
      "mean_token_accuracy": 0.9213597774505615,
      "step": 725,
      "train/kl_loss_qwen": 0.04712501242756843,
      "train/kl_loss_r1": 0.10113193169236183,
      "train/total_kl": 0.14825694710016252
    },
    {
      "epoch": 0.23057485786481363,
      "grad_norm": 5.21875,
      "learning_rate": 2.8270688566013898e-05,
      "loss": 0.4406,
      "mean_token_accuracy": 0.9167027950286866,
      "step": 730,
      "train/kl_loss_qwen": 0.04809342883527279,
      "train/kl_loss_r1": 0.11717957258224487,
      "train/total_kl": 0.1652730032801628
    },
    {
      "epoch": 0.23215413771320279,
      "grad_norm": 4.3125,
      "learning_rate": 2.825884396715098e-05,
      "loss": 0.8128,
      "mean_token_accuracy": 0.8959693789482117,
      "step": 735,
      "train/kl_loss_qwen": 0.1253665577620268,
      "train/kl_loss_r1": 0.3655559465289116,
      "train/total_kl": 0.49092249274253846
    },
    {
      "epoch": 0.2337334175615919,
      "grad_norm": 4.25,
      "learning_rate": 2.824699936828806e-05,
      "loss": 0.3931,
      "mean_token_accuracy": 0.9106727838516235,
      "step": 740,
      "train/kl_loss_qwen": 0.025178911350667476,
      "train/kl_loss_r1": 0.0726950392127037,
      "train/total_kl": 0.09787395298480987
    },
    {
      "epoch": 0.23531269740998104,
      "grad_norm": 4.71875,
      "learning_rate": 2.8235154769425144e-05,
      "loss": 0.6604,
      "mean_token_accuracy": 0.90933278799057,
      "step": 745,
      "train/kl_loss_qwen": 0.09659561701118946,
      "train/kl_loss_r1": 0.28312878832221033,
      "train/total_kl": 0.37972441166639326
    },
    {
      "epoch": 0.2368919772583702,
      "grad_norm": 3.046875,
      "learning_rate": 2.8223310170562226e-05,
      "loss": 0.3598,
      "mean_token_accuracy": 0.9175713539123536,
      "step": 750,
      "train/kl_loss_qwen": 0.025853770971298217,
      "train/kl_loss_r1": 0.07482494115829467,
      "train/total_kl": 0.1006787121295929
    },
    {
      "epoch": 0.23847125710675932,
      "grad_norm": 3.6875,
      "learning_rate": 2.8211465571699304e-05,
      "loss": 0.3764,
      "mean_token_accuracy": 0.9117866277694702,
      "step": 755,
      "train/kl_loss_qwen": 0.02956481948494911,
      "train/kl_loss_r1": 0.08544625118374824,
      "train/total_kl": 0.1150110736489296
    },
    {
      "epoch": 0.24005053695514844,
      "grad_norm": 4.40625,
      "learning_rate": 2.8199620972836386e-05,
      "loss": 0.4058,
      "mean_token_accuracy": 0.9138393759727478,
      "step": 760,
      "train/kl_loss_qwen": 0.03950389288365841,
      "train/kl_loss_r1": 0.10766673237085342,
      "train/total_kl": 0.14717062562704086
    },
    {
      "epoch": 0.2416298168035376,
      "grad_norm": 3.890625,
      "learning_rate": 2.8187776373973468e-05,
      "loss": 0.3642,
      "mean_token_accuracy": 0.9200916647911072,
      "step": 765,
      "train/kl_loss_qwen": 0.03795309662818909,
      "train/kl_loss_r1": 0.08652669936418533,
      "train/total_kl": 0.1244797945022583
    },
    {
      "epoch": 0.24320909665192672,
      "grad_norm": 4.09375,
      "learning_rate": 2.817593177511055e-05,
      "loss": 0.3802,
      "mean_token_accuracy": 0.9233215928077698,
      "step": 770,
      "train/kl_loss_qwen": 0.04185252301394939,
      "train/kl_loss_r1": 0.10097809582948684,
      "train/total_kl": 0.14283061921596527
    },
    {
      "epoch": 0.24478837650031585,
      "grad_norm": 4.25,
      "learning_rate": 2.816408717624763e-05,
      "loss": 0.3846,
      "mean_token_accuracy": 0.9269769430160523,
      "step": 775,
      "train/kl_loss_qwen": 0.04208484329283237,
      "train/kl_loss_r1": 0.10242608338594436,
      "train/total_kl": 0.14451092779636382
    },
    {
      "epoch": 0.246367656348705,
      "grad_norm": 3.734375,
      "learning_rate": 2.8152242577384714e-05,
      "loss": 0.3689,
      "mean_token_accuracy": 0.9251198291778564,
      "step": 780,
      "train/kl_loss_qwen": 0.042654494568705556,
      "train/kl_loss_r1": 0.0871317096054554,
      "train/total_kl": 0.12978620529174806
    },
    {
      "epoch": 0.24794693619709413,
      "grad_norm": 4.09375,
      "learning_rate": 2.8140397978521793e-05,
      "loss": 0.528,
      "mean_token_accuracy": 0.9183760046958923,
      "step": 785,
      "train/kl_loss_qwen": 0.08097701780498028,
      "train/kl_loss_r1": 0.18653534501791,
      "train/total_kl": 0.2675123676657677
    },
    {
      "epoch": 0.24952621604548325,
      "grad_norm": 5.46875,
      "learning_rate": 2.8128553379658878e-05,
      "loss": 0.4215,
      "mean_token_accuracy": 0.9100816130638123,
      "step": 790,
      "train/kl_loss_qwen": 0.035708888247609136,
      "train/kl_loss_r1": 0.09558346644043922,
      "train/total_kl": 0.1312923550605774
    },
    {
      "epoch": 0.2511054958938724,
      "grad_norm": 3.46875,
      "learning_rate": 2.8116708780795957e-05,
      "loss": 0.4392,
      "mean_token_accuracy": 0.9230265855789185,
      "step": 795,
      "train/kl_loss_qwen": 0.06153766680508852,
      "train/kl_loss_r1": 0.1311558410525322,
      "train/total_kl": 0.19269351065158843
    },
    {
      "epoch": 0.2526847757422615,
      "grad_norm": 3.15625,
      "learning_rate": 2.810486418193304e-05,
      "loss": 0.3672,
      "mean_token_accuracy": 0.9224226951599122,
      "step": 800,
      "train/kl_loss_qwen": 0.051443108916282655,
      "train/kl_loss_r1": 0.08029818087816239,
      "train/total_kl": 0.13174128979444505
    },
    {
      "epoch": 0.2542640555906507,
      "grad_norm": 3.640625,
      "learning_rate": 2.809301958307012e-05,
      "loss": 0.4402,
      "mean_token_accuracy": 0.9059389233589172,
      "step": 805,
      "train/kl_loss_qwen": 0.049551712721586226,
      "train/kl_loss_r1": 0.10702142640948295,
      "train/total_kl": 0.15657313615083696
    },
    {
      "epoch": 0.2558433354390398,
      "grad_norm": 4.3125,
      "learning_rate": 2.8081174984207203e-05,
      "loss": 0.3903,
      "mean_token_accuracy": 0.9165263414382935,
      "step": 810,
      "train/kl_loss_qwen": 0.019741023890674114,
      "train/kl_loss_r1": 0.07913172543048859,
      "train/total_kl": 0.0988727480173111
    },
    {
      "epoch": 0.25742261528742894,
      "grad_norm": 5.28125,
      "learning_rate": 2.8069330385344285e-05,
      "loss": 0.3909,
      "mean_token_accuracy": 0.9190155744552613,
      "step": 815,
      "train/kl_loss_qwen": 0.03204708211123943,
      "train/kl_loss_r1": 0.09303590878844262,
      "train/total_kl": 0.1250829890370369
    },
    {
      "epoch": 0.25900189513581806,
      "grad_norm": 2.875,
      "learning_rate": 2.8057485786481366e-05,
      "loss": 0.5543,
      "mean_token_accuracy": 0.9089938521385192,
      "step": 820,
      "train/kl_loss_qwen": 0.06574608832597732,
      "train/kl_loss_r1": 0.20562416017055513,
      "train/total_kl": 0.27137024998664855
    },
    {
      "epoch": 0.2605811749842072,
      "grad_norm": 4.21875,
      "learning_rate": 2.8045641187618445e-05,
      "loss": 0.392,
      "mean_token_accuracy": 0.9160252213478088,
      "step": 825,
      "train/kl_loss_qwen": 0.030706890672445298,
      "train/kl_loss_r1": 0.07311581745743752,
      "train/total_kl": 0.10382270887494087
    },
    {
      "epoch": 0.2621604548325963,
      "grad_norm": 5.3125,
      "learning_rate": 2.803379658875553e-05,
      "loss": 0.4262,
      "mean_token_accuracy": 0.9091088771820068,
      "step": 830,
      "train/kl_loss_qwen": 0.0335204778239131,
      "train/kl_loss_r1": 0.09335571229457855,
      "train/total_kl": 0.12687619179487228
    },
    {
      "epoch": 0.2637397346809855,
      "grad_norm": 3.625,
      "learning_rate": 2.802195198989261e-05,
      "loss": 0.3926,
      "mean_token_accuracy": 0.917055356502533,
      "step": 835,
      "train/kl_loss_qwen": 0.03976532854139805,
      "train/kl_loss_r1": 0.09741715267300606,
      "train/total_kl": 0.13718248456716536
    },
    {
      "epoch": 0.2653190145293746,
      "grad_norm": 5.40625,
      "learning_rate": 2.801010739102969e-05,
      "loss": 0.482,
      "mean_token_accuracy": 0.9025979518890381,
      "step": 840,
      "train/kl_loss_qwen": 0.04998598210513592,
      "train/kl_loss_r1": 0.1031669408082962,
      "train/total_kl": 0.15315292179584503
    },
    {
      "epoch": 0.26689829437776375,
      "grad_norm": 3.359375,
      "learning_rate": 2.7998262792166773e-05,
      "loss": 0.409,
      "mean_token_accuracy": 0.9056584239006042,
      "step": 845,
      "train/kl_loss_qwen": 0.03249505795538425,
      "train/kl_loss_r1": 0.07019762918353081,
      "train/total_kl": 0.10269268751144409
    },
    {
      "epoch": 0.26847757422615287,
      "grad_norm": 4.21875,
      "learning_rate": 2.7986418193303855e-05,
      "loss": 0.4027,
      "mean_token_accuracy": 0.9153900265693664,
      "step": 850,
      "train/kl_loss_qwen": 0.05629720836877823,
      "train/kl_loss_r1": 0.09726203978061676,
      "train/total_kl": 0.15355924367904664
    },
    {
      "epoch": 0.270056854074542,
      "grad_norm": 4.75,
      "learning_rate": 2.7974573594440937e-05,
      "loss": 0.4788,
      "mean_token_accuracy": 0.9220892786979675,
      "step": 855,
      "train/kl_loss_qwen": 0.0792582355439663,
      "train/kl_loss_r1": 0.1329452820122242,
      "train/total_kl": 0.21220352053642272
    },
    {
      "epoch": 0.2716361339229311,
      "grad_norm": 5.59375,
      "learning_rate": 2.796272899557802e-05,
      "loss": 0.4173,
      "mean_token_accuracy": 0.9084758758544922,
      "step": 860,
      "train/kl_loss_qwen": 0.024994323402643202,
      "train/kl_loss_r1": 0.07485152631998063,
      "train/total_kl": 0.09984585046768188
    },
    {
      "epoch": 0.2732154137713203,
      "grad_norm": 3.671875,
      "learning_rate": 2.7950884396715097e-05,
      "loss": 0.3938,
      "mean_token_accuracy": 0.9122233152389526,
      "step": 865,
      "train/kl_loss_qwen": 0.037501578032970426,
      "train/kl_loss_r1": 0.08154807239770889,
      "train/total_kl": 0.11904965043067932
    },
    {
      "epoch": 0.27479469361970943,
      "grad_norm": 4.3125,
      "learning_rate": 2.7939039797852183e-05,
      "loss": 0.4368,
      "mean_token_accuracy": 0.9168960452079773,
      "step": 870,
      "train/kl_loss_qwen": 0.052475106716156,
      "train/kl_loss_r1": 0.10713397711515427,
      "train/total_kl": 0.15960908234119414
    },
    {
      "epoch": 0.27637397346809855,
      "grad_norm": 6.9375,
      "learning_rate": 2.792719519898926e-05,
      "loss": 0.4174,
      "mean_token_accuracy": 0.9178502202033997,
      "step": 875,
      "train/kl_loss_qwen": 0.047308037802577016,
      "train/kl_loss_r1": 0.13246602714061737,
      "train/total_kl": 0.1797740638256073
    },
    {
      "epoch": 0.2779532533164877,
      "grad_norm": 4.375,
      "learning_rate": 2.791535060012634e-05,
      "loss": 0.3678,
      "mean_token_accuracy": 0.9193345308303833,
      "step": 880,
      "train/kl_loss_qwen": 0.04378705471754074,
      "train/kl_loss_r1": 0.08930878117680549,
      "train/total_kl": 0.13309583812952042
    },
    {
      "epoch": 0.2795325331648768,
      "grad_norm": 4.21875,
      "learning_rate": 2.7903506001263425e-05,
      "loss": 0.4736,
      "mean_token_accuracy": 0.8888418793678283,
      "step": 885,
      "train/kl_loss_qwen": 0.03299890570342541,
      "train/kl_loss_r1": 0.07775545120239258,
      "train/total_kl": 0.11075435876846314
    },
    {
      "epoch": 0.28111181301326593,
      "grad_norm": 6.375,
      "learning_rate": 2.7891661402400504e-05,
      "loss": 0.3939,
      "mean_token_accuracy": 0.9237197518348694,
      "step": 890,
      "train/kl_loss_qwen": 0.04137386232614517,
      "train/kl_loss_r1": 0.10401555001735688,
      "train/total_kl": 0.14538941085338591
    },
    {
      "epoch": 0.28269109286165506,
      "grad_norm": 4.46875,
      "learning_rate": 2.787981680353759e-05,
      "loss": 0.4189,
      "mean_token_accuracy": 0.9166470527648926,
      "step": 895,
      "train/kl_loss_qwen": 0.04946824684739113,
      "train/kl_loss_r1": 0.11216580867767334,
      "train/total_kl": 0.16163405478000642
    },
    {
      "epoch": 0.28427037271004424,
      "grad_norm": 4.09375,
      "learning_rate": 2.7867972204674668e-05,
      "loss": 0.4105,
      "mean_token_accuracy": 0.9106549739837646,
      "step": 900,
      "train/kl_loss_qwen": 0.050147952511906624,
      "train/kl_loss_r1": 0.09455743432044983,
      "train/total_kl": 0.14470538944005967
    },
    {
      "epoch": 0.28584965255843336,
      "grad_norm": 3.90625,
      "learning_rate": 2.785612760581175e-05,
      "loss": 0.3611,
      "mean_token_accuracy": 0.9273918032646179,
      "step": 905,
      "train/kl_loss_qwen": 0.033574892580509184,
      "train/kl_loss_r1": 0.08916549682617188,
      "train/total_kl": 0.12274039089679718
    },
    {
      "epoch": 0.2874289324068225,
      "grad_norm": 3.234375,
      "learning_rate": 2.7844283006948832e-05,
      "loss": 0.3666,
      "mean_token_accuracy": 0.927341103553772,
      "step": 910,
      "train/kl_loss_qwen": 0.046492438018321994,
      "train/kl_loss_r1": 0.08254126682877541,
      "train/total_kl": 0.1290337026119232
    },
    {
      "epoch": 0.2890082122552116,
      "grad_norm": 4.5625,
      "learning_rate": 2.7832438408085914e-05,
      "loss": 0.4609,
      "mean_token_accuracy": 0.904936671257019,
      "step": 915,
      "train/kl_loss_qwen": 0.04765871986746788,
      "train/kl_loss_r1": 0.10178251266479492,
      "train/total_kl": 0.1494412302970886
    },
    {
      "epoch": 0.29058749210360074,
      "grad_norm": 3.65625,
      "learning_rate": 2.7820593809222992e-05,
      "loss": 0.3945,
      "mean_token_accuracy": 0.9106269359588623,
      "step": 920,
      "train/kl_loss_qwen": 0.02984564509242773,
      "train/kl_loss_r1": 0.07595031410455703,
      "train/total_kl": 0.10579595863819122
    },
    {
      "epoch": 0.29216677195198987,
      "grad_norm": 3.203125,
      "learning_rate": 2.7808749210360078e-05,
      "loss": 0.3549,
      "mean_token_accuracy": 0.9191085696220398,
      "step": 925,
      "train/kl_loss_qwen": 0.03080688826739788,
      "train/kl_loss_r1": 0.0738538570702076,
      "train/total_kl": 0.10466074496507645
    },
    {
      "epoch": 0.29374605180037905,
      "grad_norm": 2.90625,
      "learning_rate": 2.7796904611497156e-05,
      "loss": 0.4382,
      "mean_token_accuracy": 0.9180479645729065,
      "step": 930,
      "train/kl_loss_qwen": 0.06083466187119484,
      "train/kl_loss_r1": 0.1260865792632103,
      "train/total_kl": 0.18692124187946318
    },
    {
      "epoch": 0.2953253316487682,
      "grad_norm": 3.828125,
      "learning_rate": 2.7785060012634242e-05,
      "loss": 0.399,
      "mean_token_accuracy": 0.9212388038635254,
      "step": 935,
      "train/kl_loss_qwen": 0.03766842745244503,
      "train/kl_loss_r1": 0.10632861033082008,
      "train/total_kl": 0.14399703592061996
    },
    {
      "epoch": 0.2969046114971573,
      "grad_norm": 4.25,
      "learning_rate": 2.777321541377132e-05,
      "loss": 0.4375,
      "mean_token_accuracy": 0.916726815700531,
      "step": 940,
      "train/kl_loss_qwen": 0.047170663997530934,
      "train/kl_loss_r1": 0.1259485125541687,
      "train/total_kl": 0.17311917692422868
    },
    {
      "epoch": 0.2984838913455464,
      "grad_norm": 3.8125,
      "learning_rate": 2.7761370814908402e-05,
      "loss": 0.4013,
      "mean_token_accuracy": 0.9178184151649476,
      "step": 945,
      "train/kl_loss_qwen": 0.03914799019694328,
      "train/kl_loss_r1": 0.09711084365844727,
      "train/total_kl": 0.1362588331103325
    },
    {
      "epoch": 0.30006317119393555,
      "grad_norm": 2.90625,
      "learning_rate": 2.7749526216045484e-05,
      "loss": 0.3779,
      "mean_token_accuracy": 0.9126025319099427,
      "step": 950,
      "train/kl_loss_qwen": 0.0351341400295496,
      "train/kl_loss_r1": 0.07189167514443398,
      "train/total_kl": 0.10702581554651261
    },
    {
      "epoch": 0.3016424510423247,
      "grad_norm": 2.5625,
      "learning_rate": 2.7737681617182566e-05,
      "loss": 0.3911,
      "mean_token_accuracy": 0.9192399621009827,
      "step": 955,
      "train/kl_loss_qwen": 0.040135934948921204,
      "train/kl_loss_r1": 0.0889073669910431,
      "train/total_kl": 0.1290433004498482
    },
    {
      "epoch": 0.30322173089071386,
      "grad_norm": 6.6875,
      "learning_rate": 2.7725837018319645e-05,
      "loss": 0.3622,
      "mean_token_accuracy": 0.9284344434738159,
      "step": 960,
      "train/kl_loss_qwen": 0.03907502368092537,
      "train/kl_loss_r1": 0.0967540517449379,
      "train/total_kl": 0.1358290746808052
    },
    {
      "epoch": 0.304801010739103,
      "grad_norm": 4.3125,
      "learning_rate": 2.771399241945673e-05,
      "loss": 0.3965,
      "mean_token_accuracy": 0.9121941685676574,
      "step": 965,
      "train/kl_loss_qwen": 0.02913835383951664,
      "train/kl_loss_r1": 0.07579499706625939,
      "train/total_kl": 0.10493335276842117
    },
    {
      "epoch": 0.3063802905874921,
      "grad_norm": 3.328125,
      "learning_rate": 2.770214782059381e-05,
      "loss": 0.4335,
      "mean_token_accuracy": 0.9208302021026611,
      "step": 970,
      "train/kl_loss_qwen": 0.05558455996215343,
      "train/kl_loss_r1": 0.13190918788313866,
      "train/total_kl": 0.18749375492334366
    },
    {
      "epoch": 0.30795957043588124,
      "grad_norm": 3.90625,
      "learning_rate": 2.7690303221730894e-05,
      "loss": 0.3862,
      "mean_token_accuracy": 0.915570068359375,
      "step": 975,
      "train/kl_loss_qwen": 0.037285884469747545,
      "train/kl_loss_r1": 0.08280575126409531,
      "train/total_kl": 0.12009163498878479
    },
    {
      "epoch": 0.30953885028427036,
      "grad_norm": 3.4375,
      "learning_rate": 2.7678458622867973e-05,
      "loss": 0.3688,
      "mean_token_accuracy": 0.9202610492706299,
      "step": 980,
      "train/kl_loss_qwen": 0.032963752374053,
      "train/kl_loss_r1": 0.07818343862891197,
      "train/total_kl": 0.11114719212055206
    },
    {
      "epoch": 0.3111181301326595,
      "grad_norm": 7.6875,
      "learning_rate": 2.7666614024005055e-05,
      "loss": 0.4478,
      "mean_token_accuracy": 0.920277726650238,
      "step": 985,
      "train/kl_loss_qwen": 0.05329423062503338,
      "train/kl_loss_r1": 0.10826097205281257,
      "train/total_kl": 0.16155520379543303
    },
    {
      "epoch": 0.31269740998104867,
      "grad_norm": 3.453125,
      "learning_rate": 2.7654769425142137e-05,
      "loss": 0.3699,
      "mean_token_accuracy": 0.9260859251022339,
      "step": 990,
      "train/kl_loss_qwen": 0.03959104530513287,
      "train/kl_loss_r1": 0.09852487146854401,
      "train/total_kl": 0.13811591565608977
    },
    {
      "epoch": 0.3142766898294378,
      "grad_norm": 3.71875,
      "learning_rate": 2.764292482627922e-05,
      "loss": 0.3693,
      "mean_token_accuracy": 0.9272526383399964,
      "step": 995,
      "train/kl_loss_qwen": 0.04407541677355766,
      "train/kl_loss_r1": 0.09461856633424759,
      "train/total_kl": 0.1386939838528633
    },
    {
      "epoch": 0.3158559696778269,
      "grad_norm": 3.046875,
      "learning_rate": 2.7631080227416297e-05,
      "loss": 0.463,
      "mean_token_accuracy": 0.9055490493774414,
      "step": 1000,
      "train/kl_loss_qwen": 0.048449795320630075,
      "train/kl_loss_r1": 0.1313101351261139,
      "train/total_kl": 0.1797599345445633
    },
    {
      "epoch": 0.31743524952621605,
      "grad_norm": 7.09375,
      "learning_rate": 2.7619235628553383e-05,
      "loss": 0.3813,
      "mean_token_accuracy": 0.9248928785324096,
      "step": 1005,
      "train/kl_loss_qwen": 0.05526920892298222,
      "train/kl_loss_r1": 0.09118993952870369,
      "train/total_kl": 0.14645914733409882
    },
    {
      "epoch": 0.31901452937460517,
      "grad_norm": 7.28125,
      "learning_rate": 2.760739102969046e-05,
      "loss": 0.409,
      "mean_token_accuracy": 0.9236547231674195,
      "step": 1010,
      "train/kl_loss_qwen": 0.05414267405867577,
      "train/kl_loss_r1": 0.12399578541517257,
      "train/total_kl": 0.17813845425844194
    },
    {
      "epoch": 0.3205938092229943,
      "grad_norm": 3.75,
      "learning_rate": 2.7595546430827543e-05,
      "loss": 0.3638,
      "mean_token_accuracy": 0.9224517822265625,
      "step": 1015,
      "train/kl_loss_qwen": 0.03951701745390892,
      "train/kl_loss_r1": 0.08569068014621735,
      "train/total_kl": 0.12520769685506822
    },
    {
      "epoch": 0.3221730890713834,
      "grad_norm": 3.734375,
      "learning_rate": 2.7583701831964625e-05,
      "loss": 0.3657,
      "mean_token_accuracy": 0.9115802645683289,
      "step": 1020,
      "train/kl_loss_qwen": 0.02692534662783146,
      "train/kl_loss_r1": 0.07535937055945396,
      "train/total_kl": 0.10228471606969833
    },
    {
      "epoch": 0.3237523689197726,
      "grad_norm": 2.9375,
      "learning_rate": 2.7571857233101704e-05,
      "loss": 0.3757,
      "mean_token_accuracy": 0.9229058980941772,
      "step": 1025,
      "train/kl_loss_qwen": 0.042735861241817476,
      "train/kl_loss_r1": 0.0955415666103363,
      "train/total_kl": 0.13827742636203766
    },
    {
      "epoch": 0.32533164876816173,
      "grad_norm": 3.859375,
      "learning_rate": 2.756001263423879e-05,
      "loss": 0.4294,
      "mean_token_accuracy": 0.9180144786834716,
      "step": 1030,
      "train/kl_loss_qwen": 0.042747174575924876,
      "train/kl_loss_r1": 0.12570120468735696,
      "train/total_kl": 0.16844838261604309
    },
    {
      "epoch": 0.32691092861655086,
      "grad_norm": 3.46875,
      "learning_rate": 2.7548168035375868e-05,
      "loss": 0.3877,
      "mean_token_accuracy": 0.9176273584365845,
      "step": 1035,
      "train/kl_loss_qwen": 0.031073962897062303,
      "train/kl_loss_r1": 0.08218887373805046,
      "train/total_kl": 0.11326283663511276
    },
    {
      "epoch": 0.32849020846494,
      "grad_norm": 3.515625,
      "learning_rate": 2.753632343651295e-05,
      "loss": 0.3954,
      "mean_token_accuracy": 0.9231147408485413,
      "step": 1040,
      "train/kl_loss_qwen": 0.05052623115479946,
      "train/kl_loss_r1": 0.11141537651419639,
      "train/total_kl": 0.16194161027669907
    },
    {
      "epoch": 0.3300694883133291,
      "grad_norm": 3.578125,
      "learning_rate": 2.752447883765003e-05,
      "loss": 0.3982,
      "mean_token_accuracy": 0.9038361191749573,
      "step": 1045,
      "train/kl_loss_qwen": 0.02011768762022257,
      "train/kl_loss_r1": 0.058029362559318544,
      "train/total_kl": 0.07814705073833465
    },
    {
      "epoch": 0.33164876816171823,
      "grad_norm": 5.375,
      "learning_rate": 2.7512634238787114e-05,
      "loss": 0.4198,
      "mean_token_accuracy": 0.9237027764320374,
      "step": 1050,
      "train/kl_loss_qwen": 0.049149174243211746,
      "train/kl_loss_r1": 0.12267882525920867,
      "train/total_kl": 0.17182800471782683
    },
    {
      "epoch": 0.3332280480101074,
      "grad_norm": 2.890625,
      "learning_rate": 2.7500789639924196e-05,
      "loss": 0.3994,
      "mean_token_accuracy": 0.9218345522880554,
      "step": 1055,
      "train/kl_loss_qwen": 0.047988555207848546,
      "train/kl_loss_r1": 0.10088327825069428,
      "train/total_kl": 0.1488718345761299
    },
    {
      "epoch": 0.33480732785849654,
      "grad_norm": 3.71875,
      "learning_rate": 2.7488945041061277e-05,
      "loss": 0.4307,
      "mean_token_accuracy": 0.9113677144050598,
      "step": 1060,
      "train/kl_loss_qwen": 0.045209009945392606,
      "train/kl_loss_r1": 0.09304542914032936,
      "train/total_kl": 0.13825443983078003
    },
    {
      "epoch": 0.33638660770688567,
      "grad_norm": 2.953125,
      "learning_rate": 2.7477100442198356e-05,
      "loss": 0.3365,
      "mean_token_accuracy": 0.9255304574966431,
      "step": 1065,
      "train/kl_loss_qwen": 0.028990533947944642,
      "train/kl_loss_r1": 0.07403095960617065,
      "train/total_kl": 0.10302149206399917
    },
    {
      "epoch": 0.3379658875552748,
      "grad_norm": 4.8125,
      "learning_rate": 2.746525584333544e-05,
      "loss": 0.4212,
      "mean_token_accuracy": 0.9104359388351441,
      "step": 1070,
      "train/kl_loss_qwen": 0.041067682951688764,
      "train/kl_loss_r1": 0.08660182803869247,
      "train/total_kl": 0.1276695102453232
    },
    {
      "epoch": 0.3395451674036639,
      "grad_norm": 4.59375,
      "learning_rate": 2.745341124447252e-05,
      "loss": 0.3825,
      "mean_token_accuracy": 0.920796000957489,
      "step": 1075,
      "train/kl_loss_qwen": 0.030382947996258736,
      "train/kl_loss_r1": 0.09011238515377044,
      "train/total_kl": 0.12049532979726792
    },
    {
      "epoch": 0.34112444725205304,
      "grad_norm": 4.625,
      "learning_rate": 2.7441566645609605e-05,
      "loss": 0.4569,
      "mean_token_accuracy": 0.9030450224876404,
      "step": 1080,
      "train/kl_loss_qwen": 0.035007006302475926,
      "train/kl_loss_r1": 0.09860124588012695,
      "train/total_kl": 0.13360825404524804
    },
    {
      "epoch": 0.3427037271004422,
      "grad_norm": 4.09375,
      "learning_rate": 2.7429722046746684e-05,
      "loss": 0.6618,
      "mean_token_accuracy": 0.9075309634208679,
      "step": 1085,
      "train/kl_loss_qwen": 0.10582229532301426,
      "train/kl_loss_r1": 0.26733664721250533,
      "train/total_kl": 0.3731589362025261
    },
    {
      "epoch": 0.34428300694883135,
      "grad_norm": 4.4375,
      "learning_rate": 2.7417877447883766e-05,
      "loss": 0.3892,
      "mean_token_accuracy": 0.9164264321327209,
      "step": 1090,
      "train/kl_loss_qwen": 0.037996953912079336,
      "train/kl_loss_r1": 0.09623216241598129,
      "train/total_kl": 0.1342291161417961
    },
    {
      "epoch": 0.3458622867972205,
      "grad_norm": 4.1875,
      "learning_rate": 2.7406032849020848e-05,
      "loss": 0.403,
      "mean_token_accuracy": 0.9166207790374756,
      "step": 1095,
      "train/kl_loss_qwen": 0.03398194424808025,
      "train/kl_loss_r1": 0.0964338980615139,
      "train/total_kl": 0.130415840446949
    },
    {
      "epoch": 0.3474415666456096,
      "grad_norm": 3.6875,
      "learning_rate": 2.739418825015793e-05,
      "loss": 0.4447,
      "mean_token_accuracy": 0.9161954879760742,
      "step": 1100,
      "train/kl_loss_qwen": 0.04720754884183407,
      "train/kl_loss_r1": 0.12633319795131684,
      "train/total_kl": 0.17354074716567994
    },
    {
      "epoch": 0.3490208464939987,
      "grad_norm": 3.625,
      "learning_rate": 2.738234365129501e-05,
      "loss": 0.3868,
      "mean_token_accuracy": 0.9128393769264221,
      "step": 1105,
      "train/kl_loss_qwen": 0.023071405291557313,
      "train/kl_loss_r1": 0.07350485995411873,
      "train/total_kl": 0.09657626301050186
    },
    {
      "epoch": 0.35060012634238785,
      "grad_norm": 5.34375,
      "learning_rate": 2.7370499052432094e-05,
      "loss": 0.3996,
      "mean_token_accuracy": 0.9165781736373901,
      "step": 1110,
      "train/kl_loss_qwen": 0.03534043170511723,
      "train/kl_loss_r1": 0.09807844460010529,
      "train/total_kl": 0.13341887593269347
    },
    {
      "epoch": 0.352179406190777,
      "grad_norm": 5.125,
      "learning_rate": 2.7358654453569172e-05,
      "loss": 0.4346,
      "mean_token_accuracy": 0.9187243461608887,
      "step": 1115,
      "train/kl_loss_qwen": 0.03706376664340496,
      "train/kl_loss_r1": 0.10244923681020737,
      "train/total_kl": 0.13951300829648972
    },
    {
      "epoch": 0.35375868603916616,
      "grad_norm": 3.0625,
      "learning_rate": 2.7346809854706258e-05,
      "loss": 0.4757,
      "mean_token_accuracy": 0.9059502005577087,
      "step": 1120,
      "train/kl_loss_qwen": 0.06461062133312226,
      "train/kl_loss_r1": 0.12058839797973633,
      "train/total_kl": 0.1851990193128586
    },
    {
      "epoch": 0.3553379658875553,
      "grad_norm": 3.65625,
      "learning_rate": 2.7334965255843336e-05,
      "loss": 0.3921,
      "mean_token_accuracy": 0.9181574821472168,
      "step": 1125,
      "train/kl_loss_qwen": 0.040164635702967647,
      "train/kl_loss_r1": 0.08374351188540459,
      "train/total_kl": 0.12390814870595931
    },
    {
      "epoch": 0.3569172457359444,
      "grad_norm": 3.46875,
      "learning_rate": 2.732312065698042e-05,
      "loss": 0.4348,
      "mean_token_accuracy": 0.9025139570236206,
      "step": 1130,
      "train/kl_loss_qwen": 0.03588338419795036,
      "train/kl_loss_r1": 0.10539090931415558,
      "train/total_kl": 0.1412742927670479
    },
    {
      "epoch": 0.35849652558433354,
      "grad_norm": 3.640625,
      "learning_rate": 2.73112760581175e-05,
      "loss": 0.3664,
      "mean_token_accuracy": 0.9221108317375183,
      "step": 1135,
      "train/kl_loss_qwen": 0.03616672344505787,
      "train/kl_loss_r1": 0.08602790832519532,
      "train/total_kl": 0.12219463288784027
    },
    {
      "epoch": 0.36007580543272266,
      "grad_norm": 29.375,
      "learning_rate": 2.729943145925458e-05,
      "loss": 0.7534,
      "mean_token_accuracy": 0.9043172121047973,
      "step": 1140,
      "train/kl_loss_qwen": 0.15733794420957564,
      "train/kl_loss_r1": 0.3016031712293625,
      "train/total_kl": 0.45894111692905426
    },
    {
      "epoch": 0.3616550852811118,
      "grad_norm": 4.125,
      "learning_rate": 2.728758686039166e-05,
      "loss": 0.4045,
      "mean_token_accuracy": 0.9101644992828369,
      "step": 1145,
      "train/kl_loss_qwen": 0.044041546434164046,
      "train/kl_loss_r1": 0.08598532974720001,
      "train/total_kl": 0.13002687841653823
    },
    {
      "epoch": 0.36323436512950097,
      "grad_norm": 3.4375,
      "learning_rate": 2.7275742261528743e-05,
      "loss": 0.3902,
      "mean_token_accuracy": 0.9123161554336547,
      "step": 1150,
      "train/kl_loss_qwen": 0.0426851898431778,
      "train/kl_loss_r1": 0.08253368511795997,
      "train/total_kl": 0.1252188727259636
    },
    {
      "epoch": 0.3648136449778901,
      "grad_norm": 4.15625,
      "learning_rate": 2.7263897662665825e-05,
      "loss": 0.3919,
      "mean_token_accuracy": 0.9181667804718018,
      "step": 1155,
      "train/kl_loss_qwen": 0.04454113356769085,
      "train/kl_loss_r1": 0.0961604356765747,
      "train/total_kl": 0.14070156663656236
    },
    {
      "epoch": 0.3663929248262792,
      "grad_norm": 5.53125,
      "learning_rate": 2.7252053063802907e-05,
      "loss": 0.4373,
      "mean_token_accuracy": 0.9129749059677124,
      "step": 1160,
      "train/kl_loss_qwen": 0.0388947818428278,
      "train/kl_loss_r1": 0.08083640336990357,
      "train/total_kl": 0.11973118484020233
    },
    {
      "epoch": 0.36797220467466835,
      "grad_norm": 3.21875,
      "learning_rate": 2.724020846493999e-05,
      "loss": 0.3833,
      "mean_token_accuracy": 0.9158246040344238,
      "step": 1165,
      "train/kl_loss_qwen": 0.039357122406363486,
      "train/kl_loss_r1": 0.09643976911902427,
      "train/total_kl": 0.13579689115285873
    },
    {
      "epoch": 0.36955148452305747,
      "grad_norm": 3.75,
      "learning_rate": 2.7228363866077067e-05,
      "loss": 0.3753,
      "mean_token_accuracy": 0.9185945749282837,
      "step": 1170,
      "train/kl_loss_qwen": 0.033802905678749086,
      "train/kl_loss_r1": 0.08716130703687668,
      "train/total_kl": 0.12096421271562577
    },
    {
      "epoch": 0.3711307643714466,
      "grad_norm": 4.375,
      "learning_rate": 2.7216519267214153e-05,
      "loss": 0.4077,
      "mean_token_accuracy": 0.9245885968208313,
      "step": 1175,
      "train/kl_loss_qwen": 0.039360678568482396,
      "train/kl_loss_r1": 0.116049362719059,
      "train/total_kl": 0.15541004091501237
    },
    {
      "epoch": 0.3727100442198358,
      "grad_norm": 3.546875,
      "learning_rate": 2.720467466835123e-05,
      "loss": 0.3904,
      "mean_token_accuracy": 0.914016854763031,
      "step": 1180,
      "train/kl_loss_qwen": 0.03059450201690197,
      "train/kl_loss_r1": 0.0753175251185894,
      "train/total_kl": 0.10591202974319458
    },
    {
      "epoch": 0.3742893240682249,
      "grad_norm": 4.09375,
      "learning_rate": 2.7192830069488313e-05,
      "loss": 0.3786,
      "mean_token_accuracy": 0.9165893793106079,
      "step": 1185,
      "train/kl_loss_qwen": 0.034191733598709105,
      "train/kl_loss_r1": 0.07712326645851135,
      "train/total_kl": 0.11131500005722046
    },
    {
      "epoch": 0.37586860391661403,
      "grad_norm": 3.90625,
      "learning_rate": 2.7180985470625395e-05,
      "loss": 0.3747,
      "mean_token_accuracy": 0.9169653058052063,
      "step": 1190,
      "train/kl_loss_qwen": 0.03833934217691422,
      "train/kl_loss_r1": 0.08487532809376716,
      "train/total_kl": 0.12321467250585556
    },
    {
      "epoch": 0.37744788376500316,
      "grad_norm": 3.890625,
      "learning_rate": 2.7169140871762477e-05,
      "loss": 0.3683,
      "mean_token_accuracy": 0.9192065000534058,
      "step": 1195,
      "train/kl_loss_qwen": 0.03011544533073902,
      "train/kl_loss_r1": 0.08271843791007996,
      "train/total_kl": 0.11283388584852219
    },
    {
      "epoch": 0.3790271636133923,
      "grad_norm": 3.734375,
      "learning_rate": 2.715729627289956e-05,
      "loss": 0.3648,
      "mean_token_accuracy": 0.921120822429657,
      "step": 1200,
      "train/kl_loss_qwen": 0.02874102145433426,
      "train/kl_loss_r1": 0.08218741416931152,
      "train/total_kl": 0.11092843562364578
    },
    {
      "epoch": 0.3806064434617814,
      "grad_norm": 2.765625,
      "learning_rate": 2.714545167403664e-05,
      "loss": 0.3493,
      "mean_token_accuracy": 0.9228572130203248,
      "step": 1205,
      "train/kl_loss_qwen": 0.028672541305422784,
      "train/kl_loss_r1": 0.08233051896095275,
      "train/total_kl": 0.11100306063890457
    },
    {
      "epoch": 0.3821857233101706,
      "grad_norm": 4.5625,
      "learning_rate": 2.713360707517372e-05,
      "loss": 0.3738,
      "mean_token_accuracy": 0.922773003578186,
      "step": 1210,
      "train/kl_loss_qwen": 0.035533550381660464,
      "train/kl_loss_r1": 0.09579968005418778,
      "train/total_kl": 0.13133323043584824
    },
    {
      "epoch": 0.3837650031585597,
      "grad_norm": 3.796875,
      "learning_rate": 2.7121762476310805e-05,
      "loss": 0.3884,
      "mean_token_accuracy": 0.9225324869155884,
      "step": 1215,
      "train/kl_loss_qwen": 0.04200977794826031,
      "train/kl_loss_r1": 0.10754130780696869,
      "train/total_kl": 0.14955108612775803
    },
    {
      "epoch": 0.38534428300694884,
      "grad_norm": 3.359375,
      "learning_rate": 2.7109917877447884e-05,
      "loss": 0.567,
      "mean_token_accuracy": 0.9059720754623413,
      "step": 1220,
      "train/kl_loss_qwen": 0.08452499322593213,
      "train/kl_loss_r1": 0.20899230241775513,
      "train/total_kl": 0.2935172960162163
    },
    {
      "epoch": 0.38692356285533797,
      "grad_norm": 3.0,
      "learning_rate": 2.7098073278584966e-05,
      "loss": 0.3548,
      "mean_token_accuracy": 0.9265740990638733,
      "step": 1225,
      "train/kl_loss_qwen": 0.03697242625057697,
      "train/kl_loss_r1": 0.08387215435504913,
      "train/total_kl": 0.12084458172321319
    },
    {
      "epoch": 0.3885028427037271,
      "grad_norm": 27.625,
      "learning_rate": 2.7086228679722048e-05,
      "loss": 0.7309,
      "mean_token_accuracy": 0.8941235780715943,
      "step": 1230,
      "train/kl_loss_qwen": 0.11973113864660263,
      "train/kl_loss_r1": 0.2907026164233685,
      "train/total_kl": 0.4104337677359581
    },
    {
      "epoch": 0.3900821225521162,
      "grad_norm": 4.1875,
      "learning_rate": 2.707438408085913e-05,
      "loss": 0.4175,
      "mean_token_accuracy": 0.9171712756156921,
      "step": 1235,
      "train/kl_loss_qwen": 0.04249858893454075,
      "train/kl_loss_r1": 0.10305260047316551,
      "train/total_kl": 0.145551186054945
    },
    {
      "epoch": 0.39166140240050534,
      "grad_norm": 4.40625,
      "learning_rate": 2.706253948199621e-05,
      "loss": 0.688,
      "mean_token_accuracy": 0.887078607082367,
      "step": 1240,
      "train/kl_loss_qwen": 0.1381281677633524,
      "train/kl_loss_r1": 0.21289537400007247,
      "train/total_kl": 0.35102354139089587
    },
    {
      "epoch": 0.3932406822488945,
      "grad_norm": 3.21875,
      "learning_rate": 2.7050694883133294e-05,
      "loss": 0.3388,
      "mean_token_accuracy": 0.9255303263664245,
      "step": 1245,
      "train/kl_loss_qwen": 0.027776998281478883,
      "train/kl_loss_r1": 0.06999287828803062,
      "train/total_kl": 0.09776987433433533
    },
    {
      "epoch": 0.39481996209728365,
      "grad_norm": 2.6875,
      "learning_rate": 2.7038850284270372e-05,
      "loss": 0.3749,
      "mean_token_accuracy": 0.9156741499900818,
      "step": 1250,
      "train/kl_loss_qwen": 0.037703001871705055,
      "train/kl_loss_r1": 0.08796408846974373,
      "train/total_kl": 0.1256670892238617
    },
    {
      "epoch": 0.3963992419456728,
      "grad_norm": 4.375,
      "learning_rate": 2.7027005685407458e-05,
      "loss": 0.3963,
      "mean_token_accuracy": 0.9046347379684448,
      "step": 1255,
      "train/kl_loss_qwen": 0.03468194045126438,
      "train/kl_loss_r1": 0.06877386793494225,
      "train/total_kl": 0.10345581024885178
    },
    {
      "epoch": 0.3979785217940619,
      "grad_norm": 4.65625,
      "learning_rate": 2.7015161086544536e-05,
      "loss": 0.4608,
      "mean_token_accuracy": 0.9224648237228393,
      "step": 1260,
      "train/kl_loss_qwen": 0.07581054866313934,
      "train/kl_loss_r1": 0.14018777906894683,
      "train/total_kl": 0.21599833071231841
    },
    {
      "epoch": 0.399557801642451,
      "grad_norm": 2.96875,
      "learning_rate": 2.7003316487681615e-05,
      "loss": 0.7225,
      "mean_token_accuracy": 0.8964391827583313,
      "step": 1265,
      "train/kl_loss_qwen": 0.12634216248989105,
      "train/kl_loss_r1": 0.2728008933365345,
      "train/total_kl": 0.39914305657148363
    },
    {
      "epoch": 0.40113708149084015,
      "grad_norm": 4.8125,
      "learning_rate": 2.69914718888187e-05,
      "loss": 0.3844,
      "mean_token_accuracy": 0.9178162336349487,
      "step": 1270,
      "train/kl_loss_qwen": 0.03329511545598507,
      "train/kl_loss_r1": 0.08622499480843544,
      "train/total_kl": 0.11952010840177536
    },
    {
      "epoch": 0.40271636133922933,
      "grad_norm": 5.25,
      "learning_rate": 2.697962728995578e-05,
      "loss": 0.3494,
      "mean_token_accuracy": 0.9268160820007324,
      "step": 1275,
      "train/kl_loss_qwen": 0.03734969459474087,
      "train/kl_loss_r1": 0.08838790357112884,
      "train/total_kl": 0.12573759853839875
    },
    {
      "epoch": 0.40429564118761846,
      "grad_norm": 2.375,
      "learning_rate": 2.6967782691092864e-05,
      "loss": 0.388,
      "mean_token_accuracy": 0.9205210566520691,
      "step": 1280,
      "train/kl_loss_qwen": 0.04101281091570854,
      "train/kl_loss_r1": 0.0930364578962326,
      "train/total_kl": 0.1340492695569992
    },
    {
      "epoch": 0.4058749210360076,
      "grad_norm": 3.921875,
      "learning_rate": 2.6955938092229943e-05,
      "loss": 0.3661,
      "mean_token_accuracy": 0.9261689424514771,
      "step": 1285,
      "train/kl_loss_qwen": 0.03783885836601257,
      "train/kl_loss_r1": 0.08813183456659317,
      "train/total_kl": 0.12597069442272185
    },
    {
      "epoch": 0.4074542008843967,
      "grad_norm": 3.46875,
      "learning_rate": 2.6944093493367025e-05,
      "loss": 0.327,
      "mean_token_accuracy": 0.9344208359718322,
      "step": 1290,
      "train/kl_loss_qwen": 0.029625180177390575,
      "train/kl_loss_r1": 0.07623563334345818,
      "train/total_kl": 0.10586080998182297
    },
    {
      "epoch": 0.40903348073278584,
      "grad_norm": 4.3125,
      "learning_rate": 2.6932248894504107e-05,
      "loss": 0.3759,
      "mean_token_accuracy": 0.9238349199295044,
      "step": 1295,
      "train/kl_loss_qwen": 0.0343143068253994,
      "train/kl_loss_r1": 0.10071116834878921,
      "train/total_kl": 0.13502547442913054
    },
    {
      "epoch": 0.41061276058117496,
      "grad_norm": 5.625,
      "learning_rate": 2.692040429564119e-05,
      "loss": 0.3576,
      "mean_token_accuracy": 0.9329147458076477,
      "step": 1300,
      "train/kl_loss_qwen": 0.0390559658408165,
      "train/kl_loss_r1": 0.09704398363828659,
      "train/total_kl": 0.1360999494791031
    },
    {
      "epoch": 0.41219204042956414,
      "grad_norm": 3.40625,
      "learning_rate": 2.6908559696778267e-05,
      "loss": 0.4384,
      "mean_token_accuracy": 0.8997732400894165,
      "step": 1305,
      "train/kl_loss_qwen": 0.03465775586664677,
      "train/kl_loss_r1": 0.07938610911369323,
      "train/total_kl": 0.11404386609792709
    },
    {
      "epoch": 0.41377132027795327,
      "grad_norm": 3.625,
      "learning_rate": 2.6896715097915352e-05,
      "loss": 0.3448,
      "mean_token_accuracy": 0.9294810771942139,
      "step": 1310,
      "train/kl_loss_qwen": 0.03695046603679657,
      "train/kl_loss_r1": 0.07044115588068962,
      "train/total_kl": 0.10739162042737008
    },
    {
      "epoch": 0.4153506001263424,
      "grad_norm": 2.65625,
      "learning_rate": 2.688487049905243e-05,
      "loss": 0.3261,
      "mean_token_accuracy": 0.9330098032951355,
      "step": 1315,
      "train/kl_loss_qwen": 0.026439381390810014,
      "train/kl_loss_r1": 0.07886992990970612,
      "train/total_kl": 0.10530931204557419
    },
    {
      "epoch": 0.4169298799747315,
      "grad_norm": 2.859375,
      "learning_rate": 2.6873025900189516e-05,
      "loss": 0.3862,
      "mean_token_accuracy": 0.9152974247932434,
      "step": 1320,
      "train/kl_loss_qwen": 0.02970890998840332,
      "train/kl_loss_r1": 0.0773049920797348,
      "train/total_kl": 0.10701390206813813
    },
    {
      "epoch": 0.41850915982312065,
      "grad_norm": 2.84375,
      "learning_rate": 2.6861181301326595e-05,
      "loss": 0.3384,
      "mean_token_accuracy": 0.9270530462265014,
      "step": 1325,
      "train/kl_loss_qwen": 0.03480346277356148,
      "train/kl_loss_r1": 0.07801690474152564,
      "train/total_kl": 0.11282036751508713
    },
    {
      "epoch": 0.4200884396715098,
      "grad_norm": 3.609375,
      "learning_rate": 2.6849336702463677e-05,
      "loss": 0.3799,
      "mean_token_accuracy": 0.9131087183952331,
      "step": 1330,
      "train/kl_loss_qwen": 0.032946069911122325,
      "train/kl_loss_r1": 0.08465765416622162,
      "train/total_kl": 0.11760372370481491
    },
    {
      "epoch": 0.42166771951989895,
      "grad_norm": 8.9375,
      "learning_rate": 2.683749210360076e-05,
      "loss": 0.4693,
      "mean_token_accuracy": 0.9012147307395935,
      "step": 1335,
      "train/kl_loss_qwen": 0.04172387607395649,
      "train/kl_loss_r1": 0.09869510754942894,
      "train/total_kl": 0.14041898250579835
    },
    {
      "epoch": 0.4232469993682881,
      "grad_norm": 3.59375,
      "learning_rate": 2.682564750473784e-05,
      "loss": 0.3644,
      "mean_token_accuracy": 0.9172945737838745,
      "step": 1340,
      "train/kl_loss_qwen": 0.023991287499666215,
      "train/kl_loss_r1": 0.07367524653673171,
      "train/total_kl": 0.097666534781456
    },
    {
      "epoch": 0.4248262792166772,
      "grad_norm": 7.625,
      "learning_rate": 2.6813802905874923e-05,
      "loss": 0.4205,
      "mean_token_accuracy": 0.9209869146347046,
      "step": 1345,
      "train/kl_loss_qwen": 0.06653879433870316,
      "train/kl_loss_r1": 0.11447124928236008,
      "train/total_kl": 0.18101004362106324
    },
    {
      "epoch": 0.42640555906506633,
      "grad_norm": 5.78125,
      "learning_rate": 2.6801958307012005e-05,
      "loss": 0.42,
      "mean_token_accuracy": 0.9204567551612854,
      "step": 1350,
      "train/kl_loss_qwen": 0.051871858164668085,
      "train/kl_loss_r1": 0.10009860694408416,
      "train/total_kl": 0.15197046399116515
    },
    {
      "epoch": 0.42798483891345546,
      "grad_norm": 2.5,
      "learning_rate": 2.6790113708149083e-05,
      "loss": 0.4712,
      "mean_token_accuracy": 0.9080936431884765,
      "step": 1355,
      "train/kl_loss_qwen": 0.04953858591616154,
      "train/kl_loss_r1": 0.0985082022845745,
      "train/total_kl": 0.14804679304361343
    },
    {
      "epoch": 0.4295641187618446,
      "grad_norm": 2.09375,
      "learning_rate": 2.677826910928617e-05,
      "loss": 0.3718,
      "mean_token_accuracy": 0.9243037343025208,
      "step": 1360,
      "train/kl_loss_qwen": 0.03089449517428875,
      "train/kl_loss_r1": 0.08510574698448181,
      "train/total_kl": 0.11600024551153183
    },
    {
      "epoch": 0.4311433986102337,
      "grad_norm": 2.640625,
      "learning_rate": 2.6766424510423247e-05,
      "loss": 0.3824,
      "mean_token_accuracy": 0.9170409560203552,
      "step": 1365,
      "train/kl_loss_qwen": 0.043358999863266946,
      "train/kl_loss_r1": 0.08913471028208733,
      "train/total_kl": 0.1324937105178833
    },
    {
      "epoch": 0.4327226784586229,
      "grad_norm": 3.390625,
      "learning_rate": 2.675457991156033e-05,
      "loss": 0.3603,
      "mean_token_accuracy": 0.9120386600494385,
      "step": 1370,
      "train/kl_loss_qwen": 0.027146858721971513,
      "train/kl_loss_r1": 0.0700517863035202,
      "train/total_kl": 0.09719864726066589
    },
    {
      "epoch": 0.434301958307012,
      "grad_norm": 5.40625,
      "learning_rate": 2.674273531269741e-05,
      "loss": 0.6147,
      "mean_token_accuracy": 0.9000384330749511,
      "step": 1375,
      "train/kl_loss_qwen": 0.10420886240899563,
      "train/kl_loss_r1": 0.19849361553788186,
      "train/total_kl": 0.3027024671435356
    },
    {
      "epoch": 0.43588123815540114,
      "grad_norm": 3.203125,
      "learning_rate": 2.6730890713834493e-05,
      "loss": 0.3956,
      "mean_token_accuracy": 0.9125915050506592,
      "step": 1380,
      "train/kl_loss_qwen": 0.032400262728333476,
      "train/kl_loss_r1": 0.08311344683170319,
      "train/total_kl": 0.11551371216773987
    },
    {
      "epoch": 0.43746051800379027,
      "grad_norm": 3.546875,
      "learning_rate": 2.6719046114971575e-05,
      "loss": 0.4215,
      "mean_token_accuracy": 0.9202767968177795,
      "step": 1385,
      "train/kl_loss_qwen": 0.042635428160429,
      "train/kl_loss_r1": 0.10554430931806565,
      "train/total_kl": 0.14817974120378494
    },
    {
      "epoch": 0.4390397978521794,
      "grad_norm": 4.0625,
      "learning_rate": 2.6707201516108654e-05,
      "loss": 0.3711,
      "mean_token_accuracy": 0.9193390846252442,
      "step": 1390,
      "train/kl_loss_qwen": 0.02584208771586418,
      "train/kl_loss_r1": 0.07836293280124665,
      "train/total_kl": 0.10420501977205276
    },
    {
      "epoch": 0.4406190777005685,
      "grad_norm": 3.796875,
      "learning_rate": 2.6695356917245736e-05,
      "loss": 0.3953,
      "mean_token_accuracy": 0.9181339263916015,
      "step": 1395,
      "train/kl_loss_qwen": 0.05590136423707008,
      "train/kl_loss_r1": 0.09515961408615112,
      "train/total_kl": 0.15106097906827926
    },
    {
      "epoch": 0.4421983575489577,
      "grad_norm": 5.78125,
      "learning_rate": 2.6683512318382818e-05,
      "loss": 0.4218,
      "mean_token_accuracy": 0.9101435780525208,
      "step": 1400,
      "train/kl_loss_qwen": 0.037794597819447516,
      "train/kl_loss_r1": 0.07928776144981384,
      "train/total_kl": 0.1170823596417904
    },
    {
      "epoch": 0.4437776373973468,
      "grad_norm": 3.515625,
      "learning_rate": 2.66716677195199e-05,
      "loss": 0.3593,
      "mean_token_accuracy": 0.9189205288887023,
      "step": 1405,
      "train/kl_loss_qwen": 0.029898228123784065,
      "train/kl_loss_r1": 0.07590225040912628,
      "train/total_kl": 0.1058004766702652
    },
    {
      "epoch": 0.44535691724573595,
      "grad_norm": 3.078125,
      "learning_rate": 2.665982312065698e-05,
      "loss": 0.3422,
      "mean_token_accuracy": 0.9274580121040344,
      "step": 1410,
      "train/kl_loss_qwen": 0.03190844841301441,
      "train/kl_loss_r1": 0.0829135850071907,
      "train/total_kl": 0.1148220345377922
    },
    {
      "epoch": 0.4469361970941251,
      "grad_norm": 5.125,
      "learning_rate": 2.6647978521794064e-05,
      "loss": 0.3528,
      "mean_token_accuracy": 0.9270729541778564,
      "step": 1415,
      "train/kl_loss_qwen": 0.03251653518527746,
      "train/kl_loss_r1": 0.07854450643062591,
      "train/total_kl": 0.11106103882193566
    },
    {
      "epoch": 0.4485154769425142,
      "grad_norm": 3.078125,
      "learning_rate": 2.6636133922931142e-05,
      "loss": 0.3354,
      "mean_token_accuracy": 0.922597074508667,
      "step": 1420,
      "train/kl_loss_qwen": 0.024554036930203436,
      "train/kl_loss_r1": 0.06484766975045204,
      "train/total_kl": 0.08940170705318451
    },
    {
      "epoch": 0.4500947567909033,
      "grad_norm": 3.671875,
      "learning_rate": 2.6624289324068228e-05,
      "loss": 0.9108,
      "mean_token_accuracy": 0.8719356179237365,
      "step": 1425,
      "train/kl_loss_qwen": 0.15624983310699464,
      "train/kl_loss_r1": 0.3716003179550171,
      "train/total_kl": 0.5278501614928246
    },
    {
      "epoch": 0.4516740366392925,
      "grad_norm": 3.234375,
      "learning_rate": 2.6612444725205306e-05,
      "loss": 0.3898,
      "mean_token_accuracy": 0.9197250843048096,
      "step": 1430,
      "train/kl_loss_qwen": 0.04201133996248245,
      "train/kl_loss_r1": 0.09729172885417939,
      "train/total_kl": 0.13930307179689408
    },
    {
      "epoch": 0.45325331648768163,
      "grad_norm": 4.25,
      "learning_rate": 2.6600600126342388e-05,
      "loss": 0.3501,
      "mean_token_accuracy": 0.9273348569869995,
      "step": 1435,
      "train/kl_loss_qwen": 0.037208805233240126,
      "train/kl_loss_r1": 0.08745781183242798,
      "train/total_kl": 0.12466661483049393
    },
    {
      "epoch": 0.45483259633607076,
      "grad_norm": 3.8125,
      "learning_rate": 2.658875552747947e-05,
      "loss": 0.3844,
      "mean_token_accuracy": 0.9211597681045532,
      "step": 1440,
      "train/kl_loss_qwen": 0.04053596295416355,
      "train/kl_loss_r1": 0.09073771387338639,
      "train/total_kl": 0.13127367794513703
    },
    {
      "epoch": 0.4564118761844599,
      "grad_norm": 4.53125,
      "learning_rate": 2.6576910928616552e-05,
      "loss": 0.437,
      "mean_token_accuracy": 0.904490876197815,
      "step": 1445,
      "train/kl_loss_qwen": 0.039917736500501635,
      "train/kl_loss_r1": 0.09061303287744522,
      "train/total_kl": 0.1305307686328888
    },
    {
      "epoch": 0.457991156032849,
      "grad_norm": 7.84375,
      "learning_rate": 2.656506632975363e-05,
      "loss": 0.4242,
      "mean_token_accuracy": 0.9134282231330871,
      "step": 1450,
      "train/kl_loss_qwen": 0.059297601506114,
      "train/kl_loss_r1": 0.11168717667460441,
      "train/total_kl": 0.17098478153347968
    },
    {
      "epoch": 0.45957043588123814,
      "grad_norm": 2.96875,
      "learning_rate": 2.6553221730890716e-05,
      "loss": 0.3963,
      "mean_token_accuracy": 0.9120491147041321,
      "step": 1455,
      "train/kl_loss_qwen": 0.03350956458598375,
      "train/kl_loss_r1": 0.08437399789690972,
      "train/total_kl": 0.1178835615515709
    },
    {
      "epoch": 0.46114971572962726,
      "grad_norm": 4.71875,
      "learning_rate": 2.6541377132027795e-05,
      "loss": 0.3851,
      "mean_token_accuracy": 0.9144776105880738,
      "step": 1460,
      "train/kl_loss_qwen": 0.03241768442094326,
      "train/kl_loss_r1": 0.07328464537858963,
      "train/total_kl": 0.10570233017206192
    },
    {
      "epoch": 0.46272899557801644,
      "grad_norm": 3.140625,
      "learning_rate": 2.652953253316488e-05,
      "loss": 0.4046,
      "mean_token_accuracy": 0.9037475705146789,
      "step": 1465,
      "train/kl_loss_qwen": 0.03153598364442587,
      "train/kl_loss_r1": 0.08471559137105941,
      "train/total_kl": 0.11625157743692398
    },
    {
      "epoch": 0.46430827542640557,
      "grad_norm": 5.625,
      "learning_rate": 2.651768793430196e-05,
      "loss": 0.3952,
      "mean_token_accuracy": 0.9182330369949341,
      "step": 1470,
      "train/kl_loss_qwen": 0.04389566034078598,
      "train/kl_loss_r1": 0.09206905364990234,
      "train/total_kl": 0.1359647125005722
    },
    {
      "epoch": 0.4658875552747947,
      "grad_norm": 3.171875,
      "learning_rate": 2.650584333543904e-05,
      "loss": 0.392,
      "mean_token_accuracy": 0.9187814831733704,
      "step": 1475,
      "train/kl_loss_qwen": 0.04668047390878201,
      "train/kl_loss_r1": 0.09733642935752869,
      "train/total_kl": 0.14401690065860748
    },
    {
      "epoch": 0.4674668351231838,
      "grad_norm": 2.90625,
      "learning_rate": 2.6493998736576123e-05,
      "loss": 0.4156,
      "mean_token_accuracy": 0.9075844049453735,
      "step": 1480,
      "train/kl_loss_qwen": 0.0375371515750885,
      "train/kl_loss_r1": 0.09291562288999558,
      "train/total_kl": 0.13045277297496796
    },
    {
      "epoch": 0.46904611497157295,
      "grad_norm": 4.03125,
      "learning_rate": 2.6482154137713205e-05,
      "loss": 0.3928,
      "mean_token_accuracy": 0.9191757678985596,
      "step": 1485,
      "train/kl_loss_qwen": 0.041358457505702974,
      "train/kl_loss_r1": 0.10835730582475663,
      "train/total_kl": 0.1497157633304596
    },
    {
      "epoch": 0.4706253948199621,
      "grad_norm": 2.96875,
      "learning_rate": 2.6470309538850283e-05,
      "loss": 0.3753,
      "mean_token_accuracy": 0.9185408234596253,
      "step": 1490,
      "train/kl_loss_qwen": 0.037813367694616316,
      "train/kl_loss_r1": 0.09808466136455536,
      "train/total_kl": 0.13589802980422974
    },
    {
      "epoch": 0.47220467466835125,
      "grad_norm": 3.65625,
      "learning_rate": 2.645846493998737e-05,
      "loss": 0.3684,
      "mean_token_accuracy": 0.9199748516082764,
      "step": 1495,
      "train/kl_loss_qwen": 0.03153411261737347,
      "train/kl_loss_r1": 0.07936602905392647,
      "train/total_kl": 0.11090013980865479
    },
    {
      "epoch": 0.4737839545167404,
      "grad_norm": 3.46875,
      "learning_rate": 2.6446620341124447e-05,
      "loss": 0.412,
      "mean_token_accuracy": 0.9242963552474975,
      "step": 1500,
      "train/kl_loss_qwen": 0.05249704606831074,
      "train/kl_loss_r1": 0.12705418169498445,
      "train/total_kl": 0.17955122590065004
    },
    {
      "epoch": 0.4753632343651295,
      "grad_norm": 17.0,
      "learning_rate": 2.6434775742261532e-05,
      "loss": 0.604,
      "mean_token_accuracy": 0.9073544025421143,
      "step": 1505,
      "train/kl_loss_qwen": 0.10196574702858925,
      "train/kl_loss_r1": 0.2354960858821869,
      "train/total_kl": 0.33746182918548584
    },
    {
      "epoch": 0.47694251421351863,
      "grad_norm": 3.546875,
      "learning_rate": 2.642293114339861e-05,
      "loss": 0.5747,
      "mean_token_accuracy": 0.8938451409339905,
      "step": 1510,
      "train/kl_loss_qwen": 0.09353167749941349,
      "train/kl_loss_r1": 0.169492270052433,
      "train/total_kl": 0.2630239427089691
    },
    {
      "epoch": 0.47852179406190776,
      "grad_norm": 3.421875,
      "learning_rate": 2.641108654453569e-05,
      "loss": 0.3283,
      "mean_token_accuracy": 0.9279634475708007,
      "step": 1515,
      "train/kl_loss_qwen": 0.03046959452331066,
      "train/kl_loss_r1": 0.07763081416487694,
      "train/total_kl": 0.10810040831565856
    },
    {
      "epoch": 0.4801010739102969,
      "grad_norm": 4.78125,
      "learning_rate": 2.6399241945672775e-05,
      "loss": 0.4028,
      "mean_token_accuracy": 0.9024159669876098,
      "step": 1520,
      "train/kl_loss_qwen": 0.02694017067551613,
      "train/kl_loss_r1": 0.07744768932461739,
      "train/total_kl": 0.10438786149024963
    },
    {
      "epoch": 0.48168035375868606,
      "grad_norm": 3.921875,
      "learning_rate": 2.6387397346809854e-05,
      "loss": 0.4096,
      "mean_token_accuracy": 0.92669677734375,
      "step": 1525,
      "train/kl_loss_qwen": 0.05647825077176094,
      "train/kl_loss_r1": 0.11235563158988952,
      "train/total_kl": 0.1688338816165924
    },
    {
      "epoch": 0.4832596336070752,
      "grad_norm": 10.75,
      "learning_rate": 2.6375552747946936e-05,
      "loss": 0.435,
      "mean_token_accuracy": 0.9233072400093079,
      "step": 1530,
      "train/kl_loss_qwen": 0.05204330384731293,
      "train/kl_loss_r1": 0.13819514811038972,
      "train/total_kl": 0.1902384489774704
    },
    {
      "epoch": 0.4848389134554643,
      "grad_norm": 5.5625,
      "learning_rate": 2.6363708149084018e-05,
      "loss": 0.4985,
      "mean_token_accuracy": 0.8852681756019593,
      "step": 1535,
      "train/kl_loss_qwen": 0.02800062149763107,
      "train/kl_loss_r1": 0.08963649645447731,
      "train/total_kl": 0.11763711720705032
    },
    {
      "epoch": 0.48641819330385344,
      "grad_norm": 4.9375,
      "learning_rate": 2.63518635502211e-05,
      "loss": 0.3968,
      "mean_token_accuracy": 0.9217487812042237,
      "step": 1540,
      "train/kl_loss_qwen": 0.05009663552045822,
      "train/kl_loss_r1": 0.10394758433103561,
      "train/total_kl": 0.15404422134160994
    },
    {
      "epoch": 0.48799747315224257,
      "grad_norm": 3.546875,
      "learning_rate": 2.634001895135818e-05,
      "loss": 0.3696,
      "mean_token_accuracy": 0.919561505317688,
      "step": 1545,
      "train/kl_loss_qwen": 0.02984781190752983,
      "train/kl_loss_r1": 0.07651440799236298,
      "train/total_kl": 0.10636222064495086
    },
    {
      "epoch": 0.4895767530006317,
      "grad_norm": 3.5,
      "learning_rate": 2.6328174352495263e-05,
      "loss": 0.3688,
      "mean_token_accuracy": 0.9252573370933532,
      "step": 1550,
      "train/kl_loss_qwen": 0.03929025419056416,
      "train/kl_loss_r1": 0.07892496287822723,
      "train/total_kl": 0.11821521669626237
    },
    {
      "epoch": 0.4911560328490209,
      "grad_norm": 5.59375,
      "learning_rate": 2.6316329753632342e-05,
      "loss": 0.3978,
      "mean_token_accuracy": 0.9241521954536438,
      "step": 1555,
      "train/kl_loss_qwen": 0.03982924781739712,
      "train/kl_loss_r1": 0.10502630621194839,
      "train/total_kl": 0.1448555514216423
    },
    {
      "epoch": 0.49273531269741,
      "grad_norm": 4.0625,
      "learning_rate": 2.6304485154769427e-05,
      "loss": 0.387,
      "mean_token_accuracy": 0.9264254808425904,
      "step": 1560,
      "train/kl_loss_qwen": 0.03859283607453108,
      "train/kl_loss_r1": 0.10121604055166245,
      "train/total_kl": 0.13980887532234193
    },
    {
      "epoch": 0.4943145925457991,
      "grad_norm": 9.3125,
      "learning_rate": 2.6292640555906506e-05,
      "loss": 0.4941,
      "mean_token_accuracy": 0.9167832970619202,
      "step": 1565,
      "train/kl_loss_qwen": 0.08327364921569824,
      "train/kl_loss_r1": 0.1620011180639267,
      "train/total_kl": 0.2452747642993927
    },
    {
      "epoch": 0.49589387239418825,
      "grad_norm": 3.921875,
      "learning_rate": 2.6280795957043588e-05,
      "loss": 0.3955,
      "mean_token_accuracy": 0.9118420720100403,
      "step": 1570,
      "train/kl_loss_qwen": 0.028081656992435457,
      "train/kl_loss_r1": 0.0909204676747322,
      "train/total_kl": 0.11900212615728378
    },
    {
      "epoch": 0.4974731522425774,
      "grad_norm": 3.890625,
      "learning_rate": 2.626895135818067e-05,
      "loss": 0.3898,
      "mean_token_accuracy": 0.9165359973907471,
      "step": 1575,
      "train/kl_loss_qwen": 0.03890275843441486,
      "train/kl_loss_r1": 0.09130221754312515,
      "train/total_kl": 0.130204975605011
    },
    {
      "epoch": 0.4990524320909665,
      "grad_norm": 6.84375,
      "learning_rate": 2.6257106759317752e-05,
      "loss": 0.3775,
      "mean_token_accuracy": 0.9173738121986389,
      "step": 1580,
      "train/kl_loss_qwen": 0.04134678840637207,
      "train/kl_loss_r1": 0.10178495571017265,
      "train/total_kl": 0.14313174784183502
    },
    {
      "epoch": 0.5006317119393556,
      "grad_norm": 3.703125,
      "learning_rate": 2.6245262160454834e-05,
      "loss": 0.3594,
      "mean_token_accuracy": 0.9282228946685791,
      "step": 1585,
      "train/kl_loss_qwen": 0.04827427752315998,
      "train/kl_loss_r1": 0.09085883349180221,
      "train/total_kl": 0.13913311064243317
    },
    {
      "epoch": 0.5022109917877448,
      "grad_norm": 3.734375,
      "learning_rate": 2.6233417561591916e-05,
      "loss": 0.648,
      "mean_token_accuracy": 0.8974647521972656,
      "step": 1590,
      "train/kl_loss_qwen": 0.11151764020323754,
      "train/kl_loss_r1": 0.2079153075814247,
      "train/total_kl": 0.3194329351186752
    },
    {
      "epoch": 0.5037902716361339,
      "grad_norm": 5.28125,
      "learning_rate": 2.6221572962728994e-05,
      "loss": 0.4011,
      "mean_token_accuracy": 0.911795699596405,
      "step": 1595,
      "train/kl_loss_qwen": 0.032901936769485475,
      "train/kl_loss_r1": 0.08797814249992371,
      "train/total_kl": 0.12088007926940918
    },
    {
      "epoch": 0.505369551484523,
      "grad_norm": 4.15625,
      "learning_rate": 2.620972836386608e-05,
      "loss": 0.3965,
      "mean_token_accuracy": 0.9190373897552491,
      "step": 1600,
      "train/kl_loss_qwen": 0.04687013104557991,
      "train/kl_loss_r1": 0.0847273200750351,
      "train/total_kl": 0.13159745186567307
    },
    {
      "epoch": 0.5069488313329122,
      "grad_norm": 6.4375,
      "learning_rate": 2.619788376500316e-05,
      "loss": 0.4367,
      "mean_token_accuracy": 0.909017550945282,
      "step": 1605,
      "train/kl_loss_qwen": 0.04429644830524922,
      "train/kl_loss_r1": 0.10758630484342575,
      "train/total_kl": 0.15188275128602982
    },
    {
      "epoch": 0.5085281111813014,
      "grad_norm": 3.9375,
      "learning_rate": 2.6186039166140244e-05,
      "loss": 0.3468,
      "mean_token_accuracy": 0.9260318636894226,
      "step": 1610,
      "train/kl_loss_qwen": 0.030309950932860374,
      "train/kl_loss_r1": 0.08060631230473518,
      "train/total_kl": 0.11091626286506653
    },
    {
      "epoch": 0.5101073910296905,
      "grad_norm": 3.953125,
      "learning_rate": 2.6174194567277322e-05,
      "loss": 0.4482,
      "mean_token_accuracy": 0.9174621224403381,
      "step": 1615,
      "train/kl_loss_qwen": 0.05694098323583603,
      "train/kl_loss_r1": 0.12819615602493287,
      "train/total_kl": 0.18513713777065277
    },
    {
      "epoch": 0.5116866708780796,
      "grad_norm": 3.203125,
      "learning_rate": 2.6162349968414404e-05,
      "loss": 0.4581,
      "mean_token_accuracy": 0.9108327150344848,
      "step": 1620,
      "train/kl_loss_qwen": 0.05476501025259495,
      "train/kl_loss_r1": 0.11412863880395889,
      "train/total_kl": 0.16889365017414093
    },
    {
      "epoch": 0.5132659507264687,
      "grad_norm": 4.21875,
      "learning_rate": 2.6150505369551486e-05,
      "loss": 0.4096,
      "mean_token_accuracy": 0.9197014570236206,
      "step": 1625,
      "train/kl_loss_qwen": 0.047441744804382326,
      "train/kl_loss_r1": 0.10399839729070663,
      "train/total_kl": 0.15144014209508896
    },
    {
      "epoch": 0.5148452305748579,
      "grad_norm": 2.859375,
      "learning_rate": 2.6138660770688568e-05,
      "loss": 0.4182,
      "mean_token_accuracy": 0.9116609930992127,
      "step": 1630,
      "train/kl_loss_qwen": 0.03003850281238556,
      "train/kl_loss_r1": 0.08903963938355446,
      "train/total_kl": 0.1190781444311142
    },
    {
      "epoch": 0.516424510423247,
      "grad_norm": 4.5625,
      "learning_rate": 2.6126816171825647e-05,
      "loss": 0.404,
      "mean_token_accuracy": 0.9100077390670777,
      "step": 1635,
      "train/kl_loss_qwen": 0.033823179081082344,
      "train/kl_loss_r1": 0.08149106130003929,
      "train/total_kl": 0.11531423628330231
    },
    {
      "epoch": 0.5180037902716361,
      "grad_norm": 5.25,
      "learning_rate": 2.611497157296273e-05,
      "loss": 0.4355,
      "mean_token_accuracy": 0.9219830870628357,
      "step": 1640,
      "train/kl_loss_qwen": 0.043203826248645785,
      "train/kl_loss_r1": 0.13423206955194472,
      "train/total_kl": 0.1774358943104744
    },
    {
      "epoch": 0.5195830701200252,
      "grad_norm": 4.15625,
      "learning_rate": 2.610312697409981e-05,
      "loss": 0.3773,
      "mean_token_accuracy": 0.9218290090560913,
      "step": 1645,
      "train/kl_loss_qwen": 0.03985932990908623,
      "train/kl_loss_r1": 0.09829081892967224,
      "train/total_kl": 0.1381501466035843
    },
    {
      "epoch": 0.5211623499684144,
      "grad_norm": 5.03125,
      "learning_rate": 2.6091282375236893e-05,
      "loss": 0.4589,
      "mean_token_accuracy": 0.9092371463775635,
      "step": 1650,
      "train/kl_loss_qwen": 0.03997619561851025,
      "train/kl_loss_r1": 0.10864420756697654,
      "train/total_kl": 0.14862040281295777
    },
    {
      "epoch": 0.5227416298168035,
      "grad_norm": 3.34375,
      "learning_rate": 2.6079437776373975e-05,
      "loss": 0.3785,
      "mean_token_accuracy": 0.918943464756012,
      "step": 1655,
      "train/kl_loss_qwen": 0.03358318917453289,
      "train/kl_loss_r1": 0.08849137127399445,
      "train/total_kl": 0.12207456082105636
    },
    {
      "epoch": 0.5243209096651926,
      "grad_norm": 3.71875,
      "learning_rate": 2.6067593177511053e-05,
      "loss": 0.389,
      "mean_token_accuracy": 0.9349912524223327,
      "step": 1660,
      "train/kl_loss_qwen": 0.058265311270952226,
      "train/kl_loss_r1": 0.12000070065259934,
      "train/total_kl": 0.17826600968837739
    },
    {
      "epoch": 0.5259001895135819,
      "grad_norm": 5.0625,
      "learning_rate": 2.605574857864814e-05,
      "loss": 0.3223,
      "mean_token_accuracy": 0.9280768990516662,
      "step": 1665,
      "train/kl_loss_qwen": 0.024833616241812705,
      "train/kl_loss_r1": 0.07256464660167694,
      "train/total_kl": 0.09739826321601867
    },
    {
      "epoch": 0.527479469361971,
      "grad_norm": 3.34375,
      "learning_rate": 2.6043903979785217e-05,
      "loss": 0.3448,
      "mean_token_accuracy": 0.9256616234779358,
      "step": 1670,
      "train/kl_loss_qwen": 0.027310534194111825,
      "train/kl_loss_r1": 0.07670907080173492,
      "train/total_kl": 0.10401960611343383
    },
    {
      "epoch": 0.5290587492103601,
      "grad_norm": 3.671875,
      "learning_rate": 2.60320593809223e-05,
      "loss": 0.3964,
      "mean_token_accuracy": 0.9237611532211304,
      "step": 1675,
      "train/kl_loss_qwen": 0.045483104698359964,
      "train/kl_loss_r1": 0.10022030845284462,
      "train/total_kl": 0.14570341557264327
    },
    {
      "epoch": 0.5306380290587492,
      "grad_norm": 6.5625,
      "learning_rate": 2.602021478205938e-05,
      "loss": 0.3719,
      "mean_token_accuracy": 0.922944700717926,
      "step": 1680,
      "train/kl_loss_qwen": 0.03418436944484711,
      "train/kl_loss_r1": 0.07859698906540871,
      "train/total_kl": 0.11278136074542999
    },
    {
      "epoch": 0.5322173089071384,
      "grad_norm": 4.84375,
      "learning_rate": 2.6008370183196463e-05,
      "loss": 0.4186,
      "mean_token_accuracy": 0.9102290987968444,
      "step": 1685,
      "train/kl_loss_qwen": 0.04204564280807972,
      "train/kl_loss_r1": 0.0995467871427536,
      "train/total_kl": 0.14159242957830429
    },
    {
      "epoch": 0.5337965887555275,
      "grad_norm": 4.125,
      "learning_rate": 2.5996525584333545e-05,
      "loss": 0.3618,
      "mean_token_accuracy": 0.9172422409057617,
      "step": 1690,
      "train/kl_loss_qwen": 0.031260650232434274,
      "train/kl_loss_r1": 0.08144056797027588,
      "train/total_kl": 0.11270121932029724
    },
    {
      "epoch": 0.5353758686039166,
      "grad_norm": 3.140625,
      "learning_rate": 2.5984680985470627e-05,
      "loss": 0.3611,
      "mean_token_accuracy": 0.9226415038108826,
      "step": 1695,
      "train/kl_loss_qwen": 0.03901412412524223,
      "train/kl_loss_r1": 0.09528855904936791,
      "train/total_kl": 0.13430268466472625
    },
    {
      "epoch": 0.5369551484523057,
      "grad_norm": 3.578125,
      "learning_rate": 2.5972836386607706e-05,
      "loss": 0.3743,
      "mean_token_accuracy": 0.9219807744026184,
      "step": 1700,
      "train/kl_loss_qwen": 0.04588426761329174,
      "train/kl_loss_r1": 0.09419769048690796,
      "train/total_kl": 0.14008195847272872
    },
    {
      "epoch": 0.5385344283006949,
      "grad_norm": 4.25,
      "learning_rate": 2.596099178774479e-05,
      "loss": 0.5013,
      "mean_token_accuracy": 0.8997799634933472,
      "step": 1705,
      "train/kl_loss_qwen": 0.06369392238557339,
      "train/kl_loss_r1": 0.13148005902767182,
      "train/total_kl": 0.19517398327589036
    },
    {
      "epoch": 0.540113708149084,
      "grad_norm": 3.84375,
      "learning_rate": 2.594914718888187e-05,
      "loss": 0.3672,
      "mean_token_accuracy": 0.9163668751716614,
      "step": 1710,
      "train/kl_loss_qwen": 0.03281854912638664,
      "train/kl_loss_r1": 0.08420985490083695,
      "train/total_kl": 0.11702840328216553
    },
    {
      "epoch": 0.5416929879974731,
      "grad_norm": 4.5,
      "learning_rate": 2.593730259001895e-05,
      "loss": 0.3758,
      "mean_token_accuracy": 0.9166756272315979,
      "step": 1715,
      "train/kl_loss_qwen": 0.03219302296638489,
      "train/kl_loss_r1": 0.07910310849547386,
      "train/total_kl": 0.11129613220691681
    },
    {
      "epoch": 0.5432722678458622,
      "grad_norm": 4.40625,
      "learning_rate": 2.5925457991156034e-05,
      "loss": 0.6668,
      "mean_token_accuracy": 0.910112178325653,
      "step": 1720,
      "train/kl_loss_qwen": 0.12527670562267304,
      "train/kl_loss_r1": 0.2639615051448345,
      "train/total_kl": 0.38923820555210115
    },
    {
      "epoch": 0.5448515476942514,
      "grad_norm": 3.359375,
      "learning_rate": 2.5913613392293116e-05,
      "loss": 0.3742,
      "mean_token_accuracy": 0.931249487400055,
      "step": 1725,
      "train/kl_loss_qwen": 0.05578612610697746,
      "train/kl_loss_r1": 0.10654140263795853,
      "train/total_kl": 0.16232752650976182
    },
    {
      "epoch": 0.5464308275426406,
      "grad_norm": 3.9375,
      "learning_rate": 2.5901768793430198e-05,
      "loss": 0.3415,
      "mean_token_accuracy": 0.9218314051628113,
      "step": 1730,
      "train/kl_loss_qwen": 0.029083089157938957,
      "train/kl_loss_r1": 0.07852198630571365,
      "train/total_kl": 0.10760507434606552
    },
    {
      "epoch": 0.5480101073910297,
      "grad_norm": 3.234375,
      "learning_rate": 2.588992419456728e-05,
      "loss": 0.3687,
      "mean_token_accuracy": 0.911465835571289,
      "step": 1735,
      "train/kl_loss_qwen": 0.028059108927845956,
      "train/kl_loss_r1": 0.0794032335281372,
      "train/total_kl": 0.10746234506368638
    },
    {
      "epoch": 0.5495893872394189,
      "grad_norm": 2.140625,
      "learning_rate": 2.5878079595704358e-05,
      "loss": 0.3173,
      "mean_token_accuracy": 0.926284384727478,
      "step": 1740,
      "train/kl_loss_qwen": 0.029044802486896514,
      "train/kl_loss_r1": 0.05795014426112175,
      "train/total_kl": 0.0869949460029602
    },
    {
      "epoch": 0.551168667087808,
      "grad_norm": 4.9375,
      "learning_rate": 2.5866234996841443e-05,
      "loss": 0.3734,
      "mean_token_accuracy": 0.926533555984497,
      "step": 1745,
      "train/kl_loss_qwen": 0.04937201365828514,
      "train/kl_loss_r1": 0.0904840111732483,
      "train/total_kl": 0.13985602855682372
    },
    {
      "epoch": 0.5527479469361971,
      "grad_norm": 3.890625,
      "learning_rate": 2.5854390397978522e-05,
      "loss": 0.3864,
      "mean_token_accuracy": 0.9201473474502564,
      "step": 1750,
      "train/kl_loss_qwen": 0.04419952854514122,
      "train/kl_loss_r1": 0.08688022941350937,
      "train/total_kl": 0.13107975721359252
    },
    {
      "epoch": 0.5543272267845862,
      "grad_norm": 3.84375,
      "learning_rate": 2.5842545799115604e-05,
      "loss": 0.4545,
      "mean_token_accuracy": 0.9016712665557861,
      "step": 1755,
      "train/kl_loss_qwen": 0.04680348336696625,
      "train/kl_loss_r1": 0.10328627973794938,
      "train/total_kl": 0.1500897616147995
    },
    {
      "epoch": 0.5559065066329754,
      "grad_norm": 4.53125,
      "learning_rate": 2.5830701200252686e-05,
      "loss": 0.4326,
      "mean_token_accuracy": 0.9069448947906494,
      "step": 1760,
      "train/kl_loss_qwen": 0.036069529876112935,
      "train/kl_loss_r1": 0.08481314927339553,
      "train/total_kl": 0.1208826795220375
    },
    {
      "epoch": 0.5574857864813645,
      "grad_norm": 3.203125,
      "learning_rate": 2.5818856601389765e-05,
      "loss": 0.3832,
      "mean_token_accuracy": 0.9130987763404846,
      "step": 1765,
      "train/kl_loss_qwen": 0.030616405233740807,
      "train/kl_loss_r1": 0.07554664462804794,
      "train/total_kl": 0.1061630517244339
    },
    {
      "epoch": 0.5590650663297536,
      "grad_norm": 4.75,
      "learning_rate": 2.580701200252685e-05,
      "loss": 0.3925,
      "mean_token_accuracy": 0.9250763654708862,
      "step": 1770,
      "train/kl_loss_qwen": 0.0430451650172472,
      "train/kl_loss_r1": 0.10602475106716155,
      "train/total_kl": 0.14906991720199586
    },
    {
      "epoch": 0.5606443461781427,
      "grad_norm": 4.0625,
      "learning_rate": 2.579516740366393e-05,
      "loss": 0.3591,
      "mean_token_accuracy": 0.9254194974899292,
      "step": 1775,
      "train/kl_loss_qwen": 0.0309088584035635,
      "train/kl_loss_r1": 0.08404684066772461,
      "train/total_kl": 0.11495569795370102
    },
    {
      "epoch": 0.5622236260265319,
      "grad_norm": 3.453125,
      "learning_rate": 2.578332280480101e-05,
      "loss": 0.3402,
      "mean_token_accuracy": 0.9290234804153442,
      "step": 1780,
      "train/kl_loss_qwen": 0.032170962914824484,
      "train/kl_loss_r1": 0.07832227051258087,
      "train/total_kl": 0.1104932352900505
    },
    {
      "epoch": 0.563802905874921,
      "grad_norm": 4.0625,
      "learning_rate": 2.5771478205938092e-05,
      "loss": 0.3883,
      "mean_token_accuracy": 0.9223749876022339,
      "step": 1785,
      "train/kl_loss_qwen": 0.045120691508054735,
      "train/kl_loss_r1": 0.10748956575989724,
      "train/total_kl": 0.15261025875806808
    },
    {
      "epoch": 0.5653821857233101,
      "grad_norm": 3.578125,
      "learning_rate": 2.5759633607075174e-05,
      "loss": 0.3363,
      "mean_token_accuracy": 0.9299935460090637,
      "step": 1790,
      "train/kl_loss_qwen": 0.028525831922888756,
      "train/kl_loss_r1": 0.07632275521755219,
      "train/total_kl": 0.10484858751296997
    },
    {
      "epoch": 0.5669614655716994,
      "grad_norm": 2.984375,
      "learning_rate": 2.5747789008212253e-05,
      "loss": 0.389,
      "mean_token_accuracy": 0.9258982181549072,
      "step": 1795,
      "train/kl_loss_qwen": 0.053332074731588366,
      "train/kl_loss_r1": 0.10729452297091484,
      "train/total_kl": 0.1606265962123871
    },
    {
      "epoch": 0.5685407454200885,
      "grad_norm": 2.734375,
      "learning_rate": 2.573594440934934e-05,
      "loss": 0.365,
      "mean_token_accuracy": 0.9245519757270813,
      "step": 1800,
      "train/kl_loss_qwen": 0.03469650223851204,
      "train/kl_loss_r1": 0.07566341459751129,
      "train/total_kl": 0.11035991609096527
    },
    {
      "epoch": 0.5701200252684776,
      "grad_norm": 3.640625,
      "learning_rate": 2.5724099810486417e-05,
      "loss": 0.3602,
      "mean_token_accuracy": 0.9156264305114746,
      "step": 1805,
      "train/kl_loss_qwen": 0.034358056262135506,
      "train/kl_loss_r1": 0.0629163920879364,
      "train/total_kl": 0.097274449467659
    },
    {
      "epoch": 0.5716993051168667,
      "grad_norm": 4.28125,
      "learning_rate": 2.5712255211623502e-05,
      "loss": 0.4306,
      "mean_token_accuracy": 0.9221442461013794,
      "step": 1810,
      "train/kl_loss_qwen": 0.06118791475892067,
      "train/kl_loss_r1": 0.11857841908931732,
      "train/total_kl": 0.17976633310317994
    },
    {
      "epoch": 0.5732785849652559,
      "grad_norm": 3.578125,
      "learning_rate": 2.570041061276058e-05,
      "loss": 0.3927,
      "mean_token_accuracy": 0.9113044977188111,
      "step": 1815,
      "train/kl_loss_qwen": 0.033102355897426605,
      "train/kl_loss_r1": 0.08294531255960465,
      "train/total_kl": 0.11604766994714737
    },
    {
      "epoch": 0.574857864813645,
      "grad_norm": 3.921875,
      "learning_rate": 2.5688566013897663e-05,
      "loss": 0.3813,
      "mean_token_accuracy": 0.9196361660957336,
      "step": 1820,
      "train/kl_loss_qwen": 0.0394960880279541,
      "train/kl_loss_r1": 0.09961815327405929,
      "train/total_kl": 0.13911423981189727
    },
    {
      "epoch": 0.5764371446620341,
      "grad_norm": 3.09375,
      "learning_rate": 2.5676721415034745e-05,
      "loss": 0.3473,
      "mean_token_accuracy": 0.9316370010375976,
      "step": 1825,
      "train/kl_loss_qwen": 0.04248550720512867,
      "train/kl_loss_r1": 0.08670611828565597,
      "train/total_kl": 0.12919162660837175
    },
    {
      "epoch": 0.5780164245104232,
      "grad_norm": 5.3125,
      "learning_rate": 2.5664876816171827e-05,
      "loss": 0.3995,
      "mean_token_accuracy": 0.9238369584083557,
      "step": 1830,
      "train/kl_loss_qwen": 0.04977034777402878,
      "train/kl_loss_r1": 0.11902538537979127,
      "train/total_kl": 0.16879573762416838
    },
    {
      "epoch": 0.5795957043588124,
      "grad_norm": 3.140625,
      "learning_rate": 2.5653032217308905e-05,
      "loss": 0.3767,
      "mean_token_accuracy": 0.918154513835907,
      "step": 1835,
      "train/kl_loss_qwen": 0.03267547190189361,
      "train/kl_loss_r1": 0.07961508929729462,
      "train/total_kl": 0.11229056268930435
    },
    {
      "epoch": 0.5811749842072015,
      "grad_norm": 5.96875,
      "learning_rate": 2.564118761844599e-05,
      "loss": 0.3975,
      "mean_token_accuracy": 0.9179840803146362,
      "step": 1840,
      "train/kl_loss_qwen": 0.04037301577627659,
      "train/kl_loss_r1": 0.08598746061325073,
      "train/total_kl": 0.12636047154664992
    },
    {
      "epoch": 0.5827542640555906,
      "grad_norm": 3.15625,
      "learning_rate": 2.562934301958307e-05,
      "loss": 0.4136,
      "mean_token_accuracy": 0.9118851542472839,
      "step": 1845,
      "train/kl_loss_qwen": 0.03349232450127602,
      "train/kl_loss_r1": 0.08008495867252349,
      "train/total_kl": 0.11357728242874146
    },
    {
      "epoch": 0.5843335439039797,
      "grad_norm": 3.421875,
      "learning_rate": 2.5617498420720155e-05,
      "loss": 0.3536,
      "mean_token_accuracy": 0.9179378032684327,
      "step": 1850,
      "train/kl_loss_qwen": 0.034217759221792224,
      "train/kl_loss_r1": 0.070881237834692,
      "train/total_kl": 0.1050989955663681
    },
    {
      "epoch": 0.585912823752369,
      "grad_norm": 3.921875,
      "learning_rate": 2.5605653821857233e-05,
      "loss": 0.3715,
      "mean_token_accuracy": 0.9173318982124329,
      "step": 1855,
      "train/kl_loss_qwen": 0.04027971848845482,
      "train/kl_loss_r1": 0.08568225651979447,
      "train/total_kl": 0.12596197575330734
    },
    {
      "epoch": 0.5874921036007581,
      "grad_norm": 3.828125,
      "learning_rate": 2.5593809222994315e-05,
      "loss": 0.3539,
      "mean_token_accuracy": 0.9180716514587403,
      "step": 1860,
      "train/kl_loss_qwen": 0.029574811831116675,
      "train/kl_loss_r1": 0.06816345825791359,
      "train/total_kl": 0.09773827046155929
    },
    {
      "epoch": 0.5890713834491472,
      "grad_norm": 3.59375,
      "learning_rate": 2.5581964624131397e-05,
      "loss": 0.3534,
      "mean_token_accuracy": 0.919114089012146,
      "step": 1865,
      "train/kl_loss_qwen": 0.027210471406579017,
      "train/kl_loss_r1": 0.08067434057593345,
      "train/total_kl": 0.10788481384515762
    },
    {
      "epoch": 0.5906506632975363,
      "grad_norm": 6.0625,
      "learning_rate": 2.557012002526848e-05,
      "loss": 0.4286,
      "mean_token_accuracy": 0.9099796056747437,
      "step": 1870,
      "train/kl_loss_qwen": 0.04612473696470261,
      "train/kl_loss_r1": 0.10810503661632538,
      "train/total_kl": 0.15422977209091188
    },
    {
      "epoch": 0.5922299431459255,
      "grad_norm": 4.09375,
      "learning_rate": 2.555827542640556e-05,
      "loss": 0.4047,
      "mean_token_accuracy": 0.9200983285903931,
      "step": 1875,
      "train/kl_loss_qwen": 0.041810973361134526,
      "train/kl_loss_r1": 0.10977670103311539,
      "train/total_kl": 0.151587675511837
    },
    {
      "epoch": 0.5938092229943146,
      "grad_norm": 3.234375,
      "learning_rate": 2.5546430827542643e-05,
      "loss": 0.3488,
      "mean_token_accuracy": 0.9246328949928284,
      "step": 1880,
      "train/kl_loss_qwen": 0.02954785097390413,
      "train/kl_loss_r1": 0.08277009725570679,
      "train/total_kl": 0.11231794655323028
    },
    {
      "epoch": 0.5953885028427037,
      "grad_norm": 3.265625,
      "learning_rate": 2.5534586228679722e-05,
      "loss": 0.401,
      "mean_token_accuracy": 0.9066407442092895,
      "step": 1885,
      "train/kl_loss_qwen": 0.031674535013735294,
      "train/kl_loss_r1": 0.08100252002477645,
      "train/total_kl": 0.1126770555973053
    },
    {
      "epoch": 0.5969677826910929,
      "grad_norm": 3.28125,
      "learning_rate": 2.5522741629816804e-05,
      "loss": 0.412,
      "mean_token_accuracy": 0.9094525218009949,
      "step": 1890,
      "train/kl_loss_qwen": 0.03353483974933624,
      "train/kl_loss_r1": 0.09072587639093399,
      "train/total_kl": 0.12426071763038635
    },
    {
      "epoch": 0.598547062539482,
      "grad_norm": 3.703125,
      "learning_rate": 2.5510897030953886e-05,
      "loss": 0.3326,
      "mean_token_accuracy": 0.9231523513793946,
      "step": 1895,
      "train/kl_loss_qwen": 0.020197218284010887,
      "train/kl_loss_r1": 0.07376486212015151,
      "train/total_kl": 0.0939620777964592
    },
    {
      "epoch": 0.6001263423878711,
      "grad_norm": 5.1875,
      "learning_rate": 2.5499052432090964e-05,
      "loss": 0.7372,
      "mean_token_accuracy": 0.8958821892738342,
      "step": 1900,
      "train/kl_loss_qwen": 0.15676370672881604,
      "train/kl_loss_r1": 0.26219249367713926,
      "train/total_kl": 0.41895620077848433
    },
    {
      "epoch": 0.6017056222362602,
      "grad_norm": 4.3125,
      "learning_rate": 2.548720783322805e-05,
      "loss": 0.4185,
      "mean_token_accuracy": 0.9271070241928101,
      "step": 1905,
      "train/kl_loss_qwen": 0.06658701561391353,
      "train/kl_loss_r1": 0.13123576790094377,
      "train/total_kl": 0.19782278388738633
    },
    {
      "epoch": 0.6032849020846494,
      "grad_norm": 3.28125,
      "learning_rate": 2.5475363234365128e-05,
      "loss": 0.3736,
      "mean_token_accuracy": 0.917541217803955,
      "step": 1910,
      "train/kl_loss_qwen": 0.026435641571879387,
      "train/kl_loss_r1": 0.0843990333378315,
      "train/total_kl": 0.11083467602729798
    },
    {
      "epoch": 0.6048641819330385,
      "grad_norm": 4.125,
      "learning_rate": 2.5463518635502214e-05,
      "loss": 0.3766,
      "mean_token_accuracy": 0.9154990196228028,
      "step": 1915,
      "train/kl_loss_qwen": 0.038460567593574524,
      "train/kl_loss_r1": 0.07671593725681305,
      "train/total_kl": 0.11517650485038758
    },
    {
      "epoch": 0.6064434617814277,
      "grad_norm": 4.40625,
      "learning_rate": 2.5451674036639292e-05,
      "loss": 0.4104,
      "mean_token_accuracy": 0.9089174151420594,
      "step": 1920,
      "train/kl_loss_qwen": 0.035858848690986635,
      "train/kl_loss_r1": 0.08351083993911743,
      "train/total_kl": 0.11936968863010407
    },
    {
      "epoch": 0.6080227416298168,
      "grad_norm": 5.09375,
      "learning_rate": 2.5439829437776374e-05,
      "loss": 0.7412,
      "mean_token_accuracy": 0.8905421614646911,
      "step": 1925,
      "train/kl_loss_qwen": 0.13072961643338205,
      "train/kl_loss_r1": 0.27700538113713263,
      "train/total_kl": 0.40773499459028245
    },
    {
      "epoch": 0.609602021478206,
      "grad_norm": 12.1875,
      "learning_rate": 2.5427984838913456e-05,
      "loss": 0.5825,
      "mean_token_accuracy": 0.9122620105743409,
      "step": 1930,
      "train/kl_loss_qwen": 0.10973418168723584,
      "train/kl_loss_r1": 0.21295054703950883,
      "train/total_kl": 0.32268472760915756
    },
    {
      "epoch": 0.6111813013265951,
      "grad_norm": 2.765625,
      "learning_rate": 2.5416140240050538e-05,
      "loss": 0.3788,
      "mean_token_accuracy": 0.9218842267990113,
      "step": 1935,
      "train/kl_loss_qwen": 0.03174293637275696,
      "train/kl_loss_r1": 0.071744754165411,
      "train/total_kl": 0.10348768830299378
    },
    {
      "epoch": 0.6127605811749842,
      "grad_norm": 4.5,
      "learning_rate": 2.5404295641187617e-05,
      "loss": 0.3861,
      "mean_token_accuracy": 0.9222492694854736,
      "step": 1940,
      "train/kl_loss_qwen": 0.039866327121853826,
      "train/kl_loss_r1": 0.087225341796875,
      "train/total_kl": 0.12709167152643203
    },
    {
      "epoch": 0.6143398610233733,
      "grad_norm": 2.8125,
      "learning_rate": 2.5392451042324702e-05,
      "loss": 0.3303,
      "mean_token_accuracy": 0.9193985939025879,
      "step": 1945,
      "train/kl_loss_qwen": 0.029998742043972015,
      "train/kl_loss_r1": 0.07054983973503112,
      "train/total_kl": 0.10054858326911927
    },
    {
      "epoch": 0.6159191408717625,
      "grad_norm": 3.71875,
      "learning_rate": 2.538060644346178e-05,
      "loss": 0.4027,
      "mean_token_accuracy": 0.9148276686668396,
      "step": 1950,
      "train/kl_loss_qwen": 0.03684010617434978,
      "train/kl_loss_r1": 0.0878615289926529,
      "train/total_kl": 0.1247016355395317
    },
    {
      "epoch": 0.6174984207201516,
      "grad_norm": 4.125,
      "learning_rate": 2.5368761844598866e-05,
      "loss": 0.4337,
      "mean_token_accuracy": 0.9122296690940856,
      "step": 1955,
      "train/kl_loss_qwen": 0.0571936696767807,
      "train/kl_loss_r1": 0.11535490155220032,
      "train/total_kl": 0.172548571228981
    },
    {
      "epoch": 0.6190777005685407,
      "grad_norm": 4.4375,
      "learning_rate": 2.5356917245735945e-05,
      "loss": 0.3553,
      "mean_token_accuracy": 0.9150336265563965,
      "step": 1960,
      "train/kl_loss_qwen": 0.027801768109202384,
      "train/kl_loss_r1": 0.06904341876506806,
      "train/total_kl": 0.09684518724679947
    },
    {
      "epoch": 0.6206569804169298,
      "grad_norm": 3.0,
      "learning_rate": 2.5345072646873027e-05,
      "loss": 0.3487,
      "mean_token_accuracy": 0.9267673015594482,
      "step": 1965,
      "train/kl_loss_qwen": 0.03857283629477024,
      "train/kl_loss_r1": 0.08039961233735085,
      "train/total_kl": 0.11897244900465012
    },
    {
      "epoch": 0.622236260265319,
      "grad_norm": 3.78125,
      "learning_rate": 2.533322804801011e-05,
      "loss": 0.3761,
      "mean_token_accuracy": 0.928635573387146,
      "step": 1970,
      "train/kl_loss_qwen": 0.058719543367624284,
      "train/kl_loss_r1": 0.09832765460014344,
      "train/total_kl": 0.15704720020294188
    },
    {
      "epoch": 0.6238155401137081,
      "grad_norm": 3.5,
      "learning_rate": 2.532138344914719e-05,
      "loss": 0.4416,
      "mean_token_accuracy": 0.9200133085250854,
      "step": 1975,
      "train/kl_loss_qwen": 0.06923877783119678,
      "train/kl_loss_r1": 0.11734992265701294,
      "train/total_kl": 0.18658870011568068
    },
    {
      "epoch": 0.6253948199620973,
      "grad_norm": 3.8125,
      "learning_rate": 2.530953885028427e-05,
      "loss": 0.3643,
      "mean_token_accuracy": 0.9298650026321411,
      "step": 1980,
      "train/kl_loss_qwen": 0.04414457306265831,
      "train/kl_loss_r1": 0.09041669070720673,
      "train/total_kl": 0.13456126749515535
    },
    {
      "epoch": 0.6269740998104865,
      "grad_norm": 3.40625,
      "learning_rate": 2.5297694251421354e-05,
      "loss": 0.3414,
      "mean_token_accuracy": 0.9145791053771972,
      "step": 1985,
      "train/kl_loss_qwen": 0.02920379638671875,
      "train/kl_loss_r1": 0.06909812465310097,
      "train/total_kl": 0.09830192029476166
    },
    {
      "epoch": 0.6285533796588756,
      "grad_norm": 35.25,
      "learning_rate": 2.5285849652558433e-05,
      "loss": 0.765,
      "mean_token_accuracy": 0.9058949947357178,
      "step": 1990,
      "train/kl_loss_qwen": 0.14503382816910743,
      "train/kl_loss_r1": 0.33853579461574557,
      "train/total_kl": 0.48356961011886596
    },
    {
      "epoch": 0.6301326595072647,
      "grad_norm": 4.40625,
      "learning_rate": 2.527400505369552e-05,
      "loss": 0.3413,
      "mean_token_accuracy": 0.9224784851074219,
      "step": 1995,
      "train/kl_loss_qwen": 0.02844323981553316,
      "train/kl_loss_r1": 0.0693981371819973,
      "train/total_kl": 0.09784137532114982
    },
    {
      "epoch": 0.6317119393556538,
      "grad_norm": 3.234375,
      "learning_rate": 2.5262160454832597e-05,
      "loss": 0.3958,
      "mean_token_accuracy": 0.9191631078720093,
      "step": 2000,
      "train/kl_loss_qwen": 0.05624440461397171,
      "train/kl_loss_r1": 0.08816543370485305,
      "train/total_kl": 0.14440983831882476
    },
    {
      "epoch": 0.633291219204043,
      "grad_norm": 5.125,
      "learning_rate": 2.525031585596968e-05,
      "loss": 0.3451,
      "mean_token_accuracy": 0.928159499168396,
      "step": 2005,
      "train/kl_loss_qwen": 0.042037079110741615,
      "train/kl_loss_r1": 0.0727903589606285,
      "train/total_kl": 0.11482744067907333
    },
    {
      "epoch": 0.6348704990524321,
      "grad_norm": 4.84375,
      "learning_rate": 2.523847125710676e-05,
      "loss": 0.4392,
      "mean_token_accuracy": 0.9190120816230773,
      "step": 2010,
      "train/kl_loss_qwen": 0.06650599539279937,
      "train/kl_loss_r1": 0.12747553884983062,
      "train/total_kl": 0.19398153424263
    },
    {
      "epoch": 0.6364497789008212,
      "grad_norm": 3.875,
      "learning_rate": 2.522662665824384e-05,
      "loss": 0.5027,
      "mean_token_accuracy": 0.9130654692649841,
      "step": 2015,
      "train/kl_loss_qwen": 0.06319191418588162,
      "train/kl_loss_r1": 0.1513908676803112,
      "train/total_kl": 0.21458278149366378
    },
    {
      "epoch": 0.6380290587492103,
      "grad_norm": 5.1875,
      "learning_rate": 2.521478205938092e-05,
      "loss": 0.3944,
      "mean_token_accuracy": 0.9206954717636109,
      "step": 2020,
      "train/kl_loss_qwen": 0.0532749243080616,
      "train/kl_loss_r1": 0.10241053998470306,
      "train/total_kl": 0.1556854635477066
    },
    {
      "epoch": 0.6396083385975995,
      "grad_norm": 4.03125,
      "learning_rate": 2.5202937460518003e-05,
      "loss": 0.3705,
      "mean_token_accuracy": 0.9214539289474487,
      "step": 2025,
      "train/kl_loss_qwen": 0.03493940606713295,
      "train/kl_loss_r1": 0.07996350005269051,
      "train/total_kl": 0.1149029091000557
    },
    {
      "epoch": 0.6411876184459886,
      "grad_norm": 4.75,
      "learning_rate": 2.5191092861655085e-05,
      "loss": 0.4112,
      "mean_token_accuracy": 0.908880615234375,
      "step": 2030,
      "train/kl_loss_qwen": 0.03760333098471165,
      "train/kl_loss_r1": 0.09668760299682617,
      "train/total_kl": 0.1342909350991249
    },
    {
      "epoch": 0.6427668982943777,
      "grad_norm": 3.75,
      "learning_rate": 2.5179248262792167e-05,
      "loss": 0.3877,
      "mean_token_accuracy": 0.9216402411460877,
      "step": 2035,
      "train/kl_loss_qwen": 0.045970536023378375,
      "train/kl_loss_r1": 0.09495973438024521,
      "train/total_kl": 0.1409302681684494
    },
    {
      "epoch": 0.6443461781427668,
      "grad_norm": 2.765625,
      "learning_rate": 2.516740366392925e-05,
      "loss": 0.3784,
      "mean_token_accuracy": 0.9260763168334961,
      "step": 2040,
      "train/kl_loss_qwen": 0.05029257386922836,
      "train/kl_loss_r1": 0.09535304754972458,
      "train/total_kl": 0.1456456184387207
    },
    {
      "epoch": 0.6459254579911561,
      "grad_norm": 2.96875,
      "learning_rate": 2.5155559065066328e-05,
      "loss": 0.4109,
      "mean_token_accuracy": 0.9137315154075623,
      "step": 2045,
      "train/kl_loss_qwen": 0.04600106477737427,
      "train/kl_loss_r1": 0.07891609966754913,
      "train/total_kl": 0.1249171644449234
    },
    {
      "epoch": 0.6475047378395452,
      "grad_norm": 3.203125,
      "learning_rate": 2.5143714466203413e-05,
      "loss": 0.3659,
      "mean_token_accuracy": 0.9146221280097961,
      "step": 2050,
      "train/kl_loss_qwen": 0.027876831218600275,
      "train/kl_loss_r1": 0.0678498163819313,
      "train/total_kl": 0.0957266479730606
    },
    {
      "epoch": 0.6490840176879343,
      "grad_norm": 3.140625,
      "learning_rate": 2.5131869867340492e-05,
      "loss": 0.4017,
      "mean_token_accuracy": 0.912306809425354,
      "step": 2055,
      "train/kl_loss_qwen": 0.04238977022469044,
      "train/kl_loss_r1": 0.08535711988806724,
      "train/total_kl": 0.1277468904852867
    },
    {
      "epoch": 0.6506632975363235,
      "grad_norm": 6.0,
      "learning_rate": 2.5120025268477574e-05,
      "loss": 0.4031,
      "mean_token_accuracy": 0.9077698230743408,
      "step": 2060,
      "train/kl_loss_qwen": 0.02936724703758955,
      "train/kl_loss_r1": 0.08423567041754723,
      "train/total_kl": 0.11360291689634323
    },
    {
      "epoch": 0.6522425773847126,
      "grad_norm": 4.625,
      "learning_rate": 2.5108180669614656e-05,
      "loss": 0.4969,
      "mean_token_accuracy": 0.9100233435630798,
      "step": 2065,
      "train/kl_loss_qwen": 0.06865382231771947,
      "train/kl_loss_r1": 0.14072849452495576,
      "train/total_kl": 0.20938232094049453
    },
    {
      "epoch": 0.6538218572331017,
      "grad_norm": 3.3125,
      "learning_rate": 2.5096336070751738e-05,
      "loss": 0.4137,
      "mean_token_accuracy": 0.91517333984375,
      "step": 2070,
      "train/kl_loss_qwen": 0.04925190731883049,
      "train/kl_loss_r1": 0.10850737392902374,
      "train/total_kl": 0.15775928497314454
    },
    {
      "epoch": 0.6554011370814908,
      "grad_norm": 2.390625,
      "learning_rate": 2.508449147188882e-05,
      "loss": 0.3333,
      "mean_token_accuracy": 0.9289180636405945,
      "step": 2075,
      "train/kl_loss_qwen": 0.033440543711185454,
      "train/kl_loss_r1": 0.07091380879282952,
      "train/total_kl": 0.1043543502688408
    },
    {
      "epoch": 0.65698041692988,
      "grad_norm": 3.203125,
      "learning_rate": 2.5072646873025902e-05,
      "loss": 0.3488,
      "mean_token_accuracy": 0.914642333984375,
      "step": 2080,
      "train/kl_loss_qwen": 0.02942515965551138,
      "train/kl_loss_r1": 0.0692016251385212,
      "train/total_kl": 0.09862678349018097
    },
    {
      "epoch": 0.6585596967782691,
      "grad_norm": 3.953125,
      "learning_rate": 2.506080227416298e-05,
      "loss": 0.4253,
      "mean_token_accuracy": 0.9137975335121155,
      "step": 2085,
      "train/kl_loss_qwen": 0.04863373413681984,
      "train/kl_loss_r1": 0.1063865803182125,
      "train/total_kl": 0.15502031296491622
    },
    {
      "epoch": 0.6601389766266582,
      "grad_norm": 3.28125,
      "learning_rate": 2.5048957675300066e-05,
      "loss": 0.3748,
      "mean_token_accuracy": 0.9185147285461426,
      "step": 2090,
      "train/kl_loss_qwen": 0.029326098412275313,
      "train/kl_loss_r1": 0.08525317460298538,
      "train/total_kl": 0.11457927525043488
    },
    {
      "epoch": 0.6617182564750473,
      "grad_norm": 4.28125,
      "learning_rate": 2.5037113076437144e-05,
      "loss": 0.3569,
      "mean_token_accuracy": 0.9235624432563782,
      "step": 2095,
      "train/kl_loss_qwen": 0.031793155893683436,
      "train/kl_loss_r1": 0.08324898630380631,
      "train/total_kl": 0.11504214257001877
    },
    {
      "epoch": 0.6632975363234365,
      "grad_norm": 4.71875,
      "learning_rate": 2.5025268477574226e-05,
      "loss": 0.3697,
      "mean_token_accuracy": 0.9141849875450134,
      "step": 2100,
      "train/kl_loss_qwen": 0.031283645704388616,
      "train/kl_loss_r1": 0.08270388022065163,
      "train/total_kl": 0.11398752480745315
    },
    {
      "epoch": 0.6648768161718257,
      "grad_norm": 2.84375,
      "learning_rate": 2.5013423878711308e-05,
      "loss": 0.3876,
      "mean_token_accuracy": 0.9249490261077881,
      "step": 2105,
      "train/kl_loss_qwen": 0.039199349284172055,
      "train/kl_loss_r1": 0.09067148938775063,
      "train/total_kl": 0.12987083941698074
    },
    {
      "epoch": 0.6664560960202148,
      "grad_norm": 5.09375,
      "learning_rate": 2.500157927984839e-05,
      "loss": 0.3379,
      "mean_token_accuracy": 0.9264767050743103,
      "step": 2110,
      "train/kl_loss_qwen": 0.03846020996570587,
      "train/kl_loss_r1": 0.07721847146749497,
      "train/total_kl": 0.11567868143320084
    },
    {
      "epoch": 0.668035375868604,
      "grad_norm": 5.53125,
      "learning_rate": 2.4989734680985472e-05,
      "loss": 0.4117,
      "mean_token_accuracy": 0.9213946700096131,
      "step": 2115,
      "train/kl_loss_qwen": 0.040704042091965674,
      "train/kl_loss_r1": 0.11017271280288696,
      "train/total_kl": 0.15087675750255586
    },
    {
      "epoch": 0.6696146557169931,
      "grad_norm": 3.515625,
      "learning_rate": 2.4977890082122554e-05,
      "loss": 0.3973,
      "mean_token_accuracy": 0.9089462161064148,
      "step": 2120,
      "train/kl_loss_qwen": 0.028432265669107438,
      "train/kl_loss_r1": 0.07672996148467064,
      "train/total_kl": 0.10516222566366196
    },
    {
      "epoch": 0.6711939355653822,
      "grad_norm": 2.90625,
      "learning_rate": 2.4966045483259633e-05,
      "loss": 0.3459,
      "mean_token_accuracy": 0.9282707929611206,
      "step": 2125,
      "train/kl_loss_qwen": 0.03086411692202091,
      "train/kl_loss_r1": 0.07920683026313782,
      "train/total_kl": 0.1100709468126297
    },
    {
      "epoch": 0.6727732154137713,
      "grad_norm": 4.375,
      "learning_rate": 2.4954200884396718e-05,
      "loss": 0.3547,
      "mean_token_accuracy": 0.9265871047973633,
      "step": 2130,
      "train/kl_loss_qwen": 0.03855619244277477,
      "train/kl_loss_r1": 0.08854741454124451,
      "train/total_kl": 0.12710361033678055
    },
    {
      "epoch": 0.6743524952621605,
      "grad_norm": 4.75,
      "learning_rate": 2.4942356285533797e-05,
      "loss": 0.4064,
      "mean_token_accuracy": 0.9149534583091736,
      "step": 2135,
      "train/kl_loss_qwen": 0.04170009195804596,
      "train/kl_loss_r1": 0.10134969353675842,
      "train/total_kl": 0.14304978549480438
    },
    {
      "epoch": 0.6759317751105496,
      "grad_norm": 2.734375,
      "learning_rate": 2.493051168667088e-05,
      "loss": 0.3346,
      "mean_token_accuracy": 0.9352992296218872,
      "step": 2140,
      "train/kl_loss_qwen": 0.042602280527353285,
      "train/kl_loss_r1": 0.08730309531092643,
      "train/total_kl": 0.1299053758382797
    },
    {
      "epoch": 0.6775110549589387,
      "grad_norm": 3.84375,
      "learning_rate": 2.491866708780796e-05,
      "loss": 0.4204,
      "mean_token_accuracy": 0.90723956823349,
      "step": 2145,
      "train/kl_loss_qwen": 0.04157908596098423,
      "train/kl_loss_r1": 0.093327946215868,
      "train/total_kl": 0.1349070340394974
    },
    {
      "epoch": 0.6790903348073278,
      "grad_norm": 3.921875,
      "learning_rate": 2.490682248894504e-05,
      "loss": 0.3712,
      "mean_token_accuracy": 0.9190201997756958,
      "step": 2150,
      "train/kl_loss_qwen": 0.0337690569460392,
      "train/kl_loss_r1": 0.07831582725048065,
      "train/total_kl": 0.11208488494157791
    },
    {
      "epoch": 0.680669614655717,
      "grad_norm": 4.125,
      "learning_rate": 2.4894977890082125e-05,
      "loss": 0.3823,
      "mean_token_accuracy": 0.9241498112678528,
      "step": 2155,
      "train/kl_loss_qwen": 0.0477649986743927,
      "train/kl_loss_r1": 0.09849860519170761,
      "train/total_kl": 0.14626360386610032
    },
    {
      "epoch": 0.6822488945041061,
      "grad_norm": 2.75,
      "learning_rate": 2.4883133291219203e-05,
      "loss": 0.3781,
      "mean_token_accuracy": 0.9280685067176819,
      "step": 2160,
      "train/kl_loss_qwen": 0.04589755013585091,
      "train/kl_loss_r1": 0.1038775958120823,
      "train/total_kl": 0.14977514892816543
    },
    {
      "epoch": 0.6838281743524952,
      "grad_norm": 4.34375,
      "learning_rate": 2.4871288692356285e-05,
      "loss": 0.6614,
      "mean_token_accuracy": 0.9021480083465576,
      "step": 2165,
      "train/kl_loss_qwen": 0.09889798164367676,
      "train/kl_loss_r1": 0.25930761396884916,
      "train/total_kl": 0.35820560306310656
    },
    {
      "epoch": 0.6854074542008844,
      "grad_norm": 3.96875,
      "learning_rate": 2.4859444093493367e-05,
      "loss": 0.3373,
      "mean_token_accuracy": 0.9282265901565552,
      "step": 2170,
      "train/kl_loss_qwen": 0.03710666820406914,
      "train/kl_loss_r1": 0.07107429429888726,
      "train/total_kl": 0.10818096101284028
    },
    {
      "epoch": 0.6869867340492736,
      "grad_norm": 3.8125,
      "learning_rate": 2.484759949463045e-05,
      "loss": 0.3368,
      "mean_token_accuracy": 0.9383379340171814,
      "step": 2175,
      "train/kl_loss_qwen": 0.04446404948830605,
      "train/kl_loss_r1": 0.0888557992875576,
      "train/total_kl": 0.13331984877586364
    },
    {
      "epoch": 0.6885660138976627,
      "grad_norm": 2.8125,
      "learning_rate": 2.483575489576753e-05,
      "loss": 0.3787,
      "mean_token_accuracy": 0.9228280782699585,
      "step": 2180,
      "train/kl_loss_qwen": 0.04515460692346096,
      "train/kl_loss_r1": 0.09792126268148423,
      "train/total_kl": 0.1430758684873581
    },
    {
      "epoch": 0.6901452937460518,
      "grad_norm": 4.09375,
      "learning_rate": 2.4823910296904613e-05,
      "loss": 0.3968,
      "mean_token_accuracy": 0.9158631920814514,
      "step": 2185,
      "train/kl_loss_qwen": 0.035012568533420566,
      "train/kl_loss_r1": 0.09668713361024857,
      "train/total_kl": 0.13169970214366913
    },
    {
      "epoch": 0.691724573594441,
      "grad_norm": 3.734375,
      "learning_rate": 2.481206569804169e-05,
      "loss": 0.3794,
      "mean_token_accuracy": 0.9185762047767639,
      "step": 2190,
      "train/kl_loss_qwen": 0.038122583553195,
      "train/kl_loss_r1": 0.08820372149348259,
      "train/total_kl": 0.12632630467414857
    },
    {
      "epoch": 0.6933038534428301,
      "grad_norm": 4.46875,
      "learning_rate": 2.4800221099178777e-05,
      "loss": 0.6912,
      "mean_token_accuracy": 0.9086827278137207,
      "step": 2195,
      "train/kl_loss_qwen": 0.09095740914344788,
      "train/kl_loss_r1": 0.3152101464569569,
      "train/total_kl": 0.40616755485534667
    },
    {
      "epoch": 0.6948831332912192,
      "grad_norm": 2.5,
      "learning_rate": 2.4788376500315856e-05,
      "loss": 0.3478,
      "mean_token_accuracy": 0.930546224117279,
      "step": 2200,
      "train/kl_loss_qwen": 0.033646737784147264,
      "train/kl_loss_r1": 0.09833809286355973,
      "train/total_kl": 0.13198482990264893
    },
    {
      "epoch": 0.6964624131396083,
      "grad_norm": 4.125,
      "learning_rate": 2.4776531901452938e-05,
      "loss": 0.3972,
      "mean_token_accuracy": 0.9240478157997132,
      "step": 2205,
      "train/kl_loss_qwen": 0.051082177087664604,
      "train/kl_loss_r1": 0.11277811676263809,
      "train/total_kl": 0.1638602912425995
    },
    {
      "epoch": 0.6980416929879975,
      "grad_norm": 3.640625,
      "learning_rate": 2.476468730259002e-05,
      "loss": 0.3479,
      "mean_token_accuracy": 0.9245799422264099,
      "step": 2210,
      "train/kl_loss_qwen": 0.03238863199949264,
      "train/kl_loss_r1": 0.08944203704595566,
      "train/total_kl": 0.1218306690454483
    },
    {
      "epoch": 0.6996209728363866,
      "grad_norm": 3.265625,
      "learning_rate": 2.47528427037271e-05,
      "loss": 0.3774,
      "mean_token_accuracy": 0.9242660999298096,
      "step": 2215,
      "train/kl_loss_qwen": 0.04103923663496971,
      "train/kl_loss_r1": 0.09031469225883484,
      "train/total_kl": 0.13135392665863038
    },
    {
      "epoch": 0.7012002526847757,
      "grad_norm": 4.03125,
      "learning_rate": 2.4740998104864184e-05,
      "loss": 0.4059,
      "mean_token_accuracy": 0.9189964056015014,
      "step": 2220,
      "train/kl_loss_qwen": 0.05655071027576923,
      "train/kl_loss_r1": 0.11072072461247444,
      "train/total_kl": 0.16727143675088882
    },
    {
      "epoch": 0.7027795325331648,
      "grad_norm": 3.75,
      "learning_rate": 2.4729153506001265e-05,
      "loss": 0.4317,
      "mean_token_accuracy": 0.9088820815086365,
      "step": 2225,
      "train/kl_loss_qwen": 0.04667430035769939,
      "train/kl_loss_r1": 0.1044918954372406,
      "train/total_kl": 0.15116619765758516
    },
    {
      "epoch": 0.704358812381554,
      "grad_norm": 3.21875,
      "learning_rate": 2.4717308907138344e-05,
      "loss": 0.403,
      "mean_token_accuracy": 0.9169308066368103,
      "step": 2230,
      "train/kl_loss_qwen": 0.041512713953852656,
      "train/kl_loss_r1": 0.08254260644316673,
      "train/total_kl": 0.12405532151460648
    },
    {
      "epoch": 0.7059380922299432,
      "grad_norm": 11.4375,
      "learning_rate": 2.470546430827543e-05,
      "loss": 0.6333,
      "mean_token_accuracy": 0.916389274597168,
      "step": 2235,
      "train/kl_loss_qwen": 0.14926290400326253,
      "train/kl_loss_r1": 0.2189345970749855,
      "train/total_kl": 0.36819748878479003
    },
    {
      "epoch": 0.7075173720783323,
      "grad_norm": 4.09375,
      "learning_rate": 2.4693619709412508e-05,
      "loss": 0.3542,
      "mean_token_accuracy": 0.9199613451957702,
      "step": 2240,
      "train/kl_loss_qwen": 0.03053949810564518,
      "train/kl_loss_r1": 0.07533566057682037,
      "train/total_kl": 0.10587515830993652
    },
    {
      "epoch": 0.7090966519267214,
      "grad_norm": 3.953125,
      "learning_rate": 2.468177511054959e-05,
      "loss": 0.3582,
      "mean_token_accuracy": 0.9274658679962158,
      "step": 2245,
      "train/kl_loss_qwen": 0.04275674633681774,
      "train/kl_loss_r1": 0.09105659648776054,
      "train/total_kl": 0.13381334245204926
    },
    {
      "epoch": 0.7106759317751106,
      "grad_norm": 3.578125,
      "learning_rate": 2.4669930511686672e-05,
      "loss": 0.3703,
      "mean_token_accuracy": 0.9179190158843994,
      "step": 2250,
      "train/kl_loss_qwen": 0.026688238605856895,
      "train/kl_loss_r1": 0.07284217327833176,
      "train/total_kl": 0.09953041076660156
    },
    {
      "epoch": 0.7122552116234997,
      "grad_norm": 4.4375,
      "learning_rate": 2.4658085912823754e-05,
      "loss": 0.4579,
      "mean_token_accuracy": 0.9159493565559387,
      "step": 2255,
      "train/kl_loss_qwen": 0.06464865319430828,
      "train/kl_loss_r1": 0.12074823081493377,
      "train/total_kl": 0.18539688885211944
    },
    {
      "epoch": 0.7138344914718888,
      "grad_norm": 3.5,
      "learning_rate": 2.4646241313960836e-05,
      "loss": 0.3164,
      "mean_token_accuracy": 0.9338837742805481,
      "step": 2260,
      "train/kl_loss_qwen": 0.028987813740968704,
      "train/kl_loss_r1": 0.07401799783110619,
      "train/total_kl": 0.10300580859184265
    },
    {
      "epoch": 0.715413771320278,
      "grad_norm": 4.78125,
      "learning_rate": 2.4634396715097915e-05,
      "loss": 0.4327,
      "mean_token_accuracy": 0.9100850582122803,
      "step": 2265,
      "train/kl_loss_qwen": 0.053746733069419864,
      "train/kl_loss_r1": 0.11145351827144623,
      "train/total_kl": 0.16520025283098222
    },
    {
      "epoch": 0.7169930511686671,
      "grad_norm": 3.125,
      "learning_rate": 2.4622552116234996e-05,
      "loss": 0.3771,
      "mean_token_accuracy": 0.9210043311119079,
      "step": 2270,
      "train/kl_loss_qwen": 0.03653993420302868,
      "train/kl_loss_r1": 0.08889604061841964,
      "train/total_kl": 0.12543597519397737
    },
    {
      "epoch": 0.7185723310170562,
      "grad_norm": 4.3125,
      "learning_rate": 2.461070751737208e-05,
      "loss": 0.3923,
      "mean_token_accuracy": 0.9234745025634765,
      "step": 2275,
      "train/kl_loss_qwen": 0.047175091505050656,
      "train/kl_loss_r1": 0.0883270338177681,
      "train/total_kl": 0.1355021268129349
    },
    {
      "epoch": 0.7201516108654453,
      "grad_norm": 4.125,
      "learning_rate": 2.459886291850916e-05,
      "loss": 0.3859,
      "mean_token_accuracy": 0.92585209608078,
      "step": 2280,
      "train/kl_loss_qwen": 0.048363619670271876,
      "train/kl_loss_r1": 0.10164024457335472,
      "train/total_kl": 0.15000386387109757
    },
    {
      "epoch": 0.7217308907138344,
      "grad_norm": 5.28125,
      "learning_rate": 2.458701831964624e-05,
      "loss": 0.4395,
      "mean_token_accuracy": 0.9216931462287903,
      "step": 2285,
      "train/kl_loss_qwen": 0.07007896527647972,
      "train/kl_loss_r1": 0.11823424100875854,
      "train/total_kl": 0.18831320405006408
    },
    {
      "epoch": 0.7233101705622236,
      "grad_norm": 4.375,
      "learning_rate": 2.4575173720783324e-05,
      "loss": 0.3599,
      "mean_token_accuracy": 0.9223866105079651,
      "step": 2290,
      "train/kl_loss_qwen": 0.033086203038692474,
      "train/kl_loss_r1": 0.08134551271796227,
      "train/total_kl": 0.1144317165017128
    },
    {
      "epoch": 0.7248894504106128,
      "grad_norm": 3.953125,
      "learning_rate": 2.4563329121920403e-05,
      "loss": 0.3376,
      "mean_token_accuracy": 0.9253845572471618,
      "step": 2295,
      "train/kl_loss_qwen": 0.027293992787599565,
      "train/kl_loss_r1": 0.07916483432054519,
      "train/total_kl": 0.10645882934331893
    },
    {
      "epoch": 0.7264687302590019,
      "grad_norm": 3.4375,
      "learning_rate": 2.455148452305749e-05,
      "loss": 0.5856,
      "mean_token_accuracy": 0.8983238816261292,
      "step": 2300,
      "train/kl_loss_qwen": 0.10275081880390644,
      "train/kl_loss_r1": 0.17683854699134827,
      "train/total_kl": 0.2795893713831902
    },
    {
      "epoch": 0.7280480101073911,
      "grad_norm": 4.09375,
      "learning_rate": 2.4539639924194567e-05,
      "loss": 0.3977,
      "mean_token_accuracy": 0.9131863713264465,
      "step": 2305,
      "train/kl_loss_qwen": 0.03624062165617943,
      "train/kl_loss_r1": 0.06860300078988076,
      "train/total_kl": 0.1048436239361763
    },
    {
      "epoch": 0.7296272899557802,
      "grad_norm": 3.578125,
      "learning_rate": 2.452779532533165e-05,
      "loss": 0.6551,
      "mean_token_accuracy": 0.908154034614563,
      "step": 2310,
      "train/kl_loss_qwen": 0.12643007561564445,
      "train/kl_loss_r1": 0.2595177531242371,
      "train/total_kl": 0.38594782799482347
    },
    {
      "epoch": 0.7312065698041693,
      "grad_norm": 3.03125,
      "learning_rate": 2.451595072646873e-05,
      "loss": 0.3593,
      "mean_token_accuracy": 0.9201328754425049,
      "step": 2315,
      "train/kl_loss_qwen": 0.036088640615344045,
      "train/kl_loss_r1": 0.08747091442346573,
      "train/total_kl": 0.12355955541133881
    },
    {
      "epoch": 0.7327858496525584,
      "grad_norm": 3.0,
      "learning_rate": 2.4504106127605813e-05,
      "loss": 0.4077,
      "mean_token_accuracy": 0.9254467248916626,
      "step": 2320,
      "train/kl_loss_qwen": 0.04676676392555237,
      "train/kl_loss_r1": 0.11094883307814599,
      "train/total_kl": 0.15771559923887252
    },
    {
      "epoch": 0.7343651295009476,
      "grad_norm": 2.84375,
      "learning_rate": 2.449226152874289e-05,
      "loss": 0.3511,
      "mean_token_accuracy": 0.9300658106803894,
      "step": 2325,
      "train/kl_loss_qwen": 0.041116179153323176,
      "train/kl_loss_r1": 0.08788876384496688,
      "train/total_kl": 0.12900494486093522
    },
    {
      "epoch": 0.7359444093493367,
      "grad_norm": 2.96875,
      "learning_rate": 2.4480416929879977e-05,
      "loss": 0.3463,
      "mean_token_accuracy": 0.9217430949211121,
      "step": 2330,
      "train/kl_loss_qwen": 0.030316578596830367,
      "train/kl_loss_r1": 0.07793957740068436,
      "train/total_kl": 0.10825615674257279
    },
    {
      "epoch": 0.7375236891977258,
      "grad_norm": 1.9921875,
      "learning_rate": 2.4468572331017055e-05,
      "loss": 0.3574,
      "mean_token_accuracy": 0.9194462418556213,
      "step": 2335,
      "train/kl_loss_qwen": 0.03517359122633934,
      "train/kl_loss_r1": 0.08479614779353142,
      "train/total_kl": 0.11996974050998688
    },
    {
      "epoch": 0.7391029690461149,
      "grad_norm": 3.671875,
      "learning_rate": 2.445672773215414e-05,
      "loss": 0.3507,
      "mean_token_accuracy": 0.9328651428222656,
      "step": 2340,
      "train/kl_loss_qwen": 0.046850771829485896,
      "train/kl_loss_r1": 0.10080698356032372,
      "train/total_kl": 0.1476577565073967
    },
    {
      "epoch": 0.7406822488945041,
      "grad_norm": 3.125,
      "learning_rate": 2.444488313329122e-05,
      "loss": 0.3274,
      "mean_token_accuracy": 0.9312006235122681,
      "step": 2345,
      "train/kl_loss_qwen": 0.03337893150746822,
      "train/kl_loss_r1": 0.07775259539484977,
      "train/total_kl": 0.11113152652978897
    },
    {
      "epoch": 0.7422615287428932,
      "grad_norm": 3.265625,
      "learning_rate": 2.44330385344283e-05,
      "loss": 0.3668,
      "mean_token_accuracy": 0.9265905499458313,
      "step": 2350,
      "train/kl_loss_qwen": 0.03836112655699253,
      "train/kl_loss_r1": 0.10450647622346879,
      "train/total_kl": 0.1428676053881645
    },
    {
      "epoch": 0.7438408085912823,
      "grad_norm": 3.140625,
      "learning_rate": 2.4421193935565383e-05,
      "loss": 0.3638,
      "mean_token_accuracy": 0.9213193655014038,
      "step": 2355,
      "train/kl_loss_qwen": 0.028002501651644706,
      "train/kl_loss_r1": 0.08434471338987351,
      "train/total_kl": 0.11234721690416336
    },
    {
      "epoch": 0.7454200884396716,
      "grad_norm": 3.015625,
      "learning_rate": 2.4409349336702465e-05,
      "loss": 0.3851,
      "mean_token_accuracy": 0.9149267911911011,
      "step": 2360,
      "train/kl_loss_qwen": 0.03710017129778862,
      "train/kl_loss_r1": 0.08855679631233215,
      "train/total_kl": 0.12565696686506272
    },
    {
      "epoch": 0.7469993682880607,
      "grad_norm": 3.671875,
      "learning_rate": 2.4397504737839544e-05,
      "loss": 0.7293,
      "mean_token_accuracy": 0.8901386499404907,
      "step": 2365,
      "train/kl_loss_qwen": 0.14339055269956588,
      "train/kl_loss_r1": 0.24662558734416962,
      "train/total_kl": 0.3900161415338516
    },
    {
      "epoch": 0.7485786481364498,
      "grad_norm": 4.25,
      "learning_rate": 2.438566013897663e-05,
      "loss": 0.4049,
      "mean_token_accuracy": 0.9148341178894043,
      "step": 2370,
      "train/kl_loss_qwen": 0.04336503557860851,
      "train/kl_loss_r1": 0.09680078327655792,
      "train/total_kl": 0.140165813267231
    },
    {
      "epoch": 0.7501579279848389,
      "grad_norm": 4.125,
      "learning_rate": 2.4373815540113708e-05,
      "loss": 0.3976,
      "mean_token_accuracy": 0.9196026206016541,
      "step": 2375,
      "train/kl_loss_qwen": 0.04883059449493885,
      "train/kl_loss_r1": 0.09221117421984673,
      "train/total_kl": 0.1410417675971985
    },
    {
      "epoch": 0.7517372078332281,
      "grad_norm": 2.734375,
      "learning_rate": 2.4361970941250793e-05,
      "loss": 0.3157,
      "mean_token_accuracy": 0.93125239610672,
      "step": 2380,
      "train/kl_loss_qwen": 0.030073122680187227,
      "train/kl_loss_r1": 0.07321802750229836,
      "train/total_kl": 0.10329115092754364
    },
    {
      "epoch": 0.7533164876816172,
      "grad_norm": 3.5625,
      "learning_rate": 2.4350126342387872e-05,
      "loss": 0.3896,
      "mean_token_accuracy": 0.9188811540603637,
      "step": 2385,
      "train/kl_loss_qwen": 0.04191601425409317,
      "train/kl_loss_r1": 0.09682500660419464,
      "train/total_kl": 0.13874102383852005
    },
    {
      "epoch": 0.7548957675300063,
      "grad_norm": 3.828125,
      "learning_rate": 2.433828174352495e-05,
      "loss": 0.7651,
      "mean_token_accuracy": 0.9011089086532593,
      "step": 2390,
      "train/kl_loss_qwen": 0.1477519679814577,
      "train/kl_loss_r1": 0.3129396215081215,
      "train/total_kl": 0.4606915906071663
    },
    {
      "epoch": 0.7564750473783954,
      "grad_norm": 3.953125,
      "learning_rate": 2.4326437144662036e-05,
      "loss": 0.4301,
      "mean_token_accuracy": 0.901045823097229,
      "step": 2395,
      "train/kl_loss_qwen": 0.03304285518825054,
      "train/kl_loss_r1": 0.08088582009077072,
      "train/total_kl": 0.11392867863178253
    },
    {
      "epoch": 0.7580543272267846,
      "grad_norm": 4.46875,
      "learning_rate": 2.4314592545799114e-05,
      "loss": 0.3525,
      "mean_token_accuracy": 0.9174474000930786,
      "step": 2400,
      "train/kl_loss_qwen": 0.028352801874279975,
      "train/kl_loss_r1": 0.07779977843165398,
      "train/total_kl": 0.10615258067846298
    },
    {
      "epoch": 0.7596336070751737,
      "grad_norm": 2.828125,
      "learning_rate": 2.43027479469362e-05,
      "loss": 0.3562,
      "mean_token_accuracy": 0.9311485767364502,
      "step": 2405,
      "train/kl_loss_qwen": 0.03659307099878788,
      "train/kl_loss_r1": 0.10589836314320564,
      "train/total_kl": 0.1424914315342903
    },
    {
      "epoch": 0.7612128869235628,
      "grad_norm": 3.640625,
      "learning_rate": 2.4290903348073278e-05,
      "loss": 0.417,
      "mean_token_accuracy": 0.9137800931930542,
      "step": 2410,
      "train/kl_loss_qwen": 0.050878635793924334,
      "train/kl_loss_r1": 0.09194393008947373,
      "train/total_kl": 0.14282256215810776
    },
    {
      "epoch": 0.7627921667719519,
      "grad_norm": 4.15625,
      "learning_rate": 2.427905874921036e-05,
      "loss": 0.3673,
      "mean_token_accuracy": 0.9246420621871948,
      "step": 2415,
      "train/kl_loss_qwen": 0.03418837860226631,
      "train/kl_loss_r1": 0.08263415396213532,
      "train/total_kl": 0.11682253032922744
    },
    {
      "epoch": 0.7643714466203412,
      "grad_norm": 4.28125,
      "learning_rate": 2.4267214150347442e-05,
      "loss": 0.3662,
      "mean_token_accuracy": 0.9210442304611206,
      "step": 2420,
      "train/kl_loss_qwen": 0.031548118032515046,
      "train/kl_loss_r1": 0.0807625450193882,
      "train/total_kl": 0.11231066435575485
    },
    {
      "epoch": 0.7659507264687303,
      "grad_norm": 4.125,
      "learning_rate": 2.4255369551484524e-05,
      "loss": 0.4058,
      "mean_token_accuracy": 0.9152141213417053,
      "step": 2425,
      "train/kl_loss_qwen": 0.04889134652912617,
      "train/kl_loss_r1": 0.09697786271572113,
      "train/total_kl": 0.14586920738220216
    },
    {
      "epoch": 0.7675300063171194,
      "grad_norm": 43.5,
      "learning_rate": 2.4243524952621603e-05,
      "loss": 0.6557,
      "mean_token_accuracy": 0.8972956538200378,
      "step": 2430,
      "train/kl_loss_qwen": 0.11973209008574486,
      "train/kl_loss_r1": 0.22609067782759668,
      "train/total_kl": 0.3458227634429932
    },
    {
      "epoch": 0.7691092861655086,
      "grad_norm": 4.46875,
      "learning_rate": 2.4231680353758688e-05,
      "loss": 0.3766,
      "mean_token_accuracy": 0.9085039854049682,
      "step": 2435,
      "train/kl_loss_qwen": 0.024388449639081954,
      "train/kl_loss_r1": 0.05907716304063797,
      "train/total_kl": 0.08346561342477798
    },
    {
      "epoch": 0.7706885660138977,
      "grad_norm": 4.21875,
      "learning_rate": 2.4219835754895767e-05,
      "loss": 0.3604,
      "mean_token_accuracy": 0.9174930930137635,
      "step": 2440,
      "train/kl_loss_qwen": 0.03186075799167156,
      "train/kl_loss_r1": 0.06832591965794563,
      "train/total_kl": 0.10018667876720429
    },
    {
      "epoch": 0.7722678458622868,
      "grad_norm": 3.71875,
      "learning_rate": 2.4207991156032852e-05,
      "loss": 0.375,
      "mean_token_accuracy": 0.9244525671005249,
      "step": 2445,
      "train/kl_loss_qwen": 0.04879983514547348,
      "train/kl_loss_r1": 0.08889396786689759,
      "train/total_kl": 0.13769380301237105
    },
    {
      "epoch": 0.7738471257106759,
      "grad_norm": 4.84375,
      "learning_rate": 2.419614655716993e-05,
      "loss": 0.3926,
      "mean_token_accuracy": 0.9218663930892944,
      "step": 2450,
      "train/kl_loss_qwen": 0.04704993050545454,
      "train/kl_loss_r1": 0.09706018269062042,
      "train/total_kl": 0.1441101089119911
    },
    {
      "epoch": 0.7754264055590651,
      "grad_norm": 2.59375,
      "learning_rate": 2.4184301958307013e-05,
      "loss": 0.3653,
      "mean_token_accuracy": 0.9215973138809204,
      "step": 2455,
      "train/kl_loss_qwen": 0.033544443733990195,
      "train/kl_loss_r1": 0.07816535606980324,
      "train/total_kl": 0.11170979887247086
    },
    {
      "epoch": 0.7770056854074542,
      "grad_norm": 2.5,
      "learning_rate": 2.4172457359444095e-05,
      "loss": 0.3562,
      "mean_token_accuracy": 0.9256291627883911,
      "step": 2460,
      "train/kl_loss_qwen": 0.029245965555310248,
      "train/kl_loss_r1": 0.08956916779279708,
      "train/total_kl": 0.1188151329755783
    },
    {
      "epoch": 0.7785849652558433,
      "grad_norm": 3.265625,
      "learning_rate": 2.4160612760581177e-05,
      "loss": 0.3379,
      "mean_token_accuracy": 0.9237740159034729,
      "step": 2465,
      "train/kl_loss_qwen": 0.028926919773221017,
      "train/kl_loss_r1": 0.06335313022136688,
      "train/total_kl": 0.09228004813194275
    },
    {
      "epoch": 0.7801642451042324,
      "grad_norm": 3.578125,
      "learning_rate": 2.4148768161718255e-05,
      "loss": 0.366,
      "mean_token_accuracy": 0.9142597675323486,
      "step": 2470,
      "train/kl_loss_qwen": 0.030956307798624037,
      "train/kl_loss_r1": 0.07973408922553063,
      "train/total_kl": 0.11069039851427079
    },
    {
      "epoch": 0.7817435249526216,
      "grad_norm": 6.4375,
      "learning_rate": 2.413692356285534e-05,
      "loss": 0.3837,
      "mean_token_accuracy": 0.9154094099998474,
      "step": 2475,
      "train/kl_loss_qwen": 0.030161214992403984,
      "train/kl_loss_r1": 0.08193599730730057,
      "train/total_kl": 0.11209721267223358
    },
    {
      "epoch": 0.7833228048010107,
      "grad_norm": 2.921875,
      "learning_rate": 2.412507896399242e-05,
      "loss": 0.3609,
      "mean_token_accuracy": 0.9128451943397522,
      "step": 2480,
      "train/kl_loss_qwen": 0.022421348467469217,
      "train/kl_loss_r1": 0.06774807944893838,
      "train/total_kl": 0.09016942828893662
    },
    {
      "epoch": 0.7849020846493999,
      "grad_norm": 5.0,
      "learning_rate": 2.4113234365129504e-05,
      "loss": 0.4012,
      "mean_token_accuracy": 0.9245272040367126,
      "step": 2485,
      "train/kl_loss_qwen": 0.05042300075292587,
      "train/kl_loss_r1": 0.11942058354616165,
      "train/total_kl": 0.1698435813188553
    },
    {
      "epoch": 0.786481364497789,
      "grad_norm": 3.828125,
      "learning_rate": 2.4101389766266583e-05,
      "loss": 0.3745,
      "mean_token_accuracy": 0.907651174068451,
      "step": 2490,
      "train/kl_loss_qwen": 0.02721784617751837,
      "train/kl_loss_r1": 0.06651914343237877,
      "train/total_kl": 0.09373699054121971
    },
    {
      "epoch": 0.7880606443461782,
      "grad_norm": 4.28125,
      "learning_rate": 2.4089545167403665e-05,
      "loss": 0.4261,
      "mean_token_accuracy": 0.9127452611923218,
      "step": 2495,
      "train/kl_loss_qwen": 0.03653549421578646,
      "train/kl_loss_r1": 0.09398385062813759,
      "train/total_kl": 0.13051934391260148
    },
    {
      "epoch": 0.7896399241945673,
      "grad_norm": 6.65625,
      "learning_rate": 2.4077700568540747e-05,
      "loss": 0.4359,
      "mean_token_accuracy": 0.9105828523635864,
      "step": 2500,
      "train/kl_loss_qwen": 0.04454654008150101,
      "train/kl_loss_r1": 0.10583894699811935,
      "train/total_kl": 0.15038548409938812
    },
    {
      "epoch": 0.7912192040429564,
      "grad_norm": 5.09375,
      "learning_rate": 2.406585596967783e-05,
      "loss": 0.3629,
      "mean_token_accuracy": 0.9238234043121338,
      "step": 2505,
      "train/kl_loss_qwen": 0.03996905367821455,
      "train/kl_loss_r1": 0.089544016122818,
      "train/total_kl": 0.12951307296752929
    },
    {
      "epoch": 0.7927984838913456,
      "grad_norm": 3.984375,
      "learning_rate": 2.4054011370814907e-05,
      "loss": 0.3951,
      "mean_token_accuracy": 0.9125975847244263,
      "step": 2510,
      "train/kl_loss_qwen": 0.03943205922842026,
      "train/kl_loss_r1": 0.08176129311323166,
      "train/total_kl": 0.12119335383176803
    },
    {
      "epoch": 0.7943777637397347,
      "grad_norm": 3.5625,
      "learning_rate": 2.404216677195199e-05,
      "loss": 0.3373,
      "mean_token_accuracy": 0.9206342339515686,
      "step": 2515,
      "train/kl_loss_qwen": 0.02708890587091446,
      "train/kl_loss_r1": 0.06473625749349594,
      "train/total_kl": 0.0918251633644104
    },
    {
      "epoch": 0.7959570435881238,
      "grad_norm": 3.765625,
      "learning_rate": 2.403032217308907e-05,
      "loss": 0.3511,
      "mean_token_accuracy": 0.9138602972030639,
      "step": 2520,
      "train/kl_loss_qwen": 0.026088026724755763,
      "train/kl_loss_r1": 0.06657440215349197,
      "train/total_kl": 0.09266242831945419
    },
    {
      "epoch": 0.7975363234365129,
      "grad_norm": 3.015625,
      "learning_rate": 2.4018477574226153e-05,
      "loss": 0.3375,
      "mean_token_accuracy": 0.9305527448654175,
      "step": 2525,
      "train/kl_loss_qwen": 0.03410542272031307,
      "train/kl_loss_r1": 0.07411143854260445,
      "train/total_kl": 0.10821686089038848
    },
    {
      "epoch": 0.799115603284902,
      "grad_norm": 2.53125,
      "learning_rate": 2.4006632975363235e-05,
      "loss": 0.3421,
      "mean_token_accuracy": 0.9249903798103333,
      "step": 2530,
      "train/kl_loss_qwen": 0.040260189399123195,
      "train/kl_loss_r1": 0.08297931253910065,
      "train/total_kl": 0.12323950231075287
    },
    {
      "epoch": 0.8006948831332912,
      "grad_norm": 3.453125,
      "learning_rate": 2.3994788376500314e-05,
      "loss": 0.3402,
      "mean_token_accuracy": 0.9303278446197509,
      "step": 2535,
      "train/kl_loss_qwen": 0.03351224288344383,
      "train/kl_loss_r1": 0.07336185872554779,
      "train/total_kl": 0.10687410235404968
    },
    {
      "epoch": 0.8022741629816803,
      "grad_norm": 4.125,
      "learning_rate": 2.39829437776374e-05,
      "loss": 0.3612,
      "mean_token_accuracy": 0.9318838477134704,
      "step": 2540,
      "train/kl_loss_qwen": 0.04090795181691646,
      "train/kl_loss_r1": 0.10184711813926697,
      "train/total_kl": 0.14275507032871246
    },
    {
      "epoch": 0.8038534428300695,
      "grad_norm": 2.796875,
      "learning_rate": 2.3971099178774478e-05,
      "loss": 0.3291,
      "mean_token_accuracy": 0.9273692488670349,
      "step": 2545,
      "train/kl_loss_qwen": 0.0339467030018568,
      "train/kl_loss_r1": 0.07381512969732285,
      "train/total_kl": 0.10776183307170868
    },
    {
      "epoch": 0.8054327226784587,
      "grad_norm": 5.5,
      "learning_rate": 2.395925457991156e-05,
      "loss": 0.3255,
      "mean_token_accuracy": 0.9312151670455933,
      "step": 2550,
      "train/kl_loss_qwen": 0.03405954949557781,
      "train/kl_loss_r1": 0.08676501587033272,
      "train/total_kl": 0.1208245649933815
    },
    {
      "epoch": 0.8070120025268478,
      "grad_norm": 3.265625,
      "learning_rate": 2.3947409981048642e-05,
      "loss": 0.3892,
      "mean_token_accuracy": 0.9220807313919067,
      "step": 2555,
      "train/kl_loss_qwen": 0.050804464519023894,
      "train/kl_loss_r1": 0.08766236901283264,
      "train/total_kl": 0.13846683651208877
    },
    {
      "epoch": 0.8085912823752369,
      "grad_norm": 3.859375,
      "learning_rate": 2.3935565382185724e-05,
      "loss": 0.4009,
      "mean_token_accuracy": 0.9117589831352234,
      "step": 2560,
      "train/kl_loss_qwen": 0.028337689489126204,
      "train/kl_loss_r1": 0.08602797985076904,
      "train/total_kl": 0.11436567157506942
    },
    {
      "epoch": 0.810170562223626,
      "grad_norm": 6.3125,
      "learning_rate": 2.3923720783322806e-05,
      "loss": 0.3888,
      "mean_token_accuracy": 0.9216238379478454,
      "step": 2565,
      "train/kl_loss_qwen": 0.05098458006978035,
      "train/kl_loss_r1": 0.08744649887084961,
      "train/total_kl": 0.13843107745051383
    },
    {
      "epoch": 0.8117498420720152,
      "grad_norm": 3.390625,
      "learning_rate": 2.3911876184459888e-05,
      "loss": 0.7176,
      "mean_token_accuracy": 0.9014476418495179,
      "step": 2570,
      "train/kl_loss_qwen": 0.13389752954244613,
      "train/kl_loss_r1": 0.3017698168754578,
      "train/total_kl": 0.4356673493981361
    },
    {
      "epoch": 0.8133291219204043,
      "grad_norm": 3.140625,
      "learning_rate": 2.3900031585596966e-05,
      "loss": 0.4306,
      "mean_token_accuracy": 0.9086627244949341,
      "step": 2575,
      "train/kl_loss_qwen": 0.04608270525932312,
      "train/kl_loss_r1": 0.09769520461559296,
      "train/total_kl": 0.14377790987491607
    },
    {
      "epoch": 0.8149084017687934,
      "grad_norm": 9.875,
      "learning_rate": 2.3888186986734052e-05,
      "loss": 0.4158,
      "mean_token_accuracy": 0.9266731977462769,
      "step": 2580,
      "train/kl_loss_qwen": 0.06154814288020134,
      "train/kl_loss_r1": 0.11556403711438179,
      "train/total_kl": 0.17711218297481537
    },
    {
      "epoch": 0.8164876816171825,
      "grad_norm": 5.0,
      "learning_rate": 2.387634238787113e-05,
      "loss": 0.388,
      "mean_token_accuracy": 0.9269638419151306,
      "step": 2585,
      "train/kl_loss_qwen": 0.05261476784944534,
      "train/kl_loss_r1": 0.10991271138191223,
      "train/total_kl": 0.1625274747610092
    },
    {
      "epoch": 0.8180669614655717,
      "grad_norm": 3.125,
      "learning_rate": 2.3864497789008212e-05,
      "loss": 0.3337,
      "mean_token_accuracy": 0.930491828918457,
      "step": 2590,
      "train/kl_loss_qwen": 0.03801763094961643,
      "train/kl_loss_r1": 0.06708204299211502,
      "train/total_kl": 0.10509967654943467
    },
    {
      "epoch": 0.8196462413139608,
      "grad_norm": 4.59375,
      "learning_rate": 2.3852653190145294e-05,
      "loss": 0.3874,
      "mean_token_accuracy": 0.9203120946884156,
      "step": 2595,
      "train/kl_loss_qwen": 0.05236174575984478,
      "train/kl_loss_r1": 0.09800608456134796,
      "train/total_kl": 0.15036782771348953
    },
    {
      "epoch": 0.8212255211623499,
      "grad_norm": 5.03125,
      "learning_rate": 2.3840808591282376e-05,
      "loss": 0.4068,
      "mean_token_accuracy": 0.9312162876129151,
      "step": 2600,
      "train/kl_loss_qwen": 0.06960356421768665,
      "train/kl_loss_r1": 0.13111656457185744,
      "train/total_kl": 0.20072013139724731
    },
    {
      "epoch": 0.822804801010739,
      "grad_norm": 3.59375,
      "learning_rate": 2.3828963992419458e-05,
      "loss": 0.4062,
      "mean_token_accuracy": 0.9156930804252624,
      "step": 2605,
      "train/kl_loss_qwen": 0.04320799596607685,
      "train/kl_loss_r1": 0.08694410771131515,
      "train/total_kl": 0.13015210628509521
    },
    {
      "epoch": 0.8243840808591283,
      "grad_norm": 3.375,
      "learning_rate": 2.381711939355654e-05,
      "loss": 0.3447,
      "mean_token_accuracy": 0.934181022644043,
      "step": 2610,
      "train/kl_loss_qwen": 0.04134991504251957,
      "train/kl_loss_r1": 0.10073989480733872,
      "train/total_kl": 0.14208980798721313
    },
    {
      "epoch": 0.8259633607075174,
      "grad_norm": 4.34375,
      "learning_rate": 2.380527479469362e-05,
      "loss": 0.3457,
      "mean_token_accuracy": 0.9206912398338318,
      "step": 2615,
      "train/kl_loss_qwen": 0.030087007954716683,
      "train/kl_loss_r1": 0.07685746923089028,
      "train/total_kl": 0.10694447755813599
    },
    {
      "epoch": 0.8275426405559065,
      "grad_norm": 3.125,
      "learning_rate": 2.3793430195830704e-05,
      "loss": 0.339,
      "mean_token_accuracy": 0.9186580896377563,
      "step": 2620,
      "train/kl_loss_qwen": 0.023380796983838083,
      "train/kl_loss_r1": 0.05962579920887947,
      "train/total_kl": 0.08300659656524659
    },
    {
      "epoch": 0.8291219204042957,
      "grad_norm": 3.625,
      "learning_rate": 2.3781585596967783e-05,
      "loss": 0.6826,
      "mean_token_accuracy": 0.8936365127563477,
      "step": 2625,
      "train/kl_loss_qwen": 0.12251255549490452,
      "train/kl_loss_r1": 0.21155278161168098,
      "train/total_kl": 0.334065343439579
    },
    {
      "epoch": 0.8307012002526848,
      "grad_norm": 4.3125,
      "learning_rate": 2.3769740998104865e-05,
      "loss": 0.3779,
      "mean_token_accuracy": 0.9149964809417724,
      "step": 2630,
      "train/kl_loss_qwen": 0.02568915206938982,
      "train/kl_loss_r1": 0.06890519857406616,
      "train/total_kl": 0.09459435045719147
    },
    {
      "epoch": 0.8322804801010739,
      "grad_norm": 3.734375,
      "learning_rate": 2.3757896399241947e-05,
      "loss": 0.3971,
      "mean_token_accuracy": 0.920639717578888,
      "step": 2635,
      "train/kl_loss_qwen": 0.04088805615901947,
      "train/kl_loss_r1": 0.1108312577009201,
      "train/total_kl": 0.15171931684017181
    },
    {
      "epoch": 0.833859759949463,
      "grad_norm": 4.03125,
      "learning_rate": 2.3746051800379025e-05,
      "loss": 0.3706,
      "mean_token_accuracy": 0.915614914894104,
      "step": 2640,
      "train/kl_loss_qwen": 0.03602025508880615,
      "train/kl_loss_r1": 0.08051187694072723,
      "train/total_kl": 0.11653213053941727
    },
    {
      "epoch": 0.8354390397978522,
      "grad_norm": 2.984375,
      "learning_rate": 2.373420720151611e-05,
      "loss": 0.3486,
      "mean_token_accuracy": 0.923656964302063,
      "step": 2645,
      "train/kl_loss_qwen": 0.02962579093873501,
      "train/kl_loss_r1": 0.06862581074237824,
      "train/total_kl": 0.09825160205364228
    },
    {
      "epoch": 0.8370183196462413,
      "grad_norm": 2.984375,
      "learning_rate": 2.372236260265319e-05,
      "loss": 0.3931,
      "mean_token_accuracy": 0.9124223470687867,
      "step": 2650,
      "train/kl_loss_qwen": 0.03956179711967707,
      "train/kl_loss_r1": 0.07831117734313012,
      "train/total_kl": 0.11787297427654267
    },
    {
      "epoch": 0.8385975994946304,
      "grad_norm": 3.703125,
      "learning_rate": 2.371051800379027e-05,
      "loss": 0.3588,
      "mean_token_accuracy": 0.9173622846603393,
      "step": 2655,
      "train/kl_loss_qwen": 0.02660207264125347,
      "train/kl_loss_r1": 0.07131514698266983,
      "train/total_kl": 0.09791721850633621
    },
    {
      "epoch": 0.8401768793430195,
      "grad_norm": 3.8125,
      "learning_rate": 2.3698673404927353e-05,
      "loss": 0.3567,
      "mean_token_accuracy": 0.9339551448822021,
      "step": 2660,
      "train/kl_loss_qwen": 0.0498113602399826,
      "train/kl_loss_r1": 0.09188093692064285,
      "train/total_kl": 0.1416922926902771
    },
    {
      "epoch": 0.8417561591914087,
      "grad_norm": 3.921875,
      "learning_rate": 2.3686828806064435e-05,
      "loss": 0.4656,
      "mean_token_accuracy": 0.9000948309898377,
      "step": 2665,
      "train/kl_loss_qwen": 0.04961310550570488,
      "train/kl_loss_r1": 0.1033561035990715,
      "train/total_kl": 0.15296921133995056
    },
    {
      "epoch": 0.8433354390397979,
      "grad_norm": 5.5625,
      "learning_rate": 2.3674984207201517e-05,
      "loss": 0.454,
      "mean_token_accuracy": 0.9090999484062194,
      "step": 2670,
      "train/kl_loss_qwen": 0.04370644688606262,
      "train/kl_loss_r1": 0.1051908329129219,
      "train/total_kl": 0.14889727979898454
    },
    {
      "epoch": 0.844914718888187,
      "grad_norm": 3.859375,
      "learning_rate": 2.36631396083386e-05,
      "loss": 0.3432,
      "mean_token_accuracy": 0.9163386821746826,
      "step": 2675,
      "train/kl_loss_qwen": 0.028805825859308243,
      "train/kl_loss_r1": 0.06775637790560722,
      "train/total_kl": 0.0965622030198574
    },
    {
      "epoch": 0.8464939987365762,
      "grad_norm": 4.53125,
      "learning_rate": 2.3651295009475678e-05,
      "loss": 0.3386,
      "mean_token_accuracy": 0.9309346675872803,
      "step": 2680,
      "train/kl_loss_qwen": 0.03741238489747047,
      "train/kl_loss_r1": 0.06686489507555962,
      "train/total_kl": 0.10427727997303009
    },
    {
      "epoch": 0.8480732785849653,
      "grad_norm": 18.25,
      "learning_rate": 2.3639450410612763e-05,
      "loss": 0.6845,
      "mean_token_accuracy": 0.9002939343452454,
      "step": 2685,
      "train/kl_loss_qwen": 0.13364123348146678,
      "train/kl_loss_r1": 0.24209992811083794,
      "train/total_kl": 0.3757411614060402
    },
    {
      "epoch": 0.8496525584333544,
      "grad_norm": 3.859375,
      "learning_rate": 2.362760581174984e-05,
      "loss": 0.3094,
      "mean_token_accuracy": 0.9339204430580139,
      "step": 2690,
      "train/kl_loss_qwen": 0.033999216929078105,
      "train/kl_loss_r1": 0.07654135674238205,
      "train/total_kl": 0.11054057478904725
    },
    {
      "epoch": 0.8512318382817435,
      "grad_norm": 2.609375,
      "learning_rate": 2.3615761212886924e-05,
      "loss": 0.3601,
      "mean_token_accuracy": 0.9307232856750488,
      "step": 2695,
      "train/kl_loss_qwen": 0.045728157833218576,
      "train/kl_loss_r1": 0.1051046222448349,
      "train/total_kl": 0.15083277821540833
    },
    {
      "epoch": 0.8528111181301327,
      "grad_norm": 3.4375,
      "learning_rate": 2.3603916614024006e-05,
      "loss": 0.3747,
      "mean_token_accuracy": 0.9224868416786194,
      "step": 2700,
      "train/kl_loss_qwen": 0.047057855501770975,
      "train/kl_loss_r1": 0.0920887365937233,
      "train/total_kl": 0.13914659172296523
    },
    {
      "epoch": 0.8543903979785218,
      "grad_norm": 4.375,
      "learning_rate": 2.3592072015161088e-05,
      "loss": 0.4153,
      "mean_token_accuracy": 0.9222820281982422,
      "step": 2705,
      "train/kl_loss_qwen": 0.04931856468319893,
      "train/kl_loss_r1": 0.11694778800010681,
      "train/total_kl": 0.16626635193824768
    },
    {
      "epoch": 0.8559696778269109,
      "grad_norm": 2.3125,
      "learning_rate": 2.358022741629817e-05,
      "loss": 0.3311,
      "mean_token_accuracy": 0.9287552952766418,
      "step": 2710,
      "train/kl_loss_qwen": 0.030334292724728585,
      "train/kl_loss_r1": 0.07026510387659073,
      "train/total_kl": 0.10059939622879029
    },
    {
      "epoch": 0.8575489576753,
      "grad_norm": 3.53125,
      "learning_rate": 2.356838281743525e-05,
      "loss": 0.3332,
      "mean_token_accuracy": 0.9307746648788452,
      "step": 2715,
      "train/kl_loss_qwen": 0.044843487441539764,
      "train/kl_loss_r1": 0.0759985588490963,
      "train/total_kl": 0.12084204852581024
    },
    {
      "epoch": 0.8591282375236892,
      "grad_norm": 3.78125,
      "learning_rate": 2.355653821857233e-05,
      "loss": 0.3934,
      "mean_token_accuracy": 0.9217597126960755,
      "step": 2720,
      "train/kl_loss_qwen": 0.03866707682609558,
      "train/kl_loss_r1": 0.0991225466132164,
      "train/total_kl": 0.13778962641954423
    },
    {
      "epoch": 0.8607075173720783,
      "grad_norm": 3.234375,
      "learning_rate": 2.3544693619709415e-05,
      "loss": 0.3434,
      "mean_token_accuracy": 0.9321288108825684,
      "step": 2725,
      "train/kl_loss_qwen": 0.039694737270474435,
      "train/kl_loss_r1": 0.09530275389552116,
      "train/total_kl": 0.1349974900484085
    },
    {
      "epoch": 0.8622867972204674,
      "grad_norm": 3.171875,
      "learning_rate": 2.3532849020846494e-05,
      "loss": 0.4235,
      "mean_token_accuracy": 0.9127881765365601,
      "step": 2730,
      "train/kl_loss_qwen": 0.04991408437490463,
      "train/kl_loss_r1": 0.10121307522058487,
      "train/total_kl": 0.15112715661525727
    },
    {
      "epoch": 0.8638660770688567,
      "grad_norm": 4.59375,
      "learning_rate": 2.3521004421983576e-05,
      "loss": 0.4212,
      "mean_token_accuracy": 0.9152961373329163,
      "step": 2735,
      "train/kl_loss_qwen": 0.04270019568502903,
      "train/kl_loss_r1": 0.09133504554629326,
      "train/total_kl": 0.1340352401137352
    },
    {
      "epoch": 0.8654453569172458,
      "grad_norm": 2.953125,
      "learning_rate": 2.3509159823120658e-05,
      "loss": 0.4234,
      "mean_token_accuracy": 0.9068126082420349,
      "step": 2740,
      "train/kl_loss_qwen": 0.035153744742274284,
      "train/kl_loss_r1": 0.07006339207291604,
      "train/total_kl": 0.10521713942289353
    },
    {
      "epoch": 0.8670246367656349,
      "grad_norm": 3.25,
      "learning_rate": 2.349731522425774e-05,
      "loss": 0.3168,
      "mean_token_accuracy": 0.9277806520462036,
      "step": 2745,
      "train/kl_loss_qwen": 0.028336655348539352,
      "train/kl_loss_r1": 0.07033251821994782,
      "train/total_kl": 0.09866917431354523
    },
    {
      "epoch": 0.868603916614024,
      "grad_norm": 5.5,
      "learning_rate": 2.3485470625394822e-05,
      "loss": 0.4091,
      "mean_token_accuracy": 0.9073657393455505,
      "step": 2750,
      "train/kl_loss_qwen": 0.043717500567436215,
      "train/kl_loss_r1": 0.09493431299924851,
      "train/total_kl": 0.1386518120765686
    },
    {
      "epoch": 0.8701831964624132,
      "grad_norm": 4.5,
      "learning_rate": 2.3473626026531904e-05,
      "loss": 0.3881,
      "mean_token_accuracy": 0.9264297723770142,
      "step": 2755,
      "train/kl_loss_qwen": 0.04793095961213112,
      "train/kl_loss_r1": 0.10894708484411239,
      "train/total_kl": 0.15687804371118547
    },
    {
      "epoch": 0.8717624763108023,
      "grad_norm": 4.625,
      "learning_rate": 2.3461781427668982e-05,
      "loss": 0.3219,
      "mean_token_accuracy": 0.9308045506477356,
      "step": 2760,
      "train/kl_loss_qwen": 0.029331354796886443,
      "train/kl_loss_r1": 0.07481529638171196,
      "train/total_kl": 0.10414665043354035
    },
    {
      "epoch": 0.8733417561591914,
      "grad_norm": 4.28125,
      "learning_rate": 2.3449936828806064e-05,
      "loss": 0.3591,
      "mean_token_accuracy": 0.9224736332893372,
      "step": 2765,
      "train/kl_loss_qwen": 0.038478368893265724,
      "train/kl_loss_r1": 0.0831017717719078,
      "train/total_kl": 0.1215801402926445
    },
    {
      "epoch": 0.8749210360075805,
      "grad_norm": 4.875,
      "learning_rate": 2.3438092229943146e-05,
      "loss": 0.3553,
      "mean_token_accuracy": 0.9307270169258117,
      "step": 2770,
      "train/kl_loss_qwen": 0.04619881398975849,
      "train/kl_loss_r1": 0.09483064562082291,
      "train/total_kl": 0.1410294622182846
    },
    {
      "epoch": 0.8765003158559697,
      "grad_norm": 3.3125,
      "learning_rate": 2.3426247631080225e-05,
      "loss": 0.3082,
      "mean_token_accuracy": 0.9232604622840881,
      "step": 2775,
      "train/kl_loss_qwen": 0.022410491853952406,
      "train/kl_loss_r1": 0.05533663854002953,
      "train/total_kl": 0.07774712964892387
    },
    {
      "epoch": 0.8780795957043588,
      "grad_norm": 4.09375,
      "learning_rate": 2.341440303221731e-05,
      "loss": 0.3484,
      "mean_token_accuracy": 0.9240765810012818,
      "step": 2780,
      "train/kl_loss_qwen": 0.02607548125088215,
      "train/kl_loss_r1": 0.07135205045342445,
      "train/total_kl": 0.09742753207683563
    },
    {
      "epoch": 0.8796588755527479,
      "grad_norm": 4.15625,
      "learning_rate": 2.340255843335439e-05,
      "loss": 0.4301,
      "mean_token_accuracy": 0.9141624212265015,
      "step": 2785,
      "train/kl_loss_qwen": 0.03876824863255024,
      "train/kl_loss_r1": 0.11168951839208603,
      "train/total_kl": 0.15045776665210725
    },
    {
      "epoch": 0.881238155401137,
      "grad_norm": 2.578125,
      "learning_rate": 2.3390713834491474e-05,
      "loss": 0.3888,
      "mean_token_accuracy": 0.9160457015037536,
      "step": 2790,
      "train/kl_loss_qwen": 0.03299585022032261,
      "train/kl_loss_r1": 0.08398881256580353,
      "train/total_kl": 0.1169846624135971
    },
    {
      "epoch": 0.8828174352495262,
      "grad_norm": 3.703125,
      "learning_rate": 2.3378869235628553e-05,
      "loss": 0.3576,
      "mean_token_accuracy": 0.9229716420173645,
      "step": 2795,
      "train/kl_loss_qwen": 0.037774191796779634,
      "train/kl_loss_r1": 0.08081903904676438,
      "train/total_kl": 0.11859322786331176
    },
    {
      "epoch": 0.8843967150979154,
      "grad_norm": 3.5,
      "learning_rate": 2.3367024636765635e-05,
      "loss": 0.3163,
      "mean_token_accuracy": 0.9317458033561706,
      "step": 2800,
      "train/kl_loss_qwen": 0.02584424801170826,
      "train/kl_loss_r1": 0.07160683199763299,
      "train/total_kl": 0.0974510818719864
    },
    {
      "epoch": 0.8859759949463045,
      "grad_norm": 4.9375,
      "learning_rate": 2.3355180037902717e-05,
      "loss": 0.3686,
      "mean_token_accuracy": 0.9181830644607544,
      "step": 2805,
      "train/kl_loss_qwen": 0.03634273037314415,
      "train/kl_loss_r1": 0.07914906591176987,
      "train/total_kl": 0.1154917985200882
    },
    {
      "epoch": 0.8875552747946936,
      "grad_norm": 5.53125,
      "learning_rate": 2.33433354390398e-05,
      "loss": 0.4767,
      "mean_token_accuracy": 0.8992764234542847,
      "step": 2810,
      "train/kl_loss_qwen": 0.045882731303572656,
      "train/kl_loss_r1": 0.09739400148391723,
      "train/total_kl": 0.14327673465013505
    },
    {
      "epoch": 0.8891345546430828,
      "grad_norm": 5.8125,
      "learning_rate": 2.3331490840176877e-05,
      "loss": 0.3317,
      "mean_token_accuracy": 0.9269258737564087,
      "step": 2815,
      "train/kl_loss_qwen": 0.0265196792781353,
      "train/kl_loss_r1": 0.07587031498551369,
      "train/total_kl": 0.10238999351859093
    },
    {
      "epoch": 0.8907138344914719,
      "grad_norm": 3.40625,
      "learning_rate": 2.3319646241313963e-05,
      "loss": 0.3306,
      "mean_token_accuracy": 0.926283848285675,
      "step": 2820,
      "train/kl_loss_qwen": 0.0317132730036974,
      "train/kl_loss_r1": 0.07335674092173576,
      "train/total_kl": 0.10507001578807831
    },
    {
      "epoch": 0.892293114339861,
      "grad_norm": 4.25,
      "learning_rate": 2.330780164245104e-05,
      "loss": 0.3311,
      "mean_token_accuracy": 0.9329625844955445,
      "step": 2825,
      "train/kl_loss_qwen": 0.04199698902666569,
      "train/kl_loss_r1": 0.07758924663066864,
      "train/total_kl": 0.11958623826503753
    },
    {
      "epoch": 0.8938723941882502,
      "grad_norm": 4.5625,
      "learning_rate": 2.3295957043588127e-05,
      "loss": 0.3686,
      "mean_token_accuracy": 0.9228216290473938,
      "step": 2830,
      "train/kl_loss_qwen": 0.031190813332796097,
      "train/kl_loss_r1": 0.07843297198414803,
      "train/total_kl": 0.10962378531694413
    },
    {
      "epoch": 0.8954516740366393,
      "grad_norm": 4.15625,
      "learning_rate": 2.3284112444725205e-05,
      "loss": 0.4179,
      "mean_token_accuracy": 0.9180848121643066,
      "step": 2835,
      "train/kl_loss_qwen": 0.047972666844725606,
      "train/kl_loss_r1": 0.1188799262046814,
      "train/total_kl": 0.16685259193181992
    },
    {
      "epoch": 0.8970309538850284,
      "grad_norm": 3.453125,
      "learning_rate": 2.3272267845862287e-05,
      "loss": 0.4663,
      "mean_token_accuracy": 0.9003514170646667,
      "step": 2840,
      "train/kl_loss_qwen": 0.03944678194820881,
      "train/kl_loss_r1": 0.0763892650604248,
      "train/total_kl": 0.1158360481262207
    },
    {
      "epoch": 0.8986102337334175,
      "grad_norm": 3.578125,
      "learning_rate": 2.326042324699937e-05,
      "loss": 0.3644,
      "mean_token_accuracy": 0.9175791263580322,
      "step": 2845,
      "train/kl_loss_qwen": 0.032439626008272174,
      "train/kl_loss_r1": 0.08110484704375268,
      "train/total_kl": 0.11354447007179261
    },
    {
      "epoch": 0.9001895135818067,
      "grad_norm": 2.71875,
      "learning_rate": 2.324857864813645e-05,
      "loss": 0.3703,
      "mean_token_accuracy": 0.9196029543876648,
      "step": 2850,
      "train/kl_loss_qwen": 0.03039756678044796,
      "train/kl_loss_r1": 0.06552705615758896,
      "train/total_kl": 0.09592462331056595
    },
    {
      "epoch": 0.9017687934301958,
      "grad_norm": 3.625,
      "learning_rate": 2.323673404927353e-05,
      "loss": 0.4119,
      "mean_token_accuracy": 0.9142733454704285,
      "step": 2855,
      "train/kl_loss_qwen": 0.0521426884457469,
      "train/kl_loss_r1": 0.102823256701231,
      "train/total_kl": 0.15496594905853273
    },
    {
      "epoch": 0.903348073278585,
      "grad_norm": 2.96875,
      "learning_rate": 2.3224889450410615e-05,
      "loss": 0.3838,
      "mean_token_accuracy": 0.921067225933075,
      "step": 2860,
      "train/kl_loss_qwen": 0.03301718980073929,
      "train/kl_loss_r1": 0.08988576382398605,
      "train/total_kl": 0.12290295213460922
    },
    {
      "epoch": 0.9049273531269741,
      "grad_norm": 3.046875,
      "learning_rate": 2.3213044851547694e-05,
      "loss": 0.315,
      "mean_token_accuracy": 0.9292155742645264,
      "step": 2865,
      "train/kl_loss_qwen": 0.025115855410695075,
      "train/kl_loss_r1": 0.06775750741362571,
      "train/total_kl": 0.09287336319684983
    },
    {
      "epoch": 0.9065066329753633,
      "grad_norm": 4.125,
      "learning_rate": 2.320120025268478e-05,
      "loss": 0.3771,
      "mean_token_accuracy": 0.9185895442962646,
      "step": 2870,
      "train/kl_loss_qwen": 0.04342378079891205,
      "train/kl_loss_r1": 0.08425913602113724,
      "train/total_kl": 0.1276829183101654
    },
    {
      "epoch": 0.9080859128237524,
      "grad_norm": 3.625,
      "learning_rate": 2.3189355653821858e-05,
      "loss": 0.3464,
      "mean_token_accuracy": 0.9226277828216553,
      "step": 2875,
      "train/kl_loss_qwen": 0.0352512426674366,
      "train/kl_loss_r1": 0.08377473652362824,
      "train/total_kl": 0.11902597993612289
    },
    {
      "epoch": 0.9096651926721415,
      "grad_norm": 3.3125,
      "learning_rate": 2.317751105495894e-05,
      "loss": 0.3193,
      "mean_token_accuracy": 0.9268807768821716,
      "step": 2880,
      "train/kl_loss_qwen": 0.02979418933391571,
      "train/kl_loss_r1": 0.07347398474812508,
      "train/total_kl": 0.10326817333698272
    },
    {
      "epoch": 0.9112444725205306,
      "grad_norm": 3.671875,
      "learning_rate": 2.316566645609602e-05,
      "loss": 0.3521,
      "mean_token_accuracy": 0.9216803193092347,
      "step": 2885,
      "train/kl_loss_qwen": 0.04764210470020771,
      "train/kl_loss_r1": 0.07351809293031693,
      "train/total_kl": 0.12116020023822785
    },
    {
      "epoch": 0.9128237523689198,
      "grad_norm": 4.0625,
      "learning_rate": 2.31538218572331e-05,
      "loss": 0.3592,
      "mean_token_accuracy": 0.9134865045547486,
      "step": 2890,
      "train/kl_loss_qwen": 0.029439922794699667,
      "train/kl_loss_r1": 0.06461982205510139,
      "train/total_kl": 0.09405974447727203
    },
    {
      "epoch": 0.9144030322173089,
      "grad_norm": 3.75,
      "learning_rate": 2.3141977258370182e-05,
      "loss": 0.3832,
      "mean_token_accuracy": 0.9185373544692993,
      "step": 2895,
      "train/kl_loss_qwen": 0.0418958630412817,
      "train/kl_loss_r1": 0.09269206523895264,
      "train/total_kl": 0.13458792716264725
    },
    {
      "epoch": 0.915982312065698,
      "grad_norm": 4.09375,
      "learning_rate": 2.3130132659507264e-05,
      "loss": 0.3702,
      "mean_token_accuracy": 0.9150678277015686,
      "step": 2900,
      "train/kl_loss_qwen": 0.03484309688210487,
      "train/kl_loss_r1": 0.074680595099926,
      "train/total_kl": 0.10952369123697281
    },
    {
      "epoch": 0.9175615919140871,
      "grad_norm": 2.734375,
      "learning_rate": 2.3118288060644346e-05,
      "loss": 0.3137,
      "mean_token_accuracy": 0.9410058617591858,
      "step": 2905,
      "train/kl_loss_qwen": 0.04197153449058533,
      "train/kl_loss_r1": 0.07785490900278091,
      "train/total_kl": 0.11982644349336624
    },
    {
      "epoch": 0.9191408717624763,
      "grad_norm": 2.75,
      "learning_rate": 2.3106443461781428e-05,
      "loss": 0.3623,
      "mean_token_accuracy": 0.9272714614868164,
      "step": 2910,
      "train/kl_loss_qwen": 0.040742290765047075,
      "train/kl_loss_r1": 0.09093925058841705,
      "train/total_kl": 0.13168154060840606
    },
    {
      "epoch": 0.9207201516108654,
      "grad_norm": 2.90625,
      "learning_rate": 2.309459886291851e-05,
      "loss": 0.3929,
      "mean_token_accuracy": 0.9142226219177246,
      "step": 2915,
      "train/kl_loss_qwen": 0.033158091455698015,
      "train/kl_loss_r1": 0.07759033590555191,
      "train/total_kl": 0.11074842512607574
    },
    {
      "epoch": 0.9222994314592545,
      "grad_norm": 3.328125,
      "learning_rate": 2.308275426405559e-05,
      "loss": 0.3376,
      "mean_token_accuracy": 0.9245054602622986,
      "step": 2920,
      "train/kl_loss_qwen": 0.032250311225652695,
      "train/kl_loss_r1": 0.07877109795808793,
      "train/total_kl": 0.11102140694856644
    },
    {
      "epoch": 0.9238787113076438,
      "grad_norm": 2.796875,
      "learning_rate": 2.3070909665192674e-05,
      "loss": 0.3918,
      "mean_token_accuracy": 0.918913733959198,
      "step": 2925,
      "train/kl_loss_qwen": 0.04415415301918983,
      "train/kl_loss_r1": 0.07916211858391761,
      "train/total_kl": 0.12331627160310746
    },
    {
      "epoch": 0.9254579911560329,
      "grad_norm": 2.71875,
      "learning_rate": 2.3059065066329753e-05,
      "loss": 0.3549,
      "mean_token_accuracy": 0.9230927944183349,
      "step": 2930,
      "train/kl_loss_qwen": 0.037778588384389876,
      "train/kl_loss_r1": 0.08441569656133652,
      "train/total_kl": 0.12219428271055222
    },
    {
      "epoch": 0.927037271004422,
      "grad_norm": 3.15625,
      "learning_rate": 2.3047220467466838e-05,
      "loss": 0.3187,
      "mean_token_accuracy": 0.9284777283668518,
      "step": 2935,
      "train/kl_loss_qwen": 0.035384738072752955,
      "train/kl_loss_r1": 0.07121852263808251,
      "train/total_kl": 0.10660326033830643
    },
    {
      "epoch": 0.9286165508528111,
      "grad_norm": 6.21875,
      "learning_rate": 2.3035375868603917e-05,
      "loss": 0.407,
      "mean_token_accuracy": 0.9145089983940125,
      "step": 2940,
      "train/kl_loss_qwen": 0.04715360701084137,
      "train/kl_loss_r1": 0.09332031607627869,
      "train/total_kl": 0.14047392457723618
    },
    {
      "epoch": 0.9301958307012003,
      "grad_norm": 2.984375,
      "learning_rate": 2.3023531269741e-05,
      "loss": 0.3454,
      "mean_token_accuracy": 0.9264336228370667,
      "step": 2945,
      "train/kl_loss_qwen": 0.04277587085962296,
      "train/kl_loss_r1": 0.08272863551974297,
      "train/total_kl": 0.125504507124424
    },
    {
      "epoch": 0.9317751105495894,
      "grad_norm": 3.109375,
      "learning_rate": 2.301168667087808e-05,
      "loss": 0.3386,
      "mean_token_accuracy": 0.9281829118728637,
      "step": 2950,
      "train/kl_loss_qwen": 0.04313252680003643,
      "train/kl_loss_r1": 0.07347260043025017,
      "train/total_kl": 0.11660512685775756
    },
    {
      "epoch": 0.9333543903979785,
      "grad_norm": 4.4375,
      "learning_rate": 2.2999842072015162e-05,
      "loss": 0.3015,
      "mean_token_accuracy": 0.9431676268577576,
      "step": 2955,
      "train/kl_loss_qwen": 0.03480535484850407,
      "train/kl_loss_r1": 0.08230780586600303,
      "train/total_kl": 0.11711316257715225
    },
    {
      "epoch": 0.9349336702463676,
      "grad_norm": 4.78125,
      "learning_rate": 2.298799747315224e-05,
      "loss": 0.4318,
      "mean_token_accuracy": 0.9192673206329346,
      "step": 2960,
      "train/kl_loss_qwen": 0.06733840852975845,
      "train/kl_loss_r1": 0.11850000619888305,
      "train/total_kl": 0.1858384132385254
    },
    {
      "epoch": 0.9365129500947568,
      "grad_norm": 3.140625,
      "learning_rate": 2.2976152874289326e-05,
      "loss": 0.3708,
      "mean_token_accuracy": 0.9182381510734559,
      "step": 2965,
      "train/kl_loss_qwen": 0.035771107859909536,
      "train/kl_loss_r1": 0.07682945877313614,
      "train/total_kl": 0.11260056495666504
    },
    {
      "epoch": 0.9380922299431459,
      "grad_norm": 4.0,
      "learning_rate": 2.2964308275426405e-05,
      "loss": 0.372,
      "mean_token_accuracy": 0.9296153545379638,
      "step": 2970,
      "train/kl_loss_qwen": 0.04844686575233936,
      "train/kl_loss_r1": 0.09364785552024842,
      "train/total_kl": 0.14209472239017487
    },
    {
      "epoch": 0.939671509791535,
      "grad_norm": 2.796875,
      "learning_rate": 2.295246367656349e-05,
      "loss": 0.3575,
      "mean_token_accuracy": 0.930724847316742,
      "step": 2975,
      "train/kl_loss_qwen": 0.0420586135238409,
      "train/kl_loss_r1": 0.1059981182217598,
      "train/total_kl": 0.1480567306280136
    },
    {
      "epoch": 0.9412507896399241,
      "grad_norm": 2.96875,
      "learning_rate": 2.294061907770057e-05,
      "loss": 0.3273,
      "mean_token_accuracy": 0.9271089792251587,
      "step": 2980,
      "train/kl_loss_qwen": 0.02205999381840229,
      "train/kl_loss_r1": 0.06332791969180107,
      "train/total_kl": 0.0853879138827324
    },
    {
      "epoch": 0.9428300694883134,
      "grad_norm": 3.1875,
      "learning_rate": 2.292877447883765e-05,
      "loss": 0.3493,
      "mean_token_accuracy": 0.9249009490013123,
      "step": 2985,
      "train/kl_loss_qwen": 0.036800478026270865,
      "train/kl_loss_r1": 0.07560685127973557,
      "train/total_kl": 0.11240732818841934
    },
    {
      "epoch": 0.9444093493367025,
      "grad_norm": 3.5625,
      "learning_rate": 2.2916929879974733e-05,
      "loss": 0.3967,
      "mean_token_accuracy": 0.9188141107559205,
      "step": 2990,
      "train/kl_loss_qwen": 0.04829471595585346,
      "train/kl_loss_r1": 0.07904288470745087,
      "train/total_kl": 0.1273376002907753
    },
    {
      "epoch": 0.9459886291850916,
      "grad_norm": 3.125,
      "learning_rate": 2.2905085281111815e-05,
      "loss": 0.3345,
      "mean_token_accuracy": 0.9291098713874817,
      "step": 2995,
      "train/kl_loss_qwen": 0.033198049664497374,
      "train/kl_loss_r1": 0.07016403153538704,
      "train/total_kl": 0.10336208194494248
    },
    {
      "epoch": 0.9475679090334808,
      "grad_norm": 3.6875,
      "learning_rate": 2.2893240682248893e-05,
      "loss": 0.4435,
      "mean_token_accuracy": 0.9104521751403809,
      "step": 3000,
      "train/kl_loss_qwen": 0.049626877158880235,
      "train/kl_loss_r1": 0.10760048627853394,
      "train/total_kl": 0.15722736418247224
    },
    {
      "epoch": 0.9491471888818699,
      "grad_norm": 3.546875,
      "learning_rate": 2.288139608338598e-05,
      "loss": 0.3387,
      "mean_token_accuracy": 0.9262456059455871,
      "step": 3005,
      "train/kl_loss_qwen": 0.034047160670161246,
      "train/kl_loss_r1": 0.07138821184635162,
      "train/total_kl": 0.10543537512421608
    },
    {
      "epoch": 0.950726468730259,
      "grad_norm": 3.25,
      "learning_rate": 2.2869551484523057e-05,
      "loss": 0.4141,
      "mean_token_accuracy": 0.9247868299484253,
      "step": 3010,
      "train/kl_loss_qwen": 0.04992226213216781,
      "train/kl_loss_r1": 0.1187254473567009,
      "train/total_kl": 0.1686477094888687
    },
    {
      "epoch": 0.9523057485786481,
      "grad_norm": 4.25,
      "learning_rate": 2.2857706885660143e-05,
      "loss": 0.415,
      "mean_token_accuracy": 0.9188775420188904,
      "step": 3015,
      "train/kl_loss_qwen": 0.05535581335425377,
      "train/kl_loss_r1": 0.10881091207265854,
      "train/total_kl": 0.164166721701622
    },
    {
      "epoch": 0.9538850284270373,
      "grad_norm": 3.765625,
      "learning_rate": 2.284586228679722e-05,
      "loss": 0.3554,
      "mean_token_accuracy": 0.9197573184967041,
      "step": 3020,
      "train/kl_loss_qwen": 0.027838678285479545,
      "train/kl_loss_r1": 0.07363396286964416,
      "train/total_kl": 0.10147264301776886
    },
    {
      "epoch": 0.9554643082754264,
      "grad_norm": 3.390625,
      "learning_rate": 2.28340176879343e-05,
      "loss": 0.3568,
      "mean_token_accuracy": 0.9237677574157714,
      "step": 3025,
      "train/kl_loss_qwen": 0.028969788551330568,
      "train/kl_loss_r1": 0.07271746098995209,
      "train/total_kl": 0.10168725252151489
    },
    {
      "epoch": 0.9570435881238155,
      "grad_norm": 4.03125,
      "learning_rate": 2.2822173089071385e-05,
      "loss": 0.3665,
      "mean_token_accuracy": 0.9218949794769287,
      "step": 3030,
      "train/kl_loss_qwen": 0.03140850998461246,
      "train/kl_loss_r1": 0.08311051279306411,
      "train/total_kl": 0.11451902389526367
    },
    {
      "epoch": 0.9586228679722046,
      "grad_norm": 2.796875,
      "learning_rate": 2.2810328490208464e-05,
      "loss": 0.3563,
      "mean_token_accuracy": 0.9198498725891113,
      "step": 3035,
      "train/kl_loss_qwen": 0.03563178218901157,
      "train/kl_loss_r1": 0.08951483368873596,
      "train/total_kl": 0.1251466155052185
    },
    {
      "epoch": 0.9602021478205938,
      "grad_norm": 2.96875,
      "learning_rate": 2.2798483891345546e-05,
      "loss": 0.3378,
      "mean_token_accuracy": 0.9267775058746338,
      "step": 3040,
      "train/kl_loss_qwen": 0.03462435491383076,
      "train/kl_loss_r1": 0.07001143991947174,
      "train/total_kl": 0.10463579595088959
    },
    {
      "epoch": 0.9617814276689829,
      "grad_norm": 3.046875,
      "learning_rate": 2.2786639292482628e-05,
      "loss": 0.3523,
      "mean_token_accuracy": 0.9350518822669983,
      "step": 3045,
      "train/kl_loss_qwen": 0.05043598674237728,
      "train/kl_loss_r1": 0.10024086683988571,
      "train/total_kl": 0.1506768524646759
    },
    {
      "epoch": 0.9633607075173721,
      "grad_norm": 2.890625,
      "learning_rate": 2.277479469361971e-05,
      "loss": 0.3147,
      "mean_token_accuracy": 0.9323802351951599,
      "step": 3050,
      "train/kl_loss_qwen": 0.03752680234611035,
      "train/kl_loss_r1": 0.06463054791092873,
      "train/total_kl": 0.10215734988451004
    },
    {
      "epoch": 0.9649399873657613,
      "grad_norm": 5.15625,
      "learning_rate": 2.2762950094756792e-05,
      "loss": 0.4081,
      "mean_token_accuracy": 0.9119511842727661,
      "step": 3055,
      "train/kl_loss_qwen": 0.03387603722512722,
      "train/kl_loss_r1": 0.08608716651797295,
      "train/total_kl": 0.11996320486068726
    },
    {
      "epoch": 0.9665192672141504,
      "grad_norm": 12.5,
      "learning_rate": 2.2751105495893874e-05,
      "loss": 0.4506,
      "mean_token_accuracy": 0.921574592590332,
      "step": 3060,
      "train/kl_loss_qwen": 0.08637710213661194,
      "train/kl_loss_r1": 0.13138798698782922,
      "train/total_kl": 0.21776508986949922
    },
    {
      "epoch": 0.9680985470625395,
      "grad_norm": 3.21875,
      "learning_rate": 2.2739260897030952e-05,
      "loss": 0.6707,
      "mean_token_accuracy": 0.8928841352462769,
      "step": 3065,
      "train/kl_loss_qwen": 0.1280539806932211,
      "train/kl_loss_r1": 0.1995810404419899,
      "train/total_kl": 0.32763502597808836
    },
    {
      "epoch": 0.9696778269109286,
      "grad_norm": 3.4375,
      "learning_rate": 2.2727416298168038e-05,
      "loss": 0.3333,
      "mean_token_accuracy": 0.9307949066162109,
      "step": 3070,
      "train/kl_loss_qwen": 0.04029642380774021,
      "train/kl_loss_r1": 0.07726048901677132,
      "train/total_kl": 0.11755691170692444
    },
    {
      "epoch": 0.9712571067593178,
      "grad_norm": 3.0,
      "learning_rate": 2.2715571699305116e-05,
      "loss": 0.3492,
      "mean_token_accuracy": 0.9251341581344604,
      "step": 3075,
      "train/kl_loss_qwen": 0.037018724530935285,
      "train/kl_loss_r1": 0.08549067825078964,
      "train/total_kl": 0.12250940203666687
    },
    {
      "epoch": 0.9728363866077069,
      "grad_norm": 2.8125,
      "learning_rate": 2.2703727100442198e-05,
      "loss": 0.6136,
      "mean_token_accuracy": 0.9143887519836426,
      "step": 3080,
      "train/kl_loss_qwen": 0.11673750467598439,
      "train/kl_loss_r1": 0.23669594377279282,
      "train/total_kl": 0.3534334346652031
    },
    {
      "epoch": 0.974415666456096,
      "grad_norm": 4.1875,
      "learning_rate": 2.269188250157928e-05,
      "loss": 0.3599,
      "mean_token_accuracy": 0.9242613196372986,
      "step": 3085,
      "train/kl_loss_qwen": 0.03269352987408638,
      "train/kl_loss_r1": 0.06747004389762878,
      "train/total_kl": 0.10016357451677323
    },
    {
      "epoch": 0.9759949463044851,
      "grad_norm": 2.5625,
      "learning_rate": 2.2680037902716362e-05,
      "loss": 0.3089,
      "mean_token_accuracy": 0.928266453742981,
      "step": 3090,
      "train/kl_loss_qwen": 0.02719697989523411,
      "train/kl_loss_r1": 0.06878720223903656,
      "train/total_kl": 0.09598418176174164
    },
    {
      "epoch": 0.9775742261528743,
      "grad_norm": 4.65625,
      "learning_rate": 2.2668193303853444e-05,
      "loss": 0.3905,
      "mean_token_accuracy": 0.9221195220947266,
      "step": 3095,
      "train/kl_loss_qwen": 0.04350057765841484,
      "train/kl_loss_r1": 0.0952111080288887,
      "train/total_kl": 0.13871168792247773
    },
    {
      "epoch": 0.9791535060012634,
      "grad_norm": 4.1875,
      "learning_rate": 2.2656348704990526e-05,
      "loss": 0.5897,
      "mean_token_accuracy": 0.898698341846466,
      "step": 3100,
      "train/kl_loss_qwen": 0.12271191589534283,
      "train/kl_loss_r1": 0.16651244312524796,
      "train/total_kl": 0.2892243444919586
    },
    {
      "epoch": 0.9807327858496525,
      "grad_norm": 3.796875,
      "learning_rate": 2.2644504106127605e-05,
      "loss": 0.3565,
      "mean_token_accuracy": 0.9222179651260376,
      "step": 3105,
      "train/kl_loss_qwen": 0.0371067862957716,
      "train/kl_loss_r1": 0.08233626335859298,
      "train/total_kl": 0.1194430485367775
    },
    {
      "epoch": 0.9823120656980417,
      "grad_norm": 3.703125,
      "learning_rate": 2.263265950726469e-05,
      "loss": 0.3615,
      "mean_token_accuracy": 0.923066246509552,
      "step": 3110,
      "train/kl_loss_qwen": 0.04022606089711189,
      "train/kl_loss_r1": 0.08371982723474503,
      "train/total_kl": 0.12394588887691498
    },
    {
      "epoch": 0.9838913455464309,
      "grad_norm": 3.21875,
      "learning_rate": 2.262081490840177e-05,
      "loss": 0.3641,
      "mean_token_accuracy": 0.9311588764190674,
      "step": 3115,
      "train/kl_loss_qwen": 0.0440082386136055,
      "train/kl_loss_r1": 0.09267747104167938,
      "train/total_kl": 0.13668570816516876
    },
    {
      "epoch": 0.98547062539482,
      "grad_norm": 3.375,
      "learning_rate": 2.260897030953885e-05,
      "loss": 0.3496,
      "mean_token_accuracy": 0.9226747393608093,
      "step": 3120,
      "train/kl_loss_qwen": 0.02800327055156231,
      "train/kl_loss_r1": 0.07155929207801819,
      "train/total_kl": 0.09956256300210953
    },
    {
      "epoch": 0.9870499052432091,
      "grad_norm": 2.515625,
      "learning_rate": 2.2597125710675933e-05,
      "loss": 0.359,
      "mean_token_accuracy": 0.9280364632606506,
      "step": 3125,
      "train/kl_loss_qwen": 0.03147074244916439,
      "train/kl_loss_r1": 0.09552234336733818,
      "train/total_kl": 0.12699308395385742
    },
    {
      "epoch": 0.9886291850915983,
      "grad_norm": 3.5625,
      "learning_rate": 2.2585281111813015e-05,
      "loss": 0.3569,
      "mean_token_accuracy": 0.920689857006073,
      "step": 3130,
      "train/kl_loss_qwen": 0.03933285139501095,
      "train/kl_loss_r1": 0.06674346774816513,
      "train/total_kl": 0.10607631802558899
    },
    {
      "epoch": 0.9902084649399874,
      "grad_norm": 3.25,
      "learning_rate": 2.2573436512950097e-05,
      "loss": 0.3731,
      "mean_token_accuracy": 0.922060751914978,
      "step": 3135,
      "train/kl_loss_qwen": 0.04032552763819695,
      "train/kl_loss_r1": 0.0840625137090683,
      "train/total_kl": 0.12438803762197495
    },
    {
      "epoch": 0.9917877447883765,
      "grad_norm": 4.71875,
      "learning_rate": 2.256159191408718e-05,
      "loss": 0.3657,
      "mean_token_accuracy": 0.9147815704345703,
      "step": 3140,
      "train/kl_loss_qwen": 0.02696884088218212,
      "train/kl_loss_r1": 0.0729674868285656,
      "train/total_kl": 0.09993632584810257
    },
    {
      "epoch": 0.9933670246367656,
      "grad_norm": 2.921875,
      "learning_rate": 2.2549747315224257e-05,
      "loss": 0.3184,
      "mean_token_accuracy": 0.9361499905586242,
      "step": 3145,
      "train/kl_loss_qwen": 0.04222720637917519,
      "train/kl_loss_r1": 0.08119570612907409,
      "train/total_kl": 0.12342291474342346
    },
    {
      "epoch": 0.9949463044851548,
      "grad_norm": 3.15625,
      "learning_rate": 2.253790271636134e-05,
      "loss": 0.3245,
      "mean_token_accuracy": 0.9341140985488892,
      "step": 3150,
      "train/kl_loss_qwen": 0.04220006763935089,
      "train/kl_loss_r1": 0.0822506494820118,
      "train/total_kl": 0.12445071637630463
    },
    {
      "epoch": 0.9965255843335439,
      "grad_norm": 3.4375,
      "learning_rate": 2.252605811749842e-05,
      "loss": 0.3819,
      "mean_token_accuracy": 0.9054730057716369,
      "step": 3155,
      "train/kl_loss_qwen": 0.021076304465532304,
      "train/kl_loss_r1": 0.07582696080207825,
      "train/total_kl": 0.0969032660126686
    },
    {
      "epoch": 0.998104864181933,
      "grad_norm": 3.234375,
      "learning_rate": 2.2514213518635503e-05,
      "loss": 0.3647,
      "mean_token_accuracy": 0.9255593419075012,
      "step": 3160,
      "train/kl_loss_qwen": 0.037224308401346204,
      "train/kl_loss_r1": 0.08729824349284172,
      "train/total_kl": 0.12452255338430404
    },
    {
      "epoch": 0.9996841440303221,
      "grad_norm": 3.15625,
      "learning_rate": 2.2502368919772585e-05,
      "loss": 0.3415,
      "mean_token_accuracy": 0.9309972524642944,
      "step": 3165,
      "train/kl_loss_qwen": 0.041233925893902776,
      "train/kl_loss_r1": 0.08098211735486985,
      "train/total_kl": 0.12221604138612747
    },
    {
      "epoch": 1.0012634238787113,
      "grad_norm": 2.65625,
      "learning_rate": 2.2490524320909664e-05,
      "loss": 0.3115,
      "mean_token_accuracy": 0.9560119390487671,
      "step": 3170,
      "train/kl_loss_qwen": 0.025640086270868778,
      "train/kl_loss_r1": 0.12074330374598503,
      "train/total_kl": 0.14638338908553122
    },
    {
      "epoch": 1.0028427037271004,
      "grad_norm": 2.28125,
      "learning_rate": 2.247867972204675e-05,
      "loss": 0.2896,
      "mean_token_accuracy": 0.9650134563446044,
      "step": 3175,
      "train/kl_loss_qwen": 0.0328770499676466,
      "train/kl_loss_r1": 0.11884289234876633,
      "train/total_kl": 0.15171994268894196
    },
    {
      "epoch": 1.0044219835754895,
      "grad_norm": 1.890625,
      "learning_rate": 2.2466835123183828e-05,
      "loss": 0.2712,
      "mean_token_accuracy": 0.9635287642478942,
      "step": 3180,
      "train/kl_loss_qwen": 0.026448529772460462,
      "train/kl_loss_r1": 0.09526190012693406,
      "train/total_kl": 0.12171043008565903
    },
    {
      "epoch": 1.0060012634238786,
      "grad_norm": 1.96875,
      "learning_rate": 2.245499052432091e-05,
      "loss": 0.2668,
      "mean_token_accuracy": 0.9690022349357605,
      "step": 3185,
      "train/kl_loss_qwen": 0.04128575399518013,
      "train/kl_loss_r1": 0.09623772948980332,
      "train/total_kl": 0.13752348572015763
    },
    {
      "epoch": 1.0075805432722678,
      "grad_norm": 5.96875,
      "learning_rate": 2.244314592545799e-05,
      "loss": 0.3071,
      "mean_token_accuracy": 0.9643856287002563,
      "step": 3190,
      "train/kl_loss_qwen": 0.044674181193113324,
      "train/kl_loss_r1": 0.11642839163541793,
      "train/total_kl": 0.16110257506370546
    },
    {
      "epoch": 1.0091598231206569,
      "grad_norm": 7.5625,
      "learning_rate": 2.2431301326595073e-05,
      "loss": 0.3642,
      "mean_token_accuracy": 0.9574234843254089,
      "step": 3195,
      "train/kl_loss_qwen": 0.07043856158852577,
      "train/kl_loss_r1": 0.13857488334178925,
      "train/total_kl": 0.20901344418525697
    },
    {
      "epoch": 1.010739102969046,
      "grad_norm": 2.34375,
      "learning_rate": 2.2419456727732155e-05,
      "loss": 0.5656,
      "mean_token_accuracy": 0.938858687877655,
      "step": 3200,
      "train/kl_loss_qwen": 0.12362933047115802,
      "train/kl_loss_r1": 0.2340306669473648,
      "train/total_kl": 0.3576599955558777
    },
    {
      "epoch": 1.0123183828174354,
      "grad_norm": 3.109375,
      "learning_rate": 2.2407612128869237e-05,
      "loss": 0.298,
      "mean_token_accuracy": 0.9657655239105225,
      "step": 3205,
      "train/kl_loss_qwen": 0.043574480712413786,
      "train/kl_loss_r1": 0.12193096578121185,
      "train/total_kl": 0.16550544500350953
    },
    {
      "epoch": 1.0138976626658245,
      "grad_norm": 2.9375,
      "learning_rate": 2.2395767530006316e-05,
      "loss": 0.2857,
      "mean_token_accuracy": 0.9667572736740112,
      "step": 3210,
      "train/kl_loss_qwen": 0.03321674019098282,
      "train/kl_loss_r1": 0.12157121300697327,
      "train/total_kl": 0.15478795170783996
    },
    {
      "epoch": 1.0154769425142136,
      "grad_norm": 2.109375,
      "learning_rate": 2.23839229311434e-05,
      "loss": 0.2583,
      "mean_token_accuracy": 0.9628570318222046,
      "step": 3215,
      "train/kl_loss_qwen": 0.028814373910427092,
      "train/kl_loss_r1": 0.0887444943189621,
      "train/total_kl": 0.11755886971950531
    },
    {
      "epoch": 1.0170562223626027,
      "grad_norm": 3.5,
      "learning_rate": 2.237207833228048e-05,
      "loss": 0.3447,
      "mean_token_accuracy": 0.9514865040779114,
      "step": 3220,
      "train/kl_loss_qwen": 0.05198634192347527,
      "train/kl_loss_r1": 0.11496023386716843,
      "train/total_kl": 0.16694657802581786
    },
    {
      "epoch": 1.0186355022109919,
      "grad_norm": 5.5,
      "learning_rate": 2.2360233733417562e-05,
      "loss": 0.3436,
      "mean_token_accuracy": 0.9510520815849304,
      "step": 3225,
      "train/kl_loss_qwen": 0.0410022359341383,
      "train/kl_loss_r1": 0.10859209150075913,
      "train/total_kl": 0.14959432929754257
    },
    {
      "epoch": 1.020214782059381,
      "grad_norm": 2.515625,
      "learning_rate": 2.2348389134554644e-05,
      "loss": 0.2742,
      "mean_token_accuracy": 0.9683983206748963,
      "step": 3230,
      "train/kl_loss_qwen": 0.03593085892498493,
      "train/kl_loss_r1": 0.10674776285886764,
      "train/total_kl": 0.14267862141132354
    },
    {
      "epoch": 1.0217940619077701,
      "grad_norm": 3.234375,
      "learning_rate": 2.2336544535691726e-05,
      "loss": 0.3569,
      "mean_token_accuracy": 0.9548901081085205,
      "step": 3235,
      "train/kl_loss_qwen": 0.05419312231242657,
      "train/kl_loss_r1": 0.1455776087939739,
      "train/total_kl": 0.199770724773407
    },
    {
      "epoch": 1.0233733417561592,
      "grad_norm": 1.9453125,
      "learning_rate": 2.2324699936828808e-05,
      "loss": 0.2734,
      "mean_token_accuracy": 0.9657649159431457,
      "step": 3240,
      "train/kl_loss_qwen": 0.03713821172714234,
      "train/kl_loss_r1": 0.11058740466833114,
      "train/total_kl": 0.1477256178855896
    },
    {
      "epoch": 1.0249526216045484,
      "grad_norm": 6.875,
      "learning_rate": 2.231285533796589e-05,
      "loss": 0.2987,
      "mean_token_accuracy": 0.9603511571884156,
      "step": 3245,
      "train/kl_loss_qwen": 0.05136045143008232,
      "train/kl_loss_r1": 0.10308634340763093,
      "train/total_kl": 0.15444678962230682
    },
    {
      "epoch": 1.0265319014529375,
      "grad_norm": 2.9375,
      "learning_rate": 2.230101073910297e-05,
      "loss": 0.3232,
      "mean_token_accuracy": 0.9601680874824524,
      "step": 3250,
      "train/kl_loss_qwen": 0.05245714895427227,
      "train/kl_loss_r1": 0.12677610963582991,
      "train/total_kl": 0.17923325747251512
    },
    {
      "epoch": 1.0281111813013266,
      "grad_norm": 2.546875,
      "learning_rate": 2.2289166140240054e-05,
      "loss": 0.2906,
      "mean_token_accuracy": 0.9582780122756958,
      "step": 3255,
      "train/kl_loss_qwen": 0.03890797644853592,
      "train/kl_loss_r1": 0.08900529518723488,
      "train/total_kl": 0.12791327089071275
    },
    {
      "epoch": 1.0296904611497157,
      "grad_norm": 1.9765625,
      "learning_rate": 2.2277321541377132e-05,
      "loss": 0.2692,
      "mean_token_accuracy": 0.9587826490402221,
      "step": 3260,
      "train/kl_loss_qwen": 0.029529448971152304,
      "train/kl_loss_r1": 0.08386245369911194,
      "train/total_kl": 0.11339190155267716
    },
    {
      "epoch": 1.0312697409981049,
      "grad_norm": 1.71875,
      "learning_rate": 2.2265476942514214e-05,
      "loss": 0.2837,
      "mean_token_accuracy": 0.9657258272171021,
      "step": 3265,
      "train/kl_loss_qwen": 0.03317863568663597,
      "train/kl_loss_r1": 0.11466053426265717,
      "train/total_kl": 0.1478391706943512
    },
    {
      "epoch": 1.032849020846494,
      "grad_norm": 1.859375,
      "learning_rate": 2.2253632343651296e-05,
      "loss": 0.2611,
      "mean_token_accuracy": 0.950425100326538,
      "step": 3270,
      "train/kl_loss_qwen": 0.02328247856348753,
      "train/kl_loss_r1": 0.07890274450182914,
      "train/total_kl": 0.10218522250652314
    },
    {
      "epoch": 1.0344283006948831,
      "grad_norm": 3.03125,
      "learning_rate": 2.2241787744788375e-05,
      "loss": 0.3167,
      "mean_token_accuracy": 0.9660818219184876,
      "step": 3275,
      "train/kl_loss_qwen": 0.051942231506109236,
      "train/kl_loss_r1": 0.13846394866704942,
      "train/total_kl": 0.1904061794281006
    },
    {
      "epoch": 1.0360075805432722,
      "grad_norm": 2.65625,
      "learning_rate": 2.222994314592546e-05,
      "loss": 0.3012,
      "mean_token_accuracy": 0.9598979115486145,
      "step": 3280,
      "train/kl_loss_qwen": 0.0401642233133316,
      "train/kl_loss_r1": 0.10353323072195053,
      "train/total_kl": 0.14369745552539825
    },
    {
      "epoch": 1.0375868603916614,
      "grad_norm": 4.34375,
      "learning_rate": 2.221809854706254e-05,
      "loss": 0.3157,
      "mean_token_accuracy": 0.9546847343444824,
      "step": 3285,
      "train/kl_loss_qwen": 0.04073624797165394,
      "train/kl_loss_r1": 0.10703556388616561,
      "train/total_kl": 0.14777181148529053
    },
    {
      "epoch": 1.0391661402400505,
      "grad_norm": 2.359375,
      "learning_rate": 2.220625394819962e-05,
      "loss": 0.2975,
      "mean_token_accuracy": 0.9523539900779724,
      "step": 3290,
      "train/kl_loss_qwen": 0.037301792204380034,
      "train/kl_loss_r1": 0.09221957325935363,
      "train/total_kl": 0.12952136248350143
    },
    {
      "epoch": 1.0407454200884396,
      "grad_norm": 3.546875,
      "learning_rate": 2.2194409349336703e-05,
      "loss": 0.3277,
      "mean_token_accuracy": 0.9523527383804321,
      "step": 3295,
      "train/kl_loss_qwen": 0.03942328318953514,
      "train/kl_loss_r1": 0.11786964982748031,
      "train/total_kl": 0.1572929322719574
    },
    {
      "epoch": 1.0423246999368287,
      "grad_norm": 2.109375,
      "learning_rate": 2.2182564750473785e-05,
      "loss": 0.2809,
      "mean_token_accuracy": 0.9592664361000061,
      "step": 3300,
      "train/kl_loss_qwen": 0.028701091930270196,
      "train/kl_loss_r1": 0.10500250160694122,
      "train/total_kl": 0.1337035983800888
    },
    {
      "epoch": 1.0439039797852179,
      "grad_norm": 2.28125,
      "learning_rate": 2.2170720151610863e-05,
      "loss": 0.2665,
      "mean_token_accuracy": 0.9650776505470275,
      "step": 3305,
      "train/kl_loss_qwen": 0.030730508454144,
      "train/kl_loss_r1": 0.10481556057929993,
      "train/total_kl": 0.13554607033729554
    },
    {
      "epoch": 1.045483259633607,
      "grad_norm": 1.9921875,
      "learning_rate": 2.215887555274795e-05,
      "loss": 0.2259,
      "mean_token_accuracy": 0.9697175025939941,
      "step": 3310,
      "train/kl_loss_qwen": 0.027501318976283072,
      "train/kl_loss_r1": 0.0796351008117199,
      "train/total_kl": 0.10713641941547394
    },
    {
      "epoch": 1.0470625394819961,
      "grad_norm": 1.8359375,
      "learning_rate": 2.2147030953885027e-05,
      "loss": 0.2426,
      "mean_token_accuracy": 0.9637905478477478,
      "step": 3315,
      "train/kl_loss_qwen": 0.02786708325147629,
      "train/kl_loss_r1": 0.07790099903941154,
      "train/total_kl": 0.10576808154582977
    },
    {
      "epoch": 1.0486418193303852,
      "grad_norm": 3.953125,
      "learning_rate": 2.2135186355022113e-05,
      "loss": 0.2859,
      "mean_token_accuracy": 0.9635117053985596,
      "step": 3320,
      "train/kl_loss_qwen": 0.04328730963170528,
      "train/kl_loss_r1": 0.09670650511980057,
      "train/total_kl": 0.13999381363391877
    },
    {
      "epoch": 1.0502210991787744,
      "grad_norm": 4.125,
      "learning_rate": 2.212334175615919e-05,
      "loss": 0.3245,
      "mean_token_accuracy": 0.959052312374115,
      "step": 3325,
      "train/kl_loss_qwen": 0.04880684055387974,
      "train/kl_loss_r1": 0.12860168218612672,
      "train/total_kl": 0.17740852534770965
    },
    {
      "epoch": 1.0518003790271635,
      "grad_norm": 2.90625,
      "learning_rate": 2.2111497157296273e-05,
      "loss": 0.316,
      "mean_token_accuracy": 0.9524584531784057,
      "step": 3330,
      "train/kl_loss_qwen": 0.048845131695270536,
      "train/kl_loss_r1": 0.10177739262580872,
      "train/total_kl": 0.15062252283096314
    },
    {
      "epoch": 1.0533796588755528,
      "grad_norm": 2.734375,
      "learning_rate": 2.2099652558433355e-05,
      "loss": 0.2758,
      "mean_token_accuracy": 0.9694467067718506,
      "step": 3335,
      "train/kl_loss_qwen": 0.04869394265115261,
      "train/kl_loss_r1": 0.10278871655464172,
      "train/total_kl": 0.15148265659809113
    },
    {
      "epoch": 1.054958938723942,
      "grad_norm": 2.0625,
      "learning_rate": 2.2087807959570437e-05,
      "loss": 0.2657,
      "mean_token_accuracy": 0.9645479083061218,
      "step": 3340,
      "train/kl_loss_qwen": 0.034964132308959964,
      "train/kl_loss_r1": 0.08698341399431228,
      "train/total_kl": 0.12194754481315613
    },
    {
      "epoch": 1.056538218572331,
      "grad_norm": 1.796875,
      "learning_rate": 2.2075963360707516e-05,
      "loss": 0.2873,
      "mean_token_accuracy": 0.9597893595695496,
      "step": 3345,
      "train/kl_loss_qwen": 0.037495244294404984,
      "train/kl_loss_r1": 0.09074312746524811,
      "train/total_kl": 0.12823836952447892
    },
    {
      "epoch": 1.0581174984207202,
      "grad_norm": 2.5625,
      "learning_rate": 2.20641187618446e-05,
      "loss": 0.304,
      "mean_token_accuracy": 0.9663947463035584,
      "step": 3350,
      "train/kl_loss_qwen": 0.0512197095900774,
      "train/kl_loss_r1": 0.11303173750638962,
      "train/total_kl": 0.1642514482140541
    },
    {
      "epoch": 1.0596967782691094,
      "grad_norm": 2.640625,
      "learning_rate": 2.205227416298168e-05,
      "loss": 0.2904,
      "mean_token_accuracy": 0.9578498125076294,
      "step": 3355,
      "train/kl_loss_qwen": 0.04140202291309834,
      "train/kl_loss_r1": 0.09443714842200279,
      "train/total_kl": 0.13583917394280434
    },
    {
      "epoch": 1.0612760581174985,
      "grad_norm": 1.6875,
      "learning_rate": 2.2040429564118765e-05,
      "loss": 0.258,
      "mean_token_accuracy": 0.9654830098152161,
      "step": 3360,
      "train/kl_loss_qwen": 0.03410881944000721,
      "train/kl_loss_r1": 0.0894771508872509,
      "train/total_kl": 0.12358597069978713
    },
    {
      "epoch": 1.0628553379658876,
      "grad_norm": 2.796875,
      "learning_rate": 2.2028584965255844e-05,
      "loss": 0.2561,
      "mean_token_accuracy": 0.9660322308540344,
      "step": 3365,
      "train/kl_loss_qwen": 0.036791940405964854,
      "train/kl_loss_r1": 0.09082298055291176,
      "train/total_kl": 0.12761492133140565
    },
    {
      "epoch": 1.0644346178142767,
      "grad_norm": 4.0625,
      "learning_rate": 2.2016740366392926e-05,
      "loss": 0.3081,
      "mean_token_accuracy": 0.9637358784675598,
      "step": 3370,
      "train/kl_loss_qwen": 0.05463301166892052,
      "train/kl_loss_r1": 0.10905377119779587,
      "train/total_kl": 0.16368678361177444
    },
    {
      "epoch": 1.0660138976626659,
      "grad_norm": 2.8125,
      "learning_rate": 2.2004895767530008e-05,
      "loss": 0.2888,
      "mean_token_accuracy": 0.9592244625091553,
      "step": 3375,
      "train/kl_loss_qwen": 0.04797116667032242,
      "train/kl_loss_r1": 0.098827163875103,
      "train/total_kl": 0.1467983305454254
    },
    {
      "epoch": 1.067593177511055,
      "grad_norm": 3.484375,
      "learning_rate": 2.199305116866709e-05,
      "loss": 0.2838,
      "mean_token_accuracy": 0.9657128334045411,
      "step": 3380,
      "train/kl_loss_qwen": 0.05274052917957306,
      "train/kl_loss_r1": 0.10606652945280075,
      "train/total_kl": 0.15880705714225768
    },
    {
      "epoch": 1.069172457359444,
      "grad_norm": 2.59375,
      "learning_rate": 2.1981206569804168e-05,
      "loss": 0.3426,
      "mean_token_accuracy": 0.950219202041626,
      "step": 3385,
      "train/kl_loss_qwen": 0.05321107804775238,
      "train/kl_loss_r1": 0.11633624285459518,
      "train/total_kl": 0.16954732239246367
    },
    {
      "epoch": 1.0707517372078332,
      "grad_norm": 4.9375,
      "learning_rate": 2.1969361970941253e-05,
      "loss": 0.3412,
      "mean_token_accuracy": 0.9603010058403015,
      "step": 3390,
      "train/kl_loss_qwen": 0.042314786836504935,
      "train/kl_loss_r1": 0.14008679687976838,
      "train/total_kl": 0.18240158557891845
    },
    {
      "epoch": 1.0723310170562224,
      "grad_norm": 2.4375,
      "learning_rate": 2.1957517372078332e-05,
      "loss": 0.2633,
      "mean_token_accuracy": 0.966342318058014,
      "step": 3395,
      "train/kl_loss_qwen": 0.03996781017631292,
      "train/kl_loss_r1": 0.09646985828876495,
      "train/total_kl": 0.13643766790628434
    },
    {
      "epoch": 1.0739102969046115,
      "grad_norm": 2.609375,
      "learning_rate": 2.1945672773215414e-05,
      "loss": 0.351,
      "mean_token_accuracy": 0.9565378308296204,
      "step": 3400,
      "train/kl_loss_qwen": 0.04315974079072475,
      "train/kl_loss_r1": 0.12781278789043427,
      "train/total_kl": 0.17097252905368804
    },
    {
      "epoch": 1.0754895767530006,
      "grad_norm": 2.53125,
      "learning_rate": 2.1933828174352496e-05,
      "loss": 0.299,
      "mean_token_accuracy": 0.9614153385162354,
      "step": 3405,
      "train/kl_loss_qwen": 0.04182270169258118,
      "train/kl_loss_r1": 0.11325979232788086,
      "train/total_kl": 0.15508249402046204
    },
    {
      "epoch": 1.0770688566013897,
      "grad_norm": 2.453125,
      "learning_rate": 2.1921983575489575e-05,
      "loss": 0.3093,
      "mean_token_accuracy": 0.9569859385490418,
      "step": 3410,
      "train/kl_loss_qwen": 0.04079798385500908,
      "train/kl_loss_r1": 0.10753391683101654,
      "train/total_kl": 0.14833189845085143
    },
    {
      "epoch": 1.0786481364497789,
      "grad_norm": 2.015625,
      "learning_rate": 2.191013897662666e-05,
      "loss": 0.2799,
      "mean_token_accuracy": 0.9653553605079651,
      "step": 3415,
      "train/kl_loss_qwen": 0.041333476826548574,
      "train/kl_loss_r1": 0.10004743486642838,
      "train/total_kl": 0.14138090908527373
    },
    {
      "epoch": 1.080227416298168,
      "grad_norm": 2.984375,
      "learning_rate": 2.189829437776374e-05,
      "loss": 0.3013,
      "mean_token_accuracy": 0.9548194408416748,
      "step": 3420,
      "train/kl_loss_qwen": 0.04067806303501129,
      "train/kl_loss_r1": 0.08417079150676728,
      "train/total_kl": 0.12484885305166245
    },
    {
      "epoch": 1.0818066961465571,
      "grad_norm": 2.90625,
      "learning_rate": 2.188644977890082e-05,
      "loss": 0.3332,
      "mean_token_accuracy": 0.9564208030700684,
      "step": 3425,
      "train/kl_loss_qwen": 0.05181012526154518,
      "train/kl_loss_r1": 0.11184037923812866,
      "train/total_kl": 0.1636505037546158
    },
    {
      "epoch": 1.0833859759949462,
      "grad_norm": 3.390625,
      "learning_rate": 2.1874605180037903e-05,
      "loss": 0.2836,
      "mean_token_accuracy": 0.961362612247467,
      "step": 3430,
      "train/kl_loss_qwen": 0.02907777689397335,
      "train/kl_loss_r1": 0.10245824903249741,
      "train/total_kl": 0.13153602480888366
    },
    {
      "epoch": 1.0849652558433354,
      "grad_norm": 2.046875,
      "learning_rate": 2.1862760581174984e-05,
      "loss": 0.2858,
      "mean_token_accuracy": 0.971900749206543,
      "step": 3435,
      "train/kl_loss_qwen": 0.043740443885326385,
      "train/kl_loss_r1": 0.11398271322250367,
      "train/total_kl": 0.15772316008806228
    },
    {
      "epoch": 1.0865445356917245,
      "grad_norm": 3.296875,
      "learning_rate": 2.1850915982312066e-05,
      "loss": 0.3217,
      "mean_token_accuracy": 0.9529153227806091,
      "step": 3440,
      "train/kl_loss_qwen": 0.039821751043200494,
      "train/kl_loss_r1": 0.10646492689847946,
      "train/total_kl": 0.146286678314209
    },
    {
      "epoch": 1.0881238155401136,
      "grad_norm": 1.7734375,
      "learning_rate": 2.183907138344915e-05,
      "loss": 0.3222,
      "mean_token_accuracy": 0.957394278049469,
      "step": 3445,
      "train/kl_loss_qwen": 0.05023112744092941,
      "train/kl_loss_r1": 0.11309699639678002,
      "train/total_kl": 0.16332812011241912
    },
    {
      "epoch": 1.0897030953885027,
      "grad_norm": 2.4375,
      "learning_rate": 2.1827226784586227e-05,
      "loss": 0.3419,
      "mean_token_accuracy": 0.9546151399612427,
      "step": 3450,
      "train/kl_loss_qwen": 0.055845167487859726,
      "train/kl_loss_r1": 0.11702072620391846,
      "train/total_kl": 0.17286589443683625
    },
    {
      "epoch": 1.0912823752368919,
      "grad_norm": 1.8671875,
      "learning_rate": 2.1815382185723312e-05,
      "loss": 0.2957,
      "mean_token_accuracy": 0.9627759575843811,
      "step": 3455,
      "train/kl_loss_qwen": 0.039809861406683925,
      "train/kl_loss_r1": 0.1049619235098362,
      "train/total_kl": 0.1447717845439911
    },
    {
      "epoch": 1.0928616550852812,
      "grad_norm": 1.6640625,
      "learning_rate": 2.180353758686039e-05,
      "loss": 0.2962,
      "mean_token_accuracy": 0.9652452349662781,
      "step": 3460,
      "train/kl_loss_qwen": 0.04810229651629925,
      "train/kl_loss_r1": 0.11666161864995957,
      "train/total_kl": 0.16476391553878783
    },
    {
      "epoch": 1.0944409349336703,
      "grad_norm": 2.15625,
      "learning_rate": 2.1791692987997476e-05,
      "loss": 0.3273,
      "mean_token_accuracy": 0.9613466382026672,
      "step": 3465,
      "train/kl_loss_qwen": 0.053610211983323094,
      "train/kl_loss_r1": 0.11490081995725632,
      "train/total_kl": 0.16851103007793428
    },
    {
      "epoch": 1.0960202147820595,
      "grad_norm": 3.1875,
      "learning_rate": 2.1779848389134555e-05,
      "loss": 0.2702,
      "mean_token_accuracy": 0.966912817955017,
      "step": 3470,
      "train/kl_loss_qwen": 0.0380069088190794,
      "train/kl_loss_r1": 0.10550742000341415,
      "train/total_kl": 0.14351432770490646
    },
    {
      "epoch": 1.0975994946304486,
      "grad_norm": 4.34375,
      "learning_rate": 2.1768003790271637e-05,
      "loss": 0.306,
      "mean_token_accuracy": 0.9612569808959961,
      "step": 3475,
      "train/kl_loss_qwen": 0.04255845062434673,
      "train/kl_loss_r1": 0.12031332403421402,
      "train/total_kl": 0.16287177056074142
    },
    {
      "epoch": 1.0991787744788377,
      "grad_norm": 2.625,
      "learning_rate": 2.175615919140872e-05,
      "loss": 0.2921,
      "mean_token_accuracy": 0.9667530417442322,
      "step": 3480,
      "train/kl_loss_qwen": 0.03182385582476854,
      "train/kl_loss_r1": 0.11894596815109253,
      "train/total_kl": 0.15076982527971267
    },
    {
      "epoch": 1.1007580543272268,
      "grad_norm": 2.15625,
      "learning_rate": 2.17443145925458e-05,
      "loss": 0.295,
      "mean_token_accuracy": 0.9644334793090821,
      "step": 3485,
      "train/kl_loss_qwen": 0.03507038224488497,
      "train/kl_loss_r1": 0.11512345522642135,
      "train/total_kl": 0.15019383281469345
    },
    {
      "epoch": 1.102337334175616,
      "grad_norm": 2.578125,
      "learning_rate": 2.173246999368288e-05,
      "loss": 0.2754,
      "mean_token_accuracy": 0.9640510082244873,
      "step": 3490,
      "train/kl_loss_qwen": 0.04499439224600792,
      "train/kl_loss_r1": 0.08394441902637481,
      "train/total_kl": 0.12893881052732467
    },
    {
      "epoch": 1.103916614024005,
      "grad_norm": 3.421875,
      "learning_rate": 2.1720625394819965e-05,
      "loss": 0.2955,
      "mean_token_accuracy": 0.9587385177612304,
      "step": 3495,
      "train/kl_loss_qwen": 0.035064952075481416,
      "train/kl_loss_r1": 0.1001215249300003,
      "train/total_kl": 0.13518647998571395
    },
    {
      "epoch": 1.1054958938723942,
      "grad_norm": 2.46875,
      "learning_rate": 2.1708780795957043e-05,
      "loss": 0.2767,
      "mean_token_accuracy": 0.9638733506202698,
      "step": 3500,
      "train/kl_loss_qwen": 0.03718983046710491,
      "train/kl_loss_r1": 0.10934945195913315,
      "train/total_kl": 0.14653928428888321
    },
    {
      "epoch": 1.1070751737207833,
      "grad_norm": 3.390625,
      "learning_rate": 2.169693619709413e-05,
      "loss": 0.2955,
      "mean_token_accuracy": 0.9529334545135498,
      "step": 3505,
      "train/kl_loss_qwen": 0.03402361758053303,
      "train/kl_loss_r1": 0.09574054777622223,
      "train/total_kl": 0.1297641634941101
    },
    {
      "epoch": 1.1086544535691725,
      "grad_norm": 2.3125,
      "learning_rate": 2.1685091598231207e-05,
      "loss": 0.273,
      "mean_token_accuracy": 0.9646507024765014,
      "step": 3510,
      "train/kl_loss_qwen": 0.03930417783558369,
      "train/kl_loss_r1": 0.09652774184942245,
      "train/total_kl": 0.13583191931247712
    },
    {
      "epoch": 1.1102337334175616,
      "grad_norm": 2.703125,
      "learning_rate": 2.167324699936829e-05,
      "loss": 0.3054,
      "mean_token_accuracy": 0.9605762362480164,
      "step": 3515,
      "train/kl_loss_qwen": 0.044501972198486325,
      "train/kl_loss_r1": 0.12194885313510895,
      "train/total_kl": 0.16645082235336303
    },
    {
      "epoch": 1.1118130132659507,
      "grad_norm": 4.71875,
      "learning_rate": 2.166140240050537e-05,
      "loss": 0.2746,
      "mean_token_accuracy": 0.9585978031158447,
      "step": 3520,
      "train/kl_loss_qwen": 0.03831633441150188,
      "train/kl_loss_r1": 0.0845057874917984,
      "train/total_kl": 0.12282212227582931
    },
    {
      "epoch": 1.1133922931143398,
      "grad_norm": 2.109375,
      "learning_rate": 2.164955780164245e-05,
      "loss": 0.2877,
      "mean_token_accuracy": 0.9707334160804748,
      "step": 3525,
      "train/kl_loss_qwen": 0.03786739856004715,
      "train/kl_loss_r1": 0.12614174336194992,
      "train/total_kl": 0.16400914192199706
    },
    {
      "epoch": 1.114971572962729,
      "grad_norm": 3.5,
      "learning_rate": 2.1637713202779532e-05,
      "loss": 0.5557,
      "mean_token_accuracy": 0.934908926486969,
      "step": 3530,
      "train/kl_loss_qwen": 0.1478494830429554,
      "train/kl_loss_r1": 0.2010779485106468,
      "train/total_kl": 0.3489274367690086
    },
    {
      "epoch": 1.116550852811118,
      "grad_norm": 2.40625,
      "learning_rate": 2.1625868603916614e-05,
      "loss": 0.3106,
      "mean_token_accuracy": 0.9638702034950256,
      "step": 3535,
      "train/kl_loss_qwen": 0.04845861755311489,
      "train/kl_loss_r1": 0.11525307148694992,
      "train/total_kl": 0.16371168941259384
    },
    {
      "epoch": 1.1181301326595072,
      "grad_norm": 1.984375,
      "learning_rate": 2.1614024005053696e-05,
      "loss": 0.2508,
      "mean_token_accuracy": 0.964933168888092,
      "step": 3540,
      "train/kl_loss_qwen": 0.03258650898933411,
      "train/kl_loss_r1": 0.07448266297578812,
      "train/total_kl": 0.10706917196512222
    },
    {
      "epoch": 1.1197094125078964,
      "grad_norm": 1.671875,
      "learning_rate": 2.1602179406190778e-05,
      "loss": 0.2804,
      "mean_token_accuracy": 0.9572123527526856,
      "step": 3545,
      "train/kl_loss_qwen": 0.03612483739852905,
      "train/kl_loss_r1": 0.0931370124220848,
      "train/total_kl": 0.12926184833049775
    },
    {
      "epoch": 1.1212886923562855,
      "grad_norm": 3.90625,
      "learning_rate": 2.159033480732786e-05,
      "loss": 0.2951,
      "mean_token_accuracy": 0.9550963878631592,
      "step": 3550,
      "train/kl_loss_qwen": 0.03587968796491623,
      "train/kl_loss_r1": 0.09598866254091262,
      "train/total_kl": 0.13186834901571273
    },
    {
      "epoch": 1.1228679722046746,
      "grad_norm": 2.953125,
      "learning_rate": 2.1578490208464938e-05,
      "loss": 0.2914,
      "mean_token_accuracy": 0.9607385158538818,
      "step": 3555,
      "train/kl_loss_qwen": 0.04809648469090462,
      "train/kl_loss_r1": 0.10686130821704865,
      "train/total_kl": 0.15495778918266295
    },
    {
      "epoch": 1.1244472520530637,
      "grad_norm": 1.9765625,
      "learning_rate": 2.1566645609602024e-05,
      "loss": 0.2934,
      "mean_token_accuracy": 0.9654027938842773,
      "step": 3560,
      "train/kl_loss_qwen": 0.04420047178864479,
      "train/kl_loss_r1": 0.10753481686115265,
      "train/total_kl": 0.15173528641462325
    },
    {
      "epoch": 1.1260265319014529,
      "grad_norm": 1.9296875,
      "learning_rate": 2.1554801010739102e-05,
      "loss": 0.6994,
      "mean_token_accuracy": 0.9156603336334228,
      "step": 3565,
      "train/kl_loss_qwen": 0.12923896722495556,
      "train/kl_loss_r1": 0.2775208577513695,
      "train/total_kl": 0.40675983130931853
    },
    {
      "epoch": 1.127605811749842,
      "grad_norm": 2.796875,
      "learning_rate": 2.1542956411876184e-05,
      "loss": 0.298,
      "mean_token_accuracy": 0.9500582814216614,
      "step": 3570,
      "train/kl_loss_qwen": 0.026421890780329704,
      "train/kl_loss_r1": 0.08942816630005837,
      "train/total_kl": 0.11585005670785904
    },
    {
      "epoch": 1.1291850915982313,
      "grad_norm": 2.3125,
      "learning_rate": 2.1531111813013266e-05,
      "loss": 0.2872,
      "mean_token_accuracy": 0.9600000381469727,
      "step": 3575,
      "train/kl_loss_qwen": 0.03754439949989319,
      "train/kl_loss_r1": 0.08909772783517837,
      "train/total_kl": 0.1266421288251877
    },
    {
      "epoch": 1.1307643714466202,
      "grad_norm": 2.03125,
      "learning_rate": 2.1519267214150348e-05,
      "loss": 0.2977,
      "mean_token_accuracy": 0.9649434804916381,
      "step": 3580,
      "train/kl_loss_qwen": 0.03301920332014561,
      "train/kl_loss_r1": 0.11933399438858032,
      "train/total_kl": 0.1523531973361969
    },
    {
      "epoch": 1.1323436512950096,
      "grad_norm": 1.828125,
      "learning_rate": 2.150742261528743e-05,
      "loss": 0.5124,
      "mean_token_accuracy": 0.9374000906944275,
      "step": 3585,
      "train/kl_loss_qwen": 0.10806241966784,
      "train/kl_loss_r1": 0.19721032977104186,
      "train/total_kl": 0.3052727609872818
    },
    {
      "epoch": 1.1339229311433987,
      "grad_norm": 1.796875,
      "learning_rate": 2.1495578016424512e-05,
      "loss": 0.2412,
      "mean_token_accuracy": 0.9704242706298828,
      "step": 3590,
      "train/kl_loss_qwen": 0.03136105835437775,
      "train/kl_loss_r1": 0.08542935848236084,
      "train/total_kl": 0.11679041683673859
    },
    {
      "epoch": 1.1355022109917878,
      "grad_norm": 2.9375,
      "learning_rate": 2.148373341756159e-05,
      "loss": 0.3057,
      "mean_token_accuracy": 0.9646150469779968,
      "step": 3595,
      "train/kl_loss_qwen": 0.04854829162359238,
      "train/kl_loss_r1": 0.11359124630689621,
      "train/total_kl": 0.16213953942060472
    },
    {
      "epoch": 1.137081490840177,
      "grad_norm": 2.921875,
      "learning_rate": 2.1471888818698676e-05,
      "loss": 0.2786,
      "mean_token_accuracy": 0.9557646870613098,
      "step": 3600,
      "train/kl_loss_qwen": 0.030383011512458324,
      "train/kl_loss_r1": 0.08102664649486542,
      "train/total_kl": 0.11140965819358825
    },
    {
      "epoch": 1.138660770688566,
      "grad_norm": 2.140625,
      "learning_rate": 2.1460044219835755e-05,
      "loss": 0.3269,
      "mean_token_accuracy": 0.9659072518348694,
      "step": 3605,
      "train/kl_loss_qwen": 0.04963378608226776,
      "train/kl_loss_r1": 0.14646055400371552,
      "train/total_kl": 0.19609434008598328
    },
    {
      "epoch": 1.1402400505369552,
      "grad_norm": 2.40625,
      "learning_rate": 2.1448199620972837e-05,
      "loss": 0.2976,
      "mean_token_accuracy": 0.9553672671318054,
      "step": 3610,
      "train/kl_loss_qwen": 0.046119637042284015,
      "train/kl_loss_r1": 0.09191006273031235,
      "train/total_kl": 0.13802969753742217
    },
    {
      "epoch": 1.1418193303853443,
      "grad_norm": 4.4375,
      "learning_rate": 2.143635502210992e-05,
      "loss": 0.3106,
      "mean_token_accuracy": 0.9574012041091919,
      "step": 3615,
      "train/kl_loss_qwen": 0.03254223205149174,
      "train/kl_loss_r1": 0.10116714090108872,
      "train/total_kl": 0.1337093695998192
    },
    {
      "epoch": 1.1433986102337335,
      "grad_norm": 2.21875,
      "learning_rate": 2.1424510423247e-05,
      "loss": 0.319,
      "mean_token_accuracy": 0.9568902850151062,
      "step": 3620,
      "train/kl_loss_qwen": 0.03041066601872444,
      "train/kl_loss_r1": 0.10932347625494003,
      "train/total_kl": 0.13973414152860641
    },
    {
      "epoch": 1.1449778900821226,
      "grad_norm": 1.7265625,
      "learning_rate": 2.1412665824384083e-05,
      "loss": 0.2825,
      "mean_token_accuracy": 0.9614522933959961,
      "step": 3625,
      "train/kl_loss_qwen": 0.03775167278945446,
      "train/kl_loss_r1": 0.0922322228550911,
      "train/total_kl": 0.12998389601707458
    },
    {
      "epoch": 1.1465571699305117,
      "grad_norm": 3.421875,
      "learning_rate": 2.1400821225521165e-05,
      "loss": 0.2851,
      "mean_token_accuracy": 0.9583948969841003,
      "step": 3630,
      "train/kl_loss_qwen": 0.03049721010029316,
      "train/kl_loss_r1": 0.08776312693953514,
      "train/total_kl": 0.11826033592224121
    },
    {
      "epoch": 1.1481364497789008,
      "grad_norm": 2.671875,
      "learning_rate": 2.1388976626658243e-05,
      "loss": 0.2614,
      "mean_token_accuracy": 0.9601945877075195,
      "step": 3635,
      "train/kl_loss_qwen": 0.0306374104693532,
      "train/kl_loss_r1": 0.09380595982074738,
      "train/total_kl": 0.12444336861371993
    },
    {
      "epoch": 1.14971572962729,
      "grad_norm": 1.71875,
      "learning_rate": 2.137713202779533e-05,
      "loss": 0.2495,
      "mean_token_accuracy": 0.9682716846466064,
      "step": 3640,
      "train/kl_loss_qwen": 0.039378654584288596,
      "train/kl_loss_r1": 0.0876434899866581,
      "train/total_kl": 0.12702214419841767
    },
    {
      "epoch": 1.151295009475679,
      "grad_norm": 2.046875,
      "learning_rate": 2.1365287428932407e-05,
      "loss": 0.2538,
      "mean_token_accuracy": 0.9663286328315734,
      "step": 3645,
      "train/kl_loss_qwen": 0.028822072595357896,
      "train/kl_loss_r1": 0.08815541788935662,
      "train/total_kl": 0.11697749197483062
    },
    {
      "epoch": 1.1528742893240682,
      "grad_norm": 2.15625,
      "learning_rate": 2.1353442830069486e-05,
      "loss": 0.2854,
      "mean_token_accuracy": 0.9580360412597656,
      "step": 3650,
      "train/kl_loss_qwen": 0.02588200829923153,
      "train/kl_loss_r1": 0.10519228652119636,
      "train/total_kl": 0.1310742937028408
    },
    {
      "epoch": 1.1544535691724573,
      "grad_norm": 2.328125,
      "learning_rate": 2.134159823120657e-05,
      "loss": 0.2544,
      "mean_token_accuracy": 0.9616833209991456,
      "step": 3655,
      "train/kl_loss_qwen": 0.02709028646349907,
      "train/kl_loss_r1": 0.08839359283447265,
      "train/total_kl": 0.11548388004302979
    },
    {
      "epoch": 1.1560328490208465,
      "grad_norm": 2.96875,
      "learning_rate": 2.132975363234365e-05,
      "loss": 0.2725,
      "mean_token_accuracy": 0.9706298470497131,
      "step": 3660,
      "train/kl_loss_qwen": 0.03371439352631569,
      "train/kl_loss_r1": 0.11454200446605682,
      "train/total_kl": 0.14825639724731446
    },
    {
      "epoch": 1.1576121288692356,
      "grad_norm": 3.21875,
      "learning_rate": 2.1317909033480735e-05,
      "loss": 0.2717,
      "mean_token_accuracy": 0.9636993527412414,
      "step": 3665,
      "train/kl_loss_qwen": 0.030815909802913665,
      "train/kl_loss_r1": 0.10840495079755783,
      "train/total_kl": 0.13922085911035537
    },
    {
      "epoch": 1.1591914087176247,
      "grad_norm": 3.796875,
      "learning_rate": 2.1306064434617814e-05,
      "loss": 0.3331,
      "mean_token_accuracy": 0.9651695013046264,
      "step": 3670,
      "train/kl_loss_qwen": 0.07103004679083824,
      "train/kl_loss_r1": 0.13202133327722548,
      "train/total_kl": 0.20305137932300568
    },
    {
      "epoch": 1.1607706885660138,
      "grad_norm": 2.359375,
      "learning_rate": 2.1294219835754895e-05,
      "loss": 0.5799,
      "mean_token_accuracy": 0.9461172699928284,
      "step": 3675,
      "train/kl_loss_qwen": 0.1307822000235319,
      "train/kl_loss_r1": 0.2560261957347393,
      "train/total_kl": 0.38680839687585833
    },
    {
      "epoch": 1.162349968414403,
      "grad_norm": 10.3125,
      "learning_rate": 2.1282375236891977e-05,
      "loss": 0.3642,
      "mean_token_accuracy": 0.9577633142471313,
      "step": 3680,
      "train/kl_loss_qwen": 0.0828935842961073,
      "train/kl_loss_r1": 0.13264236971735954,
      "train/total_kl": 0.2155359596014023
    },
    {
      "epoch": 1.163929248262792,
      "grad_norm": 2.09375,
      "learning_rate": 2.127053063802906e-05,
      "loss": 0.2643,
      "mean_token_accuracy": 0.9691606521606445,
      "step": 3685,
      "train/kl_loss_qwen": 0.0338790450245142,
      "train/kl_loss_r1": 0.11214188635349273,
      "train/total_kl": 0.14602093100547792
    },
    {
      "epoch": 1.1655085281111812,
      "grad_norm": 2.703125,
      "learning_rate": 2.125868603916614e-05,
      "loss": 0.2645,
      "mean_token_accuracy": 0.9650710463523865,
      "step": 3690,
      "train/kl_loss_qwen": 0.03991342522203922,
      "train/kl_loss_r1": 0.08417360335588456,
      "train/total_kl": 0.12408702969551086
    },
    {
      "epoch": 1.1670878079595703,
      "grad_norm": 2.390625,
      "learning_rate": 2.1246841440303223e-05,
      "loss": 0.3159,
      "mean_token_accuracy": 0.9625391840934754,
      "step": 3695,
      "train/kl_loss_qwen": 0.05647692456841469,
      "train/kl_loss_r1": 0.11663422882556915,
      "train/total_kl": 0.1731111526489258
    },
    {
      "epoch": 1.1686670878079597,
      "grad_norm": 2.96875,
      "learning_rate": 2.1234996841440302e-05,
      "loss": 0.2723,
      "mean_token_accuracy": 0.966354238986969,
      "step": 3700,
      "train/kl_loss_qwen": 0.03781019113957882,
      "train/kl_loss_r1": 0.1015682876110077,
      "train/total_kl": 0.1393784761428833
    },
    {
      "epoch": 1.1702463676563486,
      "grad_norm": 5.28125,
      "learning_rate": 2.1223152242577387e-05,
      "loss": 0.2966,
      "mean_token_accuracy": 0.9596840381622315,
      "step": 3705,
      "train/kl_loss_qwen": 0.04450843334197998,
      "train/kl_loss_r1": 0.11082954555749894,
      "train/total_kl": 0.15533798187971115
    },
    {
      "epoch": 1.171825647504738,
      "grad_norm": 2.453125,
      "learning_rate": 2.1211307643714466e-05,
      "loss": 0.2434,
      "mean_token_accuracy": 0.9707170367240906,
      "step": 3710,
      "train/kl_loss_qwen": 0.03319402188062668,
      "train/kl_loss_r1": 0.09616471081972122,
      "train/total_kl": 0.1293587327003479
    },
    {
      "epoch": 1.173404927353127,
      "grad_norm": 5.875,
      "learning_rate": 2.1199463044851548e-05,
      "loss": 0.2896,
      "mean_token_accuracy": 0.962960398197174,
      "step": 3715,
      "train/kl_loss_qwen": 0.03843155056238175,
      "train/kl_loss_r1": 0.10904917269945144,
      "train/total_kl": 0.14748072177171706
    },
    {
      "epoch": 1.1749842072015162,
      "grad_norm": 2.03125,
      "learning_rate": 2.118761844598863e-05,
      "loss": 0.235,
      "mean_token_accuracy": 0.9651673913002015,
      "step": 3720,
      "train/kl_loss_qwen": 0.029756038263440133,
      "train/kl_loss_r1": 0.07365513294935226,
      "train/total_kl": 0.10341117084026337
    },
    {
      "epoch": 1.1765634870499053,
      "grad_norm": 3.46875,
      "learning_rate": 2.1175773847125712e-05,
      "loss": 0.2733,
      "mean_token_accuracy": 0.9547190070152283,
      "step": 3725,
      "train/kl_loss_qwen": 0.03257370255887508,
      "train/kl_loss_r1": 0.0814715638756752,
      "train/total_kl": 0.1140452653169632
    },
    {
      "epoch": 1.1781427668982944,
      "grad_norm": 2.578125,
      "learning_rate": 2.1163929248262794e-05,
      "loss": 0.3165,
      "mean_token_accuracy": 0.9596776127815246,
      "step": 3730,
      "train/kl_loss_qwen": 0.04954876154661179,
      "train/kl_loss_r1": 0.11090894192457199,
      "train/total_kl": 0.16045770347118377
    },
    {
      "epoch": 1.1797220467466836,
      "grad_norm": 3.59375,
      "learning_rate": 2.1152084649399876e-05,
      "loss": 0.3087,
      "mean_token_accuracy": 0.9630552172660828,
      "step": 3735,
      "train/kl_loss_qwen": 0.04691775999963284,
      "train/kl_loss_r1": 0.1150983601808548,
      "train/total_kl": 0.16201612055301667
    },
    {
      "epoch": 1.1813013265950727,
      "grad_norm": 2.9375,
      "learning_rate": 2.1140240050536954e-05,
      "loss": 0.3318,
      "mean_token_accuracy": 0.9595719456672669,
      "step": 3740,
      "train/kl_loss_qwen": 0.04608200415968895,
      "train/kl_loss_r1": 0.129837204515934,
      "train/total_kl": 0.17591920793056487
    },
    {
      "epoch": 1.1828806064434618,
      "grad_norm": 2.296875,
      "learning_rate": 2.112839545167404e-05,
      "loss": 0.3115,
      "mean_token_accuracy": 0.9651949167251587,
      "step": 3745,
      "train/kl_loss_qwen": 0.040735474228858946,
      "train/kl_loss_r1": 0.12630822211503984,
      "train/total_kl": 0.1670436978340149
    },
    {
      "epoch": 1.184459886291851,
      "grad_norm": 1.8046875,
      "learning_rate": 2.111655085281112e-05,
      "loss": 0.288,
      "mean_token_accuracy": 0.9534114241600037,
      "step": 3750,
      "train/kl_loss_qwen": 0.02823464758694172,
      "train/kl_loss_r1": 0.07885456308722497,
      "train/total_kl": 0.10708920955657959
    },
    {
      "epoch": 1.18603916614024,
      "grad_norm": 4.71875,
      "learning_rate": 2.11047062539482e-05,
      "loss": 0.3246,
      "mean_token_accuracy": 0.957908546924591,
      "step": 3755,
      "train/kl_loss_qwen": 0.060125382244586946,
      "train/kl_loss_r1": 0.10833717584609985,
      "train/total_kl": 0.1684625566005707
    },
    {
      "epoch": 1.1876184459886292,
      "grad_norm": 3.140625,
      "learning_rate": 2.1092861655085282e-05,
      "loss": 0.2555,
      "mean_token_accuracy": 0.9636377573013306,
      "step": 3760,
      "train/kl_loss_qwen": 0.03596960641443729,
      "train/kl_loss_r1": 0.08298444673418999,
      "train/total_kl": 0.11895405203104019
    },
    {
      "epoch": 1.1891977258370183,
      "grad_norm": 3.0,
      "learning_rate": 2.1081017056222364e-05,
      "loss": 0.2982,
      "mean_token_accuracy": 0.9670576572418212,
      "step": 3765,
      "train/kl_loss_qwen": 0.048814769834280014,
      "train/kl_loss_r1": 0.10927798300981521,
      "train/total_kl": 0.15809275209903717
    },
    {
      "epoch": 1.1907770056854075,
      "grad_norm": 2.78125,
      "learning_rate": 2.1069172457359446e-05,
      "loss": 0.3142,
      "mean_token_accuracy": 0.9517574191093445,
      "step": 3770,
      "train/kl_loss_qwen": 0.036698464676737784,
      "train/kl_loss_r1": 0.09780600666999817,
      "train/total_kl": 0.13450446873903274
    },
    {
      "epoch": 1.1923562855337966,
      "grad_norm": 2.765625,
      "learning_rate": 2.1057327858496525e-05,
      "loss": 0.4333,
      "mean_token_accuracy": 0.9424691319465637,
      "step": 3775,
      "train/kl_loss_qwen": 0.08101829178631306,
      "train/kl_loss_r1": 0.16993974298238754,
      "train/total_kl": 0.2509580358862877
    },
    {
      "epoch": 1.1939355653821857,
      "grad_norm": 2.59375,
      "learning_rate": 2.1045483259633607e-05,
      "loss": 0.2912,
      "mean_token_accuracy": 0.963418436050415,
      "step": 3780,
      "train/kl_loss_qwen": 0.04738987237215042,
      "train/kl_loss_r1": 0.10172914192080498,
      "train/total_kl": 0.1491190165281296
    },
    {
      "epoch": 1.1955148452305748,
      "grad_norm": 2.578125,
      "learning_rate": 2.103363866077069e-05,
      "loss": 0.2588,
      "mean_token_accuracy": 0.9596487522125244,
      "step": 3785,
      "train/kl_loss_qwen": 0.029794510453939438,
      "train/kl_loss_r1": 0.08854113668203353,
      "train/total_kl": 0.11833564639091491
    },
    {
      "epoch": 1.197094125078964,
      "grad_norm": 1.96875,
      "learning_rate": 2.102179406190777e-05,
      "loss": 0.3002,
      "mean_token_accuracy": 0.9551068425178528,
      "step": 3790,
      "train/kl_loss_qwen": 0.03841959685087204,
      "train/kl_loss_r1": 0.08790680915117263,
      "train/total_kl": 0.12632640451192856
    },
    {
      "epoch": 1.198673404927353,
      "grad_norm": 1.9140625,
      "learning_rate": 2.100994946304485e-05,
      "loss": 0.2877,
      "mean_token_accuracy": 0.9573887467384339,
      "step": 3795,
      "train/kl_loss_qwen": 0.03675683550536633,
      "train/kl_loss_r1": 0.08911373391747475,
      "train/total_kl": 0.12587056905031205
    },
    {
      "epoch": 1.2002526847757422,
      "grad_norm": 1.4296875,
      "learning_rate": 2.0998104864181935e-05,
      "loss": 0.2865,
      "mean_token_accuracy": 0.966989803314209,
      "step": 3800,
      "train/kl_loss_qwen": 0.03910309672355652,
      "train/kl_loss_r1": 0.10924217030405999,
      "train/total_kl": 0.14834526777267457
    },
    {
      "epoch": 1.2018319646241313,
      "grad_norm": 1.953125,
      "learning_rate": 2.0986260265319013e-05,
      "loss": 0.2751,
      "mean_token_accuracy": 0.9634509205818176,
      "step": 3805,
      "train/kl_loss_qwen": 0.03598447479307652,
      "train/kl_loss_r1": 0.09990407302975654,
      "train/total_kl": 0.13588854819536209
    },
    {
      "epoch": 1.2034112444725205,
      "grad_norm": 2.609375,
      "learning_rate": 2.09744156664561e-05,
      "loss": 0.28,
      "mean_token_accuracy": 0.957117748260498,
      "step": 3810,
      "train/kl_loss_qwen": 0.02870373837649822,
      "train/kl_loss_r1": 0.08631468117237091,
      "train/total_kl": 0.11501841992139816
    },
    {
      "epoch": 1.2049905243209096,
      "grad_norm": 3.890625,
      "learning_rate": 2.0962571067593177e-05,
      "loss": 0.3373,
      "mean_token_accuracy": 0.9646350383758545,
      "step": 3815,
      "train/kl_loss_qwen": 0.05395791456103325,
      "train/kl_loss_r1": 0.14362111389636995,
      "train/total_kl": 0.19757902920246123
    },
    {
      "epoch": 1.2065698041692987,
      "grad_norm": 3.0,
      "learning_rate": 2.095072646873026e-05,
      "loss": 0.2931,
      "mean_token_accuracy": 0.9643635988235474,
      "step": 3820,
      "train/kl_loss_qwen": 0.03925386928021908,
      "train/kl_loss_r1": 0.11481977701187134,
      "train/total_kl": 0.15407364517450334
    },
    {
      "epoch": 1.208149084017688,
      "grad_norm": 6.21875,
      "learning_rate": 2.093888186986734e-05,
      "loss": 0.3318,
      "mean_token_accuracy": 0.9572894334793091,
      "step": 3825,
      "train/kl_loss_qwen": 0.06238350383937359,
      "train/kl_loss_r1": 0.110661780834198,
      "train/total_kl": 0.17304528057575225
    },
    {
      "epoch": 1.209728363866077,
      "grad_norm": 2.40625,
      "learning_rate": 2.0927037271004423e-05,
      "loss": 0.2821,
      "mean_token_accuracy": 0.9538995385169983,
      "step": 3830,
      "train/kl_loss_qwen": 0.025270987302064896,
      "train/kl_loss_r1": 0.08616796731948853,
      "train/total_kl": 0.11143895536661148
    },
    {
      "epoch": 1.2113076437144663,
      "grad_norm": 1.984375,
      "learning_rate": 2.0915192672141502e-05,
      "loss": 0.3021,
      "mean_token_accuracy": 0.9607720494270324,
      "step": 3835,
      "train/kl_loss_qwen": 0.04856106713414192,
      "train/kl_loss_r1": 0.10678209215402604,
      "train/total_kl": 0.155343160033226
    },
    {
      "epoch": 1.2128869235628554,
      "grad_norm": 2.234375,
      "learning_rate": 2.0903348073278587e-05,
      "loss": 0.2718,
      "mean_token_accuracy": 0.9585612893104554,
      "step": 3840,
      "train/kl_loss_qwen": 0.03490914385765791,
      "train/kl_loss_r1": 0.08338826522231102,
      "train/total_kl": 0.11829740703105926
    },
    {
      "epoch": 1.2144662034112446,
      "grad_norm": 2.421875,
      "learning_rate": 2.0891503474415666e-05,
      "loss": 0.2769,
      "mean_token_accuracy": 0.963493287563324,
      "step": 3845,
      "train/kl_loss_qwen": 0.03962191976606846,
      "train/kl_loss_r1": 0.10003078281879425,
      "train/total_kl": 0.13965270519256592
    },
    {
      "epoch": 1.2160454832596337,
      "grad_norm": 2.5,
      "learning_rate": 2.087965887555275e-05,
      "loss": 0.2987,
      "mean_token_accuracy": 0.960833752155304,
      "step": 3850,
      "train/kl_loss_qwen": 0.04381469190120697,
      "train/kl_loss_r1": 0.11142218261957168,
      "train/total_kl": 0.15523687303066253
    },
    {
      "epoch": 1.2176247631080228,
      "grad_norm": 2.296875,
      "learning_rate": 2.086781427668983e-05,
      "loss": 0.335,
      "mean_token_accuracy": 0.9518808007240296,
      "step": 3855,
      "train/kl_loss_qwen": 0.05659799315035343,
      "train/kl_loss_r1": 0.10317001044750214,
      "train/total_kl": 0.1597680017352104
    },
    {
      "epoch": 1.219204042956412,
      "grad_norm": 2.859375,
      "learning_rate": 2.085596967782691e-05,
      "loss": 0.2978,
      "mean_token_accuracy": 0.9504059076309204,
      "step": 3860,
      "train/kl_loss_qwen": 0.03449515588581562,
      "train/kl_loss_r1": 0.07964060232043266,
      "train/total_kl": 0.11413575857877731
    },
    {
      "epoch": 1.220783322804801,
      "grad_norm": 3.015625,
      "learning_rate": 2.0844125078963994e-05,
      "loss": 0.2818,
      "mean_token_accuracy": 0.9577629446983338,
      "step": 3865,
      "train/kl_loss_qwen": 0.04673681538552046,
      "train/kl_loss_r1": 0.08776122331619263,
      "train/total_kl": 0.13449803739786148
    },
    {
      "epoch": 1.2223626026531902,
      "grad_norm": 3.546875,
      "learning_rate": 2.0832280480101076e-05,
      "loss": 0.3006,
      "mean_token_accuracy": 0.9496965527534484,
      "step": 3870,
      "train/kl_loss_qwen": 0.02971831038594246,
      "train/kl_loss_r1": 0.09044577106833458,
      "train/total_kl": 0.12016408145427704
    },
    {
      "epoch": 1.2239418825015793,
      "grad_norm": 2.875,
      "learning_rate": 2.0820435881238154e-05,
      "loss": 0.3,
      "mean_token_accuracy": 0.9631696343421936,
      "step": 3875,
      "train/kl_loss_qwen": 0.037123050540685654,
      "train/kl_loss_r1": 0.11020882129669189,
      "train/total_kl": 0.14733187556266786
    },
    {
      "epoch": 1.2255211623499684,
      "grad_norm": 6.28125,
      "learning_rate": 2.080859128237524e-05,
      "loss": 0.341,
      "mean_token_accuracy": 0.955898118019104,
      "step": 3880,
      "train/kl_loss_qwen": 0.046445660293102264,
      "train/kl_loss_r1": 0.12775392234325408,
      "train/total_kl": 0.1741995856165886
    },
    {
      "epoch": 1.2271004421983576,
      "grad_norm": 1.703125,
      "learning_rate": 2.0796746683512318e-05,
      "loss": 0.284,
      "mean_token_accuracy": 0.959259307384491,
      "step": 3885,
      "train/kl_loss_qwen": 0.03919241987168789,
      "train/kl_loss_r1": 0.09726635962724686,
      "train/total_kl": 0.13645878285169602
    },
    {
      "epoch": 1.2286797220467467,
      "grad_norm": 2.46875,
      "learning_rate": 2.0784902084649403e-05,
      "loss": 0.2468,
      "mean_token_accuracy": 0.9625432133674622,
      "step": 3890,
      "train/kl_loss_qwen": 0.032852708175778386,
      "train/kl_loss_r1": 0.0763906329870224,
      "train/total_kl": 0.10924333930015565
    },
    {
      "epoch": 1.2302590018951358,
      "grad_norm": 2.5,
      "learning_rate": 2.0773057485786482e-05,
      "loss": 0.3135,
      "mean_token_accuracy": 0.967643141746521,
      "step": 3895,
      "train/kl_loss_qwen": 0.049361502379179,
      "train/kl_loss_r1": 0.13993418365716934,
      "train/total_kl": 0.1892956867814064
    },
    {
      "epoch": 1.231838281743525,
      "grad_norm": 2.171875,
      "learning_rate": 2.076121288692356e-05,
      "loss": 0.247,
      "mean_token_accuracy": 0.9626578092575073,
      "step": 3900,
      "train/kl_loss_qwen": 0.025497753731906415,
      "train/kl_loss_r1": 0.07994414269924163,
      "train/total_kl": 0.10544189661741257
    },
    {
      "epoch": 1.233417561591914,
      "grad_norm": 1.8125,
      "learning_rate": 2.0749368288060646e-05,
      "loss": 0.2505,
      "mean_token_accuracy": 0.9670869946479798,
      "step": 3905,
      "train/kl_loss_qwen": 0.025854425504803658,
      "train/kl_loss_r1": 0.10026284158229828,
      "train/total_kl": 0.1261172652244568
    },
    {
      "epoch": 1.2349968414403032,
      "grad_norm": 2.140625,
      "learning_rate": 2.0737523689197725e-05,
      "loss": 0.2489,
      "mean_token_accuracy": 0.9643745183944702,
      "step": 3910,
      "train/kl_loss_qwen": 0.037583792582154274,
      "train/kl_loss_r1": 0.08230764120817184,
      "train/total_kl": 0.11989143192768097
    },
    {
      "epoch": 1.2365761212886923,
      "grad_norm": 2.265625,
      "learning_rate": 2.0725679090334807e-05,
      "loss": 0.3073,
      "mean_token_accuracy": 0.9501969695091248,
      "step": 3915,
      "train/kl_loss_qwen": 0.03903764672577381,
      "train/kl_loss_r1": 0.10195644199848175,
      "train/total_kl": 0.1409940868616104
    },
    {
      "epoch": 1.2381554011370814,
      "grad_norm": 2.546875,
      "learning_rate": 2.071383449147189e-05,
      "loss": 0.3,
      "mean_token_accuracy": 0.9631790399551392,
      "step": 3920,
      "train/kl_loss_qwen": 0.03652351684868336,
      "train/kl_loss_r1": 0.10743649899959565,
      "train/total_kl": 0.14396001398563385
    },
    {
      "epoch": 1.2397346809854706,
      "grad_norm": 3.4375,
      "learning_rate": 2.070198989260897e-05,
      "loss": 0.3056,
      "mean_token_accuracy": 0.9646584153175354,
      "step": 3925,
      "train/kl_loss_qwen": 0.049883519485592845,
      "train/kl_loss_r1": 0.1171154409646988,
      "train/total_kl": 0.16699896156787872
    },
    {
      "epoch": 1.2413139608338597,
      "grad_norm": 2.8125,
      "learning_rate": 2.0690145293746052e-05,
      "loss": 0.3053,
      "mean_token_accuracy": 0.9655882596969605,
      "step": 3930,
      "train/kl_loss_qwen": 0.051488610357046126,
      "train/kl_loss_r1": 0.11833555102348328,
      "train/total_kl": 0.16982415914535523
    },
    {
      "epoch": 1.2428932406822488,
      "grad_norm": 1.625,
      "learning_rate": 2.0678300694883134e-05,
      "loss": 0.2743,
      "mean_token_accuracy": 0.9592962026596069,
      "step": 3935,
      "train/kl_loss_qwen": 0.039613519608974454,
      "train/kl_loss_r1": 0.09539609998464585,
      "train/total_kl": 0.13500962108373643
    },
    {
      "epoch": 1.244472520530638,
      "grad_norm": 1.796875,
      "learning_rate": 2.0666456096020213e-05,
      "loss": 0.2652,
      "mean_token_accuracy": 0.9555374145507812,
      "step": 3940,
      "train/kl_loss_qwen": 0.0228148864582181,
      "train/kl_loss_r1": 0.07506953105330468,
      "train/total_kl": 0.0978844165802002
    },
    {
      "epoch": 1.246051800379027,
      "grad_norm": 2.34375,
      "learning_rate": 2.06546114971573e-05,
      "loss": 0.4539,
      "mean_token_accuracy": 0.9589102149009705,
      "step": 3945,
      "train/kl_loss_qwen": 0.11135747171938419,
      "train/kl_loss_r1": 0.1950139433145523,
      "train/total_kl": 0.3063714176416397
    },
    {
      "epoch": 1.2476310802274164,
      "grad_norm": 1.7421875,
      "learning_rate": 2.0642766898294377e-05,
      "loss": 0.2905,
      "mean_token_accuracy": 0.9580713033676147,
      "step": 3950,
      "train/kl_loss_qwen": 0.038376640900969504,
      "train/kl_loss_r1": 0.09500187039375305,
      "train/total_kl": 0.13337851017713548
    },
    {
      "epoch": 1.2492103600758053,
      "grad_norm": 9.5625,
      "learning_rate": 2.063092229943146e-05,
      "loss": 0.4104,
      "mean_token_accuracy": 0.9464051008224488,
      "step": 3955,
      "train/kl_loss_qwen": 0.07086837626993656,
      "train/kl_loss_r1": 0.1600823298096657,
      "train/total_kl": 0.2309507116675377
    },
    {
      "epoch": 1.2507896399241947,
      "grad_norm": 2.046875,
      "learning_rate": 2.061907770056854e-05,
      "loss": 0.2874,
      "mean_token_accuracy": 0.9618003010749817,
      "step": 3960,
      "train/kl_loss_qwen": 0.04541564583778381,
      "train/kl_loss_r1": 0.09606570079922676,
      "train/total_kl": 0.14148134887218475
    },
    {
      "epoch": 1.2523689197725836,
      "grad_norm": 2.171875,
      "learning_rate": 2.0607233101705623e-05,
      "loss": 0.2901,
      "mean_token_accuracy": 0.9656534552574157,
      "step": 3965,
      "train/kl_loss_qwen": 0.04666389897465706,
      "train/kl_loss_r1": 0.09956516027450561,
      "train/total_kl": 0.14622905850410461
    },
    {
      "epoch": 1.253948199620973,
      "grad_norm": 2.5625,
      "learning_rate": 2.0595388502842705e-05,
      "loss": 0.2842,
      "mean_token_accuracy": 0.9577082514762878,
      "step": 3970,
      "train/kl_loss_qwen": 0.029145017638802527,
      "train/kl_loss_r1": 0.09465509355068207,
      "train/total_kl": 0.12380011230707169
    },
    {
      "epoch": 1.255527479469362,
      "grad_norm": 1.9765625,
      "learning_rate": 2.0583543903979787e-05,
      "loss": 0.2619,
      "mean_token_accuracy": 0.9576513051986695,
      "step": 3975,
      "train/kl_loss_qwen": 0.028220358490943908,
      "train/kl_loss_r1": 0.07727068066596984,
      "train/total_kl": 0.10549104064702988
    },
    {
      "epoch": 1.2571067593177512,
      "grad_norm": 2.859375,
      "learning_rate": 2.0571699305116865e-05,
      "loss": 0.3577,
      "mean_token_accuracy": 0.9487927436828614,
      "step": 3980,
      "train/kl_loss_qwen": 0.05058789849281311,
      "train/kl_loss_r1": 0.11585463136434555,
      "train/total_kl": 0.16644253432750702
    },
    {
      "epoch": 1.2586860391661403,
      "grad_norm": 1.9609375,
      "learning_rate": 2.055985470625395e-05,
      "loss": 0.272,
      "mean_token_accuracy": 0.9590231537818908,
      "step": 3985,
      "train/kl_loss_qwen": 0.03666531555354595,
      "train/kl_loss_r1": 0.08895185813307763,
      "train/total_kl": 0.1256171755492687
    },
    {
      "epoch": 1.2602653190145294,
      "grad_norm": 2.296875,
      "learning_rate": 2.054801010739103e-05,
      "loss": 0.2321,
      "mean_token_accuracy": 0.969104266166687,
      "step": 3990,
      "train/kl_loss_qwen": 0.03159029670059681,
      "train/kl_loss_r1": 0.08211300075054169,
      "train/total_kl": 0.11370329856872559
    },
    {
      "epoch": 1.2618445988629186,
      "grad_norm": 3.953125,
      "learning_rate": 2.0536165508528115e-05,
      "loss": 0.3347,
      "mean_token_accuracy": 0.9568923592567444,
      "step": 3995,
      "train/kl_loss_qwen": 0.04716028012335301,
      "train/kl_loss_r1": 0.11806784570217133,
      "train/total_kl": 0.16522812396287917
    },
    {
      "epoch": 1.2634238787113077,
      "grad_norm": 1.953125,
      "learning_rate": 2.0524320909665193e-05,
      "loss": 0.2852,
      "mean_token_accuracy": 0.9578576803207397,
      "step": 4000,
      "train/kl_loss_qwen": 0.032302948832511905,
      "train/kl_loss_r1": 0.0879747711122036,
      "train/total_kl": 0.12027771919965743
    },
    {
      "epoch": 1.2650031585596968,
      "grad_norm": 3.15625,
      "learning_rate": 2.0512476310802275e-05,
      "loss": 0.2921,
      "mean_token_accuracy": 0.9632822394371032,
      "step": 4005,
      "train/kl_loss_qwen": 0.042978232726454735,
      "train/kl_loss_r1": 0.10112003684043884,
      "train/total_kl": 0.1440982699394226
    },
    {
      "epoch": 1.266582438408086,
      "grad_norm": 1.953125,
      "learning_rate": 2.0500631711939357e-05,
      "loss": 0.2872,
      "mean_token_accuracy": 0.9644989490509033,
      "step": 4010,
      "train/kl_loss_qwen": 0.04569738656282425,
      "train/kl_loss_r1": 0.10294595807790756,
      "train/total_kl": 0.1486433431506157
    },
    {
      "epoch": 1.268161718256475,
      "grad_norm": 3.21875,
      "learning_rate": 2.048878711307644e-05,
      "loss": 0.3582,
      "mean_token_accuracy": 0.9405353665351868,
      "step": 4015,
      "train/kl_loss_qwen": 0.037640055269002916,
      "train/kl_loss_r1": 0.10135683938860893,
      "train/total_kl": 0.1389968976378441
    },
    {
      "epoch": 1.2697409981048642,
      "grad_norm": 2.234375,
      "learning_rate": 2.0476942514213518e-05,
      "loss": 0.3828,
      "mean_token_accuracy": 0.9472923159599305,
      "step": 4020,
      "train/kl_loss_qwen": 0.06062595061957836,
      "train/kl_loss_r1": 0.13352832719683647,
      "train/total_kl": 0.19415427297353743
    },
    {
      "epoch": 1.2713202779532533,
      "grad_norm": 5.65625,
      "learning_rate": 2.04650979153506e-05,
      "loss": 0.3335,
      "mean_token_accuracy": 0.9409282684326172,
      "step": 4025,
      "train/kl_loss_qwen": 0.02850969396531582,
      "train/kl_loss_r1": 0.08018419668078422,
      "train/total_kl": 0.1086938887834549
    },
    {
      "epoch": 1.2728995578016424,
      "grad_norm": 2.75,
      "learning_rate": 2.0453253316487682e-05,
      "loss": 0.2591,
      "mean_token_accuracy": 0.964190137386322,
      "step": 4030,
      "train/kl_loss_qwen": 0.03263176791369915,
      "train/kl_loss_r1": 0.09316381216049194,
      "train/total_kl": 0.12579558193683624
    },
    {
      "epoch": 1.2744788376500316,
      "grad_norm": 2.359375,
      "learning_rate": 2.0441408717624764e-05,
      "loss": 0.3166,
      "mean_token_accuracy": 0.9586971402168274,
      "step": 4035,
      "train/kl_loss_qwen": 0.04917265325784683,
      "train/kl_loss_r1": 0.10404181703925133,
      "train/total_kl": 0.1532144695520401
    },
    {
      "epoch": 1.2760581174984207,
      "grad_norm": 1.71875,
      "learning_rate": 2.0429564118761846e-05,
      "loss": 0.2605,
      "mean_token_accuracy": 0.9649809002876282,
      "step": 4040,
      "train/kl_loss_qwen": 0.03006675038486719,
      "train/kl_loss_r1": 0.08686754107475281,
      "train/total_kl": 0.11693429350852966
    },
    {
      "epoch": 1.2776373973468098,
      "grad_norm": 2.28125,
      "learning_rate": 2.0417719519898924e-05,
      "loss": 0.283,
      "mean_token_accuracy": 0.9640228748321533,
      "step": 4045,
      "train/kl_loss_qwen": 0.04162821993231773,
      "train/kl_loss_r1": 0.10463643819093704,
      "train/total_kl": 0.14626465737819672
    },
    {
      "epoch": 1.279216677195199,
      "grad_norm": 2.796875,
      "learning_rate": 2.040587492103601e-05,
      "loss": 0.3338,
      "mean_token_accuracy": 0.9540187954902649,
      "step": 4050,
      "train/kl_loss_qwen": 0.04564488455653191,
      "train/kl_loss_r1": 0.10761993229389191,
      "train/total_kl": 0.15326481461524963
    },
    {
      "epoch": 1.280795957043588,
      "grad_norm": 2.453125,
      "learning_rate": 2.0394030322173088e-05,
      "loss": 0.3248,
      "mean_token_accuracy": 0.9632748365402222,
      "step": 4055,
      "train/kl_loss_qwen": 0.04668295979499817,
      "train/kl_loss_r1": 0.13776273280382156,
      "train/total_kl": 0.18444569259881974
    },
    {
      "epoch": 1.2823752368919772,
      "grad_norm": 1.859375,
      "learning_rate": 2.038218572331017e-05,
      "loss": 0.2539,
      "mean_token_accuracy": 0.9648740410804748,
      "step": 4060,
      "train/kl_loss_qwen": 0.02704647071659565,
      "train/kl_loss_r1": 0.08119939640164375,
      "train/total_kl": 0.10824586749076844
    },
    {
      "epoch": 1.2839545167403663,
      "grad_norm": 2.796875,
      "learning_rate": 2.0370341124447252e-05,
      "loss": 0.3053,
      "mean_token_accuracy": 0.9621927738189697,
      "step": 4065,
      "train/kl_loss_qwen": 0.05116235166788101,
      "train/kl_loss_r1": 0.10800577998161316,
      "train/total_kl": 0.15916813313961028
    },
    {
      "epoch": 1.2855337965887554,
      "grad_norm": 5.5625,
      "learning_rate": 2.0358496525584334e-05,
      "loss": 0.2891,
      "mean_token_accuracy": 0.9673664927482605,
      "step": 4070,
      "train/kl_loss_qwen": 0.043795743212103846,
      "train/kl_loss_r1": 0.11381452083587647,
      "train/total_kl": 0.1576102629303932
    },
    {
      "epoch": 1.2871130764371448,
      "grad_norm": 2.75,
      "learning_rate": 2.0346651926721416e-05,
      "loss": 0.3123,
      "mean_token_accuracy": 0.9537769198417664,
      "step": 4075,
      "train/kl_loss_qwen": 0.03820946775376797,
      "train/kl_loss_r1": 0.10412556007504463,
      "train/total_kl": 0.14233502894639968
    },
    {
      "epoch": 1.2886923562855337,
      "grad_norm": 1.7578125,
      "learning_rate": 2.0334807327858498e-05,
      "loss": 0.3168,
      "mean_token_accuracy": 0.9606137752532959,
      "step": 4080,
      "train/kl_loss_qwen": 0.039086834713816646,
      "train/kl_loss_r1": 0.13003924190998079,
      "train/total_kl": 0.16912607699632645
    },
    {
      "epoch": 1.290271636133923,
      "grad_norm": 2.390625,
      "learning_rate": 2.0322962728995577e-05,
      "loss": 0.2749,
      "mean_token_accuracy": 0.9689087867736816,
      "step": 4085,
      "train/kl_loss_qwen": 0.04294727817177772,
      "train/kl_loss_r1": 0.10212097316980362,
      "train/total_kl": 0.14506825059652328
    },
    {
      "epoch": 1.291850915982312,
      "grad_norm": 2.015625,
      "learning_rate": 2.0311118130132662e-05,
      "loss": 0.2929,
      "mean_token_accuracy": 0.9664088129997254,
      "step": 4090,
      "train/kl_loss_qwen": 0.044792049750685695,
      "train/kl_loss_r1": 0.10982347279787064,
      "train/total_kl": 0.15461552441120147
    },
    {
      "epoch": 1.2934301958307013,
      "grad_norm": 7.46875,
      "learning_rate": 2.029927353126974e-05,
      "loss": 0.3477,
      "mean_token_accuracy": 0.9524661064147949,
      "step": 4095,
      "train/kl_loss_qwen": 0.05431187953799963,
      "train/kl_loss_r1": 0.11841184869408608,
      "train/total_kl": 0.1727237269282341
    },
    {
      "epoch": 1.2950094756790904,
      "grad_norm": 12.8125,
      "learning_rate": 2.0287428932406823e-05,
      "loss": 0.5786,
      "mean_token_accuracy": 0.9333510875701905,
      "step": 4100,
      "train/kl_loss_qwen": 0.13343321904540062,
      "train/kl_loss_r1": 0.23060454279184342,
      "train/total_kl": 0.36403776705265045
    },
    {
      "epoch": 1.2965887555274795,
      "grad_norm": 2.3125,
      "learning_rate": 2.0275584333543905e-05,
      "loss": 0.3197,
      "mean_token_accuracy": 0.957388722896576,
      "step": 4105,
      "train/kl_loss_qwen": 0.040214926935732365,
      "train/kl_loss_r1": 0.1146023616194725,
      "train/total_kl": 0.15481728613376616
    },
    {
      "epoch": 1.2981680353758687,
      "grad_norm": 3.375,
      "learning_rate": 2.0263739734680987e-05,
      "loss": 0.3079,
      "mean_token_accuracy": 0.957176411151886,
      "step": 4110,
      "train/kl_loss_qwen": 0.04377785883843899,
      "train/kl_loss_r1": 0.10500650852918625,
      "train/total_kl": 0.14878436774015427
    },
    {
      "epoch": 1.2997473152242578,
      "grad_norm": 1.7265625,
      "learning_rate": 2.025189513581807e-05,
      "loss": 0.2973,
      "mean_token_accuracy": 0.9548516988754272,
      "step": 4115,
      "train/kl_loss_qwen": 0.039373290911316874,
      "train/kl_loss_r1": 0.09774724543094634,
      "train/total_kl": 0.13712053447961808
    },
    {
      "epoch": 1.301326595072647,
      "grad_norm": 3.734375,
      "learning_rate": 2.024005053695515e-05,
      "loss": 0.2683,
      "mean_token_accuracy": 0.9640845060348511,
      "step": 4120,
      "train/kl_loss_qwen": 0.03353948723524809,
      "train/kl_loss_r1": 0.09169473350048066,
      "train/total_kl": 0.12523422241210938
    },
    {
      "epoch": 1.302905874921036,
      "grad_norm": 5.5625,
      "learning_rate": 2.022820593809223e-05,
      "loss": 0.3161,
      "mean_token_accuracy": 0.9540834784507751,
      "step": 4125,
      "train/kl_loss_qwen": 0.04068617094308138,
      "train/kl_loss_r1": 0.10543893724679947,
      "train/total_kl": 0.14612510949373245
    },
    {
      "epoch": 1.3044851547694252,
      "grad_norm": 3.84375,
      "learning_rate": 2.0216361339229314e-05,
      "loss": 0.3157,
      "mean_token_accuracy": 0.9515178799629211,
      "step": 4130,
      "train/kl_loss_qwen": 0.03423355743288994,
      "train/kl_loss_r1": 0.09498334378004074,
      "train/total_kl": 0.12921690344810485
    },
    {
      "epoch": 1.3060644346178143,
      "grad_norm": 2.28125,
      "learning_rate": 2.0204516740366393e-05,
      "loss": 0.2778,
      "mean_token_accuracy": 0.958819580078125,
      "step": 4135,
      "train/kl_loss_qwen": 0.036541692167520526,
      "train/kl_loss_r1": 0.08442789241671562,
      "train/total_kl": 0.12096958458423615
    },
    {
      "epoch": 1.3076437144662034,
      "grad_norm": 3.421875,
      "learning_rate": 2.0192672141503475e-05,
      "loss": 0.396,
      "mean_token_accuracy": 0.9536806344985962,
      "step": 4140,
      "train/kl_loss_qwen": 0.0870095957070589,
      "train/kl_loss_r1": 0.14437087625265121,
      "train/total_kl": 0.2313804656267166
    },
    {
      "epoch": 1.3092229943145925,
      "grad_norm": 2.390625,
      "learning_rate": 2.0180827542640557e-05,
      "loss": 0.3062,
      "mean_token_accuracy": 0.9553277611732482,
      "step": 4145,
      "train/kl_loss_qwen": 0.04446752071380615,
      "train/kl_loss_r1": 0.09840887188911437,
      "train/total_kl": 0.14287639111280442
    },
    {
      "epoch": 1.3108022741629817,
      "grad_norm": 5.0,
      "learning_rate": 2.0168982943777636e-05,
      "loss": 0.2789,
      "mean_token_accuracy": 0.963948130607605,
      "step": 4150,
      "train/kl_loss_qwen": 0.036158537678420546,
      "train/kl_loss_r1": 0.09134115278720856,
      "train/total_kl": 0.1274996891617775
    },
    {
      "epoch": 1.3123815540113708,
      "grad_norm": 1.9453125,
      "learning_rate": 2.015713834491472e-05,
      "loss": 0.279,
      "mean_token_accuracy": 0.9522378444671631,
      "step": 4155,
      "train/kl_loss_qwen": 0.02749242540448904,
      "train/kl_loss_r1": 0.08026158958673477,
      "train/total_kl": 0.10775401592254638
    },
    {
      "epoch": 1.31396083385976,
      "grad_norm": 2.78125,
      "learning_rate": 2.01452937460518e-05,
      "loss": 0.327,
      "mean_token_accuracy": 0.9519093751907348,
      "step": 4160,
      "train/kl_loss_qwen": 0.04271635785698891,
      "train/kl_loss_r1": 0.10957479178905487,
      "train/total_kl": 0.15229115039110183
    },
    {
      "epoch": 1.315540113708149,
      "grad_norm": 2.234375,
      "learning_rate": 2.013344914718888e-05,
      "loss": 0.269,
      "mean_token_accuracy": 0.9679793357849121,
      "step": 4165,
      "train/kl_loss_qwen": 0.03936488553881645,
      "train/kl_loss_r1": 0.10950679928064347,
      "train/total_kl": 0.14887168407440185
    },
    {
      "epoch": 1.3171193935565382,
      "grad_norm": 3.171875,
      "learning_rate": 2.0121604548325963e-05,
      "loss": 0.2908,
      "mean_token_accuracy": 0.954381275177002,
      "step": 4170,
      "train/kl_loss_qwen": 0.036414249986410144,
      "train/kl_loss_r1": 0.08545588180422783,
      "train/total_kl": 0.12187013179063796
    },
    {
      "epoch": 1.3186986734049273,
      "grad_norm": 2.078125,
      "learning_rate": 2.0109759949463045e-05,
      "loss": 0.2664,
      "mean_token_accuracy": 0.9635087370872497,
      "step": 4175,
      "train/kl_loss_qwen": 0.025441995076835155,
      "train/kl_loss_r1": 0.08935362249612808,
      "train/total_kl": 0.11479561775922775
    },
    {
      "epoch": 1.3202779532533164,
      "grad_norm": 2.625,
      "learning_rate": 2.0097915350600124e-05,
      "loss": 0.5055,
      "mean_token_accuracy": 0.9414554953575134,
      "step": 4180,
      "train/kl_loss_qwen": 0.11672492176294327,
      "train/kl_loss_r1": 0.18320467472076415,
      "train/total_kl": 0.29992959648370743
    },
    {
      "epoch": 1.3218572331017056,
      "grad_norm": 2.078125,
      "learning_rate": 2.008607075173721e-05,
      "loss": 0.2832,
      "mean_token_accuracy": 0.9608171105384826,
      "step": 4185,
      "train/kl_loss_qwen": 0.03656622171401978,
      "train/kl_loss_r1": 0.0980958603322506,
      "train/total_kl": 0.13466208130121232
    },
    {
      "epoch": 1.3234365129500947,
      "grad_norm": 3.546875,
      "learning_rate": 2.0074226152874288e-05,
      "loss": 0.3408,
      "mean_token_accuracy": 0.9526473760604859,
      "step": 4190,
      "train/kl_loss_qwen": 0.04521455653011799,
      "train/kl_loss_r1": 0.11633948981761932,
      "train/total_kl": 0.16155404597520828
    },
    {
      "epoch": 1.3250157927984838,
      "grad_norm": 2.109375,
      "learning_rate": 2.0062381554011373e-05,
      "loss": 0.2694,
      "mean_token_accuracy": 0.9693362832069397,
      "step": 4195,
      "train/kl_loss_qwen": 0.03555840626358986,
      "train/kl_loss_r1": 0.11213024705648422,
      "train/total_kl": 0.1476886510848999
    },
    {
      "epoch": 1.3265950726468732,
      "grad_norm": 2.21875,
      "learning_rate": 2.0050536955148452e-05,
      "loss": 0.3121,
      "mean_token_accuracy": 0.9682808995246888,
      "step": 4200,
      "train/kl_loss_qwen": 0.057590844482183455,
      "train/kl_loss_r1": 0.12797443121671676,
      "train/total_kl": 0.1855652779340744
    },
    {
      "epoch": 1.328174352495262,
      "grad_norm": 1.5703125,
      "learning_rate": 2.0038692356285534e-05,
      "loss": 0.2978,
      "mean_token_accuracy": 0.9685616612434387,
      "step": 4205,
      "train/kl_loss_qwen": 0.050023014470934865,
      "train/kl_loss_r1": 0.12326873540878296,
      "train/total_kl": 0.17329175323247908
    },
    {
      "epoch": 1.3297536323436514,
      "grad_norm": 2.15625,
      "learning_rate": 2.0026847757422616e-05,
      "loss": 0.2804,
      "mean_token_accuracy": 0.9688499093055725,
      "step": 4210,
      "train/kl_loss_qwen": 0.0400221886113286,
      "train/kl_loss_r1": 0.11215937212109565,
      "train/total_kl": 0.1521815612912178
    },
    {
      "epoch": 1.3313329121920403,
      "grad_norm": 2.4375,
      "learning_rate": 2.0015003158559698e-05,
      "loss": 0.2611,
      "mean_token_accuracy": 0.9616554379463196,
      "step": 4215,
      "train/kl_loss_qwen": 0.025236261636018754,
      "train/kl_loss_r1": 0.10048776119947433,
      "train/total_kl": 0.12572402358055115
    },
    {
      "epoch": 1.3329121920404297,
      "grad_norm": 3.734375,
      "learning_rate": 2.000315855969678e-05,
      "loss": 0.4637,
      "mean_token_accuracy": 0.931055200099945,
      "step": 4220,
      "train/kl_loss_qwen": 0.07661504447460174,
      "train/kl_loss_r1": 0.14725451618433,
      "train/total_kl": 0.22386955916881562
    },
    {
      "epoch": 1.3344914718888188,
      "grad_norm": 1.4375,
      "learning_rate": 1.9991313960833862e-05,
      "loss": 0.2572,
      "mean_token_accuracy": 0.9517345666885376,
      "step": 4225,
      "train/kl_loss_qwen": 0.021934659406542777,
      "train/kl_loss_r1": 0.06652961447834968,
      "train/total_kl": 0.08846427500247955
    },
    {
      "epoch": 1.336070751737208,
      "grad_norm": 2.1875,
      "learning_rate": 1.997946936197094e-05,
      "loss": 0.2855,
      "mean_token_accuracy": 0.9702908635139466,
      "step": 4230,
      "train/kl_loss_qwen": 0.05210568904876709,
      "train/kl_loss_r1": 0.10639957189559937,
      "train/total_kl": 0.15850525796413423
    },
    {
      "epoch": 1.337650031585597,
      "grad_norm": 2.078125,
      "learning_rate": 1.9967624763108026e-05,
      "loss": 0.4368,
      "mean_token_accuracy": 0.9394311428070068,
      "step": 4235,
      "train/kl_loss_qwen": 0.08486703373491764,
      "train/kl_loss_r1": 0.14749891608953475,
      "train/total_kl": 0.23236595690250397
    },
    {
      "epoch": 1.3392293114339862,
      "grad_norm": 5.25,
      "learning_rate": 1.9955780164245104e-05,
      "loss": 0.2688,
      "mean_token_accuracy": 0.9602742552757263,
      "step": 4240,
      "train/kl_loss_qwen": 0.03661784492433071,
      "train/kl_loss_r1": 0.08285044431686402,
      "train/total_kl": 0.11946829110383987
    },
    {
      "epoch": 1.3408085912823753,
      "grad_norm": 4.90625,
      "learning_rate": 1.9943935565382186e-05,
      "loss": 0.2992,
      "mean_token_accuracy": 0.9569628953933715,
      "step": 4245,
      "train/kl_loss_qwen": 0.03641990199685097,
      "train/kl_loss_r1": 0.09626553058624268,
      "train/total_kl": 0.13268543183803558
    },
    {
      "epoch": 1.3423878711307644,
      "grad_norm": 2.265625,
      "learning_rate": 1.9932090966519268e-05,
      "loss": 0.3146,
      "mean_token_accuracy": 0.9631345868110657,
      "step": 4250,
      "train/kl_loss_qwen": 0.06555296368896961,
      "train/kl_loss_r1": 0.11172016263008118,
      "train/total_kl": 0.17727312594652175
    },
    {
      "epoch": 1.3439671509791535,
      "grad_norm": 4.15625,
      "learning_rate": 1.992024636765635e-05,
      "loss": 0.4023,
      "mean_token_accuracy": 0.9504573822021485,
      "step": 4255,
      "train/kl_loss_qwen": 0.0771070159971714,
      "train/kl_loss_r1": 0.15099695026874543,
      "train/total_kl": 0.22810396254062654
    },
    {
      "epoch": 1.3455464308275427,
      "grad_norm": 2.796875,
      "learning_rate": 1.9908401768793432e-05,
      "loss": 0.3119,
      "mean_token_accuracy": 0.9489089369773864,
      "step": 4260,
      "train/kl_loss_qwen": 0.041877875849604604,
      "train/kl_loss_r1": 0.07895587757229805,
      "train/total_kl": 0.12083375304937363
    },
    {
      "epoch": 1.3471257106759318,
      "grad_norm": 2.09375,
      "learning_rate": 1.9896557169930514e-05,
      "loss": 0.2791,
      "mean_token_accuracy": 0.9651817083358765,
      "step": 4265,
      "train/kl_loss_qwen": 0.04200858250260353,
      "train/kl_loss_r1": 0.1165824756026268,
      "train/total_kl": 0.1585910588502884
    },
    {
      "epoch": 1.348704990524321,
      "grad_norm": 2.421875,
      "learning_rate": 1.9884712571067593e-05,
      "loss": 0.2853,
      "mean_token_accuracy": 0.9582436919212342,
      "step": 4270,
      "train/kl_loss_qwen": 0.02865699790418148,
      "train/kl_loss_r1": 0.08690985292196274,
      "train/total_kl": 0.11556684970855713
    },
    {
      "epoch": 1.35028427037271,
      "grad_norm": 2.28125,
      "learning_rate": 1.9872867972204675e-05,
      "loss": 0.2573,
      "mean_token_accuracy": 0.9659136891365051,
      "step": 4275,
      "train/kl_loss_qwen": 0.03166200630366802,
      "train/kl_loss_r1": 0.08835921287536622,
      "train/total_kl": 0.12002122104167938
    },
    {
      "epoch": 1.3518635502210992,
      "grad_norm": 2.609375,
      "learning_rate": 1.9861023373341757e-05,
      "loss": 0.3067,
      "mean_token_accuracy": 0.9566686153411865,
      "step": 4280,
      "train/kl_loss_qwen": 0.04016427770256996,
      "train/kl_loss_r1": 0.10015916377305985,
      "train/total_kl": 0.140323443710804
    },
    {
      "epoch": 1.3534428300694883,
      "grad_norm": 1.7890625,
      "learning_rate": 1.9849178774478835e-05,
      "loss": 0.2593,
      "mean_token_accuracy": 0.9647830128669739,
      "step": 4285,
      "train/kl_loss_qwen": 0.027342502400279044,
      "train/kl_loss_r1": 0.09216629415750503,
      "train/total_kl": 0.1195087969303131
    },
    {
      "epoch": 1.3550221099178774,
      "grad_norm": 3.296875,
      "learning_rate": 1.983733417561592e-05,
      "loss": 0.3324,
      "mean_token_accuracy": 0.9500020146369934,
      "step": 4290,
      "train/kl_loss_qwen": 0.04656006582081318,
      "train/kl_loss_r1": 0.09922882467508316,
      "train/total_kl": 0.14578889012336732
    },
    {
      "epoch": 1.3566013897662665,
      "grad_norm": 2.25,
      "learning_rate": 1.9825489576753e-05,
      "loss": 0.3165,
      "mean_token_accuracy": 0.9653058648109436,
      "step": 4295,
      "train/kl_loss_qwen": 0.05799228735268116,
      "train/kl_loss_r1": 0.12731897532939912,
      "train/total_kl": 0.18531126230955125
    },
    {
      "epoch": 1.3581806696146557,
      "grad_norm": 2.28125,
      "learning_rate": 1.9813644977890085e-05,
      "loss": 0.3005,
      "mean_token_accuracy": 0.9613985419273376,
      "step": 4300,
      "train/kl_loss_qwen": 0.05173634216189384,
      "train/kl_loss_r1": 0.1107184797525406,
      "train/total_kl": 0.16245481967926026
    },
    {
      "epoch": 1.3597599494630448,
      "grad_norm": 2.515625,
      "learning_rate": 1.9801800379027163e-05,
      "loss": 0.2694,
      "mean_token_accuracy": 0.9557992696762085,
      "step": 4305,
      "train/kl_loss_qwen": 0.03516281321644783,
      "train/kl_loss_r1": 0.0909169889986515,
      "train/total_kl": 0.12607980370521546
    },
    {
      "epoch": 1.361339229311434,
      "grad_norm": 2.734375,
      "learning_rate": 1.9789955780164245e-05,
      "loss": 0.2968,
      "mean_token_accuracy": 0.9657358765602112,
      "step": 4310,
      "train/kl_loss_qwen": 0.048287322372198106,
      "train/kl_loss_r1": 0.11343861669301987,
      "train/total_kl": 0.1617259383201599
    },
    {
      "epoch": 1.362918509159823,
      "grad_norm": 2.28125,
      "learning_rate": 1.9778111181301327e-05,
      "loss": 0.2499,
      "mean_token_accuracy": 0.9678225874900818,
      "step": 4315,
      "train/kl_loss_qwen": 0.034582341089844704,
      "train/kl_loss_r1": 0.08827869296073913,
      "train/total_kl": 0.12286103516817093
    },
    {
      "epoch": 1.3644977890082122,
      "grad_norm": 3.046875,
      "learning_rate": 1.976626658243841e-05,
      "loss": 0.3881,
      "mean_token_accuracy": 0.9578328609466553,
      "step": 4320,
      "train/kl_loss_qwen": 0.08166200518608094,
      "train/kl_loss_r1": 0.15269118696451187,
      "train/total_kl": 0.23435320109128951
    },
    {
      "epoch": 1.3660770688566015,
      "grad_norm": 2.46875,
      "learning_rate": 1.9754421983575488e-05,
      "loss": 0.3192,
      "mean_token_accuracy": 0.9515592217445373,
      "step": 4325,
      "train/kl_loss_qwen": 0.03780734576284885,
      "train/kl_loss_r1": 0.09073201641440391,
      "train/total_kl": 0.12853935807943345
    },
    {
      "epoch": 1.3676563487049904,
      "grad_norm": 2.28125,
      "learning_rate": 1.9742577384712573e-05,
      "loss": 0.2512,
      "mean_token_accuracy": 0.9672032594680786,
      "step": 4330,
      "train/kl_loss_qwen": 0.0351453959941864,
      "train/kl_loss_r1": 0.0955880656838417,
      "train/total_kl": 0.13073346316814421
    },
    {
      "epoch": 1.3692356285533798,
      "grad_norm": 2.3125,
      "learning_rate": 1.973073278584965e-05,
      "loss": 0.3231,
      "mean_token_accuracy": 0.9608458995819091,
      "step": 4335,
      "train/kl_loss_qwen": 0.0419995054602623,
      "train/kl_loss_r1": 0.11507322490215302,
      "train/total_kl": 0.15707273185253143
    },
    {
      "epoch": 1.3708149084017687,
      "grad_norm": 1.8515625,
      "learning_rate": 1.9718888186986737e-05,
      "loss": 0.2662,
      "mean_token_accuracy": 0.9689882516860961,
      "step": 4340,
      "train/kl_loss_qwen": 0.04353465922176838,
      "train/kl_loss_r1": 0.09924794733524323,
      "train/total_kl": 0.1427826017141342
    },
    {
      "epoch": 1.372394188250158,
      "grad_norm": 4.8125,
      "learning_rate": 1.9707043588123816e-05,
      "loss": 0.2914,
      "mean_token_accuracy": 0.9579110741615295,
      "step": 4345,
      "train/kl_loss_qwen": 0.03233658708631992,
      "train/kl_loss_r1": 0.0935288853943348,
      "train/total_kl": 0.1258654698729515
    },
    {
      "epoch": 1.373973468098547,
      "grad_norm": 2.59375,
      "learning_rate": 1.9695198989260898e-05,
      "loss": 0.29,
      "mean_token_accuracy": 0.955306100845337,
      "step": 4350,
      "train/kl_loss_qwen": 0.05330001413822174,
      "train/kl_loss_r1": 0.07963560670614242,
      "train/total_kl": 0.13293562531471254
    },
    {
      "epoch": 1.3755527479469363,
      "grad_norm": 2.765625,
      "learning_rate": 1.968335439039798e-05,
      "loss": 0.2664,
      "mean_token_accuracy": 0.9642046689987183,
      "step": 4355,
      "train/kl_loss_qwen": 0.02520008310675621,
      "train/kl_loss_r1": 0.10372688621282578,
      "train/total_kl": 0.12892697155475616
    },
    {
      "epoch": 1.3771320277953254,
      "grad_norm": 2.53125,
      "learning_rate": 1.967150979153506e-05,
      "loss": 0.2841,
      "mean_token_accuracy": 0.9661792516708374,
      "step": 4360,
      "train/kl_loss_qwen": 0.053275855630636214,
      "train/kl_loss_r1": 0.10374135822057724,
      "train/total_kl": 0.1570172131061554
    },
    {
      "epoch": 1.3787113076437145,
      "grad_norm": 2.875,
      "learning_rate": 1.965966519267214e-05,
      "loss": 0.2941,
      "mean_token_accuracy": 0.9597386956214905,
      "step": 4365,
      "train/kl_loss_qwen": 0.04098556600511074,
      "train/kl_loss_r1": 0.1051072433590889,
      "train/total_kl": 0.1460928127169609
    },
    {
      "epoch": 1.3802905874921036,
      "grad_norm": 2.15625,
      "learning_rate": 1.9647820593809225e-05,
      "loss": 0.2932,
      "mean_token_accuracy": 0.9666616320610046,
      "step": 4370,
      "train/kl_loss_qwen": 0.037358499318361285,
      "train/kl_loss_r1": 0.11763607114553451,
      "train/total_kl": 0.15499457120895385
    },
    {
      "epoch": 1.3818698673404928,
      "grad_norm": 4.03125,
      "learning_rate": 1.9635975994946304e-05,
      "loss": 0.6328,
      "mean_token_accuracy": 0.9283604264259339,
      "step": 4375,
      "train/kl_loss_qwen": 0.13976683765649794,
      "train/kl_loss_r1": 0.2582694202661514,
      "train/total_kl": 0.39803625345230104
    },
    {
      "epoch": 1.383449147188882,
      "grad_norm": 2.234375,
      "learning_rate": 1.962413139608339e-05,
      "loss": 0.2647,
      "mean_token_accuracy": 0.9676759600639343,
      "step": 4380,
      "train/kl_loss_qwen": 0.03382174223661423,
      "train/kl_loss_r1": 0.09197035953402519,
      "train/total_kl": 0.1257921040058136
    },
    {
      "epoch": 1.385028427037271,
      "grad_norm": 2.25,
      "learning_rate": 1.9612286797220468e-05,
      "loss": 0.588,
      "mean_token_accuracy": 0.929291307926178,
      "step": 4385,
      "train/kl_loss_qwen": 0.12414328008890152,
      "train/kl_loss_r1": 0.23787620812654495,
      "train/total_kl": 0.3620194911956787
    },
    {
      "epoch": 1.3866077068856602,
      "grad_norm": 1.578125,
      "learning_rate": 1.960044219835755e-05,
      "loss": 0.2605,
      "mean_token_accuracy": 0.9696667432785034,
      "step": 4390,
      "train/kl_loss_qwen": 0.03725598119199276,
      "train/kl_loss_r1": 0.09581340998411178,
      "train/total_kl": 0.13306939154863356
    },
    {
      "epoch": 1.3881869867340493,
      "grad_norm": 1.46875,
      "learning_rate": 1.9588597599494632e-05,
      "loss": 0.3135,
      "mean_token_accuracy": 0.9628771781921387,
      "step": 4395,
      "train/kl_loss_qwen": 0.06378615498542786,
      "train/kl_loss_r1": 0.11797023490071297,
      "train/total_kl": 0.18175639361143112
    },
    {
      "epoch": 1.3897662665824384,
      "grad_norm": 2.96875,
      "learning_rate": 1.957675300063171e-05,
      "loss": 0.2821,
      "mean_token_accuracy": 0.9582911014556885,
      "step": 4400,
      "train/kl_loss_qwen": 0.03567362911999226,
      "train/kl_loss_r1": 0.08875415325164795,
      "train/total_kl": 0.1244277834892273
    },
    {
      "epoch": 1.3913455464308275,
      "grad_norm": 2.40625,
      "learning_rate": 1.9564908401768792e-05,
      "loss": 0.319,
      "mean_token_accuracy": 0.9610899567604065,
      "step": 4405,
      "train/kl_loss_qwen": 0.03938590213656425,
      "train/kl_loss_r1": 0.12469573915004731,
      "train/total_kl": 0.16408164352178572
    },
    {
      "epoch": 1.3929248262792167,
      "grad_norm": 3.15625,
      "learning_rate": 1.9553063802905874e-05,
      "loss": 0.2961,
      "mean_token_accuracy": 0.954804539680481,
      "step": 4410,
      "train/kl_loss_qwen": 0.03567392658442259,
      "train/kl_loss_r1": 0.0792167030274868,
      "train/total_kl": 0.11489063203334808
    },
    {
      "epoch": 1.3945041061276058,
      "grad_norm": 2.421875,
      "learning_rate": 1.9541219204042956e-05,
      "loss": 0.2706,
      "mean_token_accuracy": 0.9622756719589234,
      "step": 4415,
      "train/kl_loss_qwen": 0.040068017691373824,
      "train/kl_loss_r1": 0.09124587774276734,
      "train/total_kl": 0.1313138946890831
    },
    {
      "epoch": 1.396083385975995,
      "grad_norm": 3.015625,
      "learning_rate": 1.952937460518004e-05,
      "loss": 0.3114,
      "mean_token_accuracy": 0.9580551862716675,
      "step": 4420,
      "train/kl_loss_qwen": 0.042500798031687735,
      "train/kl_loss_r1": 0.10348447412252426,
      "train/total_kl": 0.14598526954650878
    },
    {
      "epoch": 1.397662665824384,
      "grad_norm": 3.0,
      "learning_rate": 1.951753000631712e-05,
      "loss": 0.2958,
      "mean_token_accuracy": 0.9565512061119079,
      "step": 4425,
      "train/kl_loss_qwen": 0.04066393077373505,
      "train/kl_loss_r1": 0.091982152312994,
      "train/total_kl": 0.1326460838317871
    },
    {
      "epoch": 1.3992419456727732,
      "grad_norm": 2.265625,
      "learning_rate": 1.95056854074542e-05,
      "loss": 0.2859,
      "mean_token_accuracy": 0.9606986284255982,
      "step": 4430,
      "train/kl_loss_qwen": 0.040056885406374934,
      "train/kl_loss_r1": 0.10060959756374359,
      "train/total_kl": 0.14066648185253144
    },
    {
      "epoch": 1.4008212255211623,
      "grad_norm": 1.828125,
      "learning_rate": 1.9493840808591284e-05,
      "loss": 0.2606,
      "mean_token_accuracy": 0.965421712398529,
      "step": 4435,
      "train/kl_loss_qwen": 0.032726316899061206,
      "train/kl_loss_r1": 0.09154874086380005,
      "train/total_kl": 0.12427505850791931
    },
    {
      "epoch": 1.4024005053695514,
      "grad_norm": 3.484375,
      "learning_rate": 1.9481996209728363e-05,
      "loss": 0.3206,
      "mean_token_accuracy": 0.9573492527008056,
      "step": 4440,
      "train/kl_loss_qwen": 0.049867574125528336,
      "train/kl_loss_r1": 0.12658245712518693,
      "train/total_kl": 0.1764500319957733
    },
    {
      "epoch": 1.4039797852179405,
      "grad_norm": 2.546875,
      "learning_rate": 1.9470151610865445e-05,
      "loss": 0.3126,
      "mean_token_accuracy": 0.9518569946289063,
      "step": 4445,
      "train/kl_loss_qwen": 0.034010112285614014,
      "train/kl_loss_r1": 0.09009077996015549,
      "train/total_kl": 0.12410089522600173
    },
    {
      "epoch": 1.4055590650663299,
      "grad_norm": 3.65625,
      "learning_rate": 1.9458307012002527e-05,
      "loss": 0.3502,
      "mean_token_accuracy": 0.9409806489944458,
      "step": 4450,
      "train/kl_loss_qwen": 0.039763399958610536,
      "train/kl_loss_r1": 0.10056933388113976,
      "train/total_kl": 0.14033273756504058
    },
    {
      "epoch": 1.4071383449147188,
      "grad_norm": 2.59375,
      "learning_rate": 1.944646241313961e-05,
      "loss": 0.2967,
      "mean_token_accuracy": 0.9610785841941833,
      "step": 4455,
      "train/kl_loss_qwen": 0.0404702939093113,
      "train/kl_loss_r1": 0.11293548345565796,
      "train/total_kl": 0.1534057766199112
    },
    {
      "epoch": 1.4087176247631081,
      "grad_norm": 2.625,
      "learning_rate": 1.943461781427669e-05,
      "loss": 0.2754,
      "mean_token_accuracy": 0.9595382690429688,
      "step": 4460,
      "train/kl_loss_qwen": 0.047652697563171385,
      "train/kl_loss_r1": 0.08734358251094818,
      "train/total_kl": 0.13499628156423568
    },
    {
      "epoch": 1.410296904611497,
      "grad_norm": 2.8125,
      "learning_rate": 1.9422773215413773e-05,
      "loss": 0.3022,
      "mean_token_accuracy": 0.956800389289856,
      "step": 4465,
      "train/kl_loss_qwen": 0.03387292101979256,
      "train/kl_loss_r1": 0.09566764459013939,
      "train/total_kl": 0.12954056710004808
    },
    {
      "epoch": 1.4118761844598864,
      "grad_norm": 1.8359375,
      "learning_rate": 1.941092861655085e-05,
      "loss": 0.262,
      "mean_token_accuracy": 0.9571342587471008,
      "step": 4470,
      "train/kl_loss_qwen": 0.029099275916814805,
      "train/kl_loss_r1": 0.07780294269323348,
      "train/total_kl": 0.10690221935510635
    },
    {
      "epoch": 1.4134554643082753,
      "grad_norm": 2.09375,
      "learning_rate": 1.9399084017687937e-05,
      "loss": 0.4072,
      "mean_token_accuracy": 0.9330723881721497,
      "step": 4475,
      "train/kl_loss_qwen": 0.06049734149128198,
      "train/kl_loss_r1": 0.12258148044347764,
      "train/total_kl": 0.1830788254737854
    },
    {
      "epoch": 1.4150347441566646,
      "grad_norm": 2.25,
      "learning_rate": 1.9387239418825015e-05,
      "loss": 0.2881,
      "mean_token_accuracy": 0.9603986978530884,
      "step": 4480,
      "train/kl_loss_qwen": 0.035397449135780336,
      "train/kl_loss_r1": 0.10028458833694458,
      "train/total_kl": 0.1356820359826088
    },
    {
      "epoch": 1.4166140240050538,
      "grad_norm": 1.7109375,
      "learning_rate": 1.9375394819962097e-05,
      "loss": 0.2847,
      "mean_token_accuracy": 0.9574848532676696,
      "step": 4485,
      "train/kl_loss_qwen": 0.0404916699975729,
      "train/kl_loss_r1": 0.08413771912455559,
      "train/total_kl": 0.12462938725948333
    },
    {
      "epoch": 1.4181933038534429,
      "grad_norm": 1.96875,
      "learning_rate": 1.936355022109918e-05,
      "loss": 0.2464,
      "mean_token_accuracy": 0.9625012159347535,
      "step": 4490,
      "train/kl_loss_qwen": 0.02247481346130371,
      "train/kl_loss_r1": 0.08498267456889153,
      "train/total_kl": 0.10745749026536941
    },
    {
      "epoch": 1.419772583701832,
      "grad_norm": 2.171875,
      "learning_rate": 1.935170562223626e-05,
      "loss": 0.2871,
      "mean_token_accuracy": 0.961659824848175,
      "step": 4495,
      "train/kl_loss_qwen": 0.03160826172679663,
      "train/kl_loss_r1": 0.10406721010804176,
      "train/total_kl": 0.13567547053098677
    },
    {
      "epoch": 1.4213518635502211,
      "grad_norm": 3.0625,
      "learning_rate": 1.9339861023373343e-05,
      "loss": 0.2964,
      "mean_token_accuracy": 0.9591862678527832,
      "step": 4500,
      "train/kl_loss_qwen": 0.03796242848038674,
      "train/kl_loss_r1": 0.11592971682548522,
      "train/total_kl": 0.15389214605093002
    },
    {
      "epoch": 1.4229311433986103,
      "grad_norm": 6.09375,
      "learning_rate": 1.9328016424510425e-05,
      "loss": 0.3276,
      "mean_token_accuracy": 0.9510234594345093,
      "step": 4505,
      "train/kl_loss_qwen": 0.03979124501347542,
      "train/kl_loss_r1": 0.11464240700006485,
      "train/total_kl": 0.15443365424871444
    },
    {
      "epoch": 1.4245104232469994,
      "grad_norm": 2.890625,
      "learning_rate": 1.9316171825647504e-05,
      "loss": 0.2569,
      "mean_token_accuracy": 0.9655158400535584,
      "step": 4510,
      "train/kl_loss_qwen": 0.03996074683964253,
      "train/kl_loss_r1": 0.08993579000234604,
      "train/total_kl": 0.12989653646945953
    },
    {
      "epoch": 1.4260897030953885,
      "grad_norm": 2.96875,
      "learning_rate": 1.930432722678459e-05,
      "loss": 0.324,
      "mean_token_accuracy": 0.9515695571899414,
      "step": 4515,
      "train/kl_loss_qwen": 0.04234092421829701,
      "train/kl_loss_r1": 0.09042564034461975,
      "train/total_kl": 0.13276656717061996
    },
    {
      "epoch": 1.4276689829437776,
      "grad_norm": 2.328125,
      "learning_rate": 1.9292482627921668e-05,
      "loss": 0.2698,
      "mean_token_accuracy": 0.955196762084961,
      "step": 4520,
      "train/kl_loss_qwen": 0.02975374162197113,
      "train/kl_loss_r1": 0.07815098464488983,
      "train/total_kl": 0.10790472477674484
    },
    {
      "epoch": 1.4292482627921668,
      "grad_norm": 1.4140625,
      "learning_rate": 1.928063802905875e-05,
      "loss": 0.2501,
      "mean_token_accuracy": 0.9586687445640564,
      "step": 4525,
      "train/kl_loss_qwen": 0.024233030527830123,
      "train/kl_loss_r1": 0.0806028537452221,
      "train/total_kl": 0.10483588427305221
    },
    {
      "epoch": 1.430827542640556,
      "grad_norm": 2.015625,
      "learning_rate": 1.926879343019583e-05,
      "loss": 0.2822,
      "mean_token_accuracy": 0.9608237743377686,
      "step": 4530,
      "train/kl_loss_qwen": 0.0396702840924263,
      "train/kl_loss_r1": 0.09677752256393432,
      "train/total_kl": 0.13644780814647675
    },
    {
      "epoch": 1.432406822488945,
      "grad_norm": 5.3125,
      "learning_rate": 1.925694883133291e-05,
      "loss": 0.3437,
      "mean_token_accuracy": 0.96483393907547,
      "step": 4535,
      "train/kl_loss_qwen": 0.059776357188820836,
      "train/kl_loss_r1": 0.13778969049453735,
      "train/total_kl": 0.19756605178117753
    },
    {
      "epoch": 1.4339861023373341,
      "grad_norm": 2.34375,
      "learning_rate": 1.9245104232469996e-05,
      "loss": 0.2763,
      "mean_token_accuracy": 0.9684423327445983,
      "step": 4540,
      "train/kl_loss_qwen": 0.050443168729543686,
      "train/kl_loss_r1": 0.10166644155979157,
      "train/total_kl": 0.15210961401462555
    },
    {
      "epoch": 1.4355653821857233,
      "grad_norm": 2.140625,
      "learning_rate": 1.9233259633607074e-05,
      "loss": 0.2406,
      "mean_token_accuracy": 0.9680348992347717,
      "step": 4545,
      "train/kl_loss_qwen": 0.02775859795510769,
      "train/kl_loss_r1": 0.08512718006968498,
      "train/total_kl": 0.11288577914237977
    },
    {
      "epoch": 1.4371446620341124,
      "grad_norm": 2.75,
      "learning_rate": 1.9221415034744156e-05,
      "loss": 0.2947,
      "mean_token_accuracy": 0.9532219290733337,
      "step": 4550,
      "train/kl_loss_qwen": 0.02873414009809494,
      "train/kl_loss_r1": 0.08808910548686981,
      "train/total_kl": 0.11682324707508088
    },
    {
      "epoch": 1.4387239418825015,
      "grad_norm": 1.921875,
      "learning_rate": 1.9209570435881238e-05,
      "loss": 0.2783,
      "mean_token_accuracy": 0.9701516628265381,
      "step": 4555,
      "train/kl_loss_qwen": 0.04346229620277882,
      "train/kl_loss_r1": 0.11290783733129502,
      "train/total_kl": 0.15637013167142869
    },
    {
      "epoch": 1.4403032217308906,
      "grad_norm": 2.203125,
      "learning_rate": 1.919772583701832e-05,
      "loss": 0.2364,
      "mean_token_accuracy": 0.9655145287513733,
      "step": 4560,
      "train/kl_loss_qwen": 0.03270864635705948,
      "train/kl_loss_r1": 0.07814445123076438,
      "train/total_kl": 0.11085309684276581
    },
    {
      "epoch": 1.4418825015792798,
      "grad_norm": 1.6640625,
      "learning_rate": 1.9185881238155402e-05,
      "loss": 0.2216,
      "mean_token_accuracy": 0.9680378675460816,
      "step": 4565,
      "train/kl_loss_qwen": 0.02380184978246689,
      "train/kl_loss_r1": 0.0705703467130661,
      "train/total_kl": 0.09437219798564911
    },
    {
      "epoch": 1.443461781427669,
      "grad_norm": 1.4140625,
      "learning_rate": 1.9174036639292484e-05,
      "loss": 0.2805,
      "mean_token_accuracy": 0.9621809482574463,
      "step": 4570,
      "train/kl_loss_qwen": 0.028407613188028334,
      "train/kl_loss_r1": 0.10797795504331589,
      "train/total_kl": 0.13638556897640228
    },
    {
      "epoch": 1.4450410612760582,
      "grad_norm": 1.625,
      "learning_rate": 1.9162192040429563e-05,
      "loss": 0.259,
      "mean_token_accuracy": 0.9647883296012878,
      "step": 4575,
      "train/kl_loss_qwen": 0.026179772429168226,
      "train/kl_loss_r1": 0.0966211885213852,
      "train/total_kl": 0.12280096262693405
    },
    {
      "epoch": 1.4466203411244472,
      "grad_norm": 2.4375,
      "learning_rate": 1.9150347441566648e-05,
      "loss": 0.2676,
      "mean_token_accuracy": 0.9612613797187806,
      "step": 4580,
      "train/kl_loss_qwen": 0.031196530535817148,
      "train/kl_loss_r1": 0.09539418891072274,
      "train/total_kl": 0.12659071683883666
    },
    {
      "epoch": 1.4481996209728365,
      "grad_norm": 2.25,
      "learning_rate": 1.9138502842703727e-05,
      "loss": 0.2804,
      "mean_token_accuracy": 0.9664979577064514,
      "step": 4585,
      "train/kl_loss_qwen": 0.028669359534978865,
      "train/kl_loss_r1": 0.11709880381822586,
      "train/total_kl": 0.1457681655883789
    },
    {
      "epoch": 1.4497789008212254,
      "grad_norm": 2.15625,
      "learning_rate": 1.912665824384081e-05,
      "loss": 0.2876,
      "mean_token_accuracy": 0.9589450359344482,
      "step": 4590,
      "train/kl_loss_qwen": 0.04337799940258265,
      "train/kl_loss_r1": 0.0879811130464077,
      "train/total_kl": 0.13135911226272584
    },
    {
      "epoch": 1.4513581806696148,
      "grad_norm": 1.7890625,
      "learning_rate": 1.911481364497789e-05,
      "loss": 0.4448,
      "mean_token_accuracy": 0.9457563877105712,
      "step": 4595,
      "train/kl_loss_qwen": 0.09390295185148716,
      "train/kl_loss_r1": 0.1724180057644844,
      "train/total_kl": 0.2663209572434425
    },
    {
      "epoch": 1.4529374605180037,
      "grad_norm": 3.234375,
      "learning_rate": 1.9102969046114972e-05,
      "loss": 0.3036,
      "mean_token_accuracy": 0.9560048103332519,
      "step": 4600,
      "train/kl_loss_qwen": 0.03621966987848282,
      "train/kl_loss_r1": 0.0987351268529892,
      "train/total_kl": 0.134954796731472
    },
    {
      "epoch": 1.454516740366393,
      "grad_norm": 1.9453125,
      "learning_rate": 1.9091124447252054e-05,
      "loss": 0.3064,
      "mean_token_accuracy": 0.9470838904380798,
      "step": 4605,
      "train/kl_loss_qwen": 0.023329467698931693,
      "train/kl_loss_r1": 0.07891639471054077,
      "train/total_kl": 0.1022458627820015
    },
    {
      "epoch": 1.4560960202147821,
      "grad_norm": 2.1875,
      "learning_rate": 1.9079279848389136e-05,
      "loss": 0.3031,
      "mean_token_accuracy": 0.9595552921295166,
      "step": 4610,
      "train/kl_loss_qwen": 0.038874147459864616,
      "train/kl_loss_r1": 0.11031136065721511,
      "train/total_kl": 0.14918550699949265
    },
    {
      "epoch": 1.4576753000631713,
      "grad_norm": 2.5625,
      "learning_rate": 1.9067435249526215e-05,
      "loss": 0.2591,
      "mean_token_accuracy": 0.962556266784668,
      "step": 4615,
      "train/kl_loss_qwen": 0.033522413671016695,
      "train/kl_loss_r1": 0.08451971262693406,
      "train/total_kl": 0.11804212480783463
    },
    {
      "epoch": 1.4592545799115604,
      "grad_norm": 2.53125,
      "learning_rate": 1.90555906506633e-05,
      "loss": 0.3115,
      "mean_token_accuracy": 0.958387839794159,
      "step": 4620,
      "train/kl_loss_qwen": 0.05487663708627224,
      "train/kl_loss_r1": 0.10836037546396256,
      "train/total_kl": 0.16323701441287994
    },
    {
      "epoch": 1.4608338597599495,
      "grad_norm": 2.359375,
      "learning_rate": 1.904374605180038e-05,
      "loss": 0.3112,
      "mean_token_accuracy": 0.952399468421936,
      "step": 4625,
      "train/kl_loss_qwen": 0.036751574650406835,
      "train/kl_loss_r1": 0.09651744067668915,
      "train/total_kl": 0.13326901644468309
    },
    {
      "epoch": 1.4624131396083386,
      "grad_norm": 2.78125,
      "learning_rate": 1.903190145293746e-05,
      "loss": 0.3296,
      "mean_token_accuracy": 0.9574221491813659,
      "step": 4630,
      "train/kl_loss_qwen": 0.0442658543586731,
      "train/kl_loss_r1": 0.11011170446872712,
      "train/total_kl": 0.1543775588274002
    },
    {
      "epoch": 1.4639924194567278,
      "grad_norm": 1.9453125,
      "learning_rate": 1.9020056854074543e-05,
      "loss": 0.2735,
      "mean_token_accuracy": 0.96877601146698,
      "step": 4635,
      "train/kl_loss_qwen": 0.04193931221961975,
      "train/kl_loss_r1": 0.10275930315256118,
      "train/total_kl": 0.14469861686229707
    },
    {
      "epoch": 1.4655716993051169,
      "grad_norm": 2.734375,
      "learning_rate": 1.9008212255211625e-05,
      "loss": 0.3245,
      "mean_token_accuracy": 0.9508514881134034,
      "step": 4640,
      "train/kl_loss_qwen": 0.045588191226124765,
      "train/kl_loss_r1": 0.11133930459618568,
      "train/total_kl": 0.15692749619483948
    },
    {
      "epoch": 1.467150979153506,
      "grad_norm": 2.296875,
      "learning_rate": 1.8996367656348707e-05,
      "loss": 0.3051,
      "mean_token_accuracy": 0.9631740331649781,
      "step": 4645,
      "train/kl_loss_qwen": 0.05524761751294136,
      "train/kl_loss_r1": 0.11701711863279343,
      "train/total_kl": 0.1722647339105606
    },
    {
      "epoch": 1.4687302590018951,
      "grad_norm": 2.09375,
      "learning_rate": 1.8984523057485785e-05,
      "loss": 0.3026,
      "mean_token_accuracy": 0.94182368516922,
      "step": 4650,
      "train/kl_loss_qwen": 0.023445361852645875,
      "train/kl_loss_r1": 0.06398442462086677,
      "train/total_kl": 0.08742978572845458
    },
    {
      "epoch": 1.4703095388502843,
      "grad_norm": 2.5625,
      "learning_rate": 1.8972678458622867e-05,
      "loss": 0.3006,
      "mean_token_accuracy": 0.9629556059837341,
      "step": 4655,
      "train/kl_loss_qwen": 0.04100862182676792,
      "train/kl_loss_r1": 0.11832314133644103,
      "train/total_kl": 0.15933176279067993
    },
    {
      "epoch": 1.4718888186986734,
      "grad_norm": 2.125,
      "learning_rate": 1.896083385975995e-05,
      "loss": 0.2962,
      "mean_token_accuracy": 0.9616992592811584,
      "step": 4660,
      "train/kl_loss_qwen": 0.038057423755526544,
      "train/kl_loss_r1": 0.12337739020586014,
      "train/total_kl": 0.1614348143339157
    },
    {
      "epoch": 1.4734680985470625,
      "grad_norm": 3.0625,
      "learning_rate": 1.894898926089703e-05,
      "loss": 0.3157,
      "mean_token_accuracy": 0.9610432982444763,
      "step": 4665,
      "train/kl_loss_qwen": 0.047209664806723596,
      "train/kl_loss_r1": 0.11898027956485749,
      "train/total_kl": 0.1661899447441101
    },
    {
      "epoch": 1.4750473783954516,
      "grad_norm": 1.59375,
      "learning_rate": 1.893714466203411e-05,
      "loss": 0.3095,
      "mean_token_accuracy": 0.9631926774978637,
      "step": 4670,
      "train/kl_loss_qwen": 0.046643438190221785,
      "train/kl_loss_r1": 0.12047898471355438,
      "train/total_kl": 0.16712241917848586
    },
    {
      "epoch": 1.4766266582438408,
      "grad_norm": 2.40625,
      "learning_rate": 1.8925300063171195e-05,
      "loss": 0.3013,
      "mean_token_accuracy": 0.9559717178344727,
      "step": 4675,
      "train/kl_loss_qwen": 0.028236668184399606,
      "train/kl_loss_r1": 0.08763114362955093,
      "train/total_kl": 0.11586780995130538
    },
    {
      "epoch": 1.4782059380922299,
      "grad_norm": 1.7578125,
      "learning_rate": 1.8913455464308274e-05,
      "loss": 0.2356,
      "mean_token_accuracy": 0.9639431357383728,
      "step": 4680,
      "train/kl_loss_qwen": 0.021081164106726648,
      "train/kl_loss_r1": 0.07629505470395088,
      "train/total_kl": 0.0973762184381485
    },
    {
      "epoch": 1.479785217940619,
      "grad_norm": 3.421875,
      "learning_rate": 1.890161086544536e-05,
      "loss": 0.316,
      "mean_token_accuracy": 0.9599949002265931,
      "step": 4685,
      "train/kl_loss_qwen": 0.053360366448760035,
      "train/kl_loss_r1": 0.1247580423951149,
      "train/total_kl": 0.17811840921640396
    },
    {
      "epoch": 1.4813644977890081,
      "grad_norm": 1.859375,
      "learning_rate": 1.8889766266582438e-05,
      "loss": 0.2706,
      "mean_token_accuracy": 0.959316074848175,
      "step": 4690,
      "train/kl_loss_qwen": 0.026343658193945886,
      "train/kl_loss_r1": 0.0918030634522438,
      "train/total_kl": 0.11814672201871872
    },
    {
      "epoch": 1.4829437776373973,
      "grad_norm": 1.6953125,
      "learning_rate": 1.887792166771952e-05,
      "loss": 0.2622,
      "mean_token_accuracy": 0.9657209992408753,
      "step": 4695,
      "train/kl_loss_qwen": 0.04535595886409283,
      "train/kl_loss_r1": 0.08637072965502739,
      "train/total_kl": 0.1317266881465912
    },
    {
      "epoch": 1.4845230574857866,
      "grad_norm": 3.515625,
      "learning_rate": 1.8866077068856602e-05,
      "loss": 0.3558,
      "mean_token_accuracy": 0.9521191835403442,
      "step": 4700,
      "train/kl_loss_qwen": 0.05790814980864525,
      "train/kl_loss_r1": 0.1186299353837967,
      "train/total_kl": 0.1765380859375
    },
    {
      "epoch": 1.4861023373341755,
      "grad_norm": 2.453125,
      "learning_rate": 1.8854232469993684e-05,
      "loss": 0.2591,
      "mean_token_accuracy": 0.9664780378341675,
      "step": 4705,
      "train/kl_loss_qwen": 0.029401503130793573,
      "train/kl_loss_r1": 0.09541214406490325,
      "train/total_kl": 0.12481364905834198
    },
    {
      "epoch": 1.4876816171825649,
      "grad_norm": 2.71875,
      "learning_rate": 1.8842387871130762e-05,
      "loss": 0.2754,
      "mean_token_accuracy": 0.9662151575088501,
      "step": 4710,
      "train/kl_loss_qwen": 0.04353714622557163,
      "train/kl_loss_r1": 0.10280487835407257,
      "train/total_kl": 0.1463420256972313
    },
    {
      "epoch": 1.4892608970309538,
      "grad_norm": 2.25,
      "learning_rate": 1.8830543272267848e-05,
      "loss": 0.2611,
      "mean_token_accuracy": 0.9691568970680237,
      "step": 4715,
      "train/kl_loss_qwen": 0.03626970872282982,
      "train/kl_loss_r1": 0.09973816424608231,
      "train/total_kl": 0.13600787222385408
    },
    {
      "epoch": 1.4908401768793431,
      "grad_norm": 2.171875,
      "learning_rate": 1.8818698673404926e-05,
      "loss": 0.2925,
      "mean_token_accuracy": 0.9660473823547363,
      "step": 4720,
      "train/kl_loss_qwen": 0.050272959470748904,
      "train/kl_loss_r1": 0.10961891412734985,
      "train/total_kl": 0.15989187061786653
    },
    {
      "epoch": 1.492419456727732,
      "grad_norm": 2.515625,
      "learning_rate": 1.880685407454201e-05,
      "loss": 0.2888,
      "mean_token_accuracy": 0.9610440969467163,
      "step": 4725,
      "train/kl_loss_qwen": 0.026919344440102577,
      "train/kl_loss_r1": 0.10933726131916047,
      "train/total_kl": 0.1362566038966179
    },
    {
      "epoch": 1.4939987365761214,
      "grad_norm": 3.859375,
      "learning_rate": 1.879500947567909e-05,
      "loss": 0.2977,
      "mean_token_accuracy": 0.9589416980743408,
      "step": 4730,
      "train/kl_loss_qwen": 0.03460715226829052,
      "train/kl_loss_r1": 0.10522591918706894,
      "train/total_kl": 0.13983306884765626
    },
    {
      "epoch": 1.4955780164245105,
      "grad_norm": 1.984375,
      "learning_rate": 1.8783164876816172e-05,
      "loss": 0.2448,
      "mean_token_accuracy": 0.9633565187454224,
      "step": 4735,
      "train/kl_loss_qwen": 0.027425509691238404,
      "train/kl_loss_r1": 0.07139872536063194,
      "train/total_kl": 0.09882423430681228
    },
    {
      "epoch": 1.4971572962728996,
      "grad_norm": 2.453125,
      "learning_rate": 1.8771320277953254e-05,
      "loss": 0.2888,
      "mean_token_accuracy": 0.9608946561813354,
      "step": 4740,
      "train/kl_loss_qwen": 0.039358388632535934,
      "train/kl_loss_r1": 0.09562153220176697,
      "train/total_kl": 0.13497992157936095
    },
    {
      "epoch": 1.4987365761212887,
      "grad_norm": 4.5,
      "learning_rate": 1.8759475679090336e-05,
      "loss": 0.3381,
      "mean_token_accuracy": 0.9597094655036926,
      "step": 4745,
      "train/kl_loss_qwen": 0.06108529344201088,
      "train/kl_loss_r1": 0.13316339924931525,
      "train/total_kl": 0.1942486897110939
    },
    {
      "epoch": 1.5003158559696779,
      "grad_norm": 2.6875,
      "learning_rate": 1.8747631080227418e-05,
      "loss": 0.3044,
      "mean_token_accuracy": 0.9631342768669129,
      "step": 4750,
      "train/kl_loss_qwen": 0.053244148194789884,
      "train/kl_loss_r1": 0.11185305565595627,
      "train/total_kl": 0.16509720236063002
    },
    {
      "epoch": 1.501895135818067,
      "grad_norm": 3.046875,
      "learning_rate": 1.87357864813645e-05,
      "loss": 0.2933,
      "mean_token_accuracy": 0.9600862860679626,
      "step": 4755,
      "train/kl_loss_qwen": 0.02983510009944439,
      "train/kl_loss_r1": 0.10231602191925049,
      "train/total_kl": 0.13215111941099167
    },
    {
      "epoch": 1.5034744156664561,
      "grad_norm": 2.046875,
      "learning_rate": 1.872394188250158e-05,
      "loss": 0.2427,
      "mean_token_accuracy": 0.9619375944137574,
      "step": 4760,
      "train/kl_loss_qwen": 0.031479062885046004,
      "train/kl_loss_r1": 0.06482870429754257,
      "train/total_kl": 0.09630776792764664
    },
    {
      "epoch": 1.5050536955148452,
      "grad_norm": 3.96875,
      "learning_rate": 1.8712097283638664e-05,
      "loss": 0.3536,
      "mean_token_accuracy": 0.957589852809906,
      "step": 4765,
      "train/kl_loss_qwen": 0.05985662192106247,
      "train/kl_loss_r1": 0.13641411513090135,
      "train/total_kl": 0.1962707370519638
    },
    {
      "epoch": 1.5066329753632344,
      "grad_norm": 4.25,
      "learning_rate": 1.8700252684775743e-05,
      "loss": 0.3326,
      "mean_token_accuracy": 0.9548203825950623,
      "step": 4770,
      "train/kl_loss_qwen": 0.06138685755431652,
      "train/kl_loss_r1": 0.11700534597039222,
      "train/total_kl": 0.17839220017194748
    },
    {
      "epoch": 1.5082122552116235,
      "grad_norm": 2.140625,
      "learning_rate": 1.868840808591282e-05,
      "loss": 0.3268,
      "mean_token_accuracy": 0.9538328647613525,
      "step": 4775,
      "train/kl_loss_qwen": 0.03681586366146803,
      "train/kl_loss_r1": 0.119706629216671,
      "train/total_kl": 0.1565224915742874
    },
    {
      "epoch": 1.5097915350600126,
      "grad_norm": 2.484375,
      "learning_rate": 1.8676563487049907e-05,
      "loss": 0.3054,
      "mean_token_accuracy": 0.953168523311615,
      "step": 4780,
      "train/kl_loss_qwen": 0.03708044849336147,
      "train/kl_loss_r1": 0.10597460120916366,
      "train/total_kl": 0.1430550515651703
    },
    {
      "epoch": 1.5113708149084017,
      "grad_norm": 2.765625,
      "learning_rate": 1.8664718888186985e-05,
      "loss": 0.3293,
      "mean_token_accuracy": 0.9601280808448791,
      "step": 4785,
      "train/kl_loss_qwen": 0.06147680133581161,
      "train/kl_loss_r1": 0.11685838848352433,
      "train/total_kl": 0.17833518832921982
    },
    {
      "epoch": 1.5129500947567909,
      "grad_norm": 3.84375,
      "learning_rate": 1.865287428932407e-05,
      "loss": 0.2877,
      "mean_token_accuracy": 0.9544150233268738,
      "step": 4790,
      "train/kl_loss_qwen": 0.043860527127981185,
      "train/kl_loss_r1": 0.08360740020871163,
      "train/total_kl": 0.12746792733669282
    },
    {
      "epoch": 1.51452937460518,
      "grad_norm": 12.8125,
      "learning_rate": 1.864102969046115e-05,
      "loss": 0.4602,
      "mean_token_accuracy": 0.9398847937583923,
      "step": 4795,
      "train/kl_loss_qwen": 0.10208596438169479,
      "train/kl_loss_r1": 0.15780267715454102,
      "train/total_kl": 0.2598886415362358
    },
    {
      "epoch": 1.5161086544535691,
      "grad_norm": 2.234375,
      "learning_rate": 1.862918509159823e-05,
      "loss": 0.2928,
      "mean_token_accuracy": 0.9627715349197388,
      "step": 4800,
      "train/kl_loss_qwen": 0.03476131893694401,
      "train/kl_loss_r1": 0.11603873372077941,
      "train/total_kl": 0.15080005377531053
    },
    {
      "epoch": 1.5176879343019583,
      "grad_norm": 1.515625,
      "learning_rate": 1.8617340492735313e-05,
      "loss": 0.2556,
      "mean_token_accuracy": 0.9637080669403076,
      "step": 4805,
      "train/kl_loss_qwen": 0.026179027557373048,
      "train/kl_loss_r1": 0.0931219920516014,
      "train/total_kl": 0.11930101960897446
    },
    {
      "epoch": 1.5192672141503474,
      "grad_norm": 2.265625,
      "learning_rate": 1.8605495893872395e-05,
      "loss": 0.3038,
      "mean_token_accuracy": 0.9643283486366272,
      "step": 4810,
      "train/kl_loss_qwen": 0.04649595133960247,
      "train/kl_loss_r1": 0.11017710566520691,
      "train/total_kl": 0.15667305737733841
    },
    {
      "epoch": 1.5208464939987367,
      "grad_norm": 1.75,
      "learning_rate": 1.8593651295009474e-05,
      "loss": 0.2403,
      "mean_token_accuracy": 0.9628156065940857,
      "step": 4815,
      "train/kl_loss_qwen": 0.030744542926549913,
      "train/kl_loss_r1": 0.06841816306114197,
      "train/total_kl": 0.09916270524263382
    },
    {
      "epoch": 1.5224257738471256,
      "grad_norm": 9.0625,
      "learning_rate": 1.858180669614656e-05,
      "loss": 0.5187,
      "mean_token_accuracy": 0.9311153173446656,
      "step": 4820,
      "train/kl_loss_qwen": 0.11104062236845494,
      "train/kl_loss_r1": 0.1765412285923958,
      "train/total_kl": 0.2875818520784378
    },
    {
      "epoch": 1.524005053695515,
      "grad_norm": 3.3125,
      "learning_rate": 1.8569962097283638e-05,
      "loss": 0.3016,
      "mean_token_accuracy": 0.9618187427520752,
      "step": 4825,
      "train/kl_loss_qwen": 0.04459763504564762,
      "train/kl_loss_r1": 0.10001266300678253,
      "train/total_kl": 0.14461030066013336
    },
    {
      "epoch": 1.5255843335439039,
      "grad_norm": 2.796875,
      "learning_rate": 1.8558117498420723e-05,
      "loss": 0.5244,
      "mean_token_accuracy": 0.9358221769332886,
      "step": 4830,
      "train/kl_loss_qwen": 0.12273046188056469,
      "train/kl_loss_r1": 0.19170649871230125,
      "train/total_kl": 0.3144369661808014
    },
    {
      "epoch": 1.5271636133922932,
      "grad_norm": 2.671875,
      "learning_rate": 1.85462728995578e-05,
      "loss": 0.2527,
      "mean_token_accuracy": 0.9692996144294739,
      "step": 4835,
      "train/kl_loss_qwen": 0.03291071876883507,
      "train/kl_loss_r1": 0.09162189811468124,
      "train/total_kl": 0.12453261911869049
    },
    {
      "epoch": 1.5287428932406821,
      "grad_norm": 3.015625,
      "learning_rate": 1.8534428300694883e-05,
      "loss": 0.2862,
      "mean_token_accuracy": 0.9601655602455139,
      "step": 4840,
      "train/kl_loss_qwen": 0.037824484333395955,
      "train/kl_loss_r1": 0.09793010950088502,
      "train/total_kl": 0.13575459420681
    },
    {
      "epoch": 1.5303221730890715,
      "grad_norm": 2.0,
      "learning_rate": 1.8522583701831965e-05,
      "loss": 0.2483,
      "mean_token_accuracy": 0.9682884454727173,
      "step": 4845,
      "train/kl_loss_qwen": 0.04445100352168083,
      "train/kl_loss_r1": 0.07575062066316604,
      "train/total_kl": 0.12020162642002105
    },
    {
      "epoch": 1.5319014529374604,
      "grad_norm": 2.46875,
      "learning_rate": 1.8510739102969047e-05,
      "loss": 0.3275,
      "mean_token_accuracy": 0.9622266292572021,
      "step": 4850,
      "train/kl_loss_qwen": 0.06467283964157104,
      "train/kl_loss_r1": 0.11812219843268394,
      "train/total_kl": 0.18279504179954528
    },
    {
      "epoch": 1.5334807327858497,
      "grad_norm": 3.0,
      "learning_rate": 1.8498894504106126e-05,
      "loss": 0.2564,
      "mean_token_accuracy": 0.9586883187294006,
      "step": 4855,
      "train/kl_loss_qwen": 0.02478809133172035,
      "train/kl_loss_r1": 0.08165369182825089,
      "train/total_kl": 0.10644178241491317
    },
    {
      "epoch": 1.5350600126342386,
      "grad_norm": 2.296875,
      "learning_rate": 1.848704990524321e-05,
      "loss": 0.4091,
      "mean_token_accuracy": 0.9412274122238159,
      "step": 4860,
      "train/kl_loss_qwen": 0.0679111585021019,
      "train/kl_loss_r1": 0.1449393078684807,
      "train/total_kl": 0.2128504693508148
    },
    {
      "epoch": 1.536639292482628,
      "grad_norm": 2.890625,
      "learning_rate": 1.847520530638029e-05,
      "loss": 0.2837,
      "mean_token_accuracy": 0.955828058719635,
      "step": 4865,
      "train/kl_loss_qwen": 0.02858745530247688,
      "train/kl_loss_r1": 0.09037741273641586,
      "train/total_kl": 0.11896486729383468
    },
    {
      "epoch": 1.538218572331017,
      "grad_norm": 2.125,
      "learning_rate": 1.8463360707517375e-05,
      "loss": 0.2899,
      "mean_token_accuracy": 0.9550559282302856,
      "step": 4870,
      "train/kl_loss_qwen": 0.04102783203125,
      "train/kl_loss_r1": 0.08994400203227997,
      "train/total_kl": 0.13097183406352997
    },
    {
      "epoch": 1.5397978521794062,
      "grad_norm": 2.84375,
      "learning_rate": 1.8451516108654454e-05,
      "loss": 0.3278,
      "mean_token_accuracy": 0.9477227687835693,
      "step": 4875,
      "train/kl_loss_qwen": 0.04528830032795668,
      "train/kl_loss_r1": 0.09498817846179008,
      "train/total_kl": 0.14027647972106932
    },
    {
      "epoch": 1.5413771320277954,
      "grad_norm": 2.71875,
      "learning_rate": 1.8439671509791536e-05,
      "loss": 0.3004,
      "mean_token_accuracy": 0.9627875447273254,
      "step": 4880,
      "train/kl_loss_qwen": 0.043726333789527415,
      "train/kl_loss_r1": 0.1102189764380455,
      "train/total_kl": 0.15394531041383744
    },
    {
      "epoch": 1.5429564118761845,
      "grad_norm": 2.21875,
      "learning_rate": 1.8427826910928618e-05,
      "loss": 0.2825,
      "mean_token_accuracy": 0.9602574467658996,
      "step": 4885,
      "train/kl_loss_qwen": 0.037538789212703705,
      "train/kl_loss_r1": 0.09210009127855301,
      "train/total_kl": 0.12963888198137283
    },
    {
      "epoch": 1.5445356917245736,
      "grad_norm": 3.125,
      "learning_rate": 1.84159823120657e-05,
      "loss": 0.3077,
      "mean_token_accuracy": 0.9581627726554871,
      "step": 4890,
      "train/kl_loss_qwen": 0.0390852902084589,
      "train/kl_loss_r1": 0.11290799528360367,
      "train/total_kl": 0.15199328660964967
    },
    {
      "epoch": 1.5461149715729627,
      "grad_norm": 2.734375,
      "learning_rate": 1.840413771320278e-05,
      "loss": 0.336,
      "mean_token_accuracy": 0.9468509793281555,
      "step": 4895,
      "train/kl_loss_qwen": 0.051439995691180226,
      "train/kl_loss_r1": 0.09898207932710648,
      "train/total_kl": 0.15042207390069962
    },
    {
      "epoch": 1.5476942514213519,
      "grad_norm": 2.40625,
      "learning_rate": 1.839229311433986e-05,
      "loss": 0.2585,
      "mean_token_accuracy": 0.9686895608901978,
      "step": 4900,
      "train/kl_loss_qwen": 0.02558461166918278,
      "train/kl_loss_r1": 0.09834726378321648,
      "train/total_kl": 0.1239318773150444
    },
    {
      "epoch": 1.549273531269741,
      "grad_norm": 2.296875,
      "learning_rate": 1.8380448515476942e-05,
      "loss": 0.2468,
      "mean_token_accuracy": 0.9595552802085876,
      "step": 4905,
      "train/kl_loss_qwen": 0.022857947275042534,
      "train/kl_loss_r1": 0.07709285020828247,
      "train/total_kl": 0.09995079636573792
    },
    {
      "epoch": 1.5508528111181301,
      "grad_norm": 1.9453125,
      "learning_rate": 1.8368603916614024e-05,
      "loss": 0.3071,
      "mean_token_accuracy": 0.9588536500930787,
      "step": 4910,
      "train/kl_loss_qwen": 0.036932972446084025,
      "train/kl_loss_r1": 0.1188292808830738,
      "train/total_kl": 0.15576225370168686
    },
    {
      "epoch": 1.5524320909665192,
      "grad_norm": 1.484375,
      "learning_rate": 1.8356759317751106e-05,
      "loss": 0.2538,
      "mean_token_accuracy": 0.96286940574646,
      "step": 4915,
      "train/kl_loss_qwen": 0.02727362997829914,
      "train/kl_loss_r1": 0.0810820572078228,
      "train/total_kl": 0.10835568904876709
    },
    {
      "epoch": 1.5540113708149084,
      "grad_norm": 15.5,
      "learning_rate": 1.8344914718888185e-05,
      "loss": 0.5718,
      "mean_token_accuracy": 0.9271079421043396,
      "step": 4920,
      "train/kl_loss_qwen": 0.12666431143879892,
      "train/kl_loss_r1": 0.21426910012960435,
      "train/total_kl": 0.3409334093332291
    },
    {
      "epoch": 1.5555906506632975,
      "grad_norm": 2.15625,
      "learning_rate": 1.833307012002527e-05,
      "loss": 0.2771,
      "mean_token_accuracy": 0.9652738213539124,
      "step": 4925,
      "train/kl_loss_qwen": 0.0364518515765667,
      "train/kl_loss_r1": 0.10077509135007859,
      "train/total_kl": 0.13722694367170335
    },
    {
      "epoch": 1.5571699305116866,
      "grad_norm": 3.125,
      "learning_rate": 1.832122552116235e-05,
      "loss": 0.2792,
      "mean_token_accuracy": 0.9621742725372314,
      "step": 4930,
      "train/kl_loss_qwen": 0.03356264717876911,
      "train/kl_loss_r1": 0.0952174797654152,
      "train/total_kl": 0.1287801280617714
    },
    {
      "epoch": 1.5587492103600757,
      "grad_norm": 2.234375,
      "learning_rate": 1.830938092229943e-05,
      "loss": 0.298,
      "mean_token_accuracy": 0.9613176941871643,
      "step": 4935,
      "train/kl_loss_qwen": 0.04880597032606602,
      "train/kl_loss_r1": 0.1009241670370102,
      "train/total_kl": 0.14973013550043107
    },
    {
      "epoch": 1.560328490208465,
      "grad_norm": 3.484375,
      "learning_rate": 1.8297536323436513e-05,
      "loss": 0.2986,
      "mean_token_accuracy": 0.9533314108848572,
      "step": 4940,
      "train/kl_loss_qwen": 0.03569821380078793,
      "train/kl_loss_r1": 0.07096512839198113,
      "train/total_kl": 0.1066633403301239
    },
    {
      "epoch": 1.561907770056854,
      "grad_norm": 2.265625,
      "learning_rate": 1.8285691724573595e-05,
      "loss": 0.2965,
      "mean_token_accuracy": 0.961072564125061,
      "step": 4945,
      "train/kl_loss_qwen": 0.04849345274269581,
      "train/kl_loss_r1": 0.10982134938240051,
      "train/total_kl": 0.1583148017525673
    },
    {
      "epoch": 1.5634870499052433,
      "grad_norm": 2.328125,
      "learning_rate": 1.8273847125710677e-05,
      "loss": 0.2483,
      "mean_token_accuracy": 0.9651191830635071,
      "step": 4950,
      "train/kl_loss_qwen": 0.03232039622962475,
      "train/kl_loss_r1": 0.08500564023852349,
      "train/total_kl": 0.11732603758573532
    },
    {
      "epoch": 1.5650663297536322,
      "grad_norm": 2.234375,
      "learning_rate": 1.826200252684776e-05,
      "loss": 0.2793,
      "mean_token_accuracy": 0.9580449223518371,
      "step": 4955,
      "train/kl_loss_qwen": 0.03593952432274818,
      "train/kl_loss_r1": 0.08791730254888534,
      "train/total_kl": 0.12385682761669159
    },
    {
      "epoch": 1.5666456096020216,
      "grad_norm": 4.96875,
      "learning_rate": 1.8250157927984837e-05,
      "loss": 0.3036,
      "mean_token_accuracy": 0.9626813173294068,
      "step": 4960,
      "train/kl_loss_qwen": 0.04605437070131302,
      "train/kl_loss_r1": 0.11057046502828598,
      "train/total_kl": 0.156624835729599
    },
    {
      "epoch": 1.5682248894504105,
      "grad_norm": 3.0,
      "learning_rate": 1.8238313329121923e-05,
      "loss": 0.2694,
      "mean_token_accuracy": 0.9645307183265686,
      "step": 4965,
      "train/kl_loss_qwen": 0.03433260079473257,
      "train/kl_loss_r1": 0.08871393203735352,
      "train/total_kl": 0.12304653227329254
    },
    {
      "epoch": 1.5698041692987998,
      "grad_norm": 2.375,
      "learning_rate": 1.8226468730259e-05,
      "loss": 0.2539,
      "mean_token_accuracy": 0.9688814997673034,
      "step": 4970,
      "train/kl_loss_qwen": 0.036784827709198,
      "train/kl_loss_r1": 0.09187296479940414,
      "train/total_kl": 0.128657791018486
    },
    {
      "epoch": 1.5713834491471887,
      "grad_norm": 3.03125,
      "learning_rate": 1.8214624131396083e-05,
      "loss": 0.3085,
      "mean_token_accuracy": 0.9526297569274902,
      "step": 4975,
      "train/kl_loss_qwen": 0.04091540724039078,
      "train/kl_loss_r1": 0.0914787083864212,
      "train/total_kl": 0.13239411413669586
    },
    {
      "epoch": 1.572962728995578,
      "grad_norm": 3.203125,
      "learning_rate": 1.8202779532533165e-05,
      "loss": 0.3312,
      "mean_token_accuracy": 0.9492631554603577,
      "step": 4980,
      "train/kl_loss_qwen": 0.0332471888512373,
      "train/kl_loss_r1": 0.11046817898750305,
      "train/total_kl": 0.14371536821126937
    },
    {
      "epoch": 1.574542008843967,
      "grad_norm": 3.0625,
      "learning_rate": 1.8190934933670247e-05,
      "loss": 0.2865,
      "mean_token_accuracy": 0.9664994716644287,
      "step": 4985,
      "train/kl_loss_qwen": 0.04501254260540009,
      "train/kl_loss_r1": 0.11394366025924682,
      "train/total_kl": 0.1589562013745308
    },
    {
      "epoch": 1.5761212886923563,
      "grad_norm": 1.8359375,
      "learning_rate": 1.817909033480733e-05,
      "loss": 0.2835,
      "mean_token_accuracy": 0.9602677226066589,
      "step": 4990,
      "train/kl_loss_qwen": 0.03861864283680916,
      "train/kl_loss_r1": 0.10277867317199707,
      "train/total_kl": 0.14139731377363204
    },
    {
      "epoch": 1.5777005685407455,
      "grad_norm": 2.9375,
      "learning_rate": 1.816724573594441e-05,
      "loss": 0.2586,
      "mean_token_accuracy": 0.9615461111068726,
      "step": 4995,
      "train/kl_loss_qwen": 0.024707885272800923,
      "train/kl_loss_r1": 0.08633724376559257,
      "train/total_kl": 0.11104512810707093
    },
    {
      "epoch": 1.5792798483891346,
      "grad_norm": 3.28125,
      "learning_rate": 1.815540113708149e-05,
      "loss": 0.3348,
      "mean_token_accuracy": 0.9475095272064209,
      "step": 5000,
      "train/kl_loss_qwen": 0.04476226009428501,
      "train/kl_loss_r1": 0.09369741380214691,
      "train/total_kl": 0.138459675014019
    },
    {
      "epoch": 1.5808591282375237,
      "grad_norm": 2.671875,
      "learning_rate": 1.8143556538218575e-05,
      "loss": 0.3398,
      "mean_token_accuracy": 0.9447283625602723,
      "step": 5005,
      "train/kl_loss_qwen": 0.045039734616875646,
      "train/kl_loss_r1": 0.09138004034757614,
      "train/total_kl": 0.13641977608203887
    },
    {
      "epoch": 1.5824384080859129,
      "grad_norm": 4.21875,
      "learning_rate": 1.8131711939355654e-05,
      "loss": 0.296,
      "mean_token_accuracy": 0.9589680910110474,
      "step": 5010,
      "train/kl_loss_qwen": 0.04367956817150116,
      "train/kl_loss_r1": 0.09234432429075241,
      "train/total_kl": 0.1360238954424858
    },
    {
      "epoch": 1.584017687934302,
      "grad_norm": 2.046875,
      "learning_rate": 1.8119867340492736e-05,
      "loss": 0.2938,
      "mean_token_accuracy": 0.9643046498298645,
      "step": 5015,
      "train/kl_loss_qwen": 0.04260939061641693,
      "train/kl_loss_r1": 0.10264224112033844,
      "train/total_kl": 0.14525163024663926
    },
    {
      "epoch": 1.585596967782691,
      "grad_norm": 4.34375,
      "learning_rate": 1.8108022741629818e-05,
      "loss": 0.3205,
      "mean_token_accuracy": 0.9465242266654968,
      "step": 5020,
      "train/kl_loss_qwen": 0.03369136825203896,
      "train/kl_loss_r1": 0.07980907708406448,
      "train/total_kl": 0.1135004460811615
    },
    {
      "epoch": 1.5871762476310802,
      "grad_norm": 2.265625,
      "learning_rate": 1.80961781427669e-05,
      "loss": 0.3185,
      "mean_token_accuracy": 0.9527551531791687,
      "step": 5025,
      "train/kl_loss_qwen": 0.04622704163193703,
      "train/kl_loss_r1": 0.10520811080932617,
      "train/total_kl": 0.15143515467643737
    },
    {
      "epoch": 1.5887555274794694,
      "grad_norm": 2.265625,
      "learning_rate": 1.808433354390398e-05,
      "loss": 0.303,
      "mean_token_accuracy": 0.9567196726799011,
      "step": 5030,
      "train/kl_loss_qwen": 0.05304274894297123,
      "train/kl_loss_r1": 0.09974344223737716,
      "train/total_kl": 0.15278619080781936
    },
    {
      "epoch": 1.5903348073278585,
      "grad_norm": 1.9765625,
      "learning_rate": 1.807248894504106e-05,
      "loss": 0.3064,
      "mean_token_accuracy": 0.9459467768669129,
      "step": 5035,
      "train/kl_loss_qwen": 0.03128695879131556,
      "train/kl_loss_r1": 0.08263867348432541,
      "train/total_kl": 0.1139256328344345
    },
    {
      "epoch": 1.5919140871762476,
      "grad_norm": 19.0,
      "learning_rate": 1.8060644346178142e-05,
      "loss": 0.5892,
      "mean_token_accuracy": 0.9308818578720093,
      "step": 5040,
      "train/kl_loss_qwen": 0.14544303640723227,
      "train/kl_loss_r1": 0.22645720541477204,
      "train/total_kl": 0.37190023958683016
    },
    {
      "epoch": 1.5934933670246367,
      "grad_norm": 2.015625,
      "learning_rate": 1.8048799747315224e-05,
      "loss": 0.3161,
      "mean_token_accuracy": 0.9592323899269104,
      "step": 5045,
      "train/kl_loss_qwen": 0.03822853714227677,
      "train/kl_loss_r1": 0.11608197838068009,
      "train/total_kl": 0.15431051701307297
    },
    {
      "epoch": 1.5950726468730259,
      "grad_norm": 3.203125,
      "learning_rate": 1.8036955148452306e-05,
      "loss": 0.3135,
      "mean_token_accuracy": 0.9598764300346374,
      "step": 5050,
      "train/kl_loss_qwen": 0.048730790987610816,
      "train/kl_loss_r1": 0.10671563744544983,
      "train/total_kl": 0.15544643104076386
    },
    {
      "epoch": 1.596651926721415,
      "grad_norm": 3.0,
      "learning_rate": 1.8025110549589388e-05,
      "loss": 0.2687,
      "mean_token_accuracy": 0.9619138240814209,
      "step": 5055,
      "train/kl_loss_qwen": 0.032063265517354014,
      "train/kl_loss_r1": 0.08146318718791008,
      "train/total_kl": 0.11352645307779312
    },
    {
      "epoch": 1.598231206569804,
      "grad_norm": 2.5,
      "learning_rate": 1.801326595072647e-05,
      "loss": 0.2707,
      "mean_token_accuracy": 0.9606055498123169,
      "step": 5060,
      "train/kl_loss_qwen": 0.023444147408008577,
      "train/kl_loss_r1": 0.08306510969996453,
      "train/total_kl": 0.10650925785303116
    },
    {
      "epoch": 1.5998104864181935,
      "grad_norm": 2.71875,
      "learning_rate": 1.800142135186355e-05,
      "loss": 0.2721,
      "mean_token_accuracy": 0.962373924255371,
      "step": 5065,
      "train/kl_loss_qwen": 0.03647337518632412,
      "train/kl_loss_r1": 0.08934515714645386,
      "train/total_kl": 0.12581852823495865
    },
    {
      "epoch": 1.6013897662665824,
      "grad_norm": 2.875,
      "learning_rate": 1.7989576753000634e-05,
      "loss": 0.3289,
      "mean_token_accuracy": 0.9604684948921204,
      "step": 5070,
      "train/kl_loss_qwen": 0.05957434736192226,
      "train/kl_loss_r1": 0.12501501590013503,
      "train/total_kl": 0.18458936214447022
    },
    {
      "epoch": 1.6029690461149717,
      "grad_norm": 1.7578125,
      "learning_rate": 1.7977732154137713e-05,
      "loss": 0.2637,
      "mean_token_accuracy": 0.9688223123550415,
      "step": 5075,
      "train/kl_loss_qwen": 0.038856085017323495,
      "train/kl_loss_r1": 0.09954321533441543,
      "train/total_kl": 0.1383992999792099
    },
    {
      "epoch": 1.6045483259633606,
      "grad_norm": 1.828125,
      "learning_rate": 1.7965887555274795e-05,
      "loss": 0.274,
      "mean_token_accuracy": 0.965758740901947,
      "step": 5080,
      "train/kl_loss_qwen": 0.041578725725412366,
      "train/kl_loss_r1": 0.09290871173143386,
      "train/total_kl": 0.13448743820190429
    },
    {
      "epoch": 1.60612760581175,
      "grad_norm": 2.59375,
      "learning_rate": 1.7954042956411876e-05,
      "loss": 0.3085,
      "mean_token_accuracy": 0.9590817809104919,
      "step": 5085,
      "train/kl_loss_qwen": 0.038109660148620605,
      "train/kl_loss_r1": 0.09664418548345566,
      "train/total_kl": 0.13475384563207626
    },
    {
      "epoch": 1.6077068856601389,
      "grad_norm": 1.6015625,
      "learning_rate": 1.794219835754896e-05,
      "loss": 0.229,
      "mean_token_accuracy": 0.9735617637634277,
      "step": 5090,
      "train/kl_loss_qwen": 0.04008105918765068,
      "train/kl_loss_r1": 0.07888272106647491,
      "train/total_kl": 0.11896378099918366
    },
    {
      "epoch": 1.6092861655085282,
      "grad_norm": 2.9375,
      "learning_rate": 1.793035375868604e-05,
      "loss": 0.3543,
      "mean_token_accuracy": 0.9545761823654175,
      "step": 5095,
      "train/kl_loss_qwen": 0.06167028360068798,
      "train/kl_loss_r1": 0.11619099825620652,
      "train/total_kl": 0.17786127775907518
    },
    {
      "epoch": 1.6108654453569171,
      "grad_norm": 2.125,
      "learning_rate": 1.7918509159823122e-05,
      "loss": 0.2792,
      "mean_token_accuracy": 0.9677553415298462,
      "step": 5100,
      "train/kl_loss_qwen": 0.04271460995078087,
      "train/kl_loss_r1": 0.1137930765748024,
      "train/total_kl": 0.1565076857805252
    },
    {
      "epoch": 1.6124447252053065,
      "grad_norm": 2.109375,
      "learning_rate": 1.79066645609602e-05,
      "loss": 0.2642,
      "mean_token_accuracy": 0.9727542638778687,
      "step": 5105,
      "train/kl_loss_qwen": 0.0436590064316988,
      "train/kl_loss_r1": 0.0994985818862915,
      "train/total_kl": 0.14315758943557738
    },
    {
      "epoch": 1.6140240050536954,
      "grad_norm": 1.765625,
      "learning_rate": 1.7894819962097286e-05,
      "loss": 0.2655,
      "mean_token_accuracy": 0.963416862487793,
      "step": 5110,
      "train/kl_loss_qwen": 0.030635759234428406,
      "train/kl_loss_r1": 0.09515174329280854,
      "train/total_kl": 0.12578750103712083
    },
    {
      "epoch": 1.6156032849020847,
      "grad_norm": 3.375,
      "learning_rate": 1.7882975363234365e-05,
      "loss": 0.2646,
      "mean_token_accuracy": 0.9718488097190857,
      "step": 5115,
      "train/kl_loss_qwen": 0.03252498023211956,
      "train/kl_loss_r1": 0.10526231303811073,
      "train/total_kl": 0.1377872958779335
    },
    {
      "epoch": 1.6171825647504738,
      "grad_norm": 2.0625,
      "learning_rate": 1.7871130764371447e-05,
      "loss": 0.2723,
      "mean_token_accuracy": 0.9682145118713379,
      "step": 5120,
      "train/kl_loss_qwen": 0.038565434888005255,
      "train/kl_loss_r1": 0.09935315549373627,
      "train/total_kl": 0.1379185900092125
    },
    {
      "epoch": 1.618761844598863,
      "grad_norm": 2.28125,
      "learning_rate": 1.785928616550853e-05,
      "loss": 0.2567,
      "mean_token_accuracy": 0.9615680813789368,
      "step": 5125,
      "train/kl_loss_qwen": 0.0280915267765522,
      "train/kl_loss_r1": 0.07563170492649078,
      "train/total_kl": 0.10372323095798493
    },
    {
      "epoch": 1.620341124447252,
      "grad_norm": 3.140625,
      "learning_rate": 1.784744156664561e-05,
      "loss": 0.275,
      "mean_token_accuracy": 0.9582354426383972,
      "step": 5130,
      "train/kl_loss_qwen": 0.03761475943028927,
      "train/kl_loss_r1": 0.07994740977883338,
      "train/total_kl": 0.11756217181682586
    },
    {
      "epoch": 1.6219204042956412,
      "grad_norm": 3.84375,
      "learning_rate": 1.7835596967782693e-05,
      "loss": 0.2702,
      "mean_token_accuracy": 0.9628668069839478,
      "step": 5135,
      "train/kl_loss_qwen": 0.0333556018769741,
      "train/kl_loss_r1": 0.09749825075268745,
      "train/total_kl": 0.13085385411977768
    },
    {
      "epoch": 1.6234996841440303,
      "grad_norm": 2.859375,
      "learning_rate": 1.7823752368919775e-05,
      "loss": 0.3205,
      "mean_token_accuracy": 0.9633049845695496,
      "step": 5140,
      "train/kl_loss_qwen": 0.05835762694478035,
      "train/kl_loss_r1": 0.11656875312328338,
      "train/total_kl": 0.17492638230323793
    },
    {
      "epoch": 1.6250789639924195,
      "grad_norm": 3.078125,
      "learning_rate": 1.7811907770056853e-05,
      "loss": 0.266,
      "mean_token_accuracy": 0.9642176270484925,
      "step": 5145,
      "train/kl_loss_qwen": 0.03729858659207821,
      "train/kl_loss_r1": 0.09060643538832665,
      "train/total_kl": 0.1279050201177597
    },
    {
      "epoch": 1.6266582438408086,
      "grad_norm": 2.140625,
      "learning_rate": 1.780006317119394e-05,
      "loss": 0.3458,
      "mean_token_accuracy": 0.965228819847107,
      "step": 5150,
      "train/kl_loss_qwen": 0.06371126174926758,
      "train/kl_loss_r1": 0.15694268196821212,
      "train/total_kl": 0.22065394669771193
    },
    {
      "epoch": 1.6282375236891977,
      "grad_norm": 2.53125,
      "learning_rate": 1.7788218572331017e-05,
      "loss": 0.3077,
      "mean_token_accuracy": 0.9572019934654236,
      "step": 5155,
      "train/kl_loss_qwen": 0.042045365646481515,
      "train/kl_loss_r1": 0.10315109044313431,
      "train/total_kl": 0.14519645720720292
    },
    {
      "epoch": 1.6298168035375868,
      "grad_norm": 2.09375,
      "learning_rate": 1.7776373973468096e-05,
      "loss": 0.2955,
      "mean_token_accuracy": 0.9620293498039245,
      "step": 5160,
      "train/kl_loss_qwen": 0.036943725496530536,
      "train/kl_loss_r1": 0.11455025225877762,
      "train/total_kl": 0.15149397999048234
    },
    {
      "epoch": 1.631396083385976,
      "grad_norm": 3.53125,
      "learning_rate": 1.776452937460518e-05,
      "loss": 0.2596,
      "mean_token_accuracy": 0.9638445615768433,
      "step": 5165,
      "train/kl_loss_qwen": 0.0342584740370512,
      "train/kl_loss_r1": 0.08455819934606552,
      "train/total_kl": 0.11881667375564575
    },
    {
      "epoch": 1.632975363234365,
      "grad_norm": 2.015625,
      "learning_rate": 1.775268477574226e-05,
      "loss": 0.8246,
      "mean_token_accuracy": 0.911933982372284,
      "step": 5170,
      "train/kl_loss_qwen": 0.22300798110663891,
      "train/kl_loss_r1": 0.31449490040540695,
      "train/total_kl": 0.5375028878450394
    },
    {
      "epoch": 1.6345546430827542,
      "grad_norm": 3.015625,
      "learning_rate": 1.7740840176879345e-05,
      "loss": 0.3172,
      "mean_token_accuracy": 0.9568942666053772,
      "step": 5175,
      "train/kl_loss_qwen": 0.0462687648832798,
      "train/kl_loss_r1": 0.10913461595773696,
      "train/total_kl": 0.15540338158607483
    },
    {
      "epoch": 1.6361339229311433,
      "grad_norm": 2.671875,
      "learning_rate": 1.7728995578016424e-05,
      "loss": 0.2886,
      "mean_token_accuracy": 0.9606268525123596,
      "step": 5180,
      "train/kl_loss_qwen": 0.03397431261837482,
      "train/kl_loss_r1": 0.09145184010267257,
      "train/total_kl": 0.12542615234851837
    },
    {
      "epoch": 1.6377132027795325,
      "grad_norm": 2.703125,
      "learning_rate": 1.7717150979153506e-05,
      "loss": 0.2764,
      "mean_token_accuracy": 0.9585961699485779,
      "step": 5185,
      "train/kl_loss_qwen": 0.039028293639421466,
      "train/kl_loss_r1": 0.08939458802342415,
      "train/total_kl": 0.12842288464307786
    },
    {
      "epoch": 1.6392924826279218,
      "grad_norm": 2.046875,
      "learning_rate": 1.7705306380290588e-05,
      "loss": 0.2785,
      "mean_token_accuracy": 0.9669302225112915,
      "step": 5190,
      "train/kl_loss_qwen": 0.040070772916078565,
      "train/kl_loss_r1": 0.1030052125453949,
      "train/total_kl": 0.14307598769664764
    },
    {
      "epoch": 1.6408717624763107,
      "grad_norm": 10.375,
      "learning_rate": 1.769346178142767e-05,
      "loss": 0.2983,
      "mean_token_accuracy": 0.9601428270339966,
      "step": 5195,
      "train/kl_loss_qwen": 0.04499279856681824,
      "train/kl_loss_r1": 0.10197088420391083,
      "train/total_kl": 0.14696368277072908
    },
    {
      "epoch": 1.6424510423247,
      "grad_norm": 1.7109375,
      "learning_rate": 1.768161718256475e-05,
      "loss": 0.2954,
      "mean_token_accuracy": 0.9581763744354248,
      "step": 5200,
      "train/kl_loss_qwen": 0.030926124751567842,
      "train/kl_loss_r1": 0.09210587292909622,
      "train/total_kl": 0.12303199768066406
    },
    {
      "epoch": 1.644030322173089,
      "grad_norm": 2.28125,
      "learning_rate": 1.7669772583701834e-05,
      "loss": 0.3017,
      "mean_token_accuracy": 0.9551280021667481,
      "step": 5205,
      "train/kl_loss_qwen": 0.033274417370557786,
      "train/kl_loss_r1": 0.09771268144249916,
      "train/total_kl": 0.13098709881305695
    },
    {
      "epoch": 1.6456096020214783,
      "grad_norm": 1.8984375,
      "learning_rate": 1.7657927984838912e-05,
      "loss": 0.2674,
      "mean_token_accuracy": 0.9630029439926148,
      "step": 5210,
      "train/kl_loss_qwen": 0.02985788993537426,
      "train/kl_loss_r1": 0.08493964225053788,
      "train/total_kl": 0.11479753255844116
    },
    {
      "epoch": 1.6471888818698672,
      "grad_norm": 2.578125,
      "learning_rate": 1.7646083385975998e-05,
      "loss": 0.3292,
      "mean_token_accuracy": 0.9609008669853211,
      "step": 5215,
      "train/kl_loss_qwen": 0.05572503730654717,
      "train/kl_loss_r1": 0.12447040677070617,
      "train/total_kl": 0.18019544333219528
    },
    {
      "epoch": 1.6487681617182566,
      "grad_norm": 2.53125,
      "learning_rate": 1.7634238787113076e-05,
      "loss": 0.2769,
      "mean_token_accuracy": 0.9554272770881653,
      "step": 5220,
      "train/kl_loss_qwen": 0.03228133432567119,
      "train/kl_loss_r1": 0.07575906366109848,
      "train/total_kl": 0.10804039686918258
    },
    {
      "epoch": 1.6503474415666455,
      "grad_norm": 3.09375,
      "learning_rate": 1.7622394188250158e-05,
      "loss": 0.301,
      "mean_token_accuracy": 0.9635890364646912,
      "step": 5225,
      "train/kl_loss_qwen": 0.04877604395151138,
      "train/kl_loss_r1": 0.09367519691586494,
      "train/total_kl": 0.14245124608278276
    },
    {
      "epoch": 1.6519267214150348,
      "grad_norm": 1.828125,
      "learning_rate": 1.761054958938724e-05,
      "loss": 0.2956,
      "mean_token_accuracy": 0.9514432191848755,
      "step": 5230,
      "train/kl_loss_qwen": 0.03325424250215292,
      "train/kl_loss_r1": 0.08206587135791779,
      "train/total_kl": 0.11532011479139329
    },
    {
      "epoch": 1.6535060012634237,
      "grad_norm": 2.40625,
      "learning_rate": 1.7598704990524322e-05,
      "loss": 0.3641,
      "mean_token_accuracy": 0.9547950387001037,
      "step": 5235,
      "train/kl_loss_qwen": 0.06987997367978097,
      "train/kl_loss_r1": 0.12259236574172974,
      "train/total_kl": 0.19247234165668486
    },
    {
      "epoch": 1.655085281111813,
      "grad_norm": 2.15625,
      "learning_rate": 1.75868603916614e-05,
      "loss": 0.3008,
      "mean_token_accuracy": 0.9540727257728576,
      "step": 5240,
      "train/kl_loss_qwen": 0.04146949015557766,
      "train/kl_loss_r1": 0.09700813218951225,
      "train/total_kl": 0.138477623462677
    },
    {
      "epoch": 1.656664560960202,
      "grad_norm": 2.34375,
      "learning_rate": 1.7575015792798486e-05,
      "loss": 0.3025,
      "mean_token_accuracy": 0.9585528492927551,
      "step": 5245,
      "train/kl_loss_qwen": 0.042682092264294626,
      "train/kl_loss_r1": 0.10609034448862076,
      "train/total_kl": 0.14877243787050248
    },
    {
      "epoch": 1.6582438408085913,
      "grad_norm": 2.59375,
      "learning_rate": 1.7563171193935565e-05,
      "loss": 0.2678,
      "mean_token_accuracy": 0.9657195091247559,
      "step": 5250,
      "train/kl_loss_qwen": 0.02981017976999283,
      "train/kl_loss_r1": 0.1008859246969223,
      "train/total_kl": 0.1306961014866829
    },
    {
      "epoch": 1.6598231206569805,
      "grad_norm": 4.75,
      "learning_rate": 1.755132659507265e-05,
      "loss": 0.3052,
      "mean_token_accuracy": 0.9474646806716919,
      "step": 5255,
      "train/kl_loss_qwen": 0.03298285789787769,
      "train/kl_loss_r1": 0.08161807358264923,
      "train/total_kl": 0.11460093110799789
    },
    {
      "epoch": 1.6614024005053696,
      "grad_norm": 3.71875,
      "learning_rate": 1.753948199620973e-05,
      "loss": 0.3015,
      "mean_token_accuracy": 0.9630206108093262,
      "step": 5260,
      "train/kl_loss_qwen": 0.046903107315301895,
      "train/kl_loss_r1": 0.10307001918554307,
      "train/total_kl": 0.14997312873601915
    },
    {
      "epoch": 1.6629816803537587,
      "grad_norm": 1.5625,
      "learning_rate": 1.752763739734681e-05,
      "loss": 0.2967,
      "mean_token_accuracy": 0.9638324022293091,
      "step": 5265,
      "train/kl_loss_qwen": 0.02928678020834923,
      "train/kl_loss_r1": 0.11503093019127845,
      "train/total_kl": 0.14431770741939545
    },
    {
      "epoch": 1.6645609602021478,
      "grad_norm": 3.890625,
      "learning_rate": 1.7515792798483893e-05,
      "loss": 0.3105,
      "mean_token_accuracy": 0.9509057641029358,
      "step": 5270,
      "train/kl_loss_qwen": 0.036556807532906535,
      "train/kl_loss_r1": 0.0872780829668045,
      "train/total_kl": 0.12383489161729813
    },
    {
      "epoch": 1.666140240050537,
      "grad_norm": 2.515625,
      "learning_rate": 1.7503948199620975e-05,
      "loss": 0.2762,
      "mean_token_accuracy": 0.968660318851471,
      "step": 5275,
      "train/kl_loss_qwen": 0.04499706588685513,
      "train/kl_loss_r1": 0.1026617243885994,
      "train/total_kl": 0.14765879213809968
    },
    {
      "epoch": 1.667719519898926,
      "grad_norm": 2.671875,
      "learning_rate": 1.7492103600758057e-05,
      "loss": 0.2979,
      "mean_token_accuracy": 0.9565050601959229,
      "step": 5280,
      "train/kl_loss_qwen": 0.03636668808758259,
      "train/kl_loss_r1": 0.08962554335594178,
      "train/total_kl": 0.12599223405122756
    },
    {
      "epoch": 1.6692987997473152,
      "grad_norm": 2.734375,
      "learning_rate": 1.7480259001895135e-05,
      "loss": 0.2675,
      "mean_token_accuracy": 0.9579012870788575,
      "step": 5285,
      "train/kl_loss_qwen": 0.026114974915981293,
      "train/kl_loss_r1": 0.08535072207450867,
      "train/total_kl": 0.11146569699048996
    },
    {
      "epoch": 1.6708780795957043,
      "grad_norm": 2.421875,
      "learning_rate": 1.7468414403032217e-05,
      "loss": 0.2993,
      "mean_token_accuracy": 0.9583601951599121,
      "step": 5290,
      "train/kl_loss_qwen": 0.04099457189440727,
      "train/kl_loss_r1": 0.09433886706829071,
      "train/total_kl": 0.13533343821763993
    },
    {
      "epoch": 1.6724573594440935,
      "grad_norm": 2.53125,
      "learning_rate": 1.74565698041693e-05,
      "loss": 0.2778,
      "mean_token_accuracy": 0.9638667702674866,
      "step": 5295,
      "train/kl_loss_qwen": 0.041493486240506174,
      "train/kl_loss_r1": 0.09667314738035201,
      "train/total_kl": 0.1381666347384453
    },
    {
      "epoch": 1.6740366392924826,
      "grad_norm": 4.5625,
      "learning_rate": 1.744472520530638e-05,
      "loss": 0.2884,
      "mean_token_accuracy": 0.9534436702728272,
      "step": 5300,
      "train/kl_loss_qwen": 0.029674012772738935,
      "train/kl_loss_r1": 0.08850864917039872,
      "train/total_kl": 0.11818266361951828
    },
    {
      "epoch": 1.6756159191408717,
      "grad_norm": 5.03125,
      "learning_rate": 1.743288060644346e-05,
      "loss": 0.3526,
      "mean_token_accuracy": 0.9519017100334167,
      "step": 5305,
      "train/kl_loss_qwen": 0.05769060291349888,
      "train/kl_loss_r1": 0.1290040172636509,
      "train/total_kl": 0.18669462352991104
    },
    {
      "epoch": 1.6771951989892608,
      "grad_norm": 2.984375,
      "learning_rate": 1.7421036007580545e-05,
      "loss": 0.3003,
      "mean_token_accuracy": 0.9571117520332336,
      "step": 5310,
      "train/kl_loss_qwen": 0.03862414509057999,
      "train/kl_loss_r1": 0.09387920200824737,
      "train/total_kl": 0.13250334709882736
    },
    {
      "epoch": 1.6787744788376502,
      "grad_norm": 3.09375,
      "learning_rate": 1.7409191408717624e-05,
      "loss": 0.357,
      "mean_token_accuracy": 0.9615911841392517,
      "step": 5315,
      "train/kl_loss_qwen": 0.07259219586849212,
      "train/kl_loss_r1": 0.14197880923748016,
      "train/total_kl": 0.21457100808620452
    },
    {
      "epoch": 1.680353758686039,
      "grad_norm": 1.9453125,
      "learning_rate": 1.739734680985471e-05,
      "loss": 0.2795,
      "mean_token_accuracy": 0.9653712391853333,
      "step": 5320,
      "train/kl_loss_qwen": 0.04426160678267479,
      "train/kl_loss_r1": 0.09935070723295211,
      "train/total_kl": 0.14361231327056884
    },
    {
      "epoch": 1.6819330385344284,
      "grad_norm": 3.25,
      "learning_rate": 1.7385502210991787e-05,
      "loss": 0.3183,
      "mean_token_accuracy": 0.9515377044677734,
      "step": 5325,
      "train/kl_loss_qwen": 0.03382651805877686,
      "train/kl_loss_r1": 0.104255111515522,
      "train/total_kl": 0.1380816251039505
    },
    {
      "epoch": 1.6835123183828173,
      "grad_norm": 3.125,
      "learning_rate": 1.737365761212887e-05,
      "loss": 0.2497,
      "mean_token_accuracy": 0.961120891571045,
      "step": 5330,
      "train/kl_loss_qwen": 0.02314903140068054,
      "train/kl_loss_r1": 0.06898651793599128,
      "train/total_kl": 0.09213554859161377
    },
    {
      "epoch": 1.6850915982312067,
      "grad_norm": 3.046875,
      "learning_rate": 1.736181301326595e-05,
      "loss": 0.326,
      "mean_token_accuracy": 0.9593894004821777,
      "step": 5335,
      "train/kl_loss_qwen": 0.05705704763531685,
      "train/kl_loss_r1": 0.11691752672195435,
      "train/total_kl": 0.17397457659244536
    },
    {
      "epoch": 1.6866708780795956,
      "grad_norm": 2.796875,
      "learning_rate": 1.7349968414403033e-05,
      "loss": 0.6357,
      "mean_token_accuracy": 0.9283433079719543,
      "step": 5340,
      "train/kl_loss_qwen": 0.14514803737401963,
      "train/kl_loss_r1": 0.2588561475276947,
      "train/total_kl": 0.4040041953325272
    },
    {
      "epoch": 1.688250157927985,
      "grad_norm": 2.09375,
      "learning_rate": 1.7338123815540112e-05,
      "loss": 0.2914,
      "mean_token_accuracy": 0.9585925459861755,
      "step": 5345,
      "train/kl_loss_qwen": 0.042611193284392355,
      "train/kl_loss_r1": 0.09524012580513955,
      "train/total_kl": 0.137851320207119
    },
    {
      "epoch": 1.6898294377763738,
      "grad_norm": 3.09375,
      "learning_rate": 1.7326279216677197e-05,
      "loss": 0.2669,
      "mean_token_accuracy": 0.9637061357498169,
      "step": 5350,
      "train/kl_loss_qwen": 0.03952077925205231,
      "train/kl_loss_r1": 0.0940628319978714,
      "train/total_kl": 0.13358361274003983
    },
    {
      "epoch": 1.6914087176247632,
      "grad_norm": 5.375,
      "learning_rate": 1.7314434617814276e-05,
      "loss": 0.3152,
      "mean_token_accuracy": 0.9643152594566345,
      "step": 5355,
      "train/kl_loss_qwen": 0.05609998144209385,
      "train/kl_loss_r1": 0.12473059743642807,
      "train/total_kl": 0.18083057701587676
    },
    {
      "epoch": 1.692987997473152,
      "grad_norm": 1.796875,
      "learning_rate": 1.730259001895136e-05,
      "loss": 0.2988,
      "mean_token_accuracy": 0.9516977429389953,
      "step": 5360,
      "train/kl_loss_qwen": 0.028914622589945795,
      "train/kl_loss_r1": 0.08143803402781487,
      "train/total_kl": 0.11035265624523163
    },
    {
      "epoch": 1.6945672773215414,
      "grad_norm": 1.8125,
      "learning_rate": 1.729074542008844e-05,
      "loss": 0.2821,
      "mean_token_accuracy": 0.9646175861358642,
      "step": 5365,
      "train/kl_loss_qwen": 0.039552390575408936,
      "train/kl_loss_r1": 0.10304940491914749,
      "train/total_kl": 0.14260179549455643
    },
    {
      "epoch": 1.6961465571699303,
      "grad_norm": 1.8125,
      "learning_rate": 1.7278900821225522e-05,
      "loss": 0.2863,
      "mean_token_accuracy": 0.956343412399292,
      "step": 5370,
      "train/kl_loss_qwen": 0.0268328245729208,
      "train/kl_loss_r1": 0.10096690356731415,
      "train/total_kl": 0.12779972702264786
    },
    {
      "epoch": 1.6977258370183197,
      "grad_norm": 2.5625,
      "learning_rate": 1.7267056222362604e-05,
      "loss": 0.3248,
      "mean_token_accuracy": 0.9442677617073059,
      "step": 5375,
      "train/kl_loss_qwen": 0.045662587881088255,
      "train/kl_loss_r1": 0.0888807088136673,
      "train/total_kl": 0.13454329669475557
    },
    {
      "epoch": 1.6993051168667088,
      "grad_norm": 2.71875,
      "learning_rate": 1.7255211623499686e-05,
      "loss": 0.3994,
      "mean_token_accuracy": 0.9411943912506103,
      "step": 5380,
      "train/kl_loss_qwen": 0.0769625123590231,
      "train/kl_loss_r1": 0.11475804895162582,
      "train/total_kl": 0.19172056317329406
    },
    {
      "epoch": 1.700884396715098,
      "grad_norm": 2.421875,
      "learning_rate": 1.7243367024636764e-05,
      "loss": 0.302,
      "mean_token_accuracy": 0.9507578372955322,
      "step": 5385,
      "train/kl_loss_qwen": 0.0337438203394413,
      "train/kl_loss_r1": 0.0836663693189621,
      "train/total_kl": 0.11741019040346146
    },
    {
      "epoch": 1.702463676563487,
      "grad_norm": 3.265625,
      "learning_rate": 1.723152242577385e-05,
      "loss": 0.3082,
      "mean_token_accuracy": 0.9537096619606018,
      "step": 5390,
      "train/kl_loss_qwen": 0.03143431134521961,
      "train/kl_loss_r1": 0.09694580882787704,
      "train/total_kl": 0.12838011980056763
    },
    {
      "epoch": 1.7040429564118762,
      "grad_norm": 3.296875,
      "learning_rate": 1.721967782691093e-05,
      "loss": 0.2831,
      "mean_token_accuracy": 0.9585957169532776,
      "step": 5395,
      "train/kl_loss_qwen": 0.03435664810240269,
      "train/kl_loss_r1": 0.08732615411281586,
      "train/total_kl": 0.12168279886245728
    },
    {
      "epoch": 1.7056222362602653,
      "grad_norm": 3.65625,
      "learning_rate": 1.7207833228048014e-05,
      "loss": 0.2808,
      "mean_token_accuracy": 0.9582212448120118,
      "step": 5400,
      "train/kl_loss_qwen": 0.031883121095597745,
      "train/kl_loss_r1": 0.09409957677125931,
      "train/total_kl": 0.12598270028829575
    },
    {
      "epoch": 1.7072015161086544,
      "grad_norm": 2.171875,
      "learning_rate": 1.7195988629185092e-05,
      "loss": 0.2964,
      "mean_token_accuracy": 0.9524306654930115,
      "step": 5405,
      "train/kl_loss_qwen": 0.03191264420747757,
      "train/kl_loss_r1": 0.09875545427203178,
      "train/total_kl": 0.1306680992245674
    },
    {
      "epoch": 1.7087807959570436,
      "grad_norm": 2.5625,
      "learning_rate": 1.718414403032217e-05,
      "loss": 0.2671,
      "mean_token_accuracy": 0.9600597620010376,
      "step": 5410,
      "train/kl_loss_qwen": 0.02959233447909355,
      "train/kl_loss_r1": 0.0848437525331974,
      "train/total_kl": 0.11443608701229095
    },
    {
      "epoch": 1.7103600758054327,
      "grad_norm": 1.984375,
      "learning_rate": 1.7172299431459256e-05,
      "loss": 0.6408,
      "mean_token_accuracy": 0.9246001839637756,
      "step": 5415,
      "train/kl_loss_qwen": 0.16297519989311696,
      "train/kl_loss_r1": 0.22062818855047225,
      "train/total_kl": 0.3836033776402473
    },
    {
      "epoch": 1.7119393556538218,
      "grad_norm": 2.375,
      "learning_rate": 1.7160454832596335e-05,
      "loss": 0.3089,
      "mean_token_accuracy": 0.9589101433753967,
      "step": 5420,
      "train/kl_loss_qwen": 0.04998180791735649,
      "train/kl_loss_r1": 0.10031758025288581,
      "train/total_kl": 0.15029939115047455
    },
    {
      "epoch": 1.713518635502211,
      "grad_norm": 3.484375,
      "learning_rate": 1.7148610233733417e-05,
      "loss": 0.2799,
      "mean_token_accuracy": 0.9647825598716736,
      "step": 5425,
      "train/kl_loss_qwen": 0.046716898679733276,
      "train/kl_loss_r1": 0.09969376623630524,
      "train/total_kl": 0.14641066491603852
    },
    {
      "epoch": 1.7150979153506,
      "grad_norm": 1.84375,
      "learning_rate": 1.71367656348705e-05,
      "loss": 0.3,
      "mean_token_accuracy": 0.9561535596847535,
      "step": 5430,
      "train/kl_loss_qwen": 0.03709658198058605,
      "train/kl_loss_r1": 0.10376242101192475,
      "train/total_kl": 0.1408590018749237
    },
    {
      "epoch": 1.7166771951989892,
      "grad_norm": 2.4375,
      "learning_rate": 1.712492103600758e-05,
      "loss": 0.285,
      "mean_token_accuracy": 0.9582115411758423,
      "step": 5435,
      "train/kl_loss_qwen": 0.04015689045190811,
      "train/kl_loss_r1": 0.08389789015054702,
      "train/total_kl": 0.12405478060245514
    },
    {
      "epoch": 1.7182564750473786,
      "grad_norm": 2.765625,
      "learning_rate": 1.7113076437144663e-05,
      "loss": 0.2905,
      "mean_token_accuracy": 0.9617523312568664,
      "step": 5440,
      "train/kl_loss_qwen": 0.0418418899178505,
      "train/kl_loss_r1": 0.10740024447441102,
      "train/total_kl": 0.1492421343922615
    },
    {
      "epoch": 1.7198357548957675,
      "grad_norm": 2.3125,
      "learning_rate": 1.7101231838281745e-05,
      "loss": 0.2569,
      "mean_token_accuracy": 0.9679918050765991,
      "step": 5445,
      "train/kl_loss_qwen": 0.028497643768787384,
      "train/kl_loss_r1": 0.09242418557405471,
      "train/total_kl": 0.12092183083295822
    },
    {
      "epoch": 1.7214150347441568,
      "grad_norm": 2.90625,
      "learning_rate": 1.7089387239418823e-05,
      "loss": 0.2864,
      "mean_token_accuracy": 0.9562465786933899,
      "step": 5450,
      "train/kl_loss_qwen": 0.04152390211820602,
      "train/kl_loss_r1": 0.08986285328865051,
      "train/total_kl": 0.13138675540685654
    },
    {
      "epoch": 1.7229943145925457,
      "grad_norm": 1.546875,
      "learning_rate": 1.707754264055591e-05,
      "loss": 0.306,
      "mean_token_accuracy": 0.9552860975265502,
      "step": 5455,
      "train/kl_loss_qwen": 0.044347702711820605,
      "train/kl_loss_r1": 0.10618980377912521,
      "train/total_kl": 0.15053750872611998
    },
    {
      "epoch": 1.724573594440935,
      "grad_norm": 1.765625,
      "learning_rate": 1.7065698041692987e-05,
      "loss": 0.3015,
      "mean_token_accuracy": 0.9549282550811767,
      "step": 5460,
      "train/kl_loss_qwen": 0.028728816658258438,
      "train/kl_loss_r1": 0.08577648103237152,
      "train/total_kl": 0.11450529843568802
    },
    {
      "epoch": 1.726152874289324,
      "grad_norm": 1.59375,
      "learning_rate": 1.705385344283007e-05,
      "loss": 0.2946,
      "mean_token_accuracy": 0.9635149836540222,
      "step": 5465,
      "train/kl_loss_qwen": 0.03959320802241564,
      "train/kl_loss_r1": 0.12116357386112213,
      "train/total_kl": 0.16075678467750548
    },
    {
      "epoch": 1.7277321541377133,
      "grad_norm": 2.6875,
      "learning_rate": 1.704200884396715e-05,
      "loss": 0.2874,
      "mean_token_accuracy": 0.9489684462547302,
      "step": 5470,
      "train/kl_loss_qwen": 0.03745182454586029,
      "train/kl_loss_r1": 0.07799924463033676,
      "train/total_kl": 0.11545106768608093
    },
    {
      "epoch": 1.7293114339861022,
      "grad_norm": 4.9375,
      "learning_rate": 1.7030164245104233e-05,
      "loss": 0.3453,
      "mean_token_accuracy": 0.9493387341499329,
      "step": 5475,
      "train/kl_loss_qwen": 0.039210887253284456,
      "train/kl_loss_r1": 0.11339619904756545,
      "train/total_kl": 0.15260709077119827
    },
    {
      "epoch": 1.7308907138344916,
      "grad_norm": 1.78125,
      "learning_rate": 1.7018319646241315e-05,
      "loss": 0.2124,
      "mean_token_accuracy": 0.9698360443115235,
      "step": 5480,
      "train/kl_loss_qwen": 0.023600664734840394,
      "train/kl_loss_r1": 0.07380122914910317,
      "train/total_kl": 0.0974018931388855
    },
    {
      "epoch": 1.7324699936828805,
      "grad_norm": 2.734375,
      "learning_rate": 1.7006475047378397e-05,
      "loss": 0.3132,
      "mean_token_accuracy": 0.9635724902153016,
      "step": 5485,
      "train/kl_loss_qwen": 0.04452813044190407,
      "train/kl_loss_r1": 0.12937885224819184,
      "train/total_kl": 0.17390698492527007
    },
    {
      "epoch": 1.7340492735312698,
      "grad_norm": 2.421875,
      "learning_rate": 1.6994630448515476e-05,
      "loss": 0.2938,
      "mean_token_accuracy": 0.9638965129852295,
      "step": 5490,
      "train/kl_loss_qwen": 0.04617124870419502,
      "train/kl_loss_r1": 0.10705296769738197,
      "train/total_kl": 0.15322421491146088
    },
    {
      "epoch": 1.7356285533796587,
      "grad_norm": 2.234375,
      "learning_rate": 1.698278584965256e-05,
      "loss": 0.2947,
      "mean_token_accuracy": 0.96239013671875,
      "step": 5495,
      "train/kl_loss_qwen": 0.04116074815392494,
      "train/kl_loss_r1": 0.1035474494099617,
      "train/total_kl": 0.1447081983089447
    },
    {
      "epoch": 1.737207833228048,
      "grad_norm": 2.3125,
      "learning_rate": 1.697094125078964e-05,
      "loss": 0.3019,
      "mean_token_accuracy": 0.9590927600860596,
      "step": 5500,
      "train/kl_loss_qwen": 0.04504848662763834,
      "train/kl_loss_r1": 0.10310927331447602,
      "train/total_kl": 0.14815776348114013
    },
    {
      "epoch": 1.7387871130764372,
      "grad_norm": 1.9375,
      "learning_rate": 1.695909665192672e-05,
      "loss": 0.2526,
      "mean_token_accuracy": 0.9650665640830993,
      "step": 5505,
      "train/kl_loss_qwen": 0.03589091822504997,
      "train/kl_loss_r1": 0.08365176469087601,
      "train/total_kl": 0.11954268664121628
    },
    {
      "epoch": 1.7403663929248263,
      "grad_norm": 2.03125,
      "learning_rate": 1.6947252053063804e-05,
      "loss": 0.267,
      "mean_token_accuracy": 0.9649411678314209,
      "step": 5510,
      "train/kl_loss_qwen": 0.039500415325164795,
      "train/kl_loss_r1": 0.09261668175458908,
      "train/total_kl": 0.13211709260940552
    },
    {
      "epoch": 1.7419456727732154,
      "grad_norm": 1.609375,
      "learning_rate": 1.6935407454200886e-05,
      "loss": 0.2675,
      "mean_token_accuracy": 0.96248699426651,
      "step": 5515,
      "train/kl_loss_qwen": 0.04042439796030521,
      "train/kl_loss_r1": 0.08201149255037307,
      "train/total_kl": 0.1224358931183815
    },
    {
      "epoch": 1.7435249526216046,
      "grad_norm": 2.390625,
      "learning_rate": 1.6923562855337968e-05,
      "loss": 0.2548,
      "mean_token_accuracy": 0.9660545587539673,
      "step": 5520,
      "train/kl_loss_qwen": 0.029598009213805197,
      "train/kl_loss_r1": 0.0856412798166275,
      "train/total_kl": 0.11523928791284561
    },
    {
      "epoch": 1.7451042324699937,
      "grad_norm": 1.75,
      "learning_rate": 1.691171825647505e-05,
      "loss": 0.2421,
      "mean_token_accuracy": 0.9617631912231446,
      "step": 5525,
      "train/kl_loss_qwen": 0.029380543902516366,
      "train/kl_loss_r1": 0.07366244196891784,
      "train/total_kl": 0.1030429869890213
    },
    {
      "epoch": 1.7466835123183828,
      "grad_norm": 2.0,
      "learning_rate": 1.6899873657612128e-05,
      "loss": 0.2763,
      "mean_token_accuracy": 0.9690828561782837,
      "step": 5530,
      "train/kl_loss_qwen": 0.03207040280103683,
      "train/kl_loss_r1": 0.11489382013678551,
      "train/total_kl": 0.14696422219276428
    },
    {
      "epoch": 1.748262792166772,
      "grad_norm": 1.7265625,
      "learning_rate": 1.688802905874921e-05,
      "loss": 0.3124,
      "mean_token_accuracy": 0.9569051861763,
      "step": 5535,
      "train/kl_loss_qwen": 0.047816285863518715,
      "train/kl_loss_r1": 0.10279338657855988,
      "train/total_kl": 0.15060966908931733
    },
    {
      "epoch": 1.749842072015161,
      "grad_norm": 3.1875,
      "learning_rate": 1.6876184459886292e-05,
      "loss": 0.3317,
      "mean_token_accuracy": 0.9552275776863098,
      "step": 5540,
      "train/kl_loss_qwen": 0.04709714464843273,
      "train/kl_loss_r1": 0.12254572957754135,
      "train/total_kl": 0.16964287608861922
    },
    {
      "epoch": 1.7514213518635502,
      "grad_norm": 2.53125,
      "learning_rate": 1.6864339861023374e-05,
      "loss": 0.2747,
      "mean_token_accuracy": 0.963461983203888,
      "step": 5545,
      "train/kl_loss_qwen": 0.0385147362947464,
      "train/kl_loss_r1": 0.09029649645090103,
      "train/total_kl": 0.12881123423576354
    },
    {
      "epoch": 1.7530006317119393,
      "grad_norm": 1.859375,
      "learning_rate": 1.6852495262160456e-05,
      "loss": 0.2621,
      "mean_token_accuracy": 0.9662156462669372,
      "step": 5550,
      "train/kl_loss_qwen": 0.03789428398013115,
      "train/kl_loss_r1": 0.09142358005046844,
      "train/total_kl": 0.12931786328554154
    },
    {
      "epoch": 1.7545799115603284,
      "grad_norm": 3.796875,
      "learning_rate": 1.6840650663297535e-05,
      "loss": 0.2749,
      "mean_token_accuracy": 0.9617294430732727,
      "step": 5555,
      "train/kl_loss_qwen": 0.03047870434820652,
      "train/kl_loss_r1": 0.09475654140114784,
      "train/total_kl": 0.1252352461218834
    },
    {
      "epoch": 1.7561591914087176,
      "grad_norm": 1.90625,
      "learning_rate": 1.682880606443462e-05,
      "loss": 0.3073,
      "mean_token_accuracy": 0.9553132891654968,
      "step": 5560,
      "train/kl_loss_qwen": 0.03721968401223421,
      "train/kl_loss_r1": 0.09573861807584763,
      "train/total_kl": 0.13295830190181732
    },
    {
      "epoch": 1.757738471257107,
      "grad_norm": 2.609375,
      "learning_rate": 1.68169614655717e-05,
      "loss": 0.2845,
      "mean_token_accuracy": 0.9618149399757385,
      "step": 5565,
      "train/kl_loss_qwen": 0.03084134254604578,
      "train/kl_loss_r1": 0.09581469148397445,
      "train/total_kl": 0.12665603458881378
    },
    {
      "epoch": 1.7593177511054958,
      "grad_norm": 2.78125,
      "learning_rate": 1.680511686670878e-05,
      "loss": 0.3053,
      "mean_token_accuracy": 0.9594308614730835,
      "step": 5570,
      "train/kl_loss_qwen": 0.031242860853672026,
      "train/kl_loss_r1": 0.11332251504063606,
      "train/total_kl": 0.14456537663936614
    },
    {
      "epoch": 1.7608970309538852,
      "grad_norm": 2.46875,
      "learning_rate": 1.6793272267845862e-05,
      "loss": 0.2906,
      "mean_token_accuracy": 0.9627347350120544,
      "step": 5575,
      "train/kl_loss_qwen": 0.027759709022939207,
      "train/kl_loss_r1": 0.09678383618593216,
      "train/total_kl": 0.1245435431599617
    },
    {
      "epoch": 1.762476310802274,
      "grad_norm": 1.59375,
      "learning_rate": 1.6781427668982944e-05,
      "loss": 0.2459,
      "mean_token_accuracy": 0.9702896237373352,
      "step": 5580,
      "train/kl_loss_qwen": 0.03235093113034963,
      "train/kl_loss_r1": 0.09168237000703812,
      "train/total_kl": 0.12403329908847809
    },
    {
      "epoch": 1.7640555906506634,
      "grad_norm": 4.15625,
      "learning_rate": 1.6769583070120026e-05,
      "loss": 0.2955,
      "mean_token_accuracy": 0.9551283121109009,
      "step": 5585,
      "train/kl_loss_qwen": 0.038788830302655695,
      "train/kl_loss_r1": 0.09840352088212967,
      "train/total_kl": 0.13719235211610795
    },
    {
      "epoch": 1.7656348704990523,
      "grad_norm": 2.390625,
      "learning_rate": 1.675773847125711e-05,
      "loss": 0.2692,
      "mean_token_accuracy": 0.9665759325027465,
      "step": 5590,
      "train/kl_loss_qwen": 0.04146883487701416,
      "train/kl_loss_r1": 0.09456804394721985,
      "train/total_kl": 0.13603688180446624
    },
    {
      "epoch": 1.7672141503474417,
      "grad_norm": 4.03125,
      "learning_rate": 1.6745893872394187e-05,
      "loss": 0.2701,
      "mean_token_accuracy": 0.9619617938995362,
      "step": 5595,
      "train/kl_loss_qwen": 0.036147027462720874,
      "train/kl_loss_r1": 0.09218544065952301,
      "train/total_kl": 0.12833246886730193
    },
    {
      "epoch": 1.7687934301958306,
      "grad_norm": 2.8125,
      "learning_rate": 1.6734049273531272e-05,
      "loss": 0.2741,
      "mean_token_accuracy": 0.9605605363845825,
      "step": 5600,
      "train/kl_loss_qwen": 0.04608409740030765,
      "train/kl_loss_r1": 0.0873634472489357,
      "train/total_kl": 0.13344754427671432
    },
    {
      "epoch": 1.77037271004422,
      "grad_norm": 2.515625,
      "learning_rate": 1.672220467466835e-05,
      "loss": 0.3205,
      "mean_token_accuracy": 0.9565576791763306,
      "step": 5605,
      "train/kl_loss_qwen": 0.045622337982058524,
      "train/kl_loss_r1": 0.11839895397424698,
      "train/total_kl": 0.16402129381895064
    },
    {
      "epoch": 1.7719519898926088,
      "grad_norm": 4.25,
      "learning_rate": 1.6710360075805433e-05,
      "loss": 0.2922,
      "mean_token_accuracy": 0.9553581953048706,
      "step": 5610,
      "train/kl_loss_qwen": 0.036015570908784864,
      "train/kl_loss_r1": 0.0874337188899517,
      "train/total_kl": 0.12344928681850434
    },
    {
      "epoch": 1.7735312697409982,
      "grad_norm": 2.5625,
      "learning_rate": 1.6698515476942515e-05,
      "loss": 0.2341,
      "mean_token_accuracy": 0.968879497051239,
      "step": 5615,
      "train/kl_loss_qwen": 0.02184625491499901,
      "train/kl_loss_r1": 0.07908256128430366,
      "train/total_kl": 0.10092881470918655
    },
    {
      "epoch": 1.775110549589387,
      "grad_norm": 3.046875,
      "learning_rate": 1.6686670878079597e-05,
      "loss": 0.3069,
      "mean_token_accuracy": 0.9527592182159423,
      "step": 5620,
      "train/kl_loss_qwen": 0.032631196454167365,
      "train/kl_loss_r1": 0.09164433628320694,
      "train/total_kl": 0.12427553236484527
    },
    {
      "epoch": 1.7766898294377764,
      "grad_norm": 3.421875,
      "learning_rate": 1.667482627921668e-05,
      "loss": 0.2871,
      "mean_token_accuracy": 0.9542339086532593,
      "step": 5625,
      "train/kl_loss_qwen": 0.03521231710910797,
      "train/kl_loss_r1": 0.08778002709150315,
      "train/total_kl": 0.12299234420061111
    },
    {
      "epoch": 1.7782691092861656,
      "grad_norm": 3.25,
      "learning_rate": 1.666298168035376e-05,
      "loss": 0.3076,
      "mean_token_accuracy": 0.9599844813346863,
      "step": 5630,
      "train/kl_loss_qwen": 0.04490780346095562,
      "train/kl_loss_r1": 0.09518520981073379,
      "train/total_kl": 0.14009301066398622
    },
    {
      "epoch": 1.7798483891345547,
      "grad_norm": 3.0625,
      "learning_rate": 1.665113708149084e-05,
      "loss": 0.2693,
      "mean_token_accuracy": 0.9598155856132508,
      "step": 5635,
      "train/kl_loss_qwen": 0.03309611305594444,
      "train/kl_loss_r1": 0.0819433532655239,
      "train/total_kl": 0.11503946483135223
    },
    {
      "epoch": 1.7814276689829438,
      "grad_norm": 2.046875,
      "learning_rate": 1.6639292482627925e-05,
      "loss": 0.2588,
      "mean_token_accuracy": 0.9681211709976196,
      "step": 5640,
      "train/kl_loss_qwen": 0.03856561332941055,
      "train/kl_loss_r1": 0.09067067578434944,
      "train/total_kl": 0.12923628836870193
    },
    {
      "epoch": 1.783006948831333,
      "grad_norm": 2.125,
      "learning_rate": 1.6627447883765003e-05,
      "loss": 0.2906,
      "mean_token_accuracy": 0.9620175838470459,
      "step": 5645,
      "train/kl_loss_qwen": 0.040424026921391486,
      "train/kl_loss_r1": 0.10171355605125428,
      "train/total_kl": 0.14213758260011672
    },
    {
      "epoch": 1.784586228679722,
      "grad_norm": 1.953125,
      "learning_rate": 1.6615603284902085e-05,
      "loss": 0.5615,
      "mean_token_accuracy": 0.9238213896751404,
      "step": 5650,
      "train/kl_loss_qwen": 0.12038093991577625,
      "train/kl_loss_r1": 0.1826952166855335,
      "train/total_kl": 0.3030761435627937
    },
    {
      "epoch": 1.7861655085281112,
      "grad_norm": 4.125,
      "learning_rate": 1.6603758686039167e-05,
      "loss": 0.3002,
      "mean_token_accuracy": 0.9550010323524475,
      "step": 5655,
      "train/kl_loss_qwen": 0.02749313823878765,
      "train/kl_loss_r1": 0.08474343419075012,
      "train/total_kl": 0.11223657131195068
    },
    {
      "epoch": 1.7877447883765003,
      "grad_norm": 2.6875,
      "learning_rate": 1.6591914087176246e-05,
      "loss": 0.2473,
      "mean_token_accuracy": 0.965313708782196,
      "step": 5660,
      "train/kl_loss_qwen": 0.029064038023352622,
      "train/kl_loss_r1": 0.08924749568104744,
      "train/total_kl": 0.11831153333187103
    },
    {
      "epoch": 1.7893240682248894,
      "grad_norm": 1.828125,
      "learning_rate": 1.658006948831333e-05,
      "loss": 0.2449,
      "mean_token_accuracy": 0.9681156516075134,
      "step": 5665,
      "train/kl_loss_qwen": 0.03453796431422233,
      "train/kl_loss_r1": 0.08106319606304169,
      "train/total_kl": 0.11560116112232208
    },
    {
      "epoch": 1.7909033480732786,
      "grad_norm": 2.75,
      "learning_rate": 1.656822488945041e-05,
      "loss": 0.2897,
      "mean_token_accuracy": 0.9638298630714417,
      "step": 5670,
      "train/kl_loss_qwen": 0.04059787057340145,
      "train/kl_loss_r1": 0.10362098813056946,
      "train/total_kl": 0.14421885907649995
    },
    {
      "epoch": 1.7924826279216677,
      "grad_norm": 2.3125,
      "learning_rate": 1.6556380290587492e-05,
      "loss": 0.305,
      "mean_token_accuracy": 0.9536419749259949,
      "step": 5675,
      "train/kl_loss_qwen": 0.028426201641559602,
      "train/kl_loss_r1": 0.09434613734483718,
      "train/total_kl": 0.12277233749628066
    },
    {
      "epoch": 1.7940619077700568,
      "grad_norm": 1.8359375,
      "learning_rate": 1.6544535691724574e-05,
      "loss": 0.2555,
      "mean_token_accuracy": 0.9712459444999695,
      "step": 5680,
      "train/kl_loss_qwen": 0.03972443044185638,
      "train/kl_loss_r1": 0.09956694170832633,
      "train/total_kl": 0.13929136991500854
    },
    {
      "epoch": 1.795641187618446,
      "grad_norm": 2.078125,
      "learning_rate": 1.6532691092861656e-05,
      "loss": 0.2698,
      "mean_token_accuracy": 0.9589987635612488,
      "step": 5685,
      "train/kl_loss_qwen": 0.040585583075881004,
      "train/kl_loss_r1": 0.08037491142749786,
      "train/total_kl": 0.12096049636602402
    },
    {
      "epoch": 1.7972204674668353,
      "grad_norm": 2.78125,
      "learning_rate": 1.6520846493998734e-05,
      "loss": 0.3194,
      "mean_token_accuracy": 0.9590429067611694,
      "step": 5690,
      "train/kl_loss_qwen": 0.05336033515632153,
      "train/kl_loss_r1": 0.11085460036993026,
      "train/total_kl": 0.16421493291854858
    },
    {
      "epoch": 1.7987997473152242,
      "grad_norm": 1.7734375,
      "learning_rate": 1.650900189513582e-05,
      "loss": 0.2999,
      "mean_token_accuracy": 0.9490546107292175,
      "step": 5695,
      "train/kl_loss_qwen": 0.03574012760072946,
      "train/kl_loss_r1": 0.07749325782060623,
      "train/total_kl": 0.11323338449001312
    },
    {
      "epoch": 1.8003790271636135,
      "grad_norm": 3.875,
      "learning_rate": 1.6497157296272898e-05,
      "loss": 0.2559,
      "mean_token_accuracy": 0.9611058473587036,
      "step": 5700,
      "train/kl_loss_qwen": 0.026648689061403275,
      "train/kl_loss_r1": 0.08652327507734299,
      "train/total_kl": 0.11317196488380432
    },
    {
      "epoch": 1.8019583070120024,
      "grad_norm": 2.375,
      "learning_rate": 1.6485312697409984e-05,
      "loss": 0.2864,
      "mean_token_accuracy": 0.9685863494873047,
      "step": 5705,
      "train/kl_loss_qwen": 0.04138276837766171,
      "train/kl_loss_r1": 0.11612822860479355,
      "train/total_kl": 0.15751099586486816
    },
    {
      "epoch": 1.8035375868603918,
      "grad_norm": 3.859375,
      "learning_rate": 1.6473468098547062e-05,
      "loss": 0.5882,
      "mean_token_accuracy": 0.9331369400024414,
      "step": 5710,
      "train/kl_loss_qwen": 0.14021447133272885,
      "train/kl_loss_r1": 0.22392163425683975,
      "train/total_kl": 0.3641360908746719
    },
    {
      "epoch": 1.8051168667087807,
      "grad_norm": 2.625,
      "learning_rate": 1.6461623499684144e-05,
      "loss": 0.2706,
      "mean_token_accuracy": 0.9565688967704773,
      "step": 5715,
      "train/kl_loss_qwen": 0.03656572550535202,
      "train/kl_loss_r1": 0.0797874614596367,
      "train/total_kl": 0.11635318845510483
    },
    {
      "epoch": 1.80669614655717,
      "grad_norm": 2.8125,
      "learning_rate": 1.6449778900821226e-05,
      "loss": 0.3157,
      "mean_token_accuracy": 0.9469430327415467,
      "step": 5720,
      "train/kl_loss_qwen": 0.029701898619532586,
      "train/kl_loss_r1": 0.08662456125020981,
      "train/total_kl": 0.11632645726203919
    },
    {
      "epoch": 1.808275426405559,
      "grad_norm": 2.921875,
      "learning_rate": 1.6437934301958308e-05,
      "loss": 0.2776,
      "mean_token_accuracy": 0.9631029725074768,
      "step": 5725,
      "train/kl_loss_qwen": 0.04505093172192574,
      "train/kl_loss_r1": 0.09170481711626052,
      "train/total_kl": 0.1367557466030121
    },
    {
      "epoch": 1.8098547062539483,
      "grad_norm": 3.0,
      "learning_rate": 1.6426089703095387e-05,
      "loss": 0.4702,
      "mean_token_accuracy": 0.9393811464309693,
      "step": 5730,
      "train/kl_loss_qwen": 0.08342030048370361,
      "train/kl_loss_r1": 0.18051456212997435,
      "train/total_kl": 0.263934862613678
    },
    {
      "epoch": 1.8114339861023372,
      "grad_norm": 4.6875,
      "learning_rate": 1.6414245104232472e-05,
      "loss": 0.3027,
      "mean_token_accuracy": 0.9687475681304931,
      "step": 5735,
      "train/kl_loss_qwen": 0.05649250112473965,
      "train/kl_loss_r1": 0.12462752759456634,
      "train/total_kl": 0.18112003058195114
    },
    {
      "epoch": 1.8130132659507265,
      "grad_norm": 4.09375,
      "learning_rate": 1.640240050536955e-05,
      "loss": 0.2903,
      "mean_token_accuracy": 0.9489778876304626,
      "step": 5740,
      "train/kl_loss_qwen": 0.020765667781233788,
      "train/kl_loss_r1": 0.06855228245258331,
      "train/total_kl": 0.08931795060634613
    },
    {
      "epoch": 1.8145925457991154,
      "grad_norm": 3.15625,
      "learning_rate": 1.6390555906506636e-05,
      "loss": 0.3353,
      "mean_token_accuracy": 0.9521038651466369,
      "step": 5745,
      "train/kl_loss_qwen": 0.04363179095089435,
      "train/kl_loss_r1": 0.11793040856719017,
      "train/total_kl": 0.16156219989061354
    },
    {
      "epoch": 1.8161718256475048,
      "grad_norm": 3.0,
      "learning_rate": 1.6378711307643715e-05,
      "loss": 0.2657,
      "mean_token_accuracy": 0.9658042669296265,
      "step": 5750,
      "train/kl_loss_qwen": 0.04214836992323399,
      "train/kl_loss_r1": 0.08716838359832764,
      "train/total_kl": 0.1293167531490326
    },
    {
      "epoch": 1.817751105495894,
      "grad_norm": 4.96875,
      "learning_rate": 1.6366866708780797e-05,
      "loss": 0.2892,
      "mean_token_accuracy": 0.9653367161750793,
      "step": 5755,
      "train/kl_loss_qwen": 0.04844001904129982,
      "train/kl_loss_r1": 0.10596210062503815,
      "train/total_kl": 0.15440211892127992
    },
    {
      "epoch": 1.819330385344283,
      "grad_norm": 2.265625,
      "learning_rate": 1.635502210991788e-05,
      "loss": 0.2948,
      "mean_token_accuracy": 0.9664183259010315,
      "step": 5760,
      "train/kl_loss_qwen": 0.04634863212704658,
      "train/kl_loss_r1": 0.11950081884860993,
      "train/total_kl": 0.16584945619106292
    },
    {
      "epoch": 1.8209096651926722,
      "grad_norm": 1.8984375,
      "learning_rate": 1.634317751105496e-05,
      "loss": 0.309,
      "mean_token_accuracy": 0.9553698420524597,
      "step": 5765,
      "train/kl_loss_qwen": 0.03656389787793159,
      "train/kl_loss_r1": 0.1102949783205986,
      "train/total_kl": 0.1468588724732399
    },
    {
      "epoch": 1.8224889450410613,
      "grad_norm": 4.84375,
      "learning_rate": 1.633133291219204e-05,
      "loss": 0.3177,
      "mean_token_accuracy": 0.965743362903595,
      "step": 5770,
      "train/kl_loss_qwen": 0.049697655439376834,
      "train/kl_loss_r1": 0.12427047491073609,
      "train/total_kl": 0.17396812736988068
    },
    {
      "epoch": 1.8240682248894504,
      "grad_norm": 1.4296875,
      "learning_rate": 1.6319488313329124e-05,
      "loss": 0.2983,
      "mean_token_accuracy": 0.9514210820198059,
      "step": 5775,
      "train/kl_loss_qwen": 0.032893312722444536,
      "train/kl_loss_r1": 0.08869104385375977,
      "train/total_kl": 0.12158435583114624
    },
    {
      "epoch": 1.8256475047378395,
      "grad_norm": 5.0,
      "learning_rate": 1.6307643714466203e-05,
      "loss": 0.3081,
      "mean_token_accuracy": 0.9577853083610535,
      "step": 5780,
      "train/kl_loss_qwen": 0.03474757447838783,
      "train/kl_loss_r1": 0.11157671734690666,
      "train/total_kl": 0.1463242933154106
    },
    {
      "epoch": 1.8272267845862287,
      "grad_norm": 2.25,
      "learning_rate": 1.6295799115603285e-05,
      "loss": 0.3178,
      "mean_token_accuracy": 0.9511246800422668,
      "step": 5785,
      "train/kl_loss_qwen": 0.039079074189066886,
      "train/kl_loss_r1": 0.10459970608353615,
      "train/total_kl": 0.14367877840995788
    },
    {
      "epoch": 1.8288060644346178,
      "grad_norm": 3.125,
      "learning_rate": 1.6283954516740367e-05,
      "loss": 0.3108,
      "mean_token_accuracy": 0.9583800911903382,
      "step": 5790,
      "train/kl_loss_qwen": 0.044334367662668225,
      "train/kl_loss_r1": 0.11488586217164994,
      "train/total_kl": 0.15922023057937623
    },
    {
      "epoch": 1.830385344283007,
      "grad_norm": 2.015625,
      "learning_rate": 1.6272109917877446e-05,
      "loss": 0.3357,
      "mean_token_accuracy": 0.9565713763236999,
      "step": 5795,
      "train/kl_loss_qwen": 0.05595830902457237,
      "train/kl_loss_r1": 0.12073354721069336,
      "train/total_kl": 0.17669185549020766
    },
    {
      "epoch": 1.831964624131396,
      "grad_norm": 2.875,
      "learning_rate": 1.626026531901453e-05,
      "loss": 0.2708,
      "mean_token_accuracy": 0.9607766747474671,
      "step": 5800,
      "train/kl_loss_qwen": 0.03445180505514145,
      "train/kl_loss_r1": 0.08090831190347672,
      "train/total_kl": 0.11536011844873428
    },
    {
      "epoch": 1.8335439039797852,
      "grad_norm": 2.765625,
      "learning_rate": 1.624842072015161e-05,
      "loss": 0.3031,
      "mean_token_accuracy": 0.9587064266204834,
      "step": 5805,
      "train/kl_loss_qwen": 0.06228230819106102,
      "train/kl_loss_r1": 0.10058343708515168,
      "train/total_kl": 0.1628657504916191
    },
    {
      "epoch": 1.8351231838281743,
      "grad_norm": 2.109375,
      "learning_rate": 1.6236576121288695e-05,
      "loss": 0.2691,
      "mean_token_accuracy": 0.9599324107170105,
      "step": 5810,
      "train/kl_loss_qwen": 0.03734166398644447,
      "train/kl_loss_r1": 0.08514229506254196,
      "train/total_kl": 0.12248395830392837
    },
    {
      "epoch": 1.8367024636765636,
      "grad_norm": 2.75,
      "learning_rate": 1.6224731522425773e-05,
      "loss": 0.2527,
      "mean_token_accuracy": 0.9670416951179505,
      "step": 5815,
      "train/kl_loss_qwen": 0.03311895579099655,
      "train/kl_loss_r1": 0.07957748249173165,
      "train/total_kl": 0.11269643902778625
    },
    {
      "epoch": 1.8382817435249525,
      "grad_norm": 2.65625,
      "learning_rate": 1.6212886923562855e-05,
      "loss": 0.2658,
      "mean_token_accuracy": 0.9679569959640503,
      "step": 5820,
      "train/kl_loss_qwen": 0.030814138427376746,
      "train/kl_loss_r1": 0.09203279539942741,
      "train/total_kl": 0.12284693717956544
    },
    {
      "epoch": 1.839861023373342,
      "grad_norm": 3.328125,
      "learning_rate": 1.6201042324699937e-05,
      "loss": 0.3062,
      "mean_token_accuracy": 0.9636016726493836,
      "step": 5825,
      "train/kl_loss_qwen": 0.04324283748865128,
      "train/kl_loss_r1": 0.11632476150989532,
      "train/total_kl": 0.15956759601831436
    },
    {
      "epoch": 1.8414403032217308,
      "grad_norm": 2.15625,
      "learning_rate": 1.618919772583702e-05,
      "loss": 0.2536,
      "mean_token_accuracy": 0.9682993292808533,
      "step": 5830,
      "train/kl_loss_qwen": 0.035114158689975736,
      "train/kl_loss_r1": 0.0935179352760315,
      "train/total_kl": 0.12863209396600722
    },
    {
      "epoch": 1.8430195830701201,
      "grad_norm": 4.375,
      "learning_rate": 1.6177353126974098e-05,
      "loss": 0.5091,
      "mean_token_accuracy": 0.9356011629104615,
      "step": 5835,
      "train/kl_loss_qwen": 0.09901498965919017,
      "train/kl_loss_r1": 0.1867767319083214,
      "train/total_kl": 0.28579171448946
    },
    {
      "epoch": 1.844598862918509,
      "grad_norm": 1.7578125,
      "learning_rate": 1.6165508528111183e-05,
      "loss": 0.2907,
      "mean_token_accuracy": 0.9634949326515198,
      "step": 5840,
      "train/kl_loss_qwen": 0.03223560713231564,
      "train/kl_loss_r1": 0.11041826978325844,
      "train/total_kl": 0.14265387654304504
    },
    {
      "epoch": 1.8461781427668984,
      "grad_norm": 2.203125,
      "learning_rate": 1.6153663929248262e-05,
      "loss": 0.3842,
      "mean_token_accuracy": 0.9356426119804382,
      "step": 5845,
      "train/kl_loss_qwen": 0.05927629731595516,
      "train/kl_loss_r1": 0.10627279579639434,
      "train/total_kl": 0.16554909348487853
    },
    {
      "epoch": 1.8477574226152873,
      "grad_norm": 2.625,
      "learning_rate": 1.6141819330385347e-05,
      "loss": 0.2721,
      "mean_token_accuracy": 0.959587836265564,
      "step": 5850,
      "train/kl_loss_qwen": 0.03372352048754692,
      "train/kl_loss_r1": 0.09296881929039955,
      "train/total_kl": 0.12669233977794647
    },
    {
      "epoch": 1.8493367024636767,
      "grad_norm": 3.859375,
      "learning_rate": 1.6129974731522426e-05,
      "loss": 0.297,
      "mean_token_accuracy": 0.9599341988563538,
      "step": 5855,
      "train/kl_loss_qwen": 0.03543678373098373,
      "train/kl_loss_r1": 0.10404598563909531,
      "train/total_kl": 0.13948277086019517
    },
    {
      "epoch": 1.8509159823120656,
      "grad_norm": 5.09375,
      "learning_rate": 1.6118130132659508e-05,
      "loss": 0.4117,
      "mean_token_accuracy": 0.9453917026519776,
      "step": 5860,
      "train/kl_loss_qwen": 0.0653363436460495,
      "train/kl_loss_r1": 0.1511827141046524,
      "train/total_kl": 0.21651905179023742
    },
    {
      "epoch": 1.852495262160455,
      "grad_norm": 2.796875,
      "learning_rate": 1.610628553379659e-05,
      "loss": 0.2462,
      "mean_token_accuracy": 0.961195456981659,
      "step": 5865,
      "train/kl_loss_qwen": 0.023498565703630448,
      "train/kl_loss_r1": 0.07998940795660019,
      "train/total_kl": 0.1034879744052887
    },
    {
      "epoch": 1.8540745420088438,
      "grad_norm": 3.15625,
      "learning_rate": 1.6094440934933672e-05,
      "loss": 0.2775,
      "mean_token_accuracy": 0.9609525680541993,
      "step": 5870,
      "train/kl_loss_qwen": 0.040575283393263814,
      "train/kl_loss_r1": 0.09019716084003448,
      "train/total_kl": 0.1307724416255951
    },
    {
      "epoch": 1.8556538218572332,
      "grad_norm": 2.375,
      "learning_rate": 1.608259633607075e-05,
      "loss": 0.3145,
      "mean_token_accuracy": 0.9548525810241699,
      "step": 5875,
      "train/kl_loss_qwen": 0.04898911379277706,
      "train/kl_loss_r1": 0.10325862914323806,
      "train/total_kl": 0.15224774032831193
    },
    {
      "epoch": 1.8572331017056223,
      "grad_norm": 2.296875,
      "learning_rate": 1.6070751737207836e-05,
      "loss": 0.2966,
      "mean_token_accuracy": 0.9572304129600525,
      "step": 5880,
      "train/kl_loss_qwen": 0.03643493503332138,
      "train/kl_loss_r1": 0.09935184195637703,
      "train/total_kl": 0.13578677773475648
    },
    {
      "epoch": 1.8588123815540114,
      "grad_norm": 4.5625,
      "learning_rate": 1.6058907138344914e-05,
      "loss": 0.3039,
      "mean_token_accuracy": 0.9627833724021911,
      "step": 5885,
      "train/kl_loss_qwen": 0.05323019288480282,
      "train/kl_loss_r1": 0.11391352415084839,
      "train/total_kl": 0.16714371293783187
    },
    {
      "epoch": 1.8603916614024005,
      "grad_norm": 3.515625,
      "learning_rate": 1.6047062539482e-05,
      "loss": 0.3365,
      "mean_token_accuracy": 0.9548451900482178,
      "step": 5890,
      "train/kl_loss_qwen": 0.0407744038850069,
      "train/kl_loss_r1": 0.10744823217391967,
      "train/total_kl": 0.14822263568639754
    },
    {
      "epoch": 1.8619709412507897,
      "grad_norm": 2.71875,
      "learning_rate": 1.6035217940619078e-05,
      "loss": 0.3259,
      "mean_token_accuracy": 0.9630619287490845,
      "step": 5895,
      "train/kl_loss_qwen": 0.03142917361110449,
      "train/kl_loss_r1": 0.1302987203001976,
      "train/total_kl": 0.1617278903722763
    },
    {
      "epoch": 1.8635502210991788,
      "grad_norm": 2.25,
      "learning_rate": 1.602337334175616e-05,
      "loss": 0.299,
      "mean_token_accuracy": 0.9485944032669067,
      "step": 5900,
      "train/kl_loss_qwen": 0.03341102749109268,
      "train/kl_loss_r1": 0.08500505909323693,
      "train/total_kl": 0.11841608434915543
    },
    {
      "epoch": 1.865129500947568,
      "grad_norm": 2.40625,
      "learning_rate": 1.6011528742893242e-05,
      "loss": 0.2882,
      "mean_token_accuracy": 0.9667325258255005,
      "step": 5905,
      "train/kl_loss_qwen": 0.03421681001782417,
      "train/kl_loss_r1": 0.10960807204246521,
      "train/total_kl": 0.14382488131523133
    },
    {
      "epoch": 1.866708780795957,
      "grad_norm": 2.109375,
      "learning_rate": 1.599968414403032e-05,
      "loss": 0.5307,
      "mean_token_accuracy": 0.9375939488410949,
      "step": 5910,
      "train/kl_loss_qwen": 0.09936241395771503,
      "train/kl_loss_r1": 0.22132405191659926,
      "train/total_kl": 0.3206864669919014
    },
    {
      "epoch": 1.8682880606443462,
      "grad_norm": 1.671875,
      "learning_rate": 1.5987839545167403e-05,
      "loss": 0.278,
      "mean_token_accuracy": 0.951450788974762,
      "step": 5915,
      "train/kl_loss_qwen": 0.03564399927854538,
      "train/kl_loss_r1": 0.0800016961991787,
      "train/total_kl": 0.11564569771289826
    },
    {
      "epoch": 1.8698673404927353,
      "grad_norm": 3.84375,
      "learning_rate": 1.5975994946304485e-05,
      "loss": 0.3353,
      "mean_token_accuracy": 0.9605056881904602,
      "step": 5920,
      "train/kl_loss_qwen": 0.04541250579059124,
      "train/kl_loss_r1": 0.14195723831653595,
      "train/total_kl": 0.18736974000930787
    },
    {
      "epoch": 1.8714466203411244,
      "grad_norm": 3.0,
      "learning_rate": 1.5964150347441567e-05,
      "loss": 0.2926,
      "mean_token_accuracy": 0.9479851245880127,
      "step": 5925,
      "train/kl_loss_qwen": 0.03603140078485012,
      "train/kl_loss_r1": 0.07890548780560494,
      "train/total_kl": 0.11493688970804214
    },
    {
      "epoch": 1.8730259001895135,
      "grad_norm": 3.609375,
      "learning_rate": 1.595230574857865e-05,
      "loss": 0.298,
      "mean_token_accuracy": 0.9577114939689636,
      "step": 5930,
      "train/kl_loss_qwen": 0.03593801185488701,
      "train/kl_loss_r1": 0.09092163443565368,
      "train/total_kl": 0.12685964703559877
    },
    {
      "epoch": 1.8746051800379027,
      "grad_norm": 2.5625,
      "learning_rate": 1.594046114971573e-05,
      "loss": 0.2546,
      "mean_token_accuracy": 0.9603676795959473,
      "step": 5935,
      "train/kl_loss_qwen": 0.0346614133566618,
      "train/kl_loss_r1": 0.07691185548901558,
      "train/total_kl": 0.11157326996326447
    },
    {
      "epoch": 1.876184459886292,
      "grad_norm": 2.71875,
      "learning_rate": 1.592861655085281e-05,
      "loss": 0.2741,
      "mean_token_accuracy": 0.9572114944458008,
      "step": 5940,
      "train/kl_loss_qwen": 0.03730158545076847,
      "train/kl_loss_r1": 0.0783933475613594,
      "train/total_kl": 0.11569493263959885
    },
    {
      "epoch": 1.877763739734681,
      "grad_norm": 2.515625,
      "learning_rate": 1.5916771951989895e-05,
      "loss": 0.2872,
      "mean_token_accuracy": 0.9583325505256652,
      "step": 5945,
      "train/kl_loss_qwen": 0.03940974678844213,
      "train/kl_loss_r1": 0.09058816358447075,
      "train/total_kl": 0.12999791502952576
    },
    {
      "epoch": 1.8793430195830703,
      "grad_norm": 1.6171875,
      "learning_rate": 1.5904927353126973e-05,
      "loss": 0.2727,
      "mean_token_accuracy": 0.9635758399963379,
      "step": 5950,
      "train/kl_loss_qwen": 0.03700983375310898,
      "train/kl_loss_r1": 0.09199956431984901,
      "train/total_kl": 0.12900940030813218
    },
    {
      "epoch": 1.8809222994314592,
      "grad_norm": 3.125,
      "learning_rate": 1.5893082754264055e-05,
      "loss": 0.3381,
      "mean_token_accuracy": 0.948352324962616,
      "step": 5955,
      "train/kl_loss_qwen": 0.0415367916226387,
      "train/kl_loss_r1": 0.09995019435882568,
      "train/total_kl": 0.14148698449134828
    },
    {
      "epoch": 1.8825015792798485,
      "grad_norm": 2.15625,
      "learning_rate": 1.5881238155401137e-05,
      "loss": 0.2691,
      "mean_token_accuracy": 0.9623165965080261,
      "step": 5960,
      "train/kl_loss_qwen": 0.03245711382478476,
      "train/kl_loss_r1": 0.09266491159796715,
      "train/total_kl": 0.12512202560901642
    },
    {
      "epoch": 1.8840808591282374,
      "grad_norm": 3.015625,
      "learning_rate": 1.586939355653822e-05,
      "loss": 0.5382,
      "mean_token_accuracy": 0.9358631134033203,
      "step": 5965,
      "train/kl_loss_qwen": 0.1387790385633707,
      "train/kl_loss_r1": 0.1795283854007721,
      "train/total_kl": 0.31830742210149765
    },
    {
      "epoch": 1.8856601389766268,
      "grad_norm": 4.25,
      "learning_rate": 1.58575489576753e-05,
      "loss": 0.299,
      "mean_token_accuracy": 0.9656675457954407,
      "step": 5970,
      "train/kl_loss_qwen": 0.031575887277722356,
      "train/kl_loss_r1": 0.12694986611604692,
      "train/total_kl": 0.15852575302124022
    },
    {
      "epoch": 1.8872394188250157,
      "grad_norm": 6.90625,
      "learning_rate": 1.5845704358812383e-05,
      "loss": 0.4855,
      "mean_token_accuracy": 0.933137059211731,
      "step": 5975,
      "train/kl_loss_qwen": 0.10317446365952491,
      "train/kl_loss_r1": 0.1543087527155876,
      "train/total_kl": 0.25748321413993835
    },
    {
      "epoch": 1.888818698673405,
      "grad_norm": 2.546875,
      "learning_rate": 1.583385975994946e-05,
      "loss": 0.2766,
      "mean_token_accuracy": 0.9570620894432068,
      "step": 5980,
      "train/kl_loss_qwen": 0.035505347326397894,
      "train/kl_loss_r1": 0.0945535197854042,
      "train/total_kl": 0.13005886673927308
    },
    {
      "epoch": 1.890397978521794,
      "grad_norm": 2.625,
      "learning_rate": 1.5822015161086547e-05,
      "loss": 0.3088,
      "mean_token_accuracy": 0.9682537078857422,
      "step": 5985,
      "train/kl_loss_qwen": 0.04770698398351669,
      "train/kl_loss_r1": 0.13377987295389177,
      "train/total_kl": 0.18148685693740846
    },
    {
      "epoch": 1.8919772583701833,
      "grad_norm": 2.078125,
      "learning_rate": 1.5810170562223626e-05,
      "loss": 0.2869,
      "mean_token_accuracy": 0.9626376748085022,
      "step": 5990,
      "train/kl_loss_qwen": 0.03912109136581421,
      "train/kl_loss_r1": 0.1076616957783699,
      "train/total_kl": 0.14678278863430022
    },
    {
      "epoch": 1.8935565382185722,
      "grad_norm": 2.890625,
      "learning_rate": 1.5798325963360708e-05,
      "loss": 0.2702,
      "mean_token_accuracy": 0.9594083189964294,
      "step": 5995,
      "train/kl_loss_qwen": 0.032271938771009444,
      "train/kl_loss_r1": 0.08128715306520462,
      "train/total_kl": 0.113559091091156
    },
    {
      "epoch": 1.8951358180669615,
      "grad_norm": 2.125,
      "learning_rate": 1.578648136449779e-05,
      "loss": 0.2996,
      "mean_token_accuracy": 0.9607790470123291,
      "step": 6000,
      "train/kl_loss_qwen": 0.042495764791965485,
      "train/kl_loss_r1": 0.10100130885839462,
      "train/total_kl": 0.14349707514047622
    },
    {
      "epoch": 1.8967150979153506,
      "grad_norm": 3.0625,
      "learning_rate": 1.577463676563487e-05,
      "loss": 0.2753,
      "mean_token_accuracy": 0.9600896954536438,
      "step": 6005,
      "train/kl_loss_qwen": 0.04961861409246922,
      "train/kl_loss_r1": 0.07960631549358368,
      "train/total_kl": 0.12922492623329163
    },
    {
      "epoch": 1.8982943777637398,
      "grad_norm": 2.34375,
      "learning_rate": 1.5762792166771953e-05,
      "loss": 0.3025,
      "mean_token_accuracy": 0.9656990885734558,
      "step": 6010,
      "train/kl_loss_qwen": 0.044762665405869484,
      "train/kl_loss_r1": 0.11191230565309525,
      "train/total_kl": 0.15667496919631957
    },
    {
      "epoch": 1.899873657612129,
      "grad_norm": 2.46875,
      "learning_rate": 1.5750947567909035e-05,
      "loss": 0.2631,
      "mean_token_accuracy": 0.9646678805351258,
      "step": 6015,
      "train/kl_loss_qwen": 0.034113308787345885,
      "train/kl_loss_r1": 0.10710275620222091,
      "train/total_kl": 0.14121606349945068
    },
    {
      "epoch": 1.901452937460518,
      "grad_norm": 2.609375,
      "learning_rate": 1.5739102969046114e-05,
      "loss": 0.2412,
      "mean_token_accuracy": 0.9721398949623108,
      "step": 6020,
      "train/kl_loss_qwen": 0.03815052732825279,
      "train/kl_loss_r1": 0.09157755225896835,
      "train/total_kl": 0.12972807884216309
    },
    {
      "epoch": 1.9030322173089071,
      "grad_norm": 2.765625,
      "learning_rate": 1.57272583701832e-05,
      "loss": 0.2873,
      "mean_token_accuracy": 0.959974753856659,
      "step": 6025,
      "train/kl_loss_qwen": 0.04782358035445213,
      "train/kl_loss_r1": 0.09869266152381898,
      "train/total_kl": 0.1465162456035614
    },
    {
      "epoch": 1.9046114971572963,
      "grad_norm": 1.921875,
      "learning_rate": 1.5715413771320278e-05,
      "loss": 0.2533,
      "mean_token_accuracy": 0.9644418120384216,
      "step": 6030,
      "train/kl_loss_qwen": 0.0273905448615551,
      "train/kl_loss_r1": 0.08996914401650428,
      "train/total_kl": 0.11735968887805939
    },
    {
      "epoch": 1.9061907770056854,
      "grad_norm": 2.921875,
      "learning_rate": 1.5703569172457357e-05,
      "loss": 0.2792,
      "mean_token_accuracy": 0.9622648000717163,
      "step": 6035,
      "train/kl_loss_qwen": 0.0353506863117218,
      "train/kl_loss_r1": 0.08712354376912117,
      "train/total_kl": 0.12247423082590103
    },
    {
      "epoch": 1.9077700568540745,
      "grad_norm": 3.265625,
      "learning_rate": 1.5691724573594442e-05,
      "loss": 0.2655,
      "mean_token_accuracy": 0.9589090824127198,
      "step": 6040,
      "train/kl_loss_qwen": 0.03677756674587727,
      "train/kl_loss_r1": 0.08177397102117538,
      "train/total_kl": 0.1185515359044075
    },
    {
      "epoch": 1.9093493367024637,
      "grad_norm": 2.453125,
      "learning_rate": 1.567987997473152e-05,
      "loss": 0.3247,
      "mean_token_accuracy": 0.9487185001373291,
      "step": 6045,
      "train/kl_loss_qwen": 0.04401973783969879,
      "train/kl_loss_r1": 0.09348073750734329,
      "train/total_kl": 0.13750047385692596
    },
    {
      "epoch": 1.9109286165508528,
      "grad_norm": 2.84375,
      "learning_rate": 1.5668035375868606e-05,
      "loss": 0.2769,
      "mean_token_accuracy": 0.9640775322914124,
      "step": 6050,
      "train/kl_loss_qwen": 0.03960266709327698,
      "train/kl_loss_r1": 0.08844204097986222,
      "train/total_kl": 0.1280447080731392
    },
    {
      "epoch": 1.912507896399242,
      "grad_norm": 3.28125,
      "learning_rate": 1.5656190777005684e-05,
      "loss": 0.2954,
      "mean_token_accuracy": 0.9553475499153137,
      "step": 6055,
      "train/kl_loss_qwen": 0.0473956324160099,
      "train/kl_loss_r1": 0.08966436237096786,
      "train/total_kl": 0.1370599940419197
    },
    {
      "epoch": 1.914087176247631,
      "grad_norm": 2.65625,
      "learning_rate": 1.5644346178142766e-05,
      "loss": 0.2562,
      "mean_token_accuracy": 0.9640877485275269,
      "step": 6060,
      "train/kl_loss_qwen": 0.034355645254254344,
      "train/kl_loss_r1": 0.08359325230121613,
      "train/total_kl": 0.11794889867305755
    },
    {
      "epoch": 1.9156664560960204,
      "grad_norm": 1.71875,
      "learning_rate": 1.563250157927985e-05,
      "loss": 0.2414,
      "mean_token_accuracy": 0.9631853938102722,
      "step": 6065,
      "train/kl_loss_qwen": 0.0267809621989727,
      "train/kl_loss_r1": 0.08253797367215157,
      "train/total_kl": 0.10931893736124039
    },
    {
      "epoch": 1.9172457359444093,
      "grad_norm": 2.578125,
      "learning_rate": 1.562065698041693e-05,
      "loss": 0.2816,
      "mean_token_accuracy": 0.9615584969520569,
      "step": 6070,
      "train/kl_loss_qwen": 0.03128730282187462,
      "train/kl_loss_r1": 0.09343428835272789,
      "train/total_kl": 0.12472158670425415
    },
    {
      "epoch": 1.9188250157927986,
      "grad_norm": 4.09375,
      "learning_rate": 1.5608812381554012e-05,
      "loss": 0.514,
      "mean_token_accuracy": 0.9436793446540832,
      "step": 6075,
      "train/kl_loss_qwen": 0.11829701513051986,
      "train/kl_loss_r1": 0.20249445140361785,
      "train/total_kl": 0.3207914650440216
    },
    {
      "epoch": 1.9204042956411875,
      "grad_norm": 1.8125,
      "learning_rate": 1.5596967782691094e-05,
      "loss": 0.3195,
      "mean_token_accuracy": 0.9646710872650146,
      "step": 6080,
      "train/kl_loss_qwen": 0.05857454389333725,
      "train/kl_loss_r1": 0.11488593071699142,
      "train/total_kl": 0.17346047312021257
    },
    {
      "epoch": 1.9219835754895769,
      "grad_norm": 9.1875,
      "learning_rate": 1.5585123183828173e-05,
      "loss": 0.4524,
      "mean_token_accuracy": 0.9463179707527161,
      "step": 6085,
      "train/kl_loss_qwen": 0.10128285996615886,
      "train/kl_loss_r1": 0.16989930719137192,
      "train/total_kl": 0.2711821660399437
    },
    {
      "epoch": 1.9235628553379658,
      "grad_norm": 2.390625,
      "learning_rate": 1.5573278584965258e-05,
      "loss": 0.3068,
      "mean_token_accuracy": 0.9612534761428833,
      "step": 6090,
      "train/kl_loss_qwen": 0.0676696203649044,
      "train/kl_loss_r1": 0.1021253377199173,
      "train/total_kl": 0.16979495882987977
    },
    {
      "epoch": 1.9251421351863551,
      "grad_norm": 2.0625,
      "learning_rate": 1.5561433986102337e-05,
      "loss": 0.2628,
      "mean_token_accuracy": 0.9628806233406066,
      "step": 6095,
      "train/kl_loss_qwen": 0.039988106675446035,
      "train/kl_loss_r1": 0.07726072520017624,
      "train/total_kl": 0.11724883317947388
    },
    {
      "epoch": 1.926721415034744,
      "grad_norm": 2.203125,
      "learning_rate": 1.554958938723942e-05,
      "loss": 0.3002,
      "mean_token_accuracy": 0.9677048206329346,
      "step": 6100,
      "train/kl_loss_qwen": 0.056648113578557965,
      "train/kl_loss_r1": 0.11971210241317749,
      "train/total_kl": 0.17636021822690964
    },
    {
      "epoch": 1.9283006948831334,
      "grad_norm": 2.4375,
      "learning_rate": 1.55377447883765e-05,
      "loss": 0.2501,
      "mean_token_accuracy": 0.964553427696228,
      "step": 6105,
      "train/kl_loss_qwen": 0.03265873827040196,
      "train/kl_loss_r1": 0.07659762054681778,
      "train/total_kl": 0.10925636142492294
    },
    {
      "epoch": 1.9298799747315223,
      "grad_norm": 2.484375,
      "learning_rate": 1.5525900189513583e-05,
      "loss": 0.6134,
      "mean_token_accuracy": 0.9260523557662964,
      "step": 6110,
      "train/kl_loss_qwen": 0.13752721063792706,
      "train/kl_loss_r1": 0.2298063263297081,
      "train/total_kl": 0.3673335433006287
    },
    {
      "epoch": 1.9314592545799116,
      "grad_norm": 4.5,
      "learning_rate": 1.5514055590650665e-05,
      "loss": 0.2711,
      "mean_token_accuracy": 0.9676932573318482,
      "step": 6115,
      "train/kl_loss_qwen": 0.042959189414978026,
      "train/kl_loss_r1": 0.1012831225991249,
      "train/total_kl": 0.14424231201410292
    },
    {
      "epoch": 1.9330385344283005,
      "grad_norm": 4.125,
      "learning_rate": 1.5502210991787747e-05,
      "loss": 0.2602,
      "mean_token_accuracy": 0.9702343583106995,
      "step": 6120,
      "train/kl_loss_qwen": 0.043084985390305516,
      "train/kl_loss_r1": 0.10424283221364021,
      "train/total_kl": 0.14732781946659088
    },
    {
      "epoch": 1.9346178142766899,
      "grad_norm": 14.6875,
      "learning_rate": 1.5490366392924825e-05,
      "loss": 0.5001,
      "mean_token_accuracy": 0.9414145827293396,
      "step": 6125,
      "train/kl_loss_qwen": 0.1335492290556431,
      "train/kl_loss_r1": 0.17076152414083481,
      "train/total_kl": 0.3043107658624649
    },
    {
      "epoch": 1.936197094125079,
      "grad_norm": 2.671875,
      "learning_rate": 1.547852179406191e-05,
      "loss": 0.2972,
      "mean_token_accuracy": 0.9653166651725769,
      "step": 6130,
      "train/kl_loss_qwen": 0.041663892939686774,
      "train/kl_loss_r1": 0.11728576719760894,
      "train/total_kl": 0.1589496612548828
    },
    {
      "epoch": 1.9377763739734681,
      "grad_norm": 1.9921875,
      "learning_rate": 1.546667719519899e-05,
      "loss": 0.2569,
      "mean_token_accuracy": 0.9619585394859314,
      "step": 6135,
      "train/kl_loss_qwen": 0.04035407081246376,
      "train/kl_loss_r1": 0.08105402737855912,
      "train/total_kl": 0.1214080959558487
    },
    {
      "epoch": 1.9393556538218573,
      "grad_norm": 2.609375,
      "learning_rate": 1.545483259633607e-05,
      "loss": 0.4978,
      "mean_token_accuracy": 0.938207495212555,
      "step": 6140,
      "train/kl_loss_qwen": 0.1215125322341919,
      "train/kl_loss_r1": 0.16667343750596048,
      "train/total_kl": 0.28818596303462984
    },
    {
      "epoch": 1.9409349336702464,
      "grad_norm": 1.34375,
      "learning_rate": 1.5442987997473153e-05,
      "loss": 0.2651,
      "mean_token_accuracy": 0.9621343970298767,
      "step": 6145,
      "train/kl_loss_qwen": 0.035513009503483774,
      "train/kl_loss_r1": 0.0880920946598053,
      "train/total_kl": 0.12360510230064392
    },
    {
      "epoch": 1.9425142135186355,
      "grad_norm": 2.484375,
      "learning_rate": 1.5431143398610235e-05,
      "loss": 0.2987,
      "mean_token_accuracy": 0.9618512272834778,
      "step": 6150,
      "train/kl_loss_qwen": 0.03771014921367168,
      "train/kl_loss_r1": 0.1214379459619522,
      "train/total_kl": 0.15914809256792067
    },
    {
      "epoch": 1.9440934933670246,
      "grad_norm": 3.546875,
      "learning_rate": 1.5419298799747317e-05,
      "loss": 0.2714,
      "mean_token_accuracy": 0.9532880187034607,
      "step": 6155,
      "train/kl_loss_qwen": 0.03389046601951122,
      "train/kl_loss_r1": 0.07269448190927505,
      "train/total_kl": 0.1065849483013153
    },
    {
      "epoch": 1.9456727732154138,
      "grad_norm": 8.3125,
      "learning_rate": 1.5407454200884396e-05,
      "loss": 0.3279,
      "mean_token_accuracy": 0.9567504525184631,
      "step": 6160,
      "train/kl_loss_qwen": 0.05254775658249855,
      "train/kl_loss_r1": 0.12171603590250016,
      "train/total_kl": 0.17426379173994064
    },
    {
      "epoch": 1.947252053063803,
      "grad_norm": 2.28125,
      "learning_rate": 1.5395609602021478e-05,
      "loss": 0.3054,
      "mean_token_accuracy": 0.9538545250892639,
      "step": 6165,
      "train/kl_loss_qwen": 0.040179062820971015,
      "train/kl_loss_r1": 0.08459075018763543,
      "train/total_kl": 0.12476981207728385
    },
    {
      "epoch": 1.948831332912192,
      "grad_norm": 2.515625,
      "learning_rate": 1.538376500315856e-05,
      "loss": 0.2713,
      "mean_token_accuracy": 0.9602208375930786,
      "step": 6170,
      "train/kl_loss_qwen": 0.026572006195783614,
      "train/kl_loss_r1": 0.08811861574649811,
      "train/total_kl": 0.11469062268733979
    },
    {
      "epoch": 1.9504106127605811,
      "grad_norm": 2.296875,
      "learning_rate": 1.537192040429564e-05,
      "loss": 0.255,
      "mean_token_accuracy": 0.9642069816589356,
      "step": 6175,
      "train/kl_loss_qwen": 0.027038365602493286,
      "train/kl_loss_r1": 0.09247309863567352,
      "train/total_kl": 0.11951146274805069
    },
    {
      "epoch": 1.9519898926089703,
      "grad_norm": 2.890625,
      "learning_rate": 1.536007580543272e-05,
      "loss": 0.2988,
      "mean_token_accuracy": 0.9508386611938476,
      "step": 6180,
      "train/kl_loss_qwen": 0.0344014722853899,
      "train/kl_loss_r1": 0.09339026212692261,
      "train/total_kl": 0.12779173403978347
    },
    {
      "epoch": 1.9535691724573594,
      "grad_norm": 1.265625,
      "learning_rate": 1.5348231206569806e-05,
      "loss": 0.2556,
      "mean_token_accuracy": 0.9706254363059997,
      "step": 6185,
      "train/kl_loss_qwen": 0.031338495388627055,
      "train/kl_loss_r1": 0.09582123830914498,
      "train/total_kl": 0.12715973258018493
    },
    {
      "epoch": 1.9551484523057487,
      "grad_norm": 2.65625,
      "learning_rate": 1.5336386607706884e-05,
      "loss": 0.2721,
      "mean_token_accuracy": 0.9652405261993409,
      "step": 6190,
      "train/kl_loss_qwen": 0.04058623276650906,
      "train/kl_loss_r1": 0.09163752496242523,
      "train/total_kl": 0.1322237581014633
    },
    {
      "epoch": 1.9567277321541376,
      "grad_norm": 3.25,
      "learning_rate": 1.532454200884397e-05,
      "loss": 0.3585,
      "mean_token_accuracy": 0.9490607500076294,
      "step": 6195,
      "train/kl_loss_qwen": 0.04014108069241047,
      "train/kl_loss_r1": 0.12538652271032333,
      "train/total_kl": 0.16552760303020478
    },
    {
      "epoch": 1.958307012002527,
      "grad_norm": 1.828125,
      "learning_rate": 1.5312697409981048e-05,
      "loss": 0.2993,
      "mean_token_accuracy": 0.9656159996986389,
      "step": 6200,
      "train/kl_loss_qwen": 0.06351602151989937,
      "train/kl_loss_r1": 0.10196839570999146,
      "train/total_kl": 0.165484419465065
    },
    {
      "epoch": 1.959886291850916,
      "grad_norm": 5.09375,
      "learning_rate": 1.530085281111813e-05,
      "loss": 0.3343,
      "mean_token_accuracy": 0.936349618434906,
      "step": 6205,
      "train/kl_loss_qwen": 0.028263447061181068,
      "train/kl_loss_r1": 0.08434647247195244,
      "train/total_kl": 0.11260991990566253
    },
    {
      "epoch": 1.9614655716993052,
      "grad_norm": 2.046875,
      "learning_rate": 1.5289008212255212e-05,
      "loss": 0.254,
      "mean_token_accuracy": 0.9646383285522461,
      "step": 6210,
      "train/kl_loss_qwen": 0.02920605707913637,
      "train/kl_loss_r1": 0.08682290017604828,
      "train/total_kl": 0.11602895706892014
    },
    {
      "epoch": 1.9630448515476941,
      "grad_norm": 2.578125,
      "learning_rate": 1.5277163613392294e-05,
      "loss": 0.2281,
      "mean_token_accuracy": 0.96594717502594,
      "step": 6215,
      "train/kl_loss_qwen": 0.023864947631955145,
      "train/kl_loss_r1": 0.0619074359536171,
      "train/total_kl": 0.08577238395810127
    },
    {
      "epoch": 1.9646241313960835,
      "grad_norm": 2.15625,
      "learning_rate": 1.5265319014529373e-05,
      "loss": 0.3115,
      "mean_token_accuracy": 0.9632365703582764,
      "step": 6220,
      "train/kl_loss_qwen": 0.04294045157730579,
      "train/kl_loss_r1": 0.13019407838582991,
      "train/total_kl": 0.17313452661037446
    },
    {
      "epoch": 1.9662034112444724,
      "grad_norm": 1.8203125,
      "learning_rate": 1.5253474415666458e-05,
      "loss": 0.3089,
      "mean_token_accuracy": 0.9643009543418884,
      "step": 6225,
      "train/kl_loss_qwen": 0.04672396667301655,
      "train/kl_loss_r1": 0.12351526468992233,
      "train/total_kl": 0.17023923397064208
    },
    {
      "epoch": 1.9677826910928617,
      "grad_norm": 1.984375,
      "learning_rate": 1.5241629816803537e-05,
      "loss": 0.3167,
      "mean_token_accuracy": 0.960183835029602,
      "step": 6230,
      "train/kl_loss_qwen": 0.059415098279714584,
      "train/kl_loss_r1": 0.10755098611116409,
      "train/total_kl": 0.1669660821557045
    },
    {
      "epoch": 1.9693619709412507,
      "grad_norm": 6.1875,
      "learning_rate": 1.522978521794062e-05,
      "loss": 0.2871,
      "mean_token_accuracy": 0.9628426909446717,
      "step": 6235,
      "train/kl_loss_qwen": 0.04341960847377777,
      "train/kl_loss_r1": 0.0994553729891777,
      "train/total_kl": 0.1428749829530716
    },
    {
      "epoch": 1.97094125078964,
      "grad_norm": 2.796875,
      "learning_rate": 1.52179406190777e-05,
      "loss": 0.3193,
      "mean_token_accuracy": 0.956408679485321,
      "step": 6240,
      "train/kl_loss_qwen": 0.04886774346232414,
      "train/kl_loss_r1": 0.10023821294307708,
      "train/total_kl": 0.1491059586405754
    },
    {
      "epoch": 1.972520530638029,
      "grad_norm": 2.546875,
      "learning_rate": 1.5206096020214784e-05,
      "loss": 0.2709,
      "mean_token_accuracy": 0.9668349862098694,
      "step": 6245,
      "train/kl_loss_qwen": 0.04663791619241238,
      "train/kl_loss_r1": 0.09476924911141396,
      "train/total_kl": 0.14140716344118118
    },
    {
      "epoch": 1.9740998104864182,
      "grad_norm": 2.1875,
      "learning_rate": 1.5194251421351863e-05,
      "loss": 0.5744,
      "mean_token_accuracy": 0.9282299876213074,
      "step": 6250,
      "train/kl_loss_qwen": 0.14659463837742806,
      "train/kl_loss_r1": 0.18266348093748092,
      "train/total_kl": 0.32925812900066376
    },
    {
      "epoch": 1.9756790903348074,
      "grad_norm": 2.0625,
      "learning_rate": 1.5182406822488946e-05,
      "loss": 0.2669,
      "mean_token_accuracy": 0.9646424412727356,
      "step": 6255,
      "train/kl_loss_qwen": 0.03141367882490158,
      "train/kl_loss_r1": 0.09912259578704834,
      "train/total_kl": 0.13053627461194992
    },
    {
      "epoch": 1.9772583701831965,
      "grad_norm": 3.140625,
      "learning_rate": 1.5170562223626027e-05,
      "loss": 0.3301,
      "mean_token_accuracy": 0.9603954434394837,
      "step": 6260,
      "train/kl_loss_qwen": 0.05611117072403431,
      "train/kl_loss_r1": 0.11906250268220901,
      "train/total_kl": 0.17517367154359817
    },
    {
      "epoch": 1.9788376500315856,
      "grad_norm": 2.484375,
      "learning_rate": 1.515871762476311e-05,
      "loss": 0.2807,
      "mean_token_accuracy": 0.9593365550041199,
      "step": 6265,
      "train/kl_loss_qwen": 0.038422183319926265,
      "train/kl_loss_r1": 0.08924405574798584,
      "train/total_kl": 0.12766623944044114
    },
    {
      "epoch": 1.9804169298799748,
      "grad_norm": 2.953125,
      "learning_rate": 1.5146873025900189e-05,
      "loss": 0.3072,
      "mean_token_accuracy": 0.9570716500282288,
      "step": 6270,
      "train/kl_loss_qwen": 0.043263989686965945,
      "train/kl_loss_r1": 0.09222526848316193,
      "train/total_kl": 0.13548925817012786
    },
    {
      "epoch": 1.9819962097283639,
      "grad_norm": 2.90625,
      "learning_rate": 1.5135028427037273e-05,
      "loss": 0.2615,
      "mean_token_accuracy": 0.959635591506958,
      "step": 6275,
      "train/kl_loss_qwen": 0.03617706634104252,
      "train/kl_loss_r1": 0.08094637170433998,
      "train/total_kl": 0.11712343841791154
    },
    {
      "epoch": 1.983575489576753,
      "grad_norm": 5.25,
      "learning_rate": 1.5123183828174353e-05,
      "loss": 0.3927,
      "mean_token_accuracy": 0.9404747247695923,
      "step": 6280,
      "train/kl_loss_qwen": 0.04918486289680004,
      "train/kl_loss_r1": 0.11403940394520759,
      "train/total_kl": 0.1632242649793625
    },
    {
      "epoch": 1.9851547694251421,
      "grad_norm": 2.265625,
      "learning_rate": 1.5111339229311433e-05,
      "loss": 0.2734,
      "mean_token_accuracy": 0.9585425734519959,
      "step": 6285,
      "train/kl_loss_qwen": 0.03223748728632927,
      "train/kl_loss_r1": 0.08181583285331726,
      "train/total_kl": 0.11405331790447235
    },
    {
      "epoch": 1.9867340492735313,
      "grad_norm": 4.75,
      "learning_rate": 1.5099494630448515e-05,
      "loss": 0.3145,
      "mean_token_accuracy": 0.9465120673179627,
      "step": 6290,
      "train/kl_loss_qwen": 0.031488415971398356,
      "train/kl_loss_r1": 0.07579498887062072,
      "train/total_kl": 0.10728340446949006
    },
    {
      "epoch": 1.9883133291219204,
      "grad_norm": 2.71875,
      "learning_rate": 1.5087650031585595e-05,
      "loss": 0.27,
      "mean_token_accuracy": 0.9674764513969422,
      "step": 6295,
      "train/kl_loss_qwen": 0.039121413230896,
      "train/kl_loss_r1": 0.09238900244235992,
      "train/total_kl": 0.13151041567325591
    },
    {
      "epoch": 1.9898926089703095,
      "grad_norm": 2.296875,
      "learning_rate": 1.5075805432722679e-05,
      "loss": 0.2354,
      "mean_token_accuracy": 0.9715821623802186,
      "step": 6300,
      "train/kl_loss_qwen": 0.03151620179414749,
      "train/kl_loss_r1": 0.08403781279921532,
      "train/total_kl": 0.11555401831865311
    },
    {
      "epoch": 1.9914718888186986,
      "grad_norm": 2.21875,
      "learning_rate": 1.506396083385976e-05,
      "loss": 0.3195,
      "mean_token_accuracy": 0.9480708837509155,
      "step": 6305,
      "train/kl_loss_qwen": 0.045697296783328054,
      "train/kl_loss_r1": 0.08322953581809997,
      "train/total_kl": 0.12892683148384093
    },
    {
      "epoch": 1.9930511686670878,
      "grad_norm": 2.078125,
      "learning_rate": 1.5052116234996841e-05,
      "loss": 0.2772,
      "mean_token_accuracy": 0.9689276695251465,
      "step": 6310,
      "train/kl_loss_qwen": 0.047914443910121916,
      "train/kl_loss_r1": 0.09832399636507035,
      "train/total_kl": 0.14623844027519226
    },
    {
      "epoch": 1.994630448515477,
      "grad_norm": 3.109375,
      "learning_rate": 1.5040271636133922e-05,
      "loss": 0.3595,
      "mean_token_accuracy": 0.9613602519035339,
      "step": 6315,
      "train/kl_loss_qwen": 0.05677867867052555,
      "train/kl_loss_r1": 0.14421724826097487,
      "train/total_kl": 0.20099592953920364
    },
    {
      "epoch": 1.996209728363866,
      "grad_norm": 3.34375,
      "learning_rate": 1.5028427037271005e-05,
      "loss": 0.3156,
      "mean_token_accuracy": 0.9613821506500244,
      "step": 6320,
      "train/kl_loss_qwen": 0.0430843960493803,
      "train/kl_loss_r1": 0.12133632898330689,
      "train/total_kl": 0.16442072987556458
    },
    {
      "epoch": 1.9977890082122554,
      "grad_norm": 2.21875,
      "learning_rate": 1.5016582438408086e-05,
      "loss": 0.2633,
      "mean_token_accuracy": 0.9593652129173279,
      "step": 6325,
      "train/kl_loss_qwen": 0.027921681851148607,
      "train/kl_loss_r1": 0.07064548507332802,
      "train/total_kl": 0.09856716692447662
    },
    {
      "epoch": 1.9993682880606443,
      "grad_norm": 2.734375,
      "learning_rate": 1.5004737839545168e-05,
      "loss": 0.2851,
      "mean_token_accuracy": 0.9604188561439514,
      "step": 6330,
      "train/kl_loss_qwen": 0.02967907078564167,
      "train/kl_loss_r1": 0.0993261307477951,
      "train/total_kl": 0.12900519967079163
    }
  ],
  "logging_steps": 5,
  "max_steps": 12664,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 2.708271409126994e+17,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}