diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,4833 @@
+{
+  "best_metric": 1.7600704893538932e-07,
+  "best_model_checkpoint": "Models/t5-base-class-gen/checkpoint-32000",
+  "epoch": 14.925373134328359,
+  "eval_steps": 100,
+  "global_step": 32000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.04664179104477612,
+      "grad_norm": 1.517059087753296,
+      "learning_rate": 3.9878109452736323e-05,
+      "loss": 0.427,
+      "step": 100
+    },
+    {
+      "epoch": 0.04664179104477612,
+      "eval_loss": 0.030949920415878296,
+      "eval_runtime": 0.1617,
+      "eval_samples_per_second": 185.512,
+      "eval_steps_per_second": 24.735,
+      "step": 100
+    },
+    {
+      "epoch": 0.09328358208955224,
+      "grad_norm": 0.39070969820022583,
+      "learning_rate": 3.9753731343283585e-05,
+      "loss": 0.0403,
+      "step": 200
+    },
+    {
+      "epoch": 0.09328358208955224,
+      "eval_loss": 0.0011001491220667958,
+      "eval_runtime": 0.1623,
+      "eval_samples_per_second": 184.821,
+      "eval_steps_per_second": 24.643,
+      "step": 200
+    },
+    {
+      "epoch": 0.13992537313432835,
+      "grad_norm": 1.9688999652862549,
+      "learning_rate": 3.962935323383085e-05,
+      "loss": 0.0212,
+      "step": 300
+    },
+    {
+      "epoch": 0.13992537313432835,
+      "eval_loss": 0.0012941204477101564,
+      "eval_runtime": 0.1624,
+      "eval_samples_per_second": 184.715,
+      "eval_steps_per_second": 24.629,
+      "step": 300
+    },
+    {
+      "epoch": 0.1865671641791045,
+      "grad_norm": 0.4089069068431854,
+      "learning_rate": 3.950497512437811e-05,
+      "loss": 0.0121,
+      "step": 400
+    },
+    {
+      "epoch": 0.1865671641791045,
+      "eval_loss": 0.0006173434085212648,
+      "eval_runtime": 0.1662,
+      "eval_samples_per_second": 180.537,
+      "eval_steps_per_second": 24.072,
+      "step": 400
+    },
+    {
+      "epoch": 0.2332089552238806,
+      "grad_norm": 0.029543137177824974,
+      "learning_rate": 3.938059701492538e-05,
+      "loss": 0.0124,
+      "step": 500
+    },
+    {
+      "epoch": 0.2332089552238806,
+      "eval_loss": 0.0004312974342610687,
+      "eval_runtime": 0.1619,
+      "eval_samples_per_second": 185.304,
+      "eval_steps_per_second": 24.707,
+      "step": 500
+    },
+    {
+      "epoch": 0.2798507462686567,
+      "grad_norm": 0.03992057591676712,
+      "learning_rate": 3.925621890547264e-05,
+      "loss": 0.0078,
+      "step": 600
+    },
+    {
+      "epoch": 0.2798507462686567,
+      "eval_loss": 0.0006017092382535338,
+      "eval_runtime": 0.1668,
+      "eval_samples_per_second": 179.838,
+      "eval_steps_per_second": 23.978,
+      "step": 600
+    },
+    {
+      "epoch": 0.32649253731343286,
+      "grad_norm": 0.058661218732595444,
+      "learning_rate": 3.9131840796019907e-05,
+      "loss": 0.0052,
+      "step": 700
+    },
+    {
+      "epoch": 0.32649253731343286,
+      "eval_loss": 0.0009563259081915021,
+      "eval_runtime": 0.1598,
+      "eval_samples_per_second": 187.691,
+      "eval_steps_per_second": 25.025,
+      "step": 700
+    },
+    {
+      "epoch": 0.373134328358209,
+      "grad_norm": 0.01660950854420662,
+      "learning_rate": 3.900746268656717e-05,
+      "loss": 0.0073,
+      "step": 800
+    },
+    {
+      "epoch": 0.373134328358209,
+      "eval_loss": 0.002205046359449625,
+      "eval_runtime": 0.1627,
+      "eval_samples_per_second": 184.382,
+      "eval_steps_per_second": 24.584,
+      "step": 800
+    },
+    {
+      "epoch": 0.4197761194029851,
+      "grad_norm": 0.7414250373840332,
+      "learning_rate": 3.888308457711443e-05,
+      "loss": 0.0054,
+      "step": 900
+    },
+    {
+      "epoch": 0.4197761194029851,
+      "eval_loss": 0.0007016424206085503,
+      "eval_runtime": 0.1586,
+      "eval_samples_per_second": 189.134,
+      "eval_steps_per_second": 25.218,
+      "step": 900
+    },
+    {
+      "epoch": 0.4664179104477612,
+      "grad_norm": 0.02844785712659359,
+      "learning_rate": 3.875870646766169e-05,
+      "loss": 0.0038,
+      "step": 1000
+    },
+    {
+      "epoch": 0.4664179104477612,
+      "eval_loss": 2.6049870939459652e-05,
+      "eval_runtime": 0.1606,
+      "eval_samples_per_second": 186.806,
+      "eval_steps_per_second": 24.907,
+      "step": 1000
+    },
+    {
+      "epoch": 0.5130597014925373,
+      "grad_norm": 0.014462646096944809,
+      "learning_rate": 3.863432835820896e-05,
+      "loss": 0.0055,
+      "step": 1100
+    },
+    {
+      "epoch": 0.5130597014925373,
+      "eval_loss": 2.936022610811051e-05,
+      "eval_runtime": 0.1625,
+      "eval_samples_per_second": 184.588,
+      "eval_steps_per_second": 24.612,
+      "step": 1100
+    },
+    {
+      "epoch": 0.5597014925373134,
+      "grad_norm": 0.3922411799430847,
+      "learning_rate": 3.850995024875622e-05,
+      "loss": 0.0045,
+      "step": 1200
+    },
+    {
+      "epoch": 0.5597014925373134,
+      "eval_loss": 3.228507557651028e-05,
+      "eval_runtime": 0.16,
+      "eval_samples_per_second": 187.448,
+      "eval_steps_per_second": 24.993,
+      "step": 1200
+    },
+    {
+      "epoch": 0.6063432835820896,
+      "grad_norm": 0.09573910385370255,
+      "learning_rate": 3.838557213930348e-05,
+      "loss": 0.0039,
+      "step": 1300
+    },
+    {
+      "epoch": 0.6063432835820896,
+      "eval_loss": 3.271787500125356e-05,
+      "eval_runtime": 0.1628,
+      "eval_samples_per_second": 184.312,
+      "eval_steps_per_second": 24.575,
+      "step": 1300
+    },
+    {
+      "epoch": 0.6529850746268657,
+      "grad_norm": 0.07145769894123077,
+      "learning_rate": 3.8261194029850745e-05,
+      "loss": 0.0022,
+      "step": 1400
+    },
+    {
+      "epoch": 0.6529850746268657,
+      "eval_loss": 2.1901514628552832e-05,
+      "eval_runtime": 0.1652,
+      "eval_samples_per_second": 181.627,
+      "eval_steps_per_second": 24.217,
+      "step": 1400
+    },
+    {
+      "epoch": 0.6996268656716418,
+      "grad_norm": 0.0054560168646276,
+      "learning_rate": 3.813681592039801e-05,
+      "loss": 0.003,
+      "step": 1500
+    },
+    {
+      "epoch": 0.6996268656716418,
+      "eval_loss": 1.5640718629583716e-05,
+      "eval_runtime": 0.1629,
+      "eval_samples_per_second": 184.135,
+      "eval_steps_per_second": 24.551,
+      "step": 1500
+    },
+    {
+      "epoch": 0.746268656716418,
+      "grad_norm": 0.011433558538556099,
+      "learning_rate": 3.8012437810945275e-05,
+      "loss": 0.0028,
+      "step": 1600
+    },
+    {
+      "epoch": 0.746268656716418,
+      "eval_loss": 2.3460554075427353e-05,
+      "eval_runtime": 0.1627,
+      "eval_samples_per_second": 184.379,
+      "eval_steps_per_second": 24.584,
+      "step": 1600
+    },
+    {
+      "epoch": 0.792910447761194,
+      "grad_norm": 3.918342113494873,
+      "learning_rate": 3.788805970149254e-05,
+      "loss": 0.0024,
+      "step": 1700
+    },
+    {
+      "epoch": 0.792910447761194,
+      "eval_loss": 1.9451523257885128e-05,
+      "eval_runtime": 0.1594,
+      "eval_samples_per_second": 188.168,
+      "eval_steps_per_second": 25.089,
+      "step": 1700
+    },
+    {
+      "epoch": 0.8395522388059702,
+      "grad_norm": 0.07180186361074448,
+      "learning_rate": 3.7763681592039805e-05,
+      "loss": 0.0008,
+      "step": 1800
+    },
+    {
+      "epoch": 0.8395522388059702,
+      "eval_loss": 1.5172809071373194e-05,
+      "eval_runtime": 0.1619,
+      "eval_samples_per_second": 185.284,
+      "eval_steps_per_second": 24.705,
+      "step": 1800
+    },
+    {
+      "epoch": 0.8861940298507462,
+      "grad_norm": 0.05330910533666611,
+      "learning_rate": 3.7639303482587066e-05,
+      "loss": 0.0025,
+      "step": 1900
+    },
+    {
+      "epoch": 0.8861940298507462,
+      "eval_loss": 1.201580380438827e-05,
+      "eval_runtime": 0.162,
+      "eval_samples_per_second": 185.15,
+      "eval_steps_per_second": 24.687,
+      "step": 1900
+    },
+    {
+      "epoch": 0.9328358208955224,
+      "grad_norm": 0.05660516396164894,
+      "learning_rate": 3.751492537313433e-05,
+      "loss": 0.0019,
+      "step": 2000
+    },
+    {
+      "epoch": 0.9328358208955224,
+      "eval_loss": 1.2361979315755889e-05,
+      "eval_runtime": 0.1669,
+      "eval_samples_per_second": 179.773,
+      "eval_steps_per_second": 23.97,
+      "step": 2000
+    },
+    {
+      "epoch": 0.9794776119402985,
+      "grad_norm": 0.007570169400423765,
+      "learning_rate": 3.7390547263681596e-05,
+      "loss": 0.0025,
+      "step": 2100
+    },
+    {
+      "epoch": 0.9794776119402985,
+      "eval_loss": 1.094327490136493e-05,
+      "eval_runtime": 0.1611,
+      "eval_samples_per_second": 186.167,
+      "eval_steps_per_second": 24.822,
+      "step": 2100
+    },
+    {
+      "epoch": 1.0261194029850746,
+      "grad_norm": 0.28353750705718994,
+      "learning_rate": 3.726616915422886e-05,
+      "loss": 0.0011,
+      "step": 2200
+    },
+    {
+      "epoch": 1.0261194029850746,
+      "eval_loss": 1.550199340272229e-05,
+      "eval_runtime": 0.1612,
+      "eval_samples_per_second": 186.156,
+      "eval_steps_per_second": 24.821,
+      "step": 2200
+    },
+    {
+      "epoch": 1.0727611940298507,
+      "grad_norm": 0.09195005893707275,
+      "learning_rate": 3.7141791044776126e-05,
+      "loss": 0.0012,
+      "step": 2300
+    },
+    {
+      "epoch": 1.0727611940298507,
+      "eval_loss": 9.91811066342052e-06,
+      "eval_runtime": 0.1665,
+      "eval_samples_per_second": 180.181,
+      "eval_steps_per_second": 24.024,
+      "step": 2300
+    },
+    {
+      "epoch": 1.1194029850746268,
+      "grad_norm": 0.0037090331315994263,
+      "learning_rate": 3.701741293532339e-05,
+      "loss": 0.0012,
+      "step": 2400
+    },
+    {
+      "epoch": 1.1194029850746268,
+      "eval_loss": 1.1438472029112745e-05,
+      "eval_runtime": 0.1609,
+      "eval_samples_per_second": 186.44,
+      "eval_steps_per_second": 24.859,
+      "step": 2400
+    },
+    {
+      "epoch": 1.166044776119403,
+      "grad_norm": 0.023447873070836067,
+      "learning_rate": 3.689303482587065e-05,
+      "loss": 0.0015,
+      "step": 2500
+    },
+    {
+      "epoch": 1.166044776119403,
+      "eval_loss": 1.1437785360612907e-05,
+      "eval_runtime": 0.1647,
+      "eval_samples_per_second": 182.123,
+      "eval_steps_per_second": 24.283,
+      "step": 2500
+    },
+    {
+      "epoch": 1.212686567164179,
+      "grad_norm": 0.5180063843727112,
+      "learning_rate": 3.676865671641791e-05,
+      "loss": 0.0018,
+      "step": 2600
+    },
+    {
+      "epoch": 1.212686567164179,
+      "eval_loss": 2.4738228603382595e-05,
+      "eval_runtime": 0.1685,
+      "eval_samples_per_second": 178.021,
+      "eval_steps_per_second": 23.736,
+      "step": 2600
+    },
+    {
+      "epoch": 1.2593283582089552,
+      "grad_norm": 0.01776309497654438,
+      "learning_rate": 3.664427860696518e-05,
+      "loss": 0.0009,
+      "step": 2700
+    },
+    {
+      "epoch": 1.2593283582089552,
+      "eval_loss": 4.052605800097808e-05,
+      "eval_runtime": 0.1601,
+      "eval_samples_per_second": 187.399,
+      "eval_steps_per_second": 24.987,
+      "step": 2700
+    },
+    {
+      "epoch": 1.3059701492537314,
+      "grad_norm": 0.003942739684134722,
+      "learning_rate": 3.651990049751244e-05,
+      "loss": 0.0008,
+      "step": 2800
+    },
+    {
+      "epoch": 1.3059701492537314,
+      "eval_loss": 1.8365864889346994e-05,
+      "eval_runtime": 0.1602,
+      "eval_samples_per_second": 187.233,
+      "eval_steps_per_second": 24.964,
+      "step": 2800
+    },
+    {
+      "epoch": 1.3526119402985075,
+      "grad_norm": 0.02459954284131527,
+      "learning_rate": 3.63955223880597e-05,
+      "loss": 0.0012,
+      "step": 2900
+    },
+    {
+      "epoch": 1.3526119402985075,
+      "eval_loss": 1.1862516657856759e-05,
+      "eval_runtime": 0.1629,
+      "eval_samples_per_second": 184.179,
+      "eval_steps_per_second": 24.557,
+      "step": 2900
+    },
+    {
+      "epoch": 1.3992537313432836,
+      "grad_norm": 0.002258304273709655,
+      "learning_rate": 3.6271144278606964e-05,
+      "loss": 0.0013,
+      "step": 3000
+    },
+    {
+      "epoch": 1.3992537313432836,
+      "eval_loss": 8.7456610344816e-06,
+      "eval_runtime": 0.1649,
+      "eval_samples_per_second": 181.877,
+      "eval_steps_per_second": 24.25,
+      "step": 3000
+    },
+    {
+      "epoch": 1.4458955223880596,
+      "grad_norm": 0.0027893888764083385,
+      "learning_rate": 3.614676616915423e-05,
+      "loss": 0.001,
+      "step": 3100
+    },
+    {
+      "epoch": 1.4458955223880596,
+      "eval_loss": 1.060960585164139e-05,
+      "eval_runtime": 0.1717,
+      "eval_samples_per_second": 174.725,
+      "eval_steps_per_second": 23.297,
+      "step": 3100
+    },
+    {
+      "epoch": 1.4925373134328357,
+      "grad_norm": 0.19958974421024323,
+      "learning_rate": 3.6022388059701494e-05,
+      "loss": 0.0005,
+      "step": 3200
+    },
+    {
+      "epoch": 1.4925373134328357,
+      "eval_loss": 7.3110695666400716e-06,
+      "eval_runtime": 0.1703,
+      "eval_samples_per_second": 176.161,
+      "eval_steps_per_second": 23.488,
+      "step": 3200
+    },
+    {
+      "epoch": 1.539179104477612,
+      "grad_norm": 0.009552941657602787,
+      "learning_rate": 3.589800995024876e-05,
+      "loss": 0.0007,
+      "step": 3300
+    },
+    {
+      "epoch": 1.539179104477612,
+      "eval_loss": 7.919575182313565e-06,
+      "eval_runtime": 0.162,
+      "eval_samples_per_second": 185.238,
+      "eval_steps_per_second": 24.698,
+      "step": 3300
+    },
+    {
+      "epoch": 1.585820895522388,
+      "grad_norm": 0.0036544003523886204,
+      "learning_rate": 3.5773631840796024e-05,
+      "loss": 0.001,
+      "step": 3400
+    },
+    {
+      "epoch": 1.585820895522388,
+      "eval_loss": 9.348523235530593e-06,
+      "eval_runtime": 0.1653,
+      "eval_samples_per_second": 181.53,
+      "eval_steps_per_second": 24.204,
+      "step": 3400
+    },
+    {
+      "epoch": 1.6324626865671643,
+      "grad_norm": 0.013017863966524601,
+      "learning_rate": 3.5649253731343286e-05,
+      "loss": 0.0008,
+      "step": 3500
+    },
+    {
+      "epoch": 1.6324626865671643,
+      "eval_loss": 7.1847543949843384e-06,
+      "eval_runtime": 0.1602,
+      "eval_samples_per_second": 187.276,
+      "eval_steps_per_second": 24.97,
+      "step": 3500
+    },
+    {
+      "epoch": 1.6791044776119404,
+      "grad_norm": 0.0018999130697920918,
+      "learning_rate": 3.552487562189055e-05,
+      "loss": 0.0025,
+      "step": 3600
+    },
+    {
+      "epoch": 1.6791044776119404,
+      "eval_loss": 8.392209565499797e-06,
+      "eval_runtime": 0.1609,
+      "eval_samples_per_second": 186.495,
+      "eval_steps_per_second": 24.866,
+      "step": 3600
+    },
+    {
+      "epoch": 1.7257462686567164,
+      "grad_norm": 0.0020983312278985977,
+      "learning_rate": 3.5400497512437816e-05,
+      "loss": 0.0017,
+      "step": 3700
+    },
+    {
+      "epoch": 1.7257462686567164,
+      "eval_loss": 1.0126370398211293e-05,
+      "eval_runtime": 0.1624,
+      "eval_samples_per_second": 184.676,
+      "eval_steps_per_second": 24.623,
+      "step": 3700
+    },
+    {
+      "epoch": 1.7723880597014925,
+      "grad_norm": 0.026340099051594734,
+      "learning_rate": 3.527611940298508e-05,
+      "loss": 0.0005,
+      "step": 3800
+    },
+    {
+      "epoch": 1.7723880597014925,
+      "eval_loss": 1.1981122952420264e-05,
+      "eval_runtime": 0.1676,
+      "eval_samples_per_second": 179.033,
+      "eval_steps_per_second": 23.871,
+      "step": 3800
+    },
+    {
+      "epoch": 1.8190298507462686,
+      "grad_norm": 0.014922079630196095,
+      "learning_rate": 3.515174129353234e-05,
+      "loss": 0.0006,
+      "step": 3900
+    },
+    {
+      "epoch": 1.8190298507462686,
+      "eval_loss": 9.199016858474351e-06,
+      "eval_runtime": 0.1613,
+      "eval_samples_per_second": 185.942,
+      "eval_steps_per_second": 24.792,
+      "step": 3900
+    },
+    {
+      "epoch": 1.8656716417910446,
+      "grad_norm": 0.27961060404777527,
+      "learning_rate": 3.50273631840796e-05,
+      "loss": 0.0007,
+      "step": 4000
+    },
+    {
+      "epoch": 1.8656716417910446,
+      "eval_loss": 4.940530743624549e-06,
+      "eval_runtime": 0.1625,
+      "eval_samples_per_second": 184.614,
+      "eval_steps_per_second": 24.615,
+      "step": 4000
+    },
+    {
+      "epoch": 1.912313432835821,
+      "grad_norm": 0.017754705622792244,
+      "learning_rate": 3.490298507462687e-05,
+      "loss": 0.0007,
+      "step": 4100
+    },
+    {
+      "epoch": 1.912313432835821,
+      "eval_loss": 5.068028713139938e-06,
+      "eval_runtime": 0.16,
+      "eval_samples_per_second": 187.528,
+      "eval_steps_per_second": 25.004,
+      "step": 4100
+    },
+    {
+      "epoch": 1.9589552238805972,
+      "grad_norm": 0.27114221453666687,
+      "learning_rate": 3.477860696517413e-05,
+      "loss": 0.0004,
+      "step": 4200
+    },
+    {
+      "epoch": 1.9589552238805972,
+      "eval_loss": 4.581762823363533e-06,
+      "eval_runtime": 0.1629,
+      "eval_samples_per_second": 184.179,
+      "eval_steps_per_second": 24.557,
+      "step": 4200
+    },
+    {
+      "epoch": 2.0055970149253732,
+      "grad_norm": 0.012334506027400494,
+      "learning_rate": 3.46542288557214e-05,
+      "loss": 0.002,
+      "step": 4300
+    },
+    {
+      "epoch": 2.0055970149253732,
+      "eval_loss": 6.550972102559172e-06,
+      "eval_runtime": 0.1623,
+      "eval_samples_per_second": 184.884,
+      "eval_steps_per_second": 24.651,
+      "step": 4300
+    },
+    {
+      "epoch": 2.0522388059701493,
+      "grad_norm": 0.06736209243535995,
+      "learning_rate": 3.452985074626866e-05,
+      "loss": 0.0004,
+      "step": 4400
+    },
+    {
+      "epoch": 2.0522388059701493,
+      "eval_loss": 5.742116627516225e-06,
+      "eval_runtime": 0.1636,
+      "eval_samples_per_second": 183.421,
+      "eval_steps_per_second": 24.456,
+      "step": 4400
+    },
+    {
+      "epoch": 2.0988805970149254,
+      "grad_norm": 0.006035651080310345,
+      "learning_rate": 3.440547263681592e-05,
+      "loss": 0.0014,
+      "step": 4500
+    },
+    {
+      "epoch": 2.0988805970149254,
+      "eval_loss": 7.217912752821576e-06,
+      "eval_runtime": 0.1633,
+      "eval_samples_per_second": 183.679,
+      "eval_steps_per_second": 24.491,
+      "step": 4500
+    },
+    {
+      "epoch": 2.1455223880597014,
+      "grad_norm": 0.0010893335565924644,
+      "learning_rate": 3.4281094527363184e-05,
+      "loss": 0.0003,
+      "step": 4600
+    },
+    {
+      "epoch": 2.1455223880597014,
+      "eval_loss": 5.789845090475865e-06,
+      "eval_runtime": 0.1605,
+      "eval_samples_per_second": 186.889,
+      "eval_steps_per_second": 24.919,
+      "step": 4600
+    },
+    {
+      "epoch": 2.1921641791044775,
+      "grad_norm": 0.0023283003829419613,
+      "learning_rate": 3.415671641791045e-05,
+      "loss": 0.0003,
+      "step": 4700
+    },
+    {
+      "epoch": 2.1921641791044775,
+      "eval_loss": 4.947921297571156e-06,
+      "eval_runtime": 0.1602,
+      "eval_samples_per_second": 187.294,
+      "eval_steps_per_second": 24.973,
+      "step": 4700
+    },
+    {
+      "epoch": 2.2388059701492535,
+      "grad_norm": 0.01192499604076147,
+      "learning_rate": 3.4032338308457714e-05,
+      "loss": 0.0007,
+      "step": 4800
+    },
+    {
+      "epoch": 2.2388059701492535,
+      "eval_loss": 5.415991836343892e-06,
+      "eval_runtime": 0.1607,
+      "eval_samples_per_second": 186.643,
+      "eval_steps_per_second": 24.886,
+      "step": 4800
+    },
+    {
+      "epoch": 2.28544776119403,
+      "grad_norm": 0.001801644335500896,
+      "learning_rate": 3.390796019900498e-05,
+      "loss": 0.0003,
+      "step": 4900
+    },
+    {
+      "epoch": 2.28544776119403,
+      "eval_loss": 4.487563728616806e-06,
+      "eval_runtime": 0.1615,
+      "eval_samples_per_second": 185.729,
+      "eval_steps_per_second": 24.764,
+      "step": 4900
+    },
+    {
+      "epoch": 2.332089552238806,
+      "grad_norm": 0.0010308738565072417,
+      "learning_rate": 3.3783582089552244e-05,
+      "loss": 0.0009,
+      "step": 5000
+    },
+    {
+      "epoch": 2.332089552238806,
+      "eval_loss": 6.067215053917607e-06,
+      "eval_runtime": 0.1628,
+      "eval_samples_per_second": 184.262,
+      "eval_steps_per_second": 24.568,
+      "step": 5000
+    },
+    {
+      "epoch": 2.378731343283582,
+      "grad_norm": 0.0012472744565457106,
+      "learning_rate": 3.3659203980099505e-05,
+      "loss": 0.0003,
+      "step": 5100
+    },
+    {
+      "epoch": 2.378731343283582,
+      "eval_loss": 1.881633943412453e-05,
+      "eval_runtime": 0.1606,
+      "eval_samples_per_second": 186.836,
+      "eval_steps_per_second": 24.911,
+      "step": 5100
+    },
+    {
+      "epoch": 2.425373134328358,
+      "grad_norm": 0.7578136920928955,
+      "learning_rate": 3.353482587064677e-05,
+      "loss": 0.0009,
+      "step": 5200
+    },
+    {
+      "epoch": 2.425373134328358,
+      "eval_loss": 1.3828990631736815e-05,
+      "eval_runtime": 0.1631,
+      "eval_samples_per_second": 183.952,
+      "eval_steps_per_second": 24.527,
+      "step": 5200
+    },
+    {
+      "epoch": 2.4720149253731343,
+      "grad_norm": 0.1774366796016693,
+      "learning_rate": 3.3410447761194035e-05,
+      "loss": 0.0003,
+      "step": 5300
+    },
+    {
+      "epoch": 2.4720149253731343,
+      "eval_loss": 9.651944310462568e-06,
+      "eval_runtime": 0.1623,
+      "eval_samples_per_second": 184.83,
+      "eval_steps_per_second": 24.644,
+      "step": 5300
+    },
+    {
+      "epoch": 2.5186567164179103,
+      "grad_norm": 0.020136339589953423,
+      "learning_rate": 3.32860696517413e-05,
+      "loss": 0.0006,
+      "step": 5400
+    },
+    {
+      "epoch": 2.5186567164179103,
+      "eval_loss": 9.613087968318723e-06,
+      "eval_runtime": 0.1609,
+      "eval_samples_per_second": 186.441,
+      "eval_steps_per_second": 24.859,
+      "step": 5400
+    },
+    {
+      "epoch": 2.5652985074626864,
+      "grad_norm": 0.0021144032943993807,
+      "learning_rate": 3.316169154228856e-05,
+      "loss": 0.0002,
+      "step": 5500
+    },
+    {
+      "epoch": 2.5652985074626864,
+      "eval_loss": 6.973440122237662e-06,
+      "eval_runtime": 0.1613,
+      "eval_samples_per_second": 186.02,
+      "eval_steps_per_second": 24.803,
+      "step": 5500
+    },
+    {
+      "epoch": 2.611940298507463,
+      "grad_norm": 0.0018188412068411708,
+      "learning_rate": 3.303731343283582e-05,
+      "loss": 0.0002,
+      "step": 5600
+    },
+    {
+      "epoch": 2.611940298507463,
+      "eval_loss": 5.028930445405422e-06,
+      "eval_runtime": 0.1636,
+      "eval_samples_per_second": 183.338,
+      "eval_steps_per_second": 24.445,
+      "step": 5600
+    },
+    {
+      "epoch": 2.658582089552239,
+      "grad_norm": 0.0019931041169911623,
+      "learning_rate": 3.291293532338309e-05,
+      "loss": 0.0005,
+      "step": 5700
+    },
+    {
+      "epoch": 2.658582089552239,
+      "eval_loss": 4.870566499448614e-06,
+      "eval_runtime": 0.1657,
+      "eval_samples_per_second": 180.997,
+      "eval_steps_per_second": 24.133,
+      "step": 5700
+    },
+    {
+      "epoch": 2.705223880597015,
+      "grad_norm": 0.0018326346762478352,
+      "learning_rate": 3.278855721393035e-05,
+      "loss": 0.0002,
+      "step": 5800
+    },
+    {
+      "epoch": 2.705223880597015,
+      "eval_loss": 3.6711526263388805e-06,
+      "eval_runtime": 0.1741,
+      "eval_samples_per_second": 172.304,
+      "eval_steps_per_second": 22.974,
+      "step": 5800
+    },
+    {
+      "epoch": 2.751865671641791,
+      "grad_norm": 0.003958791960030794,
+      "learning_rate": 3.266417910447762e-05,
+      "loss": 0.0008,
+      "step": 5900
+    },
+    {
+      "epoch": 2.751865671641791,
+      "eval_loss": 8.262709343398456e-06,
+      "eval_runtime": 0.1621,
+      "eval_samples_per_second": 185.052,
+      "eval_steps_per_second": 24.674,
+      "step": 5900
+    },
+    {
+      "epoch": 2.798507462686567,
+      "grad_norm": 0.001310827792622149,
+      "learning_rate": 3.253980099502488e-05,
+      "loss": 0.0002,
+      "step": 6000
+    },
+    {
+      "epoch": 2.798507462686567,
+      "eval_loss": 5.826313099532854e-06,
+      "eval_runtime": 0.1597,
+      "eval_samples_per_second": 187.8,
+      "eval_steps_per_second": 25.04,
+      "step": 6000
+    },
+    {
+      "epoch": 2.845149253731343,
+      "grad_norm": 0.0018846031744033098,
+      "learning_rate": 3.241542288557214e-05,
+      "loss": 0.0006,
+      "step": 6100
+    },
+    {
+      "epoch": 2.845149253731343,
+      "eval_loss": 1.0719989404606167e-05,
+      "eval_runtime": 0.1635,
+      "eval_samples_per_second": 183.451,
+      "eval_steps_per_second": 24.46,
+      "step": 6100
+    },
+    {
+      "epoch": 2.8917910447761193,
+      "grad_norm": 0.0027365258429199457,
+      "learning_rate": 3.22910447761194e-05,
+      "loss": 0.0003,
+      "step": 6200
+    },
+    {
+      "epoch": 2.8917910447761193,
+      "eval_loss": 4.158847787039122e-06,
+      "eval_runtime": 0.1603,
+      "eval_samples_per_second": 187.178,
+      "eval_steps_per_second": 24.957,
+      "step": 6200
+    },
+    {
+      "epoch": 2.9384328358208958,
+      "grad_norm": 0.003453996032476425,
+      "learning_rate": 3.216666666666667e-05,
+      "loss": 0.001,
+      "step": 6300
+    },
+    {
+      "epoch": 2.9384328358208958,
+      "eval_loss": 1.029476788971806e-05,
+      "eval_runtime": 0.1606,
+      "eval_samples_per_second": 186.748,
+      "eval_steps_per_second": 24.9,
+      "step": 6300
+    },
+    {
+      "epoch": 2.9850746268656714,
+      "grad_norm": 0.036285314708948135,
+      "learning_rate": 3.204228855721393e-05,
+      "loss": 0.0006,
+      "step": 6400
+    },
+    {
+      "epoch": 2.9850746268656714,
+      "eval_loss": 5.606885224551661e-06,
+      "eval_runtime": 0.16,
+      "eval_samples_per_second": 187.553,
+      "eval_steps_per_second": 25.007,
+      "step": 6400
+    },
+    {
+      "epoch": 3.031716417910448,
+      "grad_norm": 0.46069106459617615,
+      "learning_rate": 3.1917910447761195e-05,
+      "loss": 0.0014,
+      "step": 6500
+    },
+    {
+      "epoch": 3.031716417910448,
+      "eval_loss": 6.717081760143628e-06,
+      "eval_runtime": 0.1597,
+      "eval_samples_per_second": 187.883,
+      "eval_steps_per_second": 25.051,
+      "step": 6500
+    },
+    {
+      "epoch": 3.078358208955224,
+      "grad_norm": 0.0014442165847867727,
+      "learning_rate": 3.1793532338308456e-05,
+      "loss": 0.0004,
+      "step": 6600
+    },
+    {
+      "epoch": 3.078358208955224,
+      "eval_loss": 2.4730301447561942e-05,
+      "eval_runtime": 0.1633,
+      "eval_samples_per_second": 183.726,
+      "eval_steps_per_second": 24.497,
+      "step": 6600
+    },
+    {
+      "epoch": 3.125,
+      "grad_norm": 0.0008062911801971495,
+      "learning_rate": 3.1670398009950254e-05,
+      "loss": 0.0013,
+      "step": 6700
+    },
+    {
+      "epoch": 3.125,
+      "eval_loss": 3.047320433324785e-06,
+      "eval_runtime": 0.1643,
+      "eval_samples_per_second": 182.624,
+      "eval_steps_per_second": 24.35,
+      "step": 6700
+    },
+    {
+      "epoch": 3.171641791044776,
+      "grad_norm": 0.0008228031219914556,
+      "learning_rate": 3.1546019900497516e-05,
+      "loss": 0.0007,
+      "step": 6800
+    },
+    {
+      "epoch": 3.171641791044776,
+      "eval_loss": 3.1468273391510593e-06,
+      "eval_runtime": 0.1684,
+      "eval_samples_per_second": 178.135,
+      "eval_steps_per_second": 23.751,
+      "step": 6800
+    },
+    {
+      "epoch": 3.218283582089552,
+      "grad_norm": 0.006291732657700777,
+      "learning_rate": 3.142164179104478e-05,
+      "loss": 0.0006,
+      "step": 6900
+    },
+    {
+      "epoch": 3.218283582089552,
+      "eval_loss": 3.600493528210791e-06,
+      "eval_runtime": 0.1637,
+      "eval_samples_per_second": 183.264,
+      "eval_steps_per_second": 24.435,
+      "step": 6900
+    },
+    {
+      "epoch": 3.264925373134328,
+      "grad_norm": 0.007644134573638439,
+      "learning_rate": 3.129726368159204e-05,
+      "loss": 0.0004,
+      "step": 7000
+    },
+    {
+      "epoch": 3.264925373134328,
+      "eval_loss": 3.303811354271602e-06,
+      "eval_runtime": 0.1708,
+      "eval_samples_per_second": 175.635,
+      "eval_steps_per_second": 23.418,
+      "step": 7000
+    },
+    {
+      "epoch": 3.3115671641791042,
+      "grad_norm": 0.005915221758186817,
+      "learning_rate": 3.117288557213931e-05,
+      "loss": 0.0005,
+      "step": 7100
+    },
+    {
+      "epoch": 3.3115671641791042,
+      "eval_loss": 2.879673274946981e-06,
+      "eval_runtime": 0.1654,
+      "eval_samples_per_second": 181.433,
+      "eval_steps_per_second": 24.191,
+      "step": 7100
+    },
+    {
+      "epoch": 3.3582089552238807,
+      "grad_norm": 0.0019528436241671443,
+      "learning_rate": 3.104850746268657e-05,
+      "loss": 0.0006,
+      "step": 7200
+    },
+    {
+      "epoch": 3.3582089552238807,
+      "eval_loss": 3.403750042707543e-06,
+      "eval_runtime": 0.1608,
+      "eval_samples_per_second": 186.522,
+      "eval_steps_per_second": 24.87,
+      "step": 7200
+    },
+    {
+      "epoch": 3.404850746268657,
+      "grad_norm": 0.004577342886477709,
+      "learning_rate": 3.092412935323384e-05,
+      "loss": 0.0003,
+      "step": 7300
+    },
+    {
+      "epoch": 3.404850746268657,
+      "eval_loss": 6.347925591398962e-06,
+      "eval_runtime": 0.1612,
+      "eval_samples_per_second": 186.155,
+      "eval_steps_per_second": 24.821,
+      "step": 7300
+    },
+    {
+      "epoch": 3.451492537313433,
+      "grad_norm": 0.0017618165584281087,
+      "learning_rate": 3.079975124378109e-05,
+      "loss": 0.0003,
+      "step": 7400
+    },
+    {
+      "epoch": 3.451492537313433,
+      "eval_loss": 4.75058095616987e-06,
+      "eval_runtime": 0.1761,
+      "eval_samples_per_second": 170.387,
+      "eval_steps_per_second": 22.718,
+      "step": 7400
+    },
+    {
+      "epoch": 3.498134328358209,
+      "grad_norm": 0.29626068472862244,
+      "learning_rate": 3.067537313432836e-05,
+      "loss": 0.0002,
+      "step": 7500
+    },
+    {
+      "epoch": 3.498134328358209,
+      "eval_loss": 3.206900828445214e-06,
+      "eval_runtime": 0.1606,
+      "eval_samples_per_second": 186.83,
+      "eval_steps_per_second": 24.911,
+      "step": 7500
+    },
+    {
+      "epoch": 3.544776119402985,
+      "grad_norm": 0.0010133940959349275,
+      "learning_rate": 3.055099502487562e-05,
+      "loss": 0.0002,
+      "step": 7600
+    },
+    {
+      "epoch": 3.544776119402985,
+      "eval_loss": 5.057888301962521e-06,
+      "eval_runtime": 0.1667,
+      "eval_samples_per_second": 179.969,
+      "eval_steps_per_second": 23.996,
+      "step": 7600
+    },
+    {
+      "epoch": 3.591417910447761,
+      "grad_norm": 0.001219356432557106,
+      "learning_rate": 3.0426616915422887e-05,
+      "loss": 0.0006,
+      "step": 7700
+    },
+    {
+      "epoch": 3.591417910447761,
+      "eval_loss": 4.3034724512835965e-06,
+      "eval_runtime": 0.1589,
+      "eval_samples_per_second": 188.793,
+      "eval_steps_per_second": 25.172,
+      "step": 7700
+    },
+    {
+      "epoch": 3.638059701492537,
+      "grad_norm": 0.004916150122880936,
+      "learning_rate": 3.030223880597015e-05,
+      "loss": 0.0005,
+      "step": 7800
+    },
+    {
+      "epoch": 3.638059701492537,
+      "eval_loss": 7.532180006819544e-06,
+      "eval_runtime": 0.1593,
+      "eval_samples_per_second": 188.302,
+      "eval_steps_per_second": 25.107,
+      "step": 7800
+    },
+    {
+      "epoch": 3.6847014925373136,
+      "grad_norm": 0.0009036241099238396,
+      "learning_rate": 3.0177860696517417e-05,
+      "loss": 0.0002,
+      "step": 7900
+    },
+    {
+      "epoch": 3.6847014925373136,
+      "eval_loss": 7.954057764436584e-06,
+      "eval_runtime": 0.1617,
+      "eval_samples_per_second": 185.49,
+      "eval_steps_per_second": 24.732,
+      "step": 7900
+    },
+    {
+      "epoch": 3.7313432835820897,
+      "grad_norm": 0.0004865360097028315,
+      "learning_rate": 3.005348258706468e-05,
+      "loss": 0.0002,
+      "step": 8000
+    },
+    {
+      "epoch": 3.7313432835820897,
+      "eval_loss": 5.304416390572442e-06,
+      "eval_runtime": 0.1601,
+      "eval_samples_per_second": 187.374,
+      "eval_steps_per_second": 24.983,
+      "step": 8000
+    },
+    {
+      "epoch": 3.7779850746268657,
+      "grad_norm": 0.0006289943703450263,
+      "learning_rate": 2.9929104477611944e-05,
+      "loss": 0.001,
+      "step": 8100
+    },
+    {
+      "epoch": 3.7779850746268657,
+      "eval_loss": 6.098490302974824e-06,
+      "eval_runtime": 0.1602,
+      "eval_samples_per_second": 187.228,
+      "eval_steps_per_second": 24.964,
+      "step": 8100
+    },
+    {
+      "epoch": 3.824626865671642,
+      "grad_norm": 0.0017220574663951993,
+      "learning_rate": 2.9804726368159205e-05,
+      "loss": 0.0003,
+      "step": 8200
+    },
+    {
+      "epoch": 3.824626865671642,
+      "eval_loss": 6.202932127052918e-06,
+      "eval_runtime": 0.1592,
+      "eval_samples_per_second": 188.401,
+      "eval_steps_per_second": 25.12,
+      "step": 8200
+    },
+    {
+      "epoch": 3.871268656716418,
+      "grad_norm": 0.10363404452800751,
+      "learning_rate": 2.968034825870647e-05,
+      "loss": 0.0004,
+      "step": 8300
+    },
+    {
+      "epoch": 3.871268656716418,
+      "eval_loss": 3.6653259485319722e-06,
+      "eval_runtime": 0.162,
+      "eval_samples_per_second": 185.235,
+      "eval_steps_per_second": 24.698,
+      "step": 8300
+    },
+    {
+      "epoch": 3.917910447761194,
+      "grad_norm": 0.001283775782212615,
+      "learning_rate": 2.9555970149253732e-05,
+      "loss": 0.0003,
+      "step": 8400
+    },
+    {
+      "epoch": 3.917910447761194,
+      "eval_loss": 3.4666393275983864e-06,
+      "eval_runtime": 0.1624,
+      "eval_samples_per_second": 184.777,
+      "eval_steps_per_second": 24.637,
+      "step": 8400
+    },
+    {
+      "epoch": 3.96455223880597,
+      "grad_norm": 0.0013221842236816883,
+      "learning_rate": 2.9431592039800997e-05,
+      "loss": 0.0001,
+      "step": 8500
+    },
+    {
+      "epoch": 3.96455223880597,
+      "eval_loss": 2.620423856569687e-06,
+      "eval_runtime": 0.1605,
+      "eval_samples_per_second": 186.903,
+      "eval_steps_per_second": 24.92,
+      "step": 8500
+    },
+    {
+      "epoch": 4.0111940298507465,
+      "grad_norm": 0.0007219673716463149,
+      "learning_rate": 2.930721393034826e-05,
+      "loss": 0.0004,
+      "step": 8600
+    },
+    {
+      "epoch": 4.0111940298507465,
+      "eval_loss": 2.7292528557154583e-06,
+      "eval_runtime": 0.1831,
+      "eval_samples_per_second": 163.862,
+      "eval_steps_per_second": 21.848,
+      "step": 8600
+    },
+    {
+      "epoch": 4.057835820895522,
+      "grad_norm": 0.13019587099552155,
+      "learning_rate": 2.9182835820895527e-05,
+      "loss": 0.0002,
+      "step": 8700
+    },
+    {
+      "epoch": 4.057835820895522,
+      "eval_loss": 2.8391550586093217e-06,
+      "eval_runtime": 0.1605,
+      "eval_samples_per_second": 186.893,
+      "eval_steps_per_second": 24.919,
+      "step": 8700
+    },
+    {
+      "epoch": 4.104477611940299,
+      "grad_norm": 0.0006854361272417009,
+      "learning_rate": 2.9058457711442788e-05,
+      "loss": 0.0002,
+      "step": 8800
+    },
+    {
+      "epoch": 4.104477611940299,
+      "eval_loss": 2.3670136215514503e-06,
+      "eval_runtime": 0.1616,
+      "eval_samples_per_second": 185.674,
+      "eval_steps_per_second": 24.757,
+      "step": 8800
+    },
+    {
+      "epoch": 4.151119402985074,
+      "grad_norm": 0.0259992815554142,
+      "learning_rate": 2.8934079601990053e-05,
+      "loss": 0.0001,
+      "step": 8900
+    },
+    {
+      "epoch": 4.151119402985074,
+      "eval_loss": 1.925168362504337e-06,
+      "eval_runtime": 0.1602,
+      "eval_samples_per_second": 187.21,
+      "eval_steps_per_second": 24.961,
+      "step": 8900
+    },
+    {
+      "epoch": 4.197761194029851,
+      "grad_norm": 0.002843194641172886,
+      "learning_rate": 2.8809701492537315e-05,
+      "loss": 0.0006,
+      "step": 9000
+    },
+    {
+      "epoch": 4.197761194029851,
+      "eval_loss": 1.060922113538254e-05,
+      "eval_runtime": 0.1591,
+      "eval_samples_per_second": 188.552,
+      "eval_steps_per_second": 25.14,
+      "step": 9000
+    },
+    {
+      "epoch": 4.244402985074627,
+      "grad_norm": 0.001723111025057733,
+      "learning_rate": 2.868532338308458e-05,
+      "loss": 0.0004,
+      "step": 9100
+    },
+    {
+      "epoch": 4.244402985074627,
+      "eval_loss": 6.985505024204031e-06,
+      "eval_runtime": 0.1598,
+      "eval_samples_per_second": 187.765,
+      "eval_steps_per_second": 25.035,
+      "step": 9100
+    },
+    {
+      "epoch": 4.291044776119403,
+      "grad_norm": 0.0005880256649106741,
+      "learning_rate": 2.856094527363184e-05,
+      "loss": 0.0003,
+      "step": 9200
+    },
+    {
+      "epoch": 4.291044776119403,
+      "eval_loss": 3.5329176171217114e-06,
+      "eval_runtime": 0.1632,
+      "eval_samples_per_second": 183.85,
+      "eval_steps_per_second": 24.513,
+      "step": 9200
+    },
+    {
+      "epoch": 4.337686567164179,
+      "grad_norm": 0.0004977516946382821,
+      "learning_rate": 2.8436567164179106e-05,
+      "loss": 0.0004,
+      "step": 9300
+    },
+    {
+      "epoch": 4.337686567164179,
+      "eval_loss": 4.645138687919825e-06,
+      "eval_runtime": 0.1619,
+      "eval_samples_per_second": 185.244,
+      "eval_steps_per_second": 24.699,
+      "step": 9300
+    },
+    {
+      "epoch": 4.384328358208955,
+      "grad_norm": 0.0014300497714430094,
+      "learning_rate": 2.8312189054726368e-05,
+      "loss": 0.0002,
+      "step": 9400
+    },
+    {
+      "epoch": 4.384328358208955,
+      "eval_loss": 3.881290922436165e-06,
+      "eval_runtime": 0.1601,
+      "eval_samples_per_second": 187.344,
+      "eval_steps_per_second": 24.979,
+      "step": 9400
+    },
+    {
+      "epoch": 4.4309701492537314,
+      "grad_norm": 0.0009422674193046987,
+      "learning_rate": 2.8187810945273636e-05,
+      "loss": 0.0003,
+      "step": 9500
+    },
+    {
+      "epoch": 4.4309701492537314,
+      "eval_loss": 2.567733190517174e-06,
+      "eval_runtime": 0.1599,
+      "eval_samples_per_second": 187.663,
+      "eval_steps_per_second": 25.022,
+      "step": 9500
+    },
+    {
+      "epoch": 4.477611940298507,
+      "grad_norm": 0.0015334930503740907,
+      "learning_rate": 2.8063432835820895e-05,
+      "loss": 0.0004,
+      "step": 9600
+    },
+    {
+      "epoch": 4.477611940298507,
+      "eval_loss": 4.461974185687723e-06,
+      "eval_runtime": 0.1614,
+      "eval_samples_per_second": 185.881,
+      "eval_steps_per_second": 24.784,
+      "step": 9600
+    },
+    {
+      "epoch": 4.524253731343284,
+      "grad_norm": 0.002924109809100628,
+      "learning_rate": 2.7939054726368163e-05,
+      "loss": 0.0009,
+      "step": 9700
+    },
+    {
+      "epoch": 4.524253731343284,
+      "eval_loss": 1.8840278244169895e-06,
+      "eval_runtime": 0.1609,
+      "eval_samples_per_second": 186.403,
+      "eval_steps_per_second": 24.854,
+      "step": 9700
+    },
+    {
+      "epoch": 4.57089552238806,
+      "grad_norm": 0.0012187871616333723,
+      "learning_rate": 2.7814676616915425e-05,
+      "loss": 0.0001,
+      "step": 9800
+    },
+    {
+      "epoch": 4.57089552238806,
+      "eval_loss": 1.5747774568808381e-06,
+      "eval_runtime": 0.1691,
+      "eval_samples_per_second": 177.369,
+      "eval_steps_per_second": 23.649,
+      "step": 9800
+    },
+    {
+      "epoch": 4.617537313432836,
+      "grad_norm": 0.025326132774353027,
+      "learning_rate": 2.769029850746269e-05,
+      "loss": 0.0005,
+      "step": 9900
+    },
+    {
+      "epoch": 4.617537313432836,
+      "eval_loss": 1.4370193639479112e-06,
+      "eval_runtime": 0.1606,
+      "eval_samples_per_second": 186.746,
+      "eval_steps_per_second": 24.899,
+      "step": 9900
+    },
+    {
+      "epoch": 4.664179104477612,
+      "grad_norm": 0.0006336395163089037,
+      "learning_rate": 2.756592039800995e-05,
+      "loss": 0.0002,
+      "step": 10000
+    },
+    {
+      "epoch": 4.664179104477612,
+      "eval_loss": 1.4652428035333287e-06,
+      "eval_runtime": 0.1613,
+      "eval_samples_per_second": 185.994,
+      "eval_steps_per_second": 24.799,
+      "step": 10000
+    },
+    {
+      "epoch": 4.710820895522388,
+      "grad_norm": 0.0023961891420185566,
+      "learning_rate": 2.7441542288557216e-05,
+      "loss": 0.0004,
+      "step": 10100
+    },
+    {
+      "epoch": 4.710820895522388,
+      "eval_loss": 1.3807976984026027e-06,
+      "eval_runtime": 0.1606,
+      "eval_samples_per_second": 186.745,
+      "eval_steps_per_second": 24.899,
+      "step": 10100
+    },
+    {
+      "epoch": 4.757462686567164,
+      "grad_norm": 0.0004522838571574539,
+      "learning_rate": 2.7317164179104478e-05,
+      "loss": 0.0001,
+      "step": 10200
+    },
+    {
+      "epoch": 4.757462686567164,
+      "eval_loss": 1.4655278164354968e-06,
+      "eval_runtime": 0.1607,
+      "eval_samples_per_second": 186.679,
+      "eval_steps_per_second": 24.891,
+      "step": 10200
+    },
+    {
+      "epoch": 4.80410447761194,
+      "grad_norm": 0.0003487040812615305,
+      "learning_rate": 2.7192786069651743e-05,
+      "loss": 0.0001,
+      "step": 10300
+    },
+    {
+      "epoch": 4.80410447761194,
+      "eval_loss": 1.6429443121523946e-06,
+      "eval_runtime": 0.1606,
+      "eval_samples_per_second": 186.758,
+      "eval_steps_per_second": 24.901,
+      "step": 10300
+    },
+    {
+      "epoch": 4.850746268656716,
+      "grad_norm": 0.01166547555476427,
+      "learning_rate": 2.7068407960199004e-05,
+      "loss": 0.0007,
+      "step": 10400
+    },
+    {
+      "epoch": 4.850746268656716,
+      "eval_loss": 2.425857928756159e-06,
+      "eval_runtime": 0.1599,
+      "eval_samples_per_second": 187.564,
+      "eval_steps_per_second": 25.009,
+      "step": 10400
+    },
+    {
+      "epoch": 4.897388059701493,
+      "grad_norm": 0.0007575357449240983,
+      "learning_rate": 2.6944029850746273e-05,
+      "loss": 0.0001,
+      "step": 10500
+    },
+    {
+      "epoch": 4.897388059701493,
+      "eval_loss": 2.3037405298964586e-06,
+      "eval_runtime": 0.1599,
+      "eval_samples_per_second": 187.656,
+      "eval_steps_per_second": 25.021,
+      "step": 10500
+    },
+    {
+      "epoch": 4.9440298507462686,
+      "grad_norm": 0.0067961206659674644,
+      "learning_rate": 2.6819651741293534e-05,
+      "loss": 0.0003,
+      "step": 10600
+    },
+    {
+      "epoch": 4.9440298507462686,
+      "eval_loss": 2.1566411305684596e-06,
+      "eval_runtime": 0.1671,
+      "eval_samples_per_second": 179.556,
+      "eval_steps_per_second": 23.941,
+      "step": 10600
+    },
+    {
+      "epoch": 4.990671641791045,
+      "grad_norm": 0.0007165633141994476,
+      "learning_rate": 2.6696517412935325e-05,
+      "loss": 0.0002,
+      "step": 10700
+    },
+    {
+      "epoch": 4.990671641791045,
+      "eval_loss": 2.0811207832593936e-06,
+      "eval_runtime": 0.1599,
+      "eval_samples_per_second": 187.637,
+      "eval_steps_per_second": 25.018,
+      "step": 10700
+    },
+    {
+      "epoch": 5.037313432835821,
+      "grad_norm": 0.000998039380647242,
+      "learning_rate": 2.657338308457712e-05,
+      "loss": 0.0008,
+      "step": 10800
+    },
+    {
+      "epoch": 5.037313432835821,
+      "eval_loss": 1.8726080952546909e-06,
+      "eval_runtime": 0.1624,
+      "eval_samples_per_second": 184.7,
+      "eval_steps_per_second": 24.627,
+      "step": 10800
+    },
+    {
+      "epoch": 5.083955223880597,
+      "grad_norm": 0.0009990198304876685,
+      "learning_rate": 2.6449004975124378e-05,
+      "loss": 0.0001,
+      "step": 10900
+    },
+    {
+      "epoch": 5.083955223880597,
+      "eval_loss": 3.987378022429766e-06,
+      "eval_runtime": 0.1606,
+      "eval_samples_per_second": 186.835,
+      "eval_steps_per_second": 24.911,
+      "step": 10900
+    },
+    {
+      "epoch": 5.130597014925373,
+      "grad_norm": 0.0006653439486399293,
+      "learning_rate": 2.6324626865671646e-05,
+      "loss": 0.0001,
+      "step": 11000
+    },
+    {
+      "epoch": 5.130597014925373,
+      "eval_loss": 2.8886024665553123e-06,
+      "eval_runtime": 0.1597,
+      "eval_samples_per_second": 187.847,
+      "eval_steps_per_second": 25.046,
+      "step": 11000
+    },
+    {
+      "epoch": 5.177238805970149,
+      "grad_norm": 0.0025863787159323692,
+      "learning_rate": 2.6200248756218908e-05,
+      "loss": 0.0003,
+      "step": 11100
+    },
+    {
+      "epoch": 5.177238805970149,
+      "eval_loss": 3.7123268157301936e-06,
+      "eval_runtime": 0.1611,
+      "eval_samples_per_second": 186.257,
+      "eval_steps_per_second": 24.834,
+      "step": 11100
+    },
+    {
+      "epoch": 5.223880597014926,
+      "grad_norm": 0.002973082009702921,
+      "learning_rate": 2.6075870646766173e-05,
+      "loss": 0.0003,
+      "step": 11200
+    },
+    {
+      "epoch": 5.223880597014926,
+      "eval_loss": 1.7487926697867806e-06,
+      "eval_runtime": 0.166,
+      "eval_samples_per_second": 180.691,
+      "eval_steps_per_second": 24.092,
+      "step": 11200
+    },
+    {
+      "epoch": 5.270522388059701,
+      "grad_norm": 0.0015913803363218904,
+      "learning_rate": 2.5951492537313434e-05,
+      "loss": 0.0001,
+      "step": 11300
+    },
+    {
+      "epoch": 5.270522388059701,
+      "eval_loss": 1.5435723526024958e-06,
+      "eval_runtime": 0.1604,
+      "eval_samples_per_second": 187.017,
+      "eval_steps_per_second": 24.936,
+      "step": 11300
+    },
+    {
+      "epoch": 5.317164179104478,
+      "grad_norm": 0.0005316142342053354,
+      "learning_rate": 2.58271144278607e-05,
+      "loss": 0.0001,
+      "step": 11400
+    },
+    {
+      "epoch": 5.317164179104478,
+      "eval_loss": 1.4954798643884715e-06,
+      "eval_runtime": 0.1594,
+      "eval_samples_per_second": 188.187,
+      "eval_steps_per_second": 25.092,
+      "step": 11400
+    },
+    {
+      "epoch": 5.3638059701492535,
+      "grad_norm": 0.0008541183196939528,
+      "learning_rate": 2.570273631840796e-05,
+      "loss": 0.0001,
+      "step": 11500
+    },
+    {
+      "epoch": 5.3638059701492535,
+      "eval_loss": 1.961838506758795e-06,
+      "eval_runtime": 0.1597,
+      "eval_samples_per_second": 187.843,
+      "eval_steps_per_second": 25.046,
+      "step": 11500
+    },
+    {
+      "epoch": 5.41044776119403,
+      "grad_norm": 0.0004898870829492807,
+      "learning_rate": 2.5578358208955226e-05,
+      "loss": 0.0001,
+      "step": 11600
+    },
+    {
+      "epoch": 5.41044776119403,
+      "eval_loss": 1.911733761517098e-06,
+      "eval_runtime": 0.1589,
+      "eval_samples_per_second": 188.76,
+      "eval_steps_per_second": 25.168,
+      "step": 11600
+    },
+    {
+      "epoch": 5.457089552238806,
+      "grad_norm": 0.0006626475951634347,
+      "learning_rate": 2.5453980099502488e-05,
+      "loss": 0.0001,
+      "step": 11700
+    },
+    {
+      "epoch": 5.457089552238806,
+      "eval_loss": 3.094894736932474e-06,
+      "eval_runtime": 0.1695,
+      "eval_samples_per_second": 177.033,
+      "eval_steps_per_second": 23.604,
+      "step": 11700
+    },
+    {
+      "epoch": 5.503731343283582,
+      "grad_norm": 0.005953106097877026,
+      "learning_rate": 2.5329601990049756e-05,
+      "loss": 0.0001,
+      "step": 11800
+    },
+    {
+      "epoch": 5.503731343283582,
+      "eval_loss": 3.518416178849293e-06,
+      "eval_runtime": 0.1624,
+      "eval_samples_per_second": 184.755,
+      "eval_steps_per_second": 24.634,
+      "step": 11800
+    },
+    {
+      "epoch": 5.550373134328359,
+      "grad_norm": 0.017002714797854424,
+      "learning_rate": 2.5205223880597018e-05,
+      "loss": 0.0004,
+      "step": 11900
+    },
+    {
+      "epoch": 5.550373134328359,
+      "eval_loss": 6.997787295404123e-06,
+      "eval_runtime": 0.1643,
+      "eval_samples_per_second": 182.545,
+      "eval_steps_per_second": 24.339,
+      "step": 11900
+    },
+    {
+      "epoch": 5.597014925373134,
+      "grad_norm": 0.00019286558381281793,
+      "learning_rate": 2.5080845771144283e-05,
+      "loss": 0.0001,
+      "step": 12000
+    },
+    {
+      "epoch": 5.597014925373134,
+      "eval_loss": 4.117903699807357e-06,
+      "eval_runtime": 0.1621,
+      "eval_samples_per_second": 185.078,
+      "eval_steps_per_second": 24.677,
+      "step": 12000
+    },
+    {
+      "epoch": 5.643656716417911,
+      "grad_norm": 0.0007640725816600025,
+      "learning_rate": 2.4956467661691544e-05,
+      "loss": 0.0003,
+      "step": 12100
+    },
+    {
+      "epoch": 5.643656716417911,
+      "eval_loss": 1.6209987734328024e-06,
+      "eval_runtime": 0.1629,
+      "eval_samples_per_second": 184.172,
+      "eval_steps_per_second": 24.556,
+      "step": 12100
+    },
+    {
+      "epoch": 5.690298507462686,
+      "grad_norm": 0.0009279770310968161,
+      "learning_rate": 2.483208955223881e-05,
+      "loss": 0.0001,
+      "step": 12200
+    },
+    {
+      "epoch": 5.690298507462686,
+      "eval_loss": 1.4867666777718114e-06,
+      "eval_runtime": 0.1612,
+      "eval_samples_per_second": 186.065,
+      "eval_steps_per_second": 24.809,
+      "step": 12200
+    },
+    {
+      "epoch": 5.736940298507463,
+      "grad_norm": 0.015755705535411835,
+      "learning_rate": 2.470771144278607e-05,
+      "loss": 0.0001,
+      "step": 12300
+    },
+    {
+      "epoch": 5.736940298507463,
+      "eval_loss": 1.4831955468253e-06,
+      "eval_runtime": 0.1631,
+      "eval_samples_per_second": 183.983,
+      "eval_steps_per_second": 24.531,
+      "step": 12300
+    },
+    {
+      "epoch": 5.7835820895522385,
+      "grad_norm": 0.0005079147522337735,
+      "learning_rate": 2.4583333333333336e-05,
+      "loss": 0.0001,
+      "step": 12400
+    },
+    {
+      "epoch": 5.7835820895522385,
+      "eval_loss": 1.4258381497711525e-06,
+      "eval_runtime": 0.166,
+      "eval_samples_per_second": 180.722,
+      "eval_steps_per_second": 24.096,
+      "step": 12400
+    },
+    {
+      "epoch": 5.830223880597015,
+      "grad_norm": 0.0003629447892308235,
+      "learning_rate": 2.4458955223880597e-05,
+      "loss": 0.0002,
+      "step": 12500
+    },
+    {
+      "epoch": 5.830223880597015,
+      "eval_loss": 1.1145809821755392e-06,
+      "eval_runtime": 0.1616,
+      "eval_samples_per_second": 185.699,
+      "eval_steps_per_second": 24.76,
+      "step": 12500
+    },
+    {
+      "epoch": 5.8768656716417915,
+      "grad_norm": 0.0025047562085092068,
+      "learning_rate": 2.4334577114427866e-05,
+      "loss": 0.0003,
+      "step": 12600
+    },
+    {
+      "epoch": 5.8768656716417915,
+      "eval_loss": 1.6086485175037524e-06,
+      "eval_runtime": 0.163,
+      "eval_samples_per_second": 184.018,
+      "eval_steps_per_second": 24.536,
+      "step": 12600
+    },
+    {
+      "epoch": 5.923507462686567,
+      "grad_norm": 0.0010979525977745652,
+      "learning_rate": 2.4210199004975127e-05,
+      "loss": 0.0001,
+      "step": 12700
+    },
+    {
+      "epoch": 5.923507462686567,
+      "eval_loss": 1.6331794085999718e-06,
+      "eval_runtime": 0.1638,
+      "eval_samples_per_second": 183.117,
+      "eval_steps_per_second": 24.416,
+      "step": 12700
+    },
+    {
+      "epoch": 5.970149253731344,
+      "grad_norm": 0.00042478801333345473,
+      "learning_rate": 2.4085820895522392e-05,
+      "loss": 0.0005,
+      "step": 12800
+    },
+    {
+      "epoch": 5.970149253731344,
+      "eval_loss": 2.18420427700039e-06,
+      "eval_runtime": 0.1694,
+      "eval_samples_per_second": 177.069,
+      "eval_steps_per_second": 23.609,
+      "step": 12800
+    },
+    {
+      "epoch": 6.016791044776119,
+      "grad_norm": 0.0026371392887085676,
+      "learning_rate": 2.3961442786069654e-05,
+      "loss": 0.0005,
+      "step": 12900
+    },
+    {
+      "epoch": 6.016791044776119,
+      "eval_loss": 2.724199703152408e-06,
+      "eval_runtime": 0.1645,
+      "eval_samples_per_second": 182.396,
+      "eval_steps_per_second": 24.319,
+      "step": 12900
+    },
+    {
+      "epoch": 6.063432835820896,
+      "grad_norm": 0.0005338056362234056,
+      "learning_rate": 2.383706467661692e-05,
+      "loss": 0.0007,
+      "step": 13000
+    },
+    {
+      "epoch": 6.063432835820896,
+      "eval_loss": 2.369762114540208e-06,
+      "eval_runtime": 0.1628,
+      "eval_samples_per_second": 184.255,
+      "eval_steps_per_second": 24.567,
+      "step": 13000
+    },
+    {
+      "epoch": 6.110074626865671,
+      "grad_norm": 0.0006243676762096584,
+      "learning_rate": 2.371268656716418e-05,
+      "loss": 0.0004,
+      "step": 13100
+    },
+    {
+      "epoch": 6.110074626865671,
+      "eval_loss": 4.675569016399095e-06,
+      "eval_runtime": 0.1639,
+      "eval_samples_per_second": 183.01,
+      "eval_steps_per_second": 24.401,
+      "step": 13100
+    },
+    {
+      "epoch": 6.156716417910448,
+      "grad_norm": 0.013767559081315994,
+      "learning_rate": 2.3588308457711445e-05,
+      "loss": 0.0001,
+      "step": 13200
+    },
+    {
+      "epoch": 6.156716417910448,
+      "eval_loss": 2.445647169224685e-06,
+      "eval_runtime": 0.1632,
+      "eval_samples_per_second": 183.845,
+      "eval_steps_per_second": 24.513,
+      "step": 13200
+    },
+    {
+      "epoch": 6.2033582089552235,
+      "grad_norm": 0.001806699438020587,
+      "learning_rate": 2.3463930348258707e-05,
+      "loss": 0.0001,
+      "step": 13300
+    },
+    {
+      "epoch": 6.2033582089552235,
+      "eval_loss": 2.094437604682753e-06,
+      "eval_runtime": 0.1613,
+      "eval_samples_per_second": 185.963,
+      "eval_steps_per_second": 24.795,
+      "step": 13300
+    },
+    {
+      "epoch": 6.25,
+      "grad_norm": 0.0007174393394961953,
+      "learning_rate": 2.3339552238805972e-05,
+      "loss": 0.0002,
+      "step": 13400
+    },
+    {
+      "epoch": 6.25,
+      "eval_loss": 1.5305414535760065e-06,
+      "eval_runtime": 0.1641,
+      "eval_samples_per_second": 182.796,
+      "eval_steps_per_second": 24.373,
+      "step": 13400
+    },
+    {
+      "epoch": 6.2966417910447765,
+      "grad_norm": 0.0014867472928017378,
+      "learning_rate": 2.3215174129353234e-05,
+      "loss": 0.0001,
+      "step": 13500
+    },
+    {
+      "epoch": 6.2966417910447765,
+      "eval_loss": 1.4038557765161386e-06,
+      "eval_runtime": 0.1614,
+      "eval_samples_per_second": 185.841,
+      "eval_steps_per_second": 24.779,
+      "step": 13500
+    },
+    {
+      "epoch": 6.343283582089552,
+      "grad_norm": 0.0010864852229133248,
+      "learning_rate": 2.3090796019900502e-05,
+      "loss": 0.0001,
+      "step": 13600
+    },
+    {
+      "epoch": 6.343283582089552,
+      "eval_loss": 1.2951429653185187e-06,
+      "eval_runtime": 0.1624,
+      "eval_samples_per_second": 184.725,
+      "eval_steps_per_second": 24.63,
+      "step": 13600
+    },
+    {
+      "epoch": 6.389925373134329,
+      "grad_norm": 0.0002408225554972887,
+      "learning_rate": 2.2966417910447764e-05,
+      "loss": 0.0001,
+      "step": 13700
+    },
+    {
+      "epoch": 6.389925373134329,
+      "eval_loss": 1.1805148005805677e-06,
+      "eval_runtime": 0.1646,
+      "eval_samples_per_second": 182.255,
+      "eval_steps_per_second": 24.301,
+      "step": 13700
+    },
+    {
+      "epoch": 6.436567164179104,
+      "grad_norm": 0.000278256309684366,
+      "learning_rate": 2.284203980099503e-05,
+      "loss": 0.0001,
+      "step": 13800
+    },
+    {
+      "epoch": 6.436567164179104,
+      "eval_loss": 1.1244396773690823e-06,
+      "eval_runtime": 0.1635,
+      "eval_samples_per_second": 183.524,
+      "eval_steps_per_second": 24.47,
+      "step": 13800
+    },
+    {
+      "epoch": 6.483208955223881,
+      "grad_norm": 0.0005975699750706553,
+      "learning_rate": 2.271766169154229e-05,
+      "loss": 0.0001,
+      "step": 13900
+    },
+    {
+      "epoch": 6.483208955223881,
+      "eval_loss": 9.522905202175025e-07,
+      "eval_runtime": 0.1609,
+      "eval_samples_per_second": 186.467,
+      "eval_steps_per_second": 24.862,
+      "step": 13900
+    },
+    {
+      "epoch": 6.529850746268656,
+      "grad_norm": 0.0024864040315151215,
+      "learning_rate": 2.2593283582089555e-05,
+      "loss": 0.0002,
+      "step": 14000
+    },
+    {
+      "epoch": 6.529850746268656,
+      "eval_loss": 8.695107567291416e-07,
+      "eval_runtime": 0.1606,
+      "eval_samples_per_second": 186.805,
+      "eval_steps_per_second": 24.907,
+      "step": 14000
+    },
+    {
+      "epoch": 6.576492537313433,
+      "grad_norm": 0.0012296285713091493,
+      "learning_rate": 2.2468905472636817e-05,
+      "loss": 0.0001,
+      "step": 14100
+    },
+    {
+      "epoch": 6.576492537313433,
+      "eval_loss": 7.833210702301585e-07,
+      "eval_runtime": 0.1625,
+      "eval_samples_per_second": 184.667,
+      "eval_steps_per_second": 24.622,
+      "step": 14100
+    },
+    {
+      "epoch": 6.6231343283582085,
+      "grad_norm": 0.00046841197763569653,
+      "learning_rate": 2.2344527363184082e-05,
+      "loss": 0.0001,
+      "step": 14200
+    },
+    {
+      "epoch": 6.6231343283582085,
+      "eval_loss": 7.663988981221337e-07,
+      "eval_runtime": 0.1615,
+      "eval_samples_per_second": 185.74,
+      "eval_steps_per_second": 24.765,
+      "step": 14200
+    },
+    {
+      "epoch": 6.669776119402985,
+      "grad_norm": 0.001978447660803795,
+      "learning_rate": 2.2220149253731343e-05,
+      "loss": 0.0028,
+      "step": 14300
+    },
+    {
+      "epoch": 6.669776119402985,
+      "eval_loss": 9.115080956689781e-07,
+      "eval_runtime": 0.1638,
+      "eval_samples_per_second": 183.206,
+      "eval_steps_per_second": 24.427,
+      "step": 14300
+    },
+    {
+      "epoch": 6.7164179104477615,
+      "grad_norm": 0.0005547442706301808,
+      "learning_rate": 2.2095771144278612e-05,
+      "loss": 0.0001,
+      "step": 14400
+    },
+    {
+      "epoch": 6.7164179104477615,
+      "eval_loss": 9.500566306996916e-07,
+      "eval_runtime": 0.1611,
+      "eval_samples_per_second": 186.194,
+      "eval_steps_per_second": 24.826,
+      "step": 14400
+    },
+    {
+      "epoch": 6.763059701492537,
+      "grad_norm": 0.0010947503615170717,
+      "learning_rate": 2.1971393034825873e-05,
+      "loss": 0.0004,
+      "step": 14500
+    },
+    {
+      "epoch": 6.763059701492537,
+      "eval_loss": 1.4260593843573588e-06,
+      "eval_runtime": 0.1612,
+      "eval_samples_per_second": 186.089,
+      "eval_steps_per_second": 24.812,
+      "step": 14500
+    },
+    {
+      "epoch": 6.809701492537314,
+      "grad_norm": 0.00212489883415401,
+      "learning_rate": 2.184701492537314e-05,
+      "loss": 0.0008,
+      "step": 14600
+    },
+    {
+      "epoch": 6.809701492537314,
+      "eval_loss": 1.273897169085103e-06,
+      "eval_runtime": 0.1692,
+      "eval_samples_per_second": 177.355,
+      "eval_steps_per_second": 23.647,
+      "step": 14600
+    },
+    {
+      "epoch": 6.856343283582089,
+      "grad_norm": 0.0002277992753079161,
+      "learning_rate": 2.17226368159204e-05,
+      "loss": 0.0001,
+      "step": 14700
+    },
+    {
+      "epoch": 6.856343283582089,
+      "eval_loss": 1.0682890660973499e-06,
+      "eval_runtime": 0.1623,
+      "eval_samples_per_second": 184.877,
+      "eval_steps_per_second": 24.65,
+      "step": 14700
+    },
+    {
+      "epoch": 6.902985074626866,
+      "grad_norm": 0.0003076612192671746,
+      "learning_rate": 2.1598258706467665e-05,
+      "loss": 0.0,
+      "step": 14800
+    },
+    {
+      "epoch": 6.902985074626866,
+      "eval_loss": 1.0034937076852657e-06,
+      "eval_runtime": 0.1641,
+      "eval_samples_per_second": 182.834,
+      "eval_steps_per_second": 24.378,
+      "step": 14800
+    },
+    {
+      "epoch": 6.949626865671641,
+      "grad_norm": 0.0002643874322529882,
+      "learning_rate": 2.1473880597014927e-05,
+      "loss": 0.0,
+      "step": 14900
+    },
+    {
+      "epoch": 6.949626865671641,
+      "eval_loss": 9.63684556154476e-07,
+      "eval_runtime": 0.1646,
+      "eval_samples_per_second": 182.236,
+      "eval_steps_per_second": 24.298,
+      "step": 14900
+    },
+    {
+      "epoch": 6.996268656716418,
+      "grad_norm": 0.00024292635498568416,
+      "learning_rate": 2.134950248756219e-05,
+      "loss": 0.0001,
+      "step": 15000
+    },
+    {
+      "epoch": 6.996268656716418,
+      "eval_loss": 9.160715990219614e-07,
+      "eval_runtime": 0.1631,
+      "eval_samples_per_second": 183.892,
+      "eval_steps_per_second": 24.519,
+      "step": 15000
+    },
+    {
+      "epoch": 7.042910447761194,
+      "grad_norm": 0.0030344121623784304,
+      "learning_rate": 2.1225124378109453e-05,
+      "loss": 0.0001,
+      "step": 15100
+    },
+    {
+      "epoch": 7.042910447761194,
+      "eval_loss": 8.196147973649204e-07,
+      "eval_runtime": 0.1634,
+      "eval_samples_per_second": 183.619,
+      "eval_steps_per_second": 24.482,
+      "step": 15100
+    },
+    {
+      "epoch": 7.08955223880597,
+      "grad_norm": 0.0007852727430872619,
+      "learning_rate": 2.110074626865672e-05,
+      "loss": 0.0001,
+      "step": 15200
+    },
+    {
+      "epoch": 7.08955223880597,
+      "eval_loss": 8.8703302481008e-07,
+      "eval_runtime": 0.1629,
+      "eval_samples_per_second": 184.182,
+      "eval_steps_per_second": 24.558,
+      "step": 15200
+    },
+    {
+      "epoch": 7.1361940298507465,
+      "grad_norm": 0.0011632639216259122,
+      "learning_rate": 2.097636815920398e-05,
+      "loss": 0.0001,
+      "step": 15300
+    },
+    {
+      "epoch": 7.1361940298507465,
+      "eval_loss": 7.986407126736594e-07,
+      "eval_runtime": 0.1615,
+      "eval_samples_per_second": 185.782,
+      "eval_steps_per_second": 24.771,
+      "step": 15300
+    },
+    {
+      "epoch": 7.182835820895522,
+      "grad_norm": 0.000128728206618689,
+      "learning_rate": 2.0851990049751248e-05,
+      "loss": 0.0,
+      "step": 15400
+    },
+    {
+      "epoch": 7.182835820895522,
+      "eval_loss": 7.614453352289274e-07,
+      "eval_runtime": 0.1635,
+      "eval_samples_per_second": 183.534,
+      "eval_steps_per_second": 24.471,
+      "step": 15400
+    },
+    {
+      "epoch": 7.229477611940299,
+      "grad_norm": 0.00016260573465842754,
+      "learning_rate": 2.072761194029851e-05,
+      "loss": 0.0001,
+      "step": 15500
+    },
+    {
+      "epoch": 7.229477611940299,
+      "eval_loss": 7.017587222435395e-07,
+      "eval_runtime": 0.1617,
+      "eval_samples_per_second": 185.473,
+      "eval_steps_per_second": 24.73,
+      "step": 15500
+    },
+    {
+      "epoch": 7.276119402985074,
+      "grad_norm": 0.0003854296519421041,
+      "learning_rate": 2.0603233830845775e-05,
+      "loss": 0.0002,
+      "step": 15600
+    },
+    {
+      "epoch": 7.276119402985074,
+      "eval_loss": 6.69997461955063e-07,
+      "eval_runtime": 0.163,
+      "eval_samples_per_second": 184.053,
+      "eval_steps_per_second": 24.54,
+      "step": 15600
+    },
+    {
+      "epoch": 7.322761194029851,
+      "grad_norm": 0.0005631668609566987,
+      "learning_rate": 2.0478855721393036e-05,
+      "loss": 0.0,
+      "step": 15700
+    },
+    {
+      "epoch": 7.322761194029851,
+      "eval_loss": 6.535386773975915e-07,
+      "eval_runtime": 0.1623,
+      "eval_samples_per_second": 184.837,
+      "eval_steps_per_second": 24.645,
+      "step": 15700
+    },
+    {
+      "epoch": 7.369402985074627,
+      "grad_norm": 0.00030086564947851,
+      "learning_rate": 2.03544776119403e-05,
+      "loss": 0.0001,
+      "step": 15800
+    },
+    {
+      "epoch": 7.369402985074627,
+      "eval_loss": 6.11449081588944e-07,
+      "eval_runtime": 0.17,
+      "eval_samples_per_second": 176.431,
+      "eval_steps_per_second": 23.524,
+      "step": 15800
+    },
+    {
+      "epoch": 7.416044776119403,
+      "grad_norm": 0.00038130092434585094,
+      "learning_rate": 2.0230099502487563e-05,
+      "loss": 0.0001,
+      "step": 15900
+    },
+    {
+      "epoch": 7.416044776119403,
+      "eval_loss": 5.757526650995715e-07,
+      "eval_runtime": 0.1612,
+      "eval_samples_per_second": 186.141,
+      "eval_steps_per_second": 24.819,
+      "step": 15900
+    },
+    {
+      "epoch": 7.462686567164179,
+      "grad_norm": 0.000344213709468022,
+      "learning_rate": 2.0106965174129357e-05,
+      "loss": 0.0004,
+      "step": 16000
+    },
+    {
+      "epoch": 7.462686567164179,
+      "eval_loss": 5.750833338424854e-07,
+      "eval_runtime": 0.1619,
+      "eval_samples_per_second": 185.307,
+      "eval_steps_per_second": 24.708,
+      "step": 16000
+    },
+    {
+      "epoch": 7.509328358208955,
+      "grad_norm": 0.0006042002351023257,
+      "learning_rate": 1.998258706467662e-05,
+      "loss": 0.0002,
+      "step": 16100
+    },
+    {
+      "epoch": 7.509328358208955,
+      "eval_loss": 8.297544695778925e-07,
+      "eval_runtime": 0.1687,
+      "eval_samples_per_second": 177.799,
+      "eval_steps_per_second": 23.707,
+      "step": 16100
+    },
+    {
+      "epoch": 7.5559701492537314,
+      "grad_norm": 0.0012880139984190464,
+      "learning_rate": 1.9858208955223884e-05,
+      "loss": 0.0001,
+      "step": 16200
+    },
+    {
+      "epoch": 7.5559701492537314,
+      "eval_loss": 7.933731467346661e-07,
+      "eval_runtime": 0.1631,
+      "eval_samples_per_second": 183.887,
+      "eval_steps_per_second": 24.518,
+      "step": 16200
+    },
+    {
+      "epoch": 7.602611940298507,
+      "grad_norm": 0.0001834977010730654,
+      "learning_rate": 1.9733830845771145e-05,
+      "loss": 0.0,
+      "step": 16300
+    },
+    {
+      "epoch": 7.602611940298507,
+      "eval_loss": 7.420648557854292e-07,
+      "eval_runtime": 0.1644,
+      "eval_samples_per_second": 182.491,
+      "eval_steps_per_second": 24.332,
+      "step": 16300
+    },
+    {
+      "epoch": 7.649253731343284,
+      "grad_norm": 0.0002803165989462286,
+      "learning_rate": 1.960945273631841e-05,
+      "loss": 0.0004,
+      "step": 16400
+    },
+    {
+      "epoch": 7.649253731343284,
+      "eval_loss": 8.606352821516339e-07,
+      "eval_runtime": 0.1658,
+      "eval_samples_per_second": 180.905,
+      "eval_steps_per_second": 24.121,
+      "step": 16400
+    },
+    {
+      "epoch": 7.69589552238806,
+      "grad_norm": 0.00030525182955898345,
+      "learning_rate": 1.9485074626865675e-05,
+      "loss": 0.0,
+      "step": 16500
+    },
+    {
+      "epoch": 7.69589552238806,
+      "eval_loss": 7.444876928275335e-07,
+      "eval_runtime": 0.1624,
+      "eval_samples_per_second": 184.746,
+      "eval_steps_per_second": 24.633,
+      "step": 16500
+    },
+    {
+      "epoch": 7.742537313432836,
+      "grad_norm": 0.0005061542615294456,
+      "learning_rate": 1.9360696517412937e-05,
+      "loss": 0.0,
+      "step": 16600
+    },
+    {
+      "epoch": 7.742537313432836,
+      "eval_loss": 7.410563966914196e-07,
+      "eval_runtime": 0.1632,
+      "eval_samples_per_second": 183.792,
+      "eval_steps_per_second": 24.506,
+      "step": 16600
+    },
+    {
+      "epoch": 7.789179104477612,
+      "grad_norm": 0.002078367630019784,
+      "learning_rate": 1.9236318407960202e-05,
+      "loss": 0.0,
+      "step": 16700
+    },
+    {
+      "epoch": 7.789179104477612,
+      "eval_loss": 7.394659746751131e-07,
+      "eval_runtime": 0.1672,
+      "eval_samples_per_second": 179.409,
+      "eval_steps_per_second": 23.921,
+      "step": 16700
+    },
+    {
+      "epoch": 7.835820895522388,
+      "grad_norm": 0.0005327766994014382,
+      "learning_rate": 1.9111940298507467e-05,
+      "loss": 0.0,
+      "step": 16800
+    },
+    {
+      "epoch": 7.835820895522388,
+      "eval_loss": 6.818034421485208e-07,
+      "eval_runtime": 0.1907,
+      "eval_samples_per_second": 157.346,
+      "eval_steps_per_second": 20.979,
+      "step": 16800
+    },
+    {
+      "epoch": 7.882462686567164,
+      "grad_norm": 0.00013845170906279236,
+      "learning_rate": 1.8987562189054725e-05,
+      "loss": 0.0001,
+      "step": 16900
+    },
+    {
+      "epoch": 7.882462686567164,
+      "eval_loss": 6.097284313000273e-07,
+      "eval_runtime": 0.1662,
+      "eval_samples_per_second": 180.555,
+      "eval_steps_per_second": 24.074,
+      "step": 16900
+    },
+    {
+      "epoch": 7.92910447761194,
+      "grad_norm": 0.0008268446545116603,
+      "learning_rate": 1.886318407960199e-05,
+      "loss": 0.0001,
+      "step": 17000
+    },
+    {
+      "epoch": 7.92910447761194,
+      "eval_loss": 6.165503236843506e-07,
+      "eval_runtime": 0.1612,
+      "eval_samples_per_second": 186.081,
+      "eval_steps_per_second": 24.811,
+      "step": 17000
+    },
+    {
+      "epoch": 7.975746268656716,
+      "grad_norm": 0.011116830632090569,
+      "learning_rate": 1.8738805970149255e-05,
+      "loss": 0.0002,
+      "step": 17100
+    },
+    {
+      "epoch": 7.975746268656716,
+      "eval_loss": 6.690197551506571e-07,
+      "eval_runtime": 0.1629,
+      "eval_samples_per_second": 184.18,
+      "eval_steps_per_second": 24.557,
+      "step": 17100
+    },
+    {
+      "epoch": 8.022388059701493,
+      "grad_norm": 0.00011780932982219383,
+      "learning_rate": 1.8614427860696517e-05,
+      "loss": 0.0,
+      "step": 17200
+    },
+    {
+      "epoch": 8.022388059701493,
+      "eval_loss": 6.267726462283463e-07,
+      "eval_runtime": 0.1622,
+      "eval_samples_per_second": 184.913,
+      "eval_steps_per_second": 24.655,
+      "step": 17200
+    },
+    {
+      "epoch": 8.069029850746269,
+      "grad_norm": 0.00017962571291718632,
+      "learning_rate": 1.8490049751243782e-05,
+      "loss": 0.0001,
+      "step": 17300
+    },
+    {
+      "epoch": 8.069029850746269,
+      "eval_loss": 5.683208996742906e-07,
+      "eval_runtime": 0.1617,
+      "eval_samples_per_second": 185.574,
+      "eval_steps_per_second": 24.743,
+      "step": 17300
+    },
+    {
+      "epoch": 8.115671641791044,
+      "grad_norm": 0.005691648926585913,
+      "learning_rate": 1.8365671641791047e-05,
+      "loss": 0.0,
+      "step": 17400
+    },
+    {
+      "epoch": 8.115671641791044,
+      "eval_loss": 5.686233066626301e-07,
+      "eval_runtime": 0.164,
+      "eval_samples_per_second": 182.936,
+      "eval_steps_per_second": 24.392,
+      "step": 17400
+    },
+    {
+      "epoch": 8.162313432835822,
+      "grad_norm": 0.0010604019043967128,
+      "learning_rate": 1.824129353233831e-05,
+      "loss": 0.0003,
+      "step": 17500
+    },
+    {
+      "epoch": 8.162313432835822,
+      "eval_loss": 4.841186296289379e-07,
+      "eval_runtime": 0.163,
+      "eval_samples_per_second": 184.061,
+      "eval_steps_per_second": 24.541,
+      "step": 17500
+    },
+    {
+      "epoch": 8.208955223880597,
+      "grad_norm": 0.0007983978721313179,
+      "learning_rate": 1.8116915422885573e-05,
+      "loss": 0.0,
+      "step": 17600
+    },
+    {
+      "epoch": 8.208955223880597,
+      "eval_loss": 4.732060858714249e-07,
+      "eval_runtime": 0.1617,
+      "eval_samples_per_second": 185.472,
+      "eval_steps_per_second": 24.73,
+      "step": 17600
+    },
+    {
+      "epoch": 8.255597014925373,
+      "grad_norm": 0.007192044984549284,
+      "learning_rate": 1.7992537313432835e-05,
+      "loss": 0.0,
+      "step": 17700
+    },
+    {
+      "epoch": 8.255597014925373,
+      "eval_loss": 4.6311160417644714e-07,
+      "eval_runtime": 0.165,
+      "eval_samples_per_second": 181.852,
+      "eval_steps_per_second": 24.247,
+      "step": 17700
+    },
+    {
+      "epoch": 8.302238805970148,
+      "grad_norm": 0.0002632784890010953,
+      "learning_rate": 1.78681592039801e-05,
+      "loss": 0.0001,
+      "step": 17800
+    },
+    {
+      "epoch": 8.302238805970148,
+      "eval_loss": 4.890302420790249e-07,
+      "eval_runtime": 0.1652,
+      "eval_samples_per_second": 181.562,
+      "eval_steps_per_second": 24.208,
+      "step": 17800
+    },
+    {
+      "epoch": 8.348880597014926,
+      "grad_norm": 0.00016267193132080138,
+      "learning_rate": 1.7743781094527365e-05,
+      "loss": 0.0,
+      "step": 17900
+    },
+    {
+      "epoch": 8.348880597014926,
+      "eval_loss": 4.708208507508971e-07,
+      "eval_runtime": 0.1641,
+      "eval_samples_per_second": 182.793,
+      "eval_steps_per_second": 24.372,
+      "step": 17900
+    },
+    {
+      "epoch": 8.395522388059701,
+      "grad_norm": 0.002555535174906254,
+      "learning_rate": 1.7620646766169156e-05,
+      "loss": 0.0005,
+      "step": 18000
+    },
+    {
+      "epoch": 8.395522388059701,
+      "eval_loss": 4.697274107456906e-07,
+      "eval_runtime": 0.1628,
+      "eval_samples_per_second": 184.314,
+      "eval_steps_per_second": 24.575,
+      "step": 18000
+    },
+    {
+      "epoch": 8.442164179104477,
+      "grad_norm": 0.00030655847513116896,
+      "learning_rate": 1.749626865671642e-05,
+      "loss": 0.0,
+      "step": 18100
+    },
+    {
+      "epoch": 8.442164179104477,
+      "eval_loss": 4.6684658627782483e-07,
+      "eval_runtime": 0.1639,
+      "eval_samples_per_second": 183.019,
+      "eval_steps_per_second": 24.403,
+      "step": 18100
+    },
+    {
+      "epoch": 8.488805970149254,
+      "grad_norm": 0.001360408728942275,
+      "learning_rate": 1.7371890547263682e-05,
+      "loss": 0.0,
+      "step": 18200
+    },
+    {
+      "epoch": 8.488805970149254,
+      "eval_loss": 4.7605757913515845e-07,
+      "eval_runtime": 0.1648,
+      "eval_samples_per_second": 182.057,
+      "eval_steps_per_second": 24.274,
+      "step": 18200
+    },
+    {
+      "epoch": 8.53544776119403,
+      "grad_norm": 0.00014535202353727072,
+      "learning_rate": 1.7247512437810947e-05,
+      "loss": 0.0005,
+      "step": 18300
+    },
+    {
+      "epoch": 8.53544776119403,
+      "eval_loss": 4.991737228010606e-07,
+      "eval_runtime": 0.1636,
+      "eval_samples_per_second": 183.427,
+      "eval_steps_per_second": 24.457,
+      "step": 18300
+    },
+    {
+      "epoch": 8.582089552238806,
+      "grad_norm": 0.0006497761351056397,
+      "learning_rate": 1.7123134328358212e-05,
+      "loss": 0.0001,
+      "step": 18400
+    },
+    {
+      "epoch": 8.582089552238806,
+      "eval_loss": 4.467078440484329e-07,
+      "eval_runtime": 0.1633,
+      "eval_samples_per_second": 183.659,
+      "eval_steps_per_second": 24.488,
+      "step": 18400
+    },
+    {
+      "epoch": 8.628731343283581,
+      "grad_norm": 0.00027351349126547575,
+      "learning_rate": 1.6998756218905474e-05,
+      "loss": 0.0,
+      "step": 18500
+    },
+    {
+      "epoch": 8.628731343283581,
+      "eval_loss": 4.435265168467595e-07,
+      "eval_runtime": 0.1621,
+      "eval_samples_per_second": 185.033,
+      "eval_steps_per_second": 24.671,
+      "step": 18500
+    },
+    {
+      "epoch": 8.675373134328359,
+      "grad_norm": 0.0005942362477071583,
+      "learning_rate": 1.687437810945274e-05,
+      "loss": 0.0003,
+      "step": 18600
+    },
+    {
+      "epoch": 8.675373134328359,
+      "eval_loss": 4.2622443174877844e-07,
+      "eval_runtime": 0.1645,
+      "eval_samples_per_second": 182.387,
+      "eval_steps_per_second": 24.318,
+      "step": 18600
+    },
+    {
+      "epoch": 8.722014925373134,
+      "grad_norm": 0.00017634141840972006,
+      "learning_rate": 1.675e-05,
+      "loss": 0.0006,
+      "step": 18700
+    },
+    {
+      "epoch": 8.722014925373134,
+      "eval_loss": 4.732918341687764e-07,
+      "eval_runtime": 0.1614,
+      "eval_samples_per_second": 185.867,
+      "eval_steps_per_second": 24.782,
+      "step": 18700
+    },
+    {
+      "epoch": 8.76865671641791,
+      "grad_norm": 0.0002473385538905859,
+      "learning_rate": 1.6625621890547266e-05,
+      "loss": 0.0,
+      "step": 18800
+    },
+    {
+      "epoch": 8.76865671641791,
+      "eval_loss": 4.792402705788845e-07,
+      "eval_runtime": 0.163,
+      "eval_samples_per_second": 184.093,
+      "eval_steps_per_second": 24.546,
+      "step": 18800
+    },
+    {
+      "epoch": 8.815298507462687,
+      "grad_norm": 0.0002839862136170268,
+      "learning_rate": 1.650124378109453e-05,
+      "loss": 0.0,
+      "step": 18900
+    },
+    {
+      "epoch": 8.815298507462687,
+      "eval_loss": 4.7373944767059584e-07,
+      "eval_runtime": 0.1629,
+      "eval_samples_per_second": 184.148,
+      "eval_steps_per_second": 24.553,
+      "step": 18900
+    },
+    {
+      "epoch": 8.861940298507463,
+      "grad_norm": 0.00030838322709314525,
+      "learning_rate": 1.6376865671641792e-05,
+      "loss": 0.0003,
+      "step": 19000
+    },
+    {
+      "epoch": 8.861940298507463,
+      "eval_loss": 5.06621631757298e-07,
+      "eval_runtime": 0.1629,
+      "eval_samples_per_second": 184.159,
+      "eval_steps_per_second": 24.555,
+      "step": 19000
+    },
+    {
+      "epoch": 8.908582089552239,
+      "grad_norm": 0.0008744837250560522,
+      "learning_rate": 1.6252487562189057e-05,
+      "loss": 0.0001,
+      "step": 19100
+    },
+    {
+      "epoch": 8.908582089552239,
+      "eval_loss": 4.863770755036967e-07,
+      "eval_runtime": 0.1638,
+      "eval_samples_per_second": 183.173,
+      "eval_steps_per_second": 24.423,
+      "step": 19100
+    },
+    {
+      "epoch": 8.955223880597014,
+      "grad_norm": 0.0004683129664044827,
+      "learning_rate": 1.612810945273632e-05,
+      "loss": 0.0001,
+      "step": 19200
+    },
+    {
+      "epoch": 8.955223880597014,
+      "eval_loss": 4.3130833660143253e-07,
+      "eval_runtime": 0.163,
+      "eval_samples_per_second": 184.02,
+      "eval_steps_per_second": 24.536,
+      "step": 19200
+    },
+    {
+      "epoch": 9.001865671641792,
+      "grad_norm": 0.0001069534191628918,
+      "learning_rate": 1.6003731343283584e-05,
+      "loss": 0.0,
+      "step": 19300
+    },
+    {
+      "epoch": 9.001865671641792,
+      "eval_loss": 4.232615538057871e-07,
+      "eval_runtime": 0.1637,
+      "eval_samples_per_second": 183.284,
+      "eval_steps_per_second": 24.438,
+      "step": 19300
+    },
+    {
+      "epoch": 9.048507462686567,
+      "grad_norm": 0.0001369858073303476,
+      "learning_rate": 1.587935323383085e-05,
+      "loss": 0.0001,
+      "step": 19400
+    },
+    {
+      "epoch": 9.048507462686567,
+      "eval_loss": 4.031139724247623e-07,
+      "eval_runtime": 0.1628,
+      "eval_samples_per_second": 184.23,
+      "eval_steps_per_second": 24.564,
+      "step": 19400
+    },
+    {
+      "epoch": 9.095149253731343,
+      "grad_norm": 0.0004517412162385881,
+      "learning_rate": 1.575497512437811e-05,
+      "loss": 0.0,
+      "step": 19500
+    },
+    {
+      "epoch": 9.095149253731343,
+      "eval_loss": 3.9417889752257906e-07,
+      "eval_runtime": 0.1614,
+      "eval_samples_per_second": 185.913,
+      "eval_steps_per_second": 24.788,
+      "step": 19500
+    },
+    {
+      "epoch": 9.14179104477612,
+      "grad_norm": 0.0002976813993882388,
+      "learning_rate": 1.5630597014925375e-05,
+      "loss": 0.0001,
+      "step": 19600
+    },
+    {
+      "epoch": 9.14179104477612,
+      "eval_loss": 3.6548814819070685e-07,
+      "eval_runtime": 0.1844,
+      "eval_samples_per_second": 162.664,
+      "eval_steps_per_second": 21.689,
+      "step": 19600
+    },
+    {
+      "epoch": 9.188432835820896,
+      "grad_norm": 0.0005973413935862482,
+      "learning_rate": 1.550621890547264e-05,
+      "loss": 0.0,
+      "step": 19700
+    },
+    {
+      "epoch": 9.188432835820896,
+      "eval_loss": 3.6364929201226914e-07,
+      "eval_runtime": 0.1642,
+      "eval_samples_per_second": 182.755,
+      "eval_steps_per_second": 24.367,
+      "step": 19700
+    },
+    {
+      "epoch": 9.235074626865671,
+      "grad_norm": 0.0002460335963405669,
+      "learning_rate": 1.5381840796019902e-05,
+      "loss": 0.0001,
+      "step": 19800
+    },
+    {
+      "epoch": 9.235074626865671,
+      "eval_loss": 3.593271742374782e-07,
+      "eval_runtime": 0.1686,
+      "eval_samples_per_second": 177.934,
+      "eval_steps_per_second": 23.724,
+      "step": 19800
+    },
+    {
+      "epoch": 9.281716417910447,
+      "grad_norm": 0.0013155044289305806,
+      "learning_rate": 1.5257462686567165e-05,
+      "loss": 0.0,
+      "step": 19900
+    },
+    {
+      "epoch": 9.281716417910447,
+      "eval_loss": 3.7705902400375635e-07,
+      "eval_runtime": 0.1626,
+      "eval_samples_per_second": 184.503,
+      "eval_steps_per_second": 24.6,
+      "step": 19900
+    },
+    {
+      "epoch": 9.328358208955224,
+      "grad_norm": 0.0009738287189975381,
+      "learning_rate": 1.513308457711443e-05,
+      "loss": 0.0,
+      "step": 20000
+    },
+    {
+      "epoch": 9.328358208955224,
+      "eval_loss": 3.643332604497118e-07,
+      "eval_runtime": 0.189,
+      "eval_samples_per_second": 158.751,
+      "eval_steps_per_second": 21.167,
+      "step": 20000
+    },
+    {
+      "epoch": 9.375,
+      "grad_norm": 0.0006048243958503008,
+      "learning_rate": 1.5008706467661693e-05,
+      "loss": 0.0,
+      "step": 20100
+    },
+    {
+      "epoch": 9.375,
+      "eval_loss": 3.528231502514245e-07,
+      "eval_runtime": 0.1642,
+      "eval_samples_per_second": 182.704,
+      "eval_steps_per_second": 24.36,
+      "step": 20100
+    },
+    {
+      "epoch": 9.421641791044776,
+      "grad_norm": 8.00610869191587e-05,
+      "learning_rate": 1.4884328358208957e-05,
+      "loss": 0.0,
+      "step": 20200
+    },
+    {
+      "epoch": 9.421641791044776,
+      "eval_loss": 3.4202770393676474e-07,
+      "eval_runtime": 0.1664,
+      "eval_samples_per_second": 180.277,
+      "eval_steps_per_second": 24.037,
+      "step": 20200
+    },
+    {
+      "epoch": 9.468283582089553,
+      "grad_norm": 0.00042591695091687143,
+      "learning_rate": 1.475995024875622e-05,
+      "loss": 0.0,
+      "step": 20300
+    },
+    {
+      "epoch": 9.468283582089553,
+      "eval_loss": 3.143865114907385e-07,
+      "eval_runtime": 0.165,
+      "eval_samples_per_second": 181.766,
+      "eval_steps_per_second": 24.235,
+      "step": 20300
+    },
+    {
+      "epoch": 9.514925373134329,
+      "grad_norm": 0.00014322852075565606,
+      "learning_rate": 1.4635572139303485e-05,
+      "loss": 0.0,
+      "step": 20400
+    },
+    {
+      "epoch": 9.514925373134329,
+      "eval_loss": 3.284554850324639e-07,
+      "eval_runtime": 0.1675,
+      "eval_samples_per_second": 179.113,
+      "eval_steps_per_second": 23.882,
+      "step": 20400
+    },
+    {
+      "epoch": 9.561567164179104,
+      "grad_norm": 0.005179948173463345,
+      "learning_rate": 1.4511194029850748e-05,
+      "loss": 0.0,
+      "step": 20500
+    },
+    {
+      "epoch": 9.561567164179104,
+      "eval_loss": 3.1635789810025017e-07,
+      "eval_runtime": 0.1647,
+      "eval_samples_per_second": 182.128,
+      "eval_steps_per_second": 24.284,
+      "step": 20500
+    },
+    {
+      "epoch": 9.60820895522388,
+      "grad_norm": 0.12359941005706787,
+      "learning_rate": 1.4386815920398012e-05,
+      "loss": 0.0,
+      "step": 20600
+    },
+    {
+      "epoch": 9.60820895522388,
+      "eval_loss": 3.2864005561350496e-07,
+      "eval_runtime": 0.1623,
+      "eval_samples_per_second": 184.87,
+      "eval_steps_per_second": 24.649,
+      "step": 20600
+    },
+    {
+      "epoch": 9.654850746268657,
+      "grad_norm": 0.00016295950626954436,
+      "learning_rate": 1.4262437810945275e-05,
+      "loss": 0.0,
+      "step": 20700
+    },
+    {
+      "epoch": 9.654850746268657,
+      "eval_loss": 3.228102798402688e-07,
+      "eval_runtime": 0.1635,
+      "eval_samples_per_second": 183.504,
+      "eval_steps_per_second": 24.467,
+      "step": 20700
+    },
+    {
+      "epoch": 9.701492537313433,
+      "grad_norm": 0.00019894339493475854,
+      "learning_rate": 1.413805970149254e-05,
+      "loss": 0.0,
+      "step": 20800
+    },
+    {
+      "epoch": 9.701492537313433,
+      "eval_loss": 3.1584852422383847e-07,
+      "eval_runtime": 0.1648,
+      "eval_samples_per_second": 182.082,
+      "eval_steps_per_second": 24.278,
+      "step": 20800
+    },
+    {
+      "epoch": 9.748134328358208,
+      "grad_norm": 7.664102304261178e-05,
+      "learning_rate": 1.4013681592039803e-05,
+      "loss": 0.0,
+      "step": 20900
+    },
+    {
+      "epoch": 9.748134328358208,
+      "eval_loss": 3.116034292816039e-07,
+      "eval_runtime": 0.1853,
+      "eval_samples_per_second": 161.942,
+      "eval_steps_per_second": 21.592,
+      "step": 20900
+    },
+    {
+      "epoch": 9.794776119402986,
+      "grad_norm": 0.0002929531619884074,
+      "learning_rate": 1.3889303482587067e-05,
+      "loss": 0.0,
+      "step": 21000
+    },
+    {
+      "epoch": 9.794776119402986,
+      "eval_loss": 3.048232031233056e-07,
+      "eval_runtime": 0.1652,
+      "eval_samples_per_second": 181.592,
+      "eval_steps_per_second": 24.212,
+      "step": 21000
+    },
+    {
+      "epoch": 9.841417910447761,
+      "grad_norm": 0.0021353147458285093,
+      "learning_rate": 1.376492537313433e-05,
+      "loss": 0.0,
+      "step": 21100
+    },
+    {
+      "epoch": 9.841417910447761,
+      "eval_loss": 2.998399111220351e-07,
+      "eval_runtime": 0.1725,
+      "eval_samples_per_second": 173.901,
+      "eval_steps_per_second": 23.187,
+      "step": 21100
+    },
+    {
+      "epoch": 9.888059701492537,
+      "grad_norm": 0.00020385629613883793,
+      "learning_rate": 1.3640547263681593e-05,
+      "loss": 0.0,
+      "step": 21200
+    },
+    {
+      "epoch": 9.888059701492537,
+      "eval_loss": 2.8804151952499524e-07,
+      "eval_runtime": 0.163,
+      "eval_samples_per_second": 184.019,
+      "eval_steps_per_second": 24.536,
+      "step": 21200
+    },
+    {
+      "epoch": 9.934701492537313,
+      "grad_norm": 0.00024354759079869837,
+      "learning_rate": 1.3516169154228858e-05,
+      "loss": 0.0,
+      "step": 21300
+    },
+    {
+      "epoch": 9.934701492537313,
+      "eval_loss": 2.821187479185028e-07,
+      "eval_runtime": 0.165,
+      "eval_samples_per_second": 181.842,
+      "eval_steps_per_second": 24.246,
+      "step": 21300
+    },
+    {
+      "epoch": 9.98134328358209,
+      "grad_norm": 0.00834750197827816,
+      "learning_rate": 1.3391791044776121e-05,
+      "loss": 0.0,
+      "step": 21400
+    },
+    {
+      "epoch": 9.98134328358209,
+      "eval_loss": 2.7767814003709645e-07,
+      "eval_runtime": 0.1622,
+      "eval_samples_per_second": 184.954,
+      "eval_steps_per_second": 24.661,
+      "step": 21400
+    },
+    {
+      "epoch": 10.027985074626866,
+      "grad_norm": 0.0002831424935720861,
+      "learning_rate": 1.3267412935323385e-05,
+      "loss": 0.0,
+      "step": 21500
+    },
+    {
+      "epoch": 10.027985074626866,
+      "eval_loss": 2.571897823600011e-07,
+      "eval_runtime": 0.1632,
+      "eval_samples_per_second": 183.854,
+      "eval_steps_per_second": 24.514,
+      "step": 21500
+    },
+    {
+      "epoch": 10.074626865671641,
+      "grad_norm": 0.0013846103101968765,
+      "learning_rate": 1.3143034825870648e-05,
+      "loss": 0.0,
+      "step": 21600
+    },
+    {
+      "epoch": 10.074626865671641,
+      "eval_loss": 2.5103787493208074e-07,
+      "eval_runtime": 0.1621,
+      "eval_samples_per_second": 185.048,
+      "eval_steps_per_second": 24.673,
+      "step": 21600
+    },
+    {
+      "epoch": 10.121268656716419,
+      "grad_norm": 0.0005128834745846689,
+      "learning_rate": 1.3018656716417913e-05,
+      "loss": 0.0,
+      "step": 21700
+    },
+    {
+      "epoch": 10.121268656716419,
+      "eval_loss": 2.498533717698592e-07,
+      "eval_runtime": 0.165,
+      "eval_samples_per_second": 181.785,
+      "eval_steps_per_second": 24.238,
+      "step": 21700
+    },
+    {
+      "epoch": 10.167910447761194,
+      "grad_norm": 0.00024777796352282166,
+      "learning_rate": 1.2894278606965176e-05,
+      "loss": 0.0,
+      "step": 21800
+    },
+    {
+      "epoch": 10.167910447761194,
+      "eval_loss": 2.5362717792631884e-07,
+      "eval_runtime": 0.163,
+      "eval_samples_per_second": 184.079,
+      "eval_steps_per_second": 24.544,
+      "step": 21800
+    },
+    {
+      "epoch": 10.21455223880597,
+      "grad_norm": 0.00038368601235561073,
+      "learning_rate": 1.276990049751244e-05,
+      "loss": 0.0,
+      "step": 21900
+    },
+    {
+      "epoch": 10.21455223880597,
+      "eval_loss": 2.490402550847648e-07,
+      "eval_runtime": 0.165,
+      "eval_samples_per_second": 181.841,
+      "eval_steps_per_second": 24.245,
+      "step": 21900
+    },
+    {
+      "epoch": 10.261194029850746,
+      "grad_norm": 0.0011056758230552077,
+      "learning_rate": 1.2645522388059703e-05,
+      "loss": 0.0,
+      "step": 22000
+    },
+    {
+      "epoch": 10.261194029850746,
+      "eval_loss": 2.4568615231146396e-07,
+      "eval_runtime": 0.1671,
+      "eval_samples_per_second": 179.53,
+      "eval_steps_per_second": 23.937,
+      "step": 22000
+    },
+    {
+      "epoch": 10.307835820895523,
+      "grad_norm": 0.0002836316707544029,
+      "learning_rate": 1.2521144278606966e-05,
+      "loss": 0.0,
+      "step": 22100
+    },
+    {
+      "epoch": 10.307835820895523,
+      "eval_loss": 2.3665577941756055e-07,
+      "eval_runtime": 0.1618,
+      "eval_samples_per_second": 185.41,
+      "eval_steps_per_second": 24.721,
+      "step": 22100
+    },
+    {
+      "epoch": 10.354477611940299,
+      "grad_norm": 0.00016041977505665272,
+      "learning_rate": 1.2396766169154231e-05,
+      "loss": 0.0,
+      "step": 22200
+    },
+    {
+      "epoch": 10.354477611940299,
+      "eval_loss": 2.4731053827053984e-07,
+      "eval_runtime": 0.1668,
+      "eval_samples_per_second": 179.878,
+      "eval_steps_per_second": 23.984,
+      "step": 22200
+    },
+    {
+      "epoch": 10.401119402985074,
+      "grad_norm": 0.0011778981424868107,
+      "learning_rate": 1.2272388059701494e-05,
+      "loss": 0.0,
+      "step": 22300
+    },
+    {
+      "epoch": 10.401119402985074,
+      "eval_loss": 2.5048382212844444e-07,
+      "eval_runtime": 0.164,
+      "eval_samples_per_second": 182.885,
+      "eval_steps_per_second": 24.385,
+      "step": 22300
+    },
+    {
+      "epoch": 10.447761194029852,
+      "grad_norm": 0.004293927922844887,
+      "learning_rate": 1.2149253731343285e-05,
+      "loss": 0.0007,
+      "step": 22400
+    },
+    {
+      "epoch": 10.447761194029852,
+      "eval_loss": 2.6977289735441445e-07,
+      "eval_runtime": 0.1664,
+      "eval_samples_per_second": 180.28,
+      "eval_steps_per_second": 24.037,
+      "step": 22400
+    },
+    {
+      "epoch": 10.494402985074627,
+      "grad_norm": 0.0003302557743154466,
+      "learning_rate": 1.2024875621890549e-05,
+      "loss": 0.0,
+      "step": 22500
+    },
+    {
+      "epoch": 10.494402985074627,
+      "eval_loss": 2.6683247256187315e-07,
+      "eval_runtime": 0.1641,
+      "eval_samples_per_second": 182.868,
+      "eval_steps_per_second": 24.382,
+      "step": 22500
+    },
+    {
+      "epoch": 10.541044776119403,
+      "grad_norm": 0.000583185872528702,
+      "learning_rate": 1.1900497512437812e-05,
+      "loss": 0.0001,
+      "step": 22600
+    },
+    {
+      "epoch": 10.541044776119403,
+      "eval_loss": 2.8590250167326303e-07,
+      "eval_runtime": 0.1741,
+      "eval_samples_per_second": 172.328,
+      "eval_steps_per_second": 22.977,
+      "step": 22600
+    },
+    {
+      "epoch": 10.587686567164178,
+      "grad_norm": 0.00015397944662254304,
+      "learning_rate": 1.1776119402985075e-05,
+      "loss": 0.0,
+      "step": 22700
+    },
+    {
+      "epoch": 10.587686567164178,
+      "eval_loss": 2.9392734290922817e-07,
+      "eval_runtime": 0.1658,
+      "eval_samples_per_second": 180.957,
+      "eval_steps_per_second": 24.128,
+      "step": 22700
+    },
+    {
+      "epoch": 10.634328358208956,
+      "grad_norm": 0.00023940723622217774,
+      "learning_rate": 1.1651741293532339e-05,
+      "loss": 0.0,
+      "step": 22800
+    },
+    {
+      "epoch": 10.634328358208956,
+      "eval_loss": 2.8955645348105463e-07,
+      "eval_runtime": 0.1646,
+      "eval_samples_per_second": 182.232,
+      "eval_steps_per_second": 24.298,
+      "step": 22800
+    },
+    {
+      "epoch": 10.680970149253731,
+      "grad_norm": 0.0006399727426469326,
+      "learning_rate": 1.1527363184079604e-05,
+      "loss": 0.0,
+      "step": 22900
+    },
+    {
+      "epoch": 10.680970149253731,
+      "eval_loss": 2.7836446747642185e-07,
+      "eval_runtime": 0.1653,
+      "eval_samples_per_second": 181.504,
+      "eval_steps_per_second": 24.201,
+      "step": 22900
+    },
+    {
+      "epoch": 10.727611940298507,
+      "grad_norm": 0.00027732501621358097,
+      "learning_rate": 1.1402985074626867e-05,
+      "loss": 0.0,
+      "step": 23000
+    },
+    {
+      "epoch": 10.727611940298507,
+      "eval_loss": 2.747021596860577e-07,
+      "eval_runtime": 0.1614,
+      "eval_samples_per_second": 185.849,
+      "eval_steps_per_second": 24.78,
+      "step": 23000
+    },
+    {
+      "epoch": 10.774253731343283,
+      "grad_norm": 0.000188069578143768,
+      "learning_rate": 1.127860696517413e-05,
+      "loss": 0.0,
+      "step": 23100
+    },
+    {
+      "epoch": 10.774253731343283,
+      "eval_loss": 2.720941552070144e-07,
+      "eval_runtime": 0.1657,
+      "eval_samples_per_second": 181.099,
+      "eval_steps_per_second": 24.147,
+      "step": 23100
+    },
+    {
+      "epoch": 10.82089552238806,
+      "grad_norm": 0.000323320651659742,
+      "learning_rate": 1.1154228855721393e-05,
+      "loss": 0.0,
+      "step": 23200
+    },
+    {
+      "epoch": 10.82089552238806,
+      "eval_loss": 2.7039632755077037e-07,
+      "eval_runtime": 0.1651,
+      "eval_samples_per_second": 181.694,
+      "eval_steps_per_second": 24.226,
+      "step": 23200
+    },
+    {
+      "epoch": 10.867537313432836,
+      "grad_norm": 0.00014747037494089454,
+      "learning_rate": 1.1029850746268658e-05,
+      "loss": 0.0,
+      "step": 23300
+    },
+    {
+      "epoch": 10.867537313432836,
+      "eval_loss": 2.6682712928050023e-07,
+      "eval_runtime": 0.1662,
+      "eval_samples_per_second": 180.507,
+      "eval_steps_per_second": 24.068,
+      "step": 23300
+    },
+    {
+      "epoch": 10.914179104477611,
+      "grad_norm": 7.306891347980127e-05,
+      "learning_rate": 1.0905472636815922e-05,
+      "loss": 0.0,
+      "step": 23400
+    },
+    {
+      "epoch": 10.914179104477611,
+      "eval_loss": 2.6359310822954285e-07,
+      "eval_runtime": 0.1726,
+      "eval_samples_per_second": 173.796,
+      "eval_steps_per_second": 23.173,
+      "step": 23400
+    },
+    {
+      "epoch": 10.960820895522389,
+      "grad_norm": 5.098180554341525e-05,
+      "learning_rate": 1.0781094527363185e-05,
+      "loss": 0.0,
+      "step": 23500
+    },
+    {
+      "epoch": 10.960820895522389,
+      "eval_loss": 2.554411082655861e-07,
+      "eval_runtime": 0.1637,
+      "eval_samples_per_second": 183.313,
+      "eval_steps_per_second": 24.442,
+      "step": 23500
+    },
+    {
+      "epoch": 11.007462686567164,
+      "grad_norm": 9.442290320293978e-05,
+      "learning_rate": 1.0656716417910448e-05,
+      "loss": 0.0,
+      "step": 23600
+    },
+    {
+      "epoch": 11.007462686567164,
+      "eval_loss": 2.5533600478411245e-07,
+      "eval_runtime": 0.1606,
+      "eval_samples_per_second": 186.742,
+      "eval_steps_per_second": 24.899,
+      "step": 23600
+    },
+    {
+      "epoch": 11.05410447761194,
+      "grad_norm": 0.00022185999841894954,
+      "learning_rate": 1.0532338308457712e-05,
+      "loss": 0.0,
+      "step": 23700
+    },
+    {
+      "epoch": 11.05410447761194,
+      "eval_loss": 2.524078581700451e-07,
+      "eval_runtime": 0.1638,
+      "eval_samples_per_second": 183.127,
+      "eval_steps_per_second": 24.417,
+      "step": 23700
+    },
+    {
+      "epoch": 11.100746268656716,
+      "grad_norm": 0.00503884069621563,
+      "learning_rate": 1.0407960199004977e-05,
+      "loss": 0.0004,
+      "step": 23800
+    },
+    {
+      "epoch": 11.100746268656716,
+      "eval_loss": 2.567036858636129e-07,
+      "eval_runtime": 0.1643,
+      "eval_samples_per_second": 182.544,
+      "eval_steps_per_second": 24.339,
+      "step": 23800
+    },
+    {
+      "epoch": 11.147388059701493,
+      "grad_norm": 9.032327943714336e-05,
+      "learning_rate": 1.028358208955224e-05,
+      "loss": 0.0002,
+      "step": 23900
+    },
+    {
+      "epoch": 11.147388059701493,
+      "eval_loss": 2.859654841813608e-07,
+      "eval_runtime": 0.1632,
+      "eval_samples_per_second": 183.792,
+      "eval_steps_per_second": 24.506,
+      "step": 23900
+    },
+    {
+      "epoch": 11.194029850746269,
+      "grad_norm": 0.00012014710227958858,
+      "learning_rate": 1.016044776119403e-05,
+      "loss": 0.0,
+      "step": 24000
+    },
+    {
+      "epoch": 11.194029850746269,
+      "eval_loss": 2.840646402546554e-07,
+      "eval_runtime": 0.1633,
+      "eval_samples_per_second": 183.679,
+      "eval_steps_per_second": 24.49,
+      "step": 24000
+    },
+    {
+      "epoch": 11.240671641791044,
+      "grad_norm": 0.0031628520227968693,
+      "learning_rate": 1.0036069651741294e-05,
+      "loss": 0.0,
+      "step": 24100
+    },
+    {
+      "epoch": 11.240671641791044,
+      "eval_loss": 2.8124046025368443e-07,
+      "eval_runtime": 0.1622,
+      "eval_samples_per_second": 184.928,
+      "eval_steps_per_second": 24.657,
+      "step": 24100
+    },
+    {
+      "epoch": 11.287313432835822,
+      "grad_norm": 4.342416286817752e-05,
+      "learning_rate": 9.911691542288559e-06,
+      "loss": 0.0001,
+      "step": 24200
+    },
+    {
+      "epoch": 11.287313432835822,
+      "eval_loss": 2.839091735040711e-07,
+      "eval_runtime": 0.1665,
+      "eval_samples_per_second": 180.187,
+      "eval_steps_per_second": 24.025,
+      "step": 24200
+    },
+    {
+      "epoch": 11.333955223880597,
+      "grad_norm": 9.751073230290785e-05,
+      "learning_rate": 9.787313432835822e-06,
+      "loss": 0.0,
+      "step": 24300
+    },
+    {
+      "epoch": 11.333955223880597,
+      "eval_loss": 2.823463773893309e-07,
+      "eval_runtime": 0.1652,
+      "eval_samples_per_second": 181.594,
+      "eval_steps_per_second": 24.213,
+      "step": 24300
+    },
+    {
+      "epoch": 11.380597014925373,
+      "grad_norm": 0.00015635335876140743,
+      "learning_rate": 9.662935323383086e-06,
+      "loss": 0.0,
+      "step": 24400
+    },
+    {
+      "epoch": 11.380597014925373,
+      "eval_loss": 2.809511840951018e-07,
+      "eval_runtime": 0.1669,
+      "eval_samples_per_second": 179.702,
+      "eval_steps_per_second": 23.96,
+      "step": 24400
+    },
+    {
+      "epoch": 11.427238805970148,
+      "grad_norm": 0.00012053705722792074,
+      "learning_rate": 9.538557213930349e-06,
+      "loss": 0.0,
+      "step": 24500
+    },
+    {
+      "epoch": 11.427238805970148,
+      "eval_loss": 2.786693471534818e-07,
+      "eval_runtime": 0.1641,
+      "eval_samples_per_second": 182.826,
+      "eval_steps_per_second": 24.377,
+      "step": 24500
+    },
+    {
+      "epoch": 11.473880597014926,
+      "grad_norm": 6.743449193891138e-05,
+      "learning_rate": 9.414179104477614e-06,
+      "loss": 0.0,
+      "step": 24600
+    },
+    {
+      "epoch": 11.473880597014926,
+      "eval_loss": 2.7615345743470243e-07,
+      "eval_runtime": 0.1652,
+      "eval_samples_per_second": 181.635,
+      "eval_steps_per_second": 24.218,
+      "step": 24600
+    },
+    {
+      "epoch": 11.520522388059701,
+      "grad_norm": 0.0008915641228668392,
+      "learning_rate": 9.289800995024877e-06,
+      "loss": 0.0,
+      "step": 24700
+    },
+    {
+      "epoch": 11.520522388059701,
+      "eval_loss": 2.7316258410792216e-07,
+      "eval_runtime": 0.182,
+      "eval_samples_per_second": 164.79,
+      "eval_steps_per_second": 21.972,
+      "step": 24700
+    },
+    {
+      "epoch": 11.567164179104477,
+      "grad_norm": 0.0019023872446268797,
+      "learning_rate": 9.16542288557214e-06,
+      "loss": 0.0,
+      "step": 24800
+    },
+    {
+      "epoch": 11.567164179104477,
+      "eval_loss": 2.688062750166864e-07,
+      "eval_runtime": 0.1654,
+      "eval_samples_per_second": 181.342,
+      "eval_steps_per_second": 24.179,
+      "step": 24800
+    },
+    {
+      "epoch": 11.613805970149254,
+      "grad_norm": 0.0024738411884754896,
+      "learning_rate": 9.041044776119404e-06,
+      "loss": 0.0,
+      "step": 24900
+    },
+    {
+      "epoch": 11.613805970149254,
+      "eval_loss": 2.621562771309982e-07,
+      "eval_runtime": 0.1634,
+      "eval_samples_per_second": 183.65,
+      "eval_steps_per_second": 24.487,
+      "step": 24900
+    },
+    {
+      "epoch": 11.66044776119403,
+      "grad_norm": 0.0007770381635054946,
+      "learning_rate": 8.916666666666667e-06,
+      "loss": 0.0,
+      "step": 25000
+    },
+    {
+      "epoch": 11.66044776119403,
+      "eval_loss": 2.6587503043629113e-07,
+      "eval_runtime": 0.1642,
+      "eval_samples_per_second": 182.719,
+      "eval_steps_per_second": 24.363,
+      "step": 25000
+    },
+    {
+      "epoch": 11.707089552238806,
+      "grad_norm": 9.729754674481228e-05,
+      "learning_rate": 8.792288557213932e-06,
+      "loss": 0.0,
+      "step": 25100
+    },
+    {
+      "epoch": 11.707089552238806,
+      "eval_loss": 2.6003732500612387e-07,
+      "eval_runtime": 0.1636,
+      "eval_samples_per_second": 183.389,
+      "eval_steps_per_second": 24.452,
+      "step": 25100
+    },
+    {
+      "epoch": 11.753731343283581,
+      "grad_norm": 0.00033962438465096056,
+      "learning_rate": 8.667910447761195e-06,
+      "loss": 0.0,
+      "step": 25200
+    },
+    {
+      "epoch": 11.753731343283581,
+      "eval_loss": 2.570840251792106e-07,
+      "eval_runtime": 0.1754,
+      "eval_samples_per_second": 171.067,
+      "eval_steps_per_second": 22.809,
+      "step": 25200
+    },
+    {
+      "epoch": 11.800373134328359,
+      "grad_norm": 0.0037661008536815643,
+      "learning_rate": 8.543532338308459e-06,
+      "loss": 0.0,
+      "step": 25300
+    },
+    {
+      "epoch": 11.800373134328359,
+      "eval_loss": 2.5637064027250744e-07,
+      "eval_runtime": 0.1648,
+      "eval_samples_per_second": 182.061,
+      "eval_steps_per_second": 24.275,
+      "step": 25300
+    },
+    {
+      "epoch": 11.847014925373134,
+      "grad_norm": 7.473176083294675e-05,
+      "learning_rate": 8.419154228855722e-06,
+      "loss": 0.0,
+      "step": 25400
+    },
+    {
+      "epoch": 11.847014925373134,
+      "eval_loss": 2.5390116320522793e-07,
+      "eval_runtime": 0.1659,
+      "eval_samples_per_second": 180.827,
+      "eval_steps_per_second": 24.11,
+      "step": 25400
+    },
+    {
+      "epoch": 11.89365671641791,
+      "grad_norm": 0.00015807716408744454,
+      "learning_rate": 8.294776119402985e-06,
+      "loss": 0.0,
+      "step": 25500
+    },
+    {
+      "epoch": 11.89365671641791,
+      "eval_loss": 2.5069658704524045e-07,
+      "eval_runtime": 0.1638,
+      "eval_samples_per_second": 183.2,
+      "eval_steps_per_second": 24.427,
+      "step": 25500
+    },
+    {
+      "epoch": 11.940298507462687,
+      "grad_norm": 0.0018306206911802292,
+      "learning_rate": 8.170398009950249e-06,
+      "loss": 0.0,
+      "step": 25600
+    },
+    {
+      "epoch": 11.940298507462687,
+      "eval_loss": 2.496325635092944e-07,
+      "eval_runtime": 0.1664,
+      "eval_samples_per_second": 180.237,
+      "eval_steps_per_second": 24.032,
+      "step": 25600
+    },
+    {
+      "epoch": 11.986940298507463,
+      "grad_norm": 4.205930235912092e-05,
+      "learning_rate": 8.046019900497512e-06,
+      "loss": 0.0,
+      "step": 25700
+    },
+    {
+      "epoch": 11.986940298507463,
+      "eval_loss": 2.4730249492677103e-07,
+      "eval_runtime": 0.1636,
+      "eval_samples_per_second": 183.389,
+      "eval_steps_per_second": 24.452,
+      "step": 25700
+    },
+    {
+      "epoch": 12.033582089552239,
+      "grad_norm": 0.006276166532188654,
+      "learning_rate": 7.921641791044777e-06,
+      "loss": 0.0,
+      "step": 25800
+    },
+    {
+      "epoch": 12.033582089552239,
+      "eval_loss": 2.452173930578283e-07,
+      "eval_runtime": 0.1637,
+      "eval_samples_per_second": 183.207,
+      "eval_steps_per_second": 24.428,
+      "step": 25800
+    },
+    {
+      "epoch": 12.080223880597014,
+      "grad_norm": 0.0031838086433708668,
+      "learning_rate": 7.79726368159204e-06,
+      "loss": 0.0,
+      "step": 25900
+    },
+    {
+      "epoch": 12.080223880597014,
+      "eval_loss": 2.5401092784704815e-07,
+      "eval_runtime": 0.1636,
+      "eval_samples_per_second": 183.344,
+      "eval_steps_per_second": 24.446,
+      "step": 25900
+    },
+    {
+      "epoch": 12.126865671641792,
+      "grad_norm": 6.814413063693792e-05,
+      "learning_rate": 7.672885572139303e-06,
+      "loss": 0.0,
+      "step": 26000
+    },
+    {
+      "epoch": 12.126865671641792,
+      "eval_loss": 2.5293377348134527e-07,
+      "eval_runtime": 0.1613,
+      "eval_samples_per_second": 185.984,
+      "eval_steps_per_second": 24.798,
+      "step": 26000
+    },
+    {
+      "epoch": 12.173507462686567,
+      "grad_norm": 0.0002226163778686896,
+      "learning_rate": 7.549751243781095e-06,
+      "loss": 0.0,
+      "step": 26100
+    },
+    {
+      "epoch": 12.173507462686567,
+      "eval_loss": 2.4839519596753235e-07,
+      "eval_runtime": 0.164,
+      "eval_samples_per_second": 182.882,
+      "eval_steps_per_second": 24.384,
+      "step": 26100
+    },
+    {
+      "epoch": 12.220149253731343,
+      "grad_norm": 0.00015515003178734332,
+      "learning_rate": 7.4253731343283585e-06,
+      "loss": 0.0,
+      "step": 26200
+    },
+    {
+      "epoch": 12.220149253731343,
+      "eval_loss": 2.3797457515684073e-07,
+      "eval_runtime": 0.1632,
+      "eval_samples_per_second": 183.77,
+      "eval_steps_per_second": 24.503,
+      "step": 26200
+    },
+    {
+      "epoch": 12.26679104477612,
+      "grad_norm": 0.0004189789469819516,
+      "learning_rate": 7.300995024875623e-06,
+      "loss": 0.0,
+      "step": 26300
+    },
+    {
+      "epoch": 12.26679104477612,
+      "eval_loss": 2.4495119532730314e-07,
+      "eval_runtime": 0.1641,
+      "eval_samples_per_second": 182.814,
+      "eval_steps_per_second": 24.375,
+      "step": 26300
+    },
+    {
+      "epoch": 12.313432835820896,
+      "grad_norm": 0.0008414119947701693,
+      "learning_rate": 7.176616915422886e-06,
+      "loss": 0.0,
+      "step": 26400
+    },
+    {
+      "epoch": 12.313432835820896,
+      "eval_loss": 2.375207088789466e-07,
+      "eval_runtime": 0.1644,
+      "eval_samples_per_second": 182.441,
+      "eval_steps_per_second": 24.325,
+      "step": 26400
+    },
+    {
+      "epoch": 12.360074626865671,
+      "grad_norm": 0.004033736884593964,
+      "learning_rate": 7.052238805970149e-06,
+      "loss": 0.0,
+      "step": 26500
+    },
+    {
+      "epoch": 12.360074626865671,
+      "eval_loss": 2.3545504745925427e-07,
+      "eval_runtime": 0.1634,
+      "eval_samples_per_second": 183.625,
+      "eval_steps_per_second": 24.483,
+      "step": 26500
+    },
+    {
+      "epoch": 12.406716417910447,
+      "grad_norm": 0.0006975606665946543,
+      "learning_rate": 6.927860696517413e-06,
+      "loss": 0.0002,
+      "step": 26600
+    },
+    {
+      "epoch": 12.406716417910447,
+      "eval_loss": 2.2062947380163678e-07,
+      "eval_runtime": 0.1647,
+      "eval_samples_per_second": 182.126,
+      "eval_steps_per_second": 24.283,
+      "step": 26600
+    },
+    {
+      "epoch": 12.453358208955224,
+      "grad_norm": 0.00041404165676794946,
+      "learning_rate": 6.803482587064677e-06,
+      "loss": 0.0001,
+      "step": 26700
+    },
+    {
+      "epoch": 12.453358208955224,
+      "eval_loss": 2.187805563380607e-07,
+      "eval_runtime": 0.1659,
+      "eval_samples_per_second": 180.878,
+      "eval_steps_per_second": 24.117,
+      "step": 26700
+    },
+    {
+      "epoch": 12.5,
+      "grad_norm": 0.00021756745991297066,
+      "learning_rate": 6.679104477611941e-06,
+      "loss": 0.0,
+      "step": 26800
+    },
+    {
+      "epoch": 12.5,
+      "eval_loss": 2.150947295831429e-07,
+      "eval_runtime": 0.1631,
+      "eval_samples_per_second": 183.961,
+      "eval_steps_per_second": 24.528,
+      "step": 26800
+    },
+    {
+      "epoch": 12.546641791044776,
+      "grad_norm": 0.0006945555796846747,
+      "learning_rate": 6.554726368159204e-06,
+      "loss": 0.0,
+      "step": 26900
+    },
+    {
+      "epoch": 12.546641791044776,
+      "eval_loss": 2.1574922470790625e-07,
+      "eval_runtime": 0.1662,
+      "eval_samples_per_second": 180.505,
+      "eval_steps_per_second": 24.067,
+      "step": 26900
+    },
+    {
+      "epoch": 12.593283582089553,
+      "grad_norm": 0.0018626791425049305,
+      "learning_rate": 6.430348258706468e-06,
+      "loss": 0.0,
+      "step": 27000
+    },
+    {
+      "epoch": 12.593283582089553,
+      "eval_loss": 2.1518117421237548e-07,
+      "eval_runtime": 0.1668,
+      "eval_samples_per_second": 179.805,
+      "eval_steps_per_second": 23.974,
+      "step": 27000
+    },
+    {
+      "epoch": 12.639925373134329,
+      "grad_norm": 8.379531936952844e-05,
+      "learning_rate": 6.3059701492537316e-06,
+      "loss": 0.0,
+      "step": 27100
+    },
+    {
+      "epoch": 12.639925373134329,
+      "eval_loss": 2.1498192381841363e-07,
+      "eval_runtime": 0.1636,
+      "eval_samples_per_second": 183.376,
+      "eval_steps_per_second": 24.45,
+      "step": 27100
+    },
+    {
+      "epoch": 12.686567164179104,
+      "grad_norm": 7.26276048226282e-05,
+      "learning_rate": 6.181592039800996e-06,
+      "loss": 0.0,
+      "step": 27200
+    },
+    {
+      "epoch": 12.686567164179104,
+      "eval_loss": 2.13713647667646e-07,
+      "eval_runtime": 0.1663,
+      "eval_samples_per_second": 180.45,
+      "eval_steps_per_second": 24.06,
+      "step": 27200
+    },
+    {
+      "epoch": 12.73320895522388,
+      "grad_norm": 0.0001676314859651029,
+      "learning_rate": 6.057213930348259e-06,
+      "loss": 0.0,
+      "step": 27300
+    },
+    {
+      "epoch": 12.73320895522388,
+      "eval_loss": 2.1666031102540728e-07,
+      "eval_runtime": 0.1669,
+      "eval_samples_per_second": 179.72,
+      "eval_steps_per_second": 23.963,
+      "step": 27300
+    },
+    {
+      "epoch": 12.779850746268657,
+      "grad_norm": 4.116259515285492e-05,
+      "learning_rate": 5.932835820895523e-06,
+      "loss": 0.0,
+      "step": 27400
+    },
+    {
+      "epoch": 12.779850746268657,
+      "eval_loss": 2.157899245958106e-07,
+      "eval_runtime": 0.1667,
+      "eval_samples_per_second": 179.968,
+      "eval_steps_per_second": 23.996,
+      "step": 27400
+    },
+    {
+      "epoch": 12.826492537313433,
+      "grad_norm": 0.00012961241009179503,
+      "learning_rate": 5.8084577114427864e-06,
+      "loss": 0.0,
+      "step": 27500
+    },
+    {
+      "epoch": 12.826492537313433,
+      "eval_loss": 2.1415317519313248e-07,
+      "eval_runtime": 0.1688,
+      "eval_samples_per_second": 177.772,
+      "eval_steps_per_second": 23.703,
+      "step": 27500
+    },
+    {
+      "epoch": 12.873134328358208,
+      "grad_norm": 0.00015239656204357743,
+      "learning_rate": 5.68407960199005e-06,
+      "loss": 0.0,
+      "step": 27600
+    },
+    {
+      "epoch": 12.873134328358208,
+      "eval_loss": 2.1278803785662603e-07,
+      "eval_runtime": 0.1654,
+      "eval_samples_per_second": 181.421,
+      "eval_steps_per_second": 24.189,
+      "step": 27600
+    },
+    {
+      "epoch": 12.919776119402986,
+      "grad_norm": 0.0003725362184923142,
+      "learning_rate": 5.559701492537314e-06,
+      "loss": 0.0,
+      "step": 27700
+    },
+    {
+      "epoch": 12.919776119402986,
+      "eval_loss": 2.1224755641924276e-07,
+      "eval_runtime": 0.1695,
+      "eval_samples_per_second": 176.969,
+      "eval_steps_per_second": 23.596,
+      "step": 27700
+    },
+    {
+      "epoch": 12.966417910447761,
+      "grad_norm": 0.00010239533003186807,
+      "learning_rate": 5.435323383084577e-06,
+      "loss": 0.0,
+      "step": 27800
+    },
+    {
+      "epoch": 12.966417910447761,
+      "eval_loss": 2.121297200119443e-07,
+      "eval_runtime": 0.1713,
+      "eval_samples_per_second": 175.102,
+      "eval_steps_per_second": 23.347,
+      "step": 27800
+    },
+    {
+      "epoch": 13.013059701492537,
+      "grad_norm": 0.0025495258159935474,
+      "learning_rate": 5.310945273631841e-06,
+      "loss": 0.0,
+      "step": 27900
+    },
+    {
+      "epoch": 13.013059701492537,
+      "eval_loss": 2.0902733410821384e-07,
+      "eval_runtime": 0.1664,
+      "eval_samples_per_second": 180.279,
+      "eval_steps_per_second": 24.037,
+      "step": 27900
+    },
+    {
+      "epoch": 13.059701492537313,
+      "grad_norm": 0.00011632608948275447,
+      "learning_rate": 5.187810945273632e-06,
+      "loss": 0.0,
+      "step": 28000
+    },
+    {
+      "epoch": 13.059701492537313,
+      "eval_loss": 2.0934349720391765e-07,
+      "eval_runtime": 0.1699,
+      "eval_samples_per_second": 176.597,
+      "eval_steps_per_second": 23.546,
+      "step": 28000
+    },
+    {
+      "epoch": 13.10634328358209,
+      "grad_norm": 0.0007514033932238817,
+      "learning_rate": 5.063432835820896e-06,
+      "loss": 0.0,
+      "step": 28100
+    },
+    {
+      "epoch": 13.10634328358209,
+      "eval_loss": 2.073689557846592e-07,
+      "eval_runtime": 0.1657,
+      "eval_samples_per_second": 181.029,
+      "eval_steps_per_second": 24.137,
+      "step": 28100
+    },
+    {
+      "epoch": 13.152985074626866,
+      "grad_norm": 0.0023394667077809572,
+      "learning_rate": 4.93905472636816e-06,
+      "loss": 0.0,
+      "step": 28200
+    },
+    {
+      "epoch": 13.152985074626866,
+      "eval_loss": 2.0222093155553011e-07,
+      "eval_runtime": 0.1659,
+      "eval_samples_per_second": 180.828,
+      "eval_steps_per_second": 24.11,
+      "step": 28200
+    },
+    {
+      "epoch": 13.199626865671641,
+      "grad_norm": 7.573967013740912e-05,
+      "learning_rate": 4.814676616915424e-06,
+      "loss": 0.0,
+      "step": 28300
+    },
+    {
+      "epoch": 13.199626865671641,
+      "eval_loss": 2.0147807333614765e-07,
+      "eval_runtime": 0.1637,
+      "eval_samples_per_second": 183.213,
+      "eval_steps_per_second": 24.428,
+      "step": 28300
+    },
+    {
+      "epoch": 13.246268656716419,
+      "grad_norm": 0.00038927345303818583,
+      "learning_rate": 4.690298507462687e-06,
+      "loss": 0.0,
+      "step": 28400
+    },
+    {
+      "epoch": 13.246268656716419,
+      "eval_loss": 2.0081444063180243e-07,
+      "eval_runtime": 0.1655,
+      "eval_samples_per_second": 181.315,
+      "eval_steps_per_second": 24.175,
+      "step": 28400
+    },
+    {
+      "epoch": 13.292910447761194,
+      "grad_norm": 4.6674933400936425e-05,
+      "learning_rate": 4.56592039800995e-06,
+      "loss": 0.0,
+      "step": 28500
+    },
+    {
+      "epoch": 13.292910447761194,
+      "eval_loss": 1.9368634696093068e-07,
+      "eval_runtime": 0.164,
+      "eval_samples_per_second": 182.957,
+      "eval_steps_per_second": 24.394,
+      "step": 28500
+    },
+    {
+      "epoch": 13.33955223880597,
+      "grad_norm": 0.0003617222246248275,
+      "learning_rate": 4.441542288557214e-06,
+      "loss": 0.0,
+      "step": 28600
+    },
+    {
+      "epoch": 13.33955223880597,
+      "eval_loss": 1.9277055685051891e-07,
+      "eval_runtime": 0.1664,
+      "eval_samples_per_second": 180.32,
+      "eval_steps_per_second": 24.043,
+      "step": 28600
+    },
+    {
+      "epoch": 13.386194029850746,
+      "grad_norm": 0.0003852724621538073,
+      "learning_rate": 4.317164179104478e-06,
+      "loss": 0.0,
+      "step": 28700
+    },
+    {
+      "epoch": 13.386194029850746,
+      "eval_loss": 1.9214044755244686e-07,
+      "eval_runtime": 0.1668,
+      "eval_samples_per_second": 179.9,
+      "eval_steps_per_second": 23.987,
+      "step": 28700
+    },
+    {
+      "epoch": 13.432835820895523,
+      "grad_norm": 8.544667798560113e-05,
+      "learning_rate": 4.192786069651741e-06,
+      "loss": 0.0,
+      "step": 28800
+    },
+    {
+      "epoch": 13.432835820895523,
+      "eval_loss": 1.926564152654464e-07,
+      "eval_runtime": 0.1658,
+      "eval_samples_per_second": 180.893,
+      "eval_steps_per_second": 24.119,
+      "step": 28800
+    },
+    {
+      "epoch": 13.479477611940299,
+      "grad_norm": 0.0006275973282754421,
+      "learning_rate": 4.068407960199005e-06,
+      "loss": 0.0,
+      "step": 28900
+    },
+    {
+      "epoch": 13.479477611940299,
+      "eval_loss": 1.9328433609189233e-07,
+      "eval_runtime": 0.1709,
+      "eval_samples_per_second": 175.591,
+      "eval_steps_per_second": 23.412,
+      "step": 28900
+    },
+    {
+      "epoch": 13.526119402985074,
+      "grad_norm": 0.00011571097275009379,
+      "learning_rate": 3.9440298507462686e-06,
+      "loss": 0.0,
+      "step": 29000
+    },
+    {
+      "epoch": 13.526119402985074,
+      "eval_loss": 1.924393444596717e-07,
+      "eval_runtime": 0.1653,
+      "eval_samples_per_second": 181.482,
+      "eval_steps_per_second": 24.198,
+      "step": 29000
+    },
+    {
+      "epoch": 13.572761194029852,
+      "grad_norm": 0.0030688499100506306,
+      "learning_rate": 3.819651741293533e-06,
+      "loss": 0.0,
+      "step": 29100
+    },
+    {
+      "epoch": 13.572761194029852,
+      "eval_loss": 1.9209718971069378e-07,
+      "eval_runtime": 0.1677,
+      "eval_samples_per_second": 178.88,
+      "eval_steps_per_second": 23.851,
+      "step": 29100
+    },
+    {
+      "epoch": 13.619402985074627,
+      "grad_norm": 0.00018914089014288038,
+      "learning_rate": 3.695273631840796e-06,
+      "loss": 0.0,
+      "step": 29200
+    },
+    {
+      "epoch": 13.619402985074627,
+      "eval_loss": 1.9149970853504783e-07,
+      "eval_runtime": 0.1652,
+      "eval_samples_per_second": 181.641,
+      "eval_steps_per_second": 24.219,
+      "step": 29200
+    },
+    {
+      "epoch": 13.666044776119403,
+      "grad_norm": 0.00021034800738561898,
+      "learning_rate": 3.5708955223880597e-06,
+      "loss": 0.0,
+      "step": 29300
+    },
+    {
+      "epoch": 13.666044776119403,
+      "eval_loss": 1.9132801298837876e-07,
+      "eval_runtime": 0.1656,
+      "eval_samples_per_second": 181.151,
+      "eval_steps_per_second": 24.154,
+      "step": 29300
+    },
+    {
+      "epoch": 13.712686567164178,
+      "grad_norm": 0.0001684948947513476,
+      "learning_rate": 3.4465174129353234e-06,
+      "loss": 0.0,
+      "step": 29400
+    },
+    {
+      "epoch": 13.712686567164178,
+      "eval_loss": 1.9041186760659912e-07,
+      "eval_runtime": 0.1657,
+      "eval_samples_per_second": 181.048,
+      "eval_steps_per_second": 24.14,
+      "step": 29400
+    },
+    {
+      "epoch": 13.759328358208956,
+      "grad_norm": 0.00033117341808974743,
+      "learning_rate": 3.322139303482587e-06,
+      "loss": 0.0,
+      "step": 29500
+    },
+    {
+      "epoch": 13.759328358208956,
+      "eval_loss": 1.8998919415480486e-07,
+      "eval_runtime": 0.1684,
+      "eval_samples_per_second": 178.197,
+      "eval_steps_per_second": 23.76,
+      "step": 29500
+    },
+    {
+      "epoch": 13.805970149253731,
+      "grad_norm": 7.936067413538694e-05,
+      "learning_rate": 3.197761194029851e-06,
+      "loss": 0.0001,
+      "step": 29600
+    },
+    {
+      "epoch": 13.805970149253731,
+      "eval_loss": 1.7708023847262666e-07,
+      "eval_runtime": 0.1876,
+      "eval_samples_per_second": 159.875,
+      "eval_steps_per_second": 21.317,
+      "step": 29600
+    },
+    {
+      "epoch": 13.852611940298507,
+      "grad_norm": 0.00019114046881441027,
+      "learning_rate": 3.0733830845771146e-06,
+      "loss": 0.0,
+      "step": 29700
+    },
+    {
+      "epoch": 13.852611940298507,
+      "eval_loss": 1.7666478413502773e-07,
+      "eval_runtime": 0.1663,
+      "eval_samples_per_second": 180.348,
+      "eval_steps_per_second": 24.046,
+      "step": 29700
+    },
+    {
+      "epoch": 13.899253731343283,
+      "grad_norm": 0.00010124894470209256,
+      "learning_rate": 2.9490049751243783e-06,
+      "loss": 0.0,
+      "step": 29800
+    },
+    {
+      "epoch": 13.899253731343283,
+      "eval_loss": 1.7688346076738526e-07,
+      "eval_runtime": 0.1653,
+      "eval_samples_per_second": 181.441,
+      "eval_steps_per_second": 24.192,
+      "step": 29800
+    },
+    {
+      "epoch": 13.94589552238806,
+      "grad_norm": 0.000508667784743011,
+      "learning_rate": 2.824626865671642e-06,
+      "loss": 0.0001,
+      "step": 29900
+    },
+    {
+      "epoch": 13.94589552238806,
+      "eval_loss": 1.8212197971934074e-07,
+      "eval_runtime": 0.1667,
+      "eval_samples_per_second": 179.986,
+      "eval_steps_per_second": 23.998,
+      "step": 29900
+    },
+    {
+      "epoch": 13.992537313432836,
+      "grad_norm": 0.0005743975634686649,
+      "learning_rate": 2.7002487562189058e-06,
+      "loss": 0.0,
+      "step": 30000
+    },
+    {
+      "epoch": 13.992537313432836,
+      "eval_loss": 1.8214987562714668e-07,
+      "eval_runtime": 0.1642,
+      "eval_samples_per_second": 182.734,
+      "eval_steps_per_second": 24.364,
+      "step": 30000
+    },
+    {
+      "epoch": 14.039179104477611,
+      "grad_norm": 0.0024433776270598173,
+      "learning_rate": 2.5758706467661695e-06,
+      "loss": 0.0,
+      "step": 30100
+    },
+    {
+      "epoch": 14.039179104477611,
+      "eval_loss": 1.819559116711389e-07,
+      "eval_runtime": 0.1681,
+      "eval_samples_per_second": 178.453,
+      "eval_steps_per_second": 23.794,
+      "step": 30100
+    },
+    {
+      "epoch": 14.085820895522389,
+      "grad_norm": 6.446899351431057e-05,
+      "learning_rate": 2.4514925373134328e-06,
+      "loss": 0.0,
+      "step": 30200
+    },
+    {
+      "epoch": 14.085820895522389,
+      "eval_loss": 1.817829797801096e-07,
+      "eval_runtime": 0.1665,
+      "eval_samples_per_second": 180.178,
+      "eval_steps_per_second": 24.024,
+      "step": 30200
+    },
+    {
+      "epoch": 14.132462686567164,
+      "grad_norm": 9.17048382689245e-05,
+      "learning_rate": 2.3271144278606965e-06,
+      "loss": 0.0,
+      "step": 30300
+    },
+    {
+      "epoch": 14.132462686567164,
+      "eval_loss": 1.822889856839538e-07,
+      "eval_runtime": 0.1655,
+      "eval_samples_per_second": 181.316,
+      "eval_steps_per_second": 24.175,
+      "step": 30300
+    },
+    {
+      "epoch": 14.17910447761194,
+      "grad_norm": 0.00043933966662734747,
+      "learning_rate": 2.20273631840796e-06,
+      "loss": 0.0,
+      "step": 30400
+    },
+    {
+      "epoch": 14.17910447761194,
+      "eval_loss": 1.820878310354601e-07,
+      "eval_runtime": 0.1621,
+      "eval_samples_per_second": 185.117,
+      "eval_steps_per_second": 24.682,
+      "step": 30400
+    },
+    {
+      "epoch": 14.225746268656717,
+      "grad_norm": 0.0007023545331321657,
+      "learning_rate": 2.078358208955224e-06,
+      "loss": 0.0,
+      "step": 30500
+    },
+    {
+      "epoch": 14.225746268656717,
+      "eval_loss": 1.8134495860522293e-07,
+      "eval_runtime": 0.166,
+      "eval_samples_per_second": 180.697,
+      "eval_steps_per_second": 24.093,
+      "step": 30500
+    },
+    {
+      "epoch": 14.272388059701493,
+      "grad_norm": 0.0002605569316074252,
+      "learning_rate": 1.9539800995024877e-06,
+      "loss": 0.0002,
+      "step": 30600
+    },
+    {
+      "epoch": 14.272388059701493,
+      "eval_loss": 1.7974176103052741e-07,
+      "eval_runtime": 0.1664,
+      "eval_samples_per_second": 180.282,
+      "eval_steps_per_second": 24.038,
+      "step": 30600
+    },
+    {
+      "epoch": 14.319029850746269,
+      "grad_norm": 0.00020215619588270783,
+      "learning_rate": 1.8296019900497514e-06,
+      "loss": 0.0,
+      "step": 30700
+    },
+    {
+      "epoch": 14.319029850746269,
+      "eval_loss": 1.7835911592101183e-07,
+      "eval_runtime": 0.1678,
+      "eval_samples_per_second": 178.732,
+      "eval_steps_per_second": 23.831,
+      "step": 30700
+    },
+    {
+      "epoch": 14.365671641791044,
+      "grad_norm": 0.0002130908687831834,
+      "learning_rate": 1.705223880597015e-06,
+      "loss": 0.0,
+      "step": 30800
+    },
+    {
+      "epoch": 14.365671641791044,
+      "eval_loss": 1.7787193939966528e-07,
+      "eval_runtime": 0.1641,
+      "eval_samples_per_second": 182.803,
+      "eval_steps_per_second": 24.374,
+      "step": 30800
+    },
+    {
+      "epoch": 14.412313432835822,
+      "grad_norm": 0.000937594857532531,
+      "learning_rate": 1.5808457711442788e-06,
+      "loss": 0.0,
+      "step": 30900
+    },
+    {
+      "epoch": 14.412313432835822,
+      "eval_loss": 1.7799224849568418e-07,
+      "eval_runtime": 0.1985,
+      "eval_samples_per_second": 151.106,
+      "eval_steps_per_second": 20.148,
+      "step": 30900
+    },
+    {
+      "epoch": 14.458955223880597,
+      "grad_norm": 8.0404024629388e-05,
+      "learning_rate": 1.4564676616915423e-06,
+      "loss": 0.0,
+      "step": 31000
+    },
+    {
+      "epoch": 14.458955223880597,
+      "eval_loss": 1.7809625774134474e-07,
+      "eval_runtime": 0.1637,
+      "eval_samples_per_second": 183.297,
+      "eval_steps_per_second": 24.44,
+      "step": 31000
+    },
+    {
+      "epoch": 14.505597014925373,
+      "grad_norm": 0.00047697682748548687,
+      "learning_rate": 1.332089552238806e-06,
+      "loss": 0.0,
+      "step": 31100
+    },
+    {
+      "epoch": 14.505597014925373,
+      "eval_loss": 1.7796874374198524e-07,
+      "eval_runtime": 0.1644,
+      "eval_samples_per_second": 182.438,
+      "eval_steps_per_second": 24.325,
+      "step": 31100
+    },
+    {
+      "epoch": 14.552238805970148,
+      "grad_norm": 0.0002544411108829081,
+      "learning_rate": 1.2077114427860698e-06,
+      "loss": 0.0,
+      "step": 31200
+    },
+    {
+      "epoch": 14.552238805970148,
+      "eval_loss": 1.771030895270087e-07,
+      "eval_runtime": 0.1737,
+      "eval_samples_per_second": 172.756,
+      "eval_steps_per_second": 23.034,
+      "step": 31200
+    },
+    {
+      "epoch": 14.598880597014926,
+      "grad_norm": 0.00014778469630982727,
+      "learning_rate": 1.0833333333333335e-06,
+      "loss": 0.0,
+      "step": 31300
+    },
+    {
+      "epoch": 14.598880597014926,
+      "eval_loss": 1.7678004837762273e-07,
+      "eval_runtime": 0.1628,
+      "eval_samples_per_second": 184.321,
+      "eval_steps_per_second": 24.576,
+      "step": 31300
+    },
+    {
+      "epoch": 14.645522388059701,
+      "grad_norm": 9.278374636778608e-05,
+      "learning_rate": 9.589552238805972e-07,
+      "loss": 0.0005,
+      "step": 31400
+    },
+    {
+      "epoch": 14.645522388059701,
+      "eval_loss": 1.776275126985638e-07,
+      "eval_runtime": 0.1672,
+      "eval_samples_per_second": 179.379,
+      "eval_steps_per_second": 23.917,
+      "step": 31400
+    },
+    {
+      "epoch": 14.692164179104477,
+      "grad_norm": 0.00015798135427758098,
+      "learning_rate": 8.345771144278608e-07,
+      "loss": 0.0001,
+      "step": 31500
+    },
+    {
+      "epoch": 14.692164179104477,
+      "eval_loss": 1.7723084511089837e-07,
+      "eval_runtime": 0.1647,
+      "eval_samples_per_second": 182.127,
+      "eval_steps_per_second": 24.284,
+      "step": 31500
+    },
+    {
+      "epoch": 14.738805970149254,
+      "grad_norm": 4.799765156349167e-05,
+      "learning_rate": 7.101990049751243e-07,
+      "loss": 0.0,
+      "step": 31600
+    },
+    {
+      "epoch": 14.738805970149254,
+      "eval_loss": 1.7615087699596188e-07,
+      "eval_runtime": 0.1648,
+      "eval_samples_per_second": 182.059,
+      "eval_steps_per_second": 24.275,
+      "step": 31600
+    },
+    {
+      "epoch": 14.78544776119403,
+      "grad_norm": 0.0001866283710114658,
+      "learning_rate": 5.858208955223881e-07,
+      "loss": 0.0001,
+      "step": 31700
+    },
+    {
+      "epoch": 14.78544776119403,
+      "eval_loss": 1.769698627640537e-07,
+      "eval_runtime": 0.1647,
+      "eval_samples_per_second": 182.204,
+      "eval_steps_per_second": 24.294,
+      "step": 31700
+    },
+    {
+      "epoch": 14.832089552238806,
+      "grad_norm": 0.00022171168529894203,
+      "learning_rate": 4.614427860696518e-07,
+      "loss": 0.0,
+      "step": 31800
+    },
+    {
+      "epoch": 14.832089552238806,
+      "eval_loss": 1.7630063098295068e-07,
+      "eval_runtime": 0.1655,
+      "eval_samples_per_second": 181.321,
+      "eval_steps_per_second": 24.176,
+      "step": 31800
+    },
+    {
+      "epoch": 14.878731343283581,
+      "grad_norm": 0.0003983838832937181,
+      "learning_rate": 3.370646766169155e-07,
+      "loss": 0.0,
+      "step": 31900
+    },
+    {
+      "epoch": 14.878731343283581,
+      "eval_loss": 1.7645477612404648e-07,
+      "eval_runtime": 0.1651,
+      "eval_samples_per_second": 181.747,
+      "eval_steps_per_second": 24.233,
+      "step": 31900
+    },
+    {
+      "epoch": 14.925373134328359,
+      "grad_norm": 0.00016803256585262716,
+      "learning_rate": 2.1268656716417912e-07,
+      "loss": 0.0,
+      "step": 32000
+    },
+    {
+      "epoch": 14.925373134328359,
+      "eval_loss": 1.7600704893538932e-07,
+      "eval_runtime": 0.1658,
+      "eval_samples_per_second": 180.892,
+      "eval_steps_per_second": 24.119,
+      "step": 32000
+    }
+  ],
+  "logging_steps": 100,
+  "max_steps": 32160,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 15,
+  "save_steps": 200,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.257078909708288e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}