diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -2,3218 +2,7718 @@
   "best_global_step": null,
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 10.0,
+  "epoch": 5.0,
   "eval_steps": 500,
-  "global_step": 320,
+  "global_step": 770,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "entropy": 1.179257184267044,
-      "epoch": 0.032,
-      "grad_norm": 0.6963536143302917,
+      "entropy": 1.1441087871789932,
+      "epoch": 0.006498781478472786,
+      "grad_norm": 0.22594086825847626,
       "learning_rate": 0.0002,
-      "loss": 2.5494184494018555,
-      "mean_token_accuracy": 0.5379438996315002,
-      "num_tokens": 80264.0,
+      "loss": 2.612501859664917,
+      "mean_token_accuracy": 0.5085290372371674,
+      "num_tokens": 37502.0,
       "step": 1
     },
     {
-      "entropy": 1.2070488929748535,
-      "epoch": 0.064,
-      "grad_norm": 0.41231992840766907,
+      "entropy": 1.2685535103082657,
+      "epoch": 0.012997562956945572,
+      "grad_norm": 0.1942373365163803,
       "learning_rate": 0.0002,
-      "loss": 2.0961010456085205,
-      "mean_token_accuracy": 0.5814267545938492,
-      "num_tokens": 159573.0,
+      "loss": 2.2667126655578613,
+      "mean_token_accuracy": 0.5350659266114235,
+      "num_tokens": 75406.0,
       "step": 2
     },
     {
-      "entropy": 1.324254333972931,
-      "epoch": 0.096,
-      "grad_norm": 0.40169844031333923,
+      "entropy": 1.4654520899057388,
+      "epoch": 0.01949634443541836,
+      "grad_norm": 0.13897432386875153,
       "learning_rate": 0.0002,
-      "loss": 1.8021230697631836,
-      "mean_token_accuracy": 0.595734030008316,
-      "num_tokens": 237543.0,
+      "loss": 1.8304898738861084,
+      "mean_token_accuracy": 0.5590623021125793,
+      "num_tokens": 112903.0,
       "step": 3
     },
     {
-      "entropy": 1.363574594259262,
-      "epoch": 0.128,
-      "grad_norm": 0.37209051847457886,
+      "entropy": 1.4532236605882645,
+      "epoch": 0.025995125913891144,
+      "grad_norm": 0.11201836168766022,
       "learning_rate": 0.0002,
-      "loss": 1.554048776626587,
-      "mean_token_accuracy": 0.6193354427814484,
-      "num_tokens": 316870.0,
+      "loss": 1.496299147605896,
+      "mean_token_accuracy": 0.6046230122447014,
+      "num_tokens": 150359.0,
       "step": 4
     },
     {
-      "entropy": 1.3551697731018066,
-      "epoch": 0.16,
-      "grad_norm": 0.22899997234344482,
+      "entropy": 1.4331371188163757,
+      "epoch": 0.03249390739236393,
+      "grad_norm": 0.14954882860183716,
       "learning_rate": 0.0002,
-      "loss": 1.4150879383087158,
-      "mean_token_accuracy": 0.6334295570850372,
-      "num_tokens": 395448.0,
+      "loss": 1.3865656852722168,
+      "mean_token_accuracy": 0.6091600209474564,
+      "num_tokens": 187830.0,
       "step": 5
     },
     {
-      "entropy": 1.3589494228363037,
-      "epoch": 0.192,
-      "grad_norm": 0.1690169870853424,
+      "entropy": 1.3669009655714035,
+      "epoch": 0.03899268887083672,
+      "grad_norm": 0.09565649181604385,
       "learning_rate": 0.0002,
-      "loss": 1.2825372219085693,
-      "mean_token_accuracy": 0.6543458849191666,
-      "num_tokens": 473668.0,
+      "loss": 1.2659735679626465,
+      "mean_token_accuracy": 0.6358497440814972,
+      "num_tokens": 225297.0,
       "step": 6
     },
     {
-      "entropy": 1.3066083192825317,
-      "epoch": 0.224,
-      "grad_norm": 0.12765775620937347,
+      "entropy": 1.2914617359638214,
+      "epoch": 0.045491470349309504,
+      "grad_norm": 0.0539422333240509,
       "learning_rate": 0.0002,
-      "loss": 1.1806933879852295,
-      "mean_token_accuracy": 0.6675450056791306,
-      "num_tokens": 552353.0,
+      "loss": 1.1662700176239014,
+      "mean_token_accuracy": 0.6492680311203003,
+      "num_tokens": 262906.0,
       "step": 7
     },
     {
-      "entropy": 1.267838329076767,
-      "epoch": 0.256,
-      "grad_norm": 0.13029053807258606,
+      "entropy": 1.249274119734764,
+      "epoch": 0.05199025182778229,
+      "grad_norm": 0.051388416439294815,
       "learning_rate": 0.0002,
-      "loss": 1.1103739738464355,
-      "mean_token_accuracy": 0.6742708086967468,
-      "num_tokens": 631720.0,
+      "loss": 1.1255854368209839,
+      "mean_token_accuracy": 0.6475581750273705,
+      "num_tokens": 300403.0,
       "step": 8
     },
     {
-      "entropy": 1.2157377302646637,
-      "epoch": 0.288,
-      "grad_norm": 0.1366051882505417,
+      "entropy": 1.1595238000154495,
+      "epoch": 0.05848903330625508,
+      "grad_norm": 0.057235199958086014,
       "learning_rate": 0.0002,
-      "loss": 1.0521275997161865,
-      "mean_token_accuracy": 0.6783442795276642,
-      "num_tokens": 710686.0,
+      "loss": 1.0490708351135254,
+      "mean_token_accuracy": 0.6654231324791908,
+      "num_tokens": 337487.0,
       "step": 9
     },
     {
-      "entropy": 1.1331250071525574,
-      "epoch": 0.32,
-      "grad_norm": 0.14982439577579498,
+      "entropy": 1.0685205161571503,
+      "epoch": 0.06498781478472786,
+      "grad_norm": 0.0577155165374279,
       "learning_rate": 0.0002,
-      "loss": 0.9696346521377563,
-      "mean_token_accuracy": 0.6991669833660126,
-      "num_tokens": 789991.0,
+      "loss": 0.97551429271698,
+      "mean_token_accuracy": 0.6776877418160439,
+      "num_tokens": 374847.0,
       "step": 10
     },
     {
-      "entropy": 1.0510872900485992,
-      "epoch": 0.352,
-      "grad_norm": 0.11349732428789139,
+      "entropy": 0.9968453124165535,
+      "epoch": 0.07148659626320066,
+      "grad_norm": 0.04976983740925789,
       "learning_rate": 0.0002,
-      "loss": 0.9087550640106201,
-      "mean_token_accuracy": 0.7059329897165298,
-      "num_tokens": 869400.0,
+      "loss": 0.9292951822280884,
+      "mean_token_accuracy": 0.6831405088305473,
+      "num_tokens": 412186.0,
       "step": 11
     },
     {
-      "entropy": 0.947733610868454,
-      "epoch": 0.384,
-      "grad_norm": 0.08895156532526016,
+      "entropy": 0.9424244612455368,
+      "epoch": 0.07798537774167344,
+      "grad_norm": 0.37462693452835083,
       "learning_rate": 0.0002,
-      "loss": 0.8525652885437012,
-      "mean_token_accuracy": 0.7165533006191254,
-      "num_tokens": 950044.0,
+      "loss": 0.8883110284805298,
+      "mean_token_accuracy": 0.6893546730279922,
+      "num_tokens": 449659.0,
       "step": 12
     },
     {
-      "entropy": 0.8823248594999313,
-      "epoch": 0.416,
-      "grad_norm": 0.08783440291881561,
+      "entropy": 0.9025738090276718,
+      "epoch": 0.08448415922014622,
+      "grad_norm": 0.04783688113093376,
       "learning_rate": 0.0002,
-      "loss": 0.8236583471298218,
-      "mean_token_accuracy": 0.7169439494609833,
-      "num_tokens": 1030182.0,
+      "loss": 0.8548192977905273,
+      "mean_token_accuracy": 0.6972228810191154,
+      "num_tokens": 487236.0,
       "step": 13
     },
     {
-      "entropy": 0.8259010016918182,
-      "epoch": 0.448,
-      "grad_norm": 0.10150616616010666,
+      "entropy": 0.8390183448791504,
+      "epoch": 0.09098294069861901,
+      "grad_norm": 0.06745485961437225,
       "learning_rate": 0.0002,
-      "loss": 0.8072782754898071,
-      "mean_token_accuracy": 0.719487190246582,
-      "num_tokens": 1108491.0,
+      "loss": 0.8280504941940308,
+      "mean_token_accuracy": 0.7010790929198265,
+      "num_tokens": 524798.0,
       "step": 14
     },
     {
-      "entropy": 0.7661050707101822,
-      "epoch": 0.48,
-      "grad_norm": 0.0986262708902359,
+      "entropy": 0.7939982861280441,
+      "epoch": 0.09748172217709179,
+      "grad_norm": 0.10931341350078583,
       "learning_rate": 0.0002,
-      "loss": 0.7692936062812805,
-      "mean_token_accuracy": 0.7283643782138824,
-      "num_tokens": 1186724.0,
+      "loss": 0.8073402643203735,
+      "mean_token_accuracy": 0.702845998108387,
+      "num_tokens": 561826.0,
       "step": 15
     },
     {
-      "entropy": 0.7517514675855637,
-      "epoch": 0.512,
-      "grad_norm": 0.09310254454612732,
+      "entropy": 0.762756697833538,
+      "epoch": 0.10398050365556458,
+      "grad_norm": 0.03783512860536575,
       "learning_rate": 0.0002,
-      "loss": 0.7523272037506104,
-      "mean_token_accuracy": 0.7260711044073105,
-      "num_tokens": 1264323.0,
+      "loss": 0.7596306800842285,
+      "mean_token_accuracy": 0.7148320376873016,
+      "num_tokens": 599249.0,
       "step": 16
     },
     {
-      "entropy": 0.7328858375549316,
-      "epoch": 0.544,
-      "grad_norm": 0.0957314595580101,
+      "entropy": 0.7585324719548225,
+      "epoch": 0.11047928513403736,
+      "grad_norm": 0.039289139211177826,
       "learning_rate": 0.0002,
-      "loss": 0.7284715175628662,
-      "mean_token_accuracy": 0.7308152616024017,
-      "num_tokens": 1342220.0,
+      "loss": 0.7472726106643677,
+      "mean_token_accuracy": 0.7169229537248611,
+      "num_tokens": 636495.0,
       "step": 17
     },
     {
-      "entropy": 0.719617024064064,
-      "epoch": 0.576,
-      "grad_norm": 0.08244957029819489,
+      "entropy": 0.7434245273470879,
+      "epoch": 0.11697806661251016,
+      "grad_norm": 0.04286178946495056,
       "learning_rate": 0.0002,
-      "loss": 0.7138073444366455,
-      "mean_token_accuracy": 0.7331701219081879,
-      "num_tokens": 1419705.0,
+      "loss": 0.7349382638931274,
+      "mean_token_accuracy": 0.7195246070623398,
+      "num_tokens": 674172.0,
       "step": 18
     },
     {
-      "entropy": 0.6748920828104019,
-      "epoch": 0.608,
-      "grad_norm": 0.08215262740850449,
+      "entropy": 0.7307169139385223,
+      "epoch": 0.12347684809098294,
+      "grad_norm": 0.03696495294570923,
       "learning_rate": 0.0002,
-      "loss": 0.6698633432388306,
-      "mean_token_accuracy": 0.7485965937376022,
-      "num_tokens": 1500069.0,
+      "loss": 0.7205268144607544,
+      "mean_token_accuracy": 0.7222909703850746,
+      "num_tokens": 711746.0,
       "step": 19
     },
     {
-      "entropy": 0.6863600462675095,
-      "epoch": 0.64,
-      "grad_norm": 0.07416888326406479,
+      "entropy": 0.724890224635601,
+      "epoch": 0.12997562956945571,
+      "grad_norm": 0.03647635132074356,
       "learning_rate": 0.0002,
-      "loss": 0.683246374130249,
-      "mean_token_accuracy": 0.7402395009994507,
-      "num_tokens": 1578246.0,
+      "loss": 0.7131904363632202,
+      "mean_token_accuracy": 0.7218770682811737,
+      "num_tokens": 748966.0,
       "step": 20
     },
     {
-      "entropy": 0.6788438856601715,
-      "epoch": 0.672,
-      "grad_norm": 0.06753385812044144,
+      "entropy": 0.7279537171125412,
+      "epoch": 0.1364744110479285,
+      "grad_norm": 0.033042650669813156,
       "learning_rate": 0.0002,
-      "loss": 0.6719948053359985,
-      "mean_token_accuracy": 0.7404156625270844,
-      "num_tokens": 1656022.0,
+      "loss": 0.7094431519508362,
+      "mean_token_accuracy": 0.7240697368979454,
+      "num_tokens": 786631.0,
       "step": 21
     },
     {
-      "entropy": 0.6713939607143402,
-      "epoch": 0.704,
-      "grad_norm": 0.06065193563699722,
+      "entropy": 0.7210045382380486,
+      "epoch": 0.1429731925264013,
+      "grad_norm": 0.033061351627111435,
       "learning_rate": 0.0002,
-      "loss": 0.6523999571800232,
-      "mean_token_accuracy": 0.7474960833787918,
-      "num_tokens": 1734582.0,
+      "loss": 0.7019110918045044,
+      "mean_token_accuracy": 0.7280351296067238,
+      "num_tokens": 823906.0,
       "step": 22
     },
     {
-      "entropy": 0.6675977408885956,
-      "epoch": 0.736,
-      "grad_norm": 0.06263108551502228,
+      "entropy": 0.6902946382761002,
+      "epoch": 0.14947197400487408,
+      "grad_norm": 0.03491866961121559,
       "learning_rate": 0.0002,
-      "loss": 0.649960994720459,
-      "mean_token_accuracy": 0.7477194517850876,
-      "num_tokens": 1813025.0,
+      "loss": 0.683687150478363,
+      "mean_token_accuracy": 0.7348670437932014,
+      "num_tokens": 860977.0,
       "step": 23
     },
     {
-      "entropy": 0.6574471294879913,
-      "epoch": 0.768,
-      "grad_norm": 0.06619945168495178,
+      "entropy": 0.6839202269911766,
+      "epoch": 0.15597075548334688,
+      "grad_norm": 0.030243346467614174,
       "learning_rate": 0.0002,
-      "loss": 0.639019250869751,
-      "mean_token_accuracy": 0.7521554976701736,
-      "num_tokens": 1891103.0,
+      "loss": 0.673778772354126,
+      "mean_token_accuracy": 0.7354807555675507,
+      "num_tokens": 898933.0,
       "step": 24
     },
     {
-      "entropy": 0.649686262011528,
-      "epoch": 0.8,
-      "grad_norm": 0.054676756262779236,
+      "entropy": 0.6664924547076225,
+      "epoch": 0.16246953696181965,
+      "grad_norm": 0.026247479021549225,
       "learning_rate": 0.0002,
-      "loss": 0.6350966691970825,
-      "mean_token_accuracy": 0.752074345946312,
-      "num_tokens": 1969423.0,
+      "loss": 0.6572352051734924,
+      "mean_token_accuracy": 0.7394573912024498,
+      "num_tokens": 935795.0,
       "step": 25
     },
     {
-      "entropy": 0.6216799467802048,
-      "epoch": 0.832,
-      "grad_norm": 0.053581204265356064,
+      "entropy": 0.6681515946984291,
+      "epoch": 0.16896831844029245,
+      "grad_norm": 0.03119724616408348,
       "learning_rate": 0.0002,
-      "loss": 0.613818883895874,
-      "mean_token_accuracy": 0.7612275779247284,
-      "num_tokens": 2048674.0,
+      "loss": 0.6646086573600769,
+      "mean_token_accuracy": 0.7379100769758224,
+      "num_tokens": 973280.0,
       "step": 26
     },
     {
-      "entropy": 0.609996572136879,
-      "epoch": 0.864,
-      "grad_norm": 0.06112146005034447,
+      "entropy": 0.6617112681269646,
+      "epoch": 0.17546709991876522,
+      "grad_norm": 0.031551606953144073,
       "learning_rate": 0.0002,
-      "loss": 0.6045517325401306,
-      "mean_token_accuracy": 0.766285240650177,
-      "num_tokens": 2127851.0,
+      "loss": 0.6585603952407837,
+      "mean_token_accuracy": 0.7389705181121826,
+      "num_tokens": 1010622.0,
       "step": 27
     },
     {
-      "entropy": 0.6208399832248688,
-      "epoch": 0.896,
-      "grad_norm": 0.05334703251719475,
+      "entropy": 0.6501665636897087,
+      "epoch": 0.18196588139723802,
+      "grad_norm": 0.02211320586502552,
       "learning_rate": 0.0002,
-      "loss": 0.6149610877037048,
-      "mean_token_accuracy": 0.7601087689399719,
-      "num_tokens": 2206708.0,
+      "loss": 0.6490904092788696,
+      "mean_token_accuracy": 0.7440381571650505,
+      "num_tokens": 1047900.0,
       "step": 28
     },
     {
-      "entropy": 0.5961938500404358,
-      "epoch": 0.928,
-      "grad_norm": 0.05678229779005051,
+      "entropy": 0.6470993533730507,
+      "epoch": 0.18846466287571081,
+      "grad_norm": 0.027688423171639442,
       "learning_rate": 0.0002,
-      "loss": 0.5941039323806763,
-      "mean_token_accuracy": 0.768301397562027,
-      "num_tokens": 2286280.0,
+      "loss": 0.6543895602226257,
+      "mean_token_accuracy": 0.7402380779385567,
+      "num_tokens": 1085392.0,
       "step": 29
     },
     {
-      "entropy": 0.5905821919441223,
-      "epoch": 0.96,
-      "grad_norm": 0.05276153236627579,
+      "entropy": 0.6416624039411545,
+      "epoch": 0.19496344435418358,
+      "grad_norm": 0.025217361748218536,
       "learning_rate": 0.0002,
-      "loss": 0.5909000039100647,
-      "mean_token_accuracy": 0.7674511522054672,
-      "num_tokens": 2365563.0,
+      "loss": 0.651384711265564,
+      "mean_token_accuracy": 0.7433382496237755,
+      "num_tokens": 1122465.0,
       "step": 30
     },
     {
-      "entropy": 0.6017148941755295,
-      "epoch": 0.992,
-      "grad_norm": 0.04250482842326164,
+      "entropy": 0.6523652598261833,
+      "epoch": 0.20146222583265638,
+      "grad_norm": 0.025227373465895653,
       "learning_rate": 0.0002,
-      "loss": 0.602470874786377,
-      "mean_token_accuracy": 0.7644271552562714,
-      "num_tokens": 2444186.0,
+      "loss": 0.6592613458633423,
+      "mean_token_accuracy": 0.7378358989953995,
+      "num_tokens": 1160451.0,
       "step": 31
     },
     {
-      "entropy": 0.5969142913818359,
-      "epoch": 1.0,
-      "grad_norm": 0.056718822568655014,
+      "entropy": 0.6369072422385216,
+      "epoch": 0.20796100731112915,
+      "grad_norm": 0.021375838667154312,
       "learning_rate": 0.0002,
-      "loss": 0.5996057987213135,
-      "mean_token_accuracy": 0.7648024559020996,
-      "num_tokens": 2463804.0,
+      "loss": 0.6419180631637573,
+      "mean_token_accuracy": 0.7449744045734406,
+      "num_tokens": 1198041.0,
       "step": 32
     },
     {
-      "entropy": 0.5890887081623077,
-      "epoch": 1.032,
-      "grad_norm": 0.053444478660821915,
+      "entropy": 0.6256191805005074,
+      "epoch": 0.21445978878960195,
+      "grad_norm": 0.03092559427022934,
       "learning_rate": 0.0002,
-      "loss": 0.5883615016937256,
-      "mean_token_accuracy": 0.7671252638101578,
-      "num_tokens": 2544202.0,
+      "loss": 0.6407320499420166,
+      "mean_token_accuracy": 0.7460681945085526,
+      "num_tokens": 1235291.0,
       "step": 33
     },
     {
-      "entropy": 0.5835496038198471,
-      "epoch": 1.064,
-      "grad_norm": 0.053636617958545685,
+      "entropy": 0.6299648508429527,
+      "epoch": 0.22095857026807472,
+      "grad_norm": 0.021124541759490967,
       "learning_rate": 0.0002,
-      "loss": 0.581194281578064,
-      "mean_token_accuracy": 0.7708244174718857,
-      "num_tokens": 2623607.0,
+      "loss": 0.6359473466873169,
+      "mean_token_accuracy": 0.7471542730927467,
+      "num_tokens": 1272686.0,
       "step": 34
     },
     {
-      "entropy": 0.6006538420915604,
-      "epoch": 1.096,
-      "grad_norm": 0.043986763805150986,
+      "entropy": 0.6513500586152077,
+      "epoch": 0.22745735174654752,
+      "grad_norm": 0.020855363458395004,
       "learning_rate": 0.0002,
-      "loss": 0.5970234870910645,
-      "mean_token_accuracy": 0.7643807381391525,
-      "num_tokens": 2701501.0,
+      "loss": 0.6451584100723267,
+      "mean_token_accuracy": 0.7422506660223007,
+      "num_tokens": 1310182.0,
       "step": 35
     },
     {
-      "entropy": 0.5842598974704742,
-      "epoch": 1.1280000000000001,
-      "grad_norm": 0.039326492697000504,
+      "entropy": 0.6261186450719833,
+      "epoch": 0.23395613322502032,
+      "grad_norm": 0.019606754183769226,
       "learning_rate": 0.0002,
-      "loss": 0.5772466659545898,
-      "mean_token_accuracy": 0.7730486690998077,
-      "num_tokens": 2779960.0,
+      "loss": 0.6181178092956543,
+      "mean_token_accuracy": 0.7557711079716682,
+      "num_tokens": 1347558.0,
       "step": 36
     },
     {
-      "entropy": 0.5868044942617416,
-      "epoch": 1.16,
-      "grad_norm": 0.03627580776810646,
+      "entropy": 0.6535262390971184,
+      "epoch": 0.2404549147034931,
+      "grad_norm": 0.016229776665568352,
       "learning_rate": 0.0002,
-      "loss": 0.5823642015457153,
-      "mean_token_accuracy": 0.7698265165090561,
-      "num_tokens": 2858358.0,
+      "loss": 0.6474911570549011,
+      "mean_token_accuracy": 0.7401541844010353,
+      "num_tokens": 1385217.0,
       "step": 37
     },
     {
-      "entropy": 0.584501713514328,
-      "epoch": 1.192,
-      "grad_norm": 0.03921157494187355,
+      "entropy": 0.6392895579338074,
+      "epoch": 0.2469536961819659,
+      "grad_norm": 0.01589571312069893,
       "learning_rate": 0.0002,
-      "loss": 0.5830175280570984,
-      "mean_token_accuracy": 0.7707354426383972,
-      "num_tokens": 2937490.0,
+      "loss": 0.6345689296722412,
+      "mean_token_accuracy": 0.7464647218585014,
+      "num_tokens": 1422439.0,
       "step": 38
     },
     {
-      "entropy": 0.572060227394104,
-      "epoch": 1.224,
-      "grad_norm": 0.036492444574832916,
+      "entropy": 0.6444704979658127,
+      "epoch": 0.25345247766043866,
+      "grad_norm": 0.019169991835951805,
       "learning_rate": 0.0002,
-      "loss": 0.5748717784881592,
-      "mean_token_accuracy": 0.7743632346391678,
-      "num_tokens": 3015145.0,
+      "loss": 0.6417882442474365,
+      "mean_token_accuracy": 0.7431693077087402,
+      "num_tokens": 1459951.0,
       "step": 39
     },
     {
-      "entropy": 0.5737735033035278,
-      "epoch": 1.256,
-      "grad_norm": 0.0318453311920166,
+      "entropy": 0.6366401016712189,
+      "epoch": 0.25995125913891143,
+      "grad_norm": 0.016541991382837296,
       "learning_rate": 0.0002,
-      "loss": 0.5738308429718018,
-      "mean_token_accuracy": 0.7725711613893509,
-      "num_tokens": 3094640.0,
+      "loss": 0.6387360095977783,
+      "mean_token_accuracy": 0.7458900958299637,
+      "num_tokens": 1497263.0,
       "step": 40
     },
     {
-      "entropy": 0.5824941098690033,
-      "epoch": 1.288,
-      "grad_norm": 0.03465260937809944,
+      "entropy": 0.6309419572353363,
+      "epoch": 0.26645004061738425,
+      "grad_norm": 0.02144729532301426,
       "learning_rate": 0.0002,
-      "loss": 0.5811444520950317,
-      "mean_token_accuracy": 0.7699832618236542,
-      "num_tokens": 3173371.0,
+      "loss": 0.635954737663269,
+      "mean_token_accuracy": 0.7469696402549744,
+      "num_tokens": 1534631.0,
       "step": 41
     },
     {
-      "entropy": 0.5671045929193497,
-      "epoch": 1.32,
-      "grad_norm": 0.026680685579776764,
+      "entropy": 0.6426582857966423,
+      "epoch": 0.272948822095857,
+      "grad_norm": 0.017918268218636513,
       "learning_rate": 0.0002,
-      "loss": 0.561556875705719,
-      "mean_token_accuracy": 0.7784358263015747,
-      "num_tokens": 3252583.0,
+      "loss": 0.6445891261100769,
+      "mean_token_accuracy": 0.7427951768040657,
+      "num_tokens": 1572110.0,
       "step": 42
     },
     {
-      "entropy": 0.5744841694831848,
-      "epoch": 1.3519999999999999,
-      "grad_norm": 0.030942633748054504,
+      "entropy": 0.6238497421145439,
+      "epoch": 0.2794476035743298,
+      "grad_norm": 0.016322394832968712,
       "learning_rate": 0.0002,
-      "loss": 0.5703052282333374,
-      "mean_token_accuracy": 0.7730769366025925,
-      "num_tokens": 3332485.0,
+      "loss": 0.6238967180252075,
+      "mean_token_accuracy": 0.7503543570637703,
+      "num_tokens": 1608997.0,
       "step": 43
     },
     {
-      "entropy": 0.5706202536821365,
-      "epoch": 1.384,
-      "grad_norm": 0.03452180325984955,
+      "entropy": 0.6374280378222466,
+      "epoch": 0.2859463850528026,
+      "grad_norm": 0.0153085608035326,
       "learning_rate": 0.0002,
-      "loss": 0.5691436529159546,
-      "mean_token_accuracy": 0.7742962390184402,
-      "num_tokens": 3411148.0,
+      "loss": 0.6364343762397766,
+      "mean_token_accuracy": 0.745714507997036,
+      "num_tokens": 1646556.0,
       "step": 44
     },
     {
-      "entropy": 0.5706360638141632,
-      "epoch": 1.416,
-      "grad_norm": 0.02981080673635006,
+      "entropy": 0.6222607344388962,
+      "epoch": 0.2924451665312754,
+      "grad_norm": 0.016996094956994057,
       "learning_rate": 0.0002,
-      "loss": 0.5676863193511963,
-      "mean_token_accuracy": 0.7751949429512024,
-      "num_tokens": 3490105.0,
+      "loss": 0.6201090812683105,
+      "mean_token_accuracy": 0.7516758143901825,
+      "num_tokens": 1683700.0,
       "step": 45
     },
     {
-      "entropy": 0.573886975646019,
-      "epoch": 1.448,
-      "grad_norm": 0.03274830803275108,
+      "entropy": 0.636712834239006,
+      "epoch": 0.29894394800974816,
+      "grad_norm": 0.016749687492847443,
       "learning_rate": 0.0002,
-      "loss": 0.5683993697166443,
-      "mean_token_accuracy": 0.7725706845521927,
-      "num_tokens": 3569059.0,
+      "loss": 0.6368897557258606,
+      "mean_token_accuracy": 0.746493011713028,
+      "num_tokens": 1720661.0,
       "step": 46
     },
     {
-      "entropy": 0.5753068625926971,
-      "epoch": 1.48,
-      "grad_norm": 0.027408910915255547,
+      "entropy": 0.6459397748112679,
+      "epoch": 0.30544272948822093,
+      "grad_norm": 0.011651836335659027,
       "learning_rate": 0.0002,
-      "loss": 0.5700199604034424,
-      "mean_token_accuracy": 0.7736819535493851,
-      "num_tokens": 3647511.0,
+      "loss": 0.6419233083724976,
+      "mean_token_accuracy": 0.7443490549921989,
+      "num_tokens": 1758131.0,
       "step": 47
     },
     {
-      "entropy": 0.5750648379325867,
-      "epoch": 1.512,
-      "grad_norm": 0.02696501836180687,
+      "entropy": 0.6379885226488113,
+      "epoch": 0.31194151096669376,
+      "grad_norm": 0.013901753351092339,
       "learning_rate": 0.0002,
-      "loss": 0.5703588724136353,
-      "mean_token_accuracy": 0.7729254961013794,
-      "num_tokens": 3726951.0,
+      "loss": 0.6361222267150879,
+      "mean_token_accuracy": 0.746205136179924,
+      "num_tokens": 1795397.0,
       "step": 48
     },
     {
-      "entropy": 0.5790549218654633,
-      "epoch": 1.544,
-      "grad_norm": 0.02991427481174469,
+      "entropy": 0.6377697363495827,
+      "epoch": 0.31844029244516653,
+      "grad_norm": 0.014066919684410095,
       "learning_rate": 0.0002,
-      "loss": 0.5785025358200073,
-      "mean_token_accuracy": 0.7707958817481995,
-      "num_tokens": 3804732.0,
+      "loss": 0.6356197595596313,
+      "mean_token_accuracy": 0.7455531656742096,
+      "num_tokens": 1832991.0,
       "step": 49
     },
     {
-      "entropy": 0.5549423545598984,
-      "epoch": 1.576,
-      "grad_norm": 0.03170735761523247,
+      "entropy": 0.6351528614759445,
+      "epoch": 0.3249390739236393,
+      "grad_norm": 0.013286196626722813,
       "learning_rate": 0.0002,
-      "loss": 0.5581924319267273,
-      "mean_token_accuracy": 0.7788972407579422,
-      "num_tokens": 3884688.0,
+      "loss": 0.6371215581893921,
+      "mean_token_accuracy": 0.7454295605421066,
+      "num_tokens": 1870273.0,
       "step": 50
     },
     {
-      "entropy": 0.568594440817833,
-      "epoch": 1.608,
-      "grad_norm": 0.025594916194677353,
+      "entropy": 0.6069519519805908,
+      "epoch": 0.3314378554021121,
+      "grad_norm": 0.013732723891735077,
       "learning_rate": 0.0002,
-      "loss": 0.5706242322921753,
-      "mean_token_accuracy": 0.7738734632730484,
-      "num_tokens": 3964161.0,
+      "loss": 0.6103472709655762,
+      "mean_token_accuracy": 0.7536283582448959,
+      "num_tokens": 1907528.0,
       "step": 51
     },
     {
-      "entropy": 0.5575809478759766,
-      "epoch": 1.6400000000000001,
-      "grad_norm": 0.025340234860777855,
+      "entropy": 0.6297546997666359,
+      "epoch": 0.3379366368805849,
+      "grad_norm": 0.013711776584386826,
       "learning_rate": 0.0002,
-      "loss": 0.5602903962135315,
-      "mean_token_accuracy": 0.7778308838605881,
-      "num_tokens": 4043157.0,
+      "loss": 0.6334048509597778,
+      "mean_token_accuracy": 0.745865486562252,
+      "num_tokens": 1945255.0,
       "step": 52
     },
     {
-      "entropy": 0.5458956509828568,
-      "epoch": 1.6720000000000002,
-      "grad_norm": 0.02706434205174446,
+      "entropy": 0.6280607059597969,
+      "epoch": 0.34443541835905767,
+      "grad_norm": 0.01293123047798872,
       "learning_rate": 0.0002,
-      "loss": 0.5461863875389099,
-      "mean_token_accuracy": 0.78243587911129,
-      "num_tokens": 4123675.0,
+      "loss": 0.6310492753982544,
+      "mean_token_accuracy": 0.7467462942004204,
+      "num_tokens": 1982630.0,
       "step": 53
     },
     {
-      "entropy": 0.5649830847978592,
-      "epoch": 1.704,
-      "grad_norm": 0.030553704127669334,
+      "entropy": 0.6271316409111023,
+      "epoch": 0.35093419983753044,
+      "grad_norm": 0.01478263083845377,
       "learning_rate": 0.0002,
-      "loss": 0.5619728565216064,
-      "mean_token_accuracy": 0.7764615267515182,
-      "num_tokens": 4202826.0,
+      "loss": 0.6254487037658691,
+      "mean_token_accuracy": 0.7497461065649986,
+      "num_tokens": 2020399.0,
       "step": 54
     },
     {
-      "entropy": 0.5686902850866318,
-      "epoch": 1.736,
-      "grad_norm": 0.02532677724957466,
+      "entropy": 0.6377271711826324,
+      "epoch": 0.35743298131600326,
+      "grad_norm": 0.013956044800579548,
       "learning_rate": 0.0002,
-      "loss": 0.5656942129135132,
-      "mean_token_accuracy": 0.7758781611919403,
-      "num_tokens": 4281411.0,
+      "loss": 0.6360988020896912,
+      "mean_token_accuracy": 0.7437145560979843,
+      "num_tokens": 2057803.0,
       "step": 55
     },
     {
-      "entropy": 0.5654756128787994,
-      "epoch": 1.768,
-      "grad_norm": 0.032251086086034775,
+      "entropy": 0.6247914582490921,
+      "epoch": 0.36393176279447603,
+      "grad_norm": 0.014322548173367977,
       "learning_rate": 0.0002,
-      "loss": 0.5603655576705933,
-      "mean_token_accuracy": 0.7777776569128036,
-      "num_tokens": 4361042.0,
+      "loss": 0.6240888833999634,
+      "mean_token_accuracy": 0.7484056130051613,
+      "num_tokens": 2094784.0,
       "step": 56
     },
     {
-      "entropy": 0.5811332762241364,
-      "epoch": 1.8,
-      "grad_norm": 0.0319007970392704,
+      "entropy": 0.6302464827895164,
+      "epoch": 0.3704305442729488,
+      "grad_norm": 0.011746145784854889,
       "learning_rate": 0.0002,
-      "loss": 0.576663613319397,
-      "mean_token_accuracy": 0.77076855301857,
-      "num_tokens": 4438720.0,
+      "loss": 0.633884608745575,
+      "mean_token_accuracy": 0.7437854781746864,
+      "num_tokens": 2132466.0,
       "step": 57
     },
     {
-      "entropy": 0.5682029575109482,
-      "epoch": 1.8319999999999999,
-      "grad_norm": 0.026807347312569618,
+      "entropy": 0.6274847015738487,
+      "epoch": 0.37692932575142163,
+      "grad_norm": 0.013853202573955059,
       "learning_rate": 0.0002,
-      "loss": 0.564788818359375,
-      "mean_token_accuracy": 0.7736650556325912,
-      "num_tokens": 4515989.0,
+      "loss": 0.6355714797973633,
+      "mean_token_accuracy": 0.7430111691355705,
+      "num_tokens": 2169906.0,
       "step": 58
     },
     {
-      "entropy": 0.5706409513950348,
-      "epoch": 1.8639999999999999,
-      "grad_norm": 0.02681436389684677,
+      "entropy": 0.6265148296952248,
+      "epoch": 0.3834281072298944,
+      "grad_norm": 0.01277268398553133,
       "learning_rate": 0.0002,
-      "loss": 0.5666987895965576,
-      "mean_token_accuracy": 0.7743194103240967,
-      "num_tokens": 4594420.0,
+      "loss": 0.6335763335227966,
+      "mean_token_accuracy": 0.7446934059262276,
+      "num_tokens": 2207365.0,
       "step": 59
     },
     {
-      "entropy": 0.5729261785745621,
-      "epoch": 1.896,
-      "grad_norm": 0.030915573239326477,
+      "entropy": 0.6214065626263618,
+      "epoch": 0.38992688870836717,
+      "grad_norm": 0.010742721147835255,
       "learning_rate": 0.0002,
-      "loss": 0.5717182159423828,
-      "mean_token_accuracy": 0.7706410437822342,
-      "num_tokens": 4671835.0,
+      "loss": 0.6200428009033203,
+      "mean_token_accuracy": 0.748136468231678,
+      "num_tokens": 2244334.0,
       "step": 60
     },
     {
-      "entropy": 0.5693814009428024,
-      "epoch": 1.928,
-      "grad_norm": 0.022914232686161995,
+      "entropy": 0.6408857852220535,
+      "epoch": 0.39642567018683994,
+      "grad_norm": 0.011468583717942238,
       "learning_rate": 0.0002,
-      "loss": 0.5684123039245605,
-      "mean_token_accuracy": 0.7722501307725906,
-      "num_tokens": 4749804.0,
+      "loss": 0.6358450651168823,
+      "mean_token_accuracy": 0.7437937930226326,
+      "num_tokens": 2281670.0,
       "step": 61
     },
     {
-      "entropy": 0.5578709542751312,
-      "epoch": 1.96,
-      "grad_norm": 0.02405908890068531,
+      "entropy": 0.6396703198552132,
+      "epoch": 0.40292445166531277,
+      "grad_norm": 0.011673662811517715,
       "learning_rate": 0.0002,
-      "loss": 0.5572571158409119,
-      "mean_token_accuracy": 0.7781909555196762,
-      "num_tokens": 4828428.0,
+      "loss": 0.6363998651504517,
+      "mean_token_accuracy": 0.7431989163160324,
+      "num_tokens": 2318924.0,
       "step": 62
     },
     {
-      "entropy": 0.5615971833467484,
-      "epoch": 1.992,
-      "grad_norm": 0.019845224916934967,
+      "entropy": 0.6249052062630653,
+      "epoch": 0.40942323314378554,
+      "grad_norm": 0.0118953175842762,
       "learning_rate": 0.0002,
-      "loss": 0.5606215000152588,
-      "mean_token_accuracy": 0.7773871868848801,
-      "num_tokens": 4907477.0,
+      "loss": 0.6213114261627197,
+      "mean_token_accuracy": 0.7500982582569122,
+      "num_tokens": 2356317.0,
       "step": 63
     },
     {
-      "entropy": 0.5601226091384888,
-      "epoch": 2.0,
-      "grad_norm": 0.03756139427423477,
+      "entropy": 0.6271346360445023,
+      "epoch": 0.4159220146222583,
+      "grad_norm": 0.012425282970070839,
       "learning_rate": 0.0002,
-      "loss": 0.5584151744842529,
-      "mean_token_accuracy": 0.7765219211578369,
-      "num_tokens": 4927777.0,
+      "loss": 0.6213847398757935,
+      "mean_token_accuracy": 0.7495132237672806,
+      "num_tokens": 2393601.0,
       "step": 64
     },
     {
-      "entropy": 0.5674517601728439,
-      "epoch": 2.032,
-      "grad_norm": 0.025153273716568947,
+      "entropy": 0.629973016679287,
+      "epoch": 0.42242079610073113,
+      "grad_norm": 0.011852126568555832,
       "learning_rate": 0.0002,
-      "loss": 0.5689207315444946,
-      "mean_token_accuracy": 0.7721648663282394,
-      "num_tokens": 5005271.0,
+      "loss": 0.6328772306442261,
+      "mean_token_accuracy": 0.7433357834815979,
+      "num_tokens": 2430713.0,
       "step": 65
     },
     {
-      "entropy": 0.5618810504674911,
-      "epoch": 2.064,
-      "grad_norm": 0.02671828307211399,
+      "entropy": 0.625373125076294,
+      "epoch": 0.4289195775792039,
+      "grad_norm": 0.011082636192440987,
       "learning_rate": 0.0002,
-      "loss": 0.5602548122406006,
-      "mean_token_accuracy": 0.7760673761367798,
-      "num_tokens": 5084150.0,
+      "loss": 0.6278108954429626,
+      "mean_token_accuracy": 0.7462311089038849,
+      "num_tokens": 2468332.0,
       "step": 66
     },
     {
-      "entropy": 0.5582772195339203,
-      "epoch": 2.096,
-      "grad_norm": 0.021926281973719597,
+      "entropy": 0.6213006973266602,
+      "epoch": 0.4354183590576767,
+      "grad_norm": 0.012099278159439564,
       "learning_rate": 0.0002,
-      "loss": 0.5541115999221802,
-      "mean_token_accuracy": 0.7796459347009659,
-      "num_tokens": 5164221.0,
+      "loss": 0.6325806975364685,
+      "mean_token_accuracy": 0.748037800192833,
+      "num_tokens": 2505751.0,
       "step": 67
     },
     {
-      "entropy": 0.5672923773527145,
-      "epoch": 2.128,
-      "grad_norm": 0.024038124829530716,
+      "entropy": 0.6149301454424858,
+      "epoch": 0.44191714053614944,
+      "grad_norm": 0.011624482460319996,
       "learning_rate": 0.0002,
-      "loss": 0.5617194771766663,
-      "mean_token_accuracy": 0.7748870551586151,
-      "num_tokens": 5243027.0,
+      "loss": 0.6194391250610352,
+      "mean_token_accuracy": 0.7509394213557243,
+      "num_tokens": 2543314.0,
       "step": 68
     },
     {
-      "entropy": 0.5639302432537079,
-      "epoch": 2.16,
-      "grad_norm": 0.02486521564424038,
+      "entropy": 0.6199035122990608,
+      "epoch": 0.44841592201462227,
+      "grad_norm": 0.010423528961837292,
       "learning_rate": 0.0002,
-      "loss": 0.5582059025764465,
-      "mean_token_accuracy": 0.7770267874002457,
-      "num_tokens": 5322184.0,
+      "loss": 0.6224488615989685,
+      "mean_token_accuracy": 0.749529704451561,
+      "num_tokens": 2580776.0,
       "step": 69
     },
     {
-      "entropy": 0.5651180893182755,
-      "epoch": 2.192,
-      "grad_norm": 0.025982394814491272,
+      "entropy": 0.6317284181714058,
+      "epoch": 0.45491470349309504,
+      "grad_norm": 0.010865025222301483,
       "learning_rate": 0.0002,
-      "loss": 0.5584880113601685,
-      "mean_token_accuracy": 0.7765491008758545,
-      "num_tokens": 5399749.0,
+      "loss": 0.6270354986190796,
+      "mean_token_accuracy": 0.7461089566349983,
+      "num_tokens": 2618167.0,
       "step": 70
     },
     {
-      "entropy": 0.5606539249420166,
-      "epoch": 2.224,
-      "grad_norm": 0.024214768782258034,
+      "entropy": 0.6248530969023705,
+      "epoch": 0.4614134849715678,
+      "grad_norm": 0.010545512661337852,
       "learning_rate": 0.0002,
-      "loss": 0.5596858263015747,
-      "mean_token_accuracy": 0.7755280286073685,
-      "num_tokens": 5478994.0,
+      "loss": 0.6243047714233398,
+      "mean_token_accuracy": 0.7491967007517815,
+      "num_tokens": 2655748.0,
       "step": 71
     },
     {
-      "entropy": 0.5572450906038284,
-      "epoch": 2.2560000000000002,
-      "grad_norm": 0.024621082469820976,
+      "entropy": 0.6242939457297325,
+      "epoch": 0.46791226645004064,
+      "grad_norm": 0.01175450999289751,
       "learning_rate": 0.0002,
-      "loss": 0.5587051510810852,
-      "mean_token_accuracy": 0.7774072885513306,
-      "num_tokens": 5558167.0,
+      "loss": 0.6177983283996582,
+      "mean_token_accuracy": 0.7500801458954811,
+      "num_tokens": 2693338.0,
       "step": 72
     },
     {
-      "entropy": 0.5469658970832825,
-      "epoch": 2.288,
-      "grad_norm": 0.02560459077358246,
+      "entropy": 0.6458104699850082,
+      "epoch": 0.4744110479285134,
+      "grad_norm": 0.009623173624277115,
       "learning_rate": 0.0002,
-      "loss": 0.5509265661239624,
-      "mean_token_accuracy": 0.780618280172348,
-      "num_tokens": 5637131.0,
+      "loss": 0.6407941579818726,
+      "mean_token_accuracy": 0.7437473684549332,
+      "num_tokens": 2730673.0,
       "step": 73
     },
     {
-      "entropy": 0.5541262030601501,
-      "epoch": 2.32,
-      "grad_norm": 0.02317582443356514,
+      "entropy": 0.6289036273956299,
+      "epoch": 0.4809098294069862,
+      "grad_norm": 0.010831008665263653,
       "learning_rate": 0.0002,
-      "loss": 0.5573158264160156,
-      "mean_token_accuracy": 0.7776470184326172,
-      "num_tokens": 5714968.0,
+      "loss": 0.6292392015457153,
+      "mean_token_accuracy": 0.749647282063961,
+      "num_tokens": 2768101.0,
       "step": 74
     },
     {
-      "entropy": 0.5666017383337021,
-      "epoch": 2.352,
-      "grad_norm": 0.02768876403570175,
+      "entropy": 0.6329901814460754,
+      "epoch": 0.487408610885459,
+      "grad_norm": 0.013034787029027939,
       "learning_rate": 0.0002,
-      "loss": 0.5662693977355957,
-      "mean_token_accuracy": 0.7726470977067947,
-      "num_tokens": 5794248.0,
+      "loss": 0.6349586248397827,
+      "mean_token_accuracy": 0.7437526881694794,
+      "num_tokens": 2805877.0,
       "step": 75
     },
     {
-      "entropy": 0.5658593773841858,
-      "epoch": 2.384,
-      "grad_norm": 0.022219819948077202,
+      "entropy": 0.6201201230287552,
+      "epoch": 0.4939073923639318,
+      "grad_norm": 0.010362153872847557,
       "learning_rate": 0.0002,
-      "loss": 0.5574473142623901,
-      "mean_token_accuracy": 0.7764318138360977,
-      "num_tokens": 5873673.0,
+      "loss": 0.6222230195999146,
+      "mean_token_accuracy": 0.749205470085144,
+      "num_tokens": 2843002.0,
       "step": 76
     },
     {
-      "entropy": 0.5633502453565598,
-      "epoch": 2.416,
-      "grad_norm": 0.02417032979428768,
+      "entropy": 0.6270335763692856,
+      "epoch": 0.5004061738424046,
+      "grad_norm": 0.01038055308163166,
       "learning_rate": 0.0002,
-      "loss": 0.5558116436004639,
-      "mean_token_accuracy": 0.777042031288147,
-      "num_tokens": 5953477.0,
+      "loss": 0.6268460154533386,
+      "mean_token_accuracy": 0.7455538734793663,
+      "num_tokens": 2880433.0,
       "step": 77
     },
     {
-      "entropy": 0.5666830539703369,
-      "epoch": 2.448,
-      "grad_norm": 0.025465968996286392,
+      "entropy": 0.6266148760914803,
+      "epoch": 0.5069049553208773,
+      "grad_norm": 0.011833865195512772,
       "learning_rate": 0.0002,
-      "loss": 0.5631768703460693,
-      "mean_token_accuracy": 0.775611087679863,
-      "num_tokens": 6031642.0,
+      "loss": 0.6316983699798584,
+      "mean_token_accuracy": 0.7472645714879036,
+      "num_tokens": 2917833.0,
       "step": 78
     },
     {
-      "entropy": 0.5565501898527145,
-      "epoch": 2.48,
-      "grad_norm": 0.023061588406562805,
+      "entropy": 0.6177176311612129,
+      "epoch": 0.5134037367993501,
+      "grad_norm": 0.012312863022089005,
       "learning_rate": 0.0002,
-      "loss": 0.5533760190010071,
-      "mean_token_accuracy": 0.7793847471475601,
-      "num_tokens": 6111567.0,
+      "loss": 0.620547890663147,
+      "mean_token_accuracy": 0.7489045560359955,
+      "num_tokens": 2955261.0,
       "step": 79
     },
     {
-      "entropy": 0.5474292933940887,
-      "epoch": 2.512,
-      "grad_norm": 0.02363215573132038,
+      "entropy": 0.6255826130509377,
+      "epoch": 0.5199025182778229,
+      "grad_norm": 0.011764319613575935,
       "learning_rate": 0.0002,
-      "loss": 0.5478873252868652,
-      "mean_token_accuracy": 0.7802330702543259,
-      "num_tokens": 6192161.0,
+      "loss": 0.6278033256530762,
+      "mean_token_accuracy": 0.7475135996937752,
+      "num_tokens": 2992205.0,
       "step": 80
     },
     {
-      "entropy": 0.5629004091024399,
-      "epoch": 2.544,
-      "grad_norm": 0.02275676839053631,
+      "entropy": 0.6303588896989822,
+      "epoch": 0.5264012997562957,
+      "grad_norm": 0.01056460291147232,
       "learning_rate": 0.0002,
-      "loss": 0.5638875961303711,
-      "mean_token_accuracy": 0.7731000036001205,
-      "num_tokens": 6271115.0,
+      "loss": 0.6343796253204346,
+      "mean_token_accuracy": 0.7422873452305794,
+      "num_tokens": 3029696.0,
       "step": 81
     },
     {
-      "entropy": 0.5549586117267609,
-      "epoch": 2.576,
-      "grad_norm": 0.02356044203042984,
+      "entropy": 0.6168110445141792,
+      "epoch": 0.5329000812347685,
+      "grad_norm": 0.010063163004815578,
       "learning_rate": 0.0002,
-      "loss": 0.5570619106292725,
-      "mean_token_accuracy": 0.7774660885334015,
-      "num_tokens": 6349371.0,
+      "loss": 0.6208071112632751,
+      "mean_token_accuracy": 0.7490193694829941,
+      "num_tokens": 3067080.0,
       "step": 82
     },
     {
-      "entropy": 0.5476335734128952,
-      "epoch": 2.608,
-      "grad_norm": 0.027859514579176903,
+      "entropy": 0.6252159550786018,
+      "epoch": 0.5393988627132412,
+      "grad_norm": 0.010448803193867207,
       "learning_rate": 0.0002,
-      "loss": 0.5477542877197266,
-      "mean_token_accuracy": 0.7811516225337982,
-      "num_tokens": 6427681.0,
+      "loss": 0.6253081560134888,
+      "mean_token_accuracy": 0.7478374615311623,
+      "num_tokens": 3104286.0,
       "step": 83
     },
     {
-      "entropy": 0.5653144121170044,
-      "epoch": 2.64,
-      "grad_norm": 0.022368136793375015,
+      "entropy": 0.6273334473371506,
+      "epoch": 0.545897644191714,
+      "grad_norm": 0.010500379838049412,
       "learning_rate": 0.0002,
-      "loss": 0.5637708902359009,
-      "mean_token_accuracy": 0.7758270651102066,
-      "num_tokens": 6506470.0,
+      "loss": 0.6298959255218506,
+      "mean_token_accuracy": 0.7449745014309883,
+      "num_tokens": 3141685.0,
       "step": 84
     },
     {
-      "entropy": 0.5534553080797195,
-      "epoch": 2.672,
-      "grad_norm": 0.022241799160838127,
+      "entropy": 0.6271170675754547,
+      "epoch": 0.5523964256701869,
+      "grad_norm": 0.011779806576669216,
       "learning_rate": 0.0002,
-      "loss": 0.5493363738059998,
-      "mean_token_accuracy": 0.7799386531114578,
-      "num_tokens": 6585286.0,
+      "loss": 0.6241638660430908,
+      "mean_token_accuracy": 0.7471632659435272,
+      "num_tokens": 3178925.0,
       "step": 85
     },
     {
-      "entropy": 0.567073181271553,
-      "epoch": 2.7039999999999997,
-      "grad_norm": 0.02547631412744522,
+      "entropy": 0.6358036622405052,
+      "epoch": 0.5588952071486596,
+      "grad_norm": 0.010791173204779625,
       "learning_rate": 0.0002,
-      "loss": 0.563334584236145,
-      "mean_token_accuracy": 0.7755401879549026,
-      "num_tokens": 6664430.0,
+      "loss": 0.6331329941749573,
+      "mean_token_accuracy": 0.7441665381193161,
+      "num_tokens": 3216076.0,
       "step": 86
     },
     {
-      "entropy": 0.5621544122695923,
-      "epoch": 2.7359999999999998,
-      "grad_norm": 0.02449527569115162,
+      "entropy": 0.6320113763213158,
+      "epoch": 0.5653939886271324,
+      "grad_norm": 0.009891759604215622,
       "learning_rate": 0.0002,
-      "loss": 0.5599551200866699,
-      "mean_token_accuracy": 0.7742103487253189,
-      "num_tokens": 6743061.0,
+      "loss": 0.6326582431793213,
+      "mean_token_accuracy": 0.7445697784423828,
+      "num_tokens": 3253700.0,
       "step": 87
     },
     {
-      "entropy": 0.5637608170509338,
-      "epoch": 2.768,
-      "grad_norm": 0.020537901669740677,
+      "entropy": 0.6250990331172943,
+      "epoch": 0.5718927701056052,
+      "grad_norm": 0.013541702181100845,
       "learning_rate": 0.0002,
-      "loss": 0.5627949237823486,
-      "mean_token_accuracy": 0.7747203409671783,
-      "num_tokens": 6821751.0,
+      "loss": 0.6289629936218262,
+      "mean_token_accuracy": 0.745047003030777,
+      "num_tokens": 3291250.0,
       "step": 88
     },
     {
-      "entropy": 0.5689736902713776,
-      "epoch": 2.8,
-      "grad_norm": 0.02112015336751938,
+      "entropy": 0.6287946552038193,
+      "epoch": 0.578391551584078,
+      "grad_norm": 0.009651604108512402,
       "learning_rate": 0.0002,
-      "loss": 0.565583348274231,
-      "mean_token_accuracy": 0.7746206969022751,
-      "num_tokens": 6901279.0,
+      "loss": 0.6300573348999023,
+      "mean_token_accuracy": 0.7451979964971542,
+      "num_tokens": 3328548.0,
       "step": 89
     },
     {
-      "entropy": 0.5660188943147659,
-      "epoch": 2.832,
-      "grad_norm": 0.022382516413927078,
+      "entropy": 0.6240335181355476,
+      "epoch": 0.5848903330625508,
+      "grad_norm": 0.009283720515668392,
       "learning_rate": 0.0002,
-      "loss": 0.5647023916244507,
-      "mean_token_accuracy": 0.773158609867096,
-      "num_tokens": 6981047.0,
+      "loss": 0.6263805031776428,
+      "mean_token_accuracy": 0.7456449046730995,
+      "num_tokens": 3365327.0,
       "step": 90
     },
     {
-      "entropy": 0.5572714656591415,
-      "epoch": 2.864,
-      "grad_norm": 0.022794874384999275,
+      "entropy": 0.6241980046033859,
+      "epoch": 0.5913891145410236,
+      "grad_norm": 0.010714948177337646,
       "learning_rate": 0.0002,
-      "loss": 0.5548834204673767,
-      "mean_token_accuracy": 0.7764633297920227,
-      "num_tokens": 7058727.0,
+      "loss": 0.625645637512207,
+      "mean_token_accuracy": 0.7471252828836441,
+      "num_tokens": 3402757.0,
       "step": 91
     },
     {
-      "entropy": 0.5647054761648178,
-      "epoch": 2.896,
-      "grad_norm": 0.022615451365709305,
+      "entropy": 0.628009632229805,
+      "epoch": 0.5978878960194963,
+      "grad_norm": 0.012750590220093727,
       "learning_rate": 0.0002,
-      "loss": 0.5656628608703613,
-      "mean_token_accuracy": 0.7732657492160797,
-      "num_tokens": 7135849.0,
+      "loss": 0.6261017322540283,
+      "mean_token_accuracy": 0.7465066090226173,
+      "num_tokens": 3440031.0,
       "step": 92
     },
     {
-      "entropy": 0.5565017014741898,
-      "epoch": 2.928,
-      "grad_norm": 0.023763637989759445,
+      "entropy": 0.6196997240185738,
+      "epoch": 0.6043866774979691,
+      "grad_norm": 0.011183288879692554,
       "learning_rate": 0.0002,
-      "loss": 0.5567818284034729,
-      "mean_token_accuracy": 0.7769446074962616,
-      "num_tokens": 7214895.0,
+      "loss": 0.6222946047782898,
+      "mean_token_accuracy": 0.7507361546158791,
+      "num_tokens": 3477362.0,
       "step": 93
     },
     {
-      "entropy": 0.560202345252037,
-      "epoch": 2.96,
-      "grad_norm": 0.023177897557616234,
+      "entropy": 0.6370684951543808,
+      "epoch": 0.6108854589764419,
+      "grad_norm": 0.010141120292246342,
       "learning_rate": 0.0002,
-      "loss": 0.5593446493148804,
-      "mean_token_accuracy": 0.774772509932518,
-      "num_tokens": 7294431.0,
+      "loss": 0.638361930847168,
+      "mean_token_accuracy": 0.7412139847874641,
+      "num_tokens": 3514709.0,
       "step": 94
     },
     {
-      "entropy": 0.5637674629688263,
-      "epoch": 2.992,
-      "grad_norm": 0.025069424882531166,
+      "entropy": 0.6254898384213448,
+      "epoch": 0.6173842404549147,
+      "grad_norm": 0.009656463749706745,
       "learning_rate": 0.0002,
-      "loss": 0.5618734359741211,
-      "mean_token_accuracy": 0.774286150932312,
-      "num_tokens": 7371930.0,
+      "loss": 0.6284251809120178,
+      "mean_token_accuracy": 0.7460697740316391,
+      "num_tokens": 3552355.0,
       "step": 95
     },
     {
-      "entropy": 0.5556119680404663,
-      "epoch": 3.0,
-      "grad_norm": 0.031051333993673325,
+      "entropy": 0.626992717385292,
+      "epoch": 0.6238830219333875,
+      "grad_norm": 0.01039013173431158,
       "learning_rate": 0.0002,
-      "loss": 0.550994873046875,
-      "mean_token_accuracy": 0.7802934050559998,
-      "num_tokens": 7391757.0,
+      "loss": 0.632388174533844,
+      "mean_token_accuracy": 0.7448238283395767,
+      "num_tokens": 3590160.0,
       "step": 96
     },
     {
-      "entropy": 0.5636318922042847,
-      "epoch": 3.032,
-      "grad_norm": 0.025959042832255363,
+      "entropy": 0.6212045699357986,
+      "epoch": 0.6303818034118602,
+      "grad_norm": 0.011750632897019386,
       "learning_rate": 0.0002,
-      "loss": 0.5586003065109253,
-      "mean_token_accuracy": 0.7768417745828629,
-      "num_tokens": 7470651.0,
+      "loss": 0.623359203338623,
+      "mean_token_accuracy": 0.7481106221675873,
+      "num_tokens": 3627240.0,
       "step": 97
     },
     {
-      "entropy": 0.5585259348154068,
-      "epoch": 3.064,
-      "grad_norm": 0.021750938147306442,
+      "entropy": 0.6192934885621071,
+      "epoch": 0.6368805848903331,
+      "grad_norm": 0.009845283813774586,
       "learning_rate": 0.0002,
-      "loss": 0.556551456451416,
-      "mean_token_accuracy": 0.7769591957330704,
-      "num_tokens": 7549852.0,
+      "loss": 0.6226494312286377,
+      "mean_token_accuracy": 0.7480833828449249,
+      "num_tokens": 3664279.0,
       "step": 98
     },
     {
-      "entropy": 0.5582039058208466,
-      "epoch": 3.096,
-      "grad_norm": 0.025489795953035355,
+      "entropy": 0.6310721859335899,
+      "epoch": 0.6433793663688059,
+      "grad_norm": 0.011091063730418682,
       "learning_rate": 0.0002,
-      "loss": 0.5555694103240967,
-      "mean_token_accuracy": 0.7770965248346329,
-      "num_tokens": 7629195.0,
+      "loss": 0.6305258274078369,
+      "mean_token_accuracy": 0.7463658377528191,
+      "num_tokens": 3701955.0,
       "step": 99
     },
     {
-      "entropy": 0.5573598444461823,
-      "epoch": 3.128,
-      "grad_norm": 0.023217501118779182,
+      "entropy": 0.6249227002263069,
+      "epoch": 0.6498781478472786,
+      "grad_norm": 0.009939444251358509,
       "learning_rate": 0.0002,
-      "loss": 0.5600582361221313,
-      "mean_token_accuracy": 0.7751252055168152,
-      "num_tokens": 7706266.0,
+      "loss": 0.6257964372634888,
+      "mean_token_accuracy": 0.7479037865996361,
+      "num_tokens": 3739298.0,
       "step": 100
     },
     {
-      "entropy": 0.553033709526062,
-      "epoch": 3.16,
-      "grad_norm": 0.02116326428949833,
+      "entropy": 0.6227108985185623,
+      "epoch": 0.6563769293257514,
+      "grad_norm": 0.010004580952227116,
       "learning_rate": 0.0002,
-      "loss": 0.5513657927513123,
-      "mean_token_accuracy": 0.7792606353759766,
-      "num_tokens": 7786170.0,
+      "loss": 0.6256003379821777,
+      "mean_token_accuracy": 0.7474573701620102,
+      "num_tokens": 3776806.0,
       "step": 101
     },
     {
-      "entropy": 0.5500922203063965,
-      "epoch": 3.192,
-      "grad_norm": 0.023273272439837456,
+      "entropy": 0.6166384890675545,
+      "epoch": 0.6628757108042242,
+      "grad_norm": 0.010987838730216026,
       "learning_rate": 0.0002,
-      "loss": 0.5481851696968079,
-      "mean_token_accuracy": 0.7797382026910782,
-      "num_tokens": 7866284.0,
+      "loss": 0.6182519197463989,
+      "mean_token_accuracy": 0.7499695047736168,
+      "num_tokens": 3813941.0,
       "step": 102
     },
     {
-      "entropy": 0.5571017563343048,
-      "epoch": 3.224,
-      "grad_norm": 0.024359384551644325,
+      "entropy": 0.6348938420414925,
+      "epoch": 0.669374492282697,
+      "grad_norm": 0.011195721104741096,
       "learning_rate": 0.0002,
-      "loss": 0.5538984537124634,
-      "mean_token_accuracy": 0.7775730490684509,
-      "num_tokens": 7944939.0,
+      "loss": 0.6304492950439453,
+      "mean_token_accuracy": 0.7462876811623573,
+      "num_tokens": 3851698.0,
       "step": 103
     },
     {
-      "entropy": 0.5599475651979446,
-      "epoch": 3.2560000000000002,
-      "grad_norm": 0.020094532519578934,
+      "entropy": 0.6282550320029259,
+      "epoch": 0.6758732737611698,
+      "grad_norm": 0.010595601983368397,
       "learning_rate": 0.0002,
-      "loss": 0.5542423725128174,
-      "mean_token_accuracy": 0.7767647057771683,
-      "num_tokens": 8024975.0,
+      "loss": 0.6273945569992065,
+      "mean_token_accuracy": 0.7473014816641808,
+      "num_tokens": 3888988.0,
       "step": 104
     },
     {
-      "entropy": 0.5703375190496445,
-      "epoch": 3.288,
-      "grad_norm": 0.020639436319470406,
+      "entropy": 0.6324817091226578,
+      "epoch": 0.6823720552396426,
+      "grad_norm": 0.009895245544612408,
       "learning_rate": 0.0002,
-      "loss": 0.564155638217926,
-      "mean_token_accuracy": 0.773441419005394,
-      "num_tokens": 8103041.0,
+      "loss": 0.632161557674408,
+      "mean_token_accuracy": 0.7444497495889664,
+      "num_tokens": 3925974.0,
       "step": 105
     },
     {
-      "entropy": 0.5621205866336823,
-      "epoch": 3.32,
-      "grad_norm": 0.020889636129140854,
+      "entropy": 0.6199614554643631,
+      "epoch": 0.6888708367181153,
+      "grad_norm": 0.010299347341060638,
       "learning_rate": 0.0002,
-      "loss": 0.5588844418525696,
-      "mean_token_accuracy": 0.7760203182697296,
-      "num_tokens": 8181136.0,
+      "loss": 0.6227681636810303,
+      "mean_token_accuracy": 0.746825821697712,
+      "num_tokens": 3962869.0,
       "step": 106
     },
     {
-      "entropy": 0.55582495033741,
-      "epoch": 3.352,
-      "grad_norm": 0.019500412046909332,
+      "entropy": 0.6321764960885048,
+      "epoch": 0.6953696181965882,
+      "grad_norm": 0.010949746705591679,
       "learning_rate": 0.0002,
-      "loss": 0.5550122261047363,
-      "mean_token_accuracy": 0.7787394225597382,
-      "num_tokens": 8260405.0,
+      "loss": 0.6333647966384888,
+      "mean_token_accuracy": 0.7444829195737839,
+      "num_tokens": 4000443.0,
       "step": 107
     },
     {
-      "entropy": 0.5650199949741364,
-      "epoch": 3.384,
-      "grad_norm": 0.02145702950656414,
+      "entropy": 0.6052318215370178,
+      "epoch": 0.7018683996750609,
+      "grad_norm": 0.010511154308915138,
       "learning_rate": 0.0002,
-      "loss": 0.5680341720581055,
-      "mean_token_accuracy": 0.7718839198350906,
-      "num_tokens": 8337560.0,
+      "loss": 0.6079590916633606,
+      "mean_token_accuracy": 0.7547851428389549,
+      "num_tokens": 4037560.0,
       "step": 108
     },
     {
-      "entropy": 0.5433939546346664,
-      "epoch": 3.416,
-      "grad_norm": 0.023313792422413826,
+      "entropy": 0.6213503628969193,
+      "epoch": 0.7083671811535337,
+      "grad_norm": 0.009959033690392971,
       "learning_rate": 0.0002,
-      "loss": 0.5474338531494141,
-      "mean_token_accuracy": 0.7793407142162323,
-      "num_tokens": 8416130.0,
+      "loss": 0.622962236404419,
+      "mean_token_accuracy": 0.7485458552837372,
+      "num_tokens": 4075139.0,
       "step": 109
     },
     {
-      "entropy": 0.5506273955106735,
-      "epoch": 3.448,
-      "grad_norm": 0.021077029407024384,
+      "entropy": 0.6228557601571083,
+      "epoch": 0.7148659626320065,
+      "grad_norm": 0.009467214345932007,
       "learning_rate": 0.0002,
-      "loss": 0.5539349317550659,
-      "mean_token_accuracy": 0.777793824672699,
-      "num_tokens": 8494571.0,
+      "loss": 0.6287202835083008,
+      "mean_token_accuracy": 0.7456526011228561,
+      "num_tokens": 4112628.0,
       "step": 110
     },
     {
-      "entropy": 0.5581430941820145,
-      "epoch": 3.48,
-      "grad_norm": 0.022196298465132713,
+      "entropy": 0.6176012232899666,
+      "epoch": 0.7213647441104792,
+      "grad_norm": 0.009703468531370163,
       "learning_rate": 0.0002,
-      "loss": 0.553594708442688,
-      "mean_token_accuracy": 0.7770230323076248,
-      "num_tokens": 8572656.0,
+      "loss": 0.6198960542678833,
+      "mean_token_accuracy": 0.7511189430952072,
+      "num_tokens": 4150384.0,
       "step": 111
     },
     {
-      "entropy": 0.567478135228157,
-      "epoch": 3.512,
-      "grad_norm": 0.02296704798936844,
+      "entropy": 0.6364283263683319,
+      "epoch": 0.7278635255889521,
+      "grad_norm": 0.009932373650372028,
       "learning_rate": 0.0002,
-      "loss": 0.5598067045211792,
-      "mean_token_accuracy": 0.7746621519327164,
-      "num_tokens": 8650724.0,
+      "loss": 0.6412342190742493,
+      "mean_token_accuracy": 0.742191307246685,
+      "num_tokens": 4187483.0,
       "step": 112
     },
     {
-      "entropy": 0.5627018958330154,
-      "epoch": 3.544,
-      "grad_norm": 0.020507240667939186,
+      "entropy": 0.6340491250157356,
+      "epoch": 0.7343623070674249,
+      "grad_norm": 0.009394255466759205,
       "learning_rate": 0.0002,
-      "loss": 0.5585383176803589,
-      "mean_token_accuracy": 0.7763334959745407,
-      "num_tokens": 8729104.0,
+      "loss": 0.632526695728302,
+      "mean_token_accuracy": 0.7457142248749733,
+      "num_tokens": 4224861.0,
       "step": 113
     },
     {
-      "entropy": 0.5544057041406631,
-      "epoch": 3.576,
-      "grad_norm": 0.02342178113758564,
+      "entropy": 0.626609705388546,
+      "epoch": 0.7408610885458976,
+      "grad_norm": 0.010729662142693996,
       "learning_rate": 0.0002,
-      "loss": 0.5570175647735596,
-      "mean_token_accuracy": 0.7762107402086258,
-      "num_tokens": 8807747.0,
+      "loss": 0.6251577138900757,
+      "mean_token_accuracy": 0.7473985999822617,
+      "num_tokens": 4261958.0,
       "step": 114
     },
     {
-      "entropy": 0.5425068289041519,
-      "epoch": 3.608,
-      "grad_norm": 0.02285587601363659,
+      "entropy": 0.6041073203086853,
+      "epoch": 0.7473598700243704,
+      "grad_norm": 0.010525292716920376,
       "learning_rate": 0.0002,
-      "loss": 0.545325756072998,
-      "mean_token_accuracy": 0.7807946652173996,
-      "num_tokens": 8887832.0,
+      "loss": 0.599025309085846,
+      "mean_token_accuracy": 0.7583419680595398,
+      "num_tokens": 4299462.0,
       "step": 115
     },
     {
-      "entropy": 0.553840234875679,
-      "epoch": 3.64,
-      "grad_norm": 0.022142302244901657,
+      "entropy": 0.6210100278258324,
+      "epoch": 0.7538586515028433,
+      "grad_norm": 0.009981499053537846,
       "learning_rate": 0.0002,
-      "loss": 0.5548170804977417,
-      "mean_token_accuracy": 0.7777025252580643,
-      "num_tokens": 8966925.0,
+      "loss": 0.6192715167999268,
+      "mean_token_accuracy": 0.7496819868683815,
+      "num_tokens": 4336903.0,
       "step": 116
     },
     {
-      "entropy": 0.5666903406381607,
-      "epoch": 3.672,
-      "grad_norm": 0.020184515044093132,
+      "entropy": 0.6161150112748146,
+      "epoch": 0.760357432981316,
+      "grad_norm": 0.009832478128373623,
       "learning_rate": 0.0002,
-      "loss": 0.5636236667633057,
-      "mean_token_accuracy": 0.7735326737165451,
-      "num_tokens": 9045663.0,
+      "loss": 0.6186830401420593,
+      "mean_token_accuracy": 0.7503286600112915,
+      "num_tokens": 4374156.0,
       "step": 117
     },
     {
-      "entropy": 0.5551957935094833,
-      "epoch": 3.7039999999999997,
-      "grad_norm": 0.0219353549182415,
+      "entropy": 0.6233196780085564,
+      "epoch": 0.7668562144597888,
+      "grad_norm": 0.009722237475216389,
       "learning_rate": 0.0002,
-      "loss": 0.5482048988342285,
-      "mean_token_accuracy": 0.7802202552556992,
-      "num_tokens": 9124298.0,
+      "loss": 0.6274971961975098,
+      "mean_token_accuracy": 0.7461902946233749,
+      "num_tokens": 4411885.0,
       "step": 118
     },
     {
-      "entropy": 0.5691462904214859,
-      "epoch": 3.7359999999999998,
-      "grad_norm": 0.023641400039196014,
+      "entropy": 0.6291887611150742,
+      "epoch": 0.7733549959382616,
+      "grad_norm": 0.01063536573201418,
       "learning_rate": 0.0002,
-      "loss": 0.5663072466850281,
-      "mean_token_accuracy": 0.7717752456665039,
-      "num_tokens": 9201921.0,
+      "loss": 0.6366918087005615,
+      "mean_token_accuracy": 0.7416415438055992,
+      "num_tokens": 4449209.0,
       "step": 119
     },
     {
-      "entropy": 0.564937487244606,
-      "epoch": 3.768,
-      "grad_norm": 0.02424025908112526,
+      "entropy": 0.6099276095628738,
+      "epoch": 0.7798537774167343,
+      "grad_norm": 0.010151992551982403,
       "learning_rate": 0.0002,
-      "loss": 0.5656293630599976,
-      "mean_token_accuracy": 0.7734099626541138,
-      "num_tokens": 9279520.0,
+      "loss": 0.6134237051010132,
+      "mean_token_accuracy": 0.7496887296438217,
+      "num_tokens": 4486909.0,
       "step": 120
     },
     {
-      "entropy": 0.543596014380455,
-      "epoch": 3.8,
-      "grad_norm": 0.02149128168821335,
+      "entropy": 0.6280336380004883,
+      "epoch": 0.7863525588952072,
+      "grad_norm": 0.011214104481041431,
       "learning_rate": 0.0002,
-      "loss": 0.5408511757850647,
-      "mean_token_accuracy": 0.7823390662670135,
-      "num_tokens": 9358194.0,
+      "loss": 0.6303111910820007,
+      "mean_token_accuracy": 0.743942454457283,
+      "num_tokens": 4524269.0,
       "step": 121
     },
     {
-      "entropy": 0.5443567931652069,
-      "epoch": 3.832,
-      "grad_norm": 0.02457662858068943,
+      "entropy": 0.624128632247448,
+      "epoch": 0.7928513403736799,
+      "grad_norm": 0.01028621755540371,
       "learning_rate": 0.0002,
-      "loss": 0.5437170267105103,
-      "mean_token_accuracy": 0.7806205898523331,
-      "num_tokens": 9436813.0,
+      "loss": 0.6248451471328735,
+      "mean_token_accuracy": 0.7479912042617798,
+      "num_tokens": 4561643.0,
       "step": 122
     },
     {
-      "entropy": 0.5446386188268661,
-      "epoch": 3.864,
-      "grad_norm": 0.023279579356312752,
+      "entropy": 0.6252302676439285,
+      "epoch": 0.7993501218521527,
+      "grad_norm": 0.01105236541479826,
       "learning_rate": 0.0002,
-      "loss": 0.5444090366363525,
-      "mean_token_accuracy": 0.7807519435882568,
-      "num_tokens": 9516996.0,
+      "loss": 0.6249452233314514,
+      "mean_token_accuracy": 0.7472440227866173,
+      "num_tokens": 4599086.0,
       "step": 123
     },
     {
-      "entropy": 0.5451237112283707,
-      "epoch": 3.896,
-      "grad_norm": 0.021046897396445274,
+      "entropy": 0.6296864897012711,
+      "epoch": 0.8058489033306255,
+      "grad_norm": 0.009404209442436695,
       "learning_rate": 0.0002,
-      "loss": 0.5447065830230713,
-      "mean_token_accuracy": 0.7818619459867477,
-      "num_tokens": 9597382.0,
+      "loss": 0.6316344738006592,
+      "mean_token_accuracy": 0.7425857782363892,
+      "num_tokens": 4636471.0,
       "step": 124
     },
     {
-      "entropy": 0.557760015130043,
-      "epoch": 3.928,
-      "grad_norm": 0.02131776139140129,
+      "entropy": 0.6118782311677933,
+      "epoch": 0.8123476848090982,
+      "grad_norm": 0.010075677186250687,
       "learning_rate": 0.0002,
-      "loss": 0.5580514073371887,
-      "mean_token_accuracy": 0.77658312022686,
-      "num_tokens": 9677720.0,
+      "loss": 0.6151276230812073,
+      "mean_token_accuracy": 0.7537136599421501,
+      "num_tokens": 4673814.0,
       "step": 125
     },
     {
-      "entropy": 0.544140562415123,
-      "epoch": 3.96,
-      "grad_norm": 0.020328767597675323,
+      "entropy": 0.6220783665776253,
+      "epoch": 0.8188464662875711,
+      "grad_norm": 0.01058894768357277,
       "learning_rate": 0.0002,
-      "loss": 0.5451820492744446,
-      "mean_token_accuracy": 0.7804036885499954,
-      "num_tokens": 9757131.0,
+      "loss": 0.623919665813446,
+      "mean_token_accuracy": 0.7449487373232841,
+      "num_tokens": 4711348.0,
       "step": 126
     },
     {
-      "entropy": 0.5636927336454391,
-      "epoch": 3.992,
-      "grad_norm": 0.021143252030014992,
+      "entropy": 0.6320341899991035,
+      "epoch": 0.8253452477660439,
+      "grad_norm": 0.009438078850507736,
       "learning_rate": 0.0002,
-      "loss": 0.5571653246879578,
-      "mean_token_accuracy": 0.7772841453552246,
-      "num_tokens": 9835244.0,
+      "loss": 0.635218620300293,
+      "mean_token_accuracy": 0.7428243607282639,
+      "num_tokens": 4748964.0,
       "step": 127
     },
     {
-      "entropy": 0.5422744750976562,
-      "epoch": 4.0,
-      "grad_norm": 0.0327562615275383,
+      "entropy": 0.608184851706028,
+      "epoch": 0.8318440292445166,
+      "grad_norm": 0.00950313825160265,
       "learning_rate": 0.0002,
-      "loss": 0.5362396240234375,
-      "mean_token_accuracy": 0.7841192483901978,
-      "num_tokens": 9855434.0,
+      "loss": 0.608941376209259,
+      "mean_token_accuracy": 0.7540178000926971,
+      "num_tokens": 4786015.0,
       "step": 128
     },
     {
-      "entropy": 0.5350159257650375,
-      "epoch": 4.032,
-      "grad_norm": 0.021142639219760895,
+      "entropy": 0.6191292628645897,
+      "epoch": 0.8383428107229894,
+      "grad_norm": 0.00933976098895073,
       "learning_rate": 0.0002,
-      "loss": 0.5306050181388855,
-      "mean_token_accuracy": 0.787266880273819,
-      "num_tokens": 9935799.0,
+      "loss": 0.6211662292480469,
+      "mean_token_accuracy": 0.7470189183950424,
+      "num_tokens": 4823477.0,
       "step": 129
     },
     {
-      "entropy": 0.5508315712213516,
-      "epoch": 4.064,
-      "grad_norm": 0.023039840161800385,
+      "entropy": 0.6169113591313362,
+      "epoch": 0.8448415922014623,
+      "grad_norm": 0.00967735517770052,
       "learning_rate": 0.0002,
-      "loss": 0.5510164499282837,
-      "mean_token_accuracy": 0.7783015668392181,
-      "num_tokens": 10015337.0,
+      "loss": 0.6221678256988525,
+      "mean_token_accuracy": 0.7482811883091927,
+      "num_tokens": 4860492.0,
       "step": 130
     },
     {
-      "entropy": 0.5506930351257324,
-      "epoch": 4.096,
-      "grad_norm": 0.022771824151277542,
+      "entropy": 0.6181379705667496,
+      "epoch": 0.851340373679935,
+      "grad_norm": 0.010275167413055897,
       "learning_rate": 0.0002,
-      "loss": 0.5542877912521362,
-      "mean_token_accuracy": 0.7773886919021606,
-      "num_tokens": 10092789.0,
+      "loss": 0.6256829500198364,
+      "mean_token_accuracy": 0.7461275234818459,
+      "num_tokens": 4897310.0,
       "step": 131
     },
     {
-      "entropy": 0.5481761693954468,
-      "epoch": 4.128,
-      "grad_norm": 0.020041558891534805,
+      "entropy": 0.6289433836936951,
+      "epoch": 0.8578391551584078,
+      "grad_norm": 0.010724073275923729,
       "learning_rate": 0.0002,
-      "loss": 0.5482645630836487,
-      "mean_token_accuracy": 0.7794580906629562,
-      "num_tokens": 10171363.0,
+      "loss": 0.6285852193832397,
+      "mean_token_accuracy": 0.7468666434288025,
+      "num_tokens": 4935222.0,
       "step": 132
     },
     {
-      "entropy": 0.5571236908435822,
-      "epoch": 4.16,
-      "grad_norm": 0.02578168548643589,
+      "entropy": 0.6300692111253738,
+      "epoch": 0.8643379366368806,
+      "grad_norm": 0.009074898436665535,
       "learning_rate": 0.0002,
-      "loss": 0.55533766746521,
-      "mean_token_accuracy": 0.7765619903802872,
-      "num_tokens": 10250091.0,
+      "loss": 0.6281343698501587,
+      "mean_token_accuracy": 0.7458086088299751,
+      "num_tokens": 4972549.0,
       "step": 133
     },
     {
-      "entropy": 0.5592793822288513,
-      "epoch": 4.192,
-      "grad_norm": 0.02451624907553196,
+      "entropy": 0.6445952579379082,
+      "epoch": 0.8708367181153533,
+      "grad_norm": 0.009369590319693089,
       "learning_rate": 0.0002,
-      "loss": 0.5530962944030762,
-      "mean_token_accuracy": 0.7770255208015442,
-      "num_tokens": 10328422.0,
+      "loss": 0.6381428837776184,
+      "mean_token_accuracy": 0.7420760691165924,
+      "num_tokens": 5010593.0,
       "step": 134
     },
     {
-      "entropy": 0.563023567199707,
-      "epoch": 4.224,
-      "grad_norm": 0.019786424934864044,
+      "entropy": 0.6136679574847221,
+      "epoch": 0.8773354995938262,
+      "grad_norm": 0.01232313271611929,
       "learning_rate": 0.0002,
-      "loss": 0.5598846673965454,
-      "mean_token_accuracy": 0.7750790417194366,
-      "num_tokens": 10407828.0,
+      "loss": 0.6184711456298828,
+      "mean_token_accuracy": 0.7486978471279144,
+      "num_tokens": 5047468.0,
       "step": 135
     },
     {
-      "entropy": 0.5667447596788406,
-      "epoch": 4.256,
-      "grad_norm": 0.02562200278043747,
+      "entropy": 0.6270027384161949,
+      "epoch": 0.8838342810722989,
+      "grad_norm": 0.010591822676360607,
       "learning_rate": 0.0002,
-      "loss": 0.5655050277709961,
-      "mean_token_accuracy": 0.7730535417795181,
-      "num_tokens": 10486389.0,
+      "loss": 0.6306672692298889,
+      "mean_token_accuracy": 0.7446641474962234,
+      "num_tokens": 5084586.0,
       "step": 136
     },
     {
-      "entropy": 0.5592403262853622,
-      "epoch": 4.288,
-      "grad_norm": 0.02239048294723034,
+      "entropy": 0.6240550950169563,
+      "epoch": 0.8903330625507717,
+      "grad_norm": 0.009444071911275387,
       "learning_rate": 0.0002,
-      "loss": 0.5616511702537537,
-      "mean_token_accuracy": 0.7738818973302841,
-      "num_tokens": 10565760.0,
+      "loss": 0.6233399510383606,
+      "mean_token_accuracy": 0.749994620680809,
+      "num_tokens": 5122453.0,
       "step": 137
     },
     {
-      "entropy": 0.5580462664365768,
-      "epoch": 4.32,
-      "grad_norm": 0.029557570815086365,
+      "entropy": 0.6302881464362144,
+      "epoch": 0.8968318440292445,
+      "grad_norm": 0.009541204199194908,
       "learning_rate": 0.0002,
-      "loss": 0.5544324517250061,
-      "mean_token_accuracy": 0.7771971821784973,
-      "num_tokens": 10644777.0,
+      "loss": 0.6329821348190308,
+      "mean_token_accuracy": 0.7429886758327484,
+      "num_tokens": 5160031.0,
       "step": 138
     },
     {
-      "entropy": 0.5683742761611938,
-      "epoch": 4.352,
-      "grad_norm": 0.022613991051912308,
+      "entropy": 0.6242454648017883,
+      "epoch": 0.9033306255077173,
+      "grad_norm": 0.01136871799826622,
       "learning_rate": 0.0002,
-      "loss": 0.5679054260253906,
-      "mean_token_accuracy": 0.7711237072944641,
-      "num_tokens": 10721084.0,
+      "loss": 0.628311038017273,
+      "mean_token_accuracy": 0.7444250509142876,
+      "num_tokens": 5197538.0,
       "step": 139
     },
     {
-      "entropy": 0.5512652546167374,
-      "epoch": 4.384,
-      "grad_norm": 0.023316726088523865,
+      "entropy": 0.6239208057522774,
+      "epoch": 0.9098294069861901,
+      "grad_norm": 0.009039361029863358,
       "learning_rate": 0.0002,
-      "loss": 0.5495034456253052,
-      "mean_token_accuracy": 0.7789248526096344,
-      "num_tokens": 10799528.0,
+      "loss": 0.6212806105613708,
+      "mean_token_accuracy": 0.7484918162226677,
+      "num_tokens": 5234839.0,
       "step": 140
     },
     {
-      "entropy": 0.5452638119459152,
-      "epoch": 4.416,
-      "grad_norm": 0.023327913135290146,
+      "entropy": 0.6303699910640717,
+      "epoch": 0.9163281884646629,
+      "grad_norm": 0.008452481590211391,
       "learning_rate": 0.0002,
-      "loss": 0.5428160429000854,
-      "mean_token_accuracy": 0.7811004668474197,
-      "num_tokens": 10878802.0,
+      "loss": 0.6276811361312866,
+      "mean_token_accuracy": 0.7478921040892601,
+      "num_tokens": 5272501.0,
       "step": 141
     },
     {
-      "entropy": 0.5505389273166656,
-      "epoch": 4.448,
-      "grad_norm": 0.024172673001885414,
+      "entropy": 0.6282826215028763,
+      "epoch": 0.9228269699431356,
+      "grad_norm": 0.009988308884203434,
       "learning_rate": 0.0002,
-      "loss": 0.5476949214935303,
-      "mean_token_accuracy": 0.780018538236618,
-      "num_tokens": 10957456.0,
+      "loss": 0.6308070421218872,
+      "mean_token_accuracy": 0.744203083217144,
+      "num_tokens": 5310068.0,
       "step": 142
     },
     {
-      "entropy": 0.5516746640205383,
-      "epoch": 4.48,
-      "grad_norm": 0.024863507598638535,
+      "entropy": 0.6363607943058014,
+      "epoch": 0.9293257514216084,
+      "grad_norm": 0.010694671422243118,
       "learning_rate": 0.0002,
-      "loss": 0.5560992956161499,
-      "mean_token_accuracy": 0.7761176824569702,
-      "num_tokens": 11036458.0,
+      "loss": 0.6361459493637085,
+      "mean_token_accuracy": 0.7425540909171104,
+      "num_tokens": 5347594.0,
       "step": 143
     },
     {
-      "entropy": 0.5538354814052582,
-      "epoch": 4.5120000000000005,
-      "grad_norm": 0.03181128948926926,
+      "entropy": 0.6332810074090958,
+      "epoch": 0.9358245329000813,
+      "grad_norm": 0.008342958055436611,
       "learning_rate": 0.0002,
-      "loss": 0.557499349117279,
-      "mean_token_accuracy": 0.7750093787908554,
-      "num_tokens": 11114465.0,
+      "loss": 0.6327856779098511,
+      "mean_token_accuracy": 0.7445112019777298,
+      "num_tokens": 5385240.0,
       "step": 144
     },
     {
-      "entropy": 0.5519821345806122,
-      "epoch": 4.5440000000000005,
-      "grad_norm": 0.022477390244603157,
+      "entropy": 0.622712031006813,
+      "epoch": 0.942323314378554,
+      "grad_norm": 0.009324009530246258,
       "learning_rate": 0.0002,
-      "loss": 0.5489175319671631,
-      "mean_token_accuracy": 0.7789605855941772,
-      "num_tokens": 11193250.0,
+      "loss": 0.622111439704895,
+      "mean_token_accuracy": 0.7491918131709099,
+      "num_tokens": 5423063.0,
       "step": 145
     },
     {
-      "entropy": 0.5598633885383606,
-      "epoch": 4.576,
-      "grad_norm": 0.02509782463312149,
+      "entropy": 0.6331967115402222,
+      "epoch": 0.9488220958570268,
+      "grad_norm": 0.009456835687160492,
       "learning_rate": 0.0002,
-      "loss": 0.5523159503936768,
-      "mean_token_accuracy": 0.7775833010673523,
-      "num_tokens": 11272848.0,
+      "loss": 0.635155200958252,
+      "mean_token_accuracy": 0.7427491322159767,
+      "num_tokens": 5460300.0,
       "step": 146
     },
     {
-      "entropy": 0.5443387776613235,
-      "epoch": 4.608,
-      "grad_norm": 0.02487492933869362,
+      "entropy": 0.6007715612649918,
+      "epoch": 0.9553208773354996,
+      "grad_norm": 0.009612048044800758,
       "learning_rate": 0.0002,
-      "loss": 0.5392158031463623,
-      "mean_token_accuracy": 0.7834303230047226,
-      "num_tokens": 11353910.0,
+      "loss": 0.6061209440231323,
+      "mean_token_accuracy": 0.7546330019831657,
+      "num_tokens": 5497104.0,
       "step": 147
     },
     {
-      "entropy": 0.5553220510482788,
-      "epoch": 4.64,
-      "grad_norm": 0.02597399242222309,
+      "entropy": 0.623559907078743,
+      "epoch": 0.9618196588139724,
+      "grad_norm": 0.009891507215797901,
       "learning_rate": 0.0002,
-      "loss": 0.5571849346160889,
-      "mean_token_accuracy": 0.776119738817215,
-      "num_tokens": 11430800.0,
+      "loss": 0.6299046874046326,
+      "mean_token_accuracy": 0.7460341230034828,
+      "num_tokens": 5534717.0,
       "step": 148
     },
     {
-      "entropy": 0.5394623875617981,
-      "epoch": 4.672,
-      "grad_norm": 0.02306293696165085,
+      "entropy": 0.6161656081676483,
+      "epoch": 0.9683184402924452,
+      "grad_norm": 0.009623561054468155,
       "learning_rate": 0.0002,
-      "loss": 0.5412307381629944,
-      "mean_token_accuracy": 0.7825190275907516,
-      "num_tokens": 11511445.0,
+      "loss": 0.6207596659660339,
+      "mean_token_accuracy": 0.7484462484717369,
+      "num_tokens": 5572155.0,
       "step": 149
     },
     {
-      "entropy": 0.5512889623641968,
-      "epoch": 4.704,
-      "grad_norm": 0.02301475964486599,
+      "entropy": 0.6296671330928802,
+      "epoch": 0.974817221770918,
+      "grad_norm": 0.009666231460869312,
       "learning_rate": 0.0002,
-      "loss": 0.5503370761871338,
-      "mean_token_accuracy": 0.7782942056655884,
-      "num_tokens": 11590429.0,
+      "loss": 0.6299723386764526,
+      "mean_token_accuracy": 0.7442828044295311,
+      "num_tokens": 5609434.0,
       "step": 150
     },
     {
-      "entropy": 0.5578732192516327,
-      "epoch": 4.736,
-      "grad_norm": 0.02201995998620987,
+      "entropy": 0.635062001645565,
+      "epoch": 0.9813160032493907,
+      "grad_norm": 0.009376533329486847,
       "learning_rate": 0.0002,
-      "loss": 0.5536071062088013,
-      "mean_token_accuracy": 0.7773238718509674,
-      "num_tokens": 11669487.0,
+      "loss": 0.6303079128265381,
+      "mean_token_accuracy": 0.7447433397173882,
+      "num_tokens": 5646893.0,
       "step": 151
     },
     {
-      "entropy": 0.5440615564584732,
-      "epoch": 4.768,
-      "grad_norm": 0.02202589437365532,
+      "entropy": 0.630713015794754,
+      "epoch": 0.9878147847278635,
+      "grad_norm": 0.009512092918157578,
       "learning_rate": 0.0002,
-      "loss": 0.5411347150802612,
-      "mean_token_accuracy": 0.7818420082330704,
-      "num_tokens": 11750325.0,
+      "loss": 0.6241896152496338,
+      "mean_token_accuracy": 0.745731770992279,
+      "num_tokens": 5684403.0,
       "step": 152
     },
     {
-      "entropy": 0.5498859584331512,
-      "epoch": 4.8,
-      "grad_norm": 0.026776723563671112,
+      "entropy": 0.623988039791584,
+      "epoch": 0.9943135662063363,
+      "grad_norm": 0.00946712028235197,
       "learning_rate": 0.0002,
-      "loss": 0.5520839691162109,
-      "mean_token_accuracy": 0.7781284302473068,
-      "num_tokens": 11827703.0,
+      "loss": 0.6251310110092163,
+      "mean_token_accuracy": 0.7487976476550102,
+      "num_tokens": 5721608.0,
       "step": 153
     },
     {
-      "entropy": 0.5513161569833755,
-      "epoch": 4.832,
-      "grad_norm": 0.021133486181497574,
+      "entropy": 0.5962605476379395,
+      "epoch": 1.0,
+      "grad_norm": 0.010017761029303074,
       "learning_rate": 0.0002,
-      "loss": 0.5512353181838989,
-      "mean_token_accuracy": 0.7764542549848557,
-      "num_tokens": 11904960.0,
+      "loss": 0.6195710897445679,
+      "mean_token_accuracy": 0.754073713506971,
+      "num_tokens": 5737850.0,
       "step": 154
     },
     {
-      "entropy": 0.553078830242157,
-      "epoch": 4.864,
-      "grad_norm": 0.025516293942928314,
+      "entropy": 0.6159528121352196,
+      "epoch": 1.0064987814784727,
+      "grad_norm": 0.011626223102211952,
       "learning_rate": 0.0002,
-      "loss": 0.5475062131881714,
-      "mean_token_accuracy": 0.7789602428674698,
-      "num_tokens": 11985016.0,
+      "loss": 0.6283953785896301,
+      "mean_token_accuracy": 0.7437271103262901,
+      "num_tokens": 5775226.0,
       "step": 155
     },
     {
-      "entropy": 0.5605203360319138,
-      "epoch": 4.896,
-      "grad_norm": 0.021059760823845863,
+      "entropy": 0.6398139446973801,
+      "epoch": 1.0129975629569457,
+      "grad_norm": 0.009600057266652584,
       "learning_rate": 0.0002,
-      "loss": 0.5566121339797974,
-      "mean_token_accuracy": 0.7765137851238251,
-      "num_tokens": 12063054.0,
+      "loss": 0.6418601274490356,
+      "mean_token_accuracy": 0.7394955083727837,
+      "num_tokens": 5812842.0,
       "step": 156
     },
     {
-      "entropy": 0.5402437597513199,
-      "epoch": 4.928,
-      "grad_norm": 0.0237311739474535,
+      "entropy": 0.6106663420796394,
+      "epoch": 1.0194963444354184,
+      "grad_norm": 0.01011701114475727,
       "learning_rate": 0.0002,
-      "loss": 0.5375882387161255,
-      "mean_token_accuracy": 0.7829015702009201,
-      "num_tokens": 12142425.0,
+      "loss": 0.6077284812927246,
+      "mean_token_accuracy": 0.754165381193161,
+      "num_tokens": 5850157.0,
       "step": 157
     },
     {
-      "entropy": 0.5493930131196976,
-      "epoch": 4.96,
-      "grad_norm": 0.02653537131845951,
+      "entropy": 0.6355043426156044,
+      "epoch": 1.025995125913891,
+      "grad_norm": 0.009700894355773926,
       "learning_rate": 0.0002,
-      "loss": 0.5503488779067993,
-      "mean_token_accuracy": 0.7782562971115112,
-      "num_tokens": 12220945.0,
+      "loss": 0.6273776888847351,
+      "mean_token_accuracy": 0.7454625219106674,
+      "num_tokens": 5887576.0,
       "step": 158
     },
     {
-      "entropy": 0.5531943738460541,
-      "epoch": 4.992,
-      "grad_norm": 0.02164315991103649,
+      "entropy": 0.629393994808197,
+      "epoch": 1.032493907392364,
+      "grad_norm": 0.007958617061376572,
       "learning_rate": 0.0002,
-      "loss": 0.5544183254241943,
-      "mean_token_accuracy": 0.7767967581748962,
-      "num_tokens": 12298554.0,
+      "loss": 0.6252051591873169,
+      "mean_token_accuracy": 0.7460998743772507,
+      "num_tokens": 5925165.0,
       "step": 159
     },
     {
-      "entropy": 0.5414583086967468,
-      "epoch": 5.0,
-      "grad_norm": 0.040175583213567734,
+      "entropy": 0.6266166046261787,
+      "epoch": 1.0389926888708367,
+      "grad_norm": 0.00998145341873169,
       "learning_rate": 0.0002,
-      "loss": 0.5361344218254089,
-      "mean_token_accuracy": 0.7858189940452576,
-      "num_tokens": 12319146.0,
+      "loss": 0.6290674209594727,
+      "mean_token_accuracy": 0.7430498450994492,
+      "num_tokens": 5962505.0,
       "step": 160
     },
     {
-      "entropy": 0.5582098960876465,
-      "epoch": 5.032,
-      "grad_norm": 0.022894199937582016,
+      "entropy": 0.6140859052538872,
+      "epoch": 1.0454914703493094,
+      "grad_norm": 0.010381487198174,
       "learning_rate": 0.0002,
-      "loss": 0.5531452894210815,
-      "mean_token_accuracy": 0.7776804864406586,
-      "num_tokens": 12397752.0,
+      "loss": 0.6180790662765503,
+      "mean_token_accuracy": 0.7507300451397896,
+      "num_tokens": 5999871.0,
       "step": 161
     },
     {
-      "entropy": 0.5599130690097809,
-      "epoch": 5.064,
-      "grad_norm": 0.025759946554899216,
+      "entropy": 0.6197154447436333,
+      "epoch": 1.0519902518277824,
+      "grad_norm": 0.009876130148768425,
       "learning_rate": 0.0002,
-      "loss": 0.5595382452011108,
-      "mean_token_accuracy": 0.7729753255844116,
-      "num_tokens": 12476444.0,
+      "loss": 0.6249828338623047,
+      "mean_token_accuracy": 0.7454148605465889,
+      "num_tokens": 6037575.0,
       "step": 162
     },
     {
-      "entropy": 0.5494701564311981,
-      "epoch": 5.096,
-      "grad_norm": 0.027650512754917145,
+      "entropy": 0.6122508272528648,
+      "epoch": 1.058489033306255,
+      "grad_norm": 0.008684883825480938,
       "learning_rate": 0.0002,
-      "loss": 0.552753210067749,
-      "mean_token_accuracy": 0.7775422781705856,
-      "num_tokens": 12556559.0,
+      "loss": 0.6138869524002075,
+      "mean_token_accuracy": 0.7517640963196754,
+      "num_tokens": 6074892.0,
       "step": 163
     },
     {
-      "entropy": 0.5431053638458252,
-      "epoch": 5.128,
-      "grad_norm": 0.02576346881687641,
+      "entropy": 0.6194281056523323,
+      "epoch": 1.0649878147847278,
+      "grad_norm": 0.010471724905073643,
       "learning_rate": 0.0002,
-      "loss": 0.543494701385498,
-      "mean_token_accuracy": 0.7810918986797333,
-      "num_tokens": 12635217.0,
+      "loss": 0.6223649978637695,
+      "mean_token_accuracy": 0.7462374493479729,
+      "num_tokens": 6112197.0,
       "step": 164
     },
     {
-      "entropy": 0.5598382949829102,
-      "epoch": 5.16,
-      "grad_norm": 0.022493746131658554,
+      "entropy": 0.6203417181968689,
+      "epoch": 1.0714865962632008,
+      "grad_norm": 0.00992770865559578,
       "learning_rate": 0.0002,
-      "loss": 0.5572130680084229,
-      "mean_token_accuracy": 0.7740350961685181,
-      "num_tokens": 12713524.0,
+      "loss": 0.6209006309509277,
+      "mean_token_accuracy": 0.7480901628732681,
+      "num_tokens": 6149571.0,
       "step": 165
     },
     {
-      "entropy": 0.545780748128891,
-      "epoch": 5.192,
-      "grad_norm": 0.027349475771188736,
+      "entropy": 0.6099580377340317,
+      "epoch": 1.0779853777416735,
+      "grad_norm": 0.00981567706912756,
       "learning_rate": 0.0002,
-      "loss": 0.5389693379402161,
-      "mean_token_accuracy": 0.7838597297668457,
-      "num_tokens": 12793300.0,
+      "loss": 0.6166203022003174,
+      "mean_token_accuracy": 0.7501059100031853,
+      "num_tokens": 6186570.0,
       "step": 166
     },
     {
-      "entropy": 0.5592958778142929,
-      "epoch": 5.224,
-      "grad_norm": 0.028276948258280754,
+      "entropy": 0.6217071712017059,
+      "epoch": 1.0844841592201462,
+      "grad_norm": 0.009364967234432697,
       "learning_rate": 0.0002,
-      "loss": 0.5530558824539185,
-      "mean_token_accuracy": 0.7787036746740341,
-      "num_tokens": 12872372.0,
+      "loss": 0.6240255236625671,
+      "mean_token_accuracy": 0.7475733831524849,
+      "num_tokens": 6224124.0,
       "step": 167
     },
     {
-      "entropy": 0.5440209805965424,
-      "epoch": 5.256,
-      "grad_norm": 0.029081838205456734,
+      "entropy": 0.6265001520514488,
+      "epoch": 1.090982940698619,
+      "grad_norm": 0.009150683879852295,
       "learning_rate": 0.0002,
-      "loss": 0.5472489595413208,
-      "mean_token_accuracy": 0.7790472209453583,
-      "num_tokens": 12950474.0,
+      "loss": 0.6249997615814209,
+      "mean_token_accuracy": 0.7465919181704521,
+      "num_tokens": 6261544.0,
       "step": 168
     },
     {
-      "entropy": 0.541302427649498,
-      "epoch": 5.288,
-      "grad_norm": 0.03218315914273262,
+      "entropy": 0.6337415799498558,
+      "epoch": 1.0974817221770918,
+      "grad_norm": 0.009555370546877384,
       "learning_rate": 0.0002,
-      "loss": 0.5454750657081604,
-      "mean_token_accuracy": 0.7817174345254898,
-      "num_tokens": 13029478.0,
+      "loss": 0.6337403059005737,
+      "mean_token_accuracy": 0.7431929185986519,
+      "num_tokens": 6298708.0,
       "step": 169
     },
     {
-      "entropy": 0.5527262985706329,
-      "epoch": 5.32,
-      "grad_norm": 0.023637501522898674,
+      "entropy": 0.6212407127022743,
+      "epoch": 1.1039805036555645,
+      "grad_norm": 0.009204280562698841,
       "learning_rate": 0.0002,
-      "loss": 0.5539190769195557,
-      "mean_token_accuracy": 0.7765050381422043,
-      "num_tokens": 13107874.0,
+      "loss": 0.6198785305023193,
+      "mean_token_accuracy": 0.7490911036729813,
+      "num_tokens": 6336357.0,
       "step": 170
     },
     {
-      "entropy": 0.5566678494215012,
-      "epoch": 5.352,
-      "grad_norm": 0.030741648748517036,
+      "entropy": 0.6211022287607193,
+      "epoch": 1.1104792851340373,
+      "grad_norm": 0.010933548212051392,
       "learning_rate": 0.0002,
-      "loss": 0.5524072647094727,
-      "mean_token_accuracy": 0.7761627286672592,
-      "num_tokens": 13187034.0,
+      "loss": 0.6218096017837524,
+      "mean_token_accuracy": 0.7478840723633766,
+      "num_tokens": 6373815.0,
       "step": 171
     },
     {
-      "entropy": 0.5590717494487762,
-      "epoch": 5.384,
-      "grad_norm": 0.02631162852048874,
+      "entropy": 0.6218304485082626,
+      "epoch": 1.1169780666125102,
+      "grad_norm": 0.007957971654832363,
       "learning_rate": 0.0002,
-      "loss": 0.5528249740600586,
-      "mean_token_accuracy": 0.7761536091566086,
-      "num_tokens": 13266614.0,
+      "loss": 0.6211934089660645,
+      "mean_token_accuracy": 0.7479007542133331,
+      "num_tokens": 6411469.0,
       "step": 172
     },
     {
-      "entropy": 0.54710254073143,
-      "epoch": 5.416,
-      "grad_norm": 0.030295731499791145,
+      "entropy": 0.623476430773735,
+      "epoch": 1.123476848090983,
+      "grad_norm": 0.008693872950971127,
       "learning_rate": 0.0002,
-      "loss": 0.5446665287017822,
-      "mean_token_accuracy": 0.7795543521642685,
-      "num_tokens": 13345744.0,
+      "loss": 0.628318727016449,
+      "mean_token_accuracy": 0.7435088455677032,
+      "num_tokens": 6448782.0,
       "step": 173
     },
     {
-      "entropy": 0.5439110398292542,
-      "epoch": 5.448,
-      "grad_norm": 0.028032993897795677,
+      "entropy": 0.6133372634649277,
+      "epoch": 1.1299756295694556,
+      "grad_norm": 0.010963717475533485,
       "learning_rate": 0.0002,
-      "loss": 0.5464545488357544,
-      "mean_token_accuracy": 0.7813926637172699,
-      "num_tokens": 13424891.0,
+      "loss": 0.6166828274726868,
+      "mean_token_accuracy": 0.7512464672327042,
+      "num_tokens": 6485790.0,
       "step": 174
     },
     {
-      "entropy": 0.545119971036911,
-      "epoch": 5.48,
-      "grad_norm": 0.026973431929945946,
+      "entropy": 0.6234560832381248,
+      "epoch": 1.1364744110479286,
+      "grad_norm": 0.009953116998076439,
       "learning_rate": 0.0002,
-      "loss": 0.5461452007293701,
-      "mean_token_accuracy": 0.7802118062973022,
-      "num_tokens": 13503380.0,
+      "loss": 0.6275191903114319,
+      "mean_token_accuracy": 0.7457004636526108,
+      "num_tokens": 6523433.0,
       "step": 175
     },
     {
-      "entropy": 0.5575177073478699,
-      "epoch": 5.5120000000000005,
-      "grad_norm": 0.026447977870702744,
+      "entropy": 0.612127959728241,
+      "epoch": 1.1429731925264013,
+      "grad_norm": 0.008568039163947105,
       "learning_rate": 0.0002,
-      "loss": 0.5565048456192017,
-      "mean_token_accuracy": 0.7753730118274689,
-      "num_tokens": 13581504.0,
+      "loss": 0.6111507415771484,
+      "mean_token_accuracy": 0.7523760348558426,
+      "num_tokens": 6560764.0,
       "step": 176
     },
     {
-      "entropy": 0.5443647503852844,
-      "epoch": 5.5440000000000005,
-      "grad_norm": 0.026733124628663063,
+      "entropy": 0.6384553462266922,
+      "epoch": 1.149471974004874,
+      "grad_norm": 0.010018682107329369,
       "learning_rate": 0.0002,
-      "loss": 0.5393130779266357,
-      "mean_token_accuracy": 0.7818372398614883,
-      "num_tokens": 13659491.0,
+      "loss": 0.6336829662322998,
+      "mean_token_accuracy": 0.743287555873394,
+      "num_tokens": 6598409.0,
       "step": 177
     },
     {
-      "entropy": 0.5571713000535965,
-      "epoch": 5.576,
-      "grad_norm": 0.02440747246146202,
+      "entropy": 0.6272930577397346,
+      "epoch": 1.155970755483347,
+      "grad_norm": 0.009546359069645405,
       "learning_rate": 0.0002,
-      "loss": 0.5560795664787292,
-      "mean_token_accuracy": 0.7749420404434204,
-      "num_tokens": 13736686.0,
+      "loss": 0.626544713973999,
+      "mean_token_accuracy": 0.7449387013912201,
+      "num_tokens": 6636067.0,
       "step": 178
     },
     {
-      "entropy": 0.543163537979126,
-      "epoch": 5.608,
-      "grad_norm": 0.020597316324710846,
+      "entropy": 0.6086627542972565,
+      "epoch": 1.1624695369618196,
+      "grad_norm": 0.009319686330854893,
       "learning_rate": 0.0002,
-      "loss": 0.5401595234870911,
-      "mean_token_accuracy": 0.7831601202487946,
-      "num_tokens": 13816007.0,
+      "loss": 0.6141277551651001,
+      "mean_token_accuracy": 0.7520448267459869,
+      "num_tokens": 6673446.0,
       "step": 179
     },
     {
-      "entropy": 0.5275207906961441,
-      "epoch": 5.64,
-      "grad_norm": 0.021762844175100327,
+      "entropy": 0.5994155630469322,
+      "epoch": 1.1689683184402924,
+      "grad_norm": 0.011318805627524853,
       "learning_rate": 0.0002,
-      "loss": 0.5270570516586304,
-      "mean_token_accuracy": 0.7863520830869675,
-      "num_tokens": 13895212.0,
+      "loss": 0.6087459325790405,
+      "mean_token_accuracy": 0.7519726976752281,
+      "num_tokens": 6710715.0,
       "step": 180
     },
     {
-      "entropy": 0.5509206503629684,
-      "epoch": 5.672,
-      "grad_norm": 0.025359727442264557,
+      "entropy": 0.6187063455581665,
+      "epoch": 1.1754670999187653,
+      "grad_norm": 0.009100009687244892,
       "learning_rate": 0.0002,
-      "loss": 0.5518645644187927,
-      "mean_token_accuracy": 0.7776848524808884,
-      "num_tokens": 13973672.0,
+      "loss": 0.6213230490684509,
+      "mean_token_accuracy": 0.7466496899724007,
+      "num_tokens": 6748737.0,
       "step": 181
     },
     {
-      "entropy": 0.5468782633543015,
-      "epoch": 5.704,
-      "grad_norm": 0.023991327732801437,
+      "entropy": 0.6402401477098465,
+      "epoch": 1.181965881397238,
+      "grad_norm": 0.009086531586945057,
       "learning_rate": 0.0002,
-      "loss": 0.5474581718444824,
-      "mean_token_accuracy": 0.7794180512428284,
-      "num_tokens": 14052724.0,
+      "loss": 0.636063277721405,
+      "mean_token_accuracy": 0.7429427057504654,
+      "num_tokens": 6785555.0,
       "step": 182
     },
     {
-      "entropy": 0.5526992231607437,
-      "epoch": 5.736,
-      "grad_norm": 0.01987096108496189,
+      "entropy": 0.6275543943047523,
+      "epoch": 1.1884646628757107,
+      "grad_norm": 0.00983697734773159,
       "learning_rate": 0.0002,
-      "loss": 0.5501133799552917,
-      "mean_token_accuracy": 0.7786473035812378,
-      "num_tokens": 14131904.0,
+      "loss": 0.6195461750030518,
+      "mean_token_accuracy": 0.7482333034276962,
+      "num_tokens": 6822608.0,
       "step": 183
     },
     {
-      "entropy": 0.5467601120471954,
-      "epoch": 5.768,
-      "grad_norm": 0.02320663444697857,
+      "entropy": 0.6261148154735565,
+      "epoch": 1.1949634443541837,
+      "grad_norm": 0.008595704101026058,
       "learning_rate": 0.0002,
-      "loss": 0.5411819815635681,
-      "mean_token_accuracy": 0.7823377251625061,
-      "num_tokens": 14210960.0,
+      "loss": 0.6235471963882446,
+      "mean_token_accuracy": 0.7465647235512733,
+      "num_tokens": 6860107.0,
       "step": 184
     },
     {
-      "entropy": 0.5405216664075851,
-      "epoch": 5.8,
-      "grad_norm": 0.021056123077869415,
+      "entropy": 0.615828700363636,
+      "epoch": 1.2014622258326564,
+      "grad_norm": 0.010348568670451641,
       "learning_rate": 0.0002,
-      "loss": 0.5362813472747803,
-      "mean_token_accuracy": 0.7848241031169891,
-      "num_tokens": 14291726.0,
+      "loss": 0.6194600462913513,
+      "mean_token_accuracy": 0.7490681707859039,
+      "num_tokens": 6897202.0,
       "step": 185
     },
     {
-      "entropy": 0.5541730970144272,
-      "epoch": 5.832,
-      "grad_norm": 0.02818496711552143,
+      "entropy": 0.6132261380553246,
+      "epoch": 1.207961007311129,
+      "grad_norm": 0.010876818560063839,
       "learning_rate": 0.0002,
-      "loss": 0.5557092428207397,
-      "mean_token_accuracy": 0.7762422263622284,
-      "num_tokens": 14369724.0,
+      "loss": 0.6224918365478516,
+      "mean_token_accuracy": 0.7479420974850655,
+      "num_tokens": 6934670.0,
       "step": 186
     },
     {
-      "entropy": 0.5322004407644272,
-      "epoch": 5.864,
-      "grad_norm": 0.02371285855770111,
+      "entropy": 0.6103203073143959,
+      "epoch": 1.214459788789602,
+      "grad_norm": 0.009953301399946213,
       "learning_rate": 0.0002,
-      "loss": 0.531102180480957,
-      "mean_token_accuracy": 0.7858971059322357,
-      "num_tokens": 14448931.0,
+      "loss": 0.6190877556800842,
+      "mean_token_accuracy": 0.7497604489326477,
+      "num_tokens": 6971658.0,
       "step": 187
     },
     {
-      "entropy": 0.5547012537717819,
-      "epoch": 5.896,
-      "grad_norm": 0.036903511732816696,
+      "entropy": 0.617573581635952,
+      "epoch": 1.2209585702680747,
+      "grad_norm": 0.00837781187146902,
       "learning_rate": 0.0002,
-      "loss": 0.5559762120246887,
-      "mean_token_accuracy": 0.776378944516182,
-      "num_tokens": 14526949.0,
+      "loss": 0.617588996887207,
+      "mean_token_accuracy": 0.7494269385933876,
+      "num_tokens": 7009123.0,
       "step": 188
     },
     {
-      "entropy": 0.5446574836969376,
-      "epoch": 5.928,
-      "grad_norm": 0.026452112942934036,
+      "entropy": 0.629171572625637,
+      "epoch": 1.2274573517465475,
+      "grad_norm": 0.008275787346065044,
       "learning_rate": 0.0002,
-      "loss": 0.5377969145774841,
-      "mean_token_accuracy": 0.7841950803995132,
-      "num_tokens": 14604335.0,
+      "loss": 0.6283445358276367,
+      "mean_token_accuracy": 0.7444169595837593,
+      "num_tokens": 7046906.0,
       "step": 189
     },
     {
-      "entropy": 0.5493032783269882,
-      "epoch": 5.96,
-      "grad_norm": 0.029249282553792,
+      "entropy": 0.6318405717611313,
+      "epoch": 1.2339561332250204,
+      "grad_norm": 0.008116304874420166,
       "learning_rate": 0.0002,
-      "loss": 0.5473008155822754,
-      "mean_token_accuracy": 0.7784675061702728,
-      "num_tokens": 14683487.0,
+      "loss": 0.6320441365242004,
+      "mean_token_accuracy": 0.7443558126688004,
+      "num_tokens": 7084180.0,
       "step": 190
     },
     {
-      "entropy": 0.5538213402032852,
-      "epoch": 5.992,
-      "grad_norm": 0.025735314935445786,
+      "entropy": 0.6146104484796524,
+      "epoch": 1.2404549147034931,
+      "grad_norm": 0.009803351014852524,
       "learning_rate": 0.0002,
-      "loss": 0.552562952041626,
-      "mean_token_accuracy": 0.7783134877681732,
-      "num_tokens": 14763050.0,
+      "loss": 0.6140527725219727,
+      "mean_token_accuracy": 0.7488429099321365,
+      "num_tokens": 7121475.0,
       "step": 191
     },
     {
-      "entropy": 0.5568431615829468,
-      "epoch": 6.0,
-      "grad_norm": 0.0419342927634716,
+      "entropy": 0.6153127253055573,
+      "epoch": 1.2469536961819658,
+      "grad_norm": 0.007825122214853764,
       "learning_rate": 0.0002,
-      "loss": 0.5596759915351868,
-      "mean_token_accuracy": 0.7757459878921509,
-      "num_tokens": 14783084.0,
+      "loss": 0.6195809245109558,
+      "mean_token_accuracy": 0.7493974566459656,
+      "num_tokens": 7158753.0,
       "step": 192
     },
     {
-      "entropy": 0.5499615967273712,
-      "epoch": 6.032,
-      "grad_norm": 0.030985454097390175,
+      "entropy": 0.6186842620372772,
+      "epoch": 1.2534524776604385,
+      "grad_norm": 0.00928817130625248,
       "learning_rate": 0.0002,
-      "loss": 0.5423964262008667,
-      "mean_token_accuracy": 0.7811243683099747,
-      "num_tokens": 14861223.0,
+      "loss": 0.6203854084014893,
+      "mean_token_accuracy": 0.748259611427784,
+      "num_tokens": 7195797.0,
       "step": 193
     },
     {
-      "entropy": 0.545973002910614,
-      "epoch": 6.064,
-      "grad_norm": 0.03047715686261654,
+      "entropy": 0.6210475564002991,
+      "epoch": 1.2599512591389115,
+      "grad_norm": 0.010004797019064426,
       "learning_rate": 0.0002,
-      "loss": 0.5429363250732422,
-      "mean_token_accuracy": 0.7807959467172623,
-      "num_tokens": 14939948.0,
+      "loss": 0.6249719858169556,
+      "mean_token_accuracy": 0.7474558278918266,
+      "num_tokens": 7233547.0,
       "step": 194
     },
     {
-      "entropy": 0.5407204329967499,
-      "epoch": 6.096,
-      "grad_norm": 0.026628416031599045,
+      "entropy": 0.6319233104586601,
+      "epoch": 1.2664500406173842,
+      "grad_norm": 0.010124076157808304,
       "learning_rate": 0.0002,
-      "loss": 0.5426843166351318,
-      "mean_token_accuracy": 0.7809197157621384,
-      "num_tokens": 15018808.0,
+      "loss": 0.6346510648727417,
+      "mean_token_accuracy": 0.7430363744497299,
+      "num_tokens": 7271165.0,
       "step": 195
     },
     {
-      "entropy": 0.5479950308799744,
-      "epoch": 6.128,
-      "grad_norm": 0.03580361232161522,
+      "entropy": 0.6228615492582321,
+      "epoch": 1.2729488220958571,
+      "grad_norm": 0.00920382421463728,
       "learning_rate": 0.0002,
-      "loss": 0.5534794926643372,
-      "mean_token_accuracy": 0.777429461479187,
-      "num_tokens": 15096870.0,
+      "loss": 0.6211229562759399,
+      "mean_token_accuracy": 0.7490492090582848,
+      "num_tokens": 7308659.0,
       "step": 196
     },
     {
-      "entropy": 0.5385048538446426,
-      "epoch": 6.16,
-      "grad_norm": 0.023637505248188972,
+      "entropy": 0.6310986131429672,
+      "epoch": 1.2794476035743299,
+      "grad_norm": 0.009224419482052326,
       "learning_rate": 0.0002,
-      "loss": 0.5380685329437256,
-      "mean_token_accuracy": 0.783260390162468,
-      "num_tokens": 15174757.0,
+      "loss": 0.626353919506073,
+      "mean_token_accuracy": 0.7454532235860825,
+      "num_tokens": 7346203.0,
       "step": 197
     },
     {
-      "entropy": 0.5591510534286499,
-      "epoch": 6.192,
-      "grad_norm": 0.03232690691947937,
+      "entropy": 0.610854484140873,
+      "epoch": 1.2859463850528026,
+      "grad_norm": 0.009150652214884758,
       "learning_rate": 0.0002,
-      "loss": 0.5529967546463013,
-      "mean_token_accuracy": 0.7766391485929489,
-      "num_tokens": 15252758.0,
+      "loss": 0.6110652685165405,
+      "mean_token_accuracy": 0.7504596933722496,
+      "num_tokens": 7383470.0,
       "step": 198
     },
     {
-      "entropy": 0.5531331598758698,
-      "epoch": 6.224,
-      "grad_norm": 0.027761192992329597,
+      "entropy": 0.6061088666319847,
+      "epoch": 1.2924451665312753,
+      "grad_norm": 0.008921864442527294,
       "learning_rate": 0.0002,
-      "loss": 0.5476256608963013,
-      "mean_token_accuracy": 0.7792367190122604,
-      "num_tokens": 15330164.0,
+      "loss": 0.611494779586792,
+      "mean_token_accuracy": 0.7522832676768303,
+      "num_tokens": 7420939.0,
       "step": 199
     },
     {
-      "entropy": 0.5431746989488602,
-      "epoch": 6.256,
-      "grad_norm": 0.029444098472595215,
+      "entropy": 0.6130355596542358,
+      "epoch": 1.2989439480097482,
+      "grad_norm": 0.009989106096327305,
       "learning_rate": 0.0002,
-      "loss": 0.5440195798873901,
-      "mean_token_accuracy": 0.7816928923130035,
-      "num_tokens": 15408291.0,
+      "loss": 0.6249145269393921,
+      "mean_token_accuracy": 0.7437470331788063,
+      "num_tokens": 7458037.0,
       "step": 200
     },
     {
-      "entropy": 0.5401950180530548,
-      "epoch": 6.288,
-      "grad_norm": 0.030943166464567184,
+      "entropy": 0.6245267316699028,
+      "epoch": 1.305442729488221,
+      "grad_norm": 0.008322354406118393,
       "learning_rate": 0.0002,
-      "loss": 0.5441604852676392,
-      "mean_token_accuracy": 0.7797262668609619,
-      "num_tokens": 15486855.0,
+      "loss": 0.6295627355575562,
+      "mean_token_accuracy": 0.7449559271335602,
+      "num_tokens": 7495466.0,
       "step": 201
     },
     {
-      "entropy": 0.5358632057905197,
-      "epoch": 6.32,
-      "grad_norm": 0.025611277669668198,
+      "entropy": 0.6188690215349197,
+      "epoch": 1.3119415109666939,
+      "grad_norm": 0.008306242525577545,
       "learning_rate": 0.0002,
-      "loss": 0.5336735844612122,
-      "mean_token_accuracy": 0.7844033241271973,
-      "num_tokens": 15565945.0,
+      "loss": 0.6199623942375183,
+      "mean_token_accuracy": 0.7464429587125778,
+      "num_tokens": 7533028.0,
       "step": 202
     },
     {
-      "entropy": 0.5415282398462296,
-      "epoch": 6.352,
-      "grad_norm": 0.03050980716943741,
+      "entropy": 0.6284244954586029,
+      "epoch": 1.3184402924451666,
+      "grad_norm": 0.008912610821425915,
       "learning_rate": 0.0002,
-      "loss": 0.5411715507507324,
-      "mean_token_accuracy": 0.7807140350341797,
-      "num_tokens": 15644120.0,
+      "loss": 0.6269126534461975,
+      "mean_token_accuracy": 0.7451603710651398,
+      "num_tokens": 7570454.0,
       "step": 203
     },
     {
-      "entropy": 0.5550639480352402,
-      "epoch": 6.384,
-      "grad_norm": 0.025711113587021828,
+      "entropy": 0.6253558248281479,
+      "epoch": 1.3249390739236393,
+      "grad_norm": 0.009339334443211555,
       "learning_rate": 0.0002,
-      "loss": 0.5523407459259033,
-      "mean_token_accuracy": 0.7759165465831757,
-      "num_tokens": 15721932.0,
+      "loss": 0.6178290247917175,
+      "mean_token_accuracy": 0.749000295996666,
+      "num_tokens": 7607813.0,
       "step": 204
     },
     {
-      "entropy": 0.5459660887718201,
-      "epoch": 6.416,
-      "grad_norm": 0.028625324368476868,
+      "entropy": 0.6330921202898026,
+      "epoch": 1.331437855402112,
+      "grad_norm": 0.009055222384631634,
       "learning_rate": 0.0002,
-      "loss": 0.5404061675071716,
-      "mean_token_accuracy": 0.7824661731719971,
-      "num_tokens": 15801402.0,
+      "loss": 0.627162754535675,
+      "mean_token_accuracy": 0.7448414042592049,
+      "num_tokens": 7645571.0,
       "step": 205
     },
     {
-      "entropy": 0.5531899780035019,
-      "epoch": 6.448,
-      "grad_norm": 0.026339178904891014,
+      "entropy": 0.612213708460331,
+      "epoch": 1.337936636880585,
+      "grad_norm": 0.00788091029971838,
       "learning_rate": 0.0002,
-      "loss": 0.5514405369758606,
-      "mean_token_accuracy": 0.7779718488454819,
-      "num_tokens": 15880514.0,
+      "loss": 0.6103406548500061,
+      "mean_token_accuracy": 0.7524209767580032,
+      "num_tokens": 7682827.0,
       "step": 206
     },
     {
-      "entropy": 0.5400968790054321,
-      "epoch": 6.48,
-      "grad_norm": 0.030028901994228363,
+      "entropy": 0.6196805313229561,
+      "epoch": 1.3444354183590577,
+      "grad_norm": 0.0074918679893016815,
       "learning_rate": 0.0002,
-      "loss": 0.5434284210205078,
-      "mean_token_accuracy": 0.7791207730770111,
-      "num_tokens": 15959144.0,
+      "loss": 0.6231103539466858,
+      "mean_token_accuracy": 0.7457561120390892,
+      "num_tokens": 7720532.0,
       "step": 207
     },
     {
-      "entropy": 0.5307324081659317,
-      "epoch": 6.5120000000000005,
-      "grad_norm": 0.02786673791706562,
+      "entropy": 0.6148431301116943,
+      "epoch": 1.3509341998375304,
+      "grad_norm": 0.008005033247172832,
       "learning_rate": 0.0002,
-      "loss": 0.5343931317329407,
-      "mean_token_accuracy": 0.7839188724756241,
-      "num_tokens": 16039776.0,
+      "loss": 0.6224607825279236,
+      "mean_token_accuracy": 0.7475835010409355,
+      "num_tokens": 7757914.0,
       "step": 208
     },
     {
-      "entropy": 0.5425277054309845,
-      "epoch": 6.5440000000000005,
-      "grad_norm": 0.02753874473273754,
+      "entropy": 0.6162410527467728,
+      "epoch": 1.3574329813160033,
+      "grad_norm": 0.008755719289183617,
       "learning_rate": 0.0002,
-      "loss": 0.5446346998214722,
-      "mean_token_accuracy": 0.7797478884458542,
-      "num_tokens": 16118884.0,
+      "loss": 0.6230734586715698,
+      "mean_token_accuracy": 0.7474293559789658,
+      "num_tokens": 7795111.0,
       "step": 209
     },
     {
-      "entropy": 0.5513077974319458,
-      "epoch": 6.576,
-      "grad_norm": 0.028762485831975937,
+      "entropy": 0.6132988333702087,
+      "epoch": 1.363931762794476,
+      "grad_norm": 0.008136742748320103,
       "learning_rate": 0.0002,
-      "loss": 0.5451760292053223,
-      "mean_token_accuracy": 0.7815571874380112,
-      "num_tokens": 16197259.0,
+      "loss": 0.6189343929290771,
+      "mean_token_accuracy": 0.7504551187157631,
+      "num_tokens": 7832380.0,
       "step": 210
     },
     {
-      "entropy": 0.560958668589592,
-      "epoch": 6.608,
-      "grad_norm": 0.02965443953871727,
+      "entropy": 0.62418133020401,
+      "epoch": 1.3704305442729487,
+      "grad_norm": 0.007090510334819555,
       "learning_rate": 0.0002,
-      "loss": 0.5526363849639893,
-      "mean_token_accuracy": 0.7772418111562729,
-      "num_tokens": 16276149.0,
+      "loss": 0.6275556087493896,
+      "mean_token_accuracy": 0.744689516723156,
+      "num_tokens": 7869774.0,
       "step": 211
     },
     {
-      "entropy": 0.537281408905983,
-      "epoch": 6.64,
-      "grad_norm": 0.023833326995372772,
+      "entropy": 0.6151341274380684,
+      "epoch": 1.3769293257514217,
+      "grad_norm": 0.008890950120985508,
       "learning_rate": 0.0002,
-      "loss": 0.532819390296936,
-      "mean_token_accuracy": 0.7847591936588287,
-      "num_tokens": 16356969.0,
+      "loss": 0.6140053272247314,
+      "mean_token_accuracy": 0.7508121132850647,
+      "num_tokens": 7906943.0,
       "step": 212
     },
     {
-      "entropy": 0.5362367182970047,
-      "epoch": 6.672,
-      "grad_norm": 0.02524951472878456,
+      "entropy": 0.6221983805298805,
+      "epoch": 1.3834281072298944,
+      "grad_norm": 0.009338715113699436,
       "learning_rate": 0.0002,
-      "loss": 0.5348524451255798,
-      "mean_token_accuracy": 0.7850412726402283,
-      "num_tokens": 16436635.0,
+      "loss": 0.6216065883636475,
+      "mean_token_accuracy": 0.7504482716321945,
+      "num_tokens": 7944646.0,
       "step": 213
     },
     {
-      "entropy": 0.5361433327198029,
-      "epoch": 6.704,
-      "grad_norm": 0.03111296147108078,
+      "entropy": 0.6131090745329857,
+      "epoch": 1.3899268887083671,
+      "grad_norm": 0.00849352777004242,
       "learning_rate": 0.0002,
-      "loss": 0.5393766164779663,
-      "mean_token_accuracy": 0.7813311666250229,
-      "num_tokens": 16515650.0,
+      "loss": 0.6095527410507202,
+      "mean_token_accuracy": 0.7533553466200829,
+      "num_tokens": 7982267.0,
       "step": 214
     },
     {
-      "entropy": 0.5386303067207336,
-      "epoch": 6.736,
-      "grad_norm": 0.026461219415068626,
+      "entropy": 0.6085495501756668,
+      "epoch": 1.39642567018684,
+      "grad_norm": 0.007540795486420393,
       "learning_rate": 0.0002,
-      "loss": 0.5373357534408569,
-      "mean_token_accuracy": 0.7834987640380859,
-      "num_tokens": 16595744.0,
+      "loss": 0.6092317700386047,
+      "mean_token_accuracy": 0.7533180490136147,
+      "num_tokens": 8019466.0,
       "step": 215
     },
     {
-      "entropy": 0.5392462760210037,
-      "epoch": 6.768,
-      "grad_norm": 0.027499942108988762,
+      "entropy": 0.6296884343028069,
+      "epoch": 1.4029244516653128,
+      "grad_norm": 0.009902682155370712,
       "learning_rate": 0.0002,
-      "loss": 0.5331963300704956,
-      "mean_token_accuracy": 0.7848135083913803,
-      "num_tokens": 16674509.0,
+      "loss": 0.6364089250564575,
+      "mean_token_accuracy": 0.7428499311208725,
+      "num_tokens": 8057400.0,
       "step": 216
     },
     {
-      "entropy": 0.5459368675947189,
-      "epoch": 6.8,
-      "grad_norm": 0.024176757782697678,
+      "entropy": 0.6165885776281357,
+      "epoch": 1.4094232331437855,
+      "grad_norm": 0.00911257229745388,
       "learning_rate": 0.0002,
-      "loss": 0.5415037870407104,
-      "mean_token_accuracy": 0.7818843722343445,
-      "num_tokens": 16754105.0,
+      "loss": 0.6247788071632385,
+      "mean_token_accuracy": 0.7464938163757324,
+      "num_tokens": 8094904.0,
       "step": 217
     },
     {
-      "entropy": 0.5565692782402039,
-      "epoch": 6.832,
-      "grad_norm": 0.02574721910059452,
+      "entropy": 0.6157428845763206,
+      "epoch": 1.4159220146222582,
+      "grad_norm": 0.007997944951057434,
       "learning_rate": 0.0002,
-      "loss": 0.5573182106018066,
-      "mean_token_accuracy": 0.7748913615942001,
-      "num_tokens": 16833119.0,
+      "loss": 0.6201274394989014,
+      "mean_token_accuracy": 0.7489686757326126,
+      "num_tokens": 8132290.0,
       "step": 218
     },
     {
-      "entropy": 0.5434060543775558,
-      "epoch": 6.864,
-      "grad_norm": 0.02367129735648632,
+      "entropy": 0.6217053309082985,
+      "epoch": 1.4224207961007311,
+      "grad_norm": 0.00811337772756815,
       "learning_rate": 0.0002,
-      "loss": 0.5418709516525269,
-      "mean_token_accuracy": 0.7805542647838593,
-      "num_tokens": 16913039.0,
+      "loss": 0.6197227239608765,
+      "mean_token_accuracy": 0.7485419660806656,
+      "num_tokens": 8169791.0,
       "step": 219
     },
     {
-      "entropy": 0.5493430346250534,
-      "epoch": 6.896,
-      "grad_norm": 0.02657475695014,
+      "entropy": 0.615436315536499,
+      "epoch": 1.4289195775792038,
+      "grad_norm": 0.008888707496225834,
       "learning_rate": 0.0002,
-      "loss": 0.547214686870575,
-      "mean_token_accuracy": 0.7784837335348129,
-      "num_tokens": 16991051.0,
+      "loss": 0.6124907732009888,
+      "mean_token_accuracy": 0.751116119325161,
+      "num_tokens": 8207092.0,
       "step": 220
     },
     {
-      "entropy": 0.5528377145528793,
-      "epoch": 6.928,
-      "grad_norm": 0.024269506335258484,
+      "entropy": 0.6246543675661087,
+      "epoch": 1.4354183590576768,
+      "grad_norm": 0.008732281625270844,
       "learning_rate": 0.0002,
-      "loss": 0.5508480668067932,
-      "mean_token_accuracy": 0.7769178003072739,
-      "num_tokens": 17069694.0,
+      "loss": 0.623694658279419,
+      "mean_token_accuracy": 0.7458877414464951,
+      "num_tokens": 8244309.0,
       "step": 221
     },
     {
-      "entropy": 0.5454904437065125,
-      "epoch": 6.96,
-      "grad_norm": 0.024885021150112152,
+      "entropy": 0.6304584518074989,
+      "epoch": 1.4419171405361495,
+      "grad_norm": 0.009999740868806839,
       "learning_rate": 0.0002,
-      "loss": 0.5465965270996094,
-      "mean_token_accuracy": 0.7797489613294601,
-      "num_tokens": 17147632.0,
+      "loss": 0.6301021575927734,
+      "mean_token_accuracy": 0.7443124279379845,
+      "num_tokens": 8281764.0,
       "step": 222
     },
     {
-      "entropy": 0.5398757457733154,
-      "epoch": 6.992,
-      "grad_norm": 0.02908923290669918,
+      "entropy": 0.6228891983628273,
+      "epoch": 1.4484159220146222,
+      "grad_norm": 0.009084641933441162,
       "learning_rate": 0.0002,
-      "loss": 0.5431025624275208,
-      "mean_token_accuracy": 0.779990091919899,
-      "num_tokens": 17227457.0,
+      "loss": 0.623934268951416,
+      "mean_token_accuracy": 0.746678814291954,
+      "num_tokens": 8319204.0,
       "step": 223
     },
     {
-      "entropy": 0.5505363941192627,
-      "epoch": 7.0,
-      "grad_norm": 0.03793865069746971,
+      "entropy": 0.6163553223013878,
+      "epoch": 1.454914703493095,
+      "grad_norm": 0.008567318320274353,
       "learning_rate": 0.0002,
-      "loss": 0.5468761324882507,
-      "mean_token_accuracy": 0.7777085304260254,
-      "num_tokens": 17246839.0,
+      "loss": 0.6207892894744873,
+      "mean_token_accuracy": 0.7486568316817284,
+      "num_tokens": 8356629.0,
       "step": 224
     },
     {
-      "entropy": 0.5342252403497696,
-      "epoch": 7.032,
-      "grad_norm": 0.03273385763168335,
+      "entropy": 0.6088405475020409,
+      "epoch": 1.4614134849715679,
+      "grad_norm": 0.00832637120038271,
       "learning_rate": 0.0002,
-      "loss": 0.5247677564620972,
-      "mean_token_accuracy": 0.7881321907043457,
-      "num_tokens": 17326090.0,
+      "loss": 0.612472414970398,
+      "mean_token_accuracy": 0.7514082416892052,
+      "num_tokens": 8394033.0,
       "step": 225
     },
     {
-      "entropy": 0.5373009145259857,
-      "epoch": 7.064,
-      "grad_norm": 0.025371287018060684,
+      "entropy": 0.614930123090744,
+      "epoch": 1.4679122664500406,
+      "grad_norm": 0.008623549714684486,
       "learning_rate": 0.0002,
-      "loss": 0.5323753952980042,
-      "mean_token_accuracy": 0.7845641523599625,
-      "num_tokens": 17405594.0,
+      "loss": 0.6192766427993774,
+      "mean_token_accuracy": 0.7488296180963516,
+      "num_tokens": 8431411.0,
       "step": 226
     },
     {
-      "entropy": 0.5337796360254288,
-      "epoch": 7.096,
-      "grad_norm": 0.03891294077038765,
+      "entropy": 0.6123254001140594,
+      "epoch": 1.4744110479285135,
+      "grad_norm": 0.008329757489264011,
       "learning_rate": 0.0002,
-      "loss": 0.5395870208740234,
-      "mean_token_accuracy": 0.7812530696392059,
-      "num_tokens": 17484846.0,
+      "loss": 0.614305317401886,
+      "mean_token_accuracy": 0.7501704543828964,
+      "num_tokens": 8468845.0,
       "step": 227
     },
     {
-      "entropy": 0.5363831967115402,
-      "epoch": 7.128,
-      "grad_norm": 0.03025059774518013,
+      "entropy": 0.6353933066129684,
+      "epoch": 1.4809098294069862,
+      "grad_norm": 0.010329898446798325,
       "learning_rate": 0.0002,
-      "loss": 0.538554847240448,
-      "mean_token_accuracy": 0.7826266288757324,
-      "num_tokens": 17563627.0,
+      "loss": 0.6299945712089539,
+      "mean_token_accuracy": 0.7446781396865845,
+      "num_tokens": 8506494.0,
       "step": 228
     },
     {
-      "entropy": 0.5457527041435242,
-      "epoch": 7.16,
-      "grad_norm": 0.027366969734430313,
+      "entropy": 0.6230698823928833,
+      "epoch": 1.487408610885459,
+      "grad_norm": 0.008529371581971645,
       "learning_rate": 0.0002,
-      "loss": 0.5434131622314453,
-      "mean_token_accuracy": 0.7804773300886154,
-      "num_tokens": 17642515.0,
+      "loss": 0.6179029941558838,
+      "mean_token_accuracy": 0.7506388053297997,
+      "num_tokens": 8543863.0,
       "step": 229
     },
     {
-      "entropy": 0.5363115072250366,
-      "epoch": 7.192,
-      "grad_norm": 0.03141327574849129,
+      "entropy": 0.6087015345692635,
+      "epoch": 1.4939073923639317,
+      "grad_norm": 0.008635671809315681,
       "learning_rate": 0.0002,
-      "loss": 0.5331645607948303,
-      "mean_token_accuracy": 0.7852070778608322,
-      "num_tokens": 17721303.0,
+      "loss": 0.6120889186859131,
+      "mean_token_accuracy": 0.7521615102887154,
+      "num_tokens": 8580998.0,
       "step": 230
     },
     {
-      "entropy": 0.5441312342882156,
-      "epoch": 7.224,
-      "grad_norm": 0.02731841430068016,
+      "entropy": 0.6111158281564713,
+      "epoch": 1.5004061738424046,
+      "grad_norm": 0.009943103417754173,
       "learning_rate": 0.0002,
-      "loss": 0.5440965890884399,
-      "mean_token_accuracy": 0.7794661670923233,
-      "num_tokens": 17799159.0,
+      "loss": 0.6142638921737671,
+      "mean_token_accuracy": 0.7509952187538147,
+      "num_tokens": 8618445.0,
       "step": 231
     },
     {
-      "entropy": 0.5447894930839539,
-      "epoch": 7.256,
-      "grad_norm": 0.03376320004463196,
+      "entropy": 0.6176108121871948,
+      "epoch": 1.5069049553208773,
+      "grad_norm": 0.007991382852196693,
       "learning_rate": 0.0002,
-      "loss": 0.543010950088501,
-      "mean_token_accuracy": 0.7805230468511581,
-      "num_tokens": 17877298.0,
+      "loss": 0.619983434677124,
+      "mean_token_accuracy": 0.7502688392996788,
+      "num_tokens": 8655969.0,
       "step": 232
     },
     {
-      "entropy": 0.5362501293420792,
-      "epoch": 7.288,
-      "grad_norm": 0.03332136943936348,
+      "entropy": 0.6207513734698296,
+      "epoch": 1.5134037367993503,
+      "grad_norm": 0.0076866415329277515,
       "learning_rate": 0.0002,
-      "loss": 0.5391461253166199,
-      "mean_token_accuracy": 0.7822219878435135,
-      "num_tokens": 17955795.0,
+      "loss": 0.626140296459198,
+      "mean_token_accuracy": 0.7469011545181274,
+      "num_tokens": 8693748.0,
       "step": 233
     },
     {
-      "entropy": 0.5330929756164551,
-      "epoch": 7.32,
-      "grad_norm": 0.029867252334952354,
+      "entropy": 0.6301305815577507,
+      "epoch": 1.519902518277823,
+      "grad_norm": 0.007992899045348167,
       "learning_rate": 0.0002,
-      "loss": 0.5306733846664429,
-      "mean_token_accuracy": 0.7851523607969284,
-      "num_tokens": 18035708.0,
+      "loss": 0.6301756501197815,
+      "mean_token_accuracy": 0.744783528149128,
+      "num_tokens": 8730930.0,
       "step": 234
     },
     {
-      "entropy": 0.54389688372612,
-      "epoch": 7.352,
-      "grad_norm": 0.03905443847179413,
+      "entropy": 0.6271040365099907,
+      "epoch": 1.5264012997562957,
+      "grad_norm": 0.008258827030658722,
       "learning_rate": 0.0002,
-      "loss": 0.541580080986023,
-      "mean_token_accuracy": 0.7805737406015396,
-      "num_tokens": 18112883.0,
+      "loss": 0.6266668438911438,
+      "mean_token_accuracy": 0.746114693582058,
+      "num_tokens": 8768377.0,
       "step": 235
     },
     {
-      "entropy": 0.5383184850215912,
-      "epoch": 7.384,
-      "grad_norm": 0.02882959507405758,
+      "entropy": 0.6152656748890877,
+      "epoch": 1.5329000812347684,
+      "grad_norm": 0.0098560880869627,
       "learning_rate": 0.0002,
-      "loss": 0.538108766078949,
-      "mean_token_accuracy": 0.7825261801481247,
-      "num_tokens": 18190190.0,
+      "loss": 0.6164472103118896,
+      "mean_token_accuracy": 0.7489713951945305,
+      "num_tokens": 8805779.0,
       "step": 236
     },
     {
-      "entropy": 0.5316631495952606,
-      "epoch": 7.416,
-      "grad_norm": 0.03594264015555382,
+      "entropy": 0.6294426172971725,
+      "epoch": 1.5393988627132411,
+      "grad_norm": 0.008523919619619846,
       "learning_rate": 0.0002,
-      "loss": 0.5299159288406372,
-      "mean_token_accuracy": 0.7848720252513885,
-      "num_tokens": 18269276.0,
+      "loss": 0.6326299905776978,
+      "mean_token_accuracy": 0.742644876241684,
+      "num_tokens": 8843123.0,
       "step": 237
     },
     {
-      "entropy": 0.5474216341972351,
-      "epoch": 7.448,
-      "grad_norm": 0.030706387013196945,
+      "entropy": 0.6143188178539276,
+      "epoch": 1.545897644191714,
+      "grad_norm": 0.007758335676044226,
       "learning_rate": 0.0002,
-      "loss": 0.5446693897247314,
-      "mean_token_accuracy": 0.7793966680765152,
-      "num_tokens": 18346679.0,
+      "loss": 0.6136566996574402,
+      "mean_token_accuracy": 0.7520029693841934,
+      "num_tokens": 8880249.0,
       "step": 238
     },
     {
-      "entropy": 0.5482395142316818,
-      "epoch": 7.48,
-      "grad_norm": 0.03116007335484028,
+      "entropy": 0.627038024365902,
+      "epoch": 1.552396425670187,
+      "grad_norm": 0.00911460816860199,
       "learning_rate": 0.0002,
-      "loss": 0.5429675579071045,
-      "mean_token_accuracy": 0.7807913720607758,
-      "num_tokens": 18425406.0,
+      "loss": 0.6299710869789124,
+      "mean_token_accuracy": 0.7446892932057381,
+      "num_tokens": 8917829.0,
       "step": 239
     },
     {
-      "entropy": 0.5305105894804001,
-      "epoch": 7.5120000000000005,
-      "grad_norm": 0.03157881647348404,
+      "entropy": 0.6183791384100914,
+      "epoch": 1.5588952071486597,
+      "grad_norm": 0.008657646365463734,
       "learning_rate": 0.0002,
-      "loss": 0.5327313542366028,
-      "mean_token_accuracy": 0.7842520177364349,
-      "num_tokens": 18505671.0,
+      "loss": 0.6220093965530396,
+      "mean_token_accuracy": 0.7485999315977097,
+      "num_tokens": 8955243.0,
       "step": 240
     },
     {
-      "entropy": 0.5307354480028152,
-      "epoch": 7.5440000000000005,
-      "grad_norm": 0.028445513918995857,
+      "entropy": 0.6323986351490021,
+      "epoch": 1.5653939886271324,
+      "grad_norm": 0.0085441330447793,
       "learning_rate": 0.0002,
-      "loss": 0.5331681966781616,
-      "mean_token_accuracy": 0.784986600279808,
-      "num_tokens": 18585516.0,
+      "loss": 0.6361583471298218,
+      "mean_token_accuracy": 0.7442645356059074,
+      "num_tokens": 8992960.0,
       "step": 241
     },
     {
-      "entropy": 0.5455832630395889,
-      "epoch": 7.576,
-      "grad_norm": 0.04114306718111038,
+      "entropy": 0.6108334213495255,
+      "epoch": 1.5718927701056051,
+      "grad_norm": 0.007725970819592476,
       "learning_rate": 0.0002,
-      "loss": 0.5481559038162231,
-      "mean_token_accuracy": 0.7785380184650421,
-      "num_tokens": 18663248.0,
+      "loss": 0.6101396083831787,
+      "mean_token_accuracy": 0.7506669163703918,
+      "num_tokens": 9030523.0,
       "step": 242
     },
     {
-      "entropy": 0.5406046807765961,
-      "epoch": 7.608,
-      "grad_norm": 0.033326584845781326,
+      "entropy": 0.6231397688388824,
+      "epoch": 1.5783915515840778,
+      "grad_norm": 0.008767355233430862,
       "learning_rate": 0.0002,
-      "loss": 0.5353219509124756,
-      "mean_token_accuracy": 0.7833963334560394,
-      "num_tokens": 18743695.0,
+      "loss": 0.622745156288147,
+      "mean_token_accuracy": 0.7461344972252846,
+      "num_tokens": 9067759.0,
       "step": 243
     },
     {
-      "entropy": 0.5466004610061646,
-      "epoch": 7.64,
-      "grad_norm": 0.03053242340683937,
+      "entropy": 0.627980038523674,
+      "epoch": 1.5848903330625508,
+      "grad_norm": 0.00843246839940548,
       "learning_rate": 0.0002,
-      "loss": 0.538567066192627,
-      "mean_token_accuracy": 0.7825069427490234,
-      "num_tokens": 18823876.0,
+      "loss": 0.632775604724884,
+      "mean_token_accuracy": 0.7432630062103271,
+      "num_tokens": 9105200.0,
       "step": 244
     },
     {
-      "entropy": 0.5553711205720901,
-      "epoch": 7.672,
-      "grad_norm": 0.03428297117352486,
+      "entropy": 0.6222330331802368,
+      "epoch": 1.5913891145410237,
+      "grad_norm": 0.008817244321107864,
       "learning_rate": 0.0002,
-      "loss": 0.5528771281242371,
-      "mean_token_accuracy": 0.7763674706220627,
-      "num_tokens": 18901081.0,
+      "loss": 0.625479519367218,
+      "mean_token_accuracy": 0.7454850673675537,
+      "num_tokens": 9142588.0,
       "step": 245
     },
     {
-      "entropy": 0.5412455052137375,
-      "epoch": 7.704,
-      "grad_norm": 0.03315138816833496,
+      "entropy": 0.6129688546061516,
+      "epoch": 1.5978878960194964,
+      "grad_norm": 0.009856261312961578,
       "learning_rate": 0.0002,
-      "loss": 0.5472655296325684,
-      "mean_token_accuracy": 0.7775164395570755,
-      "num_tokens": 18979461.0,
+      "loss": 0.6199933886528015,
+      "mean_token_accuracy": 0.7487855926156044,
+      "num_tokens": 9179884.0,
       "step": 246
     },
     {
-      "entropy": 0.5280198305845261,
-      "epoch": 7.736,
-      "grad_norm": 0.035438407212495804,
+      "entropy": 0.6239286884665489,
+      "epoch": 1.6043866774979691,
+      "grad_norm": 0.008906936272978783,
       "learning_rate": 0.0002,
-      "loss": 0.5257879495620728,
-      "mean_token_accuracy": 0.787413626909256,
-      "num_tokens": 19058324.0,
+      "loss": 0.6204976439476013,
+      "mean_token_accuracy": 0.7508985996246338,
+      "num_tokens": 9217103.0,
       "step": 247
     },
     {
-      "entropy": 0.5491799712181091,
-      "epoch": 7.768,
-      "grad_norm": 0.03031354583799839,
+      "entropy": 0.6336326971650124,
+      "epoch": 1.6108854589764419,
+      "grad_norm": 0.011355509981513023,
       "learning_rate": 0.0002,
-      "loss": 0.5499848127365112,
-      "mean_token_accuracy": 0.7769055813550949,
-      "num_tokens": 19136811.0,
+      "loss": 0.6262732744216919,
+      "mean_token_accuracy": 0.7465848848223686,
+      "num_tokens": 9254644.0,
       "step": 248
     },
     {
-      "entropy": 0.5516806840896606,
-      "epoch": 7.8,
-      "grad_norm": 0.03401699662208557,
+      "entropy": 0.6203413605690002,
+      "epoch": 1.6173842404549146,
+      "grad_norm": 0.007735299877822399,
       "learning_rate": 0.0002,
-      "loss": 0.5487878322601318,
-      "mean_token_accuracy": 0.7770615667104721,
-      "num_tokens": 19215642.0,
+      "loss": 0.6212904453277588,
+      "mean_token_accuracy": 0.7472884654998779,
+      "num_tokens": 9291724.0,
       "step": 249
     },
     {
-      "entropy": 0.5209604501724243,
-      "epoch": 7.832,
-      "grad_norm": 0.034633055329322815,
+      "entropy": 0.6144890040159225,
+      "epoch": 1.6238830219333875,
+      "grad_norm": 0.007746159564703703,
       "learning_rate": 0.0002,
-      "loss": 0.5194536447525024,
-      "mean_token_accuracy": 0.7909128665924072,
-      "num_tokens": 19296935.0,
+      "loss": 0.619687557220459,
+      "mean_token_accuracy": 0.7471293359994888,
+      "num_tokens": 9328841.0,
       "step": 250
     },
     {
-      "entropy": 0.5412441790103912,
-      "epoch": 7.864,
-      "grad_norm": 0.03364886716008186,
+      "entropy": 0.6305414438247681,
+      "epoch": 1.6303818034118602,
+      "grad_norm": 0.010060657747089863,
       "learning_rate": 0.0002,
-      "loss": 0.5379568934440613,
-      "mean_token_accuracy": 0.7831831127405167,
-      "num_tokens": 19374745.0,
+      "loss": 0.6419017314910889,
+      "mean_token_accuracy": 0.7386629730463028,
+      "num_tokens": 9366411.0,
       "step": 251
     },
     {
-      "entropy": 0.5517376810312271,
-      "epoch": 7.896,
-      "grad_norm": 0.026521844789385796,
+      "entropy": 0.6339073181152344,
+      "epoch": 1.6368805848903332,
+      "grad_norm": 0.007909824140369892,
       "learning_rate": 0.0002,
-      "loss": 0.5488777756690979,
-      "mean_token_accuracy": 0.7795552015304565,
-      "num_tokens": 19453532.0,
+      "loss": 0.6384083032608032,
+      "mean_token_accuracy": 0.7414835765957832,
+      "num_tokens": 9403421.0,
       "step": 252
     },
     {
-      "entropy": 0.5431555360555649,
-      "epoch": 7.928,
-      "grad_norm": 0.036673326045274734,
+      "entropy": 0.6129219457507133,
+      "epoch": 1.6433793663688059,
+      "grad_norm": 0.007407526019960642,
       "learning_rate": 0.0002,
-      "loss": 0.5410053730010986,
-      "mean_token_accuracy": 0.7819114327430725,
-      "num_tokens": 19532321.0,
+      "loss": 0.6125075817108154,
+      "mean_token_accuracy": 0.752901516854763,
+      "num_tokens": 9440793.0,
       "step": 253
     },
     {
-      "entropy": 0.5400401204824448,
-      "epoch": 7.96,
-      "grad_norm": 0.03132104128599167,
+      "entropy": 0.6255052834749222,
+      "epoch": 1.6498781478472786,
+      "grad_norm": 0.008314264938235283,
       "learning_rate": 0.0002,
-      "loss": 0.5396822690963745,
-      "mean_token_accuracy": 0.7816772311925888,
-      "num_tokens": 19611194.0,
+      "loss": 0.6215559244155884,
+      "mean_token_accuracy": 0.7486165389418602,
+      "num_tokens": 9477882.0,
       "step": 254
     },
     {
-      "entropy": 0.5364492535591125,
-      "epoch": 7.992,
-      "grad_norm": 0.03492264077067375,
+      "entropy": 0.6231755018234253,
+      "epoch": 1.6563769293257513,
+      "grad_norm": 0.0075997961685061455,
       "learning_rate": 0.0002,
-      "loss": 0.5373172163963318,
-      "mean_token_accuracy": 0.7828410267829895,
-      "num_tokens": 19690556.0,
+      "loss": 0.6211115121841431,
+      "mean_token_accuracy": 0.7495426833629608,
+      "num_tokens": 9515444.0,
       "step": 255
     },
     {
-      "entropy": 0.5176465511322021,
-      "epoch": 8.0,
-      "grad_norm": 0.04740741103887558,
+      "entropy": 0.6108886152505875,
+      "epoch": 1.6628757108042242,
+      "grad_norm": 0.007718959823250771,
       "learning_rate": 0.0002,
-      "loss": 0.5141822695732117,
-      "mean_token_accuracy": 0.7935296297073364,
-      "num_tokens": 19710992.0,
+      "loss": 0.6141905784606934,
+      "mean_token_accuracy": 0.7495709583163261,
+      "num_tokens": 9552755.0,
       "step": 256
     },
     {
-      "entropy": 0.5462612360715866,
-      "epoch": 8.032,
-      "grad_norm": 0.033493831753730774,
+      "entropy": 0.6075874269008636,
+      "epoch": 1.669374492282697,
+      "grad_norm": 0.008023767732083797,
       "learning_rate": 0.0002,
-      "loss": 0.5399432182312012,
-      "mean_token_accuracy": 0.7800467163324356,
-      "num_tokens": 19789675.0,
+      "loss": 0.6125887632369995,
+      "mean_token_accuracy": 0.7506251037120819,
+      "num_tokens": 9590261.0,
       "step": 257
     },
     {
-      "entropy": 0.5363253057003021,
-      "epoch": 8.064,
-      "grad_norm": 0.03170062601566315,
+      "entropy": 0.6148701831698418,
+      "epoch": 1.67587327376117,
+      "grad_norm": 0.008702424354851246,
       "learning_rate": 0.0002,
-      "loss": 0.5308842658996582,
-      "mean_token_accuracy": 0.7855557650327682,
-      "num_tokens": 19867187.0,
+      "loss": 0.6215870380401611,
+      "mean_token_accuracy": 0.7478265091776848,
+      "num_tokens": 9627758.0,
       "step": 258
     },
     {
-      "entropy": 0.5275291204452515,
-      "epoch": 8.096,
-      "grad_norm": 0.04521121829748154,
+      "entropy": 0.6036238297820091,
+      "epoch": 1.6823720552396426,
+      "grad_norm": 0.007766771595925093,
       "learning_rate": 0.0002,
-      "loss": 0.5327388644218445,
-      "mean_token_accuracy": 0.7848894149065018,
-      "num_tokens": 19945957.0,
+      "loss": 0.6084332466125488,
+      "mean_token_accuracy": 0.7513664737343788,
+      "num_tokens": 9664761.0,
       "step": 259
     },
     {
-      "entropy": 0.5200865566730499,
-      "epoch": 8.128,
-      "grad_norm": 0.029792284592986107,
+      "entropy": 0.6297216266393661,
+      "epoch": 1.6888708367181153,
+      "grad_norm": 0.009538715705275536,
       "learning_rate": 0.0002,
-      "loss": 0.5194977521896362,
-      "mean_token_accuracy": 0.7896896749734879,
-      "num_tokens": 20026028.0,
+      "loss": 0.6225094795227051,
+      "mean_token_accuracy": 0.7476939931511879,
+      "num_tokens": 9702161.0,
       "step": 260
     },
     {
-      "entropy": 0.5372072905302048,
-      "epoch": 8.16,
-      "grad_norm": 0.03705461695790291,
+      "entropy": 0.6292383894324303,
+      "epoch": 1.695369618196588,
+      "grad_norm": 0.00895692128688097,
       "learning_rate": 0.0002,
-      "loss": 0.5331783890724182,
-      "mean_token_accuracy": 0.7831152677536011,
-      "num_tokens": 20104507.0,
+      "loss": 0.6246432065963745,
+      "mean_token_accuracy": 0.7490787208080292,
+      "num_tokens": 9739248.0,
       "step": 261
     },
     {
-      "entropy": 0.547843337059021,
-      "epoch": 8.192,
-      "grad_norm": 0.04235028475522995,
+      "entropy": 0.6182943060994148,
+      "epoch": 1.7018683996750608,
+      "grad_norm": 0.008069152012467384,
       "learning_rate": 0.0002,
-      "loss": 0.5423978567123413,
-      "mean_token_accuracy": 0.7805410325527191,
-      "num_tokens": 20181982.0,
+      "loss": 0.6146731972694397,
+      "mean_token_accuracy": 0.7493266090750694,
+      "num_tokens": 9776948.0,
       "step": 262
     },
     {
-      "entropy": 0.5300101935863495,
-      "epoch": 8.224,
-      "grad_norm": 0.040369223803281784,
+      "entropy": 0.6093627437949181,
+      "epoch": 1.7083671811535337,
+      "grad_norm": 0.00886097177863121,
       "learning_rate": 0.0002,
-      "loss": 0.5269775390625,
-      "mean_token_accuracy": 0.786443218588829,
-      "num_tokens": 20260502.0,
+      "loss": 0.6172748804092407,
+      "mean_token_accuracy": 0.7474698126316071,
+      "num_tokens": 9814118.0,
       "step": 263
     },
     {
-      "entropy": 0.5322915017604828,
-      "epoch": 8.256,
-      "grad_norm": 0.03680511191487312,
+      "entropy": 0.599872387945652,
+      "epoch": 1.7148659626320066,
+      "grad_norm": 0.009245996363461018,
       "learning_rate": 0.0002,
-      "loss": 0.5325385332107544,
-      "mean_token_accuracy": 0.7842995822429657,
-      "num_tokens": 20338315.0,
+      "loss": 0.6081432700157166,
+      "mean_token_accuracy": 0.7528064101934433,
+      "num_tokens": 9851508.0,
       "step": 264
     },
     {
-      "entropy": 0.5264201015233994,
-      "epoch": 8.288,
-      "grad_norm": 0.033956896513700485,
+      "entropy": 0.6169805154204369,
+      "epoch": 1.7213647441104794,
+      "grad_norm": 0.009072757326066494,
       "learning_rate": 0.0002,
-      "loss": 0.5283188819885254,
-      "mean_token_accuracy": 0.7866097986698151,
-      "num_tokens": 20418550.0,
+      "loss": 0.6274439096450806,
+      "mean_token_accuracy": 0.7459535896778107,
+      "num_tokens": 9889084.0,
       "step": 265
     },
     {
-      "entropy": 0.5397807955741882,
-      "epoch": 8.32,
-      "grad_norm": 0.036973804235458374,
+      "entropy": 0.6127713099122047,
+      "epoch": 1.727863525588952,
+      "grad_norm": 0.0077358693815767765,
       "learning_rate": 0.0002,
-      "loss": 0.5362756848335266,
-      "mean_token_accuracy": 0.7828366905450821,
-      "num_tokens": 20496912.0,
+      "loss": 0.6177385449409485,
+      "mean_token_accuracy": 0.7487060204148293,
+      "num_tokens": 9926415.0,
       "step": 266
     },
     {
-      "entropy": 0.5398755520582199,
-      "epoch": 8.352,
-      "grad_norm": 0.03347233682870865,
+      "entropy": 0.6123405247926712,
+      "epoch": 1.7343623070674248,
+      "grad_norm": 0.00807071104645729,
       "learning_rate": 0.0002,
-      "loss": 0.5393750667572021,
-      "mean_token_accuracy": 0.7830456793308258,
-      "num_tokens": 20575356.0,
+      "loss": 0.6082473397254944,
+      "mean_token_accuracy": 0.7525108605623245,
+      "num_tokens": 9963968.0,
       "step": 267
     },
     {
-      "entropy": 0.5324622094631195,
-      "epoch": 8.384,
-      "grad_norm": 0.03724009171128273,
+      "entropy": 0.616507887840271,
+      "epoch": 1.7408610885458975,
+      "grad_norm": 0.008352074772119522,
       "learning_rate": 0.0002,
-      "loss": 0.5334412455558777,
-      "mean_token_accuracy": 0.7843203842639923,
-      "num_tokens": 20654961.0,
+      "loss": 0.6156948804855347,
+      "mean_token_accuracy": 0.7471591085195541,
+      "num_tokens": 10001164.0,
       "step": 268
     },
     {
-      "entropy": 0.5378891676664352,
-      "epoch": 8.416,
-      "grad_norm": 0.0343894436955452,
+      "entropy": 0.6224628537893295,
+      "epoch": 1.7473598700243704,
+      "grad_norm": 0.00863809883594513,
       "learning_rate": 0.0002,
-      "loss": 0.5370844602584839,
-      "mean_token_accuracy": 0.7837714999914169,
-      "num_tokens": 20734231.0,
+      "loss": 0.6212406158447266,
+      "mean_token_accuracy": 0.7489727661013603,
+      "num_tokens": 10038309.0,
       "step": 269
     },
     {
-      "entropy": 0.5237532556056976,
-      "epoch": 8.448,
-      "grad_norm": 0.03481991961598396,
+      "entropy": 0.6216438189148903,
+      "epoch": 1.7538586515028434,
+      "grad_norm": 0.00882699340581894,
       "learning_rate": 0.0002,
-      "loss": 0.5227078199386597,
-      "mean_token_accuracy": 0.7883624732494354,
-      "num_tokens": 20812680.0,
+      "loss": 0.6236106157302856,
+      "mean_token_accuracy": 0.7466619610786438,
+      "num_tokens": 10075167.0,
       "step": 270
     },
     {
-      "entropy": 0.5372126400470734,
-      "epoch": 8.48,
-      "grad_norm": 0.0352364219725132,
+      "entropy": 0.6243376582860947,
+      "epoch": 1.760357432981316,
+      "grad_norm": 0.008083442226052284,
       "learning_rate": 0.0002,
-      "loss": 0.5328319668769836,
-      "mean_token_accuracy": 0.7841216921806335,
-      "num_tokens": 20890719.0,
+      "loss": 0.6196388006210327,
+      "mean_token_accuracy": 0.7481562867760658,
+      "num_tokens": 10112463.0,
       "step": 271
     },
     {
-      "entropy": 0.537587359547615,
-      "epoch": 8.512,
-      "grad_norm": 0.04216265305876732,
+      "entropy": 0.6375504732131958,
+      "epoch": 1.7668562144597888,
+      "grad_norm": 0.009407855570316315,
       "learning_rate": 0.0002,
-      "loss": 0.536990761756897,
-      "mean_token_accuracy": 0.7831110805273056,
-      "num_tokens": 20969962.0,
+      "loss": 0.6351386308670044,
+      "mean_token_accuracy": 0.7419685274362564,
+      "num_tokens": 10150145.0,
       "step": 272
     },
     {
-      "entropy": 0.5395336449146271,
-      "epoch": 8.544,
-      "grad_norm": 0.03775520622730255,
+      "entropy": 0.6064025089144707,
+      "epoch": 1.7733549959382615,
+      "grad_norm": 0.008066139183938503,
       "learning_rate": 0.0002,
-      "loss": 0.5417099595069885,
-      "mean_token_accuracy": 0.7797977179288864,
-      "num_tokens": 21048464.0,
+      "loss": 0.6098955273628235,
+      "mean_token_accuracy": 0.7515333816409111,
+      "num_tokens": 10187476.0,
       "step": 273
     },
     {
-      "entropy": 0.5484501123428345,
-      "epoch": 8.576,
-      "grad_norm": 0.03163311630487442,
+      "entropy": 0.609002947807312,
+      "epoch": 1.7798537774167342,
+      "grad_norm": 0.0090776477009058,
       "learning_rate": 0.0002,
-      "loss": 0.5443800687789917,
-      "mean_token_accuracy": 0.7803587913513184,
-      "num_tokens": 21125761.0,
+      "loss": 0.6182198524475098,
+      "mean_token_accuracy": 0.749611884355545,
+      "num_tokens": 10224722.0,
       "step": 274
     },
     {
-      "entropy": 0.5378590673208237,
-      "epoch": 8.608,
-      "grad_norm": 0.03819214925169945,
+      "entropy": 0.6023535504937172,
+      "epoch": 1.7863525588952072,
+      "grad_norm": 0.008943675085902214,
       "learning_rate": 0.0002,
-      "loss": 0.5338118672370911,
-      "mean_token_accuracy": 0.7852731049060822,
-      "num_tokens": 21204791.0,
+      "loss": 0.6131588220596313,
+      "mean_token_accuracy": 0.7517130747437477,
+      "num_tokens": 10262008.0,
       "step": 275
     },
     {
-      "entropy": 0.5275268107652664,
-      "epoch": 8.64,
-      "grad_norm": 0.03613068163394928,
+      "entropy": 0.6100214198231697,
+      "epoch": 1.7928513403736799,
+      "grad_norm": 0.00780687527731061,
       "learning_rate": 0.0002,
-      "loss": 0.530475378036499,
-      "mean_token_accuracy": 0.7848686426877975,
-      "num_tokens": 21282818.0,
+      "loss": 0.6113550066947937,
+      "mean_token_accuracy": 0.7519525811076164,
+      "num_tokens": 10299484.0,
       "step": 276
     },
     {
-      "entropy": 0.5271630138158798,
-      "epoch": 8.672,
-      "grad_norm": 0.032881688326597214,
+      "entropy": 0.6407601237297058,
+      "epoch": 1.7993501218521528,
+      "grad_norm": 0.008408503606915474,
       "learning_rate": 0.0002,
-      "loss": 0.5283002257347107,
-      "mean_token_accuracy": 0.7857704162597656,
-      "num_tokens": 21363284.0,
+      "loss": 0.6413571834564209,
+      "mean_token_accuracy": 0.7412156239151955,
+      "num_tokens": 10337204.0,
       "step": 277
     },
     {
-      "entropy": 0.5359350591897964,
-      "epoch": 8.704,
-      "grad_norm": 0.037707071751356125,
+      "entropy": 0.6267746239900589,
+      "epoch": 1.8058489033306255,
+      "grad_norm": 0.00864301435649395,
       "learning_rate": 0.0002,
-      "loss": 0.5357704162597656,
-      "mean_token_accuracy": 0.783053383231163,
-      "num_tokens": 21442878.0,
+      "loss": 0.6232145428657532,
+      "mean_token_accuracy": 0.7470541223883629,
+      "num_tokens": 10374550.0,
       "step": 278
     },
     {
-      "entropy": 0.5460938811302185,
-      "epoch": 8.736,
-      "grad_norm": 0.033162783831357956,
+      "entropy": 0.6293241009116173,
+      "epoch": 1.8123476848090982,
+      "grad_norm": 0.00788764376193285,
       "learning_rate": 0.0002,
-      "loss": 0.5413697957992554,
-      "mean_token_accuracy": 0.7799932807683945,
-      "num_tokens": 21521994.0,
+      "loss": 0.624733567237854,
+      "mean_token_accuracy": 0.7457406893372536,
+      "num_tokens": 10412206.0,
       "step": 279
     },
     {
-      "entropy": 0.5444028526544571,
-      "epoch": 8.768,
-      "grad_norm": 0.03536875918507576,
+      "entropy": 0.6171272769570351,
+      "epoch": 1.818846466287571,
+      "grad_norm": 0.006805213168263435,
       "learning_rate": 0.0002,
-      "loss": 0.5372054576873779,
-      "mean_token_accuracy": 0.7830760926008224,
-      "num_tokens": 21601066.0,
+      "loss": 0.6169811487197876,
+      "mean_token_accuracy": 0.7512501701712608,
+      "num_tokens": 10449535.0,
       "step": 280
     },
     {
-      "entropy": 0.5271487236022949,
-      "epoch": 8.8,
-      "grad_norm": 0.04451414942741394,
+      "entropy": 0.6095528453588486,
+      "epoch": 1.825345247766044,
+      "grad_norm": 0.007181530352681875,
       "learning_rate": 0.0002,
-      "loss": 0.5246833562850952,
-      "mean_token_accuracy": 0.7879171818494797,
-      "num_tokens": 21680654.0,
+      "loss": 0.6091392040252686,
+      "mean_token_accuracy": 0.752260759472847,
+      "num_tokens": 10487049.0,
       "step": 281
     },
     {
-      "entropy": 0.5291329920291901,
-      "epoch": 8.832,
-      "grad_norm": 0.03518937528133392,
+      "entropy": 0.61494529992342,
+      "epoch": 1.8318440292445166,
+      "grad_norm": 0.008495395071804523,
       "learning_rate": 0.0002,
-      "loss": 0.5318595170974731,
-      "mean_token_accuracy": 0.783314660191536,
-      "num_tokens": 21759636.0,
+      "loss": 0.6144075393676758,
+      "mean_token_accuracy": 0.7521442621946335,
+      "num_tokens": 10524737.0,
       "step": 282
     },
     {
-      "entropy": 0.5357151627540588,
-      "epoch": 8.864,
-      "grad_norm": 0.035144057124853134,
+      "entropy": 0.6196796670556068,
+      "epoch": 1.8383428107229896,
+      "grad_norm": 0.00890264380723238,
       "learning_rate": 0.0002,
-      "loss": 0.5353978872299194,
-      "mean_token_accuracy": 0.7822486013174057,
-      "num_tokens": 21837957.0,
+      "loss": 0.6239107847213745,
+      "mean_token_accuracy": 0.7463233023881912,
+      "num_tokens": 10562188.0,
       "step": 283
     },
     {
-      "entropy": 0.5227695554494858,
-      "epoch": 8.896,
-      "grad_norm": 0.03644225001335144,
+      "entropy": 0.5968083441257477,
+      "epoch": 1.8448415922014623,
+      "grad_norm": 0.007078610826283693,
       "learning_rate": 0.0002,
-      "loss": 0.5201877951622009,
-      "mean_token_accuracy": 0.789432018995285,
-      "num_tokens": 21918900.0,
+      "loss": 0.5999591946601868,
+      "mean_token_accuracy": 0.7574170976877213,
+      "num_tokens": 10599566.0,
       "step": 284
     },
     {
-      "entropy": 0.5280223190784454,
-      "epoch": 8.928,
-      "grad_norm": 0.03295109048485756,
+      "entropy": 0.6290047243237495,
+      "epoch": 1.851340373679935,
+      "grad_norm": 0.0070074331015348434,
       "learning_rate": 0.0002,
-      "loss": 0.527430534362793,
-      "mean_token_accuracy": 0.7872951477766037,
-      "num_tokens": 21997073.0,
+      "loss": 0.6321331262588501,
+      "mean_token_accuracy": 0.7435007244348526,
+      "num_tokens": 10637008.0,
       "step": 285
     },
     {
-      "entropy": 0.5388615727424622,
-      "epoch": 8.96,
-      "grad_norm": 0.04273361712694168,
+      "entropy": 0.6198180019855499,
+      "epoch": 1.8578391551584077,
+      "grad_norm": 0.0070502436719834805,
       "learning_rate": 0.0002,
-      "loss": 0.53923100233078,
-      "mean_token_accuracy": 0.7823034077882767,
-      "num_tokens": 22077250.0,
+      "loss": 0.6187317967414856,
+      "mean_token_accuracy": 0.7489558085799217,
+      "num_tokens": 10674608.0,
       "step": 286
     },
     {
-      "entropy": 0.5351655334234238,
-      "epoch": 8.992,
-      "grad_norm": 0.03406304493546486,
+      "entropy": 0.6114784702658653,
+      "epoch": 1.8643379366368806,
+      "grad_norm": 0.009826047345995903,
       "learning_rate": 0.0002,
-      "loss": 0.5302039384841919,
-      "mean_token_accuracy": 0.7850055992603302,
-      "num_tokens": 22154821.0,
+      "loss": 0.6103346347808838,
+      "mean_token_accuracy": 0.7504026591777802,
+      "num_tokens": 10711947.0,
       "step": 287
     },
     {
-      "entropy": 0.5334904193878174,
-      "epoch": 9.0,
-      "grad_norm": 0.05444731563329697,
+      "entropy": 0.6145703047513962,
+      "epoch": 1.8708367181153533,
+      "grad_norm": 0.008549013175070286,
       "learning_rate": 0.0002,
-      "loss": 0.530449390411377,
-      "mean_token_accuracy": 0.7859096527099609,
-      "num_tokens": 22174693.0,
+      "loss": 0.6118488311767578,
+      "mean_token_accuracy": 0.7518711537122726,
+      "num_tokens": 10749467.0,
       "step": 288
     },
     {
-      "entropy": 0.5300431996583939,
-      "epoch": 9.032,
-      "grad_norm": 0.056497130542993546,
+      "entropy": 0.6132736206054688,
+      "epoch": 1.8773354995938263,
+      "grad_norm": 0.00803026370704174,
       "learning_rate": 0.0002,
-      "loss": 0.527766227722168,
-      "mean_token_accuracy": 0.785633996129036,
-      "num_tokens": 22251556.0,
+      "loss": 0.6166672706604004,
+      "mean_token_accuracy": 0.749387837946415,
+      "num_tokens": 10786750.0,
       "step": 289
     },
     {
-      "entropy": 0.5187714546918869,
-      "epoch": 9.064,
-      "grad_norm": 0.045192088931798935,
+      "entropy": 0.6266130581498146,
+      "epoch": 1.883834281072299,
+      "grad_norm": 0.008134995587170124,
       "learning_rate": 0.0002,
-      "loss": 0.5194220542907715,
-      "mean_token_accuracy": 0.7905025631189346,
-      "num_tokens": 22330815.0,
+      "loss": 0.6310341358184814,
+      "mean_token_accuracy": 0.743694357573986,
+      "num_tokens": 10823914.0,
       "step": 290
     },
     {
-      "entropy": 0.5319517701864243,
-      "epoch": 9.096,
-      "grad_norm": 0.04722893610596657,
+      "entropy": 0.6153042837977409,
+      "epoch": 1.8903330625507717,
+      "grad_norm": 0.009320992045104504,
       "learning_rate": 0.0002,
-      "loss": 0.5339574813842773,
-      "mean_token_accuracy": 0.7833489924669266,
-      "num_tokens": 22407165.0,
+      "loss": 0.6227383613586426,
+      "mean_token_accuracy": 0.7482397258281708,
+      "num_tokens": 10861283.0,
       "step": 291
     },
     {
-      "entropy": 0.5170899480581284,
-      "epoch": 9.128,
-      "grad_norm": 0.052741702646017075,
+      "entropy": 0.6265768557786942,
+      "epoch": 1.8968318440292444,
+      "grad_norm": 0.009208294562995434,
       "learning_rate": 0.0002,
-      "loss": 0.5152403116226196,
-      "mean_token_accuracy": 0.7910004258155823,
-      "num_tokens": 22486660.0,
+      "loss": 0.6274023056030273,
+      "mean_token_accuracy": 0.7459790334105492,
+      "num_tokens": 10898650.0,
       "step": 292
     },
     {
-      "entropy": 0.5402307510375977,
-      "epoch": 9.16,
-      "grad_norm": 0.04082133248448372,
+      "entropy": 0.6209842190146446,
+      "epoch": 1.9033306255077171,
+      "grad_norm": 0.007123781833797693,
       "learning_rate": 0.0002,
-      "loss": 0.5383718013763428,
-      "mean_token_accuracy": 0.7814162373542786,
-      "num_tokens": 22564897.0,
+      "loss": 0.6167253255844116,
+      "mean_token_accuracy": 0.7506068870425224,
+      "num_tokens": 10936405.0,
       "step": 293
     },
     {
-      "entropy": 0.5201883316040039,
-      "epoch": 9.192,
-      "grad_norm": 0.03840792551636696,
+      "entropy": 0.6177584528923035,
+      "epoch": 1.90982940698619,
+      "grad_norm": 0.0075668166391551495,
       "learning_rate": 0.0002,
-      "loss": 0.5196265578269958,
-      "mean_token_accuracy": 0.7899857312440872,
-      "num_tokens": 22644583.0,
+      "loss": 0.6163274049758911,
+      "mean_token_accuracy": 0.7489276751875877,
+      "num_tokens": 10973923.0,
       "step": 294
     },
     {
-      "entropy": 0.5268342941999435,
-      "epoch": 9.224,
-      "grad_norm": 0.04609312862157822,
+      "entropy": 0.6218736097216606,
+      "epoch": 1.916328188464663,
+      "grad_norm": 0.0085530336946249,
       "learning_rate": 0.0002,
-      "loss": 0.5226351022720337,
-      "mean_token_accuracy": 0.7885625064373016,
-      "num_tokens": 22723224.0,
+      "loss": 0.6214455366134644,
+      "mean_token_accuracy": 0.747297577559948,
+      "num_tokens": 11011183.0,
       "step": 295
     },
     {
-      "entropy": 0.5258665084838867,
-      "epoch": 9.256,
-      "grad_norm": 0.03400503844022751,
+      "entropy": 0.6203776001930237,
+      "epoch": 1.9228269699431357,
+      "grad_norm": 0.008531006053090096,
       "learning_rate": 0.0002,
-      "loss": 0.524506151676178,
-      "mean_token_accuracy": 0.7877257317304611,
-      "num_tokens": 22803901.0,
+      "loss": 0.6233081817626953,
+      "mean_token_accuracy": 0.7472613751888275,
+      "num_tokens": 11048409.0,
       "step": 296
     },
     {
-      "entropy": 0.5259871706366539,
-      "epoch": 9.288,
-      "grad_norm": 0.04140326753258705,
+      "entropy": 0.6170744299888611,
+      "epoch": 1.9293257514216084,
+      "grad_norm": 0.006794555112719536,
       "learning_rate": 0.0002,
-      "loss": 0.523682713508606,
-      "mean_token_accuracy": 0.7880030572414398,
-      "num_tokens": 22883095.0,
+      "loss": 0.6177318096160889,
+      "mean_token_accuracy": 0.7481219694018364,
+      "num_tokens": 11085854.0,
       "step": 297
     },
     {
-      "entropy": 0.5212158113718033,
-      "epoch": 9.32,
-      "grad_norm": 0.035439494997262955,
+      "entropy": 0.6205914691090584,
+      "epoch": 1.9358245329000812,
+      "grad_norm": 0.00894018355756998,
       "learning_rate": 0.0002,
-      "loss": 0.5157997608184814,
-      "mean_token_accuracy": 0.791080117225647,
-      "num_tokens": 22963722.0,
+      "loss": 0.6211514472961426,
+      "mean_token_accuracy": 0.7470626905560493,
+      "num_tokens": 11122678.0,
       "step": 298
     },
     {
-      "entropy": 0.5116661787033081,
-      "epoch": 9.352,
-      "grad_norm": 0.04255674034357071,
+      "entropy": 0.6179786175489426,
+      "epoch": 1.9423233143785539,
+      "grad_norm": 0.008218267932534218,
       "learning_rate": 0.0002,
-      "loss": 0.5151242613792419,
-      "mean_token_accuracy": 0.7906815558671951,
-      "num_tokens": 23044372.0,
+      "loss": 0.6211084723472595,
+      "mean_token_accuracy": 0.7481166496872902,
+      "num_tokens": 11160207.0,
       "step": 299
     },
     {
-      "entropy": 0.5317155867815018,
-      "epoch": 9.384,
-      "grad_norm": 0.03571729734539986,
+      "entropy": 0.6241402998566628,
+      "epoch": 1.9488220958570268,
+      "grad_norm": 0.008334561251103878,
       "learning_rate": 0.0002,
-      "loss": 0.5331252813339233,
-      "mean_token_accuracy": 0.784525066614151,
-      "num_tokens": 23121816.0,
+      "loss": 0.6270841360092163,
+      "mean_token_accuracy": 0.7478617802262306,
+      "num_tokens": 11197850.0,
       "step": 300
     },
     {
-      "entropy": 0.5284074693918228,
-      "epoch": 9.416,
-      "grad_norm": 0.038814280182123184,
+      "entropy": 0.6133561357855797,
+      "epoch": 1.9553208773354998,
+      "grad_norm": 0.0071178278885781765,
       "learning_rate": 0.0002,
-      "loss": 0.5249905586242676,
-      "mean_token_accuracy": 0.7884200662374496,
-      "num_tokens": 23200583.0,
+      "loss": 0.6169086694717407,
+      "mean_token_accuracy": 0.7505014687776566,
+      "num_tokens": 11235129.0,
       "step": 301
     },
     {
-      "entropy": 0.5218486413359642,
-      "epoch": 9.448,
-      "grad_norm": 0.03548124432563782,
+      "entropy": 0.6112445220351219,
+      "epoch": 1.9618196588139725,
+      "grad_norm": 0.00821610540151596,
       "learning_rate": 0.0002,
-      "loss": 0.5189951658248901,
-      "mean_token_accuracy": 0.7910516560077667,
-      "num_tokens": 23279098.0,
+      "loss": 0.6131574511528015,
+      "mean_token_accuracy": 0.7517394497990608,
+      "num_tokens": 11272140.0,
       "step": 302
     },
     {
-      "entropy": 0.5266124904155731,
-      "epoch": 9.48,
-      "grad_norm": 0.04210549592971802,
+      "entropy": 0.6255387738347054,
+      "epoch": 1.9683184402924452,
+      "grad_norm": 0.009108194150030613,
       "learning_rate": 0.0002,
-      "loss": 0.5285375118255615,
-      "mean_token_accuracy": 0.7852902412414551,
-      "num_tokens": 23357708.0,
+      "loss": 0.6270141005516052,
+      "mean_token_accuracy": 0.7442326322197914,
+      "num_tokens": 11309074.0,
       "step": 303
     },
     {
-      "entropy": 0.5390820801258087,
-      "epoch": 9.512,
-      "grad_norm": 0.04019097238779068,
+      "entropy": 0.6138653978705406,
+      "epoch": 1.974817221770918,
+      "grad_norm": 0.007860496640205383,
       "learning_rate": 0.0002,
-      "loss": 0.5387204885482788,
-      "mean_token_accuracy": 0.7813313454389572,
-      "num_tokens": 23435254.0,
+      "loss": 0.6152341961860657,
+      "mean_token_accuracy": 0.7488235086202621,
+      "num_tokens": 11346694.0,
       "step": 304
     },
     {
-      "entropy": 0.5226802378892899,
-      "epoch": 9.544,
-      "grad_norm": 0.045527562499046326,
+      "entropy": 0.6311385780572891,
+      "epoch": 1.9813160032493906,
+      "grad_norm": 0.008101027458906174,
       "learning_rate": 0.0002,
-      "loss": 0.5184016227722168,
-      "mean_token_accuracy": 0.7905609607696533,
-      "num_tokens": 23514397.0,
+      "loss": 0.6344274282455444,
+      "mean_token_accuracy": 0.7414661273360252,
+      "num_tokens": 11384229.0,
       "step": 305
     },
     {
-      "entropy": 0.527540773153305,
-      "epoch": 9.576,
-      "grad_norm": 0.04471491649746895,
+      "entropy": 0.6193428635597229,
+      "epoch": 1.9878147847278635,
+      "grad_norm": 0.007910770364105701,
       "learning_rate": 0.0002,
-      "loss": 0.5262195467948914,
-      "mean_token_accuracy": 0.7857372760772705,
-      "num_tokens": 23593391.0,
+      "loss": 0.6175734400749207,
+      "mean_token_accuracy": 0.7502587288618088,
+      "num_tokens": 11421788.0,
       "step": 306
     },
     {
-      "entropy": 0.5327746719121933,
-      "epoch": 9.608,
-      "grad_norm": 0.03983209654688835,
+      "entropy": 0.6222855523228645,
+      "epoch": 1.9943135662063363,
+      "grad_norm": 0.008522240445017815,
       "learning_rate": 0.0002,
-      "loss": 0.5309550762176514,
-      "mean_token_accuracy": 0.7848490923643112,
-      "num_tokens": 23671238.0,
+      "loss": 0.623339831829071,
+      "mean_token_accuracy": 0.7481605112552643,
+      "num_tokens": 11459154.0,
       "step": 307
     },
     {
-      "entropy": 0.5221132040023804,
-      "epoch": 9.64,
-      "grad_norm": 0.0506560355424881,
+      "entropy": 0.6278953552246094,
+      "epoch": 2.0,
+      "grad_norm": 0.008388028480112553,
       "learning_rate": 0.0002,
-      "loss": 0.5292017459869385,
-      "mean_token_accuracy": 0.7862422913312912,
-      "num_tokens": 23750415.0,
+      "loss": 0.6375839710235596,
+      "mean_token_accuracy": 0.7482074073382786,
+      "num_tokens": 11475381.0,
       "step": 308
     },
     {
-      "entropy": 0.5316873043775558,
-      "epoch": 9.672,
-      "grad_norm": 0.04567805677652359,
+      "entropy": 0.613355465233326,
+      "epoch": 2.0064987814784727,
+      "grad_norm": 0.00827846396714449,
       "learning_rate": 0.0002,
-      "loss": 0.5264989137649536,
-      "mean_token_accuracy": 0.7868478447198868,
-      "num_tokens": 23829328.0,
+      "loss": 0.6111053228378296,
+      "mean_token_accuracy": 0.7524029165506363,
+      "num_tokens": 11512446.0,
       "step": 309
     },
     {
-      "entropy": 0.5508539825677872,
-      "epoch": 9.704,
-      "grad_norm": 0.054871875792741776,
+      "entropy": 0.6208668798208237,
+      "epoch": 2.0129975629569454,
+      "grad_norm": 0.008557752706110477,
       "learning_rate": 0.0002,
-      "loss": 0.5446195602416992,
-      "mean_token_accuracy": 0.7784004658460617,
-      "num_tokens": 23906430.0,
+      "loss": 0.6219561100006104,
+      "mean_token_accuracy": 0.7458884716033936,
+      "num_tokens": 11549773.0,
       "step": 310
     },
     {
-      "entropy": 0.5284042209386826,
-      "epoch": 9.736,
-      "grad_norm": 0.05515004321932793,
+      "entropy": 0.6194536536931992,
+      "epoch": 2.019496344435418,
+      "grad_norm": 0.009552874602377415,
       "learning_rate": 0.0002,
-      "loss": 0.531845211982727,
-      "mean_token_accuracy": 0.7849012762308121,
-      "num_tokens": 23985699.0,
+      "loss": 0.624874472618103,
+      "mean_token_accuracy": 0.7470095753669739,
+      "num_tokens": 11587104.0,
       "step": 311
     },
     {
-      "entropy": 0.5239458531141281,
-      "epoch": 9.768,
-      "grad_norm": 0.06098300218582153,
+      "entropy": 0.6137440055608749,
+      "epoch": 2.0259951259138913,
+      "grad_norm": 0.008704678155481815,
       "learning_rate": 0.0002,
-      "loss": 0.5323519706726074,
-      "mean_token_accuracy": 0.7858945429325104,
-      "num_tokens": 24065076.0,
+      "loss": 0.6167584657669067,
+      "mean_token_accuracy": 0.7494626566767693,
+      "num_tokens": 11624387.0,
       "step": 312
     },
     {
-      "entropy": 0.5360407829284668,
-      "epoch": 9.8,
-      "grad_norm": 0.0424746535718441,
+      "entropy": 0.6213903650641441,
+      "epoch": 2.032493907392364,
+      "grad_norm": 0.007905784994363785,
       "learning_rate": 0.0002,
-      "loss": 0.5293824672698975,
-      "mean_token_accuracy": 0.7857670187950134,
-      "num_tokens": 24144078.0,
+      "loss": 0.622920036315918,
+      "mean_token_accuracy": 0.7466430589556694,
+      "num_tokens": 11661928.0,
       "step": 313
     },
     {
-      "entropy": 0.5299330949783325,
-      "epoch": 9.832,
-      "grad_norm": 0.05260491743683815,
+      "entropy": 0.608438141644001,
+      "epoch": 2.0389926888708367,
+      "grad_norm": 0.008782708086073399,
       "learning_rate": 0.0002,
-      "loss": 0.5241835117340088,
-      "mean_token_accuracy": 0.7872831225395203,
-      "num_tokens": 24224502.0,
+      "loss": 0.6084336638450623,
+      "mean_token_accuracy": 0.7526147440075874,
+      "num_tokens": 11699518.0,
       "step": 314
     },
     {
-      "entropy": 0.5187924355268478,
-      "epoch": 9.864,
-      "grad_norm": 0.07041396200656891,
+      "entropy": 0.6130177453160286,
+      "epoch": 2.0454914703493094,
+      "grad_norm": 0.009063824079930782,
       "learning_rate": 0.0002,
-      "loss": 0.5201770067214966,
-      "mean_token_accuracy": 0.7910154610872269,
-      "num_tokens": 24304691.0,
+      "loss": 0.6168864965438843,
+      "mean_token_accuracy": 0.7494284063577652,
+      "num_tokens": 11737013.0,
       "step": 315
     },
     {
-      "entropy": 0.5294213145971298,
-      "epoch": 9.896,
-      "grad_norm": 0.04269389435648918,
+      "entropy": 0.6137471869587898,
+      "epoch": 2.051990251827782,
+      "grad_norm": 0.008271057158708572,
       "learning_rate": 0.0002,
-      "loss": 0.5291866660118103,
-      "mean_token_accuracy": 0.7848200798034668,
-      "num_tokens": 24383384.0,
+      "loss": 0.6143357753753662,
+      "mean_token_accuracy": 0.7498123943805695,
+      "num_tokens": 11774346.0,
       "step": 316
     },
     {
-      "entropy": 0.5307356864213943,
-      "epoch": 9.928,
-      "grad_norm": 0.071862131357193,
+      "entropy": 0.6195183843374252,
+      "epoch": 2.058489033306255,
+      "grad_norm": 0.007240446750074625,
       "learning_rate": 0.0002,
-      "loss": 0.5289138555526733,
-      "mean_token_accuracy": 0.7856195569038391,
-      "num_tokens": 24461247.0,
+      "loss": 0.6194902062416077,
+      "mean_token_accuracy": 0.7464936003088951,
+      "num_tokens": 11811650.0,
       "step": 317
     },
     {
-      "entropy": 0.526089146733284,
-      "epoch": 9.96,
-      "grad_norm": 0.06460758298635483,
+      "entropy": 0.6273763924837112,
+      "epoch": 2.064987814784728,
+      "grad_norm": 0.009513584896922112,
       "learning_rate": 0.0002,
-      "loss": 0.5239361524581909,
-      "mean_token_accuracy": 0.7876095771789551,
-      "num_tokens": 24540321.0,
+      "loss": 0.6216360330581665,
+      "mean_token_accuracy": 0.7472178563475609,
+      "num_tokens": 11849227.0,
       "step": 318
     },
     {
-      "entropy": 0.5216004997491837,
-      "epoch": 9.992,
-      "grad_norm": 0.05195415019989014,
+      "entropy": 0.6146251261234283,
+      "epoch": 2.0714865962632008,
+      "grad_norm": 0.007890078239142895,
       "learning_rate": 0.0002,
-      "loss": 0.5229140520095825,
-      "mean_token_accuracy": 0.7891107648611069,
-      "num_tokens": 24618023.0,
+      "loss": 0.6166911721229553,
+      "mean_token_accuracy": 0.7501556724309921,
+      "num_tokens": 11886732.0,
       "step": 319
     },
     {
-      "entropy": 0.5302079319953918,
-      "epoch": 10.0,
-      "grad_norm": 0.0679510235786438,
+      "entropy": 0.5975739285349846,
+      "epoch": 2.0779853777416735,
+      "grad_norm": 0.009085185825824738,
       "learning_rate": 0.0002,
-      "loss": 0.51785808801651,
-      "mean_token_accuracy": 0.7922472953796387,
-      "num_tokens": 24637886.0,
+      "loss": 0.6020256280899048,
+      "mean_token_accuracy": 0.7525540143251419,
+      "num_tokens": 11923954.0,
       "step": 320
+    },
+    {
+      "entropy": 0.6165630891919136,
+      "epoch": 2.084484159220146,
+      "grad_norm": 0.006837225519120693,
+      "learning_rate": 0.0002,
+      "loss": 0.6179453134536743,
+      "mean_token_accuracy": 0.7491353005170822,
+      "num_tokens": 11961305.0,
+      "step": 321
+    },
+    {
+      "entropy": 0.6221051812171936,
+      "epoch": 2.090982940698619,
+      "grad_norm": 0.007578800432384014,
+      "learning_rate": 0.0002,
+      "loss": 0.6233739852905273,
+      "mean_token_accuracy": 0.7461354285478592,
+      "num_tokens": 11998561.0,
+      "step": 322
+    },
+    {
+      "entropy": 0.6175716295838356,
+      "epoch": 2.0974817221770916,
+      "grad_norm": 0.008444663137197495,
+      "learning_rate": 0.0002,
+      "loss": 0.6204410791397095,
+      "mean_token_accuracy": 0.7468695417046547,
+      "num_tokens": 12035916.0,
+      "step": 323
+    },
+    {
+      "entropy": 0.6165966615080833,
+      "epoch": 2.1039805036555648,
+      "grad_norm": 0.008185302838683128,
+      "learning_rate": 0.0002,
+      "loss": 0.6171708106994629,
+      "mean_token_accuracy": 0.7481535449624062,
+      "num_tokens": 12073704.0,
+      "step": 324
+    },
+    {
+      "entropy": 0.6210658326745033,
+      "epoch": 2.1104792851340375,
+      "grad_norm": 0.0074119968339800835,
+      "learning_rate": 0.0002,
+      "loss": 0.6183840036392212,
+      "mean_token_accuracy": 0.7473175972700119,
+      "num_tokens": 12111357.0,
+      "step": 325
+    },
+    {
+      "entropy": 0.617885060608387,
+      "epoch": 2.11697806661251,
+      "grad_norm": 0.007906505838036537,
+      "learning_rate": 0.0002,
+      "loss": 0.6161532998085022,
+      "mean_token_accuracy": 0.7496127933263779,
+      "num_tokens": 12148879.0,
+      "step": 326
+    },
+    {
+      "entropy": 0.6197918429970741,
+      "epoch": 2.123476848090983,
+      "grad_norm": 0.007352263201028109,
+      "learning_rate": 0.0002,
+      "loss": 0.6218726634979248,
+      "mean_token_accuracy": 0.7473544403910637,
+      "num_tokens": 12186546.0,
+      "step": 327
+    },
+    {
+      "entropy": 0.6142667904496193,
+      "epoch": 2.1299756295694556,
+      "grad_norm": 0.008285868912935257,
+      "learning_rate": 0.0002,
+      "loss": 0.6169737577438354,
+      "mean_token_accuracy": 0.7486495152115822,
+      "num_tokens": 12223727.0,
+      "step": 328
+    },
+    {
+      "entropy": 0.6135031282901764,
+      "epoch": 2.1364744110479283,
+      "grad_norm": 0.00797727145254612,
+      "learning_rate": 0.0002,
+      "loss": 0.6149879097938538,
+      "mean_token_accuracy": 0.7501881718635559,
+      "num_tokens": 12261432.0,
+      "step": 329
+    },
+    {
+      "entropy": 0.614725299179554,
+      "epoch": 2.1429731925264015,
+      "grad_norm": 0.0077530802227556705,
+      "learning_rate": 0.0002,
+      "loss": 0.6197780966758728,
+      "mean_token_accuracy": 0.7476382106542587,
+      "num_tokens": 12298773.0,
+      "step": 330
+    },
+    {
+      "entropy": 0.6204890534281731,
+      "epoch": 2.149471974004874,
+      "grad_norm": 0.007455397862941027,
+      "learning_rate": 0.0002,
+      "loss": 0.6217861175537109,
+      "mean_token_accuracy": 0.7472267374396324,
+      "num_tokens": 12335838.0,
+      "step": 331
+    },
+    {
+      "entropy": 0.6141664981842041,
+      "epoch": 2.155970755483347,
+      "grad_norm": 0.007735258899629116,
+      "learning_rate": 0.0002,
+      "loss": 0.619376540184021,
+      "mean_token_accuracy": 0.7473561838269234,
+      "num_tokens": 12373171.0,
+      "step": 332
+    },
+    {
+      "entropy": 0.6120809391140938,
+      "epoch": 2.1624695369618196,
+      "grad_norm": 0.008401036262512207,
+      "learning_rate": 0.0002,
+      "loss": 0.6154431700706482,
+      "mean_token_accuracy": 0.7471263483166695,
+      "num_tokens": 12410549.0,
+      "step": 333
+    },
+    {
+      "entropy": 0.6145179942250252,
+      "epoch": 2.1689683184402924,
+      "grad_norm": 0.00904724933207035,
+      "learning_rate": 0.0002,
+      "loss": 0.6087259650230408,
+      "mean_token_accuracy": 0.7501003369688988,
+      "num_tokens": 12447788.0,
+      "step": 334
+    },
+    {
+      "entropy": 0.6214614436030388,
+      "epoch": 2.175467099918765,
+      "grad_norm": 0.007604485843330622,
+      "learning_rate": 0.0002,
+      "loss": 0.6169778108596802,
+      "mean_token_accuracy": 0.7492403090000153,
+      "num_tokens": 12485368.0,
+      "step": 335
+    },
+    {
+      "entropy": 0.6146392300724983,
+      "epoch": 2.181965881397238,
+      "grad_norm": 0.008299093693494797,
+      "learning_rate": 0.0002,
+      "loss": 0.6154477596282959,
+      "mean_token_accuracy": 0.7474880516529083,
+      "num_tokens": 12522590.0,
+      "step": 336
+    },
+    {
+      "entropy": 0.6179040372371674,
+      "epoch": 2.188464662875711,
+      "grad_norm": 0.009656457230448723,
+      "learning_rate": 0.0002,
+      "loss": 0.6244622468948364,
+      "mean_token_accuracy": 0.74583899974823,
+      "num_tokens": 12559957.0,
+      "step": 337
+    },
+    {
+      "entropy": 0.6236468479037285,
+      "epoch": 2.1949634443541837,
+      "grad_norm": 0.008342688903212547,
+      "learning_rate": 0.0002,
+      "loss": 0.6273738145828247,
+      "mean_token_accuracy": 0.7441191673278809,
+      "num_tokens": 12597080.0,
+      "step": 338
+    },
+    {
+      "entropy": 0.6155964732170105,
+      "epoch": 2.2014622258326564,
+      "grad_norm": 0.008706323802471161,
+      "learning_rate": 0.0002,
+      "loss": 0.6139167547225952,
+      "mean_token_accuracy": 0.7505380213260651,
+      "num_tokens": 12634470.0,
+      "step": 339
+    },
+    {
+      "entropy": 0.6147474870085716,
+      "epoch": 2.207961007311129,
+      "grad_norm": 0.009130026213824749,
+      "learning_rate": 0.0002,
+      "loss": 0.6147286891937256,
+      "mean_token_accuracy": 0.7490043491125107,
+      "num_tokens": 12671909.0,
+      "step": 340
+    },
+    {
+      "entropy": 0.6133714243769646,
+      "epoch": 2.214459788789602,
+      "grad_norm": 0.008477689698338509,
+      "learning_rate": 0.0002,
+      "loss": 0.6187378168106079,
+      "mean_token_accuracy": 0.7469930872321129,
+      "num_tokens": 12708771.0,
+      "step": 341
+    },
+    {
+      "entropy": 0.6298847869038582,
+      "epoch": 2.2209585702680745,
+      "grad_norm": 0.009632086381316185,
+      "learning_rate": 0.0002,
+      "loss": 0.6292200684547424,
+      "mean_token_accuracy": 0.7443962469696999,
+      "num_tokens": 12746208.0,
+      "step": 342
+    },
+    {
+      "entropy": 0.6298182159662247,
+      "epoch": 2.2274573517465477,
+      "grad_norm": 0.007817580364644527,
+      "learning_rate": 0.0002,
+      "loss": 0.6289129853248596,
+      "mean_token_accuracy": 0.7439332380890846,
+      "num_tokens": 12783644.0,
+      "step": 343
+    },
+    {
+      "entropy": 0.6162709593772888,
+      "epoch": 2.2339561332250204,
+      "grad_norm": 0.009011663496494293,
+      "learning_rate": 0.0002,
+      "loss": 0.6142280101776123,
+      "mean_token_accuracy": 0.7506413981318474,
+      "num_tokens": 12821007.0,
+      "step": 344
+    },
+    {
+      "entropy": 0.6286559477448463,
+      "epoch": 2.240454914703493,
+      "grad_norm": 0.009188450872898102,
+      "learning_rate": 0.0002,
+      "loss": 0.6286423206329346,
+      "mean_token_accuracy": 0.7440641894936562,
+      "num_tokens": 12858516.0,
+      "step": 345
+    },
+    {
+      "entropy": 0.6245385706424713,
+      "epoch": 2.246953696181966,
+      "grad_norm": 0.009156906045973301,
+      "learning_rate": 0.0002,
+      "loss": 0.6278682947158813,
+      "mean_token_accuracy": 0.7443269118666649,
+      "num_tokens": 12895999.0,
+      "step": 346
+    },
+    {
+      "entropy": 0.6157897710800171,
+      "epoch": 2.2534524776604385,
+      "grad_norm": 0.007890370674431324,
+      "learning_rate": 0.0002,
+      "loss": 0.6214807033538818,
+      "mean_token_accuracy": 0.7472835406661034,
+      "num_tokens": 12933354.0,
+      "step": 347
+    },
+    {
+      "entropy": 0.6123293116688728,
+      "epoch": 2.2599512591389113,
+      "grad_norm": 0.007912621833384037,
+      "learning_rate": 0.0002,
+      "loss": 0.6164427995681763,
+      "mean_token_accuracy": 0.7487083747982979,
+      "num_tokens": 12970559.0,
+      "step": 348
+    },
+    {
+      "entropy": 0.6260915771126747,
+      "epoch": 2.2664500406173844,
+      "grad_norm": 0.008234160020947456,
+      "learning_rate": 0.0002,
+      "loss": 0.6222859621047974,
+      "mean_token_accuracy": 0.7458368241786957,
+      "num_tokens": 13008283.0,
+      "step": 349
+    },
+    {
+      "entropy": 0.6248592287302017,
+      "epoch": 2.272948822095857,
+      "grad_norm": 0.009832987561821938,
+      "learning_rate": 0.0002,
+      "loss": 0.6199322938919067,
+      "mean_token_accuracy": 0.7477703094482422,
+      "num_tokens": 13045527.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.6275637075304985,
+      "epoch": 2.27944760357433,
+      "grad_norm": 0.008727029897272587,
+      "learning_rate": 0.0002,
+      "loss": 0.632993221282959,
+      "mean_token_accuracy": 0.7412514090538025,
+      "num_tokens": 13082974.0,
+      "step": 351
+    },
+    {
+      "entropy": 0.6118641197681427,
+      "epoch": 2.2859463850528026,
+      "grad_norm": 0.009624680504202843,
+      "learning_rate": 0.0002,
+      "loss": 0.6219325065612793,
+      "mean_token_accuracy": 0.7466080039739609,
+      "num_tokens": 13120093.0,
+      "step": 352
+    },
+    {
+      "entropy": 0.5959088578820229,
+      "epoch": 2.2924451665312753,
+      "grad_norm": 0.007752889767289162,
+      "learning_rate": 0.0002,
+      "loss": 0.5983707904815674,
+      "mean_token_accuracy": 0.755429819226265,
+      "num_tokens": 13157659.0,
+      "step": 353
+    },
+    {
+      "entropy": 0.6112657487392426,
+      "epoch": 2.298943948009748,
+      "grad_norm": 0.008281555026769638,
+      "learning_rate": 0.0002,
+      "loss": 0.6082602143287659,
+      "mean_token_accuracy": 0.7516218572854996,
+      "num_tokens": 13194647.0,
+      "step": 354
+    },
+    {
+      "entropy": 0.6122025474905968,
+      "epoch": 2.305442729488221,
+      "grad_norm": 0.008959132246673107,
+      "learning_rate": 0.0002,
+      "loss": 0.6142069101333618,
+      "mean_token_accuracy": 0.7516066282987595,
+      "num_tokens": 13231933.0,
+      "step": 355
+    },
+    {
+      "entropy": 0.610342264175415,
+      "epoch": 2.311941510966694,
+      "grad_norm": 0.0076917321421206,
+      "learning_rate": 0.0002,
+      "loss": 0.6098777055740356,
+      "mean_token_accuracy": 0.7534263134002686,
+      "num_tokens": 13269472.0,
+      "step": 356
+    },
+    {
+      "entropy": 0.6236057877540588,
+      "epoch": 2.3184402924451666,
+      "grad_norm": 0.008514189161360264,
+      "learning_rate": 0.0002,
+      "loss": 0.6241997480392456,
+      "mean_token_accuracy": 0.7468095943331718,
+      "num_tokens": 13307269.0,
+      "step": 357
+    },
+    {
+      "entropy": 0.6160306110978127,
+      "epoch": 2.3249390739236393,
+      "grad_norm": 0.009143211878836155,
+      "learning_rate": 0.0002,
+      "loss": 0.6196573972702026,
+      "mean_token_accuracy": 0.7480775937438011,
+      "num_tokens": 13344551.0,
+      "step": 358
+    },
+    {
+      "entropy": 0.6120478510856628,
+      "epoch": 2.331437855402112,
+      "grad_norm": 0.007592627312988043,
+      "learning_rate": 0.0002,
+      "loss": 0.6155714392662048,
+      "mean_token_accuracy": 0.7500782683491707,
+      "num_tokens": 13382127.0,
+      "step": 359
+    },
+    {
+      "entropy": 0.6234208717942238,
+      "epoch": 2.3379366368805847,
+      "grad_norm": 0.008143738843500614,
+      "learning_rate": 0.0002,
+      "loss": 0.6278046369552612,
+      "mean_token_accuracy": 0.7446608766913414,
+      "num_tokens": 13419810.0,
+      "step": 360
+    },
+    {
+      "entropy": 0.6157349273562431,
+      "epoch": 2.3444354183590574,
+      "grad_norm": 0.008566657081246376,
+      "learning_rate": 0.0002,
+      "loss": 0.6156420707702637,
+      "mean_token_accuracy": 0.7479202896356583,
+      "num_tokens": 13457273.0,
+      "step": 361
+    },
+    {
+      "entropy": 0.6180937066674232,
+      "epoch": 2.3509341998375306,
+      "grad_norm": 0.008304375223815441,
+      "learning_rate": 0.0002,
+      "loss": 0.622611403465271,
+      "mean_token_accuracy": 0.7477732971310616,
+      "num_tokens": 13494536.0,
+      "step": 362
+    },
+    {
+      "entropy": 0.6232565492391586,
+      "epoch": 2.3574329813160033,
+      "grad_norm": 0.008868128061294556,
+      "learning_rate": 0.0002,
+      "loss": 0.6260833740234375,
+      "mean_token_accuracy": 0.7459229752421379,
+      "num_tokens": 13531864.0,
+      "step": 363
+    },
+    {
+      "entropy": 0.6254989579319954,
+      "epoch": 2.363931762794476,
+      "grad_norm": 0.008256005123257637,
+      "learning_rate": 0.0002,
+      "loss": 0.6260857582092285,
+      "mean_token_accuracy": 0.7447438836097717,
+      "num_tokens": 13569443.0,
+      "step": 364
+    },
+    {
+      "entropy": 0.6221200376749039,
+      "epoch": 2.3704305442729487,
+      "grad_norm": 0.009361873380839825,
+      "learning_rate": 0.0002,
+      "loss": 0.6180988550186157,
+      "mean_token_accuracy": 0.7470216527581215,
+      "num_tokens": 13606513.0,
+      "step": 365
+    },
+    {
+      "entropy": 0.616127498447895,
+      "epoch": 2.3769293257514215,
+      "grad_norm": 0.008144999854266644,
+      "learning_rate": 0.0002,
+      "loss": 0.6137336492538452,
+      "mean_token_accuracy": 0.7496519684791565,
+      "num_tokens": 13643706.0,
+      "step": 366
+    },
+    {
+      "entropy": 0.6194290667772293,
+      "epoch": 2.3834281072298946,
+      "grad_norm": 0.008278160355985165,
+      "learning_rate": 0.0002,
+      "loss": 0.6232280135154724,
+      "mean_token_accuracy": 0.7481147646903992,
+      "num_tokens": 13680957.0,
+      "step": 367
+    },
+    {
+      "entropy": 0.6192846670746803,
+      "epoch": 2.3899268887083673,
+      "grad_norm": 0.010223917663097382,
+      "learning_rate": 0.0002,
+      "loss": 0.6253848075866699,
+      "mean_token_accuracy": 0.744193010032177,
+      "num_tokens": 13717903.0,
+      "step": 368
+    },
+    {
+      "entropy": 0.6221625953912735,
+      "epoch": 2.39642567018684,
+      "grad_norm": 0.008200163953006268,
+      "learning_rate": 0.0002,
+      "loss": 0.6191288232803345,
+      "mean_token_accuracy": 0.74726802110672,
+      "num_tokens": 13755260.0,
+      "step": 369
+    },
+    {
+      "entropy": 0.6163566559553146,
+      "epoch": 2.4029244516653128,
+      "grad_norm": 0.007974907755851746,
+      "learning_rate": 0.0002,
+      "loss": 0.6134728193283081,
+      "mean_token_accuracy": 0.7508314996957779,
+      "num_tokens": 13792517.0,
+      "step": 370
+    },
+    {
+      "entropy": 0.6114080101251602,
+      "epoch": 2.4094232331437855,
+      "grad_norm": 0.008343557827174664,
+      "learning_rate": 0.0002,
+      "loss": 0.6111307144165039,
+      "mean_token_accuracy": 0.7514908611774445,
+      "num_tokens": 13829760.0,
+      "step": 371
+    },
+    {
+      "entropy": 0.5953814163804054,
+      "epoch": 2.415922014622258,
+      "grad_norm": 0.01898498833179474,
+      "learning_rate": 0.0002,
+      "loss": 0.5988050699234009,
+      "mean_token_accuracy": 0.7543381601572037,
+      "num_tokens": 13867370.0,
+      "step": 372
+    },
+    {
+      "entropy": 0.6071057021617889,
+      "epoch": 2.422420796100731,
+      "grad_norm": 0.023124704137444496,
+      "learning_rate": 0.0002,
+      "loss": 0.6178369522094727,
+      "mean_token_accuracy": 0.747986689209938,
+      "num_tokens": 13904625.0,
+      "step": 373
+    },
+    {
+      "entropy": 0.6107468456029892,
+      "epoch": 2.428919577579204,
+      "grad_norm": 0.01215590164065361,
+      "learning_rate": 0.0002,
+      "loss": 0.6207199096679688,
+      "mean_token_accuracy": 0.7475155666470528,
+      "num_tokens": 13942489.0,
+      "step": 374
+    },
+    {
+      "entropy": 0.6124749183654785,
+      "epoch": 2.435418359057677,
+      "grad_norm": 0.012728074565529823,
+      "learning_rate": 0.0002,
+      "loss": 0.6103756427764893,
+      "mean_token_accuracy": 0.7500820159912109,
+      "num_tokens": 13980205.0,
+      "step": 375
+    },
+    {
+      "entropy": 0.6109942868351936,
+      "epoch": 2.4419171405361495,
+      "grad_norm": 0.009368120692670345,
+      "learning_rate": 0.0002,
+      "loss": 0.6066494584083557,
+      "mean_token_accuracy": 0.7536064460873604,
+      "num_tokens": 14017641.0,
+      "step": 376
+    },
+    {
+      "entropy": 0.624424010515213,
+      "epoch": 2.448415922014622,
+      "grad_norm": 0.016890961676836014,
+      "learning_rate": 0.0002,
+      "loss": 0.614218533039093,
+      "mean_token_accuracy": 0.7494603618979454,
+      "num_tokens": 14054677.0,
+      "step": 377
+    },
+    {
+      "entropy": 0.6255820393562317,
+      "epoch": 2.454914703493095,
+      "grad_norm": 0.009696394205093384,
+      "learning_rate": 0.0002,
+      "loss": 0.6304793357849121,
+      "mean_token_accuracy": 0.7434439361095428,
+      "num_tokens": 14091764.0,
+      "step": 378
+    },
+    {
+      "entropy": 0.621983028948307,
+      "epoch": 2.4614134849715676,
+      "grad_norm": 0.020173629745841026,
+      "learning_rate": 0.0002,
+      "loss": 0.6257689595222473,
+      "mean_token_accuracy": 0.745697520673275,
+      "num_tokens": 14129278.0,
+      "step": 379
+    },
+    {
+      "entropy": 0.6207767054438591,
+      "epoch": 2.467912266450041,
+      "grad_norm": 0.008954247459769249,
+      "learning_rate": 0.0002,
+      "loss": 0.6241703033447266,
+      "mean_token_accuracy": 0.750006191432476,
+      "num_tokens": 14167232.0,
+      "step": 380
+    },
+    {
+      "entropy": 0.6112791821360588,
+      "epoch": 2.4744110479285135,
+      "grad_norm": 0.009806505404412746,
+      "learning_rate": 0.0002,
+      "loss": 0.6147319078445435,
+      "mean_token_accuracy": 0.7495506927371025,
+      "num_tokens": 14204632.0,
+      "step": 381
+    },
+    {
+      "entropy": 0.6135937720537186,
+      "epoch": 2.4809098294069862,
+      "grad_norm": 0.008928372524678707,
+      "learning_rate": 0.0002,
+      "loss": 0.6139630079269409,
+      "mean_token_accuracy": 0.7506385967135429,
+      "num_tokens": 14242352.0,
+      "step": 382
+    },
+    {
+      "entropy": 0.6150240004062653,
+      "epoch": 2.487408610885459,
+      "grad_norm": 0.015955505892634392,
+      "learning_rate": 0.0002,
+      "loss": 0.6152509450912476,
+      "mean_token_accuracy": 0.7515363991260529,
+      "num_tokens": 14280058.0,
+      "step": 383
+    },
+    {
+      "entropy": 0.6128680855035782,
+      "epoch": 2.4939073923639317,
+      "grad_norm": 0.009567083790898323,
+      "learning_rate": 0.0002,
+      "loss": 0.6167535781860352,
+      "mean_token_accuracy": 0.748919777572155,
+      "num_tokens": 14317661.0,
+      "step": 384
+    },
+    {
+      "entropy": 0.626237154006958,
+      "epoch": 2.5004061738424044,
+      "grad_norm": 0.01046669203788042,
+      "learning_rate": 0.0002,
+      "loss": 0.6324093341827393,
+      "mean_token_accuracy": 0.7444325461983681,
+      "num_tokens": 14355367.0,
+      "step": 385
+    },
+    {
+      "entropy": 0.6202276349067688,
+      "epoch": 2.506904955320877,
+      "grad_norm": 0.007664418779313564,
+      "learning_rate": 0.0002,
+      "loss": 0.6191686987876892,
+      "mean_token_accuracy": 0.7475631311535835,
+      "num_tokens": 14393315.0,
+      "step": 386
+    },
+    {
+      "entropy": 0.6242005750536919,
+      "epoch": 2.5134037367993503,
+      "grad_norm": 0.00844619981944561,
+      "learning_rate": 0.0002,
+      "loss": 0.621794581413269,
+      "mean_token_accuracy": 0.7481283098459244,
+      "num_tokens": 14430853.0,
+      "step": 387
+    },
+    {
+      "entropy": 0.6239676773548126,
+      "epoch": 2.519902518277823,
+      "grad_norm": 0.008689953945577145,
+      "learning_rate": 0.0002,
+      "loss": 0.6237965226173401,
+      "mean_token_accuracy": 0.7461953610181808,
+      "num_tokens": 14468292.0,
+      "step": 388
+    },
+    {
+      "entropy": 0.6107499524950981,
+      "epoch": 2.5264012997562957,
+      "grad_norm": 0.009505955502390862,
+      "learning_rate": 0.0002,
+      "loss": 0.6074596643447876,
+      "mean_token_accuracy": 0.7525542005896568,
+      "num_tokens": 14505921.0,
+      "step": 389
+    },
+    {
+      "entropy": 0.6142911538481712,
+      "epoch": 2.5329000812347684,
+      "grad_norm": 0.009658354334533215,
+      "learning_rate": 0.0002,
+      "loss": 0.6198441982269287,
+      "mean_token_accuracy": 0.7489058822393417,
+      "num_tokens": 14543384.0,
+      "step": 390
+    },
+    {
+      "entropy": 0.6049473881721497,
+      "epoch": 2.539398862713241,
+      "grad_norm": 0.010726661421358585,
+      "learning_rate": 0.0002,
+      "loss": 0.6118952631950378,
+      "mean_token_accuracy": 0.7526208385825157,
+      "num_tokens": 14580935.0,
+      "step": 391
+    },
+    {
+      "entropy": 0.5964333266019821,
+      "epoch": 2.5458976441917143,
+      "grad_norm": 0.00981208123266697,
+      "learning_rate": 0.0002,
+      "loss": 0.6057010889053345,
+      "mean_token_accuracy": 0.7543943077325821,
+      "num_tokens": 14618074.0,
+      "step": 392
+    },
+    {
+      "entropy": 0.6099359169602394,
+      "epoch": 2.552396425670187,
+      "grad_norm": 0.009079058654606342,
+      "learning_rate": 0.0002,
+      "loss": 0.6084835529327393,
+      "mean_token_accuracy": 0.7520028725266457,
+      "num_tokens": 14655186.0,
+      "step": 393
+    },
+    {
+      "entropy": 0.6173275411128998,
+      "epoch": 2.5588952071486597,
+      "grad_norm": 0.009335207752883434,
+      "learning_rate": 0.0002,
+      "loss": 0.6102230548858643,
+      "mean_token_accuracy": 0.7514491975307465,
+      "num_tokens": 14692383.0,
+      "step": 394
+    },
+    {
+      "entropy": 0.6305418461561203,
+      "epoch": 2.5653939886271324,
+      "grad_norm": 0.009312084876000881,
+      "learning_rate": 0.0002,
+      "loss": 0.6281323432922363,
+      "mean_token_accuracy": 0.7421899363398552,
+      "num_tokens": 14729446.0,
+      "step": 395
+    },
+    {
+      "entropy": 0.6119675636291504,
+      "epoch": 2.571892770105605,
+      "grad_norm": 0.00806971825659275,
+      "learning_rate": 0.0002,
+      "loss": 0.6118816137313843,
+      "mean_token_accuracy": 0.7502529993653297,
+      "num_tokens": 14766917.0,
+      "step": 396
+    },
+    {
+      "entropy": 0.6194342076778412,
+      "epoch": 2.578391551584078,
+      "grad_norm": 0.009102623909711838,
+      "learning_rate": 0.0002,
+      "loss": 0.625906229019165,
+      "mean_token_accuracy": 0.7444874346256256,
+      "num_tokens": 14804635.0,
+      "step": 397
+    },
+    {
+      "entropy": 0.6129268780350685,
+      "epoch": 2.5848903330625506,
+      "grad_norm": 0.010471413843333721,
+      "learning_rate": 0.0002,
+      "loss": 0.6221112012863159,
+      "mean_token_accuracy": 0.7463114410638809,
+      "num_tokens": 14842380.0,
+      "step": 398
+    },
+    {
+      "entropy": 0.6111740916967392,
+      "epoch": 2.5913891145410237,
+      "grad_norm": 0.008378157392144203,
+      "learning_rate": 0.0002,
+      "loss": 0.6132410764694214,
+      "mean_token_accuracy": 0.7493879497051239,
+      "num_tokens": 14879441.0,
+      "step": 399
+    },
+    {
+      "entropy": 0.6207928210496902,
+      "epoch": 2.5978878960194964,
+      "grad_norm": 0.008037488907575607,
+      "learning_rate": 0.0002,
+      "loss": 0.6196093559265137,
+      "mean_token_accuracy": 0.7491321563720703,
+      "num_tokens": 14916731.0,
+      "step": 400
+    },
+    {
+      "entropy": 0.6349622085690498,
+      "epoch": 2.604386677497969,
+      "grad_norm": 0.00913564208894968,
+      "learning_rate": 0.0002,
+      "loss": 0.6318038702011108,
+      "mean_token_accuracy": 0.7420522123575211,
+      "num_tokens": 14954276.0,
+      "step": 401
+    },
+    {
+      "entropy": 0.6019857749342918,
+      "epoch": 2.610885458976442,
+      "grad_norm": 0.008696068078279495,
+      "learning_rate": 0.0002,
+      "loss": 0.600864052772522,
+      "mean_token_accuracy": 0.7552035823464394,
+      "num_tokens": 14991349.0,
+      "step": 402
+    },
+    {
+      "entropy": 0.6017836257815361,
+      "epoch": 2.6173842404549146,
+      "grad_norm": 0.008207019418478012,
+      "learning_rate": 0.0002,
+      "loss": 0.6038705706596375,
+      "mean_token_accuracy": 0.7532704323530197,
+      "num_tokens": 15028601.0,
+      "step": 403
+    },
+    {
+      "entropy": 0.6120554357767105,
+      "epoch": 2.6238830219333877,
+      "grad_norm": 0.008753425441682339,
+      "learning_rate": 0.0002,
+      "loss": 0.6164306998252869,
+      "mean_token_accuracy": 0.7496815398335457,
+      "num_tokens": 15066182.0,
+      "step": 404
+    },
+    {
+      "entropy": 0.6131924018263817,
+      "epoch": 2.6303818034118605,
+      "grad_norm": 0.008548066020011902,
+      "learning_rate": 0.0002,
+      "loss": 0.6181113719940186,
+      "mean_token_accuracy": 0.7491538524627686,
+      "num_tokens": 15103340.0,
+      "step": 405
+    },
+    {
+      "entropy": 0.6061984226107597,
+      "epoch": 2.636880584890333,
+      "grad_norm": 0.009016766212880611,
+      "learning_rate": 0.0002,
+      "loss": 0.6121985912322998,
+      "mean_token_accuracy": 0.7507323697209358,
+      "num_tokens": 15140594.0,
+      "step": 406
+    },
+    {
+      "entropy": 0.6165529787540436,
+      "epoch": 2.643379366368806,
+      "grad_norm": 0.007676286157220602,
+      "learning_rate": 0.0002,
+      "loss": 0.6143813133239746,
+      "mean_token_accuracy": 0.7518013492226601,
+      "num_tokens": 15178037.0,
+      "step": 407
+    },
+    {
+      "entropy": 0.6116887852549553,
+      "epoch": 2.6498781478472786,
+      "grad_norm": 0.007873927243053913,
+      "learning_rate": 0.0002,
+      "loss": 0.6096634268760681,
+      "mean_token_accuracy": 0.7518724203109741,
+      "num_tokens": 15215362.0,
+      "step": 408
+    },
+    {
+      "entropy": 0.6075077503919601,
+      "epoch": 2.6563769293257513,
+      "grad_norm": 0.007256446871906519,
+      "learning_rate": 0.0002,
+      "loss": 0.6078198552131653,
+      "mean_token_accuracy": 0.7537789046764374,
+      "num_tokens": 15252624.0,
+      "step": 409
+    },
+    {
+      "entropy": 0.6106631234288216,
+      "epoch": 2.662875710804224,
+      "grad_norm": 0.009321639314293861,
+      "learning_rate": 0.0002,
+      "loss": 0.6120110750198364,
+      "mean_token_accuracy": 0.7512669935822487,
+      "num_tokens": 15290298.0,
+      "step": 410
+    },
+    {
+      "entropy": 0.6318610459566116,
+      "epoch": 2.6693744922826967,
+      "grad_norm": 0.008198688738048077,
+      "learning_rate": 0.0002,
+      "loss": 0.6370584964752197,
+      "mean_token_accuracy": 0.740862101316452,
+      "num_tokens": 15327374.0,
+      "step": 411
+    },
+    {
+      "entropy": 0.6139440014958382,
+      "epoch": 2.67587327376117,
+      "grad_norm": 0.008440233767032623,
+      "learning_rate": 0.0002,
+      "loss": 0.6130324602127075,
+      "mean_token_accuracy": 0.7509049624204636,
+      "num_tokens": 15364313.0,
+      "step": 412
+    },
+    {
+      "entropy": 0.6132312193512917,
+      "epoch": 2.6823720552396426,
+      "grad_norm": 0.00764568243175745,
+      "learning_rate": 0.0002,
+      "loss": 0.6143498420715332,
+      "mean_token_accuracy": 0.7499875500798225,
+      "num_tokens": 15401585.0,
+      "step": 413
+    },
+    {
+      "entropy": 0.6295821517705917,
+      "epoch": 2.6888708367181153,
+      "grad_norm": 0.00870992336422205,
+      "learning_rate": 0.0002,
+      "loss": 0.6297035217285156,
+      "mean_token_accuracy": 0.7440814226865768,
+      "num_tokens": 15439063.0,
+      "step": 414
+    },
+    {
+      "entropy": 0.6159479022026062,
+      "epoch": 2.695369618196588,
+      "grad_norm": 0.007172968704253435,
+      "learning_rate": 0.0002,
+      "loss": 0.6142174005508423,
+      "mean_token_accuracy": 0.7487895339727402,
+      "num_tokens": 15476100.0,
+      "step": 415
+    },
+    {
+      "entropy": 0.6020728126168251,
+      "epoch": 2.7018683996750608,
+      "grad_norm": 0.007704848889261484,
+      "learning_rate": 0.0002,
+      "loss": 0.6082124710083008,
+      "mean_token_accuracy": 0.752522885799408,
+      "num_tokens": 15513167.0,
+      "step": 416
+    },
+    {
+      "entropy": 0.6281116902828217,
+      "epoch": 2.708367181153534,
+      "grad_norm": 0.00924444105476141,
+      "learning_rate": 0.0002,
+      "loss": 0.6312050819396973,
+      "mean_token_accuracy": 0.7460011318325996,
+      "num_tokens": 15550739.0,
+      "step": 417
+    },
+    {
+      "entropy": 0.6233636066317558,
+      "epoch": 2.7148659626320066,
+      "grad_norm": 0.008710985071957111,
+      "learning_rate": 0.0002,
+      "loss": 0.6260718107223511,
+      "mean_token_accuracy": 0.744461327791214,
+      "num_tokens": 15587950.0,
+      "step": 418
+    },
+    {
+      "entropy": 0.6195387095212936,
+      "epoch": 2.7213647441104794,
+      "grad_norm": 0.007183417212218046,
+      "learning_rate": 0.0002,
+      "loss": 0.6212611198425293,
+      "mean_token_accuracy": 0.746867798268795,
+      "num_tokens": 15625008.0,
+      "step": 419
+    },
+    {
+      "entropy": 0.6109424605965614,
+      "epoch": 2.727863525588952,
+      "grad_norm": 0.0077791716903448105,
+      "learning_rate": 0.0002,
+      "loss": 0.6113884449005127,
+      "mean_token_accuracy": 0.7505641579627991,
+      "num_tokens": 15662159.0,
+      "step": 420
+    },
+    {
+      "entropy": 0.6187275499105453,
+      "epoch": 2.734362307067425,
+      "grad_norm": 0.008177550509572029,
+      "learning_rate": 0.0002,
+      "loss": 0.6187748908996582,
+      "mean_token_accuracy": 0.7493242397904396,
+      "num_tokens": 15699404.0,
+      "step": 421
+    },
+    {
+      "entropy": 0.6269592270255089,
+      "epoch": 2.7408610885458975,
+      "grad_norm": 0.00797547772526741,
+      "learning_rate": 0.0002,
+      "loss": 0.6266059875488281,
+      "mean_token_accuracy": 0.7470143884420395,
+      "num_tokens": 15736867.0,
+      "step": 422
+    },
+    {
+      "entropy": 0.6187637895345688,
+      "epoch": 2.74735987002437,
+      "grad_norm": 0.008009313605725765,
+      "learning_rate": 0.0002,
+      "loss": 0.6174246072769165,
+      "mean_token_accuracy": 0.748706616461277,
+      "num_tokens": 15774518.0,
+      "step": 423
+    },
+    {
+      "entropy": 0.6222201958298683,
+      "epoch": 2.7538586515028434,
+      "grad_norm": 0.00828398484736681,
+      "learning_rate": 0.0002,
+      "loss": 0.6254167556762695,
+      "mean_token_accuracy": 0.7451123148202896,
+      "num_tokens": 15811807.0,
+      "step": 424
+    },
+    {
+      "entropy": 0.6253098249435425,
+      "epoch": 2.760357432981316,
+      "grad_norm": 0.0075772651471197605,
+      "learning_rate": 0.0002,
+      "loss": 0.6270281076431274,
+      "mean_token_accuracy": 0.744781106710434,
+      "num_tokens": 15849715.0,
+      "step": 425
+    },
+    {
+      "entropy": 0.6165027767419815,
+      "epoch": 2.766856214459789,
+      "grad_norm": 0.00784947071224451,
+      "learning_rate": 0.0002,
+      "loss": 0.6173769235610962,
+      "mean_token_accuracy": 0.748318687081337,
+      "num_tokens": 15886927.0,
+      "step": 426
+    },
+    {
+      "entropy": 0.6187942251563072,
+      "epoch": 2.7733549959382615,
+      "grad_norm": 0.007960534654557705,
+      "learning_rate": 0.0002,
+      "loss": 0.6190329790115356,
+      "mean_token_accuracy": 0.7482394725084305,
+      "num_tokens": 15924820.0,
+      "step": 427
+    },
+    {
+      "entropy": 0.6049096137285233,
+      "epoch": 2.7798537774167342,
+      "grad_norm": 0.007044166326522827,
+      "learning_rate": 0.0002,
+      "loss": 0.608286440372467,
+      "mean_token_accuracy": 0.7529338300228119,
+      "num_tokens": 15962031.0,
+      "step": 428
+    },
+    {
+      "entropy": 0.6067706495523453,
+      "epoch": 2.7863525588952074,
+      "grad_norm": 0.008417103439569473,
+      "learning_rate": 0.0002,
+      "loss": 0.6071891188621521,
+      "mean_token_accuracy": 0.7529254406690598,
+      "num_tokens": 15999271.0,
+      "step": 429
+    },
+    {
+      "entropy": 0.608584851026535,
+      "epoch": 2.79285134037368,
+      "grad_norm": 0.0066911159083247185,
+      "learning_rate": 0.0002,
+      "loss": 0.6128258109092712,
+      "mean_token_accuracy": 0.7512886077165604,
+      "num_tokens": 16036782.0,
+      "step": 430
+    },
+    {
+      "entropy": 0.6173116937279701,
+      "epoch": 2.799350121852153,
+      "grad_norm": 0.007603482808917761,
+      "learning_rate": 0.0002,
+      "loss": 0.6212354898452759,
+      "mean_token_accuracy": 0.7506462633609772,
+      "num_tokens": 16074488.0,
+      "step": 431
+    },
+    {
+      "entropy": 0.612823948264122,
+      "epoch": 2.8058489033306255,
+      "grad_norm": 0.008489524014294147,
+      "learning_rate": 0.0002,
+      "loss": 0.6168191432952881,
+      "mean_token_accuracy": 0.7491662427783012,
+      "num_tokens": 16112183.0,
+      "step": 432
+    },
+    {
+      "entropy": 0.6224610134959221,
+      "epoch": 2.8123476848090982,
+      "grad_norm": 0.0072684078477323055,
+      "learning_rate": 0.0002,
+      "loss": 0.6210014820098877,
+      "mean_token_accuracy": 0.7473989799618721,
+      "num_tokens": 16149570.0,
+      "step": 433
+    },
+    {
+      "entropy": 0.6252987831830978,
+      "epoch": 2.818846466287571,
+      "grad_norm": 0.008009562268853188,
+      "learning_rate": 0.0002,
+      "loss": 0.6273704171180725,
+      "mean_token_accuracy": 0.7445142269134521,
+      "num_tokens": 16187156.0,
+      "step": 434
+    },
+    {
+      "entropy": 0.6278671100735664,
+      "epoch": 2.8253452477660437,
+      "grad_norm": 0.007072070613503456,
+      "learning_rate": 0.0002,
+      "loss": 0.6284614205360413,
+      "mean_token_accuracy": 0.743632547557354,
+      "num_tokens": 16224612.0,
+      "step": 435
+    },
+    {
+      "entropy": 0.619408093392849,
+      "epoch": 2.8318440292445164,
+      "grad_norm": 0.007810318376868963,
+      "learning_rate": 0.0002,
+      "loss": 0.6186745166778564,
+      "mean_token_accuracy": 0.7505189552903175,
+      "num_tokens": 16261848.0,
+      "step": 436
+    },
+    {
+      "entropy": 0.617984876036644,
+      "epoch": 2.8383428107229896,
+      "grad_norm": 0.007404230069369078,
+      "learning_rate": 0.0002,
+      "loss": 0.6218393445014954,
+      "mean_token_accuracy": 0.746470145881176,
+      "num_tokens": 16299536.0,
+      "step": 437
+    },
+    {
+      "entropy": 0.6115261316299438,
+      "epoch": 2.8448415922014623,
+      "grad_norm": 0.007786816451698542,
+      "learning_rate": 0.0002,
+      "loss": 0.6147224307060242,
+      "mean_token_accuracy": 0.7512907087802887,
+      "num_tokens": 16336727.0,
+      "step": 438
+    },
+    {
+      "entropy": 0.610576219856739,
+      "epoch": 2.851340373679935,
+      "grad_norm": 0.007572745438665152,
+      "learning_rate": 0.0002,
+      "loss": 0.6138074994087219,
+      "mean_token_accuracy": 0.7504170089960098,
+      "num_tokens": 16374001.0,
+      "step": 439
+    },
+    {
+      "entropy": 0.6278978139162064,
+      "epoch": 2.8578391551584077,
+      "grad_norm": 0.007563670165836811,
+      "learning_rate": 0.0002,
+      "loss": 0.6268113851547241,
+      "mean_token_accuracy": 0.7438701540231705,
+      "num_tokens": 16411469.0,
+      "step": 440
+    },
+    {
+      "entropy": 0.6234687641263008,
+      "epoch": 2.864337936636881,
+      "grad_norm": 0.007766201626509428,
+      "learning_rate": 0.0002,
+      "loss": 0.6241582632064819,
+      "mean_token_accuracy": 0.7458857893943787,
+      "num_tokens": 16449061.0,
+      "step": 441
+    },
+    {
+      "entropy": 0.6170031577348709,
+      "epoch": 2.8708367181153536,
+      "grad_norm": 0.0072515420615673065,
+      "learning_rate": 0.0002,
+      "loss": 0.6164381504058838,
+      "mean_token_accuracy": 0.7487066686153412,
+      "num_tokens": 16486116.0,
+      "step": 442
+    },
+    {
+      "entropy": 0.6071048676967621,
+      "epoch": 2.8773354995938263,
+      "grad_norm": 0.007729642558842897,
+      "learning_rate": 0.0002,
+      "loss": 0.6061218976974487,
+      "mean_token_accuracy": 0.7546169012784958,
+      "num_tokens": 16523581.0,
+      "step": 443
+    },
+    {
+      "entropy": 0.6005209162831306,
+      "epoch": 2.883834281072299,
+      "grad_norm": 0.009296285919845104,
+      "learning_rate": 0.0002,
+      "loss": 0.608015775680542,
+      "mean_token_accuracy": 0.7522506862878799,
+      "num_tokens": 16560766.0,
+      "step": 444
+    },
+    {
+      "entropy": 0.6114199981093407,
+      "epoch": 2.8903330625507717,
+      "grad_norm": 0.009450486861169338,
+      "learning_rate": 0.0002,
+      "loss": 0.61246258020401,
+      "mean_token_accuracy": 0.7513059675693512,
+      "num_tokens": 16597924.0,
+      "step": 445
+    },
+    {
+      "entropy": 0.6163241118192673,
+      "epoch": 2.8968318440292444,
+      "grad_norm": 0.007840070873498917,
+      "learning_rate": 0.0002,
+      "loss": 0.6163046360015869,
+      "mean_token_accuracy": 0.7482790276408195,
+      "num_tokens": 16635707.0,
+      "step": 446
+    },
+    {
+      "entropy": 0.6230868250131607,
+      "epoch": 2.903330625507717,
+      "grad_norm": 0.00791420228779316,
+      "learning_rate": 0.0002,
+      "loss": 0.627612829208374,
+      "mean_token_accuracy": 0.7418907359242439,
+      "num_tokens": 16673272.0,
+      "step": 447
+    },
+    {
+      "entropy": 0.6148254349827766,
+      "epoch": 2.90982940698619,
+      "grad_norm": 0.008231899701058865,
+      "learning_rate": 0.0002,
+      "loss": 0.615432620048523,
+      "mean_token_accuracy": 0.74872937053442,
+      "num_tokens": 16710307.0,
+      "step": 448
+    },
+    {
+      "entropy": 0.6145784482359886,
+      "epoch": 2.916328188464663,
+      "grad_norm": 0.008206733502447605,
+      "learning_rate": 0.0002,
+      "loss": 0.6162281036376953,
+      "mean_token_accuracy": 0.7497396990656853,
+      "num_tokens": 16747541.0,
+      "step": 449
+    },
+    {
+      "entropy": 0.6129453554749489,
+      "epoch": 2.9228269699431357,
+      "grad_norm": 0.008618887513875961,
+      "learning_rate": 0.0002,
+      "loss": 0.6113362312316895,
+      "mean_token_accuracy": 0.7538808137178421,
+      "num_tokens": 16784808.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.6318705305457115,
+      "epoch": 2.9293257514216084,
+      "grad_norm": 0.00894182175397873,
+      "learning_rate": 0.0002,
+      "loss": 0.6285560131072998,
+      "mean_token_accuracy": 0.7439467459917068,
+      "num_tokens": 16822571.0,
+      "step": 451
+    },
+    {
+      "entropy": 0.6192348748445511,
+      "epoch": 2.935824532900081,
+      "grad_norm": 0.0075249881483614445,
+      "learning_rate": 0.0002,
+      "loss": 0.6225606203079224,
+      "mean_token_accuracy": 0.7450770661234856,
+      "num_tokens": 16860236.0,
+      "step": 452
+    },
+    {
+      "entropy": 0.6156643033027649,
+      "epoch": 2.942323314378554,
+      "grad_norm": 0.008608740754425526,
+      "learning_rate": 0.0002,
+      "loss": 0.6196963787078857,
+      "mean_token_accuracy": 0.7487252280116081,
+      "num_tokens": 16897505.0,
+      "step": 453
+    },
+    {
+      "entropy": 0.6020097509026527,
+      "epoch": 2.948822095857027,
+      "grad_norm": 0.008554862812161446,
+      "learning_rate": 0.0002,
+      "loss": 0.6057189702987671,
+      "mean_token_accuracy": 0.7535376027226448,
+      "num_tokens": 16934723.0,
+      "step": 454
+    },
+    {
+      "entropy": 0.6096286401152611,
+      "epoch": 2.9553208773354998,
+      "grad_norm": 0.008194280788302422,
+      "learning_rate": 0.0002,
+      "loss": 0.6100919842720032,
+      "mean_token_accuracy": 0.7543619722127914,
+      "num_tokens": 16972160.0,
+      "step": 455
+    },
+    {
+      "entropy": 0.6101824045181274,
+      "epoch": 2.9618196588139725,
+      "grad_norm": 0.008089696057140827,
+      "learning_rate": 0.0002,
+      "loss": 0.6100078225135803,
+      "mean_token_accuracy": 0.7519838437438011,
+      "num_tokens": 17009553.0,
+      "step": 456
+    },
+    {
+      "entropy": 0.6134164482355118,
+      "epoch": 2.968318440292445,
+      "grad_norm": 0.008095222525298595,
+      "learning_rate": 0.0002,
+      "loss": 0.61534583568573,
+      "mean_token_accuracy": 0.7492899596691132,
+      "num_tokens": 17046859.0,
+      "step": 457
+    },
+    {
+      "entropy": 0.6219535619020462,
+      "epoch": 2.974817221770918,
+      "grad_norm": 0.008277404122054577,
+      "learning_rate": 0.0002,
+      "loss": 0.6226615905761719,
+      "mean_token_accuracy": 0.7486208230257034,
+      "num_tokens": 17083891.0,
+      "step": 458
+    },
+    {
+      "entropy": 0.6230077967047691,
+      "epoch": 2.9813160032493906,
+      "grad_norm": 0.007576305419206619,
+      "learning_rate": 0.0002,
+      "loss": 0.6215830445289612,
+      "mean_token_accuracy": 0.7478786036372185,
+      "num_tokens": 17121415.0,
+      "step": 459
+    },
+    {
+      "entropy": 0.6288741081953049,
+      "epoch": 2.9878147847278633,
+      "grad_norm": 0.007517755497246981,
+      "learning_rate": 0.0002,
+      "loss": 0.6266385912895203,
+      "mean_token_accuracy": 0.7459503784775734,
+      "num_tokens": 17159316.0,
+      "step": 460
+    },
+    {
+      "entropy": 0.6128450557589531,
+      "epoch": 2.994313566206336,
+      "grad_norm": 0.0082389609888196,
+      "learning_rate": 0.0002,
+      "loss": 0.6180654764175415,
+      "mean_token_accuracy": 0.7501172721385956,
+      "num_tokens": 17196686.0,
+      "step": 461
+    },
+    {
+      "entropy": 0.6104621972356524,
+      "epoch": 3.0,
+      "grad_norm": 0.009389652870595455,
+      "learning_rate": 0.0002,
+      "loss": 0.6207214593887329,
+      "mean_token_accuracy": 0.7534014242036002,
+      "num_tokens": 17212985.0,
+      "step": 462
+    },
+    {
+      "entropy": 0.6010295078158379,
+      "epoch": 3.0064987814784727,
+      "grad_norm": 0.007540151942521334,
+      "learning_rate": 0.0002,
+      "loss": 0.604315996170044,
+      "mean_token_accuracy": 0.7545758932828903,
+      "num_tokens": 17250548.0,
+      "step": 463
+    },
+    {
+      "entropy": 0.6099191755056381,
+      "epoch": 3.0129975629569454,
+      "grad_norm": 0.007964324206113815,
+      "learning_rate": 0.0002,
+      "loss": 0.6118016839027405,
+      "mean_token_accuracy": 0.7505887746810913,
+      "num_tokens": 17288088.0,
+      "step": 464
+    },
+    {
+      "entropy": 0.6099814623594284,
+      "epoch": 3.019496344435418,
+      "grad_norm": 0.009010368958115578,
+      "learning_rate": 0.0002,
+      "loss": 0.6133589744567871,
+      "mean_token_accuracy": 0.7495384216308594,
+      "num_tokens": 17325896.0,
+      "step": 465
+    },
+    {
+      "entropy": 0.6108173504471779,
+      "epoch": 3.0259951259138913,
+      "grad_norm": 0.008082750253379345,
+      "learning_rate": 0.0002,
+      "loss": 0.6106899380683899,
+      "mean_token_accuracy": 0.7496122866868973,
+      "num_tokens": 17363618.0,
+      "step": 466
+    },
+    {
+      "entropy": 0.6203465834259987,
+      "epoch": 3.032493907392364,
+      "grad_norm": 0.00794974435120821,
+      "learning_rate": 0.0002,
+      "loss": 0.62054044008255,
+      "mean_token_accuracy": 0.7457421347498894,
+      "num_tokens": 17401213.0,
+      "step": 467
+    },
+    {
+      "entropy": 0.6201630383729935,
+      "epoch": 3.0389926888708367,
+      "grad_norm": 0.00949027854949236,
+      "learning_rate": 0.0002,
+      "loss": 0.6232575178146362,
+      "mean_token_accuracy": 0.746864065527916,
+      "num_tokens": 17438437.0,
+      "step": 468
+    },
+    {
+      "entropy": 0.6282624080777168,
+      "epoch": 3.0454914703493094,
+      "grad_norm": 0.007857941091060638,
+      "learning_rate": 0.0002,
+      "loss": 0.6256210207939148,
+      "mean_token_accuracy": 0.7441679313778877,
+      "num_tokens": 17476273.0,
+      "step": 469
+    },
+    {
+      "entropy": 0.6105262786149979,
+      "epoch": 3.051990251827782,
+      "grad_norm": 0.008467303588986397,
+      "learning_rate": 0.0002,
+      "loss": 0.6100492477416992,
+      "mean_token_accuracy": 0.7501196339726448,
+      "num_tokens": 17513496.0,
+      "step": 470
+    },
+    {
+      "entropy": 0.6016805246472359,
+      "epoch": 3.058489033306255,
+      "grad_norm": 0.009371851570904255,
+      "learning_rate": 0.0002,
+      "loss": 0.604375958442688,
+      "mean_token_accuracy": 0.7524744868278503,
+      "num_tokens": 17551200.0,
+      "step": 471
+    },
+    {
+      "entropy": 0.6073808148503304,
+      "epoch": 3.064987814784728,
+      "grad_norm": 0.008492433466017246,
+      "learning_rate": 0.0002,
+      "loss": 0.6094343662261963,
+      "mean_token_accuracy": 0.752588301897049,
+      "num_tokens": 17588356.0,
+      "step": 472
+    },
+    {
+      "entropy": 0.6179896369576454,
+      "epoch": 3.0714865962632008,
+      "grad_norm": 0.008540020324289799,
+      "learning_rate": 0.0002,
+      "loss": 0.6210522055625916,
+      "mean_token_accuracy": 0.7460785582661629,
+      "num_tokens": 17625982.0,
+      "step": 473
+    },
+    {
+      "entropy": 0.6018861308693886,
+      "epoch": 3.0779853777416735,
+      "grad_norm": 0.008742019534111023,
+      "learning_rate": 0.0002,
+      "loss": 0.6078682541847229,
+      "mean_token_accuracy": 0.7513778060674667,
+      "num_tokens": 17663418.0,
+      "step": 474
+    },
+    {
+      "entropy": 0.633279986679554,
+      "epoch": 3.084484159220146,
+      "grad_norm": 0.0082003865391016,
+      "learning_rate": 0.0002,
+      "loss": 0.6310751438140869,
+      "mean_token_accuracy": 0.7428691610693932,
+      "num_tokens": 17700640.0,
+      "step": 475
+    },
+    {
+      "entropy": 0.6102183759212494,
+      "epoch": 3.090982940698619,
+      "grad_norm": 0.008503621444106102,
+      "learning_rate": 0.0002,
+      "loss": 0.6114936470985413,
+      "mean_token_accuracy": 0.7529326751828194,
+      "num_tokens": 17738330.0,
+      "step": 476
+    },
+    {
+      "entropy": 0.6162742525339127,
+      "epoch": 3.0974817221770916,
+      "grad_norm": 0.008936957456171513,
+      "learning_rate": 0.0002,
+      "loss": 0.6209756135940552,
+      "mean_token_accuracy": 0.7465755268931389,
+      "num_tokens": 17775715.0,
+      "step": 477
+    },
+    {
+      "entropy": 0.6141496077179909,
+      "epoch": 3.1039805036555648,
+      "grad_norm": 0.009372946806252003,
+      "learning_rate": 0.0002,
+      "loss": 0.617496132850647,
+      "mean_token_accuracy": 0.7488722130656242,
+      "num_tokens": 17813252.0,
+      "step": 478
+    },
+    {
+      "entropy": 0.6128992438316345,
+      "epoch": 3.1104792851340375,
+      "grad_norm": 0.009246407076716423,
+      "learning_rate": 0.0002,
+      "loss": 0.6153603792190552,
+      "mean_token_accuracy": 0.7489981204271317,
+      "num_tokens": 17850593.0,
+      "step": 479
+    },
+    {
+      "entropy": 0.6109244301915169,
+      "epoch": 3.11697806661251,
+      "grad_norm": 0.008360042236745358,
+      "learning_rate": 0.0002,
+      "loss": 0.6110379099845886,
+      "mean_token_accuracy": 0.7517814785242081,
+      "num_tokens": 17887433.0,
+      "step": 480
+    },
+    {
+      "entropy": 0.6216762438416481,
+      "epoch": 3.123476848090983,
+      "grad_norm": 0.008575825951993465,
+      "learning_rate": 0.0002,
+      "loss": 0.6188209652900696,
+      "mean_token_accuracy": 0.7479925602674484,
+      "num_tokens": 17924477.0,
+      "step": 481
+    },
+    {
+      "entropy": 0.617570586502552,
+      "epoch": 3.1299756295694556,
+      "grad_norm": 0.00846293568611145,
+      "learning_rate": 0.0002,
+      "loss": 0.6159199476242065,
+      "mean_token_accuracy": 0.7518654763698578,
+      "num_tokens": 17961735.0,
+      "step": 482
+    },
+    {
+      "entropy": 0.591516949236393,
+      "epoch": 3.1364744110479283,
+      "grad_norm": 0.009447102434933186,
+      "learning_rate": 0.0002,
+      "loss": 0.5986469984054565,
+      "mean_token_accuracy": 0.7579966932535172,
+      "num_tokens": 17998853.0,
+      "step": 483
+    },
+    {
+      "entropy": 0.6149769946932793,
+      "epoch": 3.1429731925264015,
+      "grad_norm": 0.010652805678546429,
+      "learning_rate": 0.0002,
+      "loss": 0.62315833568573,
+      "mean_token_accuracy": 0.7443839907646179,
+      "num_tokens": 18036498.0,
+      "step": 484
+    },
+    {
+      "entropy": 0.6224213317036629,
+      "epoch": 3.149471974004874,
+      "grad_norm": 0.008841714821755886,
+      "learning_rate": 0.0002,
+      "loss": 0.6213138103485107,
+      "mean_token_accuracy": 0.7460580617189407,
+      "num_tokens": 18073792.0,
+      "step": 485
+    },
+    {
+      "entropy": 0.624303974211216,
+      "epoch": 3.155970755483347,
+      "grad_norm": 0.008476318791508675,
+      "learning_rate": 0.0002,
+      "loss": 0.6206566691398621,
+      "mean_token_accuracy": 0.7465843260288239,
+      "num_tokens": 18110931.0,
+      "step": 486
+    },
+    {
+      "entropy": 0.6057487353682518,
+      "epoch": 3.1624695369618196,
+      "grad_norm": 0.007340795826166868,
+      "learning_rate": 0.0002,
+      "loss": 0.6085374355316162,
+      "mean_token_accuracy": 0.7520734146237373,
+      "num_tokens": 18148230.0,
+      "step": 487
+    },
+    {
+      "entropy": 0.5987089648842812,
+      "epoch": 3.1689683184402924,
+      "grad_norm": 0.008576547726988792,
+      "learning_rate": 0.0002,
+      "loss": 0.5994081497192383,
+      "mean_token_accuracy": 0.7560673728585243,
+      "num_tokens": 18185429.0,
+      "step": 488
+    },
+    {
+      "entropy": 0.6201451569795609,
+      "epoch": 3.175467099918765,
+      "grad_norm": 0.009309273213148117,
+      "learning_rate": 0.0002,
+      "loss": 0.6250134706497192,
+      "mean_token_accuracy": 0.7440272644162178,
+      "num_tokens": 18222845.0,
+      "step": 489
+    },
+    {
+      "entropy": 0.6149864792823792,
+      "epoch": 3.181965881397238,
+      "grad_norm": 0.00817920919507742,
+      "learning_rate": 0.0002,
+      "loss": 0.6165604591369629,
+      "mean_token_accuracy": 0.7469800040125847,
+      "num_tokens": 18260522.0,
+      "step": 490
+    },
+    {
+      "entropy": 0.6081833988428116,
+      "epoch": 3.188464662875711,
+      "grad_norm": 0.008727415464818478,
+      "learning_rate": 0.0002,
+      "loss": 0.6095278263092041,
+      "mean_token_accuracy": 0.7511461302638054,
+      "num_tokens": 18298119.0,
+      "step": 491
+    },
+    {
+      "entropy": 0.6011826619505882,
+      "epoch": 3.1949634443541837,
+      "grad_norm": 0.008403414860367775,
+      "learning_rate": 0.0002,
+      "loss": 0.6023063659667969,
+      "mean_token_accuracy": 0.7541341111063957,
+      "num_tokens": 18335516.0,
+      "step": 492
+    },
+    {
+      "entropy": 0.614462248980999,
+      "epoch": 3.2014622258326564,
+      "grad_norm": 0.008592899888753891,
+      "learning_rate": 0.0002,
+      "loss": 0.611116886138916,
+      "mean_token_accuracy": 0.7512200325727463,
+      "num_tokens": 18373224.0,
+      "step": 493
+    },
+    {
+      "entropy": 0.6118571162223816,
+      "epoch": 3.207961007311129,
+      "grad_norm": 0.008299244567751884,
+      "learning_rate": 0.0002,
+      "loss": 0.6136075258255005,
+      "mean_token_accuracy": 0.749468058347702,
+      "num_tokens": 18410845.0,
+      "step": 494
+    },
+    {
+      "entropy": 0.6049928665161133,
+      "epoch": 3.214459788789602,
+      "grad_norm": 0.010410244576632977,
+      "learning_rate": 0.0002,
+      "loss": 0.6083844304084778,
+      "mean_token_accuracy": 0.751271091401577,
+      "num_tokens": 18448238.0,
+      "step": 495
+    },
+    {
+      "entropy": 0.6233940199017525,
+      "epoch": 3.2209585702680745,
+      "grad_norm": 0.009204378351569176,
+      "learning_rate": 0.0002,
+      "loss": 0.6272657513618469,
+      "mean_token_accuracy": 0.7432866543531418,
+      "num_tokens": 18486015.0,
+      "step": 496
+    },
+    {
+      "entropy": 0.6218753755092621,
+      "epoch": 3.2274573517465477,
+      "grad_norm": 0.00954593438655138,
+      "learning_rate": 0.0002,
+      "loss": 0.6218851208686829,
+      "mean_token_accuracy": 0.7457900047302246,
+      "num_tokens": 18523424.0,
+      "step": 497
+    },
+    {
+      "entropy": 0.6145913749933243,
+      "epoch": 3.2339561332250204,
+      "grad_norm": 0.009425587020814419,
+      "learning_rate": 0.0002,
+      "loss": 0.6123300194740295,
+      "mean_token_accuracy": 0.7488762065768242,
+      "num_tokens": 18560844.0,
+      "step": 498
+    },
+    {
+      "entropy": 0.6112263351678848,
+      "epoch": 3.240454914703493,
+      "grad_norm": 0.009846451692283154,
+      "learning_rate": 0.0002,
+      "loss": 0.6109277606010437,
+      "mean_token_accuracy": 0.7503840625286102,
+      "num_tokens": 18598457.0,
+      "step": 499
+    },
+    {
+      "entropy": 0.6025180518627167,
+      "epoch": 3.246953696181966,
+      "grad_norm": 0.00901762768626213,
+      "learning_rate": 0.0002,
+      "loss": 0.6064941883087158,
+      "mean_token_accuracy": 0.7513880282640457,
+      "num_tokens": 18635799.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.6062363162636757,
+      "epoch": 3.2534524776604385,
+      "grad_norm": 0.008631935343146324,
+      "learning_rate": 0.0002,
+      "loss": 0.610137939453125,
+      "mean_token_accuracy": 0.751276396214962,
+      "num_tokens": 18673367.0,
+      "step": 501
+    },
+    {
+      "entropy": 0.6029994338750839,
+      "epoch": 3.2599512591389113,
+      "grad_norm": 0.008083198219537735,
+      "learning_rate": 0.0002,
+      "loss": 0.605480968952179,
+      "mean_token_accuracy": 0.7522158473730087,
+      "num_tokens": 18710845.0,
+      "step": 502
+    },
+    {
+      "entropy": 0.6004190221428871,
+      "epoch": 3.2664500406173844,
+      "grad_norm": 0.009547607973217964,
+      "learning_rate": 0.0002,
+      "loss": 0.5966814756393433,
+      "mean_token_accuracy": 0.7567247301340103,
+      "num_tokens": 18748254.0,
+      "step": 503
+    },
+    {
+      "entropy": 0.6217415705323219,
+      "epoch": 3.272948822095857,
+      "grad_norm": 0.008327828720211983,
+      "learning_rate": 0.0002,
+      "loss": 0.6205904483795166,
+      "mean_token_accuracy": 0.746347539126873,
+      "num_tokens": 18785876.0,
+      "step": 504
+    },
+    {
+      "entropy": 0.6177750676870346,
+      "epoch": 3.27944760357433,
+      "grad_norm": 0.008332877419888973,
+      "learning_rate": 0.0002,
+      "loss": 0.6175146102905273,
+      "mean_token_accuracy": 0.7480292990803719,
+      "num_tokens": 18823559.0,
+      "step": 505
+    },
+    {
+      "entropy": 0.6048281341791153,
+      "epoch": 3.2859463850528026,
+      "grad_norm": 0.010307500138878822,
+      "learning_rate": 0.0002,
+      "loss": 0.608547568321228,
+      "mean_token_accuracy": 0.7522715479135513,
+      "num_tokens": 18860761.0,
+      "step": 506
+    },
+    {
+      "entropy": 0.6015286520123482,
+      "epoch": 3.2924451665312753,
+      "grad_norm": 0.008793864399194717,
+      "learning_rate": 0.0002,
+      "loss": 0.6076687574386597,
+      "mean_token_accuracy": 0.7535560503602028,
+      "num_tokens": 18898017.0,
+      "step": 507
+    },
+    {
+      "entropy": 0.6203228533267975,
+      "epoch": 3.298943948009748,
+      "grad_norm": 0.008433851413428783,
+      "learning_rate": 0.0002,
+      "loss": 0.627921462059021,
+      "mean_token_accuracy": 0.743441991508007,
+      "num_tokens": 18935451.0,
+      "step": 508
+    },
+    {
+      "entropy": 0.61593446880579,
+      "epoch": 3.305442729488221,
+      "grad_norm": 0.00970753189176321,
+      "learning_rate": 0.0002,
+      "loss": 0.617050290107727,
+      "mean_token_accuracy": 0.7470974624156952,
+      "num_tokens": 18972874.0,
+      "step": 509
+    },
+    {
+      "entropy": 0.6080862656235695,
+      "epoch": 3.311941510966694,
+      "grad_norm": 0.008601192384958267,
+      "learning_rate": 0.0002,
+      "loss": 0.613648533821106,
+      "mean_token_accuracy": 0.7485402300953865,
+      "num_tokens": 19010386.0,
+      "step": 510
+    },
+    {
+      "entropy": 0.6165401563048363,
+      "epoch": 3.3184402924451666,
+      "grad_norm": 0.009453460574150085,
+      "learning_rate": 0.0002,
+      "loss": 0.6146738529205322,
+      "mean_token_accuracy": 0.7489698827266693,
+      "num_tokens": 19048086.0,
+      "step": 511
+    },
+    {
+      "entropy": 0.6097732856869698,
+      "epoch": 3.3249390739236393,
+      "grad_norm": 0.008702411316335201,
+      "learning_rate": 0.0002,
+      "loss": 0.6121372580528259,
+      "mean_token_accuracy": 0.7501634210348129,
+      "num_tokens": 19085297.0,
+      "step": 512
+    },
+    {
+      "entropy": 0.6050214320421219,
+      "epoch": 3.331437855402112,
+      "grad_norm": 0.008775223046541214,
+      "learning_rate": 0.0002,
+      "loss": 0.6089086532592773,
+      "mean_token_accuracy": 0.7513034492731094,
+      "num_tokens": 19122838.0,
+      "step": 513
+    },
+    {
+      "entropy": 0.6079926937818527,
+      "epoch": 3.3379366368805847,
+      "grad_norm": 0.009299266152083874,
+      "learning_rate": 0.0002,
+      "loss": 0.6124639511108398,
+      "mean_token_accuracy": 0.751487672328949,
+      "num_tokens": 19160495.0,
+      "step": 514
+    },
+    {
+      "entropy": 0.6141983196139336,
+      "epoch": 3.3444354183590574,
+      "grad_norm": 0.00953013077378273,
+      "learning_rate": 0.0002,
+      "loss": 0.614948570728302,
+      "mean_token_accuracy": 0.7482777014374733,
+      "num_tokens": 19197772.0,
+      "step": 515
+    },
+    {
+      "entropy": 0.6166125014424324,
+      "epoch": 3.3509341998375306,
+      "grad_norm": 0.01042372826486826,
+      "learning_rate": 0.0002,
+      "loss": 0.6166562438011169,
+      "mean_token_accuracy": 0.7489292994141579,
+      "num_tokens": 19235537.0,
+      "step": 516
+    },
+    {
+      "entropy": 0.6107529476284981,
+      "epoch": 3.3574329813160033,
+      "grad_norm": 0.009182547219097614,
+      "learning_rate": 0.0002,
+      "loss": 0.6094313263893127,
+      "mean_token_accuracy": 0.75286765396595,
+      "num_tokens": 19272557.0,
+      "step": 517
+    },
+    {
+      "entropy": 0.5976028516888618,
+      "epoch": 3.363931762794476,
+      "grad_norm": 0.00928069930523634,
+      "learning_rate": 0.0002,
+      "loss": 0.6000387668609619,
+      "mean_token_accuracy": 0.7556054145097733,
+      "num_tokens": 19309498.0,
+      "step": 518
+    },
+    {
+      "entropy": 0.6023745760321617,
+      "epoch": 3.3704305442729487,
+      "grad_norm": 0.008411030285060406,
+      "learning_rate": 0.0002,
+      "loss": 0.6067807674407959,
+      "mean_token_accuracy": 0.7529009878635406,
+      "num_tokens": 19346735.0,
+      "step": 519
+    },
+    {
+      "entropy": 0.6122680082917213,
+      "epoch": 3.3769293257514215,
+      "grad_norm": 0.008953915908932686,
+      "learning_rate": 0.0002,
+      "loss": 0.6159105896949768,
+      "mean_token_accuracy": 0.7494169026613235,
+      "num_tokens": 19384153.0,
+      "step": 520
+    },
+    {
+      "entropy": 0.6186133921146393,
+      "epoch": 3.3834281072298946,
+      "grad_norm": 0.008217048831284046,
+      "learning_rate": 0.0002,
+      "loss": 0.6203353404998779,
+      "mean_token_accuracy": 0.7473497167229652,
+      "num_tokens": 19421655.0,
+      "step": 521
+    },
+    {
+      "entropy": 0.6094240248203278,
+      "epoch": 3.3899268887083673,
+      "grad_norm": 0.00980299897491932,
+      "learning_rate": 0.0002,
+      "loss": 0.6119738221168518,
+      "mean_token_accuracy": 0.7499380633234978,
+      "num_tokens": 19458894.0,
+      "step": 522
+    },
+    {
+      "entropy": 0.6254991963505745,
+      "epoch": 3.39642567018684,
+      "grad_norm": 0.008415239863097668,
+      "learning_rate": 0.0002,
+      "loss": 0.623593807220459,
+      "mean_token_accuracy": 0.7472219467163086,
+      "num_tokens": 19496818.0,
+      "step": 523
+    },
+    {
+      "entropy": 0.6295105367898941,
+      "epoch": 3.4029244516653128,
+      "grad_norm": 0.009312564507126808,
+      "learning_rate": 0.0002,
+      "loss": 0.6262437105178833,
+      "mean_token_accuracy": 0.7448465451598167,
+      "num_tokens": 19534274.0,
+      "step": 524
+    },
+    {
+      "entropy": 0.6126851290464401,
+      "epoch": 3.4094232331437855,
+      "grad_norm": 0.00861976109445095,
+      "learning_rate": 0.0002,
+      "loss": 0.6097316145896912,
+      "mean_token_accuracy": 0.7525606751441956,
+      "num_tokens": 19571856.0,
+      "step": 525
+    },
+    {
+      "entropy": 0.6135350093245506,
+      "epoch": 3.415922014622258,
+      "grad_norm": 0.008309995755553246,
+      "learning_rate": 0.0002,
+      "loss": 0.6160346269607544,
+      "mean_token_accuracy": 0.7493128851056099,
+      "num_tokens": 19609319.0,
+      "step": 526
+    },
+    {
+      "entropy": 0.6100683584809303,
+      "epoch": 3.422420796100731,
+      "grad_norm": 0.011379679664969444,
+      "learning_rate": 0.0002,
+      "loss": 0.6197912096977234,
+      "mean_token_accuracy": 0.7495440766215324,
+      "num_tokens": 19647036.0,
+      "step": 527
+    },
+    {
+      "entropy": 0.6129766032099724,
+      "epoch": 3.428919577579204,
+      "grad_norm": 0.00879785604774952,
+      "learning_rate": 0.0002,
+      "loss": 0.6192106008529663,
+      "mean_token_accuracy": 0.7462450638413429,
+      "num_tokens": 19684528.0,
+      "step": 528
+    },
+    {
+      "entropy": 0.6183054521679878,
+      "epoch": 3.435418359057677,
+      "grad_norm": 0.009570072405040264,
+      "learning_rate": 0.0002,
+      "loss": 0.61530601978302,
+      "mean_token_accuracy": 0.7494653314352036,
+      "num_tokens": 19721829.0,
+      "step": 529
+    },
+    {
+      "entropy": 0.6237875148653984,
+      "epoch": 3.4419171405361495,
+      "grad_norm": 0.009924475103616714,
+      "learning_rate": 0.0002,
+      "loss": 0.620090126991272,
+      "mean_token_accuracy": 0.7470209151506424,
+      "num_tokens": 19759470.0,
+      "step": 530
+    },
+    {
+      "entropy": 0.6056384146213531,
+      "epoch": 3.448415922014622,
+      "grad_norm": 0.008861211128532887,
+      "learning_rate": 0.0002,
+      "loss": 0.6051967144012451,
+      "mean_token_accuracy": 0.7524856179952621,
+      "num_tokens": 19797055.0,
+      "step": 531
+    },
+    {
+      "entropy": 0.6045270711183548,
+      "epoch": 3.454914703493095,
+      "grad_norm": 0.009311981499195099,
+      "learning_rate": 0.0002,
+      "loss": 0.6111881136894226,
+      "mean_token_accuracy": 0.7494307532906532,
+      "num_tokens": 19834144.0,
+      "step": 532
+    },
+    {
+      "entropy": 0.6097549796104431,
+      "epoch": 3.4614134849715676,
+      "grad_norm": 0.010919635184109211,
+      "learning_rate": 0.0002,
+      "loss": 0.6160416007041931,
+      "mean_token_accuracy": 0.7504306137561798,
+      "num_tokens": 19871286.0,
+      "step": 533
+    },
+    {
+      "entropy": 0.6024704650044441,
+      "epoch": 3.467912266450041,
+      "grad_norm": 0.008416552096605301,
+      "learning_rate": 0.0002,
+      "loss": 0.6058158278465271,
+      "mean_token_accuracy": 0.7539756521582603,
+      "num_tokens": 19908849.0,
+      "step": 534
+    },
+    {
+      "entropy": 0.611225888133049,
+      "epoch": 3.4744110479285135,
+      "grad_norm": 0.008921737782657146,
+      "learning_rate": 0.0002,
+      "loss": 0.6157783269882202,
+      "mean_token_accuracy": 0.7485576719045639,
+      "num_tokens": 19945976.0,
+      "step": 535
+    },
+    {
+      "entropy": 0.5882637947797775,
+      "epoch": 3.4809098294069862,
+      "grad_norm": 0.009470734745264053,
+      "learning_rate": 0.0002,
+      "loss": 0.5855636596679688,
+      "mean_token_accuracy": 0.7615584507584572,
+      "num_tokens": 19983147.0,
+      "step": 536
+    },
+    {
+      "entropy": 0.607180692255497,
+      "epoch": 3.487408610885459,
+      "grad_norm": 0.00895584374666214,
+      "learning_rate": 0.0002,
+      "loss": 0.6034247875213623,
+      "mean_token_accuracy": 0.7542741522192955,
+      "num_tokens": 20020743.0,
+      "step": 537
+    },
+    {
+      "entropy": 0.6082776561379433,
+      "epoch": 3.4939073923639317,
+      "grad_norm": 0.00874309055507183,
+      "learning_rate": 0.0002,
+      "loss": 0.6111589074134827,
+      "mean_token_accuracy": 0.7497387602925301,
+      "num_tokens": 20058341.0,
+      "step": 538
+    },
+    {
+      "entropy": 0.6162533760070801,
+      "epoch": 3.5004061738424044,
+      "grad_norm": 0.009585198014974594,
+      "learning_rate": 0.0002,
+      "loss": 0.6244966387748718,
+      "mean_token_accuracy": 0.7475337833166122,
+      "num_tokens": 20095636.0,
+      "step": 539
+    },
+    {
+      "entropy": 0.6105955466628075,
+      "epoch": 3.506904955320877,
+      "grad_norm": 0.009461835958063602,
+      "learning_rate": 0.0002,
+      "loss": 0.6156846880912781,
+      "mean_token_accuracy": 0.7479056641459465,
+      "num_tokens": 20132551.0,
+      "step": 540
+    },
+    {
+      "entropy": 0.621313750743866,
+      "epoch": 3.5134037367993503,
+      "grad_norm": 0.008541502989828587,
+      "learning_rate": 0.0002,
+      "loss": 0.6199691295623779,
+      "mean_token_accuracy": 0.7473711222410202,
+      "num_tokens": 20169837.0,
+      "step": 541
+    },
+    {
+      "entropy": 0.6165328845381737,
+      "epoch": 3.519902518277823,
+      "grad_norm": 0.009770420379936695,
+      "learning_rate": 0.0002,
+      "loss": 0.6142412424087524,
+      "mean_token_accuracy": 0.7502679750323296,
+      "num_tokens": 20207307.0,
+      "step": 542
+    },
+    {
+      "entropy": 0.6032826527953148,
+      "epoch": 3.5264012997562957,
+      "grad_norm": 0.010346473194658756,
+      "learning_rate": 0.0002,
+      "loss": 0.6011192798614502,
+      "mean_token_accuracy": 0.754242517054081,
+      "num_tokens": 20244685.0,
+      "step": 543
+    },
+    {
+      "entropy": 0.6191048100590706,
+      "epoch": 3.5329000812347684,
+      "grad_norm": 0.008168449625372887,
+      "learning_rate": 0.0002,
+      "loss": 0.6207705736160278,
+      "mean_token_accuracy": 0.74729173630476,
+      "num_tokens": 20281823.0,
+      "step": 544
+    },
+    {
+      "entropy": 0.6121908649802208,
+      "epoch": 3.539398862713241,
+      "grad_norm": 0.007879077456891537,
+      "learning_rate": 0.0002,
+      "loss": 0.6121616363525391,
+      "mean_token_accuracy": 0.7506422027945518,
+      "num_tokens": 20319307.0,
+      "step": 545
+    },
+    {
+      "entropy": 0.6186290234327316,
+      "epoch": 3.5458976441917143,
+      "grad_norm": 0.00905434787273407,
+      "learning_rate": 0.0002,
+      "loss": 0.6185810565948486,
+      "mean_token_accuracy": 0.7475248873233795,
+      "num_tokens": 20356848.0,
+      "step": 546
+    },
+    {
+      "entropy": 0.6139393672347069,
+      "epoch": 3.552396425670187,
+      "grad_norm": 0.008824498392641544,
+      "learning_rate": 0.0002,
+      "loss": 0.6151487827301025,
+      "mean_token_accuracy": 0.7489016950130463,
+      "num_tokens": 20394203.0,
+      "step": 547
+    },
+    {
+      "entropy": 0.6207469254732132,
+      "epoch": 3.5588952071486597,
+      "grad_norm": 0.008396871387958527,
+      "learning_rate": 0.0002,
+      "loss": 0.6218851804733276,
+      "mean_token_accuracy": 0.7459322214126587,
+      "num_tokens": 20431870.0,
+      "step": 548
+    },
+    {
+      "entropy": 0.62169548869133,
+      "epoch": 3.5653939886271324,
+      "grad_norm": 0.00874373223632574,
+      "learning_rate": 0.0002,
+      "loss": 0.6228454113006592,
+      "mean_token_accuracy": 0.749803364276886,
+      "num_tokens": 20469261.0,
+      "step": 549
+    },
+    {
+      "entropy": 0.6220152676105499,
+      "epoch": 3.571892770105605,
+      "grad_norm": 0.008933790028095245,
+      "learning_rate": 0.0002,
+      "loss": 0.6208291053771973,
+      "mean_token_accuracy": 0.7474985867738724,
+      "num_tokens": 20506396.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.6153115481138229,
+      "epoch": 3.578391551584078,
+      "grad_norm": 0.008625194430351257,
+      "learning_rate": 0.0002,
+      "loss": 0.6176726818084717,
+      "mean_token_accuracy": 0.7483843564987183,
+      "num_tokens": 20543594.0,
+      "step": 551
+    },
+    {
+      "entropy": 0.6194273307919502,
+      "epoch": 3.5848903330625506,
+      "grad_norm": 0.009513880126178265,
+      "learning_rate": 0.0002,
+      "loss": 0.6259117126464844,
+      "mean_token_accuracy": 0.7455350533127785,
+      "num_tokens": 20580976.0,
+      "step": 552
+    },
+    {
+      "entropy": 0.6218870356678963,
+      "epoch": 3.5913891145410237,
+      "grad_norm": 0.010471499525010586,
+      "learning_rate": 0.0002,
+      "loss": 0.6255429983139038,
+      "mean_token_accuracy": 0.745422936975956,
+      "num_tokens": 20618155.0,
+      "step": 553
+    },
+    {
+      "entropy": 0.6044272035360336,
+      "epoch": 3.5978878960194964,
+      "grad_norm": 0.009206542745232582,
+      "learning_rate": 0.0002,
+      "loss": 0.6089117527008057,
+      "mean_token_accuracy": 0.7523582801222801,
+      "num_tokens": 20655356.0,
+      "step": 554
+    },
+    {
+      "entropy": 0.6034605354070663,
+      "epoch": 3.604386677497969,
+      "grad_norm": 0.01001272164285183,
+      "learning_rate": 0.0002,
+      "loss": 0.6080043315887451,
+      "mean_token_accuracy": 0.7542883977293968,
+      "num_tokens": 20692368.0,
+      "step": 555
+    },
+    {
+      "entropy": 0.6126068532466888,
+      "epoch": 3.610885458976442,
+      "grad_norm": 0.008411637507379055,
+      "learning_rate": 0.0002,
+      "loss": 0.6143564581871033,
+      "mean_token_accuracy": 0.7496707290410995,
+      "num_tokens": 20729532.0,
+      "step": 556
+    },
+    {
+      "entropy": 0.6092226430773735,
+      "epoch": 3.6173842404549146,
+      "grad_norm": 0.00788223184645176,
+      "learning_rate": 0.0002,
+      "loss": 0.6080750226974487,
+      "mean_token_accuracy": 0.7521450445055962,
+      "num_tokens": 20767316.0,
+      "step": 557
+    },
+    {
+      "entropy": 0.6103739887475967,
+      "epoch": 3.6238830219333877,
+      "grad_norm": 0.008947784081101418,
+      "learning_rate": 0.0002,
+      "loss": 0.6110090613365173,
+      "mean_token_accuracy": 0.751432754099369,
+      "num_tokens": 20804682.0,
+      "step": 558
+    },
+    {
+      "entropy": 0.6072618737816811,
+      "epoch": 3.6303818034118605,
+      "grad_norm": 0.008321145549416542,
+      "learning_rate": 0.0002,
+      "loss": 0.6072537899017334,
+      "mean_token_accuracy": 0.7519486770033836,
+      "num_tokens": 20842038.0,
+      "step": 559
+    },
+    {
+      "entropy": 0.614497110247612,
+      "epoch": 3.636880584890333,
+      "grad_norm": 0.009164758026599884,
+      "learning_rate": 0.0002,
+      "loss": 0.6098246574401855,
+      "mean_token_accuracy": 0.752671018242836,
+      "num_tokens": 20879197.0,
+      "step": 560
+    },
+    {
+      "entropy": 0.6094550266861916,
+      "epoch": 3.643379366368806,
+      "grad_norm": 0.007844182662665844,
+      "learning_rate": 0.0002,
+      "loss": 0.6068869829177856,
+      "mean_token_accuracy": 0.7533254846930504,
+      "num_tokens": 20916501.0,
+      "step": 561
+    },
+    {
+      "entropy": 0.6086015179753304,
+      "epoch": 3.6498781478472786,
+      "grad_norm": 0.008045290596783161,
+      "learning_rate": 0.0002,
+      "loss": 0.606809139251709,
+      "mean_token_accuracy": 0.7527647316455841,
+      "num_tokens": 20954134.0,
+      "step": 562
+    },
+    {
+      "entropy": 0.6285906434059143,
+      "epoch": 3.6563769293257513,
+      "grad_norm": 0.010104804299771786,
+      "learning_rate": 0.0002,
+      "loss": 0.6331174373626709,
+      "mean_token_accuracy": 0.7419461309909821,
+      "num_tokens": 20992017.0,
+      "step": 563
+    },
+    {
+      "entropy": 0.6020278707146645,
+      "epoch": 3.662875710804224,
+      "grad_norm": 0.01021239347755909,
+      "learning_rate": 0.0002,
+      "loss": 0.6128239035606384,
+      "mean_token_accuracy": 0.7514787241816521,
+      "num_tokens": 21029653.0,
+      "step": 564
+    },
+    {
+      "entropy": 0.6191821396350861,
+      "epoch": 3.6693744922826967,
+      "grad_norm": 0.009272433817386627,
+      "learning_rate": 0.0002,
+      "loss": 0.6228241324424744,
+      "mean_token_accuracy": 0.7464863955974579,
+      "num_tokens": 21067220.0,
+      "step": 565
+    },
+    {
+      "entropy": 0.6111209839582443,
+      "epoch": 3.67587327376117,
+      "grad_norm": 0.01010363083332777,
+      "learning_rate": 0.0002,
+      "loss": 0.6085663437843323,
+      "mean_token_accuracy": 0.7533030211925507,
+      "num_tokens": 21105021.0,
+      "step": 566
+    },
+    {
+      "entropy": 0.6244448721408844,
+      "epoch": 3.6823720552396426,
+      "grad_norm": 0.008229166269302368,
+      "learning_rate": 0.0002,
+      "loss": 0.6232364177703857,
+      "mean_token_accuracy": 0.7465489506721497,
+      "num_tokens": 21142672.0,
+      "step": 567
+    },
+    {
+      "entropy": 0.6002631187438965,
+      "epoch": 3.6888708367181153,
+      "grad_norm": 0.00970839336514473,
+      "learning_rate": 0.0002,
+      "loss": 0.6064876914024353,
+      "mean_token_accuracy": 0.7520908117294312,
+      "num_tokens": 21179768.0,
+      "step": 568
+    },
+    {
+      "entropy": 0.606594480574131,
+      "epoch": 3.695369618196588,
+      "grad_norm": 0.00952630303800106,
+      "learning_rate": 0.0002,
+      "loss": 0.6108591556549072,
+      "mean_token_accuracy": 0.751940131187439,
+      "num_tokens": 21217131.0,
+      "step": 569
+    },
+    {
+      "entropy": 0.6083743423223495,
+      "epoch": 3.7018683996750608,
+      "grad_norm": 0.008894224651157856,
+      "learning_rate": 0.0002,
+      "loss": 0.6050914525985718,
+      "mean_token_accuracy": 0.7556580603122711,
+      "num_tokens": 21254612.0,
+      "step": 570
+    },
+    {
+      "entropy": 0.6244188398122787,
+      "epoch": 3.708367181153534,
+      "grad_norm": 0.00945176463574171,
+      "learning_rate": 0.0002,
+      "loss": 0.6213344931602478,
+      "mean_token_accuracy": 0.7479386925697327,
+      "num_tokens": 21292008.0,
+      "step": 571
+    },
+    {
+      "entropy": 0.6202782392501831,
+      "epoch": 3.7148659626320066,
+      "grad_norm": 0.007808351423591375,
+      "learning_rate": 0.0002,
+      "loss": 0.618791937828064,
+      "mean_token_accuracy": 0.7479956671595573,
+      "num_tokens": 21329211.0,
+      "step": 572
+    },
+    {
+      "entropy": 0.6083357259631157,
+      "epoch": 3.7213647441104794,
+      "grad_norm": 0.009738109074532986,
+      "learning_rate": 0.0002,
+      "loss": 0.6082479953765869,
+      "mean_token_accuracy": 0.7511337921023369,
+      "num_tokens": 21366830.0,
+      "step": 573
+    },
+    {
+      "entropy": 0.6160146668553352,
+      "epoch": 3.727863525588952,
+      "grad_norm": 0.008708451874554157,
+      "learning_rate": 0.0002,
+      "loss": 0.6203808784484863,
+      "mean_token_accuracy": 0.749162532389164,
+      "num_tokens": 21404012.0,
+      "step": 574
+    },
+    {
+      "entropy": 0.6103571355342865,
+      "epoch": 3.734362307067425,
+      "grad_norm": 0.008490422740578651,
+      "learning_rate": 0.0002,
+      "loss": 0.6149444580078125,
+      "mean_token_accuracy": 0.7492819800972939,
+      "num_tokens": 21441418.0,
+      "step": 575
+    },
+    {
+      "entropy": 0.6097893342375755,
+      "epoch": 3.7408610885458975,
+      "grad_norm": 0.008159926161170006,
+      "learning_rate": 0.0002,
+      "loss": 0.6121622323989868,
+      "mean_token_accuracy": 0.7502710297703743,
+      "num_tokens": 21478959.0,
+      "step": 576
+    },
+    {
+      "entropy": 0.608406588435173,
+      "epoch": 3.74735987002437,
+      "grad_norm": 0.007408132776618004,
+      "learning_rate": 0.0002,
+      "loss": 0.6112476587295532,
+      "mean_token_accuracy": 0.7514019384980202,
+      "num_tokens": 21516469.0,
+      "step": 577
+    },
+    {
+      "entropy": 0.6059915721416473,
+      "epoch": 3.7538586515028434,
+      "grad_norm": 0.007707294542342424,
+      "learning_rate": 0.0002,
+      "loss": 0.6102927923202515,
+      "mean_token_accuracy": 0.7524895742535591,
+      "num_tokens": 21553955.0,
+      "step": 578
+    },
+    {
+      "entropy": 0.6110079139471054,
+      "epoch": 3.760357432981316,
+      "grad_norm": 0.008303117007017136,
+      "learning_rate": 0.0002,
+      "loss": 0.6116464734077454,
+      "mean_token_accuracy": 0.7515735402703285,
+      "num_tokens": 21591222.0,
+      "step": 579
+    },
+    {
+      "entropy": 0.6157199367880821,
+      "epoch": 3.766856214459789,
+      "grad_norm": 0.008788561448454857,
+      "learning_rate": 0.0002,
+      "loss": 0.6139419078826904,
+      "mean_token_accuracy": 0.7500691562891006,
+      "num_tokens": 21628586.0,
+      "step": 580
+    },
+    {
+      "entropy": 0.6189565733075142,
+      "epoch": 3.7733549959382615,
+      "grad_norm": 0.008861690759658813,
+      "learning_rate": 0.0002,
+      "loss": 0.6205735802650452,
+      "mean_token_accuracy": 0.7461571544408798,
+      "num_tokens": 21665902.0,
+      "step": 581
+    },
+    {
+      "entropy": 0.6169708073139191,
+      "epoch": 3.7798537774167342,
+      "grad_norm": 0.00858966913074255,
+      "learning_rate": 0.0002,
+      "loss": 0.616493284702301,
+      "mean_token_accuracy": 0.7478131875395775,
+      "num_tokens": 21703347.0,
+      "step": 582
+    },
+    {
+      "entropy": 0.6013945862650871,
+      "epoch": 3.7863525588952074,
+      "grad_norm": 0.007991503924131393,
+      "learning_rate": 0.0002,
+      "loss": 0.6024019718170166,
+      "mean_token_accuracy": 0.7541673332452774,
+      "num_tokens": 21740476.0,
+      "step": 583
+    },
+    {
+      "entropy": 0.6127390787005424,
+      "epoch": 3.79285134037368,
+      "grad_norm": 0.008759486488997936,
+      "learning_rate": 0.0002,
+      "loss": 0.6142678260803223,
+      "mean_token_accuracy": 0.7489368915557861,
+      "num_tokens": 21777903.0,
+      "step": 584
+    },
+    {
+      "entropy": 0.6033598631620407,
+      "epoch": 3.799350121852153,
+      "grad_norm": 0.008402034640312195,
+      "learning_rate": 0.0002,
+      "loss": 0.6062989234924316,
+      "mean_token_accuracy": 0.7526213079690933,
+      "num_tokens": 21815411.0,
+      "step": 585
+    },
+    {
+      "entropy": 0.6237991526722908,
+      "epoch": 3.8058489033306255,
+      "grad_norm": 0.008741356432437897,
+      "learning_rate": 0.0002,
+      "loss": 0.6236557960510254,
+      "mean_token_accuracy": 0.7458471432328224,
+      "num_tokens": 21852563.0,
+      "step": 586
+    },
+    {
+      "entropy": 0.6021682396531105,
+      "epoch": 3.8123476848090982,
+      "grad_norm": 0.009454802609980106,
+      "learning_rate": 0.0002,
+      "loss": 0.6056302785873413,
+      "mean_token_accuracy": 0.7534916624426842,
+      "num_tokens": 21889751.0,
+      "step": 587
+    },
+    {
+      "entropy": 0.6037532165646553,
+      "epoch": 3.818846466287571,
+      "grad_norm": 0.008019194938242435,
+      "learning_rate": 0.0002,
+      "loss": 0.6062440872192383,
+      "mean_token_accuracy": 0.75218416005373,
+      "num_tokens": 21927225.0,
+      "step": 588
+    },
+    {
+      "entropy": 0.6175741031765938,
+      "epoch": 3.8253452477660437,
+      "grad_norm": 0.010461182333528996,
+      "learning_rate": 0.0002,
+      "loss": 0.6219437122344971,
+      "mean_token_accuracy": 0.7449553832411766,
+      "num_tokens": 21964899.0,
+      "step": 589
+    },
+    {
+      "entropy": 0.6165783628821373,
+      "epoch": 3.8318440292445164,
+      "grad_norm": 0.007971042767167091,
+      "learning_rate": 0.0002,
+      "loss": 0.6164169311523438,
+      "mean_token_accuracy": 0.7500692084431648,
+      "num_tokens": 22001937.0,
+      "step": 590
+    },
+    {
+      "entropy": 0.6328809410333633,
+      "epoch": 3.8383428107229896,
+      "grad_norm": 0.019222697243094444,
+      "learning_rate": 0.0002,
+      "loss": 0.629861056804657,
+      "mean_token_accuracy": 0.7441875413060188,
+      "num_tokens": 22039681.0,
+      "step": 591
+    },
+    {
+      "entropy": 0.6060463637113571,
+      "epoch": 3.8448415922014623,
+      "grad_norm": 0.008333687670528889,
+      "learning_rate": 0.0002,
+      "loss": 0.6049051880836487,
+      "mean_token_accuracy": 0.7526820749044418,
+      "num_tokens": 22077083.0,
+      "step": 592
+    },
+    {
+      "entropy": 0.626244343817234,
+      "epoch": 3.851340373679935,
+      "grad_norm": 0.011930772103369236,
+      "learning_rate": 0.0002,
+      "loss": 0.6302248239517212,
+      "mean_token_accuracy": 0.7435696348547935,
+      "num_tokens": 22114449.0,
+      "step": 593
+    },
+    {
+      "entropy": 0.60886500030756,
+      "epoch": 3.8578391551584077,
+      "grad_norm": 0.00879514031112194,
+      "learning_rate": 0.0002,
+      "loss": 0.6156384944915771,
+      "mean_token_accuracy": 0.7500656992197037,
+      "num_tokens": 22151803.0,
+      "step": 594
+    },
+    {
+      "entropy": 0.6130397766828537,
+      "epoch": 3.864337936636881,
+      "grad_norm": 0.008800708688795567,
+      "learning_rate": 0.0002,
+      "loss": 0.6132546663284302,
+      "mean_token_accuracy": 0.7525083422660828,
+      "num_tokens": 22188933.0,
+      "step": 595
+    },
+    {
+      "entropy": 0.62311190366745,
+      "epoch": 3.8708367181153536,
+      "grad_norm": 0.0075268070213496685,
+      "learning_rate": 0.0002,
+      "loss": 0.624302864074707,
+      "mean_token_accuracy": 0.7459553778171539,
+      "num_tokens": 22226349.0,
+      "step": 596
+    },
+    {
+      "entropy": 0.6084966883063316,
+      "epoch": 3.8773354995938263,
+      "grad_norm": 0.008235939778387547,
+      "learning_rate": 0.0002,
+      "loss": 0.6065037250518799,
+      "mean_token_accuracy": 0.7519035413861275,
+      "num_tokens": 22263929.0,
+      "step": 597
+    },
+    {
+      "entropy": 0.599085658788681,
+      "epoch": 3.883834281072299,
+      "grad_norm": 0.008155166171491146,
+      "learning_rate": 0.0002,
+      "loss": 0.60195392370224,
+      "mean_token_accuracy": 0.7546913474798203,
+      "num_tokens": 22301001.0,
+      "step": 598
+    },
+    {
+      "entropy": 0.6163361445069313,
+      "epoch": 3.8903330625507717,
+      "grad_norm": 0.009046819061040878,
+      "learning_rate": 0.0002,
+      "loss": 0.6153905987739563,
+      "mean_token_accuracy": 0.7495288848876953,
+      "num_tokens": 22338602.0,
+      "step": 599
+    },
+    {
+      "entropy": 0.612362690269947,
+      "epoch": 3.8968318440292444,
+      "grad_norm": 0.008970526047050953,
+      "learning_rate": 0.0002,
+      "loss": 0.6125571131706238,
+      "mean_token_accuracy": 0.7503594160079956,
+      "num_tokens": 22375832.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.5970166251063347,
+      "epoch": 3.903330625507717,
+      "grad_norm": 0.010594755411148071,
+      "learning_rate": 0.0002,
+      "loss": 0.6027450561523438,
+      "mean_token_accuracy": 0.7523306459188461,
+      "num_tokens": 22412582.0,
+      "step": 601
+    },
+    {
+      "entropy": 0.6120518818497658,
+      "epoch": 3.90982940698619,
+      "grad_norm": 0.008573422208428383,
+      "learning_rate": 0.0002,
+      "loss": 0.619284451007843,
+      "mean_token_accuracy": 0.7469367831945419,
+      "num_tokens": 22449548.0,
+      "step": 602
+    },
+    {
+      "entropy": 0.6074809059500694,
+      "epoch": 3.916328188464663,
+      "grad_norm": 0.008411690592765808,
+      "learning_rate": 0.0002,
+      "loss": 0.609643280506134,
+      "mean_token_accuracy": 0.7513059079647064,
+      "num_tokens": 22486961.0,
+      "step": 603
+    },
+    {
+      "entropy": 0.6189433708786964,
+      "epoch": 3.9228269699431357,
+      "grad_norm": 0.008104969747364521,
+      "learning_rate": 0.0002,
+      "loss": 0.6152623891830444,
+      "mean_token_accuracy": 0.7489224672317505,
+      "num_tokens": 22524327.0,
+      "step": 604
+    },
+    {
+      "entropy": 0.6386607512831688,
+      "epoch": 3.9293257514216084,
+      "grad_norm": 0.007721316535025835,
+      "learning_rate": 0.0002,
+      "loss": 0.6365233659744263,
+      "mean_token_accuracy": 0.7393748313188553,
+      "num_tokens": 22561540.0,
+      "step": 605
+    },
+    {
+      "entropy": 0.6104080379009247,
+      "epoch": 3.935824532900081,
+      "grad_norm": 0.009160472080111504,
+      "learning_rate": 0.0002,
+      "loss": 0.6058955192565918,
+      "mean_token_accuracy": 0.7527115345001221,
+      "num_tokens": 22598714.0,
+      "step": 606
+    },
+    {
+      "entropy": 0.6066436171531677,
+      "epoch": 3.942323314378554,
+      "grad_norm": 0.00924257654696703,
+      "learning_rate": 0.0002,
+      "loss": 0.6112688779830933,
+      "mean_token_accuracy": 0.7510428801178932,
+      "num_tokens": 22636168.0,
+      "step": 607
+    },
+    {
+      "entropy": 0.6113385632634163,
+      "epoch": 3.948822095857027,
+      "grad_norm": 0.009213161654770374,
+      "learning_rate": 0.0002,
+      "loss": 0.6169242858886719,
+      "mean_token_accuracy": 0.7488819509744644,
+      "num_tokens": 22673222.0,
+      "step": 608
+    },
+    {
+      "entropy": 0.6157431975007057,
+      "epoch": 3.9553208773354998,
+      "grad_norm": 0.00805346667766571,
+      "learning_rate": 0.0002,
+      "loss": 0.6147987842559814,
+      "mean_token_accuracy": 0.7497100383043289,
+      "num_tokens": 22710362.0,
+      "step": 609
+    },
+    {
+      "entropy": 0.6155053228139877,
+      "epoch": 3.9618196588139725,
+      "grad_norm": 0.008978918194770813,
+      "learning_rate": 0.0002,
+      "loss": 0.614108681678772,
+      "mean_token_accuracy": 0.7489725723862648,
+      "num_tokens": 22747892.0,
+      "step": 610
+    },
+    {
+      "entropy": 0.6131211444735527,
+      "epoch": 3.968318440292445,
+      "grad_norm": 0.010324759408831596,
+      "learning_rate": 0.0002,
+      "loss": 0.6139857769012451,
+      "mean_token_accuracy": 0.7501854747533798,
+      "num_tokens": 22784987.0,
+      "step": 611
+    },
+    {
+      "entropy": 0.6161521598696709,
+      "epoch": 3.974817221770918,
+      "grad_norm": 0.009645137935876846,
+      "learning_rate": 0.0002,
+      "loss": 0.6184964179992676,
+      "mean_token_accuracy": 0.748919740319252,
+      "num_tokens": 22822604.0,
+      "step": 612
+    },
+    {
+      "entropy": 0.6195979118347168,
+      "epoch": 3.9813160032493906,
+      "grad_norm": 0.008733006194233894,
+      "learning_rate": 0.0002,
+      "loss": 0.6204901933670044,
+      "mean_token_accuracy": 0.7458331510424614,
+      "num_tokens": 22859577.0,
+      "step": 613
+    },
+    {
+      "entropy": 0.6223805174231529,
+      "epoch": 3.9878147847278633,
+      "grad_norm": 0.008579275570809841,
+      "learning_rate": 0.0002,
+      "loss": 0.6251469850540161,
+      "mean_token_accuracy": 0.7441524565219879,
+      "num_tokens": 22897065.0,
+      "step": 614
+    },
+    {
+      "entropy": 0.6099046468734741,
+      "epoch": 3.994313566206336,
+      "grad_norm": 0.009405579417943954,
+      "learning_rate": 0.0002,
+      "loss": 0.6134316921234131,
+      "mean_token_accuracy": 0.7505489960312843,
+      "num_tokens": 22934240.0,
+      "step": 615
+    },
+    {
+      "entropy": 0.5790954487664359,
+      "epoch": 4.0,
+      "grad_norm": 0.011152846738696098,
+      "learning_rate": 0.0002,
+      "loss": 0.6086491346359253,
+      "mean_token_accuracy": 0.7593175172805786,
+      "num_tokens": 22950408.0,
+      "step": 616
+    },
+    {
+      "entropy": 0.6060308367013931,
+      "epoch": 4.006498781478473,
+      "grad_norm": 0.008177882991731167,
+      "learning_rate": 0.0002,
+      "loss": 0.6047600507736206,
+      "mean_token_accuracy": 0.7523083239793777,
+      "num_tokens": 22987854.0,
+      "step": 617
+    },
+    {
+      "entropy": 0.60086639970541,
+      "epoch": 4.012997562956945,
+      "grad_norm": 0.009073281660676003,
+      "learning_rate": 0.0002,
+      "loss": 0.5952266454696655,
+      "mean_token_accuracy": 0.7580123767256737,
+      "num_tokens": 23025377.0,
+      "step": 618
+    },
+    {
+      "entropy": 0.5975999683141708,
+      "epoch": 4.019496344435418,
+      "grad_norm": 0.00987546518445015,
+      "learning_rate": 0.0002,
+      "loss": 0.596405565738678,
+      "mean_token_accuracy": 0.7551270872354507,
+      "num_tokens": 23062634.0,
+      "step": 619
+    },
+    {
+      "entropy": 0.6007415279746056,
+      "epoch": 4.025995125913891,
+      "grad_norm": 0.008907251060009003,
+      "learning_rate": 0.0002,
+      "loss": 0.6056969165802002,
+      "mean_token_accuracy": 0.752616822719574,
+      "num_tokens": 23099758.0,
+      "step": 620
+    },
+    {
+      "entropy": 0.5893639698624611,
+      "epoch": 4.032493907392364,
+      "grad_norm": 0.010224535129964352,
+      "learning_rate": 0.0002,
+      "loss": 0.5967025756835938,
+      "mean_token_accuracy": 0.7561892941594124,
+      "num_tokens": 23137072.0,
+      "step": 621
+    },
+    {
+      "entropy": 0.60489272326231,
+      "epoch": 4.038992688870836,
+      "grad_norm": 0.011304826475679874,
+      "learning_rate": 0.0002,
+      "loss": 0.6063746213912964,
+      "mean_token_accuracy": 0.7524363398551941,
+      "num_tokens": 23174098.0,
+      "step": 622
+    },
+    {
+      "entropy": 0.6080682426691055,
+      "epoch": 4.04549147034931,
+      "grad_norm": 0.009452976286411285,
+      "learning_rate": 0.0002,
+      "loss": 0.6047728061676025,
+      "mean_token_accuracy": 0.7509661018848419,
+      "num_tokens": 23211396.0,
+      "step": 623
+    },
+    {
+      "entropy": 0.6045035719871521,
+      "epoch": 4.051990251827783,
+      "grad_norm": 0.010957913473248482,
+      "learning_rate": 0.0002,
+      "loss": 0.6045751571655273,
+      "mean_token_accuracy": 0.7534721195697784,
+      "num_tokens": 23248640.0,
+      "step": 624
+    },
+    {
+      "entropy": 0.6037070825695992,
+      "epoch": 4.058489033306255,
+      "grad_norm": 0.011129945516586304,
+      "learning_rate": 0.0002,
+      "loss": 0.6084102392196655,
+      "mean_token_accuracy": 0.7514655515551567,
+      "num_tokens": 23286103.0,
+      "step": 625
+    },
+    {
+      "entropy": 0.6091515198349953,
+      "epoch": 4.064987814784728,
+      "grad_norm": 0.011563125997781754,
+      "learning_rate": 0.0002,
+      "loss": 0.6175334453582764,
+      "mean_token_accuracy": 0.7468038350343704,
+      "num_tokens": 23323689.0,
+      "step": 626
+    },
+    {
+      "entropy": 0.624121755361557,
+      "epoch": 4.071486596263201,
+      "grad_norm": 0.010375406593084335,
+      "learning_rate": 0.0002,
+      "loss": 0.6241302490234375,
+      "mean_token_accuracy": 0.7438605725765228,
+      "num_tokens": 23361119.0,
+      "step": 627
+    },
+    {
+      "entropy": 0.6100181490182877,
+      "epoch": 4.0779853777416735,
+      "grad_norm": 0.011141455732285976,
+      "learning_rate": 0.0002,
+      "loss": 0.6070010662078857,
+      "mean_token_accuracy": 0.7539649158716202,
+      "num_tokens": 23398396.0,
+      "step": 628
+    },
+    {
+      "entropy": 0.6109626740217209,
+      "epoch": 4.084484159220146,
+      "grad_norm": 0.008930398151278496,
+      "learning_rate": 0.0002,
+      "loss": 0.610953688621521,
+      "mean_token_accuracy": 0.7503283470869064,
+      "num_tokens": 23435949.0,
+      "step": 629
+    },
+    {
+      "entropy": 0.5975475162267685,
+      "epoch": 4.090982940698619,
+      "grad_norm": 0.00977004412561655,
+      "learning_rate": 0.0002,
+      "loss": 0.6004647612571716,
+      "mean_token_accuracy": 0.755259171128273,
+      "num_tokens": 23473365.0,
+      "step": 630
+    },
+    {
+      "entropy": 0.5996551141142845,
+      "epoch": 4.097481722177092,
+      "grad_norm": 0.009881756268441677,
+      "learning_rate": 0.0002,
+      "loss": 0.6020142436027527,
+      "mean_token_accuracy": 0.7552994042634964,
+      "num_tokens": 23510883.0,
+      "step": 631
+    },
+    {
+      "entropy": 0.6100725084543228,
+      "epoch": 4.103980503655564,
+      "grad_norm": 0.009554548189043999,
+      "learning_rate": 0.0002,
+      "loss": 0.60805743932724,
+      "mean_token_accuracy": 0.749746523797512,
+      "num_tokens": 23548279.0,
+      "step": 632
+    },
+    {
+      "entropy": 0.604118101298809,
+      "epoch": 4.110479285134037,
+      "grad_norm": 0.0090833380818367,
+      "learning_rate": 0.0002,
+      "loss": 0.6060141324996948,
+      "mean_token_accuracy": 0.7535399347543716,
+      "num_tokens": 23585460.0,
+      "step": 633
+    },
+    {
+      "entropy": 0.6112289279699326,
+      "epoch": 4.11697806661251,
+      "grad_norm": 0.009132446721196175,
+      "learning_rate": 0.0002,
+      "loss": 0.6089117527008057,
+      "mean_token_accuracy": 0.7522810772061348,
+      "num_tokens": 23623069.0,
+      "step": 634
+    },
+    {
+      "entropy": 0.5981876626610756,
+      "epoch": 4.123476848090983,
+      "grad_norm": 0.009694000706076622,
+      "learning_rate": 0.0002,
+      "loss": 0.6000718474388123,
+      "mean_token_accuracy": 0.7525889277458191,
+      "num_tokens": 23660201.0,
+      "step": 635
+    },
+    {
+      "entropy": 0.6080240532755852,
+      "epoch": 4.129975629569456,
+      "grad_norm": 0.009349804371595383,
+      "learning_rate": 0.0002,
+      "loss": 0.6115583181381226,
+      "mean_token_accuracy": 0.7498094886541367,
+      "num_tokens": 23697467.0,
+      "step": 636
+    },
+    {
+      "entropy": 0.6051702946424484,
+      "epoch": 4.136474411047929,
+      "grad_norm": 0.00995174515992403,
+      "learning_rate": 0.0002,
+      "loss": 0.6054593324661255,
+      "mean_token_accuracy": 0.7512011528015137,
+      "num_tokens": 23734887.0,
+      "step": 637
+    },
+    {
+      "entropy": 0.6021575108170509,
+      "epoch": 4.1429731925264015,
+      "grad_norm": 0.011656476184725761,
+      "learning_rate": 0.0002,
+      "loss": 0.6069491505622864,
+      "mean_token_accuracy": 0.7523794695734978,
+      "num_tokens": 23772003.0,
+      "step": 638
+    },
+    {
+      "entropy": 0.6030121594667435,
+      "epoch": 4.149471974004874,
+      "grad_norm": 0.009891415014863014,
+      "learning_rate": 0.0002,
+      "loss": 0.604573667049408,
+      "mean_token_accuracy": 0.7512457966804504,
+      "num_tokens": 23809089.0,
+      "step": 639
+    },
+    {
+      "entropy": 0.5972474366426468,
+      "epoch": 4.155970755483347,
+      "grad_norm": 0.011574115604162216,
+      "learning_rate": 0.0002,
+      "loss": 0.5955284833908081,
+      "mean_token_accuracy": 0.7579837068915367,
+      "num_tokens": 23846256.0,
+      "step": 640
+    },
+    {
+      "entropy": 0.6166966333985329,
+      "epoch": 4.16246953696182,
+      "grad_norm": 0.010048413649201393,
+      "learning_rate": 0.0002,
+      "loss": 0.6198529005050659,
+      "mean_token_accuracy": 0.7490389570593834,
+      "num_tokens": 23884254.0,
+      "step": 641
+    },
+    {
+      "entropy": 0.5962498784065247,
+      "epoch": 4.168968318440292,
+      "grad_norm": 0.011094493791460991,
+      "learning_rate": 0.0002,
+      "loss": 0.6002663969993591,
+      "mean_token_accuracy": 0.7531006932258606,
+      "num_tokens": 23921748.0,
+      "step": 642
+    },
+    {
+      "entropy": 0.6103070974349976,
+      "epoch": 4.175467099918765,
+      "grad_norm": 0.0102005023509264,
+      "learning_rate": 0.0002,
+      "loss": 0.6120426654815674,
+      "mean_token_accuracy": 0.7472726628184319,
+      "num_tokens": 23959118.0,
+      "step": 643
+    },
+    {
+      "entropy": 0.6111981943249702,
+      "epoch": 4.181965881397238,
+      "grad_norm": 0.010743658058345318,
+      "learning_rate": 0.0002,
+      "loss": 0.608763575553894,
+      "mean_token_accuracy": 0.7516906931996346,
+      "num_tokens": 23996874.0,
+      "step": 644
+    },
+    {
+      "entropy": 0.6073967814445496,
+      "epoch": 4.1884646628757105,
+      "grad_norm": 0.011524532921612263,
+      "learning_rate": 0.0002,
+      "loss": 0.6078468561172485,
+      "mean_token_accuracy": 0.750710740685463,
+      "num_tokens": 24034180.0,
+      "step": 645
+    },
+    {
+      "entropy": 0.6122067421674728,
+      "epoch": 4.194963444354183,
+      "grad_norm": 0.009897702373564243,
+      "learning_rate": 0.0002,
+      "loss": 0.613964319229126,
+      "mean_token_accuracy": 0.7508190050721169,
+      "num_tokens": 24071959.0,
+      "step": 646
+    },
+    {
+      "entropy": 0.6176895946264267,
+      "epoch": 4.201462225832657,
+      "grad_norm": 0.010724211111664772,
+      "learning_rate": 0.0002,
+      "loss": 0.6206725835800171,
+      "mean_token_accuracy": 0.7463300824165344,
+      "num_tokens": 24109459.0,
+      "step": 647
+    },
+    {
+      "entropy": 0.6166289448738098,
+      "epoch": 4.2079610073111295,
+      "grad_norm": 0.011503348127007484,
+      "learning_rate": 0.0002,
+      "loss": 0.6165927052497864,
+      "mean_token_accuracy": 0.7491984441876411,
+      "num_tokens": 24146901.0,
+      "step": 648
+    },
+    {
+      "entropy": 0.6203835979104042,
+      "epoch": 4.214459788789602,
+      "grad_norm": 0.010957739315927029,
+      "learning_rate": 0.0002,
+      "loss": 0.6207370758056641,
+      "mean_token_accuracy": 0.7465796917676926,
+      "num_tokens": 24184105.0,
+      "step": 649
+    },
+    {
+      "entropy": 0.608573280274868,
+      "epoch": 4.220958570268075,
+      "grad_norm": 0.012403266504406929,
+      "learning_rate": 0.0002,
+      "loss": 0.6109322309494019,
+      "mean_token_accuracy": 0.7499670162796974,
+      "num_tokens": 24221373.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.61744424700737,
+      "epoch": 4.227457351746548,
+      "grad_norm": 0.010664467699825764,
+      "learning_rate": 0.0002,
+      "loss": 0.6191272735595703,
+      "mean_token_accuracy": 0.7458655536174774,
+      "num_tokens": 24259291.0,
+      "step": 651
+    },
+    {
+      "entropy": 0.601327195763588,
+      "epoch": 4.23395613322502,
+      "grad_norm": 0.012125416658818722,
+      "learning_rate": 0.0002,
+      "loss": 0.6103509664535522,
+      "mean_token_accuracy": 0.7481491416692734,
+      "num_tokens": 24296467.0,
+      "step": 652
+    },
+    {
+      "entropy": 0.6011637076735497,
+      "epoch": 4.240454914703493,
+      "grad_norm": 0.011278907768428326,
+      "learning_rate": 0.0002,
+      "loss": 0.6051167845726013,
+      "mean_token_accuracy": 0.7522369399666786,
+      "num_tokens": 24334038.0,
+      "step": 653
+    },
+    {
+      "entropy": 0.624407596886158,
+      "epoch": 4.246953696181966,
+      "grad_norm": 0.011394547298550606,
+      "learning_rate": 0.0002,
+      "loss": 0.6223498582839966,
+      "mean_token_accuracy": 0.7456056624650955,
+      "num_tokens": 24371361.0,
+      "step": 654
+    },
+    {
+      "entropy": 0.6008864194154739,
+      "epoch": 4.2534524776604385,
+      "grad_norm": 0.010803118348121643,
+      "learning_rate": 0.0002,
+      "loss": 0.595673680305481,
+      "mean_token_accuracy": 0.7541809007525444,
+      "num_tokens": 24408760.0,
+      "step": 655
+    },
+    {
+      "entropy": 0.6180669814348221,
+      "epoch": 4.259951259138911,
+      "grad_norm": 0.010789727792143822,
+      "learning_rate": 0.0002,
+      "loss": 0.6144995093345642,
+      "mean_token_accuracy": 0.7488504275679588,
+      "num_tokens": 24446067.0,
+      "step": 656
+    },
+    {
+      "entropy": 0.5908245071768761,
+      "epoch": 4.266450040617384,
+      "grad_norm": 0.013646950013935566,
+      "learning_rate": 0.0002,
+      "loss": 0.5992537140846252,
+      "mean_token_accuracy": 0.7573117613792419,
+      "num_tokens": 24483475.0,
+      "step": 657
+    },
+    {
+      "entropy": 0.6028355807065964,
+      "epoch": 4.272948822095857,
+      "grad_norm": 0.012632039375603199,
+      "learning_rate": 0.0002,
+      "loss": 0.6093482971191406,
+      "mean_token_accuracy": 0.7510194182395935,
+      "num_tokens": 24520599.0,
+      "step": 658
+    },
+    {
+      "entropy": 0.6133154481649399,
+      "epoch": 4.279447603574329,
+      "grad_norm": 0.009593473747372627,
+      "learning_rate": 0.0002,
+      "loss": 0.6142933964729309,
+      "mean_token_accuracy": 0.749172069132328,
+      "num_tokens": 24557970.0,
+      "step": 659
+    },
+    {
+      "entropy": 0.6059748083353043,
+      "epoch": 4.285946385052803,
+      "grad_norm": 0.010930807329714298,
+      "learning_rate": 0.0002,
+      "loss": 0.6034088134765625,
+      "mean_token_accuracy": 0.7542562857270241,
+      "num_tokens": 24595447.0,
+      "step": 660
+    },
+    {
+      "entropy": 0.6150760650634766,
+      "epoch": 4.292445166531276,
+      "grad_norm": 0.012249140068888664,
+      "learning_rate": 0.0002,
+      "loss": 0.6127879023551941,
+      "mean_token_accuracy": 0.7484256699681282,
+      "num_tokens": 24632972.0,
+      "step": 661
+    },
+    {
+      "entropy": 0.6088320016860962,
+      "epoch": 4.298943948009748,
+      "grad_norm": 0.010698465630412102,
+      "learning_rate": 0.0002,
+      "loss": 0.611052930355072,
+      "mean_token_accuracy": 0.7510727941989899,
+      "num_tokens": 24670189.0,
+      "step": 662
+    },
+    {
+      "entropy": 0.5984915345907211,
+      "epoch": 4.305442729488221,
+      "grad_norm": 0.012718550860881805,
+      "learning_rate": 0.0002,
+      "loss": 0.6050142049789429,
+      "mean_token_accuracy": 0.7525518760085106,
+      "num_tokens": 24706956.0,
+      "step": 663
+    },
+    {
+      "entropy": 0.6053661406040192,
+      "epoch": 4.311941510966694,
+      "grad_norm": 0.00997387245297432,
+      "learning_rate": 0.0002,
+      "loss": 0.6083434820175171,
+      "mean_token_accuracy": 0.7501671761274338,
+      "num_tokens": 24744572.0,
+      "step": 664
+    },
+    {
+      "entropy": 0.5996779799461365,
+      "epoch": 4.318440292445167,
+      "grad_norm": 0.009505977854132652,
+      "learning_rate": 0.0002,
+      "loss": 0.6021357774734497,
+      "mean_token_accuracy": 0.753576286137104,
+      "num_tokens": 24782008.0,
+      "step": 665
+    },
+    {
+      "entropy": 0.6148371547460556,
+      "epoch": 4.324939073923639,
+      "grad_norm": 0.010230735875666142,
+      "learning_rate": 0.0002,
+      "loss": 0.6132264137268066,
+      "mean_token_accuracy": 0.7497095540165901,
+      "num_tokens": 24819250.0,
+      "step": 666
+    },
+    {
+      "entropy": 0.6186807677149773,
+      "epoch": 4.331437855402112,
+      "grad_norm": 0.011359174735844135,
+      "learning_rate": 0.0002,
+      "loss": 0.616733193397522,
+      "mean_token_accuracy": 0.7471146136522293,
+      "num_tokens": 24856525.0,
+      "step": 667
+    },
+    {
+      "entropy": 0.5949462577700615,
+      "epoch": 4.337936636880585,
+      "grad_norm": 0.009301742538809776,
+      "learning_rate": 0.0002,
+      "loss": 0.5954026579856873,
+      "mean_token_accuracy": 0.7579122558236122,
+      "num_tokens": 24893833.0,
+      "step": 668
+    },
+    {
+      "entropy": 0.6139709204435349,
+      "epoch": 4.344435418359057,
+      "grad_norm": 0.009777146391570568,
+      "learning_rate": 0.0002,
+      "loss": 0.6152162551879883,
+      "mean_token_accuracy": 0.7482587099075317,
+      "num_tokens": 24931302.0,
+      "step": 669
+    },
+    {
+      "entropy": 0.6105548739433289,
+      "epoch": 4.35093419983753,
+      "grad_norm": 0.009922128170728683,
+      "learning_rate": 0.0002,
+      "loss": 0.6112467646598816,
+      "mean_token_accuracy": 0.7494839727878571,
+      "num_tokens": 24968727.0,
+      "step": 670
+    },
+    {
+      "entropy": 0.6040442362427711,
+      "epoch": 4.357432981316003,
+      "grad_norm": 0.009872304275631905,
+      "learning_rate": 0.0002,
+      "loss": 0.6058943271636963,
+      "mean_token_accuracy": 0.7517568320035934,
+      "num_tokens": 25006181.0,
+      "step": 671
+    },
+    {
+      "entropy": 0.6142793446779251,
+      "epoch": 4.363931762794476,
+      "grad_norm": 0.009624022990465164,
+      "learning_rate": 0.0002,
+      "loss": 0.6160279512405396,
+      "mean_token_accuracy": 0.7462561428546906,
+      "num_tokens": 25043335.0,
+      "step": 672
+    },
+    {
+      "entropy": 0.6127363592386246,
+      "epoch": 4.370430544272949,
+      "grad_norm": 0.00913744606077671,
+      "learning_rate": 0.0002,
+      "loss": 0.6129778027534485,
+      "mean_token_accuracy": 0.750290147960186,
+      "num_tokens": 25081246.0,
+      "step": 673
+    },
+    {
+      "entropy": 0.6002702713012695,
+      "epoch": 4.376929325751422,
+      "grad_norm": 0.010623721405863762,
+      "learning_rate": 0.0002,
+      "loss": 0.6051443815231323,
+      "mean_token_accuracy": 0.7518057897686958,
+      "num_tokens": 25118440.0,
+      "step": 674
+    },
+    {
+      "entropy": 0.6069729775190353,
+      "epoch": 4.383428107229895,
+      "grad_norm": 0.011354091577231884,
+      "learning_rate": 0.0002,
+      "loss": 0.6139703989028931,
+      "mean_token_accuracy": 0.7482916563749313,
+      "num_tokens": 25155694.0,
+      "step": 675
+    },
+    {
+      "entropy": 0.6122787147760391,
+      "epoch": 4.389926888708367,
+      "grad_norm": 0.009671922773122787,
+      "learning_rate": 0.0002,
+      "loss": 0.6111292839050293,
+      "mean_token_accuracy": 0.7500254139304161,
+      "num_tokens": 25192885.0,
+      "step": 676
+    },
+    {
+      "entropy": 0.6137835532426834,
+      "epoch": 4.39642567018684,
+      "grad_norm": 0.009973683394491673,
+      "learning_rate": 0.0002,
+      "loss": 0.6136457920074463,
+      "mean_token_accuracy": 0.7491521015763283,
+      "num_tokens": 25230234.0,
+      "step": 677
+    },
+    {
+      "entropy": 0.6145277097821236,
+      "epoch": 4.402924451665313,
+      "grad_norm": 0.010024932213127613,
+      "learning_rate": 0.0002,
+      "loss": 0.612713634967804,
+      "mean_token_accuracy": 0.7496964707970619,
+      "num_tokens": 25267657.0,
+      "step": 678
+    },
+    {
+      "entropy": 0.6107024550437927,
+      "epoch": 4.4094232331437855,
+      "grad_norm": 0.009390750899910927,
+      "learning_rate": 0.0002,
+      "loss": 0.6097205281257629,
+      "mean_token_accuracy": 0.7498994544148445,
+      "num_tokens": 25305087.0,
+      "step": 679
+    },
+    {
+      "entropy": 0.6041795536875725,
+      "epoch": 4.415922014622258,
+      "grad_norm": 0.012812652625143528,
+      "learning_rate": 0.0002,
+      "loss": 0.6109946370124817,
+      "mean_token_accuracy": 0.7516659647226334,
+      "num_tokens": 25342187.0,
+      "step": 680
+    },
+    {
+      "entropy": 0.5976850837469101,
+      "epoch": 4.422420796100731,
+      "grad_norm": 0.010710278525948524,
+      "learning_rate": 0.0002,
+      "loss": 0.6007599830627441,
+      "mean_token_accuracy": 0.7541299387812614,
+      "num_tokens": 25379496.0,
+      "step": 681
+    },
+    {
+      "entropy": 0.6042672917246819,
+      "epoch": 4.428919577579204,
+      "grad_norm": 0.011890453286468983,
+      "learning_rate": 0.0002,
+      "loss": 0.6084088087081909,
+      "mean_token_accuracy": 0.7506248950958252,
+      "num_tokens": 25416619.0,
+      "step": 682
+    },
+    {
+      "entropy": 0.5986599251627922,
+      "epoch": 4.435418359057676,
+      "grad_norm": 0.010659754276275635,
+      "learning_rate": 0.0002,
+      "loss": 0.5982859134674072,
+      "mean_token_accuracy": 0.7548331841826439,
+      "num_tokens": 25453693.0,
+      "step": 683
+    },
+    {
+      "entropy": 0.616942897439003,
+      "epoch": 4.441917140536149,
+      "grad_norm": 0.01076586078852415,
+      "learning_rate": 0.0002,
+      "loss": 0.6122304201126099,
+      "mean_token_accuracy": 0.7499786019325256,
+      "num_tokens": 25491062.0,
+      "step": 684
+    },
+    {
+      "entropy": 0.6103186085820198,
+      "epoch": 4.448415922014623,
+      "grad_norm": 0.010572830215096474,
+      "learning_rate": 0.0002,
+      "loss": 0.6094003915786743,
+      "mean_token_accuracy": 0.7506365403532982,
+      "num_tokens": 25528185.0,
+      "step": 685
+    },
+    {
+      "entropy": 0.5939375460147858,
+      "epoch": 4.454914703493095,
+      "grad_norm": 0.011964102275669575,
+      "learning_rate": 0.0002,
+      "loss": 0.6006197333335876,
+      "mean_token_accuracy": 0.7550608739256859,
+      "num_tokens": 25565262.0,
+      "step": 686
+    },
+    {
+      "entropy": 0.6017919331789017,
+      "epoch": 4.461413484971568,
+      "grad_norm": 0.013583175837993622,
+      "learning_rate": 0.0002,
+      "loss": 0.6119198203086853,
+      "mean_token_accuracy": 0.7524380683898926,
+      "num_tokens": 25602632.0,
+      "step": 687
+    },
+    {
+      "entropy": 0.6157395392656326,
+      "epoch": 4.467912266450041,
+      "grad_norm": 0.009645634330809116,
+      "learning_rate": 0.0002,
+      "loss": 0.6124662160873413,
+      "mean_token_accuracy": 0.7506381198763847,
+      "num_tokens": 25640141.0,
+      "step": 688
+    },
+    {
+      "entropy": 0.6060995981097221,
+      "epoch": 4.4744110479285135,
+      "grad_norm": 0.011023740284144878,
+      "learning_rate": 0.0002,
+      "loss": 0.604090690612793,
+      "mean_token_accuracy": 0.7523544281721115,
+      "num_tokens": 25676930.0,
+      "step": 689
+    },
+    {
+      "entropy": 0.6346146166324615,
+      "epoch": 4.480909829406986,
+      "grad_norm": 0.010563536547124386,
+      "learning_rate": 0.0002,
+      "loss": 0.6303510665893555,
+      "mean_token_accuracy": 0.7432904541492462,
+      "num_tokens": 25714523.0,
+      "step": 690
+    },
+    {
+      "entropy": 0.6144333481788635,
+      "epoch": 4.487408610885459,
+      "grad_norm": 0.010604371316730976,
+      "learning_rate": 0.0002,
+      "loss": 0.6150377988815308,
+      "mean_token_accuracy": 0.74857547134161,
+      "num_tokens": 25751872.0,
+      "step": 691
+    },
+    {
+      "entropy": 0.5893413797020912,
+      "epoch": 4.493907392363932,
+      "grad_norm": 0.010374541394412518,
+      "learning_rate": 0.0002,
+      "loss": 0.59273362159729,
+      "mean_token_accuracy": 0.7577084898948669,
+      "num_tokens": 25789255.0,
+      "step": 692
+    },
+    {
+      "entropy": 0.5985084772109985,
+      "epoch": 4.500406173842404,
+      "grad_norm": 0.010325289331376553,
+      "learning_rate": 0.0002,
+      "loss": 0.5997466444969177,
+      "mean_token_accuracy": 0.7552184760570526,
+      "num_tokens": 25826911.0,
+      "step": 693
+    },
+    {
+      "entropy": 0.6087296083569527,
+      "epoch": 4.506904955320877,
+      "grad_norm": 0.012259433045983315,
+      "learning_rate": 0.0002,
+      "loss": 0.6142268776893616,
+      "mean_token_accuracy": 0.7504933401942253,
+      "num_tokens": 25864195.0,
+      "step": 694
+    },
+    {
+      "entropy": 0.5923497229814529,
+      "epoch": 4.51340373679935,
+      "grad_norm": 0.011389669962227345,
+      "learning_rate": 0.0002,
+      "loss": 0.597306489944458,
+      "mean_token_accuracy": 0.7565679550170898,
+      "num_tokens": 25901656.0,
+      "step": 695
+    },
+    {
+      "entropy": 0.6001605242490768,
+      "epoch": 4.5199025182778225,
+      "grad_norm": 0.009456605650484562,
+      "learning_rate": 0.0002,
+      "loss": 0.6000933051109314,
+      "mean_token_accuracy": 0.7530366033315659,
+      "num_tokens": 25939491.0,
+      "step": 696
+    },
+    {
+      "entropy": 0.5900749191641808,
+      "epoch": 4.526401299756296,
+      "grad_norm": 0.011240561492741108,
+      "learning_rate": 0.0002,
+      "loss": 0.59348464012146,
+      "mean_token_accuracy": 0.7567196115851402,
+      "num_tokens": 25976506.0,
+      "step": 697
+    },
+    {
+      "entropy": 0.5967278108000755,
+      "epoch": 4.532900081234769,
+      "grad_norm": 0.010507924482226372,
+      "learning_rate": 0.0002,
+      "loss": 0.6000585556030273,
+      "mean_token_accuracy": 0.7546230331063271,
+      "num_tokens": 26013826.0,
+      "step": 698
+    },
+    {
+      "entropy": 0.5980387553572655,
+      "epoch": 4.5393988627132416,
+      "grad_norm": 0.01707952283322811,
+      "learning_rate": 0.0002,
+      "loss": 0.6007162928581238,
+      "mean_token_accuracy": 0.7539133131504059,
+      "num_tokens": 26051676.0,
+      "step": 699
+    },
+    {
+      "entropy": 0.6002839356660843,
+      "epoch": 4.545897644191714,
+      "grad_norm": 0.01148629654198885,
+      "learning_rate": 0.0002,
+      "loss": 0.6052761077880859,
+      "mean_token_accuracy": 0.7534726038575172,
+      "num_tokens": 26088813.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.6063856557011604,
+      "epoch": 4.552396425670187,
+      "grad_norm": 0.04600154981017113,
+      "learning_rate": 0.0002,
+      "loss": 0.6116973161697388,
+      "mean_token_accuracy": 0.7503002285957336,
+      "num_tokens": 26126381.0,
+      "step": 701
+    },
+    {
+      "entropy": 0.6121505573391914,
+      "epoch": 4.55889520714866,
+      "grad_norm": 0.01381687168031931,
+      "learning_rate": 0.0002,
+      "loss": 0.6069159507751465,
+      "mean_token_accuracy": 0.7519886940717697,
+      "num_tokens": 26163482.0,
+      "step": 702
+    },
+    {
+      "entropy": 0.6286856532096863,
+      "epoch": 4.565393988627132,
+      "grad_norm": 0.010802995413541794,
+      "learning_rate": 0.0002,
+      "loss": 0.6213472485542297,
+      "mean_token_accuracy": 0.7473803609609604,
+      "num_tokens": 26200948.0,
+      "step": 703
+    },
+    {
+      "entropy": 0.6292322054505348,
+      "epoch": 4.571892770105605,
+      "grad_norm": 0.01318847481161356,
+      "learning_rate": 0.0002,
+      "loss": 0.6286896467208862,
+      "mean_token_accuracy": 0.7422280386090279,
+      "num_tokens": 26238501.0,
+      "step": 704
+    },
+    {
+      "entropy": 0.61526670306921,
+      "epoch": 4.578391551584078,
+      "grad_norm": 0.012318129651248455,
+      "learning_rate": 0.0002,
+      "loss": 0.6191307306289673,
+      "mean_token_accuracy": 0.7456179112195969,
+      "num_tokens": 26276241.0,
+      "step": 705
+    },
+    {
+      "entropy": 0.5996731072664261,
+      "epoch": 4.584890333062551,
+      "grad_norm": 0.01126960851252079,
+      "learning_rate": 0.0002,
+      "loss": 0.5998624563217163,
+      "mean_token_accuracy": 0.7562123388051987,
+      "num_tokens": 26313858.0,
+      "step": 706
+    },
+    {
+      "entropy": 0.6051791980862617,
+      "epoch": 4.591389114541023,
+      "grad_norm": 0.008872485719621181,
+      "learning_rate": 0.0002,
+      "loss": 0.6033820509910583,
+      "mean_token_accuracy": 0.7548251152038574,
+      "num_tokens": 26351577.0,
+      "step": 707
+    },
+    {
+      "entropy": 0.603697344660759,
+      "epoch": 4.597887896019496,
+      "grad_norm": 0.01154793705791235,
+      "learning_rate": 0.0002,
+      "loss": 0.6016978025436401,
+      "mean_token_accuracy": 0.7543343156576157,
+      "num_tokens": 26388978.0,
+      "step": 708
+    },
+    {
+      "entropy": 0.6170104667544365,
+      "epoch": 4.604386677497969,
+      "grad_norm": 0.01096999179571867,
+      "learning_rate": 0.0002,
+      "loss": 0.622772216796875,
+      "mean_token_accuracy": 0.7449462041258812,
+      "num_tokens": 26426264.0,
+      "step": 709
+    },
+    {
+      "entropy": 0.6072003245353699,
+      "epoch": 4.610885458976442,
+      "grad_norm": 0.010877830907702446,
+      "learning_rate": 0.0002,
+      "loss": 0.6076858639717102,
+      "mean_token_accuracy": 0.7529691979289055,
+      "num_tokens": 26463914.0,
+      "step": 710
+    },
+    {
+      "entropy": 0.6025268062949181,
+      "epoch": 4.617384240454915,
+      "grad_norm": 0.011258685030043125,
+      "learning_rate": 0.0002,
+      "loss": 0.6081684231758118,
+      "mean_token_accuracy": 0.751194529235363,
+      "num_tokens": 26501178.0,
+      "step": 711
+    },
+    {
+      "entropy": 0.6161645799875259,
+      "epoch": 4.623883021933388,
+      "grad_norm": 0.011774537153542042,
+      "learning_rate": 0.0002,
+      "loss": 0.6219759583473206,
+      "mean_token_accuracy": 0.7475455403327942,
+      "num_tokens": 26538587.0,
+      "step": 712
+    },
+    {
+      "entropy": 0.61099823564291,
+      "epoch": 4.6303818034118605,
+      "grad_norm": 0.011626939289271832,
+      "learning_rate": 0.0002,
+      "loss": 0.6097919940948486,
+      "mean_token_accuracy": 0.7494744956493378,
+      "num_tokens": 26576055.0,
+      "step": 713
+    },
+    {
+      "entropy": 0.5967102572321892,
+      "epoch": 4.636880584890333,
+      "grad_norm": 0.011615315452218056,
+      "learning_rate": 0.0002,
+      "loss": 0.5964145660400391,
+      "mean_token_accuracy": 0.7566935420036316,
+      "num_tokens": 26613206.0,
+      "step": 714
+    },
+    {
+      "entropy": 0.6024897918105125,
+      "epoch": 4.643379366368806,
+      "grad_norm": 0.011359010823071003,
+      "learning_rate": 0.0002,
+      "loss": 0.6033364534378052,
+      "mean_token_accuracy": 0.75516776740551,
+      "num_tokens": 26650906.0,
+      "step": 715
+    },
+    {
+      "entropy": 0.6127264201641083,
+      "epoch": 4.649878147847279,
+      "grad_norm": 0.011858063749969006,
+      "learning_rate": 0.0002,
+      "loss": 0.6169857382774353,
+      "mean_token_accuracy": 0.7504541873931885,
+      "num_tokens": 26688277.0,
+      "step": 716
+    },
+    {
+      "entropy": 0.6103435754776001,
+      "epoch": 4.656376929325751,
+      "grad_norm": 0.012304591946303844,
+      "learning_rate": 0.0002,
+      "loss": 0.6171730756759644,
+      "mean_token_accuracy": 0.7490719556808472,
+      "num_tokens": 26725704.0,
+      "step": 717
+    },
+    {
+      "entropy": 0.6082026064395905,
+      "epoch": 4.662875710804224,
+      "grad_norm": 0.00955925416201353,
+      "learning_rate": 0.0002,
+      "loss": 0.6081550121307373,
+      "mean_token_accuracy": 0.7512648478150368,
+      "num_tokens": 26762897.0,
+      "step": 718
+    },
+    {
+      "entropy": 0.615670770406723,
+      "epoch": 4.669374492282697,
+      "grad_norm": 0.011302441358566284,
+      "learning_rate": 0.0002,
+      "loss": 0.6137043833732605,
+      "mean_token_accuracy": 0.7484593465924263,
+      "num_tokens": 26799617.0,
+      "step": 719
+    },
+    {
+      "entropy": 0.6008039489388466,
+      "epoch": 4.6758732737611695,
+      "grad_norm": 0.010860615409910679,
+      "learning_rate": 0.0002,
+      "loss": 0.5949652194976807,
+      "mean_token_accuracy": 0.7573561295866966,
+      "num_tokens": 26836935.0,
+      "step": 720
+    },
+    {
+      "entropy": 0.6153427511453629,
+      "epoch": 4.682372055239643,
+      "grad_norm": 0.01038694754242897,
+      "learning_rate": 0.0002,
+      "loss": 0.6163858771324158,
+      "mean_token_accuracy": 0.7477854266762733,
+      "num_tokens": 26874482.0,
+      "step": 721
+    },
+    {
+      "entropy": 0.6078703328967094,
+      "epoch": 4.688870836718115,
+      "grad_norm": 0.011593160219490528,
+      "learning_rate": 0.0002,
+      "loss": 0.6154011487960815,
+      "mean_token_accuracy": 0.7504984140396118,
+      "num_tokens": 26912012.0,
+      "step": 722
+    },
+    {
+      "entropy": 0.6097044795751572,
+      "epoch": 4.6953696181965885,
+      "grad_norm": 0.011452315375208855,
+      "learning_rate": 0.0002,
+      "loss": 0.6159889698028564,
+      "mean_token_accuracy": 0.7497701123356819,
+      "num_tokens": 26949190.0,
+      "step": 723
+    },
+    {
+      "entropy": 0.6035362407565117,
+      "epoch": 4.701868399675061,
+      "grad_norm": 0.009807482361793518,
+      "learning_rate": 0.0002,
+      "loss": 0.6073694229125977,
+      "mean_token_accuracy": 0.7511623874306679,
+      "num_tokens": 26986593.0,
+      "step": 724
+    },
+    {
+      "entropy": 0.6145569458603859,
+      "epoch": 4.708367181153534,
+      "grad_norm": 0.009589890018105507,
+      "learning_rate": 0.0002,
+      "loss": 0.615104079246521,
+      "mean_token_accuracy": 0.7482927516102791,
+      "num_tokens": 27023613.0,
+      "step": 725
+    },
+    {
+      "entropy": 0.6154572740197182,
+      "epoch": 4.714865962632007,
+      "grad_norm": 0.01081397570669651,
+      "learning_rate": 0.0002,
+      "loss": 0.6108368635177612,
+      "mean_token_accuracy": 0.7491915374994278,
+      "num_tokens": 27061197.0,
+      "step": 726
+    },
+    {
+      "entropy": 0.620484970510006,
+      "epoch": 4.721364744110479,
+      "grad_norm": 0.009981570765376091,
+      "learning_rate": 0.0002,
+      "loss": 0.6177104711532593,
+      "mean_token_accuracy": 0.7493368536233902,
+      "num_tokens": 27098608.0,
+      "step": 727
+    },
+    {
+      "entropy": 0.6061017587780952,
+      "epoch": 4.727863525588952,
+      "grad_norm": 0.011618540622293949,
+      "learning_rate": 0.0002,
+      "loss": 0.6062725186347961,
+      "mean_token_accuracy": 0.7519984692335129,
+      "num_tokens": 27136501.0,
+      "step": 728
+    },
+    {
+      "entropy": 0.6067364290356636,
+      "epoch": 4.734362307067425,
+      "grad_norm": 0.009954583831131458,
+      "learning_rate": 0.0002,
+      "loss": 0.6102150678634644,
+      "mean_token_accuracy": 0.7502684965729713,
+      "num_tokens": 27173703.0,
+      "step": 729
+    },
+    {
+      "entropy": 0.605129785835743,
+      "epoch": 4.7408610885458975,
+      "grad_norm": 0.012882952578365803,
+      "learning_rate": 0.0002,
+      "loss": 0.6149674654006958,
+      "mean_token_accuracy": 0.7470449358224869,
+      "num_tokens": 27210993.0,
+      "step": 730
+    },
+    {
+      "entropy": 0.6143843829631805,
+      "epoch": 4.74735987002437,
+      "grad_norm": 0.010333128273487091,
+      "learning_rate": 0.0002,
+      "loss": 0.6135010719299316,
+      "mean_token_accuracy": 0.7495505511760712,
+      "num_tokens": 27248544.0,
+      "step": 731
+    },
+    {
+      "entropy": 0.619462713599205,
+      "epoch": 4.753858651502843,
+      "grad_norm": 0.009751244448125362,
+      "learning_rate": 0.0002,
+      "loss": 0.6161588430404663,
+      "mean_token_accuracy": 0.7501519843935966,
+      "num_tokens": 27286017.0,
+      "step": 732
+    },
+    {
+      "entropy": 0.6085372492671013,
+      "epoch": 4.760357432981316,
+      "grad_norm": 0.01087701041251421,
+      "learning_rate": 0.0002,
+      "loss": 0.6058393716812134,
+      "mean_token_accuracy": 0.7525414749979973,
+      "num_tokens": 27323456.0,
+      "step": 733
+    },
+    {
+      "entropy": 0.6052304580807686,
+      "epoch": 4.766856214459789,
+      "grad_norm": 0.010626486502587795,
+      "learning_rate": 0.0002,
+      "loss": 0.6062231063842773,
+      "mean_token_accuracy": 0.7521145865321159,
+      "num_tokens": 27360812.0,
+      "step": 734
+    },
+    {
+      "entropy": 0.609903909265995,
+      "epoch": 4.773354995938262,
+      "grad_norm": 0.010470789857208729,
+      "learning_rate": 0.0002,
+      "loss": 0.6119512319564819,
+      "mean_token_accuracy": 0.7512121498584747,
+      "num_tokens": 27398510.0,
+      "step": 735
+    },
+    {
+      "entropy": 0.6010274812579155,
+      "epoch": 4.779853777416735,
+      "grad_norm": 0.009013401344418526,
+      "learning_rate": 0.0002,
+      "loss": 0.6045113205909729,
+      "mean_token_accuracy": 0.7538518905639648,
+      "num_tokens": 27435677.0,
+      "step": 736
+    },
+    {
+      "entropy": 0.6044503897428513,
+      "epoch": 4.786352558895207,
+      "grad_norm": 0.009969563223421574,
+      "learning_rate": 0.0002,
+      "loss": 0.6042656898498535,
+      "mean_token_accuracy": 0.7533875405788422,
+      "num_tokens": 27472951.0,
+      "step": 737
+    },
+    {
+      "entropy": 0.6085273697972298,
+      "epoch": 4.79285134037368,
+      "grad_norm": 0.010333449579775333,
+      "learning_rate": 0.0002,
+      "loss": 0.6119939088821411,
+      "mean_token_accuracy": 0.7497808411717415,
+      "num_tokens": 27510678.0,
+      "step": 738
+    },
+    {
+      "entropy": 0.5953918546438217,
+      "epoch": 4.799350121852153,
+      "grad_norm": 0.011067725718021393,
+      "learning_rate": 0.0002,
+      "loss": 0.6043837070465088,
+      "mean_token_accuracy": 0.7549910023808479,
+      "num_tokens": 27548061.0,
+      "step": 739
+    },
+    {
+      "entropy": 0.6027686223387718,
+      "epoch": 4.8058489033306255,
+      "grad_norm": 0.010080584324896336,
+      "learning_rate": 0.0002,
+      "loss": 0.6062842011451721,
+      "mean_token_accuracy": 0.752067320048809,
+      "num_tokens": 27585301.0,
+      "step": 740
+    },
+    {
+      "entropy": 0.6061441823840141,
+      "epoch": 4.812347684809098,
+      "grad_norm": 0.009884923696517944,
+      "learning_rate": 0.0002,
+      "loss": 0.6060526371002197,
+      "mean_token_accuracy": 0.7508253157138824,
+      "num_tokens": 27622471.0,
+      "step": 741
+    },
+    {
+      "entropy": 0.6114162281155586,
+      "epoch": 4.818846466287571,
+      "grad_norm": 0.009803572669625282,
+      "learning_rate": 0.0002,
+      "loss": 0.6056950092315674,
+      "mean_token_accuracy": 0.7509299144148827,
+      "num_tokens": 27659725.0,
+      "step": 742
+    },
+    {
+      "entropy": 0.604758232831955,
+      "epoch": 4.825345247766044,
+      "grad_norm": 0.009157917462289333,
+      "learning_rate": 0.0002,
+      "loss": 0.6060429811477661,
+      "mean_token_accuracy": 0.7535065114498138,
+      "num_tokens": 27697579.0,
+      "step": 743
+    },
+    {
+      "entropy": 0.6063171178102493,
+      "epoch": 4.831844029244516,
+      "grad_norm": 0.010433568619191647,
+      "learning_rate": 0.0002,
+      "loss": 0.6129876971244812,
+      "mean_token_accuracy": 0.7488505020737648,
+      "num_tokens": 27735152.0,
+      "step": 744
+    },
+    {
+      "entropy": 0.6220549941062927,
+      "epoch": 4.838342810722989,
+      "grad_norm": 0.009133824147284031,
+      "learning_rate": 0.0002,
+      "loss": 0.6260920166969299,
+      "mean_token_accuracy": 0.7444146424531937,
+      "num_tokens": 27772809.0,
+      "step": 745
+    },
+    {
+      "entropy": 0.6151563301682472,
+      "epoch": 4.844841592201462,
+      "grad_norm": 0.009275715798139572,
+      "learning_rate": 0.0002,
+      "loss": 0.6143773794174194,
+      "mean_token_accuracy": 0.7501434609293938,
+      "num_tokens": 27810585.0,
+      "step": 746
+    },
+    {
+      "entropy": 0.6030779108405113,
+      "epoch": 4.851340373679935,
+      "grad_norm": 0.009752030484378338,
+      "learning_rate": 0.0002,
+      "loss": 0.6026537418365479,
+      "mean_token_accuracy": 0.7538170889019966,
+      "num_tokens": 27848177.0,
+      "step": 747
+    },
+    {
+      "entropy": 0.6034507304430008,
+      "epoch": 4.857839155158408,
+      "grad_norm": 0.009275935590267181,
+      "learning_rate": 0.0002,
+      "loss": 0.6020852327346802,
+      "mean_token_accuracy": 0.7548267766833305,
+      "num_tokens": 27885820.0,
+      "step": 748
+    },
+    {
+      "entropy": 0.5959972590208054,
+      "epoch": 4.864337936636881,
+      "grad_norm": 0.01019821036607027,
+      "learning_rate": 0.0002,
+      "loss": 0.5997313857078552,
+      "mean_token_accuracy": 0.7544100061058998,
+      "num_tokens": 27923215.0,
+      "step": 749
+    },
+    {
+      "entropy": 0.5984265580773354,
+      "epoch": 4.870836718115354,
+      "grad_norm": 0.010381966829299927,
+      "learning_rate": 0.0002,
+      "loss": 0.5997567176818848,
+      "mean_token_accuracy": 0.754194863140583,
+      "num_tokens": 27960378.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.6121873781085014,
+      "epoch": 4.877335499593826,
+      "grad_norm": 0.010476244613528252,
+      "learning_rate": 0.0002,
+      "loss": 0.6164151430130005,
+      "mean_token_accuracy": 0.747443437576294,
+      "num_tokens": 27997671.0,
+      "step": 751
+    },
+    {
+      "entropy": 0.6084479093551636,
+      "epoch": 4.883834281072299,
+      "grad_norm": 0.011203959584236145,
+      "learning_rate": 0.0002,
+      "loss": 0.608171820640564,
+      "mean_token_accuracy": 0.75231072306633,
+      "num_tokens": 28034897.0,
+      "step": 752
+    },
+    {
+      "entropy": 0.6147375106811523,
+      "epoch": 4.890333062550772,
+      "grad_norm": 0.009496328420937061,
+      "learning_rate": 0.0002,
+      "loss": 0.6129316091537476,
+      "mean_token_accuracy": 0.7494983300566673,
+      "num_tokens": 28072034.0,
+      "step": 753
+    },
+    {
+      "entropy": 0.6106340512633324,
+      "epoch": 4.896831844029244,
+      "grad_norm": 0.011393910273909569,
+      "learning_rate": 0.0002,
+      "loss": 0.6084151268005371,
+      "mean_token_accuracy": 0.7524324804544449,
+      "num_tokens": 28109258.0,
+      "step": 754
+    },
+    {
+      "entropy": 0.600101962685585,
+      "epoch": 4.903330625507717,
+      "grad_norm": 0.010883449576795101,
+      "learning_rate": 0.0002,
+      "loss": 0.6043680906295776,
+      "mean_token_accuracy": 0.7542835772037506,
+      "num_tokens": 28146570.0,
+      "step": 755
+    },
+    {
+      "entropy": 0.6109775528311729,
+      "epoch": 4.90982940698619,
+      "grad_norm": 0.010207289829850197,
+      "learning_rate": 0.0002,
+      "loss": 0.6150296926498413,
+      "mean_token_accuracy": 0.7493007630109787,
+      "num_tokens": 28183579.0,
+      "step": 756
+    },
+    {
+      "entropy": 0.597320593893528,
+      "epoch": 4.916328188464663,
+      "grad_norm": 0.010998498648405075,
+      "learning_rate": 0.0002,
+      "loss": 0.599211573600769,
+      "mean_token_accuracy": 0.7554142326116562,
+      "num_tokens": 28220767.0,
+      "step": 757
+    },
+    {
+      "entropy": 0.6176287531852722,
+      "epoch": 4.922826969943135,
+      "grad_norm": 0.012422211468219757,
+      "learning_rate": 0.0002,
+      "loss": 0.6129187941551208,
+      "mean_token_accuracy": 0.7506537437438965,
+      "num_tokens": 28258324.0,
+      "step": 758
+    },
+    {
+      "entropy": 0.604163758456707,
+      "epoch": 4.929325751421608,
+      "grad_norm": 0.010133393108844757,
+      "learning_rate": 0.0002,
+      "loss": 0.6023921966552734,
+      "mean_token_accuracy": 0.7550497874617577,
+      "num_tokens": 28295946.0,
+      "step": 759
+    },
+    {
+      "entropy": 0.6072142794728279,
+      "epoch": 4.935824532900082,
+      "grad_norm": 0.011114662513136864,
+      "learning_rate": 0.0002,
+      "loss": 0.6146305799484253,
+      "mean_token_accuracy": 0.7498172372579575,
+      "num_tokens": 28333200.0,
+      "step": 760
+    },
+    {
+      "entropy": 0.6026475057005882,
+      "epoch": 4.942323314378554,
+      "grad_norm": 0.010693948715925217,
+      "learning_rate": 0.0002,
+      "loss": 0.6080969572067261,
+      "mean_token_accuracy": 0.7521899044513702,
+      "num_tokens": 28370788.0,
+      "step": 761
+    },
+    {
+      "entropy": 0.6020209938287735,
+      "epoch": 4.948822095857027,
+      "grad_norm": 0.010837409645318985,
+      "learning_rate": 0.0002,
+      "loss": 0.6070795059204102,
+      "mean_token_accuracy": 0.7505692467093468,
+      "num_tokens": 28407932.0,
+      "step": 762
+    },
+    {
+      "entropy": 0.612369492650032,
+      "epoch": 4.9553208773355,
+      "grad_norm": 0.010154753923416138,
+      "learning_rate": 0.0002,
+      "loss": 0.6116279363632202,
+      "mean_token_accuracy": 0.7497418820858002,
+      "num_tokens": 28445493.0,
+      "step": 763
+    },
+    {
+      "entropy": 0.6173591017723083,
+      "epoch": 4.9618196588139725,
+      "grad_norm": 0.008918640203773975,
+      "learning_rate": 0.0002,
+      "loss": 0.61656254529953,
+      "mean_token_accuracy": 0.7493520677089691,
+      "num_tokens": 28483286.0,
+      "step": 764
+    },
+    {
+      "entropy": 0.6011045426130295,
+      "epoch": 4.968318440292445,
+      "grad_norm": 0.011289622634649277,
+      "learning_rate": 0.0002,
+      "loss": 0.6005913019180298,
+      "mean_token_accuracy": 0.7565959766507149,
+      "num_tokens": 28520772.0,
+      "step": 765
+    },
+    {
+      "entropy": 0.6143112033605576,
+      "epoch": 4.974817221770918,
+      "grad_norm": 0.009968056343495846,
+      "learning_rate": 0.0002,
+      "loss": 0.6112393140792847,
+      "mean_token_accuracy": 0.7505759075284004,
+      "num_tokens": 28558593.0,
+      "step": 766
+    },
+    {
+      "entropy": 0.609484076499939,
+      "epoch": 4.981316003249391,
+      "grad_norm": 0.00863439030945301,
+      "learning_rate": 0.0002,
+      "loss": 0.6094443798065186,
+      "mean_token_accuracy": 0.753460243344307,
+      "num_tokens": 28596214.0,
+      "step": 767
+    },
+    {
+      "entropy": 0.6114853024482727,
+      "epoch": 4.987814784727863,
+      "grad_norm": 0.008814731612801552,
+      "learning_rate": 0.0002,
+      "loss": 0.6125534772872925,
+      "mean_token_accuracy": 0.7496102303266525,
+      "num_tokens": 28634065.0,
+      "step": 768
+    },
+    {
+      "entropy": 0.6083709001541138,
+      "epoch": 4.994313566206336,
+      "grad_norm": 0.012475132942199707,
+      "learning_rate": 0.0002,
+      "loss": 0.6144940853118896,
+      "mean_token_accuracy": 0.7477302476763725,
+      "num_tokens": 28671577.0,
+      "step": 769
+    },
+    {
+      "entropy": 0.6027948686054775,
+      "epoch": 5.0,
+      "grad_norm": 0.009683027863502502,
+      "learning_rate": 0.0002,
+      "loss": 0.6097056865692139,
+      "mean_token_accuracy": 0.7535298126084464,
+      "num_tokens": 28688115.0,
+      "step": 770
     }
   ],
   "logging_steps": 1,
-  "max_steps": 320,
+  "max_steps": 770,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 10,
+  "num_train_epochs": 5,
   "save_steps": 500,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -3227,8 +7727,8 @@
       "attributes": {}
     }
   },
-  "total_flos": 1.7190992994946253e+18,
-  "train_batch_size": 16,
+  "total_flos": 2.685759473330422e+18,
+  "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null
 }