{
  "best_global_step": 562,
  "best_metric": 0.15477769076824188,
  "best_model_checkpoint": "/app/output/bf-router-v0.5/checkpoint-562",
  "epoch": 2.0,
  "eval_steps": 500,
  "global_step": 562,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "entropy": 1.4157426357269287,
      "epoch": 0.003561887800534283,
      "grad_norm": 1.203125,
      "learning_rate": 0.0,
      "loss": 4.336674213409424,
      "mean_token_accuracy": 0.4680640399456024,
      "num_tokens": 3092.0,
      "step": 1
    },
    {
      "entropy": 1.5884604156017303,
      "epoch": 0.03561887800534283,
      "grad_norm": 0.59375,
      "learning_rate": 6.923076923076924e-05,
      "loss": 3.986017862955729,
      "mean_token_accuracy": 0.4735516524977154,
      "num_tokens": 30650.0,
      "step": 10
    },
    {
      "entropy": 2.2315500736236573,
      "epoch": 0.07123775601068566,
      "grad_norm": 0.26953125,
      "learning_rate": 0.00014615384615384615,
      "loss": 2.16544189453125,
      "mean_token_accuracy": 0.6272616922855377,
      "num_tokens": 61735.0,
      "step": 20
    },
    {
      "entropy": 0.8066075079143047,
      "epoch": 0.10685663401602849,
      "grad_norm": 0.298828125,
      "learning_rate": 0.00019999334629824895,
      "loss": 0.7935717582702637,
      "mean_token_accuracy": 0.857068446278572,
      "num_tokens": 92484.0,
      "step": 30
    },
    {
      "entropy": 0.5340980306267739,
      "epoch": 0.14247551202137132,
      "grad_norm": 0.10693359375,
      "learning_rate": 0.00019987508289735716,
      "loss": 0.5261529922485352,
      "mean_token_accuracy": 0.9004581302404404,
      "num_tokens": 123010.0,
      "step": 40
    },
    {
      "entropy": 0.4001643668860197,
      "epoch": 0.17809439002671415,
      "grad_norm": 0.08251953125,
      "learning_rate": 0.0001996091607179287,
      "loss": 0.3956298351287842,
      "mean_token_accuracy": 0.9209299519658088,
      "num_tokens": 153710.0,
      "step": 50
    },
    {
      "entropy": 0.34828933551907537,
      "epoch": 0.21371326803205698,
      "grad_norm": 0.10009765625,
      "learning_rate": 0.00019919597290851538,
      "loss": 0.3648662090301514,
      "mean_token_accuracy": 0.9263768374919892,
      "num_tokens": 184577.0,
      "step": 60
    },
    {
      "entropy": 0.2983129292726517,
      "epoch": 0.2493321460373998,
      "grad_norm": 0.08447265625,
      "learning_rate": 0.00019863613034027224,
      "loss": 0.29642860889434813,
      "mean_token_accuracy": 0.9371836110949516,
      "num_tokens": 215040.0,
      "step": 70
    },
    {
      "entropy": 0.280751570686698,
      "epoch": 0.28495102404274264,
      "grad_norm": 0.099609375,
      "learning_rate": 0.00019793046070382437,
      "loss": 0.2813178300857544,
      "mean_token_accuracy": 0.9388988897204399,
      "num_tokens": 245683.0,
      "step": 80
    },
    {
      "entropy": 0.2960826952010393,
      "epoch": 0.3205699020480855,
      "grad_norm": 0.11962890625,
      "learning_rate": 0.00019708000728558064,
      "loss": 0.31115410327911375,
      "mean_token_accuracy": 0.9335825845599175,
      "num_tokens": 276430.0,
      "step": 90
    },
    {
      "entropy": 0.28416929207742214,
      "epoch": 0.3561887800534283,
      "grad_norm": 0.103515625,
      "learning_rate": 0.00019608602742530283,
      "loss": 0.27938365936279297,
      "mean_token_accuracy": 0.9393619552254677,
      "num_tokens": 307296.0,
      "step": 100
    },
    {
      "entropy": 0.23417662251740695,
      "epoch": 0.39180765805877116,
      "grad_norm": 0.06298828125,
      "learning_rate": 0.00019494999065721108,
      "loss": 0.23734774589538574,
      "mean_token_accuracy": 0.9487677246332169,
      "num_tokens": 337962.0,
      "step": 110
    },
    {
      "entropy": 0.24019648060202597,
      "epoch": 0.42742653606411396,
      "grad_norm": 0.06640625,
      "learning_rate": 0.0001936735765373737,
      "loss": 0.245324969291687,
      "mean_token_accuracy": 0.9457737937569618,
      "num_tokens": 368782.0,
      "step": 120
    },
    {
      "entropy": 0.269666058011353,
      "epoch": 0.4630454140694568,
      "grad_norm": 0.09765625,
      "learning_rate": 0.00019225867216059325,
      "loss": 0.2686375617980957,
      "mean_token_accuracy": 0.9398508563637733,
      "num_tokens": 399329.0,
      "step": 130
    },
    {
      "entropy": 0.18328354582190515,
      "epoch": 0.4986642920747996,
      "grad_norm": 0.06298828125,
      "learning_rate": 0.00019070736937046035,
      "loss": 0.18070420026779174,
      "mean_token_accuracy": 0.9568557634949684,
      "num_tokens": 429759.0,
      "step": 140
    },
    {
      "entropy": 0.21165105439722537,
      "epoch": 0.5342831700801425,
      "grad_norm": 0.0810546875,
      "learning_rate": 0.0001890219616666997,
      "loss": 0.22004930973052977,
      "mean_token_accuracy": 0.9494227185845375,
      "num_tokens": 460409.0,
      "step": 150
    },
    {
      "entropy": 0.23916025627404452,
      "epoch": 0.5699020480854853,
      "grad_norm": 0.07568359375,
      "learning_rate": 0.00018720494081438078,
      "loss": 0.2379377841949463,
      "mean_token_accuracy": 0.9455192387104034,
      "num_tokens": 491344.0,
      "step": 160
    },
    {
      "entropy": 0.2238690486177802,
      "epoch": 0.6055209260908282,
      "grad_norm": 0.08984375,
      "learning_rate": 0.00018525899316000608,
      "loss": 0.2245168685913086,
      "mean_token_accuracy": 0.9485938593745231,
      "num_tokens": 522029.0,
      "step": 170
    },
    {
      "entropy": 0.17032606173306702,
      "epoch": 0.641139804096171,
      "grad_norm": 0.080078125,
      "learning_rate": 0.00018318699565992357,
      "loss": 0.17549347877502441,
      "mean_token_accuracy": 0.9574712902307511,
      "num_tokens": 552306.0,
      "step": 180
    },
    {
      "entropy": 0.21890966054052113,
      "epoch": 0.6767586821015138,
      "grad_norm": 0.0673828125,
      "learning_rate": 0.00018099201162693476,
      "loss": 0.21968293190002441,
      "mean_token_accuracy": 0.9482985377311707,
      "num_tokens": 583181.0,
      "step": 190
    },
    {
      "entropy": 0.20646683853119613,
      "epoch": 0.7123775601068566,
      "grad_norm": 0.09765625,
      "learning_rate": 0.00017867728620138708,
      "loss": 0.20968315601348878,
      "mean_token_accuracy": 0.9519409075379371,
      "num_tokens": 613818.0,
      "step": 200
    },
    {
      "entropy": 0.23000301327556372,
      "epoch": 0.7479964381121995,
      "grad_norm": 0.08935546875,
      "learning_rate": 0.00017624624155344626,
      "loss": 0.2383474349975586,
      "mean_token_accuracy": 0.9458372846245766,
      "num_tokens": 644576.0,
      "step": 210
    },
    {
      "entropy": 0.21420854832977057,
      "epoch": 0.7836153161175423,
      "grad_norm": 0.0966796875,
      "learning_rate": 0.0001737024718236413,
      "loss": 0.21017465591430665,
      "mean_token_accuracy": 0.9511569887399673,
      "num_tokens": 675564.0,
      "step": 220
    },
    {
      "entropy": 0.19280126914381981,
      "epoch": 0.8192341941228851,
      "grad_norm": 0.08154296875,
      "learning_rate": 0.00017104973780916294,
      "loss": 0.1936139941215515,
      "mean_token_accuracy": 0.954449312388897,
      "num_tokens": 706309.0,
      "step": 230
    },
    {
      "entropy": 0.22524890769273043,
      "epoch": 0.8548530721282279,
      "grad_norm": 0.10546875,
      "learning_rate": 0.00016829196140377085,
      "loss": 0.22950620651245118,
      "mean_token_accuracy": 0.9476036429405212,
      "num_tokens": 737135.0,
      "step": 240
    },
    {
      "entropy": 0.18842433094978334,
      "epoch": 0.8904719501335708,
      "grad_norm": 0.087890625,
      "learning_rate": 0.00016543321979953007,
      "loss": 0.19049547910690307,
      "mean_token_accuracy": 0.9543813273310662,
      "num_tokens": 767853.0,
      "step": 250
    },
    {
      "entropy": 0.1991994746029377,
      "epoch": 0.9260908281389136,
      "grad_norm": 0.07763671875,
      "learning_rate": 0.00016247773945894962,
      "loss": 0.19527161121368408,
      "mean_token_accuracy": 0.9535173490643502,
      "num_tokens": 798511.0,
      "step": 260
    },
    {
      "entropy": 0.18196075949817897,
      "epoch": 0.9617097061442564,
      "grad_norm": 0.0771484375,
      "learning_rate": 0.00015942988986643352,
      "loss": 0.19091761112213135,
      "mean_token_accuracy": 0.9535432115197182,
      "num_tokens": 829447.0,
      "step": 270
    },
    {
      "entropy": 0.18307169582694768,
      "epoch": 0.9973285841495992,
      "grad_norm": 0.08837890625,
      "learning_rate": 0.00015629417706828423,
      "loss": 0.1818631410598755,
      "mean_token_accuracy": 0.955574706196785,
      "num_tokens": 860142.0,
      "step": 280
    },
    {
      "epoch": 1.0,
      "eval_entropy": 0.18207345770817276,
      "eval_loss": 0.18149949610233307,
      "eval_mean_token_accuracy": 0.9555717660180221,
      "eval_num_tokens": 862512.0,
      "eval_runtime": 35.3256,
      "eval_samples_per_second": 15.881,
      "eval_steps_per_second": 3.991,
      "step": 281
    },
    {
      "entropy": 0.1545313375118451,
      "epoch": 1.0320569902048085,
      "grad_norm": 0.064453125,
      "learning_rate": 0.00015307523701080768,
      "loss": 0.14675980806350708,
      "mean_token_accuracy": 0.9623546004295349,
      "num_tokens": 889907.0,
      "step": 290
    },
    {
      "entropy": 0.13382284864783286,
      "epoch": 1.0676758682101515,
      "grad_norm": 0.060546875,
      "learning_rate": 0.00014977782868636999,
      "loss": 0.13706474304199218,
      "mean_token_accuracy": 0.9633408606052398,
      "num_tokens": 920639.0,
      "step": 300
    },
    {
      "entropy": 0.16596811451017857,
      "epoch": 1.1032947462154943,
      "grad_norm": 0.0791015625,
      "learning_rate": 0.00014640682709753832,
      "loss": 0.16318607330322266,
      "mean_token_accuracy": 0.9578790530562401,
      "num_tokens": 951323.0,
      "step": 310
    },
    {
      "entropy": 0.1480622159317136,
      "epoch": 1.138913624220837,
      "grad_norm": 0.057861328125,
      "learning_rate": 0.0001429672160497085,
      "loss": 0.1488279938697815,
      "mean_token_accuracy": 0.9609164595603943,
      "num_tokens": 981911.0,
      "step": 320
    },
    {
      "entropy": 0.1655621325597167,
      "epoch": 1.1745325022261799,
      "grad_norm": 0.07275390625,
      "learning_rate": 0.00013946408078287462,
      "loss": 0.16427644491195678,
      "mean_token_accuracy": 0.9576473370194435,
      "num_tokens": 1012535.0,
      "step": 330
    },
    {
      "entropy": 0.18122829273343086,
      "epoch": 1.2101513802315227,
      "grad_norm": 0.06689453125,
      "learning_rate": 0.00013590260045343432,
      "loss": 0.176006543636322,
      "mean_token_accuracy": 0.9551453411579132,
      "num_tokens": 1043503.0,
      "step": 340
    },
    {
      "entropy": 0.15644419118762015,
      "epoch": 1.2457702582368655,
      "grad_norm": 0.06201171875,
      "learning_rate": 0.00013228804047714463,
      "loss": 0.1652477502822876,
      "mean_token_accuracy": 0.9575047269463539,
      "num_tokens": 1074168.0,
      "step": 350
    },
    {
      "entropy": 0.14828295167535543,
      "epoch": 1.2813891362422083,
      "grad_norm": 0.06982421875,
      "learning_rate": 0.00012862574474454928,
      "loss": 0.14066768884658815,
      "mean_token_accuracy": 0.9639188721776009,
      "num_tokens": 1104681.0,
      "step": 360
    },
    {
      "entropy": 0.19175196047872306,
      "epoch": 1.317008014247551,
      "grad_norm": 0.09619140625,
      "learning_rate": 0.0001249211277203859,
      "loss": 0.1917089343070984,
      "mean_token_accuracy": 0.9522068575024605,
      "num_tokens": 1135361.0,
      "step": 370
    },
    {
      "entropy": 0.1644950734451413,
      "epoch": 1.3526268922528941,
      "grad_norm": 0.0849609375,
      "learning_rate": 0.00012117966643865398,
      "loss": 0.16864393949508666,
      "mean_token_accuracy": 0.9572859182953835,
      "num_tokens": 1166125.0,
      "step": 380
    },
    {
      "entropy": 0.16431492734700442,
      "epoch": 1.388245770258237,
      "grad_norm": 0.08056640625,
      "learning_rate": 0.00011740689240517837,
      "loss": 0.1597315788269043,
      "mean_token_accuracy": 0.9589557304978371,
      "num_tokens": 1197232.0,
      "step": 390
    },
    {
      "entropy": 0.15399955678731203,
      "epoch": 1.4238646482635797,
      "grad_norm": 0.07421875,
      "learning_rate": 0.00011360838341963964,
      "loss": 0.15647656917572023,
      "mean_token_accuracy": 0.9591503396630288,
      "num_tokens": 1227801.0,
      "step": 400
    },
    {
      "entropy": 0.15702628958970308,
      "epoch": 1.4594835262689225,
      "grad_norm": 0.0888671875,
      "learning_rate": 0.00010978975532916189,
      "loss": 0.16044070720672607,
      "mean_token_accuracy": 0.958300518989563,
      "num_tokens": 1258629.0,
      "step": 410
    },
    {
      "entropy": 0.12639004811644555,
      "epoch": 1.4951024042742653,
      "grad_norm": 0.0654296875,
      "learning_rate": 0.00010595665372565027,
      "loss": 0.12638626098632813,
      "mean_token_accuracy": 0.9659075498580932,
      "num_tokens": 1289072.0,
      "step": 420
    },
    {
      "entropy": 0.14171069134026765,
      "epoch": 1.5307212822796084,
      "grad_norm": 0.08056640625,
      "learning_rate": 0.00010211474559915233,
      "loss": 0.13931651115417482,
      "mean_token_accuracy": 0.9626615524291993,
      "num_tokens": 1319747.0,
      "step": 430
    },
    {
      "entropy": 0.16094463262706996,
      "epoch": 1.566340160284951,
      "grad_norm": 0.07958984375,
      "learning_rate": 9.826971095958395e-05,
      "loss": 0.15868637561798096,
      "mean_token_accuracy": 0.9590194016695023,
      "num_tokens": 1350664.0,
      "step": 440
    },
    {
      "entropy": 0.15832340456545352,
      "epoch": 1.601959038290294,
      "grad_norm": 0.0673828125,
      "learning_rate": 9.442723443920623e-05,
      "loss": 0.16317789554595946,
      "mean_token_accuracy": 0.9585595563054085,
      "num_tokens": 1381628.0,
      "step": 450
    },
    {
      "entropy": 0.16448966227471828,
      "epoch": 1.6375779162956365,
      "grad_norm": 0.0751953125,
      "learning_rate": 9.059299688826816e-05,
      "loss": 0.16371761560440062,
      "mean_token_accuracy": 0.9589671149849892,
      "num_tokens": 1412506.0,
      "step": 460
    },
    {
      "entropy": 0.15921370945870877,
      "epoch": 1.6731967943009796,
      "grad_norm": 0.0732421875,
      "learning_rate": 8.677266697624138e-05,
      "loss": 0.15553101301193237,
      "mean_token_accuracy": 0.9597976416349411,
      "num_tokens": 1442986.0,
      "step": 470
    },
    {
      "entropy": 0.14739455822855235,
      "epoch": 1.7088156723063224,
      "grad_norm": 0.0751953125,
      "learning_rate": 8.297189281106278e-05,
      "loss": 0.14005433320999144,
      "mean_token_accuracy": 0.9631642028689384,
      "num_tokens": 1473444.0,
      "step": 480
    },
    {
      "entropy": 0.1320339234545827,
      "epoch": 1.7444345503116652,
      "grad_norm": 0.07275390625,
      "learning_rate": 7.919629358877657e-05,
      "loss": 0.13607435226440429,
      "mean_token_accuracy": 0.9643332988023758,
      "num_tokens": 1504098.0,
      "step": 490
    },
    {
      "entropy": 0.12228272054344416,
      "epoch": 1.780053428317008,
      "grad_norm": 0.0888671875,
      "learning_rate": 7.54514512859201e-05,
      "loss": 0.12318435907363892,
      "mean_token_accuracy": 0.9668563097715378,
      "num_tokens": 1534785.0,
      "step": 500
    },
    {
      "entropy": 0.14428242221474646,
      "epoch": 1.8156723063223508,
      "grad_norm": 0.07861328125,
      "learning_rate": 7.174290240693689e-05,
      "loss": 0.14451712369918823,
      "mean_token_accuracy": 0.9615424692630767,
      "num_tokens": 1565500.0,
      "step": 510
    },
    {
      "entropy": 0.1427522897720337,
      "epoch": 1.8512911843276938,
      "grad_norm": 0.056884765625,
      "learning_rate": 6.807612979881661e-05,
      "loss": 0.1388334035873413,
      "mean_token_accuracy": 0.9621888875961304,
      "num_tokens": 1596446.0,
      "step": 520
    },
    {
      "entropy": 0.15157688688486814,
      "epoch": 1.8869100623330364,
      "grad_norm": 0.08349609375,
      "learning_rate": 6.445655454506465e-05,
      "loss": 0.15446548461914061,
      "mean_token_accuracy": 0.9599128782749176,
      "num_tokens": 1627189.0,
      "step": 530
    },
    {
      "entropy": 0.1300442773848772,
      "epoch": 1.9225289403383794,
      "grad_norm": 0.059326171875,
      "learning_rate": 6.0889527950984416e-05,
      "loss": 0.1254338026046753,
      "mean_token_accuracy": 0.9661760002374649,
      "num_tokens": 1657858.0,
      "step": 540
    },
    {
      "entropy": 0.1557828625664115,
      "epoch": 1.9581478183437222,
      "grad_norm": 0.06689453125,
      "learning_rate": 5.738032363212258e-05,
      "loss": 0.16083067655563354,
      "mean_token_accuracy": 0.9595108240842819,
      "num_tokens": 1688776.0,
      "step": 550
    },
    {
      "entropy": 0.13151839561760426,
      "epoch": 1.993766696349065,
      "grad_norm": 0.0810546875,
      "learning_rate": 5.3934129717573165e-05,
      "loss": 0.12607554197311402,
      "mean_token_accuracy": 0.966105441749096,
      "num_tokens": 1719573.0,
      "step": 560
    },
    {
      "epoch": 2.0,
      "eval_entropy": 0.14480842649936676,
      "eval_loss": 0.15477769076824188,
      "eval_mean_token_accuracy": 0.9607759993127052,
      "eval_num_tokens": 1725024.0,
      "eval_runtime": 35.306,
      "eval_samples_per_second": 15.89,
      "eval_steps_per_second": 3.994,
      "step": 562
    }
  ],
  "logging_steps": 10,
  "max_steps": 843,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "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": 4.057603725318144e+16,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}