{
  "report": {
    "held": {
      "over_refusal": 0.02034428794992175,
      "miss_rate": 0.007722007722007722,
      "accuracy": 0.9788813727107738,
      "macro_f1": 0.9781646456301699,
      "n": 6061,
      "per_class": {
        "sexual_content": {
          "acc": 0.9896049896049897,
          "n": 481
        },
        "jailbreak": {
          "acc": 0.9749303621169917,
          "n": 359
        },
        "self_harm": {
          "acc": 0.9883268482490273,
          "n": 514
        },
        "misinformation": {
          "acc": 0.9661971830985916,
          "n": 355
        },
        "nonviolent_crime": {
          "acc": 0.9577464788732394,
          "n": 355
        },
        "safe_sensitive": {
          "acc": 0.9793956043956044,
          "n": 728
        },
        "prompt_injection": {
          "acc": 0.9802631578947368,
          "n": 304
        },
        "hate_harassment": {
          "acc": 0.9896265560165975,
          "n": 482
        },
        "specialized_advice": {
          "acc": 0.9841897233201581,
          "n": 506
        },
        "safe": {
          "acc": 0.9747687132043734,
          "n": 1189
        },
        "privacy_pii": {
          "acc": 0.9889867841409692,
          "n": 454
        },
        "violent_weapons": {
          "acc": 0.9640718562874252,
          "n": 334
        }
      },
      "confusions": {
        "safe->nonviolent_crime": 14,
        "violent_weapons->safe": 8,
        "nonviolent_crime->safe": 7,
        "safe->specialized_advice": 5,
        "self_harm->safe": 4,
        "safe_sensitive->safe": 4,
        "safe->violent_weapons": 4,
        "jailbreak->prompt_injection": 4,
        "misinformation->privacy_pii": 4,
        "misinformation->safe": 3,
        "prompt_injection->jailbreak": 3,
        "jailbreak->hate_harassment": 3,
        "misinformation->specialized_advice": 3,
        "nonviolent_crime->hate_harassment": 3,
        "safe->safe_sensitive": 2,
        "specialized_advice->safe": 2,
        "safe_sensitive->nonviolent_crime": 2,
        "violent_weapons->nonviolent_crime": 2,
        "hate_harassment->safe": 2,
        "self_harm->specialized_advice": 2
      }
    },
    "unseen_phrasing": {
      "over_refusal": 0.0,
      "miss_rate": 0.04952424624555772,
      "accuracy": 0.8821506362489969,
      "macro_f1": 0.15407220696385895,
      "n": 8723,
      "per_class": {
        "prompt_injection": {
          "acc": 0.9823788546255506,
          "n": 3859
        },
        "jailbreak": {
          "acc": 0.8026315789473685,
          "n": 4864
        }
      },
      "confusions": {
        "jailbreak->safe": 396,
        "jailbreak->prompt_injection": 234,
        "jailbreak->hate_harassment": 160,
        "jailbreak->self_harm": 43,
        "jailbreak->sexual_content": 38,
        "prompt_injection->jailbreak": 33,
        "jailbreak->privacy_pii": 30,
        "jailbreak->misinformation": 28,
        "jailbreak->safe_sensitive": 22,
        "prompt_injection->privacy_pii": 19,
        "prompt_injection->safe_sensitive": 8,
        "prompt_injection->safe": 6,
        "jailbreak->violent_weapons": 5,
        "jailbreak->nonviolent_crime": 3,
        "prompt_injection->hate_harassment": 1,
        "prompt_injection->nonviolent_crime": 1,
        "jailbreak->specialized_advice": 1
      }
    },
    "unseen_dialect": {
      "over_refusal": 0.07542579075425791,
      "miss_rate": 0.012905107360285809,
      "accuracy": 0.9397650964512965,
      "macro_f1": 0.9399695311054369,
      "n": 39761,
      "per_class": {
        "safe": {
          "acc": 0.8633740437525165,
          "n": 7451
        },
        "safe_sensitive": {
          "acc": 0.9838081574093052,
          "n": 4879
        },
        "self_harm": {
          "acc": 0.9266173157386547,
          "n": 3107
        },
        "violent_weapons": {
          "acc": 0.9035357417371253,
          "n": 2602
        },
        "nonviolent_crime": {
          "acc": 0.8909435751541015,
          "n": 2109
        },
        "prompt_injection": {
          "acc": 0.9738955823293173,
          "n": 2490
        },
        "sexual_content": {
          "acc": 0.9875,
          "n": 2960
        },
        "hate_harassment": {
          "acc": 0.9551760939167556,
          "n": 2811
        },
        "privacy_pii": {
          "acc": 0.9679393762751385,
          "n": 3431
        },
        "specialized_advice": {
          "acc": 0.9515931372549019,
          "n": 3264
        },
        "misinformation": {
          "acc": 0.9799017418490398,
          "n": 2239
        },
        "jailbreak": {
          "acc": 0.9801488833746899,
          "n": 2418
        }
      },
      "confusions": {
        "safe->violent_weapons": 167,
        "safe->specialized_advice": 164,
        "safe->safe_sensitive": 157,
        "safe->misinformation": 146,
        "safe->nonviolent_crime": 140,
        "safe->sexual_content": 61,
        "safe->privacy_pii": 56,
        "violent_weapons->safe_sensitive": 53,
        "nonviolent_crime->hate_harassment": 49,
        "self_harm->hate_harassment": 47,
        "nonviolent_crime->safe_sensitive": 43,
        "self_harm->violent_weapons": 43,
        "violent_weapons->hate_harassment": 43,
        "safe->self_harm": 41,
        "safe->hate_harassment": 38,
        "prompt_injection->jailbreak": 38,
        "self_harm->jailbreak": 38,
        "violent_weapons->nonviolent_crime": 37,
        "nonviolent_crime->safe": 37,
        "self_harm->safe_sensitive": 37
      }
    },
    "hard": {
      "over_refusal": 0.01510989010989011,
      "miss_rate": 0.0,
      "accuracy": 0.9793956043956044,
      "macro_f1": 0.08246588017580384,
      "n": 728,
      "per_class": {
        "safe_sensitive": {
          "acc": 0.9793956043956044,
          "n": 728
        }
      },
      "confusions": {
        "safe_sensitive->safe": 4,
        "safe_sensitive->nonviolent_crime": 2,
        "safe_sensitive->jailbreak": 2,
        "safe_sensitive->sexual_content": 2,
        "safe_sensitive->self_harm": 2,
        "safe_sensitive->prompt_injection": 1,
        "safe_sensitive->misinformation": 1,
        "safe_sensitive->specialized_advice": 1
      }
    }
  },
  "base_model": "oddadmix/Emhotob-500K-v2",
  "log_history": [
    {
      "loss": 2.1952140808105467,
      "grad_norm": 23.127119064331055,
      "learning_rate": 0.0007462500000000001,
      "epoch": 0.0651890482398957,
      "step": 200
    },
    {
      "loss": 1.0176493835449218,
      "grad_norm": 16.538270950317383,
      "learning_rate": 0.0014962500000000002,
      "epoch": 0.1303780964797914,
      "step": 400
    },
    {
      "loss": 0.5449490737915039,
      "grad_norm": 6.46068000793457,
      "learning_rate": 0.00224625,
      "epoch": 0.19556714471968709,
      "step": 600
    },
    {
      "loss": 0.382957763671875,
      "grad_norm": 4.831028461456299,
      "learning_rate": 0.00299625,
      "epoch": 0.2607561929595828,
      "step": 800
    },
    {
      "loss": 0.29860366821289064,
      "grad_norm": 2.7325191497802734,
      "learning_rate": 0.002999774022447062,
      "epoch": 0.3259452411994785,
      "step": 1000
    },
    {
      "loss": 0.23355463027954101,
      "grad_norm": 3.112755298614502,
      "learning_rate": 0.0029990916107089597,
      "epoch": 0.39113428943937417,
      "step": 1200
    },
    {
      "loss": 0.21465389251708986,
      "grad_norm": 3.836207389831543,
      "learning_rate": 0.002997952966763772,
      "epoch": 0.45632333767926986,
      "step": 1400
    },
    {
      "eval_held_loss": 0.19275572896003723,
      "eval_held_accuracy": 0.9415937964032338,
      "eval_held_macro_f1": 0.9405305791201086,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1943,
      "eval_held_samples_per_second": 31190.672,
      "eval_held_steps_per_second": 247.014,
      "epoch": 0.48891786179921776,
      "step": 1500
    },
    {
      "eval_unseen_phrasing_loss": 0.49662351608276367,
      "eval_unseen_phrasing_accuracy": 0.8563567579961022,
      "eval_unseen_phrasing_macro_f1": 0.15061125684004806,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2097,
      "eval_unseen_phrasing_samples_per_second": 41591.493,
      "eval_unseen_phrasing_steps_per_second": 328.994,
      "epoch": 0.48891786179921776,
      "step": 1500
    },
    {
      "eval_unseen_dialect_loss": 0.3042675852775574,
      "eval_unseen_dialect_accuracy": 0.9102386760896356,
      "eval_unseen_dialect_macro_f1": 0.9054147320120141,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6424,
      "eval_unseen_dialect_samples_per_second": 61897.603,
      "eval_unseen_dialect_steps_per_second": 484.147,
      "epoch": 0.48891786179921776,
      "step": 1500
    },
    {
      "eval_hard_loss": 0.21103380620479584,
      "eval_hard_accuracy": 0.9381868131868132,
      "eval_hard_macro_f1": 0.0806756437514765,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0718,
      "eval_hard_samples_per_second": 10143.386,
      "eval_hard_steps_per_second": 83.599,
      "epoch": 0.48891786179921776,
      "step": 1500
    },
    {
      "loss": 0.1889168357849121,
      "grad_norm": 2.3584070205688477,
      "learning_rate": 0.002996358437145384,
      "epoch": 0.5215123859191656,
      "step": 1600
    },
    {
      "loss": 0.1676872444152832,
      "grad_norm": 1.3204810619354248,
      "learning_rate": 0.0029943085071315197,
      "epoch": 0.5867014341590613,
      "step": 1800
    },
    {
      "loss": 0.15570733070373535,
      "grad_norm": 1.8337198495864868,
      "learning_rate": 0.002991803800596058,
      "epoch": 0.651890482398957,
      "step": 2000
    },
    {
      "loss": 0.1408362579345703,
      "grad_norm": 1.2878652811050415,
      "learning_rate": 0.002988845079819156,
      "epoch": 0.7170795306388527,
      "step": 2200
    },
    {
      "loss": 0.1328596591949463,
      "grad_norm": 2.169991970062256,
      "learning_rate": 0.0029854332452552636,
      "epoch": 0.7822685788787483,
      "step": 2400
    },
    {
      "loss": 0.1266844940185547,
      "grad_norm": 1.003103256225586,
      "learning_rate": 0.0029815693352590756,
      "epoch": 0.847457627118644,
      "step": 2600
    },
    {
      "loss": 0.11402256011962891,
      "grad_norm": 1.5363110303878784,
      "learning_rate": 0.002977254525769523,
      "epoch": 0.9126466753585397,
      "step": 2800
    },
    {
      "loss": 0.11763209342956543,
      "grad_norm": 1.188177466392517,
      "learning_rate": 0.002972490129951887,
      "epoch": 0.9778357235984355,
      "step": 3000
    },
    {
      "eval_held_loss": 0.10533597320318222,
      "eval_held_accuracy": 0.9670021448605841,
      "eval_held_macro_f1": 0.9660969193353056,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1899,
      "eval_held_samples_per_second": 31911.042,
      "eval_held_steps_per_second": 252.719,
      "epoch": 0.9778357235984355,
      "step": 3000
    },
    {
      "eval_unseen_phrasing_loss": 0.44238102436065674,
      "eval_unseen_phrasing_accuracy": 0.8843287859681302,
      "eval_unseen_phrasing_macro_f1": 0.15345430620363873,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.21,
      "eval_unseen_phrasing_samples_per_second": 41530.638,
      "eval_unseen_phrasing_steps_per_second": 328.512,
      "epoch": 0.9778357235984355,
      "step": 3000
    },
    {
      "eval_unseen_dialect_loss": 0.2203148454427719,
      "eval_unseen_dialect_accuracy": 0.9360680063378688,
      "eval_unseen_dialect_macro_f1": 0.9341579999900128,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6576,
      "eval_unseen_dialect_samples_per_second": 60465.568,
      "eval_unseen_dialect_steps_per_second": 472.946,
      "epoch": 0.9778357235984355,
      "step": 3000
    },
    {
      "eval_hard_loss": 0.12471413612365723,
      "eval_hard_accuracy": 0.9601648351648352,
      "eval_hard_macro_f1": 0.08163980378416258,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0974,
      "eval_hard_samples_per_second": 7471.794,
      "eval_hard_steps_per_second": 61.581,
      "epoch": 0.9778357235984355,
      "step": 3000
    },
    {
      "loss": 0.08490500450134278,
      "grad_norm": 0.037526536732912064,
      "learning_rate": 0.002967277597798152,
      "epoch": 1.0430247718383312,
      "step": 3200
    },
    {
      "loss": 0.06376803398132325,
      "grad_norm": 0.25600579380989075,
      "learning_rate": 0.002961618515685716,
      "epoch": 1.1082138200782268,
      "step": 3400
    },
    {
      "loss": 0.06879507064819336,
      "grad_norm": 0.3666667640209198,
      "learning_rate": 0.0029555146058945943,
      "epoch": 1.1734028683181226,
      "step": 3600
    },
    {
      "loss": 0.06678017139434815,
      "grad_norm": 1.0908209085464478,
      "learning_rate": 0.0029489677260832628,
      "epoch": 1.2385919165580184,
      "step": 3800
    },
    {
      "loss": 0.058169617652893066,
      "grad_norm": 1.3792606592178345,
      "learning_rate": 0.0029419798687232977,
      "epoch": 1.303780964797914,
      "step": 4000
    },
    {
      "loss": 0.05968661308288574,
      "grad_norm": 1.0037418603897095,
      "learning_rate": 0.0029345531604929908,
      "epoch": 1.3689700130378095,
      "step": 4200
    },
    {
      "loss": 0.0638724946975708,
      "grad_norm": 0.8174710273742676,
      "learning_rate": 0.002926689861630119,
      "epoch": 1.4341590612777053,
      "step": 4400
    },
    {
      "eval_held_loss": 0.10515185445547104,
      "eval_held_accuracy": 0.9701369410988285,
      "eval_held_macro_f1": 0.9692092843051797,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1926,
      "eval_held_samples_per_second": 31466.21,
      "eval_held_steps_per_second": 249.196,
      "epoch": 1.4667535853976532,
      "step": 4500
    },
    {
      "eval_unseen_phrasing_loss": 0.42798590660095215,
      "eval_unseen_phrasing_accuracy": 0.8852459016393442,
      "eval_unseen_phrasing_macro_f1": 0.15392992589063928,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.211,
      "eval_unseen_phrasing_samples_per_second": 41348.901,
      "eval_unseen_phrasing_steps_per_second": 327.075,
      "epoch": 1.4667535853976532,
      "step": 4500
    },
    {
      "eval_unseen_dialect_loss": 0.27985629439353943,
      "eval_unseen_dialect_accuracy": 0.9270642086466638,
      "eval_unseen_dialect_macro_f1": 0.9260390752552076,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6652,
      "eval_unseen_dialect_samples_per_second": 59776.6,
      "eval_unseen_dialect_steps_per_second": 467.557,
      "epoch": 1.4667535853976532,
      "step": 4500
    },
    {
      "eval_hard_loss": 0.08975941687822342,
      "eval_hard_accuracy": 0.9739010989010989,
      "eval_hard_macro_f1": 0.08223150081187659,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0962,
      "eval_hard_samples_per_second": 7568.375,
      "eval_hard_steps_per_second": 62.377,
      "epoch": 1.4667535853976532,
      "step": 4500
    },
    {
      "loss": 0.0591568660736084,
      "grad_norm": 1.0088281631469727,
      "learning_rate": 0.002918392365244061,
      "epoch": 1.4993481095176011,
      "step": 4600
    },
    {
      "loss": 0.06349584102630615,
      "grad_norm": 0.24618516862392426,
      "learning_rate": 0.0029096631965874853,
      "epoch": 1.5645371577574967,
      "step": 4800
    },
    {
      "loss": 0.06407528400421142,
      "grad_norm": 0.9340533018112183,
      "learning_rate": 0.0029005050122878147,
      "epoch": 1.6297262059973925,
      "step": 5000
    },
    {
      "loss": 0.061186332702636716,
      "grad_norm": 0.9107056856155396,
      "learning_rate": 0.0028909205995387103,
      "epoch": 1.694915254237288,
      "step": 5200
    },
    {
      "loss": 0.054206957817077635,
      "grad_norm": 0.5927267074584961,
      "learning_rate": 0.002880912875251822,
      "epoch": 1.7601043024771839,
      "step": 5400
    },
    {
      "loss": 0.06110722541809082,
      "grad_norm": 0.5050310492515564,
      "learning_rate": 0.0028704848851690556,
      "epoch": 1.8252933507170797,
      "step": 5600
    },
    {
      "loss": 0.06215485572814941,
      "grad_norm": 0.9473423957824707,
      "learning_rate": 0.002859639802935635,
      "epoch": 1.8904823989569752,
      "step": 5800
    },
    {
      "loss": 0.05840580940246582,
      "grad_norm": 0.8732274174690247,
      "learning_rate": 0.002848380929134238,
      "epoch": 1.9556714471968708,
      "step": 6000
    },
    {
      "eval_held_loss": 0.091562420129776,
      "eval_held_accuracy": 0.9739316944398614,
      "eval_held_macro_f1": 0.9733442004204589,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1923,
      "eval_held_samples_per_second": 31513.837,
      "eval_held_steps_per_second": 249.573,
      "epoch": 1.9556714471968708,
      "step": 6000
    },
    {
      "eval_unseen_phrasing_loss": 0.6874486207962036,
      "eval_unseen_phrasing_accuracy": 0.8426000229278918,
      "eval_unseen_phrasing_macro_f1": 0.15042589845771517,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2158,
      "eval_unseen_phrasing_samples_per_second": 40423.643,
      "eval_unseen_phrasing_steps_per_second": 319.756,
      "epoch": 1.9556714471968708,
      "step": 6000
    },
    {
      "eval_unseen_dialect_loss": 0.2561938166618347,
      "eval_unseen_dialect_accuracy": 0.9321445637685168,
      "eval_unseen_dialect_macro_f1": 0.934623923757742,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6412,
      "eval_unseen_dialect_samples_per_second": 62005.906,
      "eval_unseen_dialect_steps_per_second": 484.994,
      "epoch": 1.9556714471968708,
      "step": 6000
    },
    {
      "eval_hard_loss": 0.050711967051029205,
      "eval_hard_accuracy": 0.9862637362637363,
      "eval_hard_macro_f1": 0.08275703088981097,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.1008,
      "eval_hard_samples_per_second": 7222.732,
      "eval_hard_steps_per_second": 59.528,
      "epoch": 1.9556714471968708,
      "step": 6000
    },
    {
      "loss": 0.04839906215667725,
      "grad_norm": 0.2468060553073883,
      "learning_rate": 0.002836711690280501,
      "epoch": 2.0208604954367666,
      "step": 6200
    },
    {
      "loss": 0.01925981044769287,
      "grad_norm": 0.06194070354104042,
      "learning_rate": 0.0028246356377801944,
      "epoch": 2.0860495436766624,
      "step": 6400
    },
    {
      "loss": 0.01897188901901245,
      "grad_norm": 1.0675151348114014,
      "learning_rate": 0.002812156446848393,
      "epoch": 2.151238591916558,
      "step": 6600
    },
    {
      "loss": 0.02150702476501465,
      "grad_norm": 0.3522121012210846,
      "learning_rate": 0.002799277915390961,
      "epoch": 2.2164276401564535,
      "step": 6800
    },
    {
      "loss": 0.02550948619842529,
      "grad_norm": 0.3296761214733124,
      "learning_rate": 0.002786003962848706,
      "epoch": 2.2816166883963493,
      "step": 7000
    },
    {
      "loss": 0.021695730686187745,
      "grad_norm": 1.0783436298370361,
      "learning_rate": 0.0027723386290045347,
      "epoch": 2.346805736636245,
      "step": 7200
    },
    {
      "loss": 0.02679917573928833,
      "grad_norm": 1.2645609378814697,
      "learning_rate": 0.0027582860727539908,
      "epoch": 2.411994784876141,
      "step": 7400
    },
    {
      "eval_held_loss": 0.1025511845946312,
      "eval_held_accuracy": 0.9734367266127701,
      "eval_held_macro_f1": 0.9722225212274557,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.2,
      "eval_held_samples_per_second": 30311.136,
      "eval_held_steps_per_second": 240.049,
      "epoch": 2.444589308996089,
      "step": 7500
    },
    {
      "eval_unseen_phrasing_loss": 0.6239086985588074,
      "eval_unseen_phrasing_accuracy": 0.8631204860713058,
      "eval_unseen_phrasing_macro_f1": 0.15031679209778467,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2158,
      "eval_unseen_phrasing_samples_per_second": 40428.512,
      "eval_unseen_phrasing_steps_per_second": 319.794,
      "epoch": 2.444589308996089,
      "step": 7500
    },
    {
      "eval_unseen_dialect_loss": 0.2581072449684143,
      "eval_unseen_dialect_accuracy": 0.9421292221020598,
      "eval_unseen_dialect_macro_f1": 0.9400861215677966,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6554,
      "eval_unseen_dialect_samples_per_second": 60662.396,
      "eval_unseen_dialect_steps_per_second": 474.485,
      "epoch": 2.444589308996089,
      "step": 7500
    },
    {
      "eval_hard_loss": 0.10843586176633835,
      "eval_hard_accuracy": 0.9725274725274725,
      "eval_hard_macro_f1": 0.08217270194986072,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.068,
      "eval_hard_samples_per_second": 10710.827,
      "eval_hard_steps_per_second": 88.276,
      "epoch": 2.444589308996089,
      "step": 7500
    },
    {
      "loss": 0.024594943523406982,
      "grad_norm": 0.06156105548143387,
      "learning_rate": 0.002743850570839542,
      "epoch": 2.4771838331160367,
      "step": 7600
    },
    {
      "loss": 0.027882373332977294,
      "grad_norm": 1.0311678647994995,
      "learning_rate": 0.0027290365165489965,
      "epoch": 2.542372881355932,
      "step": 7800
    },
    {
      "loss": 0.025555319786071777,
      "grad_norm": 0.1658269762992859,
      "learning_rate": 0.002713848418378459,
      "epoch": 2.607561929595828,
      "step": 8000
    },
    {
      "loss": 0.02656315565109253,
      "grad_norm": 0.03239389881491661,
      "learning_rate": 0.0026982908986602134,
      "epoch": 2.6727509778357237,
      "step": 8200
    },
    {
      "loss": 0.030051681995391846,
      "grad_norm": 0.17788361012935638,
      "learning_rate": 0.0026823686921559693,
      "epoch": 2.737940026075619,
      "step": 8400
    },
    {
      "loss": 0.029807121753692628,
      "grad_norm": 0.15262019634246826,
      "learning_rate": 0.0026660866446158882,
      "epoch": 2.803129074315515,
      "step": 8600
    },
    {
      "loss": 0.02619988441467285,
      "grad_norm": 1.711606740951538,
      "learning_rate": 0.002649449711303834,
      "epoch": 2.8683181225554106,
      "step": 8800
    },
    {
      "loss": 0.031050331592559814,
      "grad_norm": 1.0433506965637207,
      "learning_rate": 0.0026324629554892874,
      "epoch": 2.9335071707953064,
      "step": 9000
    },
    {
      "eval_held_loss": 0.10007420927286148,
      "eval_held_accuracy": 0.9727767695099818,
      "eval_held_macro_f1": 0.9723198672543146,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.2012,
      "eval_held_samples_per_second": 30124.931,
      "eval_held_steps_per_second": 238.574,
      "epoch": 2.9335071707953064,
      "step": 9000
    },
    {
      "eval_unseen_phrasing_loss": 0.44695401191711426,
      "eval_unseen_phrasing_accuracy": 0.8987733577897512,
      "eval_unseen_phrasing_macro_f1": 0.1555783687064858,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2123,
      "eval_unseen_phrasing_samples_per_second": 41093.609,
      "eval_unseen_phrasing_steps_per_second": 325.055,
      "epoch": 2.9335071707953064,
      "step": 9000
    },
    {
      "eval_unseen_dialect_loss": 0.3041954040527344,
      "eval_unseen_dialect_accuracy": 0.9305097960312869,
      "eval_unseen_dialect_macro_f1": 0.9275125781125492,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6423,
      "eval_unseen_dialect_samples_per_second": 61901.118,
      "eval_unseen_dialect_steps_per_second": 484.174,
      "epoch": 2.9335071707953064,
      "step": 9000
    },
    {
      "eval_hard_loss": 0.12261448055505753,
      "eval_hard_accuracy": 0.967032967032967,
      "eval_hard_macro_f1": 0.08193668528864059,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0944,
      "eval_hard_samples_per_second": 7715.124,
      "eval_hard_steps_per_second": 63.586,
      "epoch": 2.9335071707953064,
      "step": 9000
    },
    {
      "loss": 0.025427730083465577,
      "grad_norm": 0.16791117191314697,
      "learning_rate": 0.002615131546906401,
      "epoch": 2.9986962190352022,
      "step": 9200
    },
    {
      "loss": 0.010956484079360961,
      "grad_norm": 0.14401958882808685,
      "learning_rate": 0.002597460760180644,
      "epoch": 3.0638852672750976,
      "step": 9400
    },
    {
      "loss": 0.009910619258880616,
      "grad_norm": 0.09711571037769318,
      "learning_rate": 0.0025794559732235265,
      "epoch": 3.1290743155149934,
      "step": 9600
    },
    {
      "loss": 0.011872271299362183,
      "grad_norm": 0.027869893237948418,
      "learning_rate": 0.0025611226655958925,
      "epoch": 3.194263363754889,
      "step": 9800
    },
    {
      "loss": 0.01215682625770569,
      "grad_norm": 0.0019678683020174503,
      "learning_rate": 0.0025424664168402747,
      "epoch": 3.259452411994785,
      "step": 10000
    },
    {
      "loss": 0.013592723608016968,
      "grad_norm": 0.11185621470212936,
      "learning_rate": 0.0025234929047828207,
      "epoch": 3.3246414602346808,
      "step": 10200
    },
    {
      "loss": 0.011327080726623535,
      "grad_norm": 0.15672415494918823,
      "learning_rate": 0.00250420790380531,
      "epoch": 3.389830508474576,
      "step": 10400
    },
    {
      "eval_held_loss": 0.11069366335868835,
      "eval_held_accuracy": 0.9732717373370731,
      "eval_held_macro_f1": 0.9725045194495406,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.191,
      "eval_held_samples_per_second": 31740.515,
      "eval_held_steps_per_second": 251.369,
      "epoch": 3.422425032594524,
      "step": 10500
    },
    {
      "eval_unseen_phrasing_loss": 0.8103329539299011,
      "eval_unseen_phrasing_accuracy": 0.8514272612633268,
      "eval_unseen_phrasing_macro_f1": 0.15040481480414067,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2182,
      "eval_unseen_phrasing_samples_per_second": 39978.707,
      "eval_unseen_phrasing_steps_per_second": 316.236,
      "epoch": 3.422425032594524,
      "step": 10500
    },
    {
      "eval_unseen_dialect_loss": 0.30133235454559326,
      "eval_unseen_dialect_accuracy": 0.9396644953597747,
      "eval_unseen_dialect_macro_f1": 0.9393537114525294,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6388,
      "eval_unseen_dialect_samples_per_second": 62243.44,
      "eval_unseen_dialect_steps_per_second": 486.852,
      "epoch": 3.422425032594524,
      "step": 10500
    },
    {
      "eval_hard_loss": 0.1576533317565918,
      "eval_hard_accuracy": 0.9656593406593407,
      "eval_hard_macro_f1": 0.08187747495923596,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0904,
      "eval_hard_samples_per_second": 8050.572,
      "eval_hard_steps_per_second": 66.351,
      "epoch": 3.422425032594524,
      "step": 10500
    },
    {
      "loss": 0.011172587871551514,
      "grad_norm": 0.6160985827445984,
      "learning_rate": 0.002484617283087783,
      "epoch": 3.455019556714472,
      "step": 10600
    },
    {
      "loss": 0.018457911014556884,
      "grad_norm": 0.5318465828895569,
      "learning_rate": 0.002464727004822319,
      "epoch": 3.5202086049543677,
      "step": 10800
    },
    {
      "loss": 0.015429561138153075,
      "grad_norm": 1.0316749811172485,
      "learning_rate": 0.0024445431223985093,
      "epoch": 3.5853976531942635,
      "step": 11000
    },
    {
      "loss": 0.013867697715759277,
      "grad_norm": 1.1682846546173096,
      "learning_rate": 0.0024240717785611737,
      "epoch": 3.6505867014341593,
      "step": 11200
    },
    {
      "loss": 0.017523043155670166,
      "grad_norm": 0.08472991734743118,
      "learning_rate": 0.0024033192035408817,
      "epoch": 3.7157757496740547,
      "step": 11400
    },
    {
      "loss": 0.010207840204238892,
      "grad_norm": 0.1447364240884781,
      "learning_rate": 0.002382291713157853,
      "epoch": 3.7809647979139505,
      "step": 11600
    },
    {
      "loss": 0.015066323280334472,
      "grad_norm": 0.08443405479192734,
      "learning_rate": 0.0023609957068998024,
      "epoch": 3.8461538461538463,
      "step": 11800
    },
    {
      "loss": 0.012173240184783935,
      "grad_norm": 0.43619173765182495,
      "learning_rate": 0.0023394376659743275,
      "epoch": 3.9113428943937416,
      "step": 12000
    },
    {
      "eval_held_loss": 0.0987815111875534,
      "eval_held_accuracy": 0.9773964692295001,
      "eval_held_macro_f1": 0.97685223860291,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.19,
      "eval_held_samples_per_second": 31893.707,
      "eval_held_steps_per_second": 252.582,
      "epoch": 3.9113428943937416,
      "step": 12000
    },
    {
      "eval_unseen_phrasing_loss": 0.9891596436500549,
      "eval_unseen_phrasing_accuracy": 0.8440903358936146,
      "eval_unseen_phrasing_macro_f1": 0.14909741620054304,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2129,
      "eval_unseen_phrasing_samples_per_second": 40969.504,
      "eval_unseen_phrasing_steps_per_second": 324.074,
      "epoch": 3.9113428943937416,
      "step": 12000
    },
    {
      "eval_unseen_dialect_loss": 0.3277216851711273,
      "eval_unseen_dialect_accuracy": 0.9389854379920022,
      "eval_unseen_dialect_macro_f1": 0.9395415018723355,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.647,
      "eval_unseen_dialect_samples_per_second": 61452.745,
      "eval_unseen_dialect_steps_per_second": 480.667,
      "epoch": 3.9113428943937416,
      "step": 12000
    },
    {
      "eval_hard_loss": 0.07751184701919556,
      "eval_hard_accuracy": 0.9835164835164835,
      "eval_hard_macro_f1": 0.08264081255771007,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0699,
      "eval_hard_samples_per_second": 10416.259,
      "eval_hard_steps_per_second": 85.848,
      "epoch": 3.9113428943937416,
      "step": 12000
    },
    {
      "loss": 0.015515915155410766,
      "grad_norm": 1.3443394899368286,
      "learning_rate": 0.002317624151336424,
      "epoch": 3.9765319426336374,
      "step": 12200
    },
    {
      "loss": 0.009629806876182556,
      "grad_norm": 0.017426181584596634,
      "learning_rate": 0.0022955618016917297,
      "epoch": 4.041720990873533,
      "step": 12400
    },
    {
      "loss": 0.004908042252063751,
      "grad_norm": 0.004287226591259241,
      "learning_rate": 0.0022732573314761073,
      "epoch": 4.106910039113429,
      "step": 12600
    },
    {
      "loss": 0.002916430830955505,
      "grad_norm": 0.0017758934991434216,
      "learning_rate": 0.002250717528812178,
      "epoch": 4.172099087353325,
      "step": 12800
    },
    {
      "loss": 0.005083450078964233,
      "grad_norm": 0.03615580126643181,
      "learning_rate": 0.00222794925344343,
      "epoch": 4.237288135593221,
      "step": 13000
    },
    {
      "loss": 0.005427567362785339,
      "grad_norm": 0.9530061483383179,
      "learning_rate": 0.002204959434646535,
      "epoch": 4.302477183833116,
      "step": 13200
    },
    {
      "loss": 0.006234148740768433,
      "grad_norm": 0.0016657921951264143,
      "learning_rate": 0.00218175506912249,
      "epoch": 4.367666232073011,
      "step": 13400
    },
    {
      "eval_held_loss": 0.11707107722759247,
      "eval_held_accuracy": 0.9755815871968322,
      "eval_held_macro_f1": 0.9746913263378923,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1943,
      "eval_held_samples_per_second": 31193.504,
      "eval_held_steps_per_second": 247.036,
      "epoch": 4.400260756192959,
      "step": 13500
    },
    {
      "eval_unseen_phrasing_loss": 0.7506057620048523,
      "eval_unseen_phrasing_accuracy": 0.8761893843861057,
      "eval_unseen_phrasing_macro_f1": 0.15342070919463482,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2103,
      "eval_unseen_phrasing_samples_per_second": 41488.488,
      "eval_unseen_phrasing_steps_per_second": 328.179,
      "epoch": 4.400260756192959,
      "step": 13500
    },
    {
      "eval_unseen_dialect_loss": 0.36537283658981323,
      "eval_unseen_dialect_accuracy": 0.9392369407208068,
      "eval_unseen_dialect_macro_f1": 0.9387424661370556,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6474,
      "eval_unseen_dialect_samples_per_second": 61416.851,
      "eval_unseen_dialect_steps_per_second": 480.386,
      "epoch": 4.400260756192959,
      "step": 13500
    },
    {
      "eval_hard_loss": 0.15271884202957153,
      "eval_hard_accuracy": 0.9752747252747253,
      "eval_hard_macro_f1": 0.08229021789522485,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0706,
      "eval_hard_samples_per_second": 10312.272,
      "eval_hard_steps_per_second": 84.991,
      "epoch": 4.400260756192959,
      "step": 13500
    },
    {
      "loss": 0.00779943585395813,
      "grad_norm": 1.3296419382095337,
      "learning_rate": 0.0021583432188672547,
      "epoch": 4.432855280312907,
      "step": 13600
    },
    {
      "loss": 0.005994245409965515,
      "grad_norm": 0.00252769747748971,
      "learning_rate": 0.0021347310090225067,
      "epoch": 4.498044328552803,
      "step": 13800
    },
    {
      "loss": 0.0077133625745773315,
      "grad_norm": 0.03387093544006348,
      "learning_rate": 0.0021109256257071823,
      "epoch": 4.563233376792699,
      "step": 14000
    },
    {
      "loss": 0.009878547191619872,
      "grad_norm": 0.07391855865716934,
      "learning_rate": 0.0020869343138304607,
      "epoch": 4.6284224250325945,
      "step": 14200
    },
    {
      "loss": 0.008405025601387024,
      "grad_norm": 0.42607390880584717,
      "learning_rate": 0.0020627643748868545,
      "epoch": 4.69361147327249,
      "step": 14400
    },
    {
      "loss": 0.00548877239227295,
      "grad_norm": 0.015178793109953403,
      "learning_rate": 0.002038423164734077,
      "epoch": 4.758800521512386,
      "step": 14600
    },
    {
      "loss": 0.006401987671852112,
      "grad_norm": 0.005941878538578749,
      "learning_rate": 0.0020139180913543706,
      "epoch": 4.823989569752282,
      "step": 14800
    },
    {
      "loss": 0.007755057215690613,
      "grad_norm": 0.9141596555709839,
      "learning_rate": 0.001989256612599966,
      "epoch": 4.889178617992178,
      "step": 15000
    },
    {
      "eval_held_loss": 0.1293095350265503,
      "eval_held_accuracy": 0.9739316944398614,
      "eval_held_macro_f1": 0.9732220534325807,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1905,
      "eval_held_samples_per_second": 31816.548,
      "eval_held_steps_per_second": 251.971,
      "epoch": 4.889178617992178,
      "step": 15000
    },
    {
      "eval_unseen_phrasing_loss": 0.8130305409431458,
      "eval_unseen_phrasing_accuracy": 0.8730941189957584,
      "eval_unseen_phrasing_macro_f1": 0.15290569249338104,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2142,
      "eval_unseen_phrasing_samples_per_second": 40726.747,
      "eval_unseen_phrasing_steps_per_second": 322.154,
      "epoch": 4.889178617992178,
      "step": 15000
    },
    {
      "eval_unseen_dialect_loss": 0.3997732996940613,
      "eval_unseen_dialect_accuracy": 0.935640451698901,
      "eval_unseen_dialect_macro_f1": 0.9361924662852501,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6425,
      "eval_unseen_dialect_samples_per_second": 61888.415,
      "eval_unseen_dialect_steps_per_second": 484.075,
      "epoch": 4.889178617992178,
      "step": 15000
    },
    {
      "eval_hard_loss": 0.16973277926445007,
      "eval_hard_accuracy": 0.967032967032967,
      "eval_hard_macro_f1": 0.08193668528864059,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0991,
      "eval_hard_samples_per_second": 7346.778,
      "eval_hard_steps_per_second": 60.55,
      "epoch": 4.889178617992178,
      "step": 15000
    },
    {
      "loss": 0.007403180599212646,
      "grad_norm": 0.004417558200657368,
      "learning_rate": 0.0019644462339233667,
      "epoch": 4.9543676662320735,
      "step": 15200
    },
    {
      "loss": 0.005648733377456665,
      "grad_norm": 0.003013500478118658,
      "learning_rate": 0.0019394945060931454,
      "epoch": 5.019556714471968,
      "step": 15400
    },
    {
      "loss": 0.0020281124114990235,
      "grad_norm": 0.03553903475403786,
      "learning_rate": 0.0019144090228959505,
      "epoch": 5.084745762711864,
      "step": 15600
    },
    {
      "loss": 0.0023620997369289396,
      "grad_norm": 0.05907260999083519,
      "learning_rate": 0.0018891974188254203,
      "epoch": 5.14993481095176,
      "step": 15800
    },
    {
      "loss": 0.0015316706895828246,
      "grad_norm": 0.00022171845193952322,
      "learning_rate": 0.0018638673667587078,
      "epoch": 5.215123859191656,
      "step": 16000
    },
    {
      "loss": 0.0017865265905857087,
      "grad_norm": 0.00469543831422925,
      "learning_rate": 0.0018384265756213227,
      "epoch": 5.280312907431552,
      "step": 16200
    },
    {
      "loss": 0.0016463798284530639,
      "grad_norm": 0.0022215796634554863,
      "learning_rate": 0.001812882788041005,
      "epoch": 5.345501955671447,
      "step": 16400
    },
    {
      "eval_held_loss": 0.1353742480278015,
      "eval_held_accuracy": 0.9749216300940439,
      "eval_held_macro_f1": 0.9735729685113267,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1983,
      "eval_held_samples_per_second": 30571.839,
      "eval_held_steps_per_second": 242.113,
      "epoch": 5.378096479791395,
      "step": 16500
    },
    {
      "eval_unseen_phrasing_loss": 0.8409979939460754,
      "eval_unseen_phrasing_accuracy": 0.8759601054683022,
      "eval_unseen_phrasing_macro_f1": 0.1535842422700813,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2107,
      "eval_unseen_phrasing_samples_per_second": 41391.798,
      "eval_unseen_phrasing_steps_per_second": 327.414,
      "epoch": 5.378096479791395,
      "step": 16500
    },
    {
      "eval_unseen_dialect_loss": 0.5059581398963928,
      "eval_unseen_dialect_accuracy": 0.930786449032972,
      "eval_unseen_dialect_macro_f1": 0.9325400650201624,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6584,
      "eval_unseen_dialect_samples_per_second": 60387.033,
      "eval_unseen_dialect_steps_per_second": 472.331,
      "epoch": 5.378096479791395,
      "step": 16500
    },
    {
      "eval_hard_loss": 0.13117605447769165,
      "eval_hard_accuracy": 0.9752747252747253,
      "eval_hard_macro_f1": 0.08229021789522485,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0982,
      "eval_hard_samples_per_second": 7413.598,
      "eval_hard_steps_per_second": 61.101,
      "epoch": 5.378096479791395,
      "step": 16500
    },
    {
      "loss": 0.0033564266562461855,
      "grad_norm": 0.00723454961553216,
      "learning_rate": 0.0017872437779913428,
      "epoch": 5.410691003911343,
      "step": 16600
    },
    {
      "loss": 0.0022684647142887115,
      "grad_norm": 0.016895825043320656,
      "learning_rate": 0.0017615173484258418,
      "epoch": 5.475880052151239,
      "step": 16800
    },
    {
      "loss": 0.0026698514819145204,
      "grad_norm": 0.000877934624440968,
      "learning_rate": 0.0017357113289031892,
      "epoch": 5.541069100391134,
      "step": 17000
    },
    {
      "loss": 0.0015284459292888642,
      "grad_norm": 0.0027167557273060083,
      "learning_rate": 0.0017098335732044065,
      "epoch": 5.60625814863103,
      "step": 17200
    },
    {
      "loss": 0.002852696180343628,
      "grad_norm": 0.035916391760110855,
      "learning_rate": 0.0016838919569426402,
      "epoch": 5.6714471968709255,
      "step": 17400
    },
    {
      "loss": 0.003841307759284973,
      "grad_norm": 0.036015864461660385,
      "learning_rate": 0.001657894375166304,
      "epoch": 5.736636245110821,
      "step": 17600
    },
    {
      "loss": 0.0060520821809768675,
      "grad_norm": 0.015460636466741562,
      "learning_rate": 0.0016318487399563054,
      "epoch": 5.801825293350717,
      "step": 17800
    },
    {
      "loss": 0.003214528262615204,
      "grad_norm": 0.003109982004389167,
      "learning_rate": 0.0016057629780180922,
      "epoch": 5.867014341590613,
      "step": 18000
    },
    {
      "eval_held_loss": 0.1356932818889618,
      "eval_held_accuracy": 0.9770664906781059,
      "eval_held_macro_f1": 0.9762518625413512,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1949,
      "eval_held_samples_per_second": 31094.798,
      "eval_held_steps_per_second": 246.255,
      "epoch": 5.867014341590613,
      "step": 18000
    },
    {
      "eval_unseen_phrasing_loss": 0.7676151990890503,
      "eval_unseen_phrasing_accuracy": 0.8855898200160496,
      "eval_unseen_phrasing_macro_f1": 0.15482806984558908,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2178,
      "eval_unseen_phrasing_samples_per_second": 40053.68,
      "eval_unseen_phrasing_steps_per_second": 316.83,
      "epoch": 5.867014341590613,
      "step": 18000
    },
    {
      "eval_unseen_dialect_loss": 0.44217127561569214,
      "eval_unseen_dialect_accuracy": 0.9346595910565629,
      "eval_unseen_dialect_macro_f1": 0.93632479470951,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6435,
      "eval_unseen_dialect_samples_per_second": 61792.883,
      "eval_unseen_dialect_steps_per_second": 483.328,
      "epoch": 5.867014341590613,
      "step": 18000
    },
    {
      "eval_hard_loss": 0.10372380167245865,
      "eval_hard_accuracy": 0.9862637362637363,
      "eval_hard_macro_f1": 0.08275703088981097,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0732,
      "eval_hard_samples_per_second": 9946.296,
      "eval_hard_steps_per_second": 81.975,
      "epoch": 5.867014341590613,
      "step": 18000
    },
    {
      "loss": 0.0028575989603996277,
      "grad_norm": 0.00823401752859354,
      "learning_rate": 0.0015796450282692448,
      "epoch": 5.932203389830509,
      "step": 18200
    },
    {
      "loss": 0.004153192639350891,
      "grad_norm": 0.06483903527259827,
      "learning_rate": 0.0015535028394233527,
      "epoch": 5.9973924380704045,
      "step": 18400
    },
    {
      "loss": 0.0011822809278964996,
      "grad_norm": 0.012307186610996723,
      "learning_rate": 0.001527344367570911,
      "epoch": 6.0625814863103,
      "step": 18600
    },
    {
      "loss": 0.00023670084774494172,
      "grad_norm": 0.0008011198951862752,
      "learning_rate": 0.0015011775737579708,
      "epoch": 6.127770534550195,
      "step": 18800
    },
    {
      "loss": 0.00024879436939954756,
      "grad_norm": 0.002788627753034234,
      "learning_rate": 0.001475010421563281,
      "epoch": 6.192959582790091,
      "step": 19000
    },
    {
      "loss": 0.0006249099969863892,
      "grad_norm": 0.016984356567263603,
      "learning_rate": 0.0014488508746746593,
      "epoch": 6.258148631029987,
      "step": 19200
    },
    {
      "loss": 0.0004946313798427582,
      "grad_norm": 0.010328582488000393,
      "learning_rate": 0.0014227068944653327,
      "epoch": 6.3233376792698825,
      "step": 19400
    },
    {
      "eval_held_loss": 0.1532353013753891,
      "eval_held_accuracy": 0.9765715228510147,
      "eval_held_macro_f1": 0.976447494258056,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1938,
      "eval_held_samples_per_second": 31276.823,
      "eval_held_steps_per_second": 247.696,
      "epoch": 6.3559322033898304,
      "step": 19500
    },
    {
      "eval_unseen_phrasing_loss": 1.0850872993469238,
      "eval_unseen_phrasing_accuracy": 0.8652986357904391,
      "eval_unseen_phrasing_macro_f1": 0.1527593348892912,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2103,
      "eval_unseen_phrasing_samples_per_second": 41483.972,
      "eval_unseen_phrasing_steps_per_second": 328.143,
      "epoch": 6.3559322033898304,
      "step": 19500
    },
    {
      "eval_unseen_dialect_loss": 0.41693857312202454,
      "eval_unseen_dialect_accuracy": 0.9399411483614597,
      "eval_unseen_dialect_macro_f1": 0.9395283426436881,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.665,
      "eval_unseen_dialect_samples_per_second": 59790.788,
      "eval_unseen_dialect_steps_per_second": 467.668,
      "epoch": 6.3559322033898304,
      "step": 19500
    },
    {
      "eval_hard_loss": 0.23710276186466217,
      "eval_hard_accuracy": 0.9629120879120879,
      "eval_hard_macro_f1": 0.08175880569162584,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0721,
      "eval_hard_samples_per_second": 10095.128,
      "eval_hard_steps_per_second": 83.202,
      "epoch": 6.3559322033898304,
      "step": 19500
    },
    {
      "loss": 0.0015800869464874267,
      "grad_norm": 0.0011451613390818238,
      "learning_rate": 0.001396586437570978,
      "epoch": 6.388526727509778,
      "step": 19600
    },
    {
      "loss": 0.0010473155230283737,
      "grad_norm": 0.0833016186952591,
      "learning_rate": 0.001370497453468208,
      "epoch": 6.453715775749674,
      "step": 19800
    },
    {
      "loss": 0.0006658817827701569,
      "grad_norm": 0.0005158233689144254,
      "learning_rate": 0.001344447882055233,
      "epoch": 6.51890482398957,
      "step": 20000
    },
    {
      "loss": 0.0011239054799079896,
      "grad_norm": 0.0006993607967160642,
      "learning_rate": 0.001318445651235441,
      "epoch": 6.584093872229466,
      "step": 20200
    },
    {
      "loss": 0.0009567379206418991,
      "grad_norm": 0.005306244362145662,
      "learning_rate": 0.001292498674504626,
      "epoch": 6.6492829204693615,
      "step": 20400
    },
    {
      "loss": 0.002421462684869766,
      "grad_norm": 0.0023437021300196648,
      "learning_rate": 0.0012666148485426017,
      "epoch": 6.7144719687092564,
      "step": 20600
    },
    {
      "loss": 0.002153696715831757,
      "grad_norm": 0.0007901916978880763,
      "learning_rate": 0.0012408020508099334,
      "epoch": 6.779661016949152,
      "step": 20800
    },
    {
      "loss": 0.0012323959171772004,
      "grad_norm": 0.0017923114355653524,
      "learning_rate": 0.001215068137150517,
      "epoch": 6.844850065189048,
      "step": 21000
    },
    {
      "eval_held_loss": 0.15646591782569885,
      "eval_held_accuracy": 0.9722818016828906,
      "eval_held_macro_f1": 0.9732346766031365,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1877,
      "eval_held_samples_per_second": 32283.339,
      "eval_held_steps_per_second": 255.667,
      "epoch": 6.844850065189048,
      "step": 21000
    },
    {
      "eval_unseen_phrasing_loss": 0.9185952544212341,
      "eval_unseen_phrasing_accuracy": 0.8755015476326952,
      "eval_unseen_phrasing_macro_f1": 0.15405079929911678,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.209,
      "eval_unseen_phrasing_samples_per_second": 41736.963,
      "eval_unseen_phrasing_steps_per_second": 330.144,
      "epoch": 6.844850065189048,
      "step": 21000
    },
    {
      "eval_unseen_dialect_loss": 0.3974415957927704,
      "eval_unseen_dialect_accuracy": 0.9438645909308117,
      "eval_unseen_dialect_macro_f1": 0.9435824256304857,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6298,
      "eval_unseen_dialect_samples_per_second": 63136.843,
      "eval_unseen_dialect_steps_per_second": 493.84,
      "epoch": 6.844850065189048,
      "step": 21000
    },
    {
      "eval_hard_loss": 0.24147410690784454,
      "eval_hard_accuracy": 0.9574175824175825,
      "eval_hard_macro_f1": 0.08152046783625731,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0987,
      "eval_hard_samples_per_second": 7378.824,
      "eval_hard_steps_per_second": 60.814,
      "epoch": 6.844850065189048,
      "step": 21000
    },
    {
      "loss": 0.0020152950286865235,
      "grad_norm": 0.03562590479850769,
      "learning_rate": 0.001189420939400741,
      "epoch": 6.910039113428944,
      "step": 21200
    },
    {
      "loss": 0.0015001146495342255,
      "grad_norm": 0.005310706794261932,
      "learning_rate": 0.0011638682630059508,
      "epoch": 6.97522816166884,
      "step": 21400
    },
    {
      "loss": 0.0007807877659797668,
      "grad_norm": 0.32783180475234985,
      "learning_rate": 0.0011384178846449461,
      "epoch": 7.040417209908735,
      "step": 21600
    },
    {
      "loss": 0.00022133855149149894,
      "grad_norm": 0.0022341725416481495,
      "learning_rate": 0.001113077549863235,
      "epoch": 7.105606258148631,
      "step": 21800
    },
    {
      "loss": 0.00011499571613967419,
      "grad_norm": 0.00012046092888340354,
      "learning_rate": 0.0010878549707157602,
      "epoch": 7.170795306388527,
      "step": 22000
    },
    {
      "loss": 0.0001301498617976904,
      "grad_norm": 0.0006064897170290351,
      "learning_rate": 0.0010627578234198192,
      "epoch": 7.235984354628423,
      "step": 22200
    },
    {
      "loss": 0.00017196711152791977,
      "grad_norm": 7.804731285432354e-05,
      "learning_rate": 0.0010377937460188907,
      "epoch": 7.301173402868318,
      "step": 22400
    },
    {
      "eval_held_loss": 0.14113515615463257,
      "eval_held_accuracy": 0.9780564263322884,
      "eval_held_macro_f1": 0.976986382087114,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1908,
      "eval_held_samples_per_second": 31768.399,
      "eval_held_steps_per_second": 251.589,
      "epoch": 7.333767926988266,
      "step": 22500
    },
    {
      "eval_unseen_phrasing_loss": 0.9802523255348206,
      "eval_unseen_phrasing_accuracy": 0.8789407313997478,
      "eval_unseen_phrasing_macro_f1": 0.1537979714714063,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2119,
      "eval_unseen_phrasing_samples_per_second": 41170.972,
      "eval_unseen_phrasing_steps_per_second": 325.667,
      "epoch": 7.333767926988266,
      "step": 22500
    },
    {
      "eval_unseen_dialect_loss": 0.4899526536464691,
      "eval_unseen_dialect_accuracy": 0.9392117904479264,
      "eval_unseen_dialect_macro_f1": 0.9396435854458236,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6642,
      "eval_unseen_dialect_samples_per_second": 59865.781,
      "eval_unseen_dialect_steps_per_second": 468.254,
      "epoch": 7.333767926988266,
      "step": 22500
    },
    {
      "eval_hard_loss": 0.16756650805473328,
      "eval_hard_accuracy": 0.9821428571428571,
      "eval_hard_macro_f1": 0.08258258258258257,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0693,
      "eval_hard_samples_per_second": 10509.218,
      "eval_hard_steps_per_second": 86.614,
      "epoch": 7.333767926988266,
      "step": 22500
    },
    {
      "loss": 0.00012516571208834648,
      "grad_norm": 0.023826057091355324,
      "learning_rate": 0.001012970336058082,
      "epoch": 7.3663624511082135,
      "step": 22600
    },
    {
      "loss": 0.00010503841564059257,
      "grad_norm": 0.0013365615159273148,
      "learning_rate": 0.0009882951482718945,
      "epoch": 7.431551499348109,
      "step": 22800
    },
    {
      "loss": 0.00014142820611596108,
      "grad_norm": 0.0003191770811099559,
      "learning_rate": 0.0009637756922850269,
      "epoch": 7.496740547588005,
      "step": 23000
    },
    {
      "loss": 0.00013418926857411862,
      "grad_norm": 6.237982597667724e-05,
      "learning_rate": 0.0009394194303268989,
      "epoch": 7.561929595827901,
      "step": 23200
    },
    {
      "loss": 0.0003565175831317902,
      "grad_norm": 0.00036435143556445837,
      "learning_rate": 0.0009152337749606063,
      "epoch": 7.627118644067797,
      "step": 23400
    },
    {
      "loss": 0.0003813246637582779,
      "grad_norm": 5.814128599013202e-05,
      "learning_rate": 0.0008912260868269847,
      "epoch": 7.6923076923076925,
      "step": 23600
    },
    {
      "loss": 0.00010491352528333663,
      "grad_norm": 0.0005376170738600194,
      "learning_rate": 0.0008674036724044822,
      "epoch": 7.757496740547588,
      "step": 23800
    },
    {
      "loss": 0.000499812588095665,
      "grad_norm": 0.006998493801802397,
      "learning_rate": 0.0008437737817855095,
      "epoch": 7.822685788787483,
      "step": 24000
    },
    {
      "eval_held_loss": 0.1482814997434616,
      "eval_held_accuracy": 0.977231479953803,
      "eval_held_macro_f1": 0.977377433110548,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.202,
      "eval_held_samples_per_second": 30006.311,
      "eval_held_steps_per_second": 237.635,
      "epoch": 7.822685788787483,
      "step": 24000
    },
    {
      "eval_unseen_phrasing_loss": 1.0274180173873901,
      "eval_unseen_phrasing_accuracy": 0.8768772211395163,
      "eval_unseen_phrasing_macro_f1": 0.1535257702487184,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2125,
      "eval_unseen_phrasing_samples_per_second": 41044.834,
      "eval_unseen_phrasing_steps_per_second": 324.67,
      "epoch": 7.822685788787483,
      "step": 24000
    },
    {
      "eval_unseen_dialect_loss": 0.48141905665397644,
      "eval_unseen_dialect_accuracy": 0.9400417494529816,
      "eval_unseen_dialect_macro_f1": 0.9398712618597806,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6429,
      "eval_unseen_dialect_samples_per_second": 61850.474,
      "eval_unseen_dialect_steps_per_second": 483.778,
      "epoch": 7.822685788787483,
      "step": 24000
    },
    {
      "eval_hard_loss": 0.22298771142959595,
      "eval_hard_accuracy": 0.9684065934065934,
      "eval_hard_macro_f1": 0.08199581297976273,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0708,
      "eval_hard_samples_per_second": 10289.58,
      "eval_hard_steps_per_second": 84.804,
      "epoch": 7.822685788787483,
      "step": 24000
    },
    {
      "loss": 0.0003143933415412903,
      "grad_norm": 0.01763354055583477,
      "learning_rate": 0.0008203436064699542,
      "epoch": 7.887874837027379,
      "step": 24200
    },
    {
      "loss": 0.00017375987023115157,
      "grad_norm": 0.00020164674788247794,
      "learning_rate": 0.0007971202771765272,
      "epoch": 7.953063885267275,
      "step": 24400
    },
    {
      "loss": 0.0002944411523640156,
      "grad_norm": 0.0006192452274262905,
      "learning_rate": 0.000774110861672603,
      "epoch": 8.01825293350717,
      "step": 24600
    },
    {
      "loss": 7.585027255117893e-05,
      "grad_norm": 0.00010485780512681231,
      "learning_rate": 0.0007513223626232257,
      "epoch": 8.083441981747066,
      "step": 24800
    },
    {
      "loss": 2.1936502307653427e-05,
      "grad_norm": 0.00032462243689224124,
      "learning_rate": 0.0007287617154599214,
      "epoch": 8.148631029986962,
      "step": 25000
    },
    {
      "loss": 3.8847094401717186e-05,
      "grad_norm": 0.0007482388755306602,
      "learning_rate": 0.0007064357862699758,
      "epoch": 8.213820078226858,
      "step": 25200
    },
    {
      "loss": 1.7666900530457495e-05,
      "grad_norm": 0.001964268973097205,
      "learning_rate": 0.0006843513697068162,
      "epoch": 8.279009126466754,
      "step": 25400
    },
    {
      "eval_held_loss": 0.15441980957984924,
      "eval_held_accuracy": 0.9773964692295001,
      "eval_held_macro_f1": 0.9770415452913186,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.189,
      "eval_held_samples_per_second": 32065.08,
      "eval_held_steps_per_second": 253.939,
      "epoch": 8.3116036505867,
      "step": 25500
    },
    {
      "eval_unseen_phrasing_loss": 1.0938074588775635,
      "eval_unseen_phrasing_accuracy": 0.8665596698383584,
      "eval_unseen_phrasing_macro_f1": 0.15259715641632401,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2219,
      "eval_unseen_phrasing_samples_per_second": 39312.192,
      "eval_unseen_phrasing_steps_per_second": 310.964,
      "epoch": 8.3116036505867,
      "step": 25500
    },
    {
      "eval_unseen_dialect_loss": 0.4920024573802948,
      "eval_unseen_dialect_accuracy": 0.9411232111868414,
      "eval_unseen_dialect_macro_f1": 0.9411037219947037,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6468,
      "eval_unseen_dialect_samples_per_second": 61475.828,
      "eval_unseen_dialect_steps_per_second": 480.848,
      "epoch": 8.3116036505867,
      "step": 25500
    },
    {
      "eval_hard_loss": 0.18859992921352386,
      "eval_hard_accuracy": 0.9752747252747253,
      "eval_hard_macro_f1": 0.08229021789522485,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0995,
      "eval_hard_samples_per_second": 7317.92,
      "eval_hard_steps_per_second": 60.313,
      "epoch": 8.3116036505867,
      "step": 25500
    },
    {
      "loss": 8.592988364398479e-05,
      "grad_norm": 0.0018528795335441828,
      "learning_rate": 0.0006625151869221302,
      "epoch": 8.34419817470665,
      "step": 25600
    },
    {
      "loss": 3.838728182017803e-05,
      "grad_norm": 6.19717247900553e-05,
      "learning_rate": 0.0006409338835203597,
      "epoch": 8.409387222946545,
      "step": 25800
    },
    {
      "loss": 1.7936211079359055e-05,
      "grad_norm": 0.0003067332727368921,
      "learning_rate": 0.0006196140275361816,
      "epoch": 8.474576271186441,
      "step": 26000
    },
    {
      "loss": 2.3439964279532434e-05,
      "grad_norm": 2.731915628828574e-05,
      "learning_rate": 0.0005985621074355978,
      "epoch": 8.539765319426337,
      "step": 26200
    },
    {
      "loss": 1.97579525411129e-05,
      "grad_norm": 4.191105836071074e-05,
      "learning_rate": 0.0005777845301412427,
      "epoch": 8.604954367666233,
      "step": 26400
    },
    {
      "loss": 2.1651322022080422e-05,
      "grad_norm": 9.114162821788341e-05,
      "learning_rate": 0.0005572876190825047,
      "epoch": 8.670143415906129,
      "step": 26600
    },
    {
      "loss": 1.93751510232687e-05,
      "grad_norm": 0.0029533361084759235,
      "learning_rate": 0.0005370776122710588,
      "epoch": 8.735332464146023,
      "step": 26800
    },
    {
      "loss": 1.8145898357033728e-05,
      "grad_norm": 0.000156165478983894,
      "learning_rate": 0.000517160660402398,
      "epoch": 8.800521512385918,
      "step": 27000
    },
    {
      "eval_held_loss": 0.1512400209903717,
      "eval_held_accuracy": 0.9782214156079855,
      "eval_held_macro_f1": 0.9777303881326141,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1981,
      "eval_held_samples_per_second": 30589.349,
      "eval_held_steps_per_second": 242.252,
      "epoch": 8.800521512385918,
      "step": 27000
    },
    {
      "eval_unseen_phrasing_loss": 1.0126179456710815,
      "eval_unseen_phrasing_accuracy": 0.8780236157285337,
      "eval_unseen_phrasing_macro_f1": 0.1535370480342466,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2131,
      "eval_unseen_phrasing_samples_per_second": 40933.43,
      "eval_unseen_phrasing_steps_per_second": 323.788,
      "epoch": 8.800521512385918,
      "step": 27000
    },
    {
      "eval_unseen_dialect_loss": 0.5041497945785522,
      "eval_unseen_dialect_accuracy": 0.9404693040919494,
      "eval_unseen_dialect_macro_f1": 0.9407379024830506,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6313,
      "eval_unseen_dialect_samples_per_second": 62986.614,
      "eval_unseen_dialect_steps_per_second": 492.665,
      "epoch": 8.800521512385918,
      "step": 27000
    },
    {
      "eval_hard_loss": 0.18605831265449524,
      "eval_hard_accuracy": 0.9793956043956044,
      "eval_hard_macro_f1": 0.08246588017580384,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0727,
      "eval_hard_samples_per_second": 10019.765,
      "eval_hard_steps_per_second": 82.58,
      "epoch": 8.800521512385918,
      "step": 27000
    },
    {
      "loss": 1.3405987992882728e-05,
      "grad_norm": 0.00017298512102570385,
      "learning_rate": 0.0004975428249839343,
      "epoch": 8.865710560625814,
      "step": 27200
    },
    {
      "loss": 1.4411471784114838e-05,
      "grad_norm": 0.0003377518441993743,
      "learning_rate": 0.00047823007649024497,
      "epoch": 8.93089960886571,
      "step": 27400
    },
    {
      "loss": 1.4787353575229644e-05,
      "grad_norm": 4.055908721056767e-05,
      "learning_rate": 0.00045922829254602506,
      "epoch": 8.996088657105606,
      "step": 27600
    },
    {
      "loss": 1.0408128146082163e-05,
      "grad_norm": 0.000422323850216344,
      "learning_rate": 0.0004405432561372935,
      "epoch": 9.061277705345502,
      "step": 27800
    },
    {
      "loss": 3.9707869291305545e-06,
      "grad_norm": 0.00012347962183412164,
      "learning_rate": 0.0004221806538514079,
      "epoch": 9.126466753585397,
      "step": 28000
    },
    {
      "loss": 4.812553524971008e-06,
      "grad_norm": 0.00021798703528475016,
      "learning_rate": 0.0004041460741464107,
      "epoch": 9.191655801825293,
      "step": 28200
    },
    {
      "loss": 4.756934940814972e-06,
      "grad_norm": 0.00022741024440620095,
      "learning_rate": 0.0003864450056502445,
      "epoch": 9.256844850065189,
      "step": 28400
    },
    {
      "eval_held_loss": 0.15126317739486694,
      "eval_held_accuracy": 0.9782214156079855,
      "eval_held_macro_f1": 0.9776621918961469,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1929,
      "eval_held_samples_per_second": 31415.27,
      "eval_held_steps_per_second": 248.793,
      "epoch": 9.289439374185136,
      "step": 28500
    },
    {
      "eval_unseen_phrasing_loss": 0.9854173064231873,
      "eval_unseen_phrasing_accuracy": 0.8827238335435057,
      "eval_unseen_phrasing_macro_f1": 0.1540440702746768,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2114,
      "eval_unseen_phrasing_samples_per_second": 41266.539,
      "eval_unseen_phrasing_steps_per_second": 326.423,
      "epoch": 9.289439374185136,
      "step": 28500
    },
    {
      "eval_unseen_dialect_loss": 0.5147712230682373,
      "eval_unseen_dialect_accuracy": 0.9400417494529816,
      "eval_unseen_dialect_macro_f1": 0.940186034501196,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6268,
      "eval_unseen_dialect_samples_per_second": 63438.714,
      "eval_unseen_dialect_steps_per_second": 496.201,
      "epoch": 9.289439374185136,
      "step": 28500
    },
    {
      "eval_hard_loss": 0.18782676756381989,
      "eval_hard_accuracy": 0.9807692307692307,
      "eval_hard_macro_f1": 0.08252427184466019,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0679,
      "eval_hard_samples_per_second": 10727.12,
      "eval_hard_steps_per_second": 88.41,
      "epoch": 9.289439374185136,
      "step": 28500
    },
    {
      "loss": 4.3132249265909195e-06,
      "grad_norm": 0.00037367508048191667,
      "learning_rate": 0.0003690828354903469,
      "epoch": 9.322033898305085,
      "step": 28600
    },
    {
      "loss": 5.851099267601967e-06,
      "grad_norm": 0.0006751463515684009,
      "learning_rate": 0.0003520648476541379,
      "epoch": 9.38722294654498,
      "step": 28800
    },
    {
      "loss": 3.5041105002164843e-06,
      "grad_norm": 5.1099781558150426e-05,
      "learning_rate": 0.00033539622138089426,
      "epoch": 9.452411994784876,
      "step": 29000
    },
    {
      "loss": 3.877617418766022e-06,
      "grad_norm": 3.238074350520037e-05,
      "learning_rate": 0.00031908202958550667,
      "epoch": 9.517601043024772,
      "step": 29200
    },
    {
      "loss": 3.754002973437309e-06,
      "grad_norm": 0.00022564925893675536,
      "learning_rate": 0.0003031272373145915,
      "epoch": 9.582790091264668,
      "step": 29400
    },
    {
      "loss": 3.7502124905586245e-06,
      "grad_norm": 8.494478970533237e-05,
      "learning_rate": 0.0002875367002354337,
      "epoch": 9.647979139504564,
      "step": 29600
    },
    {
      "loss": 2.9243528842926025e-06,
      "grad_norm": 4.193716449663043e-06,
      "learning_rate": 0.00027231516315821786,
      "epoch": 9.71316818774446,
      "step": 29800
    },
    {
      "loss": 4.24572266638279e-06,
      "grad_norm": 0.0001473828888265416,
      "learning_rate": 0.0002574672585919948,
      "epoch": 9.778357235984355,
      "step": 30000
    },
    {
      "eval_held_loss": 0.15340350568294525,
      "eval_held_accuracy": 0.9788813727107738,
      "eval_held_macro_f1": 0.9781646456301698,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1976,
      "eval_held_samples_per_second": 30680.831,
      "eval_held_steps_per_second": 242.976,
      "epoch": 9.778357235984355,
      "step": 30000
    },
    {
      "eval_unseen_phrasing_loss": 0.9964732527732849,
      "eval_unseen_phrasing_accuracy": 0.8821506362489969,
      "eval_unseen_phrasing_macro_f1": 0.15407220696385895,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2136,
      "eval_unseen_phrasing_samples_per_second": 40842.635,
      "eval_unseen_phrasing_steps_per_second": 323.07,
      "epoch": 9.778357235984355,
      "step": 30000
    },
    {
      "eval_unseen_dialect_loss": 0.5279785990715027,
      "eval_unseen_dialect_accuracy": 0.9397650964512965,
      "eval_unseen_dialect_macro_f1": 0.9399695311054369,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6451,
      "eval_unseen_dialect_samples_per_second": 61636.852,
      "eval_unseen_dialect_steps_per_second": 482.107,
      "epoch": 9.778357235984355,
      "step": 30000
    },
    {
      "eval_hard_loss": 0.19260503351688385,
      "eval_hard_accuracy": 0.9793956043956044,
      "eval_hard_macro_f1": 0.08246588017580384,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.068,
      "eval_hard_samples_per_second": 10701.405,
      "eval_hard_steps_per_second": 88.198,
      "epoch": 9.778357235984355,
      "step": 30000
    },
    {
      "loss": 3.5035330802202224e-06,
      "grad_norm": 8.517535752616823e-05,
      "learning_rate": 0.0002429975053348308,
      "epoch": 9.843546284224251,
      "step": 30200
    },
    {
      "loss": 3.36955301463604e-06,
      "grad_norm": 0.00014641178131569177,
      "learning_rate": 0.00022891030709855858,
      "epoch": 9.908735332464147,
      "step": 30400
    },
    {
      "loss": 3.6497227847576143e-06,
      "grad_norm": 0.00019411781977396458,
      "learning_rate": 0.00021520995116855614,
      "epoch": 9.973924380704041,
      "step": 30600
    },
    {
      "loss": 3.276835777796805e-06,
      "grad_norm": 8.209656698454637e-06,
      "learning_rate": 0.00020190060709895818,
      "epoch": 10.039113428943937,
      "step": 30800
    },
    {
      "loss": 2.113180235028267e-06,
      "grad_norm": 0.00045076521928422153,
      "learning_rate": 0.0001889863254436982,
      "epoch": 10.104302477183833,
      "step": 31000
    },
    {
      "loss": 2.7205795049667357e-06,
      "grad_norm": 0.00012407796748448163,
      "learning_rate": 0.000176471036523765,
      "epoch": 10.169491525423728,
      "step": 31200
    },
    {
      "loss": 2.047289162874222e-06,
      "grad_norm": 8.700434409547597e-05,
      "learning_rate": 0.0001643585492310538,
      "epoch": 10.234680573663624,
      "step": 31400
    },
    {
      "eval_held_loss": 0.15544317662715912,
      "eval_held_accuracy": 0.9785513941593796,
      "eval_held_macro_f1": 0.977789708896006,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1974,
      "eval_held_samples_per_second": 30700.765,
      "eval_held_steps_per_second": 243.134,
      "epoch": 10.267275097783573,
      "step": 31500
    },
    {
      "eval_unseen_phrasing_loss": 1.017669916152954,
      "eval_unseen_phrasing_accuracy": 0.8815774389544881,
      "eval_unseen_phrasing_macro_f1": 0.15397885271779754,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2112,
      "eval_unseen_phrasing_samples_per_second": 41306.0,
      "eval_unseen_phrasing_steps_per_second": 326.736,
      "epoch": 10.267275097783573,
      "step": 31500
    },
    {
      "eval_unseen_dialect_loss": 0.539714515209198,
      "eval_unseen_dialect_accuracy": 0.9396141948140138,
      "eval_unseen_dialect_macro_f1": 0.9399248879885271,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6584,
      "eval_unseen_dialect_samples_per_second": 60393.09,
      "eval_unseen_dialect_steps_per_second": 472.379,
      "epoch": 10.267275097783573,
      "step": 31500
    },
    {
      "eval_hard_loss": 0.19299156963825226,
      "eval_hard_accuracy": 0.978021978021978,
      "eval_hard_macro_f1": 0.0824074074074074,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0984,
      "eval_hard_samples_per_second": 7397.63,
      "eval_hard_steps_per_second": 60.969,
      "epoch": 10.267275097783573,
      "step": 31500
    },
    {
      "loss": 2.323649823665619e-06,
      "grad_norm": 4.971117959939875e-06,
      "learning_rate": 0.00015265254986917133,
      "epoch": 10.29986962190352,
      "step": 31600
    },
    {
      "loss": 2.3561064153909682e-06,
      "grad_norm": 1.534152943349909e-05,
      "learning_rate": 0.00014135660103154846,
      "epoch": 10.365058670143416,
      "step": 31800
    },
    {
      "loss": 2.536037936806679e-06,
      "grad_norm": 0.00017727049998939037,
      "learning_rate": 0.00013047414051720502,
      "epoch": 10.430247718383312,
      "step": 32000
    },
    {
      "loss": 2.1120160818099978e-06,
      "grad_norm": 2.0081704860785976e-05,
      "learning_rate": 0.00012000848028449157,
      "epoch": 10.495436766623207,
      "step": 32200
    },
    {
      "loss": 2.114633098244667e-06,
      "grad_norm": 2.7619544198387302e-05,
      "learning_rate": 0.00010996280544313225,
      "epoch": 10.560625814863103,
      "step": 32400
    },
    {
      "loss": 2.2349413484334946e-06,
      "grad_norm": 5.3831998229725286e-05,
      "learning_rate": 0.000100340173284871,
      "epoch": 10.625814863102999,
      "step": 32600
    },
    {
      "loss": 2.1641142666339873e-06,
      "grad_norm": 2.118246629834175e-05,
      "learning_rate": 9.114351235301766e-05,
      "epoch": 10.691003911342895,
      "step": 32800
    },
    {
      "loss": 2.128463238477707e-06,
      "grad_norm": 1.19837059173733e-05,
      "learning_rate": 8.237562155117828e-05,
      "epoch": 10.75619295958279,
      "step": 33000
    },
    {
      "eval_held_loss": 0.1571933925151825,
      "eval_held_accuracy": 0.9785513941593796,
      "eval_held_macro_f1": 0.977789708896006,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.2017,
      "eval_held_samples_per_second": 30045.179,
      "eval_held_steps_per_second": 237.942,
      "epoch": 10.75619295958279,
      "step": 33000
    },
    {
      "eval_unseen_phrasing_loss": 1.0116543769836426,
      "eval_unseen_phrasing_accuracy": 0.8828384730024075,
      "eval_unseen_phrasing_macro_f1": 0.15414686477971626,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2218,
      "eval_unseen_phrasing_samples_per_second": 39324.742,
      "eval_unseen_phrasing_steps_per_second": 311.064,
      "epoch": 10.75619295958279,
      "step": 33000
    },
    {
      "eval_unseen_dialect_loss": 0.5479562878608704,
      "eval_unseen_dialect_accuracy": 0.93941299263097,
      "eval_unseen_dialect_macro_f1": 0.9397113320624547,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6578,
      "eval_unseen_dialect_samples_per_second": 60442.164,
      "eval_unseen_dialect_steps_per_second": 472.763,
      "epoch": 10.75619295958279,
      "step": 33000
    },
    {
      "eval_hard_loss": 0.19669130444526672,
      "eval_hard_accuracy": 0.978021978021978,
      "eval_hard_macro_f1": 0.0824074074074074,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0686,
      "eval_hard_samples_per_second": 10616.977,
      "eval_hard_steps_per_second": 87.503,
      "epoch": 10.75619295958279,
      "step": 33000
    },
    {
      "loss": 1.7195940017700196e-06,
      "grad_norm": 6.149806722532958e-05,
      "learning_rate": 7.403916929143851e-05,
      "epoch": 10.821382007822686,
      "step": 33200
    },
    {
      "loss": 2.02791765332222e-06,
      "grad_norm": 1.4440478480537422e-05,
      "learning_rate": 6.613669268226247e-05,
      "epoch": 10.886571056062582,
      "step": 33400
    },
    {
      "loss": 1.7724744975566865e-06,
      "grad_norm": 1.6741221770644188e-05,
      "learning_rate": 5.8670596756350667e-05,
      "epoch": 10.951760104302478,
      "step": 33600
    },
    {
      "loss": 2.0193826640024783e-06,
      "grad_norm": 7.293969974853098e-05,
      "learning_rate": 5.164315373869477e-05,
      "epoch": 11.016949152542374,
      "step": 33800
    },
    {
      "loss": 1.4859996736049651e-06,
      "grad_norm": 0.00028648285660892725,
      "learning_rate": 4.50565023550506e-05,
      "epoch": 11.082138200782268,
      "step": 34000
    },
    {
      "loss": 1.589478924870491e-06,
      "grad_norm": 7.355489651672542e-05,
      "learning_rate": 3.891264718104015e-05,
      "epoch": 11.147327249022164,
      "step": 34200
    },
    {
      "loss": 1.5578418970108032e-06,
      "grad_norm": 0.00012585640070028603,
      "learning_rate": 3.321345803208092e-05,
      "epoch": 11.21251629726206,
      "step": 34400
    },
    {
      "eval_held_loss": 0.1582455188035965,
      "eval_held_accuracy": 0.9788813727107738,
      "eval_held_macro_f1": 0.9780138491772182,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1915,
      "eval_held_samples_per_second": 31651.085,
      "eval_held_steps_per_second": 250.66,
      "epoch": 11.245110821382008,
      "step": 34500
    },
    {
      "eval_unseen_phrasing_loss": 1.0285286903381348,
      "eval_unseen_phrasing_accuracy": 0.8824945546257021,
      "eval_unseen_phrasing_macro_f1": 0.1541046721624911,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2192,
      "eval_unseen_phrasing_samples_per_second": 39791.308,
      "eval_unseen_phrasing_steps_per_second": 314.754,
      "epoch": 11.245110821382008,
      "step": 34500
    },
    {
      "eval_unseen_dialect_loss": 0.5534936189651489,
      "eval_unseen_dialect_accuracy": 0.9393123915394482,
      "eval_unseen_dialect_macro_f1": 0.9396581191535277,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.642,
      "eval_unseen_dialect_samples_per_second": 61935.648,
      "eval_unseen_dialect_steps_per_second": 484.444,
      "epoch": 11.245110821382008,
      "step": 34500
    },
    {
      "eval_hard_loss": 0.1963353455066681,
      "eval_hard_accuracy": 0.9793956043956044,
      "eval_hard_macro_f1": 0.08246588017580384,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.07,
      "eval_hard_samples_per_second": 10407.028,
      "eval_hard_steps_per_second": 85.772,
      "epoch": 11.245110821382008,
      "step": 34500
    },
    {
      "loss": 1.466916874051094e-06,
      "grad_norm": 1.183496988232946e-05,
      "learning_rate": 2.796066939432812e-05,
      "epoch": 11.277705345501955,
      "step": 34600
    },
    {
      "loss": 1.4404859393835068e-06,
      "grad_norm": 0.00010413937707198784,
      "learning_rate": 2.3155879896802566e-05,
      "epoch": 11.342894393741851,
      "step": 34800
    },
    {
      "loss": 1.8354412168264389e-06,
      "grad_norm": 0.00011732364509953186,
      "learning_rate": 1.88005518248649e-05,
      "epoch": 11.408083441981747,
      "step": 35000
    },
    {
      "loss": 1.6290787607431412e-06,
      "grad_norm": 7.742628076812252e-05,
      "learning_rate": 1.4896010675185868e-05,
      "epoch": 11.473272490221643,
      "step": 35200
    },
    {
      "loss": 1.570470631122589e-06,
      "grad_norm": 3.107869270024821e-05,
      "learning_rate": 1.1443444752345211e-05,
      "epoch": 11.538461538461538,
      "step": 35400
    },
    {
      "loss": 1.4190562069416047e-06,
      "grad_norm": 5.619867442874238e-05,
      "learning_rate": 8.443904807183933e-06,
      "epoch": 11.603650586701434,
      "step": 35600
    },
    {
      "loss": 1.3390928506851196e-06,
      "grad_norm": 3.6323285712569486e-06,
      "learning_rate": 5.898303717020603e-06,
      "epoch": 11.66883963494133,
      "step": 35800
    },
    {
      "loss": 1.2829061597585678e-06,
      "grad_norm": 8.561168942833319e-05,
      "learning_rate": 3.807416207825931e-06,
      "epoch": 11.734028683181226,
      "step": 36000
    },
    {
      "eval_held_loss": 0.15885284543037415,
      "eval_held_accuracy": 0.9787163834350767,
      "eval_held_macro_f1": 0.9779075397528817,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1887,
      "eval_held_samples_per_second": 32119.002,
      "eval_held_steps_per_second": 254.366,
      "epoch": 11.734028683181226,
      "step": 36000
    },
    {
      "eval_unseen_phrasing_loss": 1.0259946584701538,
      "eval_unseen_phrasing_accuracy": 0.8828384730024075,
      "eval_unseen_phrasing_macro_f1": 0.15413669596810647,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2117,
      "eval_unseen_phrasing_samples_per_second": 41210.947,
      "eval_unseen_phrasing_steps_per_second": 325.984,
      "epoch": 11.734028683181226,
      "step": 36000
    },
    {
      "eval_unseen_dialect_loss": 0.5587638020515442,
      "eval_unseen_dialect_accuracy": 0.9392620909936873,
      "eval_unseen_dialect_macro_f1": 0.9396129141890838,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6501,
      "eval_unseen_dialect_samples_per_second": 61162.078,
      "eval_unseen_dialect_steps_per_second": 478.394,
      "epoch": 11.734028683181226,
      "step": 36000
    },
    {
      "eval_hard_loss": 0.19653108716011047,
      "eval_hard_accuracy": 0.978021978021978,
      "eval_hard_macro_f1": 0.0824074074074074,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.0681,
      "eval_hard_samples_per_second": 10697.918,
      "eval_hard_steps_per_second": 88.17,
      "epoch": 11.734028683181226,
      "step": 36000
    },
    {
      "loss": 1.525282859802246e-06,
      "grad_norm": 8.671376417623833e-05,
      "learning_rate": 2.171878618444534e-06,
      "epoch": 11.799217731421122,
      "step": 36200
    },
    {
      "loss": 1.4903955161571502e-06,
      "grad_norm": 3.6585981888492825e-06,
      "learning_rate": 9.921887069311852e-07,
      "epoch": 11.864406779661017,
      "step": 36400
    },
    {
      "loss": 1.5045423060655594e-06,
      "grad_norm": 0.00010040865163318813,
      "learning_rate": 2.687054990636595e-07,
      "epoch": 11.929595827900913,
      "step": 36600
    },
    {
      "loss": 1.4826934784650802e-06,
      "grad_norm": 2.545448842283804e-05,
      "learning_rate": 1.6491790783002002e-09,
      "epoch": 11.994784876140809,
      "step": 36800
    },
    {
      "eval_held_loss": 0.15902547538280487,
      "eval_held_accuracy": 0.9787163834350767,
      "eval_held_macro_f1": 0.9779075397528817,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1824,
      "eval_held_samples_per_second": 33231.732,
      "eval_held_steps_per_second": 263.178,
      "epoch": 12.0,
      "step": 36816
    },
    {
      "eval_unseen_phrasing_loss": 1.0271514654159546,
      "eval_unseen_phrasing_accuracy": 0.8826091940846039,
      "eval_unseen_phrasing_macro_f1": 0.15409486603366546,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2202,
      "eval_unseen_phrasing_samples_per_second": 39620.754,
      "eval_unseen_phrasing_steps_per_second": 313.405,
      "epoch": 12.0,
      "step": 36816
    },
    {
      "eval_unseen_dialect_loss": 0.5586314797401428,
      "eval_unseen_dialect_accuracy": 0.9391866401750459,
      "eval_unseen_dialect_macro_f1": 0.9395719822455334,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6482,
      "eval_unseen_dialect_samples_per_second": 61340.452,
      "eval_unseen_dialect_steps_per_second": 479.789,
      "epoch": 12.0,
      "step": 36816
    },
    {
      "eval_hard_loss": 0.19733069837093353,
      "eval_hard_accuracy": 0.978021978021978,
      "eval_hard_macro_f1": 0.0824074074074074,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.1052,
      "eval_hard_samples_per_second": 6918.408,
      "eval_hard_steps_per_second": 57.02,
      "epoch": 12.0,
      "step": 36816
    },
    {
      "train_runtime": 173.5675,
      "train_samples_per_second": 13571.454,
      "train_steps_per_second": 212.113,
      "total_flos": 11123400221472.0,
      "train_loss": 0.04227041617808625,
      "epoch": 12.0,
      "step": 36816
    },
    {
      "eval_held_loss": 0.15340350568294525,
      "eval_held_accuracy": 0.9788813727107738,
      "eval_held_macro_f1": 0.9781646456301698,
      "eval_held_majority_baseline": 0.19617224880382775,
      "eval_held_runtime": 0.1777,
      "eval_held_samples_per_second": 34114.127,
      "eval_held_steps_per_second": 270.166,
      "epoch": 12.0,
      "step": 36816
    },
    {
      "eval_unseen_phrasing_loss": 0.9964732527732849,
      "eval_unseen_phrasing_accuracy": 0.8821506362489969,
      "eval_unseen_phrasing_macro_f1": 0.15407220696385895,
      "eval_unseen_phrasing_majority_baseline": 0.5576063280981314,
      "eval_unseen_phrasing_runtime": 0.2235,
      "eval_unseen_phrasing_samples_per_second": 39027.04,
      "eval_unseen_phrasing_steps_per_second": 308.709,
      "epoch": 12.0,
      "step": 36816
    },
    {
      "eval_unseen_dialect_loss": 0.5279785990715027,
      "eval_unseen_dialect_accuracy": 0.9397650964512965,
      "eval_unseen_dialect_macro_f1": 0.9399695311054369,
      "eval_unseen_dialect_majority_baseline": 0.18739468323231306,
      "eval_unseen_dialect_runtime": 0.6569,
      "eval_unseen_dialect_samples_per_second": 60524.006,
      "eval_unseen_dialect_steps_per_second": 473.403,
      "epoch": 12.0,
      "step": 36816
    },
    {
      "eval_hard_loss": 0.19260503351688385,
      "eval_hard_accuracy": 0.9793956043956044,
      "eval_hard_macro_f1": 0.08246588017580384,
      "eval_hard_majority_baseline": 1.0,
      "eval_hard_runtime": 0.1007,
      "eval_hard_samples_per_second": 7229.025,
      "eval_hard_steps_per_second": 59.58,
      "epoch": 12.0,
      "step": 36816
    }
  ]
}