diff --git "a/checkpoint-8800/trainer_state.json" "b/checkpoint-8800/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-8800/trainer_state.json"
@@ -0,0 +1,3914 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.1912416663846492,
+  "eval_steps": 100,
+  "global_step": 8800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0001353683711800738,
+      "eval_loss": 0.05271552875638008,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 121.9881,
+      "eval_samples_per_second": 4.099,
+      "eval_steps_per_second": 1.025,
+      "step": 1
+    },
+    {
+      "epoch": 0.0027073674236014755,
+      "grad_norm": 4.584434986114502,
+      "learning_rate": 1e-05,
+      "loss": 0.1125,
+      "step": 20
+    },
+    {
+      "epoch": 0.005414734847202951,
+      "grad_norm": 8.236103057861328,
+      "learning_rate": 2e-05,
+      "loss": 0.1108,
+      "step": 40
+    },
+    {
+      "epoch": 0.008122102270804427,
+      "grad_norm": 29.931549072265625,
+      "learning_rate": 3e-05,
+      "loss": 0.1274,
+      "step": 60
+    },
+    {
+      "epoch": 0.010829469694405902,
+      "grad_norm": 0.10951080918312073,
+      "learning_rate": 4e-05,
+      "loss": 0.096,
+      "step": 80
+    },
+    {
+      "epoch": 0.013536837118007378,
+      "grad_norm": 3.041987419128418,
+      "learning_rate": 5e-05,
+      "loss": 0.1144,
+      "step": 100
+    },
+    {
+      "epoch": 0.013536837118007378,
+      "eval_loss": 0.05268385633826256,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 126.9017,
+      "eval_samples_per_second": 3.94,
+      "eval_steps_per_second": 0.985,
+      "step": 100
+    },
+    {
+      "epoch": 0.016244204541608853,
+      "grad_norm": 5.450573444366455,
+      "learning_rate": 4.997285190715352e-05,
+      "loss": 0.1957,
+      "step": 120
+    },
+    {
+      "epoch": 0.01895157196521033,
+      "grad_norm": 10.190834999084473,
+      "learning_rate": 4.994570381430705e-05,
+      "loss": 0.0732,
+      "step": 140
+    },
+    {
+      "epoch": 0.021658939388811804,
+      "grad_norm": 0.6384797096252441,
+      "learning_rate": 4.991855572146057e-05,
+      "loss": 0.2031,
+      "step": 160
+    },
+    {
+      "epoch": 0.02436630681241328,
+      "grad_norm": 33.95022964477539,
+      "learning_rate": 4.989140762861409e-05,
+      "loss": 0.3639,
+      "step": 180
+    },
+    {
+      "epoch": 0.027073674236014755,
+      "grad_norm": 0.5610571503639221,
+      "learning_rate": 4.9864259535767615e-05,
+      "loss": 0.2497,
+      "step": 200
+    },
+    {
+      "epoch": 0.027073674236014755,
+      "eval_loss": 0.05859573185443878,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 126.668,
+      "eval_samples_per_second": 3.947,
+      "eval_steps_per_second": 0.987,
+      "step": 200
+    },
+    {
+      "epoch": 0.02978104165961623,
+      "grad_norm": 27.80652618408203,
+      "learning_rate": 4.983711144292114e-05,
+      "loss": 0.2845,
+      "step": 220
+    },
+    {
+      "epoch": 0.032488409083217706,
+      "grad_norm": 0.5620715022087097,
+      "learning_rate": 4.980996335007466e-05,
+      "loss": 0.11,
+      "step": 240
+    },
+    {
+      "epoch": 0.03519577650681918,
+      "grad_norm": 0.33994266390800476,
+      "learning_rate": 4.978281525722818e-05,
+      "loss": 0.3338,
+      "step": 260
+    },
+    {
+      "epoch": 0.03790314393042066,
+      "grad_norm": 0.8733130097389221,
+      "learning_rate": 4.9755667164381704e-05,
+      "loss": 0.2284,
+      "step": 280
+    },
+    {
+      "epoch": 0.04061051135402213,
+      "grad_norm": 6.556461811065674,
+      "learning_rate": 4.972851907153523e-05,
+      "loss": 0.1527,
+      "step": 300
+    },
+    {
+      "epoch": 0.04061051135402213,
+      "eval_loss": 0.053429730236530304,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 127.1913,
+      "eval_samples_per_second": 3.931,
+      "eval_steps_per_second": 0.983,
+      "step": 300
+    },
+    {
+      "epoch": 0.04331787877762361,
+      "grad_norm": 0.0169974397867918,
+      "learning_rate": 4.970137097868875e-05,
+      "loss": 0.0819,
+      "step": 320
+    },
+    {
+      "epoch": 0.046025246201225084,
+      "grad_norm": 10.627715110778809,
+      "learning_rate": 4.967422288584227e-05,
+      "loss": 0.2767,
+      "step": 340
+    },
+    {
+      "epoch": 0.04873261362482656,
+      "grad_norm": 3.286104440689087,
+      "learning_rate": 4.964707479299579e-05,
+      "loss": 0.1822,
+      "step": 360
+    },
+    {
+      "epoch": 0.051439981048428035,
+      "grad_norm": 9.183119773864746,
+      "learning_rate": 4.9619926700149316e-05,
+      "loss": 0.2516,
+      "step": 380
+    },
+    {
+      "epoch": 0.05414734847202951,
+      "grad_norm": 31.08264923095703,
+      "learning_rate": 4.9592778607302834e-05,
+      "loss": 0.1377,
+      "step": 400
+    },
+    {
+      "epoch": 0.05414734847202951,
+      "eval_loss": 0.06496748328208923,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 126.6592,
+      "eval_samples_per_second": 3.948,
+      "eval_steps_per_second": 0.987,
+      "step": 400
+    },
+    {
+      "epoch": 0.056854715895630986,
+      "grad_norm": 0.8989093899726868,
+      "learning_rate": 4.9565630514456364e-05,
+      "loss": 0.0798,
+      "step": 420
+    },
+    {
+      "epoch": 0.05956208331923246,
+      "grad_norm": 5.862271785736084,
+      "learning_rate": 4.953848242160988e-05,
+      "loss": 0.0596,
+      "step": 440
+    },
+    {
+      "epoch": 0.06226945074283394,
+      "grad_norm": 29.281267166137695,
+      "learning_rate": 4.951133432876341e-05,
+      "loss": 0.2643,
+      "step": 460
+    },
+    {
+      "epoch": 0.06497681816643541,
+      "grad_norm": 1.005914330482483,
+      "learning_rate": 4.948418623591693e-05,
+      "loss": 0.4338,
+      "step": 480
+    },
+    {
+      "epoch": 0.06768418559003689,
+      "grad_norm": 1.3645344972610474,
+      "learning_rate": 4.945703814307045e-05,
+      "loss": 0.0815,
+      "step": 500
+    },
+    {
+      "epoch": 0.06768418559003689,
+      "eval_loss": 0.06759323924779892,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 135.4497,
+      "eval_samples_per_second": 3.691,
+      "eval_steps_per_second": 0.923,
+      "step": 500
+    },
+    {
+      "epoch": 0.07039155301363836,
+      "grad_norm": 16.54131317138672,
+      "learning_rate": 4.942989005022398e-05,
+      "loss": 0.156,
+      "step": 520
+    },
+    {
+      "epoch": 0.07309892043723984,
+      "grad_norm": 1.9615036249160767,
+      "learning_rate": 4.9402741957377494e-05,
+      "loss": 0.2292,
+      "step": 540
+    },
+    {
+      "epoch": 0.07580628786084131,
+      "grad_norm": 1.1833112239837646,
+      "learning_rate": 4.937559386453102e-05,
+      "loss": 0.1009,
+      "step": 560
+    },
+    {
+      "epoch": 0.07851365528444279,
+      "grad_norm": 0.9788858294487,
+      "learning_rate": 4.934844577168454e-05,
+      "loss": 0.047,
+      "step": 580
+    },
+    {
+      "epoch": 0.08122102270804427,
+      "grad_norm": 0.8096566200256348,
+      "learning_rate": 4.9321297678838066e-05,
+      "loss": 0.1603,
+      "step": 600
+    },
+    {
+      "epoch": 0.08122102270804427,
+      "eval_loss": 0.0586598739027977,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 126.1816,
+      "eval_samples_per_second": 3.963,
+      "eval_steps_per_second": 0.991,
+      "step": 600
+    },
+    {
+      "epoch": 0.08392839013164574,
+      "grad_norm": 0.018350690603256226,
+      "learning_rate": 4.929414958599159e-05,
+      "loss": 0.1014,
+      "step": 620
+    },
+    {
+      "epoch": 0.08663575755524722,
+      "grad_norm": 1.8620424270629883,
+      "learning_rate": 4.9267001493145107e-05,
+      "loss": 0.2682,
+      "step": 640
+    },
+    {
+      "epoch": 0.08934312497884869,
+      "grad_norm": 22.248580932617188,
+      "learning_rate": 4.923985340029863e-05,
+      "loss": 0.1485,
+      "step": 660
+    },
+    {
+      "epoch": 0.09205049240245017,
+      "grad_norm": 0.05200627073645592,
+      "learning_rate": 4.9212705307452154e-05,
+      "loss": 0.1227,
+      "step": 680
+    },
+    {
+      "epoch": 0.09475785982605164,
+      "grad_norm": 1.1178234815597534,
+      "learning_rate": 4.918555721460568e-05,
+      "loss": 0.1507,
+      "step": 700
+    },
+    {
+      "epoch": 0.09475785982605164,
+      "eval_loss": 0.05151496082544327,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 126.9972,
+      "eval_samples_per_second": 3.937,
+      "eval_steps_per_second": 0.984,
+      "step": 700
+    },
+    {
+      "epoch": 0.09746522724965312,
+      "grad_norm": 1.1798577308654785,
+      "learning_rate": 4.9158409121759195e-05,
+      "loss": 0.2476,
+      "step": 720
+    },
+    {
+      "epoch": 0.1001725946732546,
+      "grad_norm": 21.11642074584961,
+      "learning_rate": 4.9131261028912726e-05,
+      "loss": 0.2516,
+      "step": 740
+    },
+    {
+      "epoch": 0.10287996209685607,
+      "grad_norm": 21.654647827148438,
+      "learning_rate": 4.910411293606624e-05,
+      "loss": 0.2303,
+      "step": 760
+    },
+    {
+      "epoch": 0.10558732952045755,
+      "grad_norm": 27.137983322143555,
+      "learning_rate": 4.907696484321977e-05,
+      "loss": 0.1943,
+      "step": 780
+    },
+    {
+      "epoch": 0.10829469694405902,
+      "grad_norm": 12.374953269958496,
+      "learning_rate": 4.904981675037329e-05,
+      "loss": 0.2747,
+      "step": 800
+    },
+    {
+      "epoch": 0.10829469694405902,
+      "eval_loss": 0.04837380722165108,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 120.5107,
+      "eval_samples_per_second": 4.149,
+      "eval_steps_per_second": 1.037,
+      "step": 800
+    },
+    {
+      "epoch": 0.1110020643676605,
+      "grad_norm": 22.375734329223633,
+      "learning_rate": 4.902266865752681e-05,
+      "loss": 0.3168,
+      "step": 820
+    },
+    {
+      "epoch": 0.11370943179126197,
+      "grad_norm": 0.05682824179530144,
+      "learning_rate": 4.899552056468034e-05,
+      "loss": 0.065,
+      "step": 840
+    },
+    {
+      "epoch": 0.11641679921486345,
+      "grad_norm": 8.457867622375488,
+      "learning_rate": 4.8968372471833856e-05,
+      "loss": 0.1069,
+      "step": 860
+    },
+    {
+      "epoch": 0.11912416663846492,
+      "grad_norm": 0.8650844693183899,
+      "learning_rate": 4.894122437898738e-05,
+      "loss": 0.0799,
+      "step": 880
+    },
+    {
+      "epoch": 0.1218315340620664,
+      "grad_norm": 10.979653358459473,
+      "learning_rate": 4.8914076286140903e-05,
+      "loss": 0.3638,
+      "step": 900
+    },
+    {
+      "epoch": 0.1218315340620664,
+      "eval_loss": 0.048302676528692245,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 125.5881,
+      "eval_samples_per_second": 3.981,
+      "eval_steps_per_second": 0.995,
+      "step": 900
+    },
+    {
+      "epoch": 0.12453890148566787,
+      "grad_norm": 1.0572680234909058,
+      "learning_rate": 4.888692819329442e-05,
+      "loss": 0.1154,
+      "step": 920
+    },
+    {
+      "epoch": 0.12724626890926935,
+      "grad_norm": 0.3924155533313751,
+      "learning_rate": 4.8859780100447944e-05,
+      "loss": 0.1085,
+      "step": 940
+    },
+    {
+      "epoch": 0.12995363633287083,
+      "grad_norm": 0.07971794903278351,
+      "learning_rate": 4.883263200760147e-05,
+      "loss": 0.1827,
+      "step": 960
+    },
+    {
+      "epoch": 0.1326610037564723,
+      "grad_norm": 1.397789478302002,
+      "learning_rate": 4.880548391475499e-05,
+      "loss": 0.0695,
+      "step": 980
+    },
+    {
+      "epoch": 0.13536837118007378,
+      "grad_norm": 3.120443105697632,
+      "learning_rate": 4.877833582190851e-05,
+      "loss": 0.149,
+      "step": 1000
+    },
+    {
+      "epoch": 0.13536837118007378,
+      "eval_loss": 0.057031989097595215,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.6549,
+      "eval_samples_per_second": 4.323,
+      "eval_steps_per_second": 1.081,
+      "step": 1000
+    },
+    {
+      "epoch": 0.13807573860367525,
+      "grad_norm": 1.3726035356521606,
+      "learning_rate": 4.875118772906204e-05,
+      "loss": 0.0567,
+      "step": 1020
+    },
+    {
+      "epoch": 0.14078310602727673,
+      "grad_norm": 0.16208282113075256,
+      "learning_rate": 4.872403963621556e-05,
+      "loss": 0.2078,
+      "step": 1040
+    },
+    {
+      "epoch": 0.1434904734508782,
+      "grad_norm": 1.425825834274292,
+      "learning_rate": 4.869689154336908e-05,
+      "loss": 0.353,
+      "step": 1060
+    },
+    {
+      "epoch": 0.14619784087447968,
+      "grad_norm": 16.413312911987305,
+      "learning_rate": 4.8669743450522605e-05,
+      "loss": 0.0656,
+      "step": 1080
+    },
+    {
+      "epoch": 0.14890520829808115,
+      "grad_norm": 9.234221458435059,
+      "learning_rate": 4.864259535767612e-05,
+      "loss": 0.2405,
+      "step": 1100
+    },
+    {
+      "epoch": 0.14890520829808115,
+      "eval_loss": 0.04857669770717621,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.5724,
+      "eval_samples_per_second": 4.326,
+      "eval_steps_per_second": 1.082,
+      "step": 1100
+    },
+    {
+      "epoch": 0.15161257572168263,
+      "grad_norm": 39.68949890136719,
+      "learning_rate": 4.861544726482965e-05,
+      "loss": 0.2318,
+      "step": 1120
+    },
+    {
+      "epoch": 0.1543199431452841,
+      "grad_norm": 0.022096823900938034,
+      "learning_rate": 4.858829917198317e-05,
+      "loss": 0.1425,
+      "step": 1140
+    },
+    {
+      "epoch": 0.15702731056888558,
+      "grad_norm": 23.582176208496094,
+      "learning_rate": 4.8561151079136694e-05,
+      "loss": 0.2259,
+      "step": 1160
+    },
+    {
+      "epoch": 0.15973467799248706,
+      "grad_norm": 17.384206771850586,
+      "learning_rate": 4.853400298629022e-05,
+      "loss": 0.1154,
+      "step": 1180
+    },
+    {
+      "epoch": 0.16244204541608853,
+      "grad_norm": 14.885477066040039,
+      "learning_rate": 4.8506854893443735e-05,
+      "loss": 0.2201,
+      "step": 1200
+    },
+    {
+      "epoch": 0.16244204541608853,
+      "eval_loss": 0.043531883507966995,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.2235,
+      "eval_samples_per_second": 4.339,
+      "eval_steps_per_second": 1.085,
+      "step": 1200
+    },
+    {
+      "epoch": 0.16514941283969,
+      "grad_norm": 4.196330547332764,
+      "learning_rate": 4.847970680059726e-05,
+      "loss": 0.1639,
+      "step": 1220
+    },
+    {
+      "epoch": 0.16785678026329148,
+      "grad_norm": 27.09263038635254,
+      "learning_rate": 4.845255870775078e-05,
+      "loss": 0.1695,
+      "step": 1240
+    },
+    {
+      "epoch": 0.17056414768689296,
+      "grad_norm": 31.63964080810547,
+      "learning_rate": 4.8425410614904306e-05,
+      "loss": 0.1279,
+      "step": 1260
+    },
+    {
+      "epoch": 0.17327151511049443,
+      "grad_norm": 0.0727367028594017,
+      "learning_rate": 4.839826252205783e-05,
+      "loss": 0.1283,
+      "step": 1280
+    },
+    {
+      "epoch": 0.1759788825340959,
+      "grad_norm": 5.623673915863037,
+      "learning_rate": 4.837111442921135e-05,
+      "loss": 0.195,
+      "step": 1300
+    },
+    {
+      "epoch": 0.1759788825340959,
+      "eval_loss": 0.040112994611263275,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.6438,
+      "eval_samples_per_second": 4.324,
+      "eval_steps_per_second": 1.081,
+      "step": 1300
+    },
+    {
+      "epoch": 0.17868624995769739,
+      "grad_norm": 6.618963241577148,
+      "learning_rate": 4.834396633636487e-05,
+      "loss": 0.1477,
+      "step": 1320
+    },
+    {
+      "epoch": 0.18139361738129886,
+      "grad_norm": 0.15700267255306244,
+      "learning_rate": 4.8316818243518395e-05,
+      "loss": 0.082,
+      "step": 1340
+    },
+    {
+      "epoch": 0.18410098480490034,
+      "grad_norm": 0.07716820389032364,
+      "learning_rate": 4.828967015067192e-05,
+      "loss": 0.1526,
+      "step": 1360
+    },
+    {
+      "epoch": 0.1868083522285018,
+      "grad_norm": 0.1871941089630127,
+      "learning_rate": 4.8262522057825436e-05,
+      "loss": 0.2056,
+      "step": 1380
+    },
+    {
+      "epoch": 0.1895157196521033,
+      "grad_norm": 7.145862579345703,
+      "learning_rate": 4.823537396497897e-05,
+      "loss": 0.186,
+      "step": 1400
+    },
+    {
+      "epoch": 0.1895157196521033,
+      "eval_loss": 0.04851188883185387,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.2328,
+      "eval_samples_per_second": 4.339,
+      "eval_steps_per_second": 1.085,
+      "step": 1400
+    },
+    {
+      "epoch": 0.19222308707570476,
+      "grad_norm": 0.10455508530139923,
+      "learning_rate": 4.8208225872132484e-05,
+      "loss": 0.3346,
+      "step": 1420
+    },
+    {
+      "epoch": 0.19493045449930624,
+      "grad_norm": 6.885641098022461,
+      "learning_rate": 4.818107777928601e-05,
+      "loss": 0.2167,
+      "step": 1440
+    },
+    {
+      "epoch": 0.1976378219229077,
+      "grad_norm": 12.304588317871094,
+      "learning_rate": 4.815392968643953e-05,
+      "loss": 0.1419,
+      "step": 1460
+    },
+    {
+      "epoch": 0.2003451893465092,
+      "grad_norm": 12.884013175964355,
+      "learning_rate": 4.812678159359305e-05,
+      "loss": 0.2035,
+      "step": 1480
+    },
+    {
+      "epoch": 0.20305255677011066,
+      "grad_norm": 22.836992263793945,
+      "learning_rate": 4.809963350074658e-05,
+      "loss": 0.1116,
+      "step": 1500
+    },
+    {
+      "epoch": 0.20305255677011066,
+      "eval_loss": 0.055405810475349426,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.8109,
+      "eval_samples_per_second": 4.317,
+      "eval_steps_per_second": 1.079,
+      "step": 1500
+    },
+    {
+      "epoch": 0.20575992419371214,
+      "grad_norm": 2.097397565841675,
+      "learning_rate": 4.8072485407900096e-05,
+      "loss": 0.0493,
+      "step": 1520
+    },
+    {
+      "epoch": 0.20846729161731362,
+      "grad_norm": 8.2213773727417,
+      "learning_rate": 4.804533731505362e-05,
+      "loss": 0.0844,
+      "step": 1540
+    },
+    {
+      "epoch": 0.2111746590409151,
+      "grad_norm": 0.014365073293447495,
+      "learning_rate": 4.8018189222207144e-05,
+      "loss": 0.2243,
+      "step": 1560
+    },
+    {
+      "epoch": 0.21388202646451657,
+      "grad_norm": 0.2695819139480591,
+      "learning_rate": 4.799104112936066e-05,
+      "loss": 0.1934,
+      "step": 1580
+    },
+    {
+      "epoch": 0.21658939388811804,
+      "grad_norm": 0.25374430418014526,
+      "learning_rate": 4.7963893036514185e-05,
+      "loss": 0.2522,
+      "step": 1600
+    },
+    {
+      "epoch": 0.21658939388811804,
+      "eval_loss": 0.059707771986722946,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.274,
+      "eval_samples_per_second": 4.375,
+      "eval_steps_per_second": 1.094,
+      "step": 1600
+    },
+    {
+      "epoch": 0.21929676131171952,
+      "grad_norm": 0.003937189467251301,
+      "learning_rate": 4.793674494366771e-05,
+      "loss": 0.0801,
+      "step": 1620
+    },
+    {
+      "epoch": 0.222004128735321,
+      "grad_norm": 0.9062768220901489,
+      "learning_rate": 4.790959685082123e-05,
+      "loss": 0.2022,
+      "step": 1640
+    },
+    {
+      "epoch": 0.22471149615892247,
+      "grad_norm": 3.008561849594116,
+      "learning_rate": 4.788244875797476e-05,
+      "loss": 0.06,
+      "step": 1660
+    },
+    {
+      "epoch": 0.22741886358252394,
+      "grad_norm": 0.8405902981758118,
+      "learning_rate": 4.785530066512828e-05,
+      "loss": 0.2472,
+      "step": 1680
+    },
+    {
+      "epoch": 0.23012623100612542,
+      "grad_norm": 8.165879249572754,
+      "learning_rate": 4.78281525722818e-05,
+      "loss": 0.1708,
+      "step": 1700
+    },
+    {
+      "epoch": 0.23012623100612542,
+      "eval_loss": 0.057815808802843094,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.0717,
+      "eval_samples_per_second": 4.345,
+      "eval_steps_per_second": 1.086,
+      "step": 1700
+    },
+    {
+      "epoch": 0.2328335984297269,
+      "grad_norm": 0.13060545921325684,
+      "learning_rate": 4.780100447943532e-05,
+      "loss": 0.117,
+      "step": 1720
+    },
+    {
+      "epoch": 0.23554096585332837,
+      "grad_norm": 12.249765396118164,
+      "learning_rate": 4.7773856386588846e-05,
+      "loss": 0.1507,
+      "step": 1740
+    },
+    {
+      "epoch": 0.23824833327692985,
+      "grad_norm": 1.8489207029342651,
+      "learning_rate": 4.774670829374236e-05,
+      "loss": 0.3488,
+      "step": 1760
+    },
+    {
+      "epoch": 0.24095570070053132,
+      "grad_norm": 0.9456131458282471,
+      "learning_rate": 4.771956020089589e-05,
+      "loss": 0.3084,
+      "step": 1780
+    },
+    {
+      "epoch": 0.2436630681241328,
+      "grad_norm": 24.736726760864258,
+      "learning_rate": 4.769241210804941e-05,
+      "loss": 0.2627,
+      "step": 1800
+    },
+    {
+      "epoch": 0.2436630681241328,
+      "eval_loss": 0.06722074747085571,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.4272,
+      "eval_samples_per_second": 4.332,
+      "eval_steps_per_second": 1.083,
+      "step": 1800
+    },
+    {
+      "epoch": 0.24637043554773427,
+      "grad_norm": 2.1300015449523926,
+      "learning_rate": 4.7665264015202934e-05,
+      "loss": 0.1497,
+      "step": 1820
+    },
+    {
+      "epoch": 0.24907780297133575,
+      "grad_norm": 7.431568145751953,
+      "learning_rate": 4.763811592235646e-05,
+      "loss": 0.1287,
+      "step": 1840
+    },
+    {
+      "epoch": 0.2517851703949372,
+      "grad_norm": 0.017447171732783318,
+      "learning_rate": 4.7610967829509975e-05,
+      "loss": 0.0981,
+      "step": 1860
+    },
+    {
+      "epoch": 0.2544925378185387,
+      "grad_norm": 0.7597976922988892,
+      "learning_rate": 4.7583819736663506e-05,
+      "loss": 0.1814,
+      "step": 1880
+    },
+    {
+      "epoch": 0.25719990524214015,
+      "grad_norm": 2.8532488346099854,
+      "learning_rate": 4.755667164381702e-05,
+      "loss": 0.2901,
+      "step": 1900
+    },
+    {
+      "epoch": 0.25719990524214015,
+      "eval_loss": 0.05738656595349312,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.4019,
+      "eval_samples_per_second": 4.371,
+      "eval_steps_per_second": 1.093,
+      "step": 1900
+    },
+    {
+      "epoch": 0.25990727266574165,
+      "grad_norm": 6.662034511566162,
+      "learning_rate": 4.752952355097055e-05,
+      "loss": 0.1815,
+      "step": 1920
+    },
+    {
+      "epoch": 0.2626146400893431,
+      "grad_norm": 0.4220564365386963,
+      "learning_rate": 4.750237545812407e-05,
+      "loss": 0.0649,
+      "step": 1940
+    },
+    {
+      "epoch": 0.2653220075129446,
+      "grad_norm": 7.456377029418945,
+      "learning_rate": 4.7475227365277595e-05,
+      "loss": 0.1509,
+      "step": 1960
+    },
+    {
+      "epoch": 0.26802937493654605,
+      "grad_norm": 0.16843333840370178,
+      "learning_rate": 4.744807927243111e-05,
+      "loss": 0.1159,
+      "step": 1980
+    },
+    {
+      "epoch": 0.27073674236014755,
+      "grad_norm": 0.038046594709157944,
+      "learning_rate": 4.7420931179584636e-05,
+      "loss": 0.0354,
+      "step": 2000
+    },
+    {
+      "epoch": 0.27073674236014755,
+      "eval_loss": 0.09768073260784149,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.1145,
+      "eval_samples_per_second": 4.344,
+      "eval_steps_per_second": 1.086,
+      "step": 2000
+    },
+    {
+      "epoch": 0.273444109783749,
+      "grad_norm": 19.501468658447266,
+      "learning_rate": 4.739378308673816e-05,
+      "loss": 0.1081,
+      "step": 2020
+    },
+    {
+      "epoch": 0.2761514772073505,
+      "grad_norm": 11.927206993103027,
+      "learning_rate": 4.736663499389168e-05,
+      "loss": 0.0767,
+      "step": 2040
+    },
+    {
+      "epoch": 0.27885884463095195,
+      "grad_norm": 1.6448673009872437,
+      "learning_rate": 4.733948690104521e-05,
+      "loss": 0.1237,
+      "step": 2060
+    },
+    {
+      "epoch": 0.28156621205455346,
+      "grad_norm": 8.10880184173584,
+      "learning_rate": 4.7312338808198724e-05,
+      "loss": 0.0313,
+      "step": 2080
+    },
+    {
+      "epoch": 0.2842735794781549,
+      "grad_norm": 10.024765968322754,
+      "learning_rate": 4.728519071535225e-05,
+      "loss": 0.156,
+      "step": 2100
+    },
+    {
+      "epoch": 0.2842735794781549,
+      "eval_loss": 0.08133851736783981,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.2526,
+      "eval_samples_per_second": 4.338,
+      "eval_steps_per_second": 1.085,
+      "step": 2100
+    },
+    {
+      "epoch": 0.2869809469017564,
+      "grad_norm": 0.131832554936409,
+      "learning_rate": 4.725804262250577e-05,
+      "loss": 0.1318,
+      "step": 2120
+    },
+    {
+      "epoch": 0.28968831432535785,
+      "grad_norm": 12.999658584594727,
+      "learning_rate": 4.723089452965929e-05,
+      "loss": 0.2504,
+      "step": 2140
+    },
+    {
+      "epoch": 0.29239568174895936,
+      "grad_norm": 12.960088729858398,
+      "learning_rate": 4.720374643681282e-05,
+      "loss": 0.2561,
+      "step": 2160
+    },
+    {
+      "epoch": 0.2951030491725608,
+      "grad_norm": 2.730931043624878,
+      "learning_rate": 4.717659834396634e-05,
+      "loss": 0.2609,
+      "step": 2180
+    },
+    {
+      "epoch": 0.2978104165961623,
+      "grad_norm": 28.357595443725586,
+      "learning_rate": 4.714945025111986e-05,
+      "loss": 0.1156,
+      "step": 2200
+    },
+    {
+      "epoch": 0.2978104165961623,
+      "eval_loss": 0.06201322376728058,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.4336,
+      "eval_samples_per_second": 4.331,
+      "eval_steps_per_second": 1.083,
+      "step": 2200
+    },
+    {
+      "epoch": 0.30051778401976376,
+      "grad_norm": 21.61467170715332,
+      "learning_rate": 4.7122302158273385e-05,
+      "loss": 0.139,
+      "step": 2220
+    },
+    {
+      "epoch": 0.30322515144336526,
+      "grad_norm": 1.2531449794769287,
+      "learning_rate": 4.70951540654269e-05,
+      "loss": 0.1351,
+      "step": 2240
+    },
+    {
+      "epoch": 0.3059325188669667,
+      "grad_norm": 0.1177072823047638,
+      "learning_rate": 4.706800597258043e-05,
+      "loss": 0.2042,
+      "step": 2260
+    },
+    {
+      "epoch": 0.3086398862905682,
+      "grad_norm": 0.035650696605443954,
+      "learning_rate": 4.704085787973395e-05,
+      "loss": 0.165,
+      "step": 2280
+    },
+    {
+      "epoch": 0.31134725371416966,
+      "grad_norm": 0.03751590847969055,
+      "learning_rate": 4.7013709786887474e-05,
+      "loss": 0.1089,
+      "step": 2300
+    },
+    {
+      "epoch": 0.31134725371416966,
+      "eval_loss": 0.06733442097902298,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.3253,
+      "eval_samples_per_second": 4.336,
+      "eval_steps_per_second": 1.084,
+      "step": 2300
+    },
+    {
+      "epoch": 0.31405462113777116,
+      "grad_norm": 61.547889709472656,
+      "learning_rate": 4.6986561694041e-05,
+      "loss": 0.319,
+      "step": 2320
+    },
+    {
+      "epoch": 0.3167619885613726,
+      "grad_norm": 20.737930297851562,
+      "learning_rate": 4.695941360119452e-05,
+      "loss": 0.1578,
+      "step": 2340
+    },
+    {
+      "epoch": 0.3194693559849741,
+      "grad_norm": 0.3770120143890381,
+      "learning_rate": 4.693226550834804e-05,
+      "loss": 0.1038,
+      "step": 2360
+    },
+    {
+      "epoch": 0.32217672340857556,
+      "grad_norm": 12.355575561523438,
+      "learning_rate": 4.690511741550156e-05,
+      "loss": 0.1896,
+      "step": 2380
+    },
+    {
+      "epoch": 0.32488409083217706,
+      "grad_norm": 7.065931797027588,
+      "learning_rate": 4.6877969322655086e-05,
+      "loss": 0.1368,
+      "step": 2400
+    },
+    {
+      "epoch": 0.32488409083217706,
+      "eval_loss": 0.058929502964019775,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.0859,
+      "eval_samples_per_second": 4.345,
+      "eval_steps_per_second": 1.086,
+      "step": 2400
+    },
+    {
+      "epoch": 0.3275914582557785,
+      "grad_norm": 1.2074625492095947,
+      "learning_rate": 4.68508212298086e-05,
+      "loss": 0.2128,
+      "step": 2420
+    },
+    {
+      "epoch": 0.33029882567938,
+      "grad_norm": 24.548015594482422,
+      "learning_rate": 4.6823673136962134e-05,
+      "loss": 0.1701,
+      "step": 2440
+    },
+    {
+      "epoch": 0.33300619310298146,
+      "grad_norm": 1.5518879890441895,
+      "learning_rate": 4.679652504411565e-05,
+      "loss": 0.2455,
+      "step": 2460
+    },
+    {
+      "epoch": 0.33571356052658297,
+      "grad_norm": 0.4867059886455536,
+      "learning_rate": 4.6769376951269175e-05,
+      "loss": 0.2979,
+      "step": 2480
+    },
+    {
+      "epoch": 0.3384209279501844,
+      "grad_norm": 2.8492958545684814,
+      "learning_rate": 4.67422288584227e-05,
+      "loss": 0.1853,
+      "step": 2500
+    },
+    {
+      "epoch": 0.3384209279501844,
+      "eval_loss": 0.06687568873167038,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.1592,
+      "eval_samples_per_second": 4.342,
+      "eval_steps_per_second": 1.085,
+      "step": 2500
+    },
+    {
+      "epoch": 0.3411282953737859,
+      "grad_norm": 1.295285940170288,
+      "learning_rate": 4.6715080765576216e-05,
+      "loss": 0.0603,
+      "step": 2520
+    },
+    {
+      "epoch": 0.34383566279738736,
+      "grad_norm": 0.0439661480486393,
+      "learning_rate": 4.6687932672729747e-05,
+      "loss": 0.0869,
+      "step": 2540
+    },
+    {
+      "epoch": 0.34654303022098887,
+      "grad_norm": 42.599430084228516,
+      "learning_rate": 4.6660784579883264e-05,
+      "loss": 0.0603,
+      "step": 2560
+    },
+    {
+      "epoch": 0.3492503976445903,
+      "grad_norm": 0.09535373747348785,
+      "learning_rate": 4.663363648703679e-05,
+      "loss": 0.3583,
+      "step": 2580
+    },
+    {
+      "epoch": 0.3519577650681918,
+      "grad_norm": 0.08492514491081238,
+      "learning_rate": 4.660648839419031e-05,
+      "loss": 0.1109,
+      "step": 2600
+    },
+    {
+      "epoch": 0.3519577650681918,
+      "eval_loss": 0.07154419273138046,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.7194,
+      "eval_samples_per_second": 4.321,
+      "eval_steps_per_second": 1.08,
+      "step": 2600
+    },
+    {
+      "epoch": 0.35466513249179327,
+      "grad_norm": 0.145077183842659,
+      "learning_rate": 4.6579340301343835e-05,
+      "loss": 0.0875,
+      "step": 2620
+    },
+    {
+      "epoch": 0.35737249991539477,
+      "grad_norm": 9.414113998413086,
+      "learning_rate": 4.655219220849735e-05,
+      "loss": 0.2708,
+      "step": 2640
+    },
+    {
+      "epoch": 0.3600798673389962,
+      "grad_norm": 0.012831827625632286,
+      "learning_rate": 4.6525044115650876e-05,
+      "loss": 0.3477,
+      "step": 2660
+    },
+    {
+      "epoch": 0.3627872347625977,
+      "grad_norm": 0.06755544990301132,
+      "learning_rate": 4.64978960228044e-05,
+      "loss": 0.2059,
+      "step": 2680
+    },
+    {
+      "epoch": 0.36549460218619917,
+      "grad_norm": 0.40635791420936584,
+      "learning_rate": 4.6470747929957924e-05,
+      "loss": 0.1477,
+      "step": 2700
+    },
+    {
+      "epoch": 0.36549460218619917,
+      "eval_loss": 0.08114828914403915,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.4236,
+      "eval_samples_per_second": 4.332,
+      "eval_steps_per_second": 1.083,
+      "step": 2700
+    },
+    {
+      "epoch": 0.3682019696098007,
+      "grad_norm": 21.508838653564453,
+      "learning_rate": 4.644359983711145e-05,
+      "loss": 0.1263,
+      "step": 2720
+    },
+    {
+      "epoch": 0.3709093370334021,
+      "grad_norm": 0.01996573433279991,
+      "learning_rate": 4.6416451744264965e-05,
+      "loss": 0.0615,
+      "step": 2740
+    },
+    {
+      "epoch": 0.3736167044570036,
+      "grad_norm": 0.048173803836107254,
+      "learning_rate": 4.638930365141849e-05,
+      "loss": 0.1578,
+      "step": 2760
+    },
+    {
+      "epoch": 0.37632407188060507,
+      "grad_norm": 9.07430362701416,
+      "learning_rate": 4.636215555857201e-05,
+      "loss": 0.1826,
+      "step": 2780
+    },
+    {
+      "epoch": 0.3790314393042066,
+      "grad_norm": 0.015813810750842094,
+      "learning_rate": 4.633500746572553e-05,
+      "loss": 0.0999,
+      "step": 2800
+    },
+    {
+      "epoch": 0.3790314393042066,
+      "eval_loss": 0.07192689180374146,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.7893,
+      "eval_samples_per_second": 4.356,
+      "eval_steps_per_second": 1.089,
+      "step": 2800
+    },
+    {
+      "epoch": 0.381738806727808,
+      "grad_norm": 17.83676528930664,
+      "learning_rate": 4.630785937287906e-05,
+      "loss": 0.3131,
+      "step": 2820
+    },
+    {
+      "epoch": 0.3844461741514095,
+      "grad_norm": 0.2090553194284439,
+      "learning_rate": 4.628071128003258e-05,
+      "loss": 0.1114,
+      "step": 2840
+    },
+    {
+      "epoch": 0.387153541575011,
+      "grad_norm": 15.477974891662598,
+      "learning_rate": 4.62535631871861e-05,
+      "loss": 0.1856,
+      "step": 2860
+    },
+    {
+      "epoch": 0.3898609089986125,
+      "grad_norm": 2.8544585704803467,
+      "learning_rate": 4.6226415094339625e-05,
+      "loss": 0.1231,
+      "step": 2880
+    },
+    {
+      "epoch": 0.3925682764222139,
+      "grad_norm": 4.374979019165039,
+      "learning_rate": 4.619926700149315e-05,
+      "loss": 0.0525,
+      "step": 2900
+    },
+    {
+      "epoch": 0.3925682764222139,
+      "eval_loss": 0.07568126916885376,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.4166,
+      "eval_samples_per_second": 4.332,
+      "eval_steps_per_second": 1.083,
+      "step": 2900
+    },
+    {
+      "epoch": 0.3952756438458154,
+      "grad_norm": 17.247142791748047,
+      "learning_rate": 4.617211890864667e-05,
+      "loss": 0.1124,
+      "step": 2920
+    },
+    {
+      "epoch": 0.3979830112694169,
+      "grad_norm": 0.016074227169156075,
+      "learning_rate": 4.614497081580019e-05,
+      "loss": 0.0538,
+      "step": 2940
+    },
+    {
+      "epoch": 0.4006903786930184,
+      "grad_norm": 0.5515703558921814,
+      "learning_rate": 4.6117822722953714e-05,
+      "loss": 0.1552,
+      "step": 2960
+    },
+    {
+      "epoch": 0.4033977461166198,
+      "grad_norm": 0.03307110071182251,
+      "learning_rate": 4.609067463010724e-05,
+      "loss": 0.0419,
+      "step": 2980
+    },
+    {
+      "epoch": 0.40610511354022133,
+      "grad_norm": 0.1128387525677681,
+      "learning_rate": 4.606352653726076e-05,
+      "loss": 0.0513,
+      "step": 3000
+    },
+    {
+      "epoch": 0.40610511354022133,
+      "eval_loss": 0.0807776153087616,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.7987,
+      "eval_samples_per_second": 4.355,
+      "eval_steps_per_second": 1.089,
+      "step": 3000
+    },
+    {
+      "epoch": 0.4088124809638228,
+      "grad_norm": 7.236224174499512,
+      "learning_rate": 4.603637844441428e-05,
+      "loss": 0.1509,
+      "step": 3020
+    },
+    {
+      "epoch": 0.4115198483874243,
+      "grad_norm": 13.330012321472168,
+      "learning_rate": 4.60092303515678e-05,
+      "loss": 0.2394,
+      "step": 3040
+    },
+    {
+      "epoch": 0.41422721581102573,
+      "grad_norm": 16.844871520996094,
+      "learning_rate": 4.598208225872133e-05,
+      "loss": 0.1604,
+      "step": 3060
+    },
+    {
+      "epoch": 0.41693458323462723,
+      "grad_norm": 0.01411959808319807,
+      "learning_rate": 4.595493416587485e-05,
+      "loss": 0.0778,
+      "step": 3080
+    },
+    {
+      "epoch": 0.4196419506582287,
+      "grad_norm": 5.0137529373168945,
+      "learning_rate": 4.5927786073028375e-05,
+      "loss": 0.1562,
+      "step": 3100
+    },
+    {
+      "epoch": 0.4196419506582287,
+      "eval_loss": 0.06787914782762527,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.2644,
+      "eval_samples_per_second": 4.338,
+      "eval_steps_per_second": 1.084,
+      "step": 3100
+    },
+    {
+      "epoch": 0.4223493180818302,
+      "grad_norm": 2.870209217071533,
+      "learning_rate": 4.590063798018189e-05,
+      "loss": 0.1097,
+      "step": 3120
+    },
+    {
+      "epoch": 0.42505668550543163,
+      "grad_norm": 0.18988296389579773,
+      "learning_rate": 4.587348988733542e-05,
+      "loss": 0.1236,
+      "step": 3140
+    },
+    {
+      "epoch": 0.42776405292903313,
+      "grad_norm": 2.6342244148254395,
+      "learning_rate": 4.584634179448894e-05,
+      "loss": 0.1644,
+      "step": 3160
+    },
+    {
+      "epoch": 0.4304714203526346,
+      "grad_norm": 1.666049838066101,
+      "learning_rate": 4.581919370164246e-05,
+      "loss": 0.0629,
+      "step": 3180
+    },
+    {
+      "epoch": 0.4331787877762361,
+      "grad_norm": 0.0062801106832921505,
+      "learning_rate": 4.579204560879599e-05,
+      "loss": 0.1523,
+      "step": 3200
+    },
+    {
+      "epoch": 0.4331787877762361,
+      "eval_loss": 0.07749966531991959,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.8371,
+      "eval_samples_per_second": 4.354,
+      "eval_steps_per_second": 1.088,
+      "step": 3200
+    },
+    {
+      "epoch": 0.43588615519983753,
+      "grad_norm": 0.9460217356681824,
+      "learning_rate": 4.5764897515949504e-05,
+      "loss": 0.1426,
+      "step": 3220
+    },
+    {
+      "epoch": 0.43859352262343904,
+      "grad_norm": 6.54037618637085,
+      "learning_rate": 4.573774942310303e-05,
+      "loss": 0.1225,
+      "step": 3240
+    },
+    {
+      "epoch": 0.4413008900470405,
+      "grad_norm": 6.7528395652771,
+      "learning_rate": 4.571060133025655e-05,
+      "loss": 0.1678,
+      "step": 3260
+    },
+    {
+      "epoch": 0.444008257470642,
+      "grad_norm": 0.906831681728363,
+      "learning_rate": 4.5683453237410076e-05,
+      "loss": 0.0716,
+      "step": 3280
+    },
+    {
+      "epoch": 0.44671562489424343,
+      "grad_norm": 0.9793009161949158,
+      "learning_rate": 4.56563051445636e-05,
+      "loss": 0.1616,
+      "step": 3300
+    },
+    {
+      "epoch": 0.44671562489424343,
+      "eval_loss": 0.08035814017057419,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.7485,
+      "eval_samples_per_second": 4.357,
+      "eval_steps_per_second": 1.089,
+      "step": 3300
+    },
+    {
+      "epoch": 0.44942299231784494,
+      "grad_norm": 0.014806450344622135,
+      "learning_rate": 4.562915705171712e-05,
+      "loss": 0.1744,
+      "step": 3320
+    },
+    {
+      "epoch": 0.4521303597414464,
+      "grad_norm": 20.00578498840332,
+      "learning_rate": 4.560200895887064e-05,
+      "loss": 0.0343,
+      "step": 3340
+    },
+    {
+      "epoch": 0.4548377271650479,
+      "grad_norm": 0.03944426402449608,
+      "learning_rate": 4.5574860866024165e-05,
+      "loss": 0.1867,
+      "step": 3360
+    },
+    {
+      "epoch": 0.45754509458864934,
+      "grad_norm": 5.614328861236572,
+      "learning_rate": 4.554771277317769e-05,
+      "loss": 0.0663,
+      "step": 3380
+    },
+    {
+      "epoch": 0.46025246201225084,
+      "grad_norm": 0.06043994799256325,
+      "learning_rate": 4.5520564680331206e-05,
+      "loss": 0.1641,
+      "step": 3400
+    },
+    {
+      "epoch": 0.46025246201225084,
+      "eval_loss": 0.08000614494085312,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.1794,
+      "eval_samples_per_second": 4.341,
+      "eval_steps_per_second": 1.085,
+      "step": 3400
+    },
+    {
+      "epoch": 0.4629598294358523,
+      "grad_norm": 24.133304595947266,
+      "learning_rate": 4.549341658748473e-05,
+      "loss": 0.1472,
+      "step": 3420
+    },
+    {
+      "epoch": 0.4656671968594538,
+      "grad_norm": 0.7912468314170837,
+      "learning_rate": 4.5466268494638254e-05,
+      "loss": 0.2045,
+      "step": 3440
+    },
+    {
+      "epoch": 0.46837456428305524,
+      "grad_norm": 21.022037506103516,
+      "learning_rate": 4.543912040179177e-05,
+      "loss": 0.242,
+      "step": 3460
+    },
+    {
+      "epoch": 0.47108193170665674,
+      "grad_norm": 3.1126956939697266,
+      "learning_rate": 4.54119723089453e-05,
+      "loss": 0.1122,
+      "step": 3480
+    },
+    {
+      "epoch": 0.4737892991302582,
+      "grad_norm": 0.03333674371242523,
+      "learning_rate": 4.538482421609882e-05,
+      "loss": 0.1162,
+      "step": 3500
+    },
+    {
+      "epoch": 0.4737892991302582,
+      "eval_loss": 0.06670122593641281,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.6776,
+      "eval_samples_per_second": 4.322,
+      "eval_steps_per_second": 1.081,
+      "step": 3500
+    },
+    {
+      "epoch": 0.4764966665538597,
+      "grad_norm": 0.3281305134296417,
+      "learning_rate": 4.535767612325235e-05,
+      "loss": 0.0489,
+      "step": 3520
+    },
+    {
+      "epoch": 0.47920403397746114,
+      "grad_norm": 2.830657720565796,
+      "learning_rate": 4.5330528030405866e-05,
+      "loss": 0.1212,
+      "step": 3540
+    },
+    {
+      "epoch": 0.48191140140106264,
+      "grad_norm": 1.2127472162246704,
+      "learning_rate": 4.530337993755939e-05,
+      "loss": 0.0906,
+      "step": 3560
+    },
+    {
+      "epoch": 0.4846187688246641,
+      "grad_norm": 33.909671783447266,
+      "learning_rate": 4.5276231844712914e-05,
+      "loss": 0.2042,
+      "step": 3580
+    },
+    {
+      "epoch": 0.4873261362482656,
+      "grad_norm": 0.00924315769225359,
+      "learning_rate": 4.524908375186643e-05,
+      "loss": 0.1062,
+      "step": 3600
+    },
+    {
+      "epoch": 0.4873261362482656,
+      "eval_loss": 0.06282315403223038,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.5799,
+      "eval_samples_per_second": 4.326,
+      "eval_steps_per_second": 1.082,
+      "step": 3600
+    },
+    {
+      "epoch": 0.49003350367186704,
+      "grad_norm": 0.005617950111627579,
+      "learning_rate": 4.5221935659019955e-05,
+      "loss": 0.1927,
+      "step": 3620
+    },
+    {
+      "epoch": 0.49274087109546855,
+      "grad_norm": 0.00801560003310442,
+      "learning_rate": 4.519478756617348e-05,
+      "loss": 0.1406,
+      "step": 3640
+    },
+    {
+      "epoch": 0.49544823851907,
+      "grad_norm": 23.532512664794922,
+      "learning_rate": 4.5167639473327e-05,
+      "loss": 0.0821,
+      "step": 3660
+    },
+    {
+      "epoch": 0.4981556059426715,
+      "grad_norm": 3.598512649536133,
+      "learning_rate": 4.514049138048052e-05,
+      "loss": 0.0751,
+      "step": 3680
+    },
+    {
+      "epoch": 0.500862973366273,
+      "grad_norm": 0.09781412780284882,
+      "learning_rate": 4.5113343287634044e-05,
+      "loss": 0.2411,
+      "step": 3700
+    },
+    {
+      "epoch": 0.500862973366273,
+      "eval_loss": 0.05527465417981148,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.763,
+      "eval_samples_per_second": 4.319,
+      "eval_steps_per_second": 1.08,
+      "step": 3700
+    },
+    {
+      "epoch": 0.5035703407898744,
+      "grad_norm": 0.02711186371743679,
+      "learning_rate": 4.508619519478757e-05,
+      "loss": 0.1311,
+      "step": 3720
+    },
+    {
+      "epoch": 0.506277708213476,
+      "grad_norm": 0.5090947151184082,
+      "learning_rate": 4.505904710194109e-05,
+      "loss": 0.1294,
+      "step": 3740
+    },
+    {
+      "epoch": 0.5089850756370774,
+      "grad_norm": 0.09400524944067001,
+      "learning_rate": 4.5031899009094615e-05,
+      "loss": 0.3008,
+      "step": 3760
+    },
+    {
+      "epoch": 0.5116924430606788,
+      "grad_norm": 0.27078312635421753,
+      "learning_rate": 4.500475091624813e-05,
+      "loss": 0.1703,
+      "step": 3780
+    },
+    {
+      "epoch": 0.5143998104842803,
+      "grad_norm": 0.035410646349191666,
+      "learning_rate": 4.497760282340166e-05,
+      "loss": 0.1955,
+      "step": 3800
+    },
+    {
+      "epoch": 0.5143998104842803,
+      "eval_loss": 0.06636335700750351,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.6171,
+      "eval_samples_per_second": 4.325,
+      "eval_steps_per_second": 1.081,
+      "step": 3800
+    },
+    {
+      "epoch": 0.5171071779078819,
+      "grad_norm": 2.16019344329834,
+      "learning_rate": 4.495045473055518e-05,
+      "loss": 0.3026,
+      "step": 3820
+    },
+    {
+      "epoch": 0.5198145453314833,
+      "grad_norm": 0.03840683028101921,
+      "learning_rate": 4.4923306637708704e-05,
+      "loss": 0.1793,
+      "step": 3840
+    },
+    {
+      "epoch": 0.5225219127550847,
+      "grad_norm": 23.82236099243164,
+      "learning_rate": 4.489615854486223e-05,
+      "loss": 0.1801,
+      "step": 3860
+    },
+    {
+      "epoch": 0.5252292801786862,
+      "grad_norm": 1.3276259899139404,
+      "learning_rate": 4.4869010452015745e-05,
+      "loss": 0.1351,
+      "step": 3880
+    },
+    {
+      "epoch": 0.5279366476022878,
+      "grad_norm": 0.8841078877449036,
+      "learning_rate": 4.4841862359169276e-05,
+      "loss": 0.1596,
+      "step": 3900
+    },
+    {
+      "epoch": 0.5279366476022878,
+      "eval_loss": 0.06823331862688065,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.4954,
+      "eval_samples_per_second": 4.329,
+      "eval_steps_per_second": 1.082,
+      "step": 3900
+    },
+    {
+      "epoch": 0.5306440150258892,
+      "grad_norm": 0.6198431253433228,
+      "learning_rate": 4.481471426632279e-05,
+      "loss": 0.0834,
+      "step": 3920
+    },
+    {
+      "epoch": 0.5333513824494907,
+      "grad_norm": 0.3783263564109802,
+      "learning_rate": 4.478756617347632e-05,
+      "loss": 0.171,
+      "step": 3940
+    },
+    {
+      "epoch": 0.5360587498730921,
+      "grad_norm": 0.045466240495443344,
+      "learning_rate": 4.476041808062984e-05,
+      "loss": 0.049,
+      "step": 3960
+    },
+    {
+      "epoch": 0.5387661172966937,
+      "grad_norm": 4.255916118621826,
+      "learning_rate": 4.473326998778336e-05,
+      "loss": 0.1866,
+      "step": 3980
+    },
+    {
+      "epoch": 0.5414734847202951,
+      "grad_norm": 0.548932671546936,
+      "learning_rate": 4.470612189493688e-05,
+      "loss": 0.1338,
+      "step": 4000
+    },
+    {
+      "epoch": 0.5414734847202951,
+      "eval_loss": 0.05355897173285484,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.7915,
+      "eval_samples_per_second": 4.356,
+      "eval_steps_per_second": 1.089,
+      "step": 4000
+    },
+    {
+      "epoch": 0.5441808521438966,
+      "grad_norm": 3.635627508163452,
+      "learning_rate": 4.4678973802090405e-05,
+      "loss": 0.2088,
+      "step": 4020
+    },
+    {
+      "epoch": 0.546888219567498,
+      "grad_norm": 1.3527976274490356,
+      "learning_rate": 4.465182570924393e-05,
+      "loss": 0.1462,
+      "step": 4040
+    },
+    {
+      "epoch": 0.5495955869910996,
+      "grad_norm": 35.63922882080078,
+      "learning_rate": 4.4624677616397446e-05,
+      "loss": 0.1028,
+      "step": 4060
+    },
+    {
+      "epoch": 0.552302954414701,
+      "grad_norm": 1.668758511543274,
+      "learning_rate": 4.459752952355098e-05,
+      "loss": 0.1783,
+      "step": 4080
+    },
+    {
+      "epoch": 0.5550103218383025,
+      "grad_norm": 0.10880416631698608,
+      "learning_rate": 4.4570381430704494e-05,
+      "loss": 0.1616,
+      "step": 4100
+    },
+    {
+      "epoch": 0.5550103218383025,
+      "eval_loss": 0.05879218131303787,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.8074,
+      "eval_samples_per_second": 4.355,
+      "eval_steps_per_second": 1.089,
+      "step": 4100
+    },
+    {
+      "epoch": 0.5577176892619039,
+      "grad_norm": 4.690010070800781,
+      "learning_rate": 4.454323333785802e-05,
+      "loss": 0.122,
+      "step": 4120
+    },
+    {
+      "epoch": 0.5604250566855055,
+      "grad_norm": 0.06354126334190369,
+      "learning_rate": 4.451608524501154e-05,
+      "loss": 0.1858,
+      "step": 4140
+    },
+    {
+      "epoch": 0.5631324241091069,
+      "grad_norm": 7.339563846588135,
+      "learning_rate": 4.448893715216506e-05,
+      "loss": 0.092,
+      "step": 4160
+    },
+    {
+      "epoch": 0.5658397915327084,
+      "grad_norm": 0.001140138367190957,
+      "learning_rate": 4.446178905931859e-05,
+      "loss": 0.1959,
+      "step": 4180
+    },
+    {
+      "epoch": 0.5685471589563098,
+      "grad_norm": 0.7119511365890503,
+      "learning_rate": 4.443464096647211e-05,
+      "loss": 0.134,
+      "step": 4200
+    },
+    {
+      "epoch": 0.5685471589563098,
+      "eval_loss": 0.052126504480838776,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.3761,
+      "eval_samples_per_second": 4.334,
+      "eval_steps_per_second": 1.083,
+      "step": 4200
+    },
+    {
+      "epoch": 0.5712545263799114,
+      "grad_norm": 0.4891607463359833,
+      "learning_rate": 4.440749287362563e-05,
+      "loss": 0.0518,
+      "step": 4220
+    },
+    {
+      "epoch": 0.5739618938035128,
+      "grad_norm": 16.58160400390625,
+      "learning_rate": 4.4380344780779155e-05,
+      "loss": 0.1553,
+      "step": 4240
+    },
+    {
+      "epoch": 0.5766692612271143,
+      "grad_norm": 10.971266746520996,
+      "learning_rate": 4.435319668793267e-05,
+      "loss": 0.1916,
+      "step": 4260
+    },
+    {
+      "epoch": 0.5793766286507157,
+      "grad_norm": 0.2085973471403122,
+      "learning_rate": 4.4326048595086196e-05,
+      "loss": 0.1525,
+      "step": 4280
+    },
+    {
+      "epoch": 0.5820839960743173,
+      "grad_norm": 2.1405222415924072,
+      "learning_rate": 4.429890050223972e-05,
+      "loss": 0.1593,
+      "step": 4300
+    },
+    {
+      "epoch": 0.5820839960743173,
+      "eval_loss": 0.06334167718887329,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 112.7876,
+      "eval_samples_per_second": 4.433,
+      "eval_steps_per_second": 1.108,
+      "step": 4300
+    },
+    {
+      "epoch": 0.5847913634979187,
+      "grad_norm": 0.03152284771203995,
+      "learning_rate": 4.427175240939324e-05,
+      "loss": 0.2384,
+      "step": 4320
+    },
+    {
+      "epoch": 0.5874987309215202,
+      "grad_norm": 1.505940556526184,
+      "learning_rate": 4.424460431654677e-05,
+      "loss": 0.1345,
+      "step": 4340
+    },
+    {
+      "epoch": 0.5902060983451216,
+      "grad_norm": 2.4096856117248535,
+      "learning_rate": 4.4217456223700284e-05,
+      "loss": 0.1783,
+      "step": 4360
+    },
+    {
+      "epoch": 0.5929134657687232,
+      "grad_norm": 9.761833190917969,
+      "learning_rate": 4.419030813085381e-05,
+      "loss": 0.1601,
+      "step": 4380
+    },
+    {
+      "epoch": 0.5956208331923246,
+      "grad_norm": 0.4443033039569855,
+      "learning_rate": 4.416316003800733e-05,
+      "loss": 0.1302,
+      "step": 4400
+    },
+    {
+      "epoch": 0.5956208331923246,
+      "eval_loss": 0.08022167533636093,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.5736,
+      "eval_samples_per_second": 4.364,
+      "eval_steps_per_second": 1.091,
+      "step": 4400
+    },
+    {
+      "epoch": 0.5983282006159261,
+      "grad_norm": 2.908780574798584,
+      "learning_rate": 4.4136011945160856e-05,
+      "loss": 0.2055,
+      "step": 4420
+    },
+    {
+      "epoch": 0.6010355680395275,
+      "grad_norm": 48.880680084228516,
+      "learning_rate": 4.410886385231437e-05,
+      "loss": 0.2362,
+      "step": 4440
+    },
+    {
+      "epoch": 0.6037429354631291,
+      "grad_norm": 5.180681228637695,
+      "learning_rate": 4.4081715759467904e-05,
+      "loss": 0.2119,
+      "step": 4460
+    },
+    {
+      "epoch": 0.6064503028867305,
+      "grad_norm": 19.13489532470703,
+      "learning_rate": 4.405456766662142e-05,
+      "loss": 0.185,
+      "step": 4480
+    },
+    {
+      "epoch": 0.609157670310332,
+      "grad_norm": 11.651802062988281,
+      "learning_rate": 4.4027419573774945e-05,
+      "loss": 0.0693,
+      "step": 4500
+    },
+    {
+      "epoch": 0.609157670310332,
+      "eval_loss": 0.06786096096038818,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.8317,
+      "eval_samples_per_second": 4.317,
+      "eval_steps_per_second": 1.079,
+      "step": 4500
+    },
+    {
+      "epoch": 0.6118650377339334,
+      "grad_norm": 0.008749375119805336,
+      "learning_rate": 4.400027148092847e-05,
+      "loss": 0.1525,
+      "step": 4520
+    },
+    {
+      "epoch": 0.614572405157535,
+      "grad_norm": 1.9322203397750854,
+      "learning_rate": 4.3973123388081986e-05,
+      "loss": 0.1135,
+      "step": 4540
+    },
+    {
+      "epoch": 0.6172797725811364,
+      "grad_norm": 7.96317195892334,
+      "learning_rate": 4.3945975295235516e-05,
+      "loss": 0.1073,
+      "step": 4560
+    },
+    {
+      "epoch": 0.6199871400047379,
+      "grad_norm": 0.061754632741212845,
+      "learning_rate": 4.3918827202389034e-05,
+      "loss": 0.1121,
+      "step": 4580
+    },
+    {
+      "epoch": 0.6226945074283393,
+      "grad_norm": 20.79821014404297,
+      "learning_rate": 4.389167910954256e-05,
+      "loss": 0.2335,
+      "step": 4600
+    },
+    {
+      "epoch": 0.6226945074283393,
+      "eval_loss": 0.05240228772163391,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 113.8695,
+      "eval_samples_per_second": 4.391,
+      "eval_steps_per_second": 1.098,
+      "step": 4600
+    },
+    {
+      "epoch": 0.6254018748519409,
+      "grad_norm": 0.7108599543571472,
+      "learning_rate": 4.386453101669608e-05,
+      "loss": 0.1733,
+      "step": 4620
+    },
+    {
+      "epoch": 0.6281092422755423,
+      "grad_norm": 5.500114917755127,
+      "learning_rate": 4.38373829238496e-05,
+      "loss": 0.0439,
+      "step": 4640
+    },
+    {
+      "epoch": 0.6308166096991438,
+      "grad_norm": 0.4140441119670868,
+      "learning_rate": 4.381023483100312e-05,
+      "loss": 0.1013,
+      "step": 4660
+    },
+    {
+      "epoch": 0.6335239771227452,
+      "grad_norm": 27.898061752319336,
+      "learning_rate": 4.3783086738156646e-05,
+      "loss": 0.1334,
+      "step": 4680
+    },
+    {
+      "epoch": 0.6362313445463468,
+      "grad_norm": 0.002663947641849518,
+      "learning_rate": 4.375593864531017e-05,
+      "loss": 0.0407,
+      "step": 4700
+    },
+    {
+      "epoch": 0.6362313445463468,
+      "eval_loss": 0.08497549593448639,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.0355,
+      "eval_samples_per_second": 4.346,
+      "eval_steps_per_second": 1.087,
+      "step": 4700
+    },
+    {
+      "epoch": 0.6389387119699482,
+      "grad_norm": 0.06536103039979935,
+      "learning_rate": 4.3728790552463694e-05,
+      "loss": 0.064,
+      "step": 4720
+    },
+    {
+      "epoch": 0.6416460793935497,
+      "grad_norm": 1.3059289455413818,
+      "learning_rate": 4.370164245961722e-05,
+      "loss": 0.2077,
+      "step": 4740
+    },
+    {
+      "epoch": 0.6443534468171511,
+      "grad_norm": 0.029632871970534325,
+      "learning_rate": 4.3674494366770735e-05,
+      "loss": 0.1241,
+      "step": 4760
+    },
+    {
+      "epoch": 0.6470608142407527,
+      "grad_norm": 0.18283845484256744,
+      "learning_rate": 4.364734627392426e-05,
+      "loss": 0.1071,
+      "step": 4780
+    },
+    {
+      "epoch": 0.6497681816643541,
+      "grad_norm": 0.02363455854356289,
+      "learning_rate": 4.362019818107778e-05,
+      "loss": 0.0742,
+      "step": 4800
+    },
+    {
+      "epoch": 0.6497681816643541,
+      "eval_loss": 0.08371000736951828,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.0517,
+      "eval_samples_per_second": 4.346,
+      "eval_steps_per_second": 1.086,
+      "step": 4800
+    },
+    {
+      "epoch": 0.6524755490879556,
+      "grad_norm": 15.434126853942871,
+      "learning_rate": 4.35930500882313e-05,
+      "loss": 0.3661,
+      "step": 4820
+    },
+    {
+      "epoch": 0.655182916511557,
+      "grad_norm": 0.0013404142810031772,
+      "learning_rate": 4.356590199538483e-05,
+      "loss": 0.0634,
+      "step": 4840
+    },
+    {
+      "epoch": 0.6578902839351586,
+      "grad_norm": 9.277833938598633,
+      "learning_rate": 4.353875390253835e-05,
+      "loss": 0.2201,
+      "step": 4860
+    },
+    {
+      "epoch": 0.66059765135876,
+      "grad_norm": 9.396682739257812,
+      "learning_rate": 4.351160580969187e-05,
+      "loss": 0.2939,
+      "step": 4880
+    },
+    {
+      "epoch": 0.6633050187823615,
+      "grad_norm": 13.4754638671875,
+      "learning_rate": 4.3484457716845395e-05,
+      "loss": 0.1117,
+      "step": 4900
+    },
+    {
+      "epoch": 0.6633050187823615,
+      "eval_loss": 0.06177794933319092,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 116.1116,
+      "eval_samples_per_second": 4.306,
+      "eval_steps_per_second": 1.077,
+      "step": 4900
+    },
+    {
+      "epoch": 0.6660123862059629,
+      "grad_norm": 0.6623813509941101,
+      "learning_rate": 4.345730962399891e-05,
+      "loss": 0.1209,
+      "step": 4920
+    },
+    {
+      "epoch": 0.6687197536295645,
+      "grad_norm": 0.0009588833199813962,
+      "learning_rate": 4.343016153115244e-05,
+      "loss": 0.1701,
+      "step": 4940
+    },
+    {
+      "epoch": 0.6714271210531659,
+      "grad_norm": 0.20336271822452545,
+      "learning_rate": 4.340301343830596e-05,
+      "loss": 0.28,
+      "step": 4960
+    },
+    {
+      "epoch": 0.6741344884767674,
+      "grad_norm": 3.593808174133301,
+      "learning_rate": 4.3375865345459484e-05,
+      "loss": 0.1269,
+      "step": 4980
+    },
+    {
+      "epoch": 0.6768418559003688,
+      "grad_norm": 0.030866390094161034,
+      "learning_rate": 4.334871725261301e-05,
+      "loss": 0.0896,
+      "step": 5000
+    },
+    {
+      "epoch": 0.6768418559003688,
+      "eval_loss": 0.05724571645259857,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.0334,
+      "eval_samples_per_second": 4.347,
+      "eval_steps_per_second": 1.087,
+      "step": 5000
+    },
+    {
+      "epoch": 0.6795492233239704,
+      "grad_norm": 0.6822009086608887,
+      "learning_rate": 4.332156915976653e-05,
+      "loss": 0.1471,
+      "step": 5020
+    },
+    {
+      "epoch": 0.6822565907475718,
+      "grad_norm": 0.5282585620880127,
+      "learning_rate": 4.329442106692005e-05,
+      "loss": 0.1834,
+      "step": 5040
+    },
+    {
+      "epoch": 0.6849639581711733,
+      "grad_norm": 1.8356233835220337,
+      "learning_rate": 4.326727297407357e-05,
+      "loss": 0.1108,
+      "step": 5060
+    },
+    {
+      "epoch": 0.6876713255947747,
+      "grad_norm": 22.017763137817383,
+      "learning_rate": 4.32401248812271e-05,
+      "loss": 0.1741,
+      "step": 5080
+    },
+    {
+      "epoch": 0.6903786930183763,
+      "grad_norm": 22.150096893310547,
+      "learning_rate": 4.3212976788380614e-05,
+      "loss": 0.1629,
+      "step": 5100
+    },
+    {
+      "epoch": 0.6903786930183763,
+      "eval_loss": 0.06724371761083603,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.1009,
+      "eval_samples_per_second": 4.344,
+      "eval_steps_per_second": 1.086,
+      "step": 5100
+    },
+    {
+      "epoch": 0.6930860604419777,
+      "grad_norm": 1.9554486274719238,
+      "learning_rate": 4.3185828695534144e-05,
+      "loss": 0.2799,
+      "step": 5120
+    },
+    {
+      "epoch": 0.6957934278655792,
+      "grad_norm": 2.001250743865967,
+      "learning_rate": 4.315868060268766e-05,
+      "loss": 0.1485,
+      "step": 5140
+    },
+    {
+      "epoch": 0.6985007952891806,
+      "grad_norm": 13.129595756530762,
+      "learning_rate": 4.3131532509841185e-05,
+      "loss": 0.1245,
+      "step": 5160
+    },
+    {
+      "epoch": 0.7012081627127822,
+      "grad_norm": 0.695175051689148,
+      "learning_rate": 4.310438441699471e-05,
+      "loss": 0.2419,
+      "step": 5180
+    },
+    {
+      "epoch": 0.7039155301363836,
+      "grad_norm": 1.0608196258544922,
+      "learning_rate": 4.3077236324148226e-05,
+      "loss": 0.1922,
+      "step": 5200
+    },
+    {
+      "epoch": 0.7039155301363836,
+      "eval_loss": 0.06558438390493393,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 121.1736,
+      "eval_samples_per_second": 4.126,
+      "eval_steps_per_second": 1.032,
+      "step": 5200
+    },
+    {
+      "epoch": 0.7066228975599851,
+      "grad_norm": 17.941152572631836,
+      "learning_rate": 4.305008823130176e-05,
+      "loss": 0.2501,
+      "step": 5220
+    },
+    {
+      "epoch": 0.7093302649835865,
+      "grad_norm": 0.3917510509490967,
+      "learning_rate": 4.3022940138455274e-05,
+      "loss": 0.0669,
+      "step": 5240
+    },
+    {
+      "epoch": 0.7120376324071881,
+      "grad_norm": 2.205418825149536,
+      "learning_rate": 4.29957920456088e-05,
+      "loss": 0.1968,
+      "step": 5260
+    },
+    {
+      "epoch": 0.7147449998307895,
+      "grad_norm": 1.7837224006652832,
+      "learning_rate": 4.296864395276232e-05,
+      "loss": 0.1016,
+      "step": 5280
+    },
+    {
+      "epoch": 0.717452367254391,
+      "grad_norm": 0.15405257046222687,
+      "learning_rate": 4.294149585991584e-05,
+      "loss": 0.2557,
+      "step": 5300
+    },
+    {
+      "epoch": 0.717452367254391,
+      "eval_loss": 0.07933684438467026,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.6706,
+      "eval_samples_per_second": 4.36,
+      "eval_steps_per_second": 1.09,
+      "step": 5300
+    },
+    {
+      "epoch": 0.7201597346779924,
+      "grad_norm": 16.733394622802734,
+      "learning_rate": 4.291434776706936e-05,
+      "loss": 0.079,
+      "step": 5320
+    },
+    {
+      "epoch": 0.722867102101594,
+      "grad_norm": 0.18252775073051453,
+      "learning_rate": 4.288719967422289e-05,
+      "loss": 0.1165,
+      "step": 5340
+    },
+    {
+      "epoch": 0.7255744695251954,
+      "grad_norm": 0.0028702563140541315,
+      "learning_rate": 4.286005158137641e-05,
+      "loss": 0.2177,
+      "step": 5360
+    },
+    {
+      "epoch": 0.7282818369487969,
+      "grad_norm": 0.00213778973557055,
+      "learning_rate": 4.2832903488529935e-05,
+      "loss": 0.0568,
+      "step": 5380
+    },
+    {
+      "epoch": 0.7309892043723983,
+      "grad_norm": 0.09297258406877518,
+      "learning_rate": 4.280575539568346e-05,
+      "loss": 0.0597,
+      "step": 5400
+    },
+    {
+      "epoch": 0.7309892043723983,
+      "eval_loss": 0.07073411345481873,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.2895,
+      "eval_samples_per_second": 4.337,
+      "eval_steps_per_second": 1.084,
+      "step": 5400
+    },
+    {
+      "epoch": 0.7336965717959999,
+      "grad_norm": 0.8263906836509705,
+      "learning_rate": 4.2778607302836976e-05,
+      "loss": 0.2522,
+      "step": 5420
+    },
+    {
+      "epoch": 0.7364039392196013,
+      "grad_norm": 0.753628671169281,
+      "learning_rate": 4.27514592099905e-05,
+      "loss": 0.1264,
+      "step": 5440
+    },
+    {
+      "epoch": 0.7391113066432028,
+      "grad_norm": 20.244163513183594,
+      "learning_rate": 4.272431111714402e-05,
+      "loss": 0.1499,
+      "step": 5460
+    },
+    {
+      "epoch": 0.7418186740668042,
+      "grad_norm": 0.2985801100730896,
+      "learning_rate": 4.269716302429754e-05,
+      "loss": 0.0601,
+      "step": 5480
+    },
+    {
+      "epoch": 0.7445260414904058,
+      "grad_norm": 9.927957534790039,
+      "learning_rate": 4.267001493145107e-05,
+      "loss": 0.2035,
+      "step": 5500
+    },
+    {
+      "epoch": 0.7445260414904058,
+      "eval_loss": 0.082557812333107,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 116.3116,
+      "eval_samples_per_second": 4.299,
+      "eval_steps_per_second": 1.075,
+      "step": 5500
+    },
+    {
+      "epoch": 0.7472334089140072,
+      "grad_norm": 1.3682934045791626,
+      "learning_rate": 4.264286683860459e-05,
+      "loss": 0.1115,
+      "step": 5520
+    },
+    {
+      "epoch": 0.7499407763376087,
+      "grad_norm": 1.9332202672958374,
+      "learning_rate": 4.261571874575811e-05,
+      "loss": 0.2087,
+      "step": 5540
+    },
+    {
+      "epoch": 0.7526481437612101,
+      "grad_norm": 1.8556005954742432,
+      "learning_rate": 4.2588570652911636e-05,
+      "loss": 0.2302,
+      "step": 5560
+    },
+    {
+      "epoch": 0.7553555111848117,
+      "grad_norm": 0.06130724772810936,
+      "learning_rate": 4.256142256006515e-05,
+      "loss": 0.1079,
+      "step": 5580
+    },
+    {
+      "epoch": 0.7580628786084131,
+      "grad_norm": 0.0025592364836484194,
+      "learning_rate": 4.2534274467218684e-05,
+      "loss": 0.2651,
+      "step": 5600
+    },
+    {
+      "epoch": 0.7580628786084131,
+      "eval_loss": 0.07107679545879364,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.1747,
+      "eval_samples_per_second": 4.341,
+      "eval_steps_per_second": 1.085,
+      "step": 5600
+    },
+    {
+      "epoch": 0.7607702460320146,
+      "grad_norm": 0.1967998445034027,
+      "learning_rate": 4.25071263743722e-05,
+      "loss": 0.1033,
+      "step": 5620
+    },
+    {
+      "epoch": 0.763477613455616,
+      "grad_norm": 0.05981509014964104,
+      "learning_rate": 4.2479978281525725e-05,
+      "loss": 0.0682,
+      "step": 5640
+    },
+    {
+      "epoch": 0.7661849808792176,
+      "grad_norm": 0.013984935358166695,
+      "learning_rate": 4.245283018867925e-05,
+      "loss": 0.1213,
+      "step": 5660
+    },
+    {
+      "epoch": 0.768892348302819,
+      "grad_norm": 0.6655880808830261,
+      "learning_rate": 4.242568209583277e-05,
+      "loss": 0.3236,
+      "step": 5680
+    },
+    {
+      "epoch": 0.7715997157264205,
+      "grad_norm": 3.138312816619873,
+      "learning_rate": 4.239853400298629e-05,
+      "loss": 0.0734,
+      "step": 5700
+    },
+    {
+      "epoch": 0.7715997157264205,
+      "eval_loss": 0.09912095963954926,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.2143,
+      "eval_samples_per_second": 4.34,
+      "eval_steps_per_second": 1.085,
+      "step": 5700
+    },
+    {
+      "epoch": 0.774307083150022,
+      "grad_norm": 0.02605035901069641,
+      "learning_rate": 4.2371385910139813e-05,
+      "loss": 0.2264,
+      "step": 5720
+    },
+    {
+      "epoch": 0.7770144505736235,
+      "grad_norm": 0.10319148004055023,
+      "learning_rate": 4.234423781729334e-05,
+      "loss": 0.0924,
+      "step": 5740
+    },
+    {
+      "epoch": 0.779721817997225,
+      "grad_norm": 1.5601540803909302,
+      "learning_rate": 4.231708972444686e-05,
+      "loss": 0.2175,
+      "step": 5760
+    },
+    {
+      "epoch": 0.7824291854208264,
+      "grad_norm": 1.938568115234375,
+      "learning_rate": 4.2289941631600385e-05,
+      "loss": 0.2074,
+      "step": 5780
+    },
+    {
+      "epoch": 0.7851365528444278,
+      "grad_norm": 0.013515871949493885,
+      "learning_rate": 4.22627935387539e-05,
+      "loss": 0.1029,
+      "step": 5800
+    },
+    {
+      "epoch": 0.7851365528444278,
+      "eval_loss": 0.07735220342874527,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.5718,
+      "eval_samples_per_second": 4.326,
+      "eval_steps_per_second": 1.082,
+      "step": 5800
+    },
+    {
+      "epoch": 0.7878439202680294,
+      "grad_norm": 0.0078099388629198074,
+      "learning_rate": 4.2235645445907426e-05,
+      "loss": 0.2075,
+      "step": 5820
+    },
+    {
+      "epoch": 0.7905512876916309,
+      "grad_norm": 0.2638203203678131,
+      "learning_rate": 4.220849735306095e-05,
+      "loss": 0.1487,
+      "step": 5840
+    },
+    {
+      "epoch": 0.7932586551152323,
+      "grad_norm": 0.005258066579699516,
+      "learning_rate": 4.218134926021447e-05,
+      "loss": 0.1111,
+      "step": 5860
+    },
+    {
+      "epoch": 0.7959660225388338,
+      "grad_norm": 0.22548271715641022,
+      "learning_rate": 4.2154201167368e-05,
+      "loss": 0.0303,
+      "step": 5880
+    },
+    {
+      "epoch": 0.7986733899624353,
+      "grad_norm": 0.26001954078674316,
+      "learning_rate": 4.2127053074521515e-05,
+      "loss": 0.1185,
+      "step": 5900
+    },
+    {
+      "epoch": 0.7986733899624353,
+      "eval_loss": 0.0766243264079094,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.8216,
+      "eval_samples_per_second": 4.317,
+      "eval_steps_per_second": 1.079,
+      "step": 5900
+    },
+    {
+      "epoch": 0.8013807573860368,
+      "grad_norm": 2.299499988555908,
+      "learning_rate": 4.209990498167504e-05,
+      "loss": 0.4069,
+      "step": 5920
+    },
+    {
+      "epoch": 0.8040881248096382,
+      "grad_norm": 0.04289079084992409,
+      "learning_rate": 4.207275688882856e-05,
+      "loss": 0.0866,
+      "step": 5940
+    },
+    {
+      "epoch": 0.8067954922332397,
+      "grad_norm": 0.13593335449695587,
+      "learning_rate": 4.2045608795982086e-05,
+      "loss": 0.11,
+      "step": 5960
+    },
+    {
+      "epoch": 0.8095028596568412,
+      "grad_norm": 0.02351505681872368,
+      "learning_rate": 4.201846070313561e-05,
+      "loss": 0.2065,
+      "step": 5980
+    },
+    {
+      "epoch": 0.8122102270804427,
+      "grad_norm": 0.8734444975852966,
+      "learning_rate": 4.199131261028913e-05,
+      "loss": 0.1191,
+      "step": 6000
+    },
+    {
+      "epoch": 0.8122102270804427,
+      "eval_loss": 0.0750957652926445,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.1813,
+      "eval_samples_per_second": 4.341,
+      "eval_steps_per_second": 1.085,
+      "step": 6000
+    },
+    {
+      "epoch": 0.8149175945040441,
+      "grad_norm": 21.489639282226562,
+      "learning_rate": 4.196416451744265e-05,
+      "loss": 0.2108,
+      "step": 6020
+    },
+    {
+      "epoch": 0.8176249619276456,
+      "grad_norm": 14.320918083190918,
+      "learning_rate": 4.1937016424596175e-05,
+      "loss": 0.0872,
+      "step": 6040
+    },
+    {
+      "epoch": 0.8203323293512471,
+      "grad_norm": 0.14705882966518402,
+      "learning_rate": 4.19098683317497e-05,
+      "loss": 0.1089,
+      "step": 6060
+    },
+    {
+      "epoch": 0.8230396967748486,
+      "grad_norm": 0.018130503594875336,
+      "learning_rate": 4.1882720238903216e-05,
+      "loss": 0.0833,
+      "step": 6080
+    },
+    {
+      "epoch": 0.82574706419845,
+      "grad_norm": 0.07059603184461594,
+      "learning_rate": 4.185557214605674e-05,
+      "loss": 0.0807,
+      "step": 6100
+    },
+    {
+      "epoch": 0.82574706419845,
+      "eval_loss": 0.07839541137218475,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.0296,
+      "eval_samples_per_second": 4.347,
+      "eval_steps_per_second": 1.087,
+      "step": 6100
+    },
+    {
+      "epoch": 0.8284544316220515,
+      "grad_norm": 2.696398973464966,
+      "learning_rate": 4.1828424053210264e-05,
+      "loss": 0.1347,
+      "step": 6120
+    },
+    {
+      "epoch": 0.831161799045653,
+      "grad_norm": 0.7686747908592224,
+      "learning_rate": 4.180127596036378e-05,
+      "loss": 0.1131,
+      "step": 6140
+    },
+    {
+      "epoch": 0.8338691664692545,
+      "grad_norm": 0.569398045539856,
+      "learning_rate": 4.177412786751731e-05,
+      "loss": 0.103,
+      "step": 6160
+    },
+    {
+      "epoch": 0.8365765338928559,
+      "grad_norm": 17.672176361083984,
+      "learning_rate": 4.174697977467083e-05,
+      "loss": 0.1845,
+      "step": 6180
+    },
+    {
+      "epoch": 0.8392839013164574,
+      "grad_norm": 3.657409906387329,
+      "learning_rate": 4.171983168182436e-05,
+      "loss": 0.1087,
+      "step": 6200
+    },
+    {
+      "epoch": 0.8392839013164574,
+      "eval_loss": 0.09643953293561935,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 116.3381,
+      "eval_samples_per_second": 4.298,
+      "eval_steps_per_second": 1.074,
+      "step": 6200
+    },
+    {
+      "epoch": 0.8419912687400589,
+      "grad_norm": 0.23462818562984467,
+      "learning_rate": 4.169268358897788e-05,
+      "loss": 0.1078,
+      "step": 6220
+    },
+    {
+      "epoch": 0.8446986361636604,
+      "grad_norm": 6.32927131652832,
+      "learning_rate": 4.1665535496131394e-05,
+      "loss": 0.1914,
+      "step": 6240
+    },
+    {
+      "epoch": 0.8474060035872618,
+      "grad_norm": 1.2845321893692017,
+      "learning_rate": 4.1638387403284924e-05,
+      "loss": 0.0184,
+      "step": 6260
+    },
+    {
+      "epoch": 0.8501133710108633,
+      "grad_norm": 0.045799922198057175,
+      "learning_rate": 4.161123931043844e-05,
+      "loss": 0.0568,
+      "step": 6280
+    },
+    {
+      "epoch": 0.8528207384344648,
+      "grad_norm": 0.01587361842393875,
+      "learning_rate": 4.1584091217591965e-05,
+      "loss": 0.1459,
+      "step": 6300
+    },
+    {
+      "epoch": 0.8528207384344648,
+      "eval_loss": 0.0922488272190094,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.5988,
+      "eval_samples_per_second": 4.325,
+      "eval_steps_per_second": 1.081,
+      "step": 6300
+    },
+    {
+      "epoch": 0.8555281058580663,
+      "grad_norm": 0.11717058718204498,
+      "learning_rate": 4.155694312474549e-05,
+      "loss": 0.1182,
+      "step": 6320
+    },
+    {
+      "epoch": 0.8582354732816677,
+      "grad_norm": 1.3142938613891602,
+      "learning_rate": 4.152979503189901e-05,
+      "loss": 0.1918,
+      "step": 6340
+    },
+    {
+      "epoch": 0.8609428407052692,
+      "grad_norm": 0.06714503467082977,
+      "learning_rate": 4.150264693905254e-05,
+      "loss": 0.0725,
+      "step": 6360
+    },
+    {
+      "epoch": 0.8636502081288707,
+      "grad_norm": 12.575489044189453,
+      "learning_rate": 4.1475498846206054e-05,
+      "loss": 0.201,
+      "step": 6380
+    },
+    {
+      "epoch": 0.8663575755524722,
+      "grad_norm": 0.9065766930580139,
+      "learning_rate": 4.144835075335958e-05,
+      "loss": 0.1276,
+      "step": 6400
+    },
+    {
+      "epoch": 0.8663575755524722,
+      "eval_loss": 0.08115123957395554,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 117.9244,
+      "eval_samples_per_second": 4.24,
+      "eval_steps_per_second": 1.06,
+      "step": 6400
+    },
+    {
+      "epoch": 0.8690649429760736,
+      "grad_norm": 0.10184822231531143,
+      "learning_rate": 4.14212026605131e-05,
+      "loss": 0.1413,
+      "step": 6420
+    },
+    {
+      "epoch": 0.8717723103996751,
+      "grad_norm": 0.0267459936439991,
+      "learning_rate": 4.1394054567666626e-05,
+      "loss": 0.1017,
+      "step": 6440
+    },
+    {
+      "epoch": 0.8744796778232766,
+      "grad_norm": 0.059938717633485794,
+      "learning_rate": 4.136690647482014e-05,
+      "loss": 0.0481,
+      "step": 6460
+    },
+    {
+      "epoch": 0.8771870452468781,
+      "grad_norm": 24.361602783203125,
+      "learning_rate": 4.133975838197367e-05,
+      "loss": 0.1917,
+      "step": 6480
+    },
+    {
+      "epoch": 0.8798944126704795,
+      "grad_norm": 0.9935716986656189,
+      "learning_rate": 4.131261028912719e-05,
+      "loss": 0.1553,
+      "step": 6500
+    },
+    {
+      "epoch": 0.8798944126704795,
+      "eval_loss": 0.10655944794416428,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.5718,
+      "eval_samples_per_second": 4.364,
+      "eval_steps_per_second": 1.091,
+      "step": 6500
+    },
+    {
+      "epoch": 0.882601780094081,
+      "grad_norm": 17.362058639526367,
+      "learning_rate": 4.128546219628071e-05,
+      "loss": 0.2953,
+      "step": 6520
+    },
+    {
+      "epoch": 0.8853091475176825,
+      "grad_norm": 5.4117865562438965,
+      "learning_rate": 4.125831410343424e-05,
+      "loss": 0.1114,
+      "step": 6540
+    },
+    {
+      "epoch": 0.888016514941284,
+      "grad_norm": 0.013206146657466888,
+      "learning_rate": 4.1231166010587756e-05,
+      "loss": 0.2183,
+      "step": 6560
+    },
+    {
+      "epoch": 0.8907238823648854,
+      "grad_norm": 0.501082718372345,
+      "learning_rate": 4.1204017917741286e-05,
+      "loss": 0.0674,
+      "step": 6580
+    },
+    {
+      "epoch": 0.8934312497884869,
+      "grad_norm": 0.08269740641117096,
+      "learning_rate": 4.11768698248948e-05,
+      "loss": 0.0716,
+      "step": 6600
+    },
+    {
+      "epoch": 0.8934312497884869,
+      "eval_loss": 0.08250731229782104,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.7008,
+      "eval_samples_per_second": 4.359,
+      "eval_steps_per_second": 1.09,
+      "step": 6600
+    },
+    {
+      "epoch": 0.8961386172120884,
+      "grad_norm": 0.02271423488855362,
+      "learning_rate": 4.114972173204833e-05,
+      "loss": 0.1032,
+      "step": 6620
+    },
+    {
+      "epoch": 0.8988459846356899,
+      "grad_norm": 25.167842864990234,
+      "learning_rate": 4.112257363920185e-05,
+      "loss": 0.0961,
+      "step": 6640
+    },
+    {
+      "epoch": 0.9015533520592913,
+      "grad_norm": 0.049321677535772324,
+      "learning_rate": 4.109542554635537e-05,
+      "loss": 0.2755,
+      "step": 6660
+    },
+    {
+      "epoch": 0.9042607194828928,
+      "grad_norm": 18.843111038208008,
+      "learning_rate": 4.106827745350889e-05,
+      "loss": 0.1792,
+      "step": 6680
+    },
+    {
+      "epoch": 0.9069680869064943,
+      "grad_norm": 1.2584052085876465,
+      "learning_rate": 4.1041129360662416e-05,
+      "loss": 0.0299,
+      "step": 6700
+    },
+    {
+      "epoch": 0.9069680869064943,
+      "eval_loss": 0.07415125519037247,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.619,
+      "eval_samples_per_second": 4.325,
+      "eval_steps_per_second": 1.081,
+      "step": 6700
+    },
+    {
+      "epoch": 0.9096754543300958,
+      "grad_norm": 0.1885201781988144,
+      "learning_rate": 4.101398126781594e-05,
+      "loss": 0.1009,
+      "step": 6720
+    },
+    {
+      "epoch": 0.9123828217536972,
+      "grad_norm": 0.022826386615633965,
+      "learning_rate": 4.098683317496946e-05,
+      "loss": 0.0213,
+      "step": 6740
+    },
+    {
+      "epoch": 0.9150901891772987,
+      "grad_norm": 0.9626327157020569,
+      "learning_rate": 4.095968508212298e-05,
+      "loss": 0.2421,
+      "step": 6760
+    },
+    {
+      "epoch": 0.9177975566009002,
+      "grad_norm": 0.1741826981306076,
+      "learning_rate": 4.0932536989276505e-05,
+      "loss": 0.0697,
+      "step": 6780
+    },
+    {
+      "epoch": 0.9205049240245017,
+      "grad_norm": 12.775446891784668,
+      "learning_rate": 4.090538889643003e-05,
+      "loss": 0.0762,
+      "step": 6800
+    },
+    {
+      "epoch": 0.9205049240245017,
+      "eval_loss": 0.07394778728485107,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.0305,
+      "eval_samples_per_second": 4.347,
+      "eval_steps_per_second": 1.087,
+      "step": 6800
+    },
+    {
+      "epoch": 0.9232122914481031,
+      "grad_norm": 17.907012939453125,
+      "learning_rate": 4.087824080358355e-05,
+      "loss": 0.2498,
+      "step": 6820
+    },
+    {
+      "epoch": 0.9259196588717046,
+      "grad_norm": 0.01809736154973507,
+      "learning_rate": 4.085109271073707e-05,
+      "loss": 0.1663,
+      "step": 6840
+    },
+    {
+      "epoch": 0.9286270262953061,
+      "grad_norm": 0.7653181552886963,
+      "learning_rate": 4.08239446178906e-05,
+      "loss": 0.2626,
+      "step": 6860
+    },
+    {
+      "epoch": 0.9313343937189076,
+      "grad_norm": 0.08846959471702576,
+      "learning_rate": 4.079679652504412e-05,
+      "loss": 0.0397,
+      "step": 6880
+    },
+    {
+      "epoch": 0.934041761142509,
+      "grad_norm": 0.0012274185428395867,
+      "learning_rate": 4.076964843219764e-05,
+      "loss": 0.069,
+      "step": 6900
+    },
+    {
+      "epoch": 0.934041761142509,
+      "eval_loss": 0.05653444677591324,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.7193,
+      "eval_samples_per_second": 4.358,
+      "eval_steps_per_second": 1.09,
+      "step": 6900
+    },
+    {
+      "epoch": 0.9367491285661105,
+      "grad_norm": 0.34094715118408203,
+      "learning_rate": 4.0742500339351165e-05,
+      "loss": 0.0899,
+      "step": 6920
+    },
+    {
+      "epoch": 0.939456495989712,
+      "grad_norm": 0.6107985377311707,
+      "learning_rate": 4.071535224650468e-05,
+      "loss": 0.193,
+      "step": 6940
+    },
+    {
+      "epoch": 0.9421638634133135,
+      "grad_norm": 29.315858840942383,
+      "learning_rate": 4.0688204153658206e-05,
+      "loss": 0.2523,
+      "step": 6960
+    },
+    {
+      "epoch": 0.9448712308369149,
+      "grad_norm": 0.030056612566113472,
+      "learning_rate": 4.066105606081173e-05,
+      "loss": 0.042,
+      "step": 6980
+    },
+    {
+      "epoch": 0.9475785982605164,
+      "grad_norm": 3.3578696250915527,
+      "learning_rate": 4.0633907967965254e-05,
+      "loss": 0.1456,
+      "step": 7000
+    },
+    {
+      "epoch": 0.9475785982605164,
+      "eval_loss": 0.07208700478076935,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.5581,
+      "eval_samples_per_second": 4.327,
+      "eval_steps_per_second": 1.082,
+      "step": 7000
+    },
+    {
+      "epoch": 0.9502859656841179,
+      "grad_norm": 0.08940195292234421,
+      "learning_rate": 4.060675987511878e-05,
+      "loss": 0.1813,
+      "step": 7020
+    },
+    {
+      "epoch": 0.9529933331077194,
+      "grad_norm": 0.05165726691484451,
+      "learning_rate": 4.0579611782272295e-05,
+      "loss": 0.0218,
+      "step": 7040
+    },
+    {
+      "epoch": 0.9557007005313208,
+      "grad_norm": 4.932803153991699,
+      "learning_rate": 4.055246368942582e-05,
+      "loss": 0.1261,
+      "step": 7060
+    },
+    {
+      "epoch": 0.9584080679549223,
+      "grad_norm": 0.024785397574305534,
+      "learning_rate": 4.052531559657934e-05,
+      "loss": 0.1756,
+      "step": 7080
+    },
+    {
+      "epoch": 0.9611154353785238,
+      "grad_norm": 4.420279026031494,
+      "learning_rate": 4.0498167503732866e-05,
+      "loss": 0.1332,
+      "step": 7100
+    },
+    {
+      "epoch": 0.9611154353785238,
+      "eval_loss": 0.06981143355369568,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.8045,
+      "eval_samples_per_second": 4.355,
+      "eval_steps_per_second": 1.089,
+      "step": 7100
+    },
+    {
+      "epoch": 0.9638228028021253,
+      "grad_norm": 0.11680381745100021,
+      "learning_rate": 4.0471019410886384e-05,
+      "loss": 0.1575,
+      "step": 7120
+    },
+    {
+      "epoch": 0.9665301702257267,
+      "grad_norm": 0.8033447265625,
+      "learning_rate": 4.0443871318039914e-05,
+      "loss": 0.1011,
+      "step": 7140
+    },
+    {
+      "epoch": 0.9692375376493282,
+      "grad_norm": 0.29050976037979126,
+      "learning_rate": 4.041672322519343e-05,
+      "loss": 0.192,
+      "step": 7160
+    },
+    {
+      "epoch": 0.9719449050729297,
+      "grad_norm": 0.09221057593822479,
+      "learning_rate": 4.0389575132346955e-05,
+      "loss": 0.1809,
+      "step": 7180
+    },
+    {
+      "epoch": 0.9746522724965312,
+      "grad_norm": 0.0891430452466011,
+      "learning_rate": 4.036242703950048e-05,
+      "loss": 0.233,
+      "step": 7200
+    },
+    {
+      "epoch": 0.9746522724965312,
+      "eval_loss": 0.04730875417590141,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.9057,
+      "eval_samples_per_second": 4.351,
+      "eval_steps_per_second": 1.088,
+      "step": 7200
+    },
+    {
+      "epoch": 0.9773596399201326,
+      "grad_norm": 0.8720383644104004,
+      "learning_rate": 4.0335278946653996e-05,
+      "loss": 0.1189,
+      "step": 7220
+    },
+    {
+      "epoch": 0.9800670073437341,
+      "grad_norm": 0.02585684508085251,
+      "learning_rate": 4.030813085380753e-05,
+      "loss": 0.1412,
+      "step": 7240
+    },
+    {
+      "epoch": 0.9827743747673356,
+      "grad_norm": 0.012327141128480434,
+      "learning_rate": 4.0280982760961044e-05,
+      "loss": 0.1259,
+      "step": 7260
+    },
+    {
+      "epoch": 0.9854817421909371,
+      "grad_norm": 15.601494789123535,
+      "learning_rate": 4.025383466811457e-05,
+      "loss": 0.2373,
+      "step": 7280
+    },
+    {
+      "epoch": 0.9881891096145385,
+      "grad_norm": 2.5966286659240723,
+      "learning_rate": 4.022668657526809e-05,
+      "loss": 0.1293,
+      "step": 7300
+    },
+    {
+      "epoch": 0.9881891096145385,
+      "eval_loss": 0.06829043477773666,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.6143,
+      "eval_samples_per_second": 4.325,
+      "eval_steps_per_second": 1.081,
+      "step": 7300
+    },
+    {
+      "epoch": 0.99089647703814,
+      "grad_norm": 0.3134647607803345,
+      "learning_rate": 4.019953848242161e-05,
+      "loss": 0.0711,
+      "step": 7320
+    },
+    {
+      "epoch": 0.9936038444617415,
+      "grad_norm": 32.37504959106445,
+      "learning_rate": 4.017239038957513e-05,
+      "loss": 0.1052,
+      "step": 7340
+    },
+    {
+      "epoch": 0.996311211885343,
+      "grad_norm": 0.11635546386241913,
+      "learning_rate": 4.0145242296728657e-05,
+      "loss": 0.2258,
+      "step": 7360
+    },
+    {
+      "epoch": 0.9990185793089444,
+      "grad_norm": 0.011782504618167877,
+      "learning_rate": 4.011809420388218e-05,
+      "loss": 0.0385,
+      "step": 7380
+    },
+    {
+      "epoch": 1.001725946732546,
+      "grad_norm": 0.7141312956809998,
+      "learning_rate": 4.0090946111035704e-05,
+      "loss": 0.416,
+      "step": 7400
+    },
+    {
+      "epoch": 1.001725946732546,
+      "eval_loss": 0.07393650710582733,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.9306,
+      "eval_samples_per_second": 4.35,
+      "eval_steps_per_second": 1.088,
+      "step": 7400
+    },
+    {
+      "epoch": 1.0044333141561474,
+      "grad_norm": 0.0028216764330863953,
+      "learning_rate": 4.006379801818922e-05,
+      "loss": 0.103,
+      "step": 7420
+    },
+    {
+      "epoch": 1.0071406815797488,
+      "grad_norm": 0.07318446040153503,
+      "learning_rate": 4.0036649925342745e-05,
+      "loss": 0.046,
+      "step": 7440
+    },
+    {
+      "epoch": 1.0098480490033503,
+      "grad_norm": 0.06094740703701973,
+      "learning_rate": 4.000950183249627e-05,
+      "loss": 0.1924,
+      "step": 7460
+    },
+    {
+      "epoch": 1.012555416426952,
+      "grad_norm": 0.0017843634122982621,
+      "learning_rate": 3.998235373964979e-05,
+      "loss": 0.1091,
+      "step": 7480
+    },
+    {
+      "epoch": 1.0152627838505532,
+      "grad_norm": 0.830274760723114,
+      "learning_rate": 3.995520564680331e-05,
+      "loss": 0.0678,
+      "step": 7500
+    },
+    {
+      "epoch": 1.0152627838505532,
+      "eval_loss": 0.08040212839841843,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.4828,
+      "eval_samples_per_second": 4.33,
+      "eval_steps_per_second": 1.082,
+      "step": 7500
+    },
+    {
+      "epoch": 1.0179701512741548,
+      "grad_norm": 0.16321413218975067,
+      "learning_rate": 3.992805755395684e-05,
+      "loss": 0.0646,
+      "step": 7520
+    },
+    {
+      "epoch": 1.0206775186977564,
+      "grad_norm": 0.03706735000014305,
+      "learning_rate": 3.990090946111036e-05,
+      "loss": 0.1927,
+      "step": 7540
+    },
+    {
+      "epoch": 1.0233848861213577,
+      "grad_norm": 0.01869605854153633,
+      "learning_rate": 3.987376136826388e-05,
+      "loss": 0.0437,
+      "step": 7560
+    },
+    {
+      "epoch": 1.0260922535449593,
+      "grad_norm": 5.068511486053467,
+      "learning_rate": 3.9846613275417406e-05,
+      "loss": 0.1424,
+      "step": 7580
+    },
+    {
+      "epoch": 1.0287996209685606,
+      "grad_norm": 0.08981607854366302,
+      "learning_rate": 3.981946518257092e-05,
+      "loss": 0.2282,
+      "step": 7600
+    },
+    {
+      "epoch": 1.0287996209685606,
+      "eval_loss": 0.07650502771139145,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 114.7492,
+      "eval_samples_per_second": 4.357,
+      "eval_steps_per_second": 1.089,
+      "step": 7600
+    },
+    {
+      "epoch": 1.0315069883921621,
+      "grad_norm": 1.7685487270355225,
+      "learning_rate": 3.9792317089724454e-05,
+      "loss": 0.0231,
+      "step": 7620
+    },
+    {
+      "epoch": 1.0342143558157637,
+      "grad_norm": 0.3977906107902527,
+      "learning_rate": 3.976516899687797e-05,
+      "loss": 0.0819,
+      "step": 7640
+    },
+    {
+      "epoch": 1.036921723239365,
+      "grad_norm": 4.576267719268799,
+      "learning_rate": 3.9738020904031495e-05,
+      "loss": 0.1249,
+      "step": 7660
+    },
+    {
+      "epoch": 1.0396290906629666,
+      "grad_norm": 22.835840225219727,
+      "learning_rate": 3.971087281118502e-05,
+      "loss": 0.1024,
+      "step": 7680
+    },
+    {
+      "epoch": 1.0423364580865682,
+      "grad_norm": 2.8661038875579834,
+      "learning_rate": 3.9683724718338535e-05,
+      "loss": 0.1666,
+      "step": 7700
+    },
+    {
+      "epoch": 1.0423364580865682,
+      "eval_loss": 0.08426863700151443,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.3533,
+      "eval_samples_per_second": 4.335,
+      "eval_steps_per_second": 1.084,
+      "step": 7700
+    },
+    {
+      "epoch": 1.0450438255101695,
+      "grad_norm": 0.08293268084526062,
+      "learning_rate": 3.965657662549206e-05,
+      "loss": 0.0538,
+      "step": 7720
+    },
+    {
+      "epoch": 1.047751192933771,
+      "grad_norm": 0.09539427608251572,
+      "learning_rate": 3.962942853264558e-05,
+      "loss": 0.2608,
+      "step": 7740
+    },
+    {
+      "epoch": 1.0504585603573724,
+      "grad_norm": 8.918827056884766,
+      "learning_rate": 3.960228043979911e-05,
+      "loss": 0.2025,
+      "step": 7760
+    },
+    {
+      "epoch": 1.053165927780974,
+      "grad_norm": 1.5636457204818726,
+      "learning_rate": 3.9575132346952624e-05,
+      "loss": 0.134,
+      "step": 7780
+    },
+    {
+      "epoch": 1.0558732952045755,
+      "grad_norm": 1.9283950328826904,
+      "learning_rate": 3.9547984254106155e-05,
+      "loss": 0.1059,
+      "step": 7800
+    },
+    {
+      "epoch": 1.0558732952045755,
+      "eval_loss": 0.09331218898296356,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 116.4095,
+      "eval_samples_per_second": 4.295,
+      "eval_steps_per_second": 1.074,
+      "step": 7800
+    },
+    {
+      "epoch": 1.0585806626281768,
+      "grad_norm": 3.447388172149658,
+      "learning_rate": 3.952083616125967e-05,
+      "loss": 0.2881,
+      "step": 7820
+    },
+    {
+      "epoch": 1.0612880300517784,
+      "grad_norm": 0.010725182481110096,
+      "learning_rate": 3.9493688068413196e-05,
+      "loss": 0.1261,
+      "step": 7840
+    },
+    {
+      "epoch": 1.06399539747538,
+      "grad_norm": 0.0917816236615181,
+      "learning_rate": 3.946653997556672e-05,
+      "loss": 0.246,
+      "step": 7860
+    },
+    {
+      "epoch": 1.0667027648989813,
+      "grad_norm": 1.0219537019729614,
+      "learning_rate": 3.943939188272024e-05,
+      "loss": 0.0927,
+      "step": 7880
+    },
+    {
+      "epoch": 1.0694101323225829,
+      "grad_norm": 0.059237148612737656,
+      "learning_rate": 3.941224378987377e-05,
+      "loss": 0.1514,
+      "step": 7900
+    },
+    {
+      "epoch": 1.0694101323225829,
+      "eval_loss": 0.08446906507015228,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.3793,
+      "eval_samples_per_second": 4.334,
+      "eval_steps_per_second": 1.083,
+      "step": 7900
+    },
+    {
+      "epoch": 1.0721174997461844,
+      "grad_norm": 34.07741165161133,
+      "learning_rate": 3.9385095697027285e-05,
+      "loss": 0.0838,
+      "step": 7920
+    },
+    {
+      "epoch": 1.0748248671697858,
+      "grad_norm": 0.4585834741592407,
+      "learning_rate": 3.935794760418081e-05,
+      "loss": 0.1584,
+      "step": 7940
+    },
+    {
+      "epoch": 1.0775322345933873,
+      "grad_norm": 27.79207420349121,
+      "learning_rate": 3.933079951133433e-05,
+      "loss": 0.0999,
+      "step": 7960
+    },
+    {
+      "epoch": 1.0802396020169887,
+      "grad_norm": 0.5480608344078064,
+      "learning_rate": 3.930365141848785e-05,
+      "loss": 0.0918,
+      "step": 7980
+    },
+    {
+      "epoch": 1.0829469694405902,
+      "grad_norm": 17.493274688720703,
+      "learning_rate": 3.9276503325641373e-05,
+      "loss": 0.1292,
+      "step": 8000
+    },
+    {
+      "epoch": 1.0829469694405902,
+      "eval_loss": 0.06774063408374786,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.224,
+      "eval_samples_per_second": 4.339,
+      "eval_steps_per_second": 1.085,
+      "step": 8000
+    },
+    {
+      "epoch": 1.0856543368641918,
+      "grad_norm": 9.466602325439453,
+      "learning_rate": 3.92493552327949e-05,
+      "loss": 0.1294,
+      "step": 8020
+    },
+    {
+      "epoch": 1.088361704287793,
+      "grad_norm": 0.09533428400754929,
+      "learning_rate": 3.922220713994842e-05,
+      "loss": 0.0924,
+      "step": 8040
+    },
+    {
+      "epoch": 1.0910690717113947,
+      "grad_norm": 0.18998439610004425,
+      "learning_rate": 3.9195059047101945e-05,
+      "loss": 0.0505,
+      "step": 8060
+    },
+    {
+      "epoch": 1.093776439134996,
+      "grad_norm": 0.0072032492607831955,
+      "learning_rate": 3.916791095425547e-05,
+      "loss": 0.0635,
+      "step": 8080
+    },
+    {
+      "epoch": 1.0964838065585976,
+      "grad_norm": 0.26891425251960754,
+      "learning_rate": 3.9140762861408986e-05,
+      "loss": 0.0963,
+      "step": 8100
+    },
+    {
+      "epoch": 1.0964838065585976,
+      "eval_loss": 0.07367356866598129,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.0833,
+      "eval_samples_per_second": 4.345,
+      "eval_steps_per_second": 1.086,
+      "step": 8100
+    },
+    {
+      "epoch": 1.0991911739821991,
+      "grad_norm": 11.478422164916992,
+      "learning_rate": 3.911361476856251e-05,
+      "loss": 0.1339,
+      "step": 8120
+    },
+    {
+      "epoch": 1.1018985414058005,
+      "grad_norm": 0.5796290636062622,
+      "learning_rate": 3.9086466675716034e-05,
+      "loss": 0.2033,
+      "step": 8140
+    },
+    {
+      "epoch": 1.104605908829402,
+      "grad_norm": 0.02336627058684826,
+      "learning_rate": 3.905931858286955e-05,
+      "loss": 0.0583,
+      "step": 8160
+    },
+    {
+      "epoch": 1.1073132762530036,
+      "grad_norm": 1.0814790725708008,
+      "learning_rate": 3.903217049002308e-05,
+      "loss": 0.124,
+      "step": 8180
+    },
+    {
+      "epoch": 1.110020643676605,
+      "grad_norm": 0.42775824666023254,
+      "learning_rate": 3.90050223971766e-05,
+      "loss": 0.1272,
+      "step": 8200
+    },
+    {
+      "epoch": 1.110020643676605,
+      "eval_loss": 0.06376690417528152,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 115.2277,
+      "eval_samples_per_second": 4.339,
+      "eval_steps_per_second": 1.085,
+      "step": 8200
+    },
+    {
+      "epoch": 1.1127280111002065,
+      "grad_norm": 0.19235482811927795,
+      "learning_rate": 3.897787430433012e-05,
+      "loss": 0.1596,
+      "step": 8220
+    },
+    {
+      "epoch": 1.115435378523808,
+      "grad_norm": 0.5556570291519165,
+      "learning_rate": 3.8950726211483646e-05,
+      "loss": 0.2195,
+      "step": 8240
+    },
+    {
+      "epoch": 1.1181427459474094,
+      "grad_norm": 0.026464879512786865,
+      "learning_rate": 3.8923578118637164e-05,
+      "loss": 0.0748,
+      "step": 8260
+    },
+    {
+      "epoch": 1.120850113371011,
+      "grad_norm": 7.5615034103393555,
+      "learning_rate": 3.8896430025790694e-05,
+      "loss": 0.047,
+      "step": 8280
+    },
+    {
+      "epoch": 1.1235574807946123,
+      "grad_norm": 0.41822749376296997,
+      "learning_rate": 3.886928193294421e-05,
+      "loss": 0.0948,
+      "step": 8300
+    },
+    {
+      "epoch": 1.1235574807946123,
+      "eval_loss": 0.060398317873477936,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 104.2866,
+      "eval_samples_per_second": 4.794,
+      "eval_steps_per_second": 1.199,
+      "step": 8300
+    },
+    {
+      "epoch": 1.1262648482182138,
+      "grad_norm": 0.034160107374191284,
+      "learning_rate": 3.8842133840097735e-05,
+      "loss": 0.1445,
+      "step": 8320
+    },
+    {
+      "epoch": 1.1289722156418154,
+      "grad_norm": 0.40724292397499084,
+      "learning_rate": 3.881498574725126e-05,
+      "loss": 0.0616,
+      "step": 8340
+    },
+    {
+      "epoch": 1.1316795830654167,
+      "grad_norm": 0.1533338874578476,
+      "learning_rate": 3.8787837654404776e-05,
+      "loss": 0.0982,
+      "step": 8360
+    },
+    {
+      "epoch": 1.1343869504890183,
+      "grad_norm": 0.003590109059587121,
+      "learning_rate": 3.87606895615583e-05,
+      "loss": 0.1408,
+      "step": 8380
+    },
+    {
+      "epoch": 1.1370943179126196,
+      "grad_norm": 0.04526267945766449,
+      "learning_rate": 3.8733541468711824e-05,
+      "loss": 0.1955,
+      "step": 8400
+    },
+    {
+      "epoch": 1.1370943179126196,
+      "eval_loss": 0.06311442703008652,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 104.2656,
+      "eval_samples_per_second": 4.795,
+      "eval_steps_per_second": 1.199,
+      "step": 8400
+    },
+    {
+      "epoch": 1.1398016853362212,
+      "grad_norm": 3.631730318069458,
+      "learning_rate": 3.870639337586535e-05,
+      "loss": 0.1025,
+      "step": 8420
+    },
+    {
+      "epoch": 1.1425090527598227,
+      "grad_norm": 0.006099845748394728,
+      "learning_rate": 3.867924528301887e-05,
+      "loss": 0.0781,
+      "step": 8440
+    },
+    {
+      "epoch": 1.145216420183424,
+      "grad_norm": 0.0006692282040603459,
+      "learning_rate": 3.8652097190172396e-05,
+      "loss": 0.076,
+      "step": 8460
+    },
+    {
+      "epoch": 1.1479237876070256,
+      "grad_norm": 0.023256322368979454,
+      "learning_rate": 3.862494909732591e-05,
+      "loss": 0.1354,
+      "step": 8480
+    },
+    {
+      "epoch": 1.1506311550306272,
+      "grad_norm": 28.583494186401367,
+      "learning_rate": 3.8597801004479437e-05,
+      "loss": 0.1174,
+      "step": 8500
+    },
+    {
+      "epoch": 1.1506311550306272,
+      "eval_loss": 0.07084568589925766,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 105.2689,
+      "eval_samples_per_second": 4.75,
+      "eval_steps_per_second": 1.187,
+      "step": 8500
+    },
+    {
+      "epoch": 1.1533385224542285,
+      "grad_norm": 10.414568901062012,
+      "learning_rate": 3.857065291163296e-05,
+      "loss": 0.1554,
+      "step": 8520
+    },
+    {
+      "epoch": 1.15604588987783,
+      "grad_norm": 8.933186531066895,
+      "learning_rate": 3.854350481878648e-05,
+      "loss": 0.0685,
+      "step": 8540
+    },
+    {
+      "epoch": 1.1587532573014316,
+      "grad_norm": 2.6918463706970215,
+      "learning_rate": 3.851635672594001e-05,
+      "loss": 0.1234,
+      "step": 8560
+    },
+    {
+      "epoch": 1.161460624725033,
+      "grad_norm": 1.297598123550415,
+      "learning_rate": 3.8489208633093525e-05,
+      "loss": 0.0685,
+      "step": 8580
+    },
+    {
+      "epoch": 1.1641679921486345,
+      "grad_norm": 0.10164650529623032,
+      "learning_rate": 3.846206054024705e-05,
+      "loss": 0.0544,
+      "step": 8600
+    },
+    {
+      "epoch": 1.1641679921486345,
+      "eval_loss": 0.07730449736118317,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 104.8653,
+      "eval_samples_per_second": 4.768,
+      "eval_steps_per_second": 1.192,
+      "step": 8600
+    },
+    {
+      "epoch": 1.1668753595722359,
+      "grad_norm": 7.305291652679443,
+      "learning_rate": 3.843491244740057e-05,
+      "loss": 0.0794,
+      "step": 8620
+    },
+    {
+      "epoch": 1.1695827269958374,
+      "grad_norm": 0.07961925864219666,
+      "learning_rate": 3.840776435455409e-05,
+      "loss": 0.044,
+      "step": 8640
+    },
+    {
+      "epoch": 1.172290094419439,
+      "grad_norm": 0.023566043004393578,
+      "learning_rate": 3.838061626170762e-05,
+      "loss": 0.1417,
+      "step": 8660
+    },
+    {
+      "epoch": 1.1749974618430403,
+      "grad_norm": 0.15135520696640015,
+      "learning_rate": 3.835346816886114e-05,
+      "loss": 0.1456,
+      "step": 8680
+    },
+    {
+      "epoch": 1.1777048292666419,
+      "grad_norm": 0.07481474429368973,
+      "learning_rate": 3.832632007601466e-05,
+      "loss": 0.1329,
+      "step": 8700
+    },
+    {
+      "epoch": 1.1777048292666419,
+      "eval_loss": 0.07221387326717377,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 104.6128,
+      "eval_samples_per_second": 4.78,
+      "eval_steps_per_second": 1.195,
+      "step": 8700
+    },
+    {
+      "epoch": 1.1804121966902432,
+      "grad_norm": 1.943707823753357,
+      "learning_rate": 3.8299171983168186e-05,
+      "loss": 0.1643,
+      "step": 8720
+    },
+    {
+      "epoch": 1.1831195641138448,
+      "grad_norm": 0.20479469001293182,
+      "learning_rate": 3.827202389032171e-05,
+      "loss": 0.1327,
+      "step": 8740
+    },
+    {
+      "epoch": 1.1858269315374463,
+      "grad_norm": 1.116450548171997,
+      "learning_rate": 3.824487579747523e-05,
+      "loss": 0.1332,
+      "step": 8760
+    },
+    {
+      "epoch": 1.1885342989610477,
+      "grad_norm": 1.1040481328964233,
+      "learning_rate": 3.821772770462875e-05,
+      "loss": 0.1719,
+      "step": 8780
+    },
+    {
+      "epoch": 1.1912416663846492,
+      "grad_norm": 0.03437604010105133,
+      "learning_rate": 3.8190579611782274e-05,
+      "loss": 0.0317,
+      "step": 8800
+    },
+    {
+      "epoch": 1.1912416663846492,
+      "eval_loss": 0.09452595561742783,
+      "eval_model_preparation_time": 0.0086,
+      "eval_runtime": 104.2972,
+      "eval_samples_per_second": 4.794,
+      "eval_steps_per_second": 1.198,
+      "step": 8800
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 36935,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "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": 2.119610830408488e+18,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}