diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,4606 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 100,
+  "global_step": 2703,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.001849796522382538,
+      "grad_norm": 6.71875,
+      "learning_rate": 3.690036900369004e-07,
+      "loss": 1.0359,
+      "mean_token_accuracy": 0.7334607805116029,
+      "step": 5
+    },
+    {
+      "epoch": 0.003699593044765076,
+      "grad_norm": 6.6875,
+      "learning_rate": 7.380073800738008e-07,
+      "loss": 1.0642,
+      "mean_token_accuracy": 0.725717600508523,
+      "step": 10
+    },
+    {
+      "epoch": 0.005549389567147614,
+      "grad_norm": 6.40625,
+      "learning_rate": 1.1070110701107011e-06,
+      "loss": 1.0002,
+      "mean_token_accuracy": 0.7400892254527123,
+      "step": 15
+    },
+    {
+      "epoch": 0.007399186089530152,
+      "grad_norm": 6.53125,
+      "learning_rate": 1.4760147601476015e-06,
+      "loss": 1.0207,
+      "mean_token_accuracy": 0.7351857063698171,
+      "step": 20
+    },
+    {
+      "epoch": 0.00924898261191269,
+      "grad_norm": 6.34375,
+      "learning_rate": 1.845018450184502e-06,
+      "loss": 1.0441,
+      "mean_token_accuracy": 0.7217523568187054,
+      "step": 25
+    },
+    {
+      "epoch": 0.011098779134295227,
+      "grad_norm": 4.875,
+      "learning_rate": 2.2140221402214023e-06,
+      "loss": 1.0411,
+      "mean_token_accuracy": 0.7219991375630554,
+      "step": 30
+    },
+    {
+      "epoch": 0.012948575656677765,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.5830258302583027e-06,
+      "loss": 0.9179,
+      "mean_token_accuracy": 0.7535858713311748,
+      "step": 35
+    },
+    {
+      "epoch": 0.014798372179060304,
+      "grad_norm": 3.71875,
+      "learning_rate": 2.952029520295203e-06,
+      "loss": 0.9195,
+      "mean_token_accuracy": 0.7462933295487761,
+      "step": 40
+    },
+    {
+      "epoch": 0.016648168701442843,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.3210332103321034e-06,
+      "loss": 0.9707,
+      "mean_token_accuracy": 0.7322681096111887,
+      "step": 45
+    },
+    {
+      "epoch": 0.01849796522382538,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.690036900369004e-06,
+      "loss": 0.9302,
+      "mean_token_accuracy": 0.7402578597480863,
+      "step": 50
+    },
+    {
+      "epoch": 0.020347761746207917,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.059040590405905e-06,
+      "loss": 0.8534,
+      "mean_token_accuracy": 0.7605697968174225,
+      "step": 55
+    },
+    {
+      "epoch": 0.022197558268590455,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.428044280442805e-06,
+      "loss": 0.8845,
+      "mean_token_accuracy": 0.7471328709683942,
+      "step": 60
+    },
+    {
+      "epoch": 0.024047354790972992,
+      "grad_norm": 2.09375,
+      "learning_rate": 4.797047970479705e-06,
+      "loss": 0.8936,
+      "mean_token_accuracy": 0.7483969556481461,
+      "step": 65
+    },
+    {
+      "epoch": 0.02589715131335553,
+      "grad_norm": 2.40625,
+      "learning_rate": 5.166051660516605e-06,
+      "loss": 0.8834,
+      "mean_token_accuracy": 0.746596679268768,
+      "step": 70
+    },
+    {
+      "epoch": 0.02774694783573807,
+      "grad_norm": 2.03125,
+      "learning_rate": 5.535055350553506e-06,
+      "loss": 0.838,
+      "mean_token_accuracy": 0.7577640358784391,
+      "step": 75
+    },
+    {
+      "epoch": 0.029596744358120607,
+      "grad_norm": 1.828125,
+      "learning_rate": 5.904059040590406e-06,
+      "loss": 0.8034,
+      "mean_token_accuracy": 0.7679854500593037,
+      "step": 80
+    },
+    {
+      "epoch": 0.031446540880503145,
+      "grad_norm": 1.6796875,
+      "learning_rate": 6.273062730627307e-06,
+      "loss": 0.7772,
+      "mean_token_accuracy": 0.772276831995462,
+      "step": 85
+    },
+    {
+      "epoch": 0.033296337402885685,
+      "grad_norm": 1.7578125,
+      "learning_rate": 6.642066420664207e-06,
+      "loss": 0.8176,
+      "mean_token_accuracy": 0.7610055654799306,
+      "step": 90
+    },
+    {
+      "epoch": 0.03514613392526822,
+      "grad_norm": 1.765625,
+      "learning_rate": 7.011070110701108e-06,
+      "loss": 0.7514,
+      "mean_token_accuracy": 0.7759203412810342,
+      "step": 95
+    },
+    {
+      "epoch": 0.03699593044765076,
+      "grad_norm": 1.9375,
+      "learning_rate": 7.380073800738008e-06,
+      "loss": 0.7801,
+      "mean_token_accuracy": 0.7682134590113303,
+      "step": 100
+    },
+    {
+      "epoch": 0.03699593044765076,
+      "eval_loss": 0.8076974153518677,
+      "eval_mean_token_accuracy": 0.7595044873390318,
+      "eval_runtime": 89.7602,
+      "eval_samples_per_second": 1.437,
+      "eval_steps_per_second": 0.368,
+      "step": 100
+    },
+    {
+      "epoch": 0.038845726970033294,
+      "grad_norm": 1.640625,
+      "learning_rate": 7.749077490774908e-06,
+      "loss": 0.7558,
+      "mean_token_accuracy": 0.7753679217287962,
+      "step": 105
+    },
+    {
+      "epoch": 0.040695523492415835,
+      "grad_norm": 1.5390625,
+      "learning_rate": 8.11808118081181e-06,
+      "loss": 0.729,
+      "mean_token_accuracy": 0.7821930031916854,
+      "step": 110
+    },
+    {
+      "epoch": 0.042545320014798375,
+      "grad_norm": 1.8515625,
+      "learning_rate": 8.48708487084871e-06,
+      "loss": 0.7523,
+      "mean_token_accuracy": 0.7764337635681507,
+      "step": 115
+    },
+    {
+      "epoch": 0.04439511653718091,
+      "grad_norm": 1.578125,
+      "learning_rate": 8.85608856088561e-06,
+      "loss": 0.7732,
+      "mean_token_accuracy": 0.7691583061445552,
+      "step": 120
+    },
+    {
+      "epoch": 0.04624491305956345,
+      "grad_norm": 1.6640625,
+      "learning_rate": 9.22509225092251e-06,
+      "loss": 0.7767,
+      "mean_token_accuracy": 0.7670524519817176,
+      "step": 125
+    },
+    {
+      "epoch": 0.048094709581945984,
+      "grad_norm": 1.5859375,
+      "learning_rate": 9.59409594095941e-06,
+      "loss": 0.7615,
+      "mean_token_accuracy": 0.7714133141734332,
+      "step": 130
+    },
+    {
+      "epoch": 0.049944506104328525,
+      "grad_norm": 1.5390625,
+      "learning_rate": 9.963099630996312e-06,
+      "loss": 0.7463,
+      "mean_token_accuracy": 0.7763736692059899,
+      "step": 135
+    },
+    {
+      "epoch": 0.05179430262671106,
+      "grad_norm": 1.90625,
+      "learning_rate": 1.033210332103321e-05,
+      "loss": 0.8003,
+      "mean_token_accuracy": 0.7584582137233424,
+      "step": 140
+    },
+    {
+      "epoch": 0.0536440991490936,
+      "grad_norm": 1.75,
+      "learning_rate": 1.0701107011070112e-05,
+      "loss": 0.7528,
+      "mean_token_accuracy": 0.774144136850804,
+      "step": 145
+    },
+    {
+      "epoch": 0.05549389567147614,
+      "grad_norm": 1.7734375,
+      "learning_rate": 1.1070110701107012e-05,
+      "loss": 0.7758,
+      "mean_token_accuracy": 0.7652208817571029,
+      "step": 150
+    },
+    {
+      "epoch": 0.057343692193858674,
+      "grad_norm": 1.6796875,
+      "learning_rate": 1.1439114391143913e-05,
+      "loss": 0.7026,
+      "mean_token_accuracy": 0.787824234302027,
+      "step": 155
+    },
+    {
+      "epoch": 0.059193488716241215,
+      "grad_norm": 1.828125,
+      "learning_rate": 1.1808118081180812e-05,
+      "loss": 0.741,
+      "mean_token_accuracy": 0.7761717028282358,
+      "step": 160
+    },
+    {
+      "epoch": 0.06104328523862375,
+      "grad_norm": 2.09375,
+      "learning_rate": 1.2177121771217713e-05,
+      "loss": 0.743,
+      "mean_token_accuracy": 0.7752838302609522,
+      "step": 165
+    },
+    {
+      "epoch": 0.06289308176100629,
+      "grad_norm": 1.7109375,
+      "learning_rate": 1.2546125461254614e-05,
+      "loss": 0.7011,
+      "mean_token_accuracy": 0.7859840694454219,
+      "step": 170
+    },
+    {
+      "epoch": 0.06474287828338883,
+      "grad_norm": 1.703125,
+      "learning_rate": 1.2915129151291515e-05,
+      "loss": 0.728,
+      "mean_token_accuracy": 0.7778929361410125,
+      "step": 175
+    },
+    {
+      "epoch": 0.06659267480577137,
+      "grad_norm": 1.7734375,
+      "learning_rate": 1.3284132841328414e-05,
+      "loss": 0.7416,
+      "mean_token_accuracy": 0.7746633811319505,
+      "step": 180
+    },
+    {
+      "epoch": 0.0684424713281539,
+      "grad_norm": 1.5625,
+      "learning_rate": 1.3653136531365315e-05,
+      "loss": 0.6821,
+      "mean_token_accuracy": 0.7921394169470078,
+      "step": 185
+    },
+    {
+      "epoch": 0.07029226785053644,
+      "grad_norm": 1.640625,
+      "learning_rate": 1.4022140221402215e-05,
+      "loss": 0.6716,
+      "mean_token_accuracy": 0.7925588833966197,
+      "step": 190
+    },
+    {
+      "epoch": 0.07214206437291898,
+      "grad_norm": 1.703125,
+      "learning_rate": 1.4391143911439116e-05,
+      "loss": 0.7218,
+      "mean_token_accuracy": 0.777652988384069,
+      "step": 195
+    },
+    {
+      "epoch": 0.07399186089530152,
+      "grad_norm": 1.671875,
+      "learning_rate": 1.4760147601476015e-05,
+      "loss": 0.7133,
+      "mean_token_accuracy": 0.7806573763255718,
+      "step": 200
+    },
+    {
+      "epoch": 0.07399186089530152,
+      "eval_loss": 0.729367733001709,
+      "eval_mean_token_accuracy": 0.774526536718017,
+      "eval_runtime": 89.5112,
+      "eval_samples_per_second": 1.441,
+      "eval_steps_per_second": 0.369,
+      "step": 200
+    },
+    {
+      "epoch": 0.07584165741768406,
+      "grad_norm": 1.875,
+      "learning_rate": 1.5129151291512916e-05,
+      "loss": 0.7466,
+      "mean_token_accuracy": 0.76975331301781,
+      "step": 205
+    },
+    {
+      "epoch": 0.07769145394006659,
+      "grad_norm": 1.921875,
+      "learning_rate": 1.5498154981549817e-05,
+      "loss": 0.7262,
+      "mean_token_accuracy": 0.7780428598112316,
+      "step": 210
+    },
+    {
+      "epoch": 0.07954125046244913,
+      "grad_norm": 1.5546875,
+      "learning_rate": 1.5867158671586716e-05,
+      "loss": 0.679,
+      "mean_token_accuracy": 0.7881943840544794,
+      "step": 215
+    },
+    {
+      "epoch": 0.08139104698483167,
+      "grad_norm": 1.5703125,
+      "learning_rate": 1.623616236162362e-05,
+      "loss": 0.6941,
+      "mean_token_accuracy": 0.7853721991173701,
+      "step": 220
+    },
+    {
+      "epoch": 0.08324084350721421,
+      "grad_norm": 1.921875,
+      "learning_rate": 1.6605166051660518e-05,
+      "loss": 0.7008,
+      "mean_token_accuracy": 0.7854988029624879,
+      "step": 225
+    },
+    {
+      "epoch": 0.08509064002959675,
+      "grad_norm": 1.671875,
+      "learning_rate": 1.697416974169742e-05,
+      "loss": 0.7267,
+      "mean_token_accuracy": 0.7753679242661929,
+      "step": 230
+    },
+    {
+      "epoch": 0.08694043655197928,
+      "grad_norm": 1.796875,
+      "learning_rate": 1.734317343173432e-05,
+      "loss": 0.6766,
+      "mean_token_accuracy": 0.7937584401564536,
+      "step": 235
+    },
+    {
+      "epoch": 0.08879023307436182,
+      "grad_norm": 1.640625,
+      "learning_rate": 1.771217712177122e-05,
+      "loss": 0.7177,
+      "mean_token_accuracy": 0.7791101067821142,
+      "step": 240
+    },
+    {
+      "epoch": 0.09064002959674436,
+      "grad_norm": 1.7109375,
+      "learning_rate": 1.8081180811808117e-05,
+      "loss": 0.6934,
+      "mean_token_accuracy": 0.7868716189473924,
+      "step": 245
+    },
+    {
+      "epoch": 0.0924898261191269,
+      "grad_norm": 1.671875,
+      "learning_rate": 1.845018450184502e-05,
+      "loss": 0.7334,
+      "mean_token_accuracy": 0.7720347284120163,
+      "step": 250
+    },
+    {
+      "epoch": 0.09433962264150944,
+      "grad_norm": 1.515625,
+      "learning_rate": 1.8819188191881922e-05,
+      "loss": 0.6676,
+      "mean_token_accuracy": 0.7933765371867549,
+      "step": 255
+    },
+    {
+      "epoch": 0.09618941916389197,
+      "grad_norm": 1.625,
+      "learning_rate": 1.918819188191882e-05,
+      "loss": 0.6851,
+      "mean_token_accuracy": 0.7883584436465398,
+      "step": 260
+    },
+    {
+      "epoch": 0.09803921568627451,
+      "grad_norm": 1.5625,
+      "learning_rate": 1.955719557195572e-05,
+      "loss": 0.6871,
+      "mean_token_accuracy": 0.7864127735992346,
+      "step": 265
+    },
+    {
+      "epoch": 0.09988901220865705,
+      "grad_norm": 1.75,
+      "learning_rate": 1.9926199261992623e-05,
+      "loss": 0.629,
+      "mean_token_accuracy": 0.806636058952831,
+      "step": 270
+    },
+    {
+      "epoch": 0.10173880873103959,
+      "grad_norm": 1.8203125,
+      "learning_rate": 1.9999866506037346e-05,
+      "loss": 0.6945,
+      "mean_token_accuracy": 0.7838493749400353,
+      "step": 275
+    },
+    {
+      "epoch": 0.10358860525342212,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.9999324192922497e-05,
+      "loss": 0.6385,
+      "mean_token_accuracy": 0.8004213087756744,
+      "step": 280
+    },
+    {
+      "epoch": 0.10543840177580466,
+      "grad_norm": 1.6015625,
+      "learning_rate": 1.9998364739888954e-05,
+      "loss": 0.6762,
+      "mean_token_accuracy": 0.7904962508201961,
+      "step": 285
+    },
+    {
+      "epoch": 0.1072881982981872,
+      "grad_norm": 1.7421875,
+      "learning_rate": 1.9996988186962044e-05,
+      "loss": 0.6415,
+      "mean_token_accuracy": 0.8004900897835986,
+      "step": 290
+    },
+    {
+      "epoch": 0.10913799482056974,
+      "grad_norm": 1.671875,
+      "learning_rate": 1.999519459156716e-05,
+      "loss": 0.6506,
+      "mean_token_accuracy": 0.7948352166495436,
+      "step": 295
+    },
+    {
+      "epoch": 0.11098779134295228,
+      "grad_norm": 2.125,
+      "learning_rate": 1.9992984028527388e-05,
+      "loss": 0.6665,
+      "mean_token_accuracy": 0.7933757594308829,
+      "step": 300
+    },
+    {
+      "epoch": 0.11098779134295228,
+      "eval_loss": 0.6985837817192078,
+      "eval_mean_token_accuracy": 0.7809261873521474,
+      "eval_runtime": 89.4893,
+      "eval_samples_per_second": 1.442,
+      "eval_steps_per_second": 0.369,
+      "step": 300
+    },
+    {
+      "epoch": 0.1128375878653348,
+      "grad_norm": 1.8828125,
+      "learning_rate": 1.9990356590060363e-05,
+      "loss": 0.6942,
+      "mean_token_accuracy": 0.7864369174879459,
+      "step": 305
+    },
+    {
+      "epoch": 0.11468738438771735,
+      "grad_norm": 1.578125,
+      "learning_rate": 1.9987312385774444e-05,
+      "loss": 0.6221,
+      "mean_token_accuracy": 0.8062137325040547,
+      "step": 310
+    },
+    {
+      "epoch": 0.11653718091009989,
+      "grad_norm": 1.4921875,
+      "learning_rate": 1.9983851542664125e-05,
+      "loss": 0.6562,
+      "mean_token_accuracy": 0.7950893934122119,
+      "step": 315
+    },
+    {
+      "epoch": 0.11838697743248243,
+      "grad_norm": 1.5625,
+      "learning_rate": 1.9979974205104743e-05,
+      "loss": 0.6837,
+      "mean_token_accuracy": 0.7870972154039886,
+      "step": 320
+    },
+    {
+      "epoch": 0.12023677395486497,
+      "grad_norm": 1.6640625,
+      "learning_rate": 1.9975680534846457e-05,
+      "loss": 0.6589,
+      "mean_token_accuracy": 0.7941569505845114,
+      "step": 325
+    },
+    {
+      "epoch": 0.1220865704772475,
+      "grad_norm": 1.8359375,
+      "learning_rate": 1.99709707110075e-05,
+      "loss": 0.6705,
+      "mean_token_accuracy": 0.7909794585714891,
+      "step": 330
+    },
+    {
+      "epoch": 0.12393636699963004,
+      "grad_norm": 1.7734375,
+      "learning_rate": 1.99658449300667e-05,
+      "loss": 0.6895,
+      "mean_token_accuracy": 0.784911454898636,
+      "step": 335
+    },
+    {
+      "epoch": 0.12578616352201258,
+      "grad_norm": 2.03125,
+      "learning_rate": 1.9960303405855295e-05,
+      "loss": 0.6546,
+      "mean_token_accuracy": 0.795888251245202,
+      "step": 340
+    },
+    {
+      "epoch": 0.12763596004439512,
+      "grad_norm": 1.5703125,
+      "learning_rate": 1.9954346369548002e-05,
+      "loss": 0.6035,
+      "mean_token_accuracy": 0.8116251687595014,
+      "step": 345
+    },
+    {
+      "epoch": 0.12948575656677766,
+      "grad_norm": 1.6640625,
+      "learning_rate": 1.994797406965338e-05,
+      "loss": 0.6989,
+      "mean_token_accuracy": 0.7816051787441078,
+      "step": 350
+    },
+    {
+      "epoch": 0.1313355530891602,
+      "grad_norm": 1.734375,
+      "learning_rate": 1.9941186772003463e-05,
+      "loss": 0.7093,
+      "mean_token_accuracy": 0.7770323463799637,
+      "step": 355
+    },
+    {
+      "epoch": 0.13318534961154274,
+      "grad_norm": 1.5546875,
+      "learning_rate": 1.9933984759742664e-05,
+      "loss": 0.6309,
+      "mean_token_accuracy": 0.8021016459934547,
+      "step": 360
+    },
+    {
+      "epoch": 0.13503514613392525,
+      "grad_norm": 1.640625,
+      "learning_rate": 1.9926368333315964e-05,
+      "loss": 0.6334,
+      "mean_token_accuracy": 0.8028749333879306,
+      "step": 365
+    },
+    {
+      "epoch": 0.1368849426563078,
+      "grad_norm": 1.40625,
+      "learning_rate": 1.991833781045639e-05,
+      "loss": 0.6192,
+      "mean_token_accuracy": 0.8062173263658489,
+      "step": 370
+    },
+    {
+      "epoch": 0.13873473917869034,
+      "grad_norm": 1.7265625,
+      "learning_rate": 1.9909893526171745e-05,
+      "loss": 0.6346,
+      "mean_token_accuracy": 0.8001201518790418,
+      "step": 375
+    },
+    {
+      "epoch": 0.14058453570107288,
+      "grad_norm": 1.7421875,
+      "learning_rate": 1.9901035832730647e-05,
+      "loss": 0.6924,
+      "mean_token_accuracy": 0.7839783110482117,
+      "step": 380
+    },
+    {
+      "epoch": 0.14243433222345542,
+      "grad_norm": 1.6640625,
+      "learning_rate": 1.989176509964781e-05,
+      "loss": 0.6274,
+      "mean_token_accuracy": 0.8020502526581922,
+      "step": 385
+    },
+    {
+      "epoch": 0.14428412874583796,
+      "grad_norm": 1.7265625,
+      "learning_rate": 1.9882081713668665e-05,
+      "loss": 0.6963,
+      "mean_token_accuracy": 0.780670543584479,
+      "step": 390
+    },
+    {
+      "epoch": 0.1461339252682205,
+      "grad_norm": 1.671875,
+      "learning_rate": 1.987198607875319e-05,
+      "loss": 0.6081,
+      "mean_token_accuracy": 0.8095117915692551,
+      "step": 395
+    },
+    {
+      "epoch": 0.14798372179060304,
+      "grad_norm": 1.609375,
+      "learning_rate": 1.986147861605908e-05,
+      "loss": 0.6897,
+      "mean_token_accuracy": 0.783279006742849,
+      "step": 400
+    },
+    {
+      "epoch": 0.14798372179060304,
+      "eval_loss": 0.682924747467041,
+      "eval_mean_token_accuracy": 0.7849004804347127,
+      "eval_runtime": 89.4925,
+      "eval_samples_per_second": 1.441,
+      "eval_steps_per_second": 0.369,
+      "step": 400
+    },
+    {
+      "epoch": 0.14983351831298558,
+      "grad_norm": 1.6875,
+      "learning_rate": 1.9850559763924176e-05,
+      "loss": 0.6672,
+      "mean_token_accuracy": 0.7912924188301151,
+      "step": 405
+    },
+    {
+      "epoch": 0.15168331483536812,
+      "grad_norm": 1.578125,
+      "learning_rate": 1.9839229977848164e-05,
+      "loss": 0.6945,
+      "mean_token_accuracy": 0.7821084146257229,
+      "step": 410
+    },
+    {
+      "epoch": 0.15353311135775063,
+      "grad_norm": 1.5390625,
+      "learning_rate": 1.9827489730473597e-05,
+      "loss": 0.613,
+      "mean_token_accuracy": 0.8071375346357137,
+      "step": 415
+    },
+    {
+      "epoch": 0.15538290788013318,
+      "grad_norm": 1.3984375,
+      "learning_rate": 1.9815339511566148e-05,
+      "loss": 0.6847,
+      "mean_token_accuracy": 0.7854940613809729,
+      "step": 420
+    },
+    {
+      "epoch": 0.15723270440251572,
+      "grad_norm": 1.625,
+      "learning_rate": 1.9802779827994214e-05,
+      "loss": 0.6728,
+      "mean_token_accuracy": 0.7893773416974063,
+      "step": 425
+    },
+    {
+      "epoch": 0.15908250092489826,
+      "grad_norm": 1.6640625,
+      "learning_rate": 1.9789811203707736e-05,
+      "loss": 0.6687,
+      "mean_token_accuracy": 0.7905696239482544,
+      "step": 430
+    },
+    {
+      "epoch": 0.1609322974472808,
+      "grad_norm": 1.5859375,
+      "learning_rate": 1.9776434179716365e-05,
+      "loss": 0.6728,
+      "mean_token_accuracy": 0.788163378580386,
+      "step": 435
+    },
+    {
+      "epoch": 0.16278209396966334,
+      "grad_norm": 1.6640625,
+      "learning_rate": 1.9762649314066888e-05,
+      "loss": 0.6084,
+      "mean_token_accuracy": 0.8095275063816718,
+      "step": 440
+    },
+    {
+      "epoch": 0.16463189049204588,
+      "grad_norm": 1.8671875,
+      "learning_rate": 1.9748457181819937e-05,
+      "loss": 0.669,
+      "mean_token_accuracy": 0.78985704348369,
+      "step": 445
+    },
+    {
+      "epoch": 0.16648168701442842,
+      "grad_norm": 1.625,
+      "learning_rate": 1.9733858375026015e-05,
+      "loss": 0.6191,
+      "mean_token_accuracy": 0.8057819896532706,
+      "step": 450
+    },
+    {
+      "epoch": 0.16833148353681096,
+      "grad_norm": 1.6015625,
+      "learning_rate": 1.9718853502700783e-05,
+      "loss": 0.659,
+      "mean_token_accuracy": 0.7931944531220467,
+      "step": 455
+    },
+    {
+      "epoch": 0.1701812800591935,
+      "grad_norm": 1.5,
+      "learning_rate": 1.9703443190799673e-05,
+      "loss": 0.6287,
+      "mean_token_accuracy": 0.803461639572949,
+      "step": 460
+    },
+    {
+      "epoch": 0.17203107658157601,
+      "grad_norm": 1.875,
+      "learning_rate": 1.9687628082191748e-05,
+      "loss": 0.6705,
+      "mean_token_accuracy": 0.7901391390899182,
+      "step": 465
+    },
+    {
+      "epoch": 0.17388087310395856,
+      "grad_norm": 1.5625,
+      "learning_rate": 1.9671408836632915e-05,
+      "loss": 0.6063,
+      "mean_token_accuracy": 0.8097209966430643,
+      "step": 470
+    },
+    {
+      "epoch": 0.1757306696263411,
+      "grad_norm": 1.5078125,
+      "learning_rate": 1.9654786130738372e-05,
+      "loss": 0.6598,
+      "mean_token_accuracy": 0.7930358795856463,
+      "step": 475
+    },
+    {
+      "epoch": 0.17758046614872364,
+      "grad_norm": 1.59375,
+      "learning_rate": 1.96377606579544e-05,
+      "loss": 0.6852,
+      "mean_token_accuracy": 0.7850989878863648,
+      "step": 480
+    },
+    {
+      "epoch": 0.17943026267110618,
+      "grad_norm": 1.3671875,
+      "learning_rate": 1.9620333128529436e-05,
+      "loss": 0.6104,
+      "mean_token_accuracy": 0.8066077363544224,
+      "step": 485
+    },
+    {
+      "epoch": 0.18128005919348872,
+      "grad_norm": 1.6875,
+      "learning_rate": 1.9602504269484432e-05,
+      "loss": 0.6775,
+      "mean_token_accuracy": 0.7856656520351726,
+      "step": 490
+    },
+    {
+      "epoch": 0.18312985571587126,
+      "grad_norm": 2.125,
+      "learning_rate": 1.958427482458253e-05,
+      "loss": 0.6659,
+      "mean_token_accuracy": 0.7897591112434519,
+      "step": 495
+    },
+    {
+      "epoch": 0.1849796522382538,
+      "grad_norm": 1.3515625,
+      "learning_rate": 1.9565645554298043e-05,
+      "loss": 0.5895,
+      "mean_token_accuracy": 0.8136673273200001,
+      "step": 500
+    },
+    {
+      "epoch": 0.1849796522382538,
+      "eval_loss": 0.6728442311286926,
+      "eval_mean_token_accuracy": 0.7870013854916261,
+      "eval_runtime": 89.4951,
+      "eval_samples_per_second": 1.441,
+      "eval_steps_per_second": 0.369,
+      "step": 500
+    },
+    {
+      "epoch": 0.18682944876063634,
+      "grad_norm": 1.671875,
+      "learning_rate": 1.9546617235784716e-05,
+      "loss": 0.6845,
+      "mean_token_accuracy": 0.7845499567371429,
+      "step": 505
+    },
+    {
+      "epoch": 0.18867924528301888,
+      "grad_norm": 1.515625,
+      "learning_rate": 1.952719066284333e-05,
+      "loss": 0.6439,
+      "mean_token_accuracy": 0.7964992558951397,
+      "step": 510
+    },
+    {
+      "epoch": 0.1905290418054014,
+      "grad_norm": 1.4921875,
+      "learning_rate": 1.9507366645888544e-05,
+      "loss": 0.6343,
+      "mean_token_accuracy": 0.8003737420912586,
+      "step": 515
+    },
+    {
+      "epoch": 0.19237883832778394,
+      "grad_norm": 1.765625,
+      "learning_rate": 1.9487146011915142e-05,
+      "loss": 0.6354,
+      "mean_token_accuracy": 0.799852064852044,
+      "step": 520
+    },
+    {
+      "epoch": 0.19422863485016648,
+      "grad_norm": 1.234375,
+      "learning_rate": 1.9466529604463484e-05,
+      "loss": 0.6417,
+      "mean_token_accuracy": 0.7962427798077195,
+      "step": 525
+    },
+    {
+      "epoch": 0.19607843137254902,
+      "grad_norm": 1.546875,
+      "learning_rate": 1.9445518283584345e-05,
+      "loss": 0.6577,
+      "mean_token_accuracy": 0.7906289290077481,
+      "step": 530
+    },
+    {
+      "epoch": 0.19792822789493156,
+      "grad_norm": 1.4765625,
+      "learning_rate": 1.942411292580304e-05,
+      "loss": 0.605,
+      "mean_token_accuracy": 0.8097480409725485,
+      "step": 535
+    },
+    {
+      "epoch": 0.1997780244173141,
+      "grad_norm": 1.5859375,
+      "learning_rate": 1.9402314424082824e-05,
+      "loss": 0.6549,
+      "mean_token_accuracy": 0.7933861121852547,
+      "step": 540
+    },
+    {
+      "epoch": 0.20162782093969664,
+      "grad_norm": 1.6015625,
+      "learning_rate": 1.938012368778768e-05,
+      "loss": 0.6636,
+      "mean_token_accuracy": 0.7931950932288816,
+      "step": 545
+    },
+    {
+      "epoch": 0.20347761746207918,
+      "grad_norm": 1.453125,
+      "learning_rate": 1.935754164264437e-05,
+      "loss": 0.6346,
+      "mean_token_accuracy": 0.7994820879327266,
+      "step": 550
+    },
+    {
+      "epoch": 0.20532741398446172,
+      "grad_norm": 1.4765625,
+      "learning_rate": 1.9334569230703794e-05,
+      "loss": 0.6399,
+      "mean_token_accuracy": 0.7978100597227022,
+      "step": 555
+    },
+    {
+      "epoch": 0.20717721050684423,
+      "grad_norm": 1.6171875,
+      "learning_rate": 1.9311207410301728e-05,
+      "loss": 0.6746,
+      "mean_token_accuracy": 0.7888646042677985,
+      "step": 560
+    },
+    {
+      "epoch": 0.20902700702922677,
+      "grad_norm": 1.5546875,
+      "learning_rate": 1.9287457156018824e-05,
+      "loss": 0.6039,
+      "mean_token_accuracy": 0.8089380042612957,
+      "step": 565
+    },
+    {
+      "epoch": 0.21087680355160932,
+      "grad_norm": 1.46875,
+      "learning_rate": 1.926331945863995e-05,
+      "loss": 0.6364,
+      "mean_token_accuracy": 0.8001672921052979,
+      "step": 570
+    },
+    {
+      "epoch": 0.21272660007399186,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.9238795325112867e-05,
+      "loss": 0.6256,
+      "mean_token_accuracy": 0.8013039708225026,
+      "step": 575
+    },
+    {
+      "epoch": 0.2145763965963744,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.9213885778506228e-05,
+      "loss": 0.6234,
+      "mean_token_accuracy": 0.8027936688811925,
+      "step": 580
+    },
+    {
+      "epoch": 0.21642619311875694,
+      "grad_norm": 1.5625,
+      "learning_rate": 1.9188591857966875e-05,
+      "loss": 0.6393,
+      "mean_token_accuracy": 0.797021279033377,
+      "step": 585
+    },
+    {
+      "epoch": 0.21827598964113948,
+      "grad_norm": 1.6875,
+      "learning_rate": 1.916291461867652e-05,
+      "loss": 0.6194,
+      "mean_token_accuracy": 0.8039651707452858,
+      "step": 590
+    },
+    {
+      "epoch": 0.22012578616352202,
+      "grad_norm": 1.6875,
+      "learning_rate": 1.9136855131807705e-05,
+      "loss": 0.6422,
+      "mean_token_accuracy": 0.7964879657172526,
+      "step": 595
+    },
+    {
+      "epoch": 0.22197558268590456,
+      "grad_norm": 1.7734375,
+      "learning_rate": 1.9110414484479116e-05,
+      "loss": 0.6485,
+      "mean_token_accuracy": 0.7939532095667089,
+      "step": 600
+    },
+    {
+      "epoch": 0.22197558268590456,
+      "eval_loss": 0.6659713387489319,
+      "eval_mean_token_accuracy": 0.7885845413767406,
+      "eval_runtime": 89.4566,
+      "eval_samples_per_second": 1.442,
+      "eval_steps_per_second": 0.369,
+      "step": 600
+    },
+    {
+      "epoch": 0.2238253792082871,
+      "grad_norm": 1.609375,
+      "learning_rate": 1.908359377971025e-05,
+      "loss": 0.6202,
+      "mean_token_accuracy": 0.8055169054389353,
+      "step": 605
+    },
+    {
+      "epoch": 0.2256751757306696,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.9056394136375377e-05,
+      "loss": 0.6085,
+      "mean_token_accuracy": 0.8058548960531426,
+      "step": 610
+    },
+    {
+      "epoch": 0.22752497225305215,
+      "grad_norm": 1.6640625,
+      "learning_rate": 1.902881668915688e-05,
+      "loss": 0.6517,
+      "mean_token_accuracy": 0.7945962886280218,
+      "step": 615
+    },
+    {
+      "epoch": 0.2293747687754347,
+      "grad_norm": 1.546875,
+      "learning_rate": 1.9000862588497918e-05,
+      "loss": 0.6279,
+      "mean_token_accuracy": 0.8027668838672385,
+      "step": 620
+    },
+    {
+      "epoch": 0.23122456529781724,
+      "grad_norm": 1.46875,
+      "learning_rate": 1.897253300055443e-05,
+      "loss": 0.6427,
+      "mean_token_accuracy": 0.7986463435658552,
+      "step": 625
+    },
+    {
+      "epoch": 0.23307436182019978,
+      "grad_norm": 1.5,
+      "learning_rate": 1.8943829107146485e-05,
+      "loss": 0.6542,
+      "mean_token_accuracy": 0.7935039108063646,
+      "step": 630
+    },
+    {
+      "epoch": 0.23492415834258232,
+      "grad_norm": 1.6953125,
+      "learning_rate": 1.891475210570898e-05,
+      "loss": 0.6367,
+      "mean_token_accuracy": 0.7989973408686958,
+      "step": 635
+    },
+    {
+      "epoch": 0.23677395486496486,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.8885303209241694e-05,
+      "loss": 0.6595,
+      "mean_token_accuracy": 0.7922220691420713,
+      "step": 640
+    },
+    {
+      "epoch": 0.2386237513873474,
+      "grad_norm": 1.8125,
+      "learning_rate": 1.8855483646258677e-05,
+      "loss": 0.6051,
+      "mean_token_accuracy": 0.8069901773227958,
+      "step": 645
+    },
+    {
+      "epoch": 0.24047354790972994,
+      "grad_norm": 1.6484375,
+      "learning_rate": 1.882529466073701e-05,
+      "loss": 0.6163,
+      "mean_token_accuracy": 0.8040134740299975,
+      "step": 650
+    },
+    {
+      "epoch": 0.24232334443211248,
+      "grad_norm": 1.5390625,
+      "learning_rate": 1.879473751206489e-05,
+      "loss": 0.6412,
+      "mean_token_accuracy": 0.7975903966665798,
+      "step": 655
+    },
+    {
+      "epoch": 0.244173140954495,
+      "grad_norm": 1.5859375,
+      "learning_rate": 1.876381347498913e-05,
+      "loss": 0.6542,
+      "mean_token_accuracy": 0.794887916894406,
+      "step": 660
+    },
+    {
+      "epoch": 0.24602293747687753,
+      "grad_norm": 1.7109375,
+      "learning_rate": 1.8732523839561934e-05,
+      "loss": 0.6701,
+      "mean_token_accuracy": 0.7888407861035162,
+      "step": 665
+    },
+    {
+      "epoch": 0.24787273399926008,
+      "grad_norm": 1.6015625,
+      "learning_rate": 1.8700869911087115e-05,
+      "loss": 0.6618,
+      "mean_token_accuracy": 0.7917902683217066,
+      "step": 670
+    },
+    {
+      "epoch": 0.24972253052164262,
+      "grad_norm": 1.359375,
+      "learning_rate": 1.8668853010065633e-05,
+      "loss": 0.6276,
+      "mean_token_accuracy": 0.801046681903658,
+      "step": 675
+    },
+    {
+      "epoch": 0.25157232704402516,
+      "grad_norm": 1.5390625,
+      "learning_rate": 1.8636474472140494e-05,
+      "loss": 0.659,
+      "mean_token_accuracy": 0.7910063925799713,
+      "step": 680
+    },
+    {
+      "epoch": 0.2534221235664077,
+      "grad_norm": 1.6875,
+      "learning_rate": 1.8603735648041054e-05,
+      "loss": 0.6316,
+      "mean_token_accuracy": 0.8003829521978973,
+      "step": 685
+    },
+    {
+      "epoch": 0.25527192008879024,
+      "grad_norm": 1.421875,
+      "learning_rate": 1.857063790352665e-05,
+      "loss": 0.5663,
+      "mean_token_accuracy": 0.8194859375177957,
+      "step": 690
+    },
+    {
+      "epoch": 0.2571217166111728,
+      "grad_norm": 1.7265625,
+      "learning_rate": 1.853718261932964e-05,
+      "loss": 0.6162,
+      "mean_token_accuracy": 0.8068691848035947,
+      "step": 695
+    },
+    {
+      "epoch": 0.2589715131335553,
+      "grad_norm": 1.859375,
+      "learning_rate": 1.85033711910978e-05,
+      "loss": 0.6148,
+      "mean_token_accuracy": 0.8047477553037756,
+      "step": 700
+    },
+    {
+      "epoch": 0.2589715131335553,
+      "eval_loss": 0.661170244216919,
+      "eval_mean_token_accuracy": 0.789346045267158,
+      "eval_runtime": 89.4864,
+      "eval_samples_per_second": 1.442,
+      "eval_steps_per_second": 0.369,
+      "step": 700
+    },
+    {
+      "epoch": 0.26082130965593786,
+      "grad_norm": 1.640625,
+      "learning_rate": 1.846920502933609e-05,
+      "loss": 0.6062,
+      "mean_token_accuracy": 0.8073728975404844,
+      "step": 705
+    },
+    {
+      "epoch": 0.2626711061783204,
+      "grad_norm": 1.5859375,
+      "learning_rate": 1.843468555934783e-05,
+      "loss": 0.6485,
+      "mean_token_accuracy": 0.7950890513352781,
+      "step": 710
+    },
+    {
+      "epoch": 0.26452090270070294,
+      "grad_norm": 1.5078125,
+      "learning_rate": 1.839981422117523e-05,
+      "loss": 0.6436,
+      "mean_token_accuracy": 0.7963245807858615,
+      "step": 715
+    },
+    {
+      "epoch": 0.2663706992230855,
+      "grad_norm": 1.5546875,
+      "learning_rate": 1.836459246953931e-05,
+      "loss": 0.6265,
+      "mean_token_accuracy": 0.799983712114828,
+      "step": 720
+    },
+    {
+      "epoch": 0.268220495745468,
+      "grad_norm": 1.5,
+      "learning_rate": 1.8329021773779242e-05,
+      "loss": 0.6384,
+      "mean_token_accuracy": 0.798832024376296,
+      "step": 725
+    },
+    {
+      "epoch": 0.2700702922678505,
+      "grad_norm": 1.53125,
+      "learning_rate": 1.8293103617791016e-05,
+      "loss": 0.6449,
+      "mean_token_accuracy": 0.794099280788323,
+      "step": 730
+    },
+    {
+      "epoch": 0.27192008879023305,
+      "grad_norm": 1.5078125,
+      "learning_rate": 1.825683949996556e-05,
+      "loss": 0.6598,
+      "mean_token_accuracy": 0.7926169244804246,
+      "step": 735
+    },
+    {
+      "epoch": 0.2737698853126156,
+      "grad_norm": 1.59375,
+      "learning_rate": 1.8220230933126234e-05,
+      "loss": 0.5907,
+      "mean_token_accuracy": 0.8115653285474618,
+      "step": 740
+    },
+    {
+      "epoch": 0.27561968183499813,
+      "grad_norm": 1.7890625,
+      "learning_rate": 1.81832794444657e-05,
+      "loss": 0.6218,
+      "mean_token_accuracy": 0.8015434097801298,
+      "step": 745
+    },
+    {
+      "epoch": 0.27746947835738067,
+      "grad_norm": 1.5859375,
+      "learning_rate": 1.814598657548223e-05,
+      "loss": 0.6166,
+      "mean_token_accuracy": 0.8066213903295013,
+      "step": 750
+    },
+    {
+      "epoch": 0.2793192748797632,
+      "grad_norm": 1.6328125,
+      "learning_rate": 1.8108353881915403e-05,
+      "loss": 0.6241,
+      "mean_token_accuracy": 0.8010177723486601,
+      "step": 755
+    },
+    {
+      "epoch": 0.28116907140214575,
+      "grad_norm": 1.578125,
+      "learning_rate": 1.8070382933681185e-05,
+      "loss": 0.6361,
+      "mean_token_accuracy": 0.7971330814719428,
+      "step": 760
+    },
+    {
+      "epoch": 0.2830188679245283,
+      "grad_norm": 1.6875,
+      "learning_rate": 1.803207531480645e-05,
+      "loss": 0.6308,
+      "mean_token_accuracy": 0.7973929924777202,
+      "step": 765
+    },
+    {
+      "epoch": 0.28486866444691084,
+      "grad_norm": 1.6953125,
+      "learning_rate": 1.79934326233629e-05,
+      "loss": 0.6063,
+      "mean_token_accuracy": 0.806685878743868,
+      "step": 770
+    },
+    {
+      "epoch": 0.2867184609692934,
+      "grad_norm": 1.6328125,
+      "learning_rate": 1.7954456471400393e-05,
+      "loss": 0.6463,
+      "mean_token_accuracy": 0.7944443259506135,
+      "step": 775
+    },
+    {
+      "epoch": 0.2885682574916759,
+      "grad_norm": 1.3671875,
+      "learning_rate": 1.7915148484879706e-05,
+      "loss": 0.6356,
+      "mean_token_accuracy": 0.7991871275852491,
+      "step": 780
+    },
+    {
+      "epoch": 0.29041805401405846,
+      "grad_norm": 1.4453125,
+      "learning_rate": 1.7875510303604678e-05,
+      "loss": 0.6061,
+      "mean_token_accuracy": 0.809440939198223,
+      "step": 785
+    },
+    {
+      "epoch": 0.292267850536441,
+      "grad_norm": 1.625,
+      "learning_rate": 1.783554358115384e-05,
+      "loss": 0.6644,
+      "mean_token_accuracy": 0.7890826509644704,
+      "step": 790
+    },
+    {
+      "epoch": 0.29411764705882354,
+      "grad_norm": 1.4765625,
+      "learning_rate": 1.7795249984811397e-05,
+      "loss": 0.6003,
+      "mean_token_accuracy": 0.8101999171186793,
+      "step": 795
+    },
+    {
+      "epoch": 0.2959674435812061,
+      "grad_norm": 1.8046875,
+      "learning_rate": 1.77546311954977e-05,
+      "loss": 0.6533,
+      "mean_token_accuracy": 0.7934218494709553,
+      "step": 800
+    },
+    {
+      "epoch": 0.2959674435812061,
+      "eval_loss": 0.6562904715538025,
+      "eval_mean_token_accuracy": 0.7902898340765919,
+      "eval_runtime": 89.6164,
+      "eval_samples_per_second": 1.439,
+      "eval_steps_per_second": 0.368,
+      "step": 800
+    },
+    {
+      "epoch": 0.2978172401035886,
+      "grad_norm": 1.515625,
+      "learning_rate": 1.7713688907699107e-05,
+      "loss": 0.6735,
+      "mean_token_accuracy": 0.7867014949644692,
+      "step": 805
+    },
+    {
+      "epoch": 0.29966703662597116,
+      "grad_norm": 1.5078125,
+      "learning_rate": 1.767242482939731e-05,
+      "loss": 0.6398,
+      "mean_token_accuracy": 0.7978769080711773,
+      "step": 810
+    },
+    {
+      "epoch": 0.3015168331483537,
+      "grad_norm": 1.6328125,
+      "learning_rate": 1.7630840681998068e-05,
+      "loss": 0.6013,
+      "mean_token_accuracy": 0.8096431348991004,
+      "step": 815
+    },
+    {
+      "epoch": 0.30336662967073624,
+      "grad_norm": 1.5390625,
+      "learning_rate": 1.75889382002594e-05,
+      "loss": 0.6409,
+      "mean_token_accuracy": 0.7962791229258687,
+      "step": 820
+    },
+    {
+      "epoch": 0.3052164261931188,
+      "grad_norm": 1.640625,
+      "learning_rate": 1.754671913221923e-05,
+      "loss": 0.6381,
+      "mean_token_accuracy": 0.7973464796497893,
+      "step": 825
+    },
+    {
+      "epoch": 0.30706622271550127,
+      "grad_norm": 1.5,
+      "learning_rate": 1.7504185239122432e-05,
+      "loss": 0.6114,
+      "mean_token_accuracy": 0.805420972567774,
+      "step": 830
+    },
+    {
+      "epoch": 0.3089160192378838,
+      "grad_norm": 1.4921875,
+      "learning_rate": 1.7461338295347404e-05,
+      "loss": 0.6016,
+      "mean_token_accuracy": 0.8097418734044766,
+      "step": 835
+    },
+    {
+      "epoch": 0.31076581576026635,
+      "grad_norm": 1.40625,
+      "learning_rate": 1.7418180088332005e-05,
+      "loss": 0.6147,
+      "mean_token_accuracy": 0.8036096487014749,
+      "step": 840
+    },
+    {
+      "epoch": 0.3126156122826489,
+      "grad_norm": 1.40625,
+      "learning_rate": 1.7374712418498997e-05,
+      "loss": 0.6181,
+      "mean_token_accuracy": 0.8032291560159276,
+      "step": 845
+    },
+    {
+      "epoch": 0.31446540880503143,
+      "grad_norm": 1.6171875,
+      "learning_rate": 1.7330937099180956e-05,
+      "loss": 0.6508,
+      "mean_token_accuracy": 0.7932453588501767,
+      "step": 850
+    },
+    {
+      "epoch": 0.316315205327414,
+      "grad_norm": 1.640625,
+      "learning_rate": 1.7286855956544616e-05,
+      "loss": 0.6167,
+      "mean_token_accuracy": 0.8034507035193359,
+      "step": 855
+    },
+    {
+      "epoch": 0.3181650018497965,
+      "grad_norm": 1.5859375,
+      "learning_rate": 1.7242470829514674e-05,
+      "loss": 0.6474,
+      "mean_token_accuracy": 0.7931935779688848,
+      "step": 860
+    },
+    {
+      "epoch": 0.32001479837217905,
+      "grad_norm": 1.5546875,
+      "learning_rate": 1.7197783569697084e-05,
+      "loss": 0.6247,
+      "mean_token_accuracy": 0.8033568872388832,
+      "step": 865
+    },
+    {
+      "epoch": 0.3218645948945616,
+      "grad_norm": 1.7734375,
+      "learning_rate": 1.715279604130183e-05,
+      "loss": 0.6624,
+      "mean_token_accuracy": 0.7916663261272019,
+      "step": 870
+    },
+    {
+      "epoch": 0.32371439141694414,
+      "grad_norm": 1.578125,
+      "learning_rate": 1.7107510121065138e-05,
+      "loss": 0.6115,
+      "mean_token_accuracy": 0.8060779428709661,
+      "step": 875
+    },
+    {
+      "epoch": 0.3255641879393267,
+      "grad_norm": 1.6875,
+      "learning_rate": 1.7061927698171187e-05,
+      "loss": 0.57,
+      "mean_token_accuracy": 0.8167656282687918,
+      "step": 880
+    },
+    {
+      "epoch": 0.3274139844617092,
+      "grad_norm": 1.7109375,
+      "learning_rate": 1.7016050674173304e-05,
+      "loss": 0.6309,
+      "mean_token_accuracy": 0.8030734112570179,
+      "step": 885
+    },
+    {
+      "epoch": 0.32926378098409176,
+      "grad_norm": 1.5234375,
+      "learning_rate": 1.696988096291463e-05,
+      "loss": 0.6566,
+      "mean_token_accuracy": 0.7914602262945556,
+      "step": 890
+    },
+    {
+      "epoch": 0.3311135775064743,
+      "grad_norm": 1.4453125,
+      "learning_rate": 1.6923420490448298e-05,
+      "loss": 0.6315,
+      "mean_token_accuracy": 0.8005812904170252,
+      "step": 895
+    },
+    {
+      "epoch": 0.33296337402885684,
+      "grad_norm": 1.5078125,
+      "learning_rate": 1.687667119495705e-05,
+      "loss": 0.6575,
+      "mean_token_accuracy": 0.7923108787065782,
+      "step": 900
+    },
+    {
+      "epoch": 0.33296337402885684,
+      "eval_loss": 0.6524114012718201,
+      "eval_mean_token_accuracy": 0.7915192679536375,
+      "eval_runtime": 89.606,
+      "eval_samples_per_second": 1.44,
+      "eval_steps_per_second": 0.368,
+      "step": 900
+    },
+    {
+      "epoch": 0.3348131705512394,
+      "grad_norm": 1.6875,
+      "learning_rate": 1.6829635026672432e-05,
+      "loss": 0.62,
+      "mean_token_accuracy": 0.8020949300485505,
+      "step": 905
+    },
+    {
+      "epoch": 0.3366629670736219,
+      "grad_norm": 1.5,
+      "learning_rate": 1.6782313947793377e-05,
+      "loss": 0.6299,
+      "mean_token_accuracy": 0.799654433976644,
+      "step": 910
+    },
+    {
+      "epoch": 0.33851276359600446,
+      "grad_norm": 1.46875,
+      "learning_rate": 1.6734709932404404e-05,
+      "loss": 0.6298,
+      "mean_token_accuracy": 0.8003587858084458,
+      "step": 915
+    },
+    {
+      "epoch": 0.340362560118387,
+      "grad_norm": 1.609375,
+      "learning_rate": 1.668682496639323e-05,
+      "loss": 0.6328,
+      "mean_token_accuracy": 0.7990015427815018,
+      "step": 920
+    },
+    {
+      "epoch": 0.3422123566407695,
+      "grad_norm": 1.40625,
+      "learning_rate": 1.663866104736793e-05,
+      "loss": 0.5802,
+      "mean_token_accuracy": 0.814437940730177,
+      "step": 925
+    },
+    {
+      "epoch": 0.34406215316315203,
+      "grad_norm": 1.609375,
+      "learning_rate": 1.6590220184573623e-05,
+      "loss": 0.661,
+      "mean_token_accuracy": 0.7891311462047851,
+      "step": 930
+    },
+    {
+      "epoch": 0.34591194968553457,
+      "grad_norm": 1.6015625,
+      "learning_rate": 1.6541504398808633e-05,
+      "loss": 0.6383,
+      "mean_token_accuracy": 0.7998139182847727,
+      "step": 935
+    },
+    {
+      "epoch": 0.3477617462079171,
+      "grad_norm": 1.421875,
+      "learning_rate": 1.6492515722340193e-05,
+      "loss": 0.6745,
+      "mean_token_accuracy": 0.7854493413477951,
+      "step": 940
+    },
+    {
+      "epoch": 0.34961154273029965,
+      "grad_norm": 1.625,
+      "learning_rate": 1.6443256198819665e-05,
+      "loss": 0.6558,
+      "mean_token_accuracy": 0.7913892138004563,
+      "step": 945
+    },
+    {
+      "epoch": 0.3514613392526822,
+      "grad_norm": 1.5078125,
+      "learning_rate": 1.6393727883197292e-05,
+      "loss": 0.6191,
+      "mean_token_accuracy": 0.8010231204738876,
+      "step": 950
+    },
+    {
+      "epoch": 0.35331113577506473,
+      "grad_norm": 1.546875,
+      "learning_rate": 1.6343932841636455e-05,
+      "loss": 0.6404,
+      "mean_token_accuracy": 0.7966922303266114,
+      "step": 955
+    },
+    {
+      "epoch": 0.3551609322974473,
+      "grad_norm": 1.453125,
+      "learning_rate": 1.6293873151427505e-05,
+      "loss": 0.6513,
+      "mean_token_accuracy": 0.7944516515365347,
+      "step": 960
+    },
+    {
+      "epoch": 0.3570107288198298,
+      "grad_norm": 1.6171875,
+      "learning_rate": 1.6243550900901076e-05,
+      "loss": 0.6496,
+      "mean_token_accuracy": 0.7923123310559392,
+      "step": 965
+    },
+    {
+      "epoch": 0.35886052534221236,
+      "grad_norm": 1.4921875,
+      "learning_rate": 1.6192968189340993e-05,
+      "loss": 0.6442,
+      "mean_token_accuracy": 0.7955391652128261,
+      "step": 970
+    },
+    {
+      "epoch": 0.3607103218645949,
+      "grad_norm": 1.390625,
+      "learning_rate": 1.6142127126896682e-05,
+      "loss": 0.6501,
+      "mean_token_accuracy": 0.7932307318742777,
+      "step": 975
+    },
+    {
+      "epoch": 0.36256011838697744,
+      "grad_norm": 1.3359375,
+      "learning_rate": 1.6091029834495133e-05,
+      "loss": 0.6096,
+      "mean_token_accuracy": 0.8066235320811712,
+      "step": 980
+    },
+    {
+      "epoch": 0.36440991490936,
+      "grad_norm": 1.4921875,
+      "learning_rate": 1.603967844375245e-05,
+      "loss": 0.6257,
+      "mean_token_accuracy": 0.8004245838839372,
+      "step": 985
+    },
+    {
+      "epoch": 0.3662597114317425,
+      "grad_norm": 1.5859375,
+      "learning_rate": 1.59880750968849e-05,
+      "loss": 0.6175,
+      "mean_token_accuracy": 0.8011132737724307,
+      "step": 990
+    },
+    {
+      "epoch": 0.36810950795412506,
+      "grad_norm": 1.609375,
+      "learning_rate": 1.593622194661956e-05,
+      "loss": 0.585,
+      "mean_token_accuracy": 0.814000299442783,
+      "step": 995
+    },
+    {
+      "epoch": 0.3699593044765076,
+      "grad_norm": 1.40625,
+      "learning_rate": 1.5884121156104496e-05,
+      "loss": 0.6174,
+      "mean_token_accuracy": 0.8035369649193376,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3699593044765076,
+      "eval_loss": 0.6492791771888733,
+      "eval_mean_token_accuracy": 0.7922599270220881,
+      "eval_runtime": 89.6554,
+      "eval_samples_per_second": 1.439,
+      "eval_steps_per_second": 0.368,
+      "step": 1000
+    },
+    {
+      "epoch": 0.37180910099889014,
+      "grad_norm": 1.6015625,
+      "learning_rate": 1.5831774898818558e-05,
+      "loss": 0.6548,
+      "mean_token_accuracy": 0.7927545199348696,
+      "step": 1005
+    },
+    {
+      "epoch": 0.3736588975212727,
+      "grad_norm": 1.609375,
+      "learning_rate": 1.5779185358480666e-05,
+      "loss": 0.6348,
+      "mean_token_accuracy": 0.798432584436958,
+      "step": 1010
+    },
+    {
+      "epoch": 0.3755086940436552,
+      "grad_norm": 1.484375,
+      "learning_rate": 1.5726354728958736e-05,
+      "loss": 0.6221,
+      "mean_token_accuracy": 0.8041254372881184,
+      "step": 1015
+    },
+    {
+      "epoch": 0.37735849056603776,
+      "grad_norm": 1.515625,
+      "learning_rate": 1.5673285214178176e-05,
+      "loss": 0.6165,
+      "mean_token_accuracy": 0.8033880504490309,
+      "step": 1020
+    },
+    {
+      "epoch": 0.37920828708842025,
+      "grad_norm": 1.46875,
+      "learning_rate": 1.5619979028029898e-05,
+      "loss": 0.6068,
+      "mean_token_accuracy": 0.8045155094889667,
+      "step": 1025
+    },
+    {
+      "epoch": 0.3810580836108028,
+      "grad_norm": 1.5234375,
+      "learning_rate": 1.5566438394278005e-05,
+      "loss": 0.5815,
+      "mean_token_accuracy": 0.8127077124432907,
+      "step": 1030
+    },
+    {
+      "epoch": 0.38290788013318533,
+      "grad_norm": 1.5078125,
+      "learning_rate": 1.5512665546467008e-05,
+      "loss": 0.6809,
+      "mean_token_accuracy": 0.7824735840973288,
+      "step": 1035
+    },
+    {
+      "epoch": 0.38475767665556787,
+      "grad_norm": 1.375,
+      "learning_rate": 1.5458662727828633e-05,
+      "loss": 0.6113,
+      "mean_token_accuracy": 0.8053544662689619,
+      "step": 1040
+    },
+    {
+      "epoch": 0.3866074731779504,
+      "grad_norm": 1.453125,
+      "learning_rate": 1.540443219118827e-05,
+      "loss": 0.6322,
+      "mean_token_accuracy": 0.7981511576366589,
+      "step": 1045
+    },
+    {
+      "epoch": 0.38845726970033295,
+      "grad_norm": 1.5234375,
+      "learning_rate": 1.5349976198870974e-05,
+      "loss": 0.6353,
+      "mean_token_accuracy": 0.799042859122671,
+      "step": 1050
+    },
+    {
+      "epoch": 0.3903070662227155,
+      "grad_norm": 1.5859375,
+      "learning_rate": 1.529529702260709e-05,
+      "loss": 0.6361,
+      "mean_token_accuracy": 0.7966232102060518,
+      "step": 1055
+    },
+    {
+      "epoch": 0.39215686274509803,
+      "grad_norm": 1.46875,
+      "learning_rate": 1.5240396943437483e-05,
+      "loss": 0.6607,
+      "mean_token_accuracy": 0.7911870086371723,
+      "step": 1060
+    },
+    {
+      "epoch": 0.3940066592674806,
+      "grad_norm": 1.40625,
+      "learning_rate": 1.5185278251618391e-05,
+      "loss": 0.6352,
+      "mean_token_accuracy": 0.7971305063080498,
+      "step": 1065
+    },
+    {
+      "epoch": 0.3958564557898631,
+      "grad_norm": 1.4453125,
+      "learning_rate": 1.5129943246525871e-05,
+      "loss": 0.6403,
+      "mean_token_accuracy": 0.7959993996736948,
+      "step": 1070
+    },
+    {
+      "epoch": 0.39770625231224566,
+      "grad_norm": 1.3671875,
+      "learning_rate": 1.5074394236559871e-05,
+      "loss": 0.5777,
+      "mean_token_accuracy": 0.8149735020480074,
+      "step": 1075
+    },
+    {
+      "epoch": 0.3995560488346282,
+      "grad_norm": 1.578125,
+      "learning_rate": 1.5018633539047955e-05,
+      "loss": 0.6486,
+      "mean_token_accuracy": 0.7940828638311908,
+      "step": 1080
+    },
+    {
+      "epoch": 0.40140584535701074,
+      "grad_norm": 1.453125,
+      "learning_rate": 1.4962663480148606e-05,
+      "loss": 0.6047,
+      "mean_token_accuracy": 0.8072560422872016,
+      "step": 1085
+    },
+    {
+      "epoch": 0.4032556418793933,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.4906486394754198e-05,
+      "loss": 0.5876,
+      "mean_token_accuracy": 0.8125326436899881,
+      "step": 1090
+    },
+    {
+      "epoch": 0.4051054384017758,
+      "grad_norm": 1.5703125,
+      "learning_rate": 1.4850104626393598e-05,
+      "loss": 0.6059,
+      "mean_token_accuracy": 0.8075963725338712,
+      "step": 1095
+    },
+    {
+      "epoch": 0.40695523492415836,
+      "grad_norm": 1.6328125,
+      "learning_rate": 1.4793520527134381e-05,
+      "loss": 0.6157,
+      "mean_token_accuracy": 0.8048102235227077,
+      "step": 1100
+    },
+    {
+      "epoch": 0.40695523492415836,
+      "eval_loss": 0.6465983986854553,
+      "eval_mean_token_accuracy": 0.7927890480652175,
+      "eval_runtime": 88.662,
+      "eval_samples_per_second": 1.455,
+      "eval_steps_per_second": 0.372,
+      "step": 1100
+    },
+    {
+      "epoch": 0.4088050314465409,
+      "grad_norm": 1.546875,
+      "learning_rate": 1.473673645748473e-05,
+      "loss": 0.6242,
+      "mean_token_accuracy": 0.7996578198270112,
+      "step": 1105
+    },
+    {
+      "epoch": 0.41065482796892344,
+      "grad_norm": 1.46875,
+      "learning_rate": 1.4679754786294954e-05,
+      "loss": 0.6093,
+      "mean_token_accuracy": 0.8039842588845403,
+      "step": 1110
+    },
+    {
+      "epoch": 0.412504624491306,
+      "grad_norm": 1.359375,
+      "learning_rate": 1.4622577890658668e-05,
+      "loss": 0.6044,
+      "mean_token_accuracy": 0.8057674562401933,
+      "step": 1115
+    },
+    {
+      "epoch": 0.41435442101368847,
+      "grad_norm": 1.640625,
+      "learning_rate": 1.4565208155813626e-05,
+      "loss": 0.6008,
+      "mean_token_accuracy": 0.8088222481664176,
+      "step": 1120
+    },
+    {
+      "epoch": 0.416204217536071,
+      "grad_norm": 1.4765625,
+      "learning_rate": 1.4507647975042221e-05,
+      "loss": 0.6232,
+      "mean_token_accuracy": 0.80052939021627,
+      "step": 1125
+    },
+    {
+      "epoch": 0.41805401405845355,
+      "grad_norm": 1.421875,
+      "learning_rate": 1.4449899749571643e-05,
+      "loss": 0.6108,
+      "mean_token_accuracy": 0.8042650462769881,
+      "step": 1130
+    },
+    {
+      "epoch": 0.4199038105808361,
+      "grad_norm": 1.3203125,
+      "learning_rate": 1.4391965888473705e-05,
+      "loss": 0.5983,
+      "mean_token_accuracy": 0.8074625864491232,
+      "step": 1135
+    },
+    {
+      "epoch": 0.42175360710321863,
+      "grad_norm": 1.4609375,
+      "learning_rate": 1.4333848808564353e-05,
+      "loss": 0.6306,
+      "mean_token_accuracy": 0.7985940566693959,
+      "step": 1140
+    },
+    {
+      "epoch": 0.42360340362560117,
+      "grad_norm": 1.5859375,
+      "learning_rate": 1.4275550934302822e-05,
+      "loss": 0.5966,
+      "mean_token_accuracy": 0.8099505664815225,
+      "step": 1145
+    },
+    {
+      "epoch": 0.4254532001479837,
+      "grad_norm": 1.6328125,
+      "learning_rate": 1.4217074697690532e-05,
+      "loss": 0.6462,
+      "mean_token_accuracy": 0.7943994488547826,
+      "step": 1150
+    },
+    {
+      "epoch": 0.42730299667036625,
+      "grad_norm": 1.40625,
+      "learning_rate": 1.4158422538169596e-05,
+      "loss": 0.6319,
+      "mean_token_accuracy": 0.7977675460969558,
+      "step": 1155
+    },
+    {
+      "epoch": 0.4291527931927488,
+      "grad_norm": 1.640625,
+      "learning_rate": 1.4099596902521073e-05,
+      "loss": 0.6288,
+      "mean_token_accuracy": 0.8011337245764597,
+      "step": 1160
+    },
+    {
+      "epoch": 0.43100258971513133,
+      "grad_norm": 1.515625,
+      "learning_rate": 1.40406002447629e-05,
+      "loss": 0.6465,
+      "mean_token_accuracy": 0.7952717477194938,
+      "step": 1165
+    },
+    {
+      "epoch": 0.4328523862375139,
+      "grad_norm": 1.4375,
+      "learning_rate": 1.39814350260475e-05,
+      "loss": 0.6123,
+      "mean_token_accuracy": 0.8032346756924053,
+      "step": 1170
+    },
+    {
+      "epoch": 0.4347021827598964,
+      "grad_norm": 1.640625,
+      "learning_rate": 1.392210371455913e-05,
+      "loss": 0.6206,
+      "mean_token_accuracy": 0.8022290885180814,
+      "step": 1175
+    },
+    {
+      "epoch": 0.43655197928227896,
+      "grad_norm": 1.4765625,
+      "learning_rate": 1.3862608785410903e-05,
+      "loss": 0.6037,
+      "mean_token_accuracy": 0.8084956493131734,
+      "step": 1180
+    },
+    {
+      "epoch": 0.4384017758046615,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.3802952720541543e-05,
+      "loss": 0.5795,
+      "mean_token_accuracy": 0.8150522721445338,
+      "step": 1185
+    },
+    {
+      "epoch": 0.44025157232704404,
+      "grad_norm": 1.4140625,
+      "learning_rate": 1.3743138008611838e-05,
+      "loss": 0.6086,
+      "mean_token_accuracy": 0.8058791995891041,
+      "step": 1190
+    },
+    {
+      "epoch": 0.4421013688494266,
+      "grad_norm": 1.3984375,
+      "learning_rate": 1.3683167144900833e-05,
+      "loss": 0.6097,
+      "mean_token_accuracy": 0.8070925245148421,
+      "step": 1195
+    },
+    {
+      "epoch": 0.4439511653718091,
+      "grad_norm": 1.59375,
+      "learning_rate": 1.3623042631201713e-05,
+      "loss": 0.6497,
+      "mean_token_accuracy": 0.7923035658124328,
+      "step": 1200
+    },
+    {
+      "epoch": 0.4439511653718091,
+      "eval_loss": 0.6446292996406555,
+      "eval_mean_token_accuracy": 0.793524480067428,
+      "eval_runtime": 89.2797,
+      "eval_samples_per_second": 1.445,
+      "eval_steps_per_second": 0.37,
+      "step": 1200
+    },
+    {
+      "epoch": 0.44580096189419166,
+      "grad_norm": 1.59375,
+      "learning_rate": 1.3562766975717468e-05,
+      "loss": 0.612,
+      "mean_token_accuracy": 0.8019620322322162,
+      "step": 1205
+    },
+    {
+      "epoch": 0.4476507584165742,
+      "grad_norm": 1.2421875,
+      "learning_rate": 1.3502342692956229e-05,
+      "loss": 0.5514,
+      "mean_token_accuracy": 0.8231528326487867,
+      "step": 1210
+    },
+    {
+      "epoch": 0.44950055493895674,
+      "grad_norm": 1.625,
+      "learning_rate": 1.3441772303626387e-05,
+      "loss": 0.6378,
+      "mean_token_accuracy": 0.7959821084867433,
+      "step": 1215
+    },
+    {
+      "epoch": 0.4513503514613392,
+      "grad_norm": 1.5390625,
+      "learning_rate": 1.3381058334531436e-05,
+      "loss": 0.5745,
+      "mean_token_accuracy": 0.8149386671282464,
+      "step": 1220
+    },
+    {
+      "epoch": 0.45320014798372177,
+      "grad_norm": 1.640625,
+      "learning_rate": 1.3320203318464552e-05,
+      "loss": 0.6321,
+      "mean_token_accuracy": 0.7992501223396398,
+      "step": 1225
+    },
+    {
+      "epoch": 0.4550499445061043,
+      "grad_norm": 1.5078125,
+      "learning_rate": 1.3259209794102952e-05,
+      "loss": 0.6111,
+      "mean_token_accuracy": 0.8055151366474279,
+      "step": 1230
+    },
+    {
+      "epoch": 0.45689974102848685,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.319808030590197e-05,
+      "loss": 0.6038,
+      "mean_token_accuracy": 0.8075011003179897,
+      "step": 1235
+    },
+    {
+      "epoch": 0.4587495375508694,
+      "grad_norm": 1.3671875,
+      "learning_rate": 1.3136817403988918e-05,
+      "loss": 0.6388,
+      "mean_token_accuracy": 0.7966051838393937,
+      "step": 1240
+    },
+    {
+      "epoch": 0.46059933407325193,
+      "grad_norm": 1.46875,
+      "learning_rate": 1.3075423644056699e-05,
+      "loss": 0.6615,
+      "mean_token_accuracy": 0.78788567921951,
+      "step": 1245
+    },
+    {
+      "epoch": 0.46244913059563447,
+      "grad_norm": 1.5390625,
+      "learning_rate": 1.3013901587257207e-05,
+      "loss": 0.6253,
+      "mean_token_accuracy": 0.8015557549939769,
+      "step": 1250
+    },
+    {
+      "epoch": 0.464298927118017,
+      "grad_norm": 1.40625,
+      "learning_rate": 1.2952253800094467e-05,
+      "loss": 0.6275,
+      "mean_token_accuracy": 0.798606470568502,
+      "step": 1255
+    },
+    {
+      "epoch": 0.46614872364039955,
+      "grad_norm": 1.609375,
+      "learning_rate": 1.2890482854317565e-05,
+      "loss": 0.6834,
+      "mean_token_accuracy": 0.7821290938501924,
+      "step": 1260
+    },
+    {
+      "epoch": 0.4679985201627821,
+      "grad_norm": 1.5390625,
+      "learning_rate": 1.2828591326813382e-05,
+      "loss": 0.6397,
+      "mean_token_accuracy": 0.7969137820925224,
+      "step": 1265
+    },
+    {
+      "epoch": 0.46984831668516464,
+      "grad_norm": 1.4609375,
+      "learning_rate": 1.2766581799499078e-05,
+      "loss": 0.6202,
+      "mean_token_accuracy": 0.8020725569528727,
+      "step": 1270
+    },
+    {
+      "epoch": 0.4716981132075472,
+      "grad_norm": 1.5625,
+      "learning_rate": 1.2704456859214397e-05,
+      "loss": 0.657,
+      "mean_token_accuracy": 0.7897297173350418,
+      "step": 1275
+    },
+    {
+      "epoch": 0.4735479097299297,
+      "grad_norm": 1.453125,
+      "learning_rate": 1.264221909761374e-05,
+      "loss": 0.5664,
+      "mean_token_accuracy": 0.8193003194015439,
+      "step": 1280
+    },
+    {
+      "epoch": 0.47539770625231226,
+      "grad_norm": 1.421875,
+      "learning_rate": 1.2579871111058042e-05,
+      "loss": 0.6545,
+      "mean_token_accuracy": 0.7930202188051163,
+      "step": 1285
+    },
+    {
+      "epoch": 0.4772475027746948,
+      "grad_norm": 1.453125,
+      "learning_rate": 1.251741550050649e-05,
+      "loss": 0.5864,
+      "mean_token_accuracy": 0.8118832096191291,
+      "step": 1290
+    },
+    {
+      "epoch": 0.47909729929707734,
+      "grad_norm": 1.3203125,
+      "learning_rate": 1.2454854871407993e-05,
+      "loss": 0.6385,
+      "mean_token_accuracy": 0.7980835441016285,
+      "step": 1295
+    },
+    {
+      "epoch": 0.4809470958194599,
+      "grad_norm": 1.4921875,
+      "learning_rate": 1.2392191833592495e-05,
+      "loss": 0.5965,
+      "mean_token_accuracy": 0.8100436231727027,
+      "step": 1300
+    },
+    {
+      "epoch": 0.4809470958194599,
+      "eval_loss": 0.6423260569572449,
+      "eval_mean_token_accuracy": 0.7935656088119338,
+      "eval_runtime": 89.3328,
+      "eval_samples_per_second": 1.444,
+      "eval_steps_per_second": 0.369,
+      "step": 1300
+    },
+    {
+      "epoch": 0.4827968923418424,
+      "grad_norm": 1.375,
+      "learning_rate": 1.2329429001162114e-05,
+      "loss": 0.6025,
+      "mean_token_accuracy": 0.8073580633493688,
+      "step": 1305
+    },
+    {
+      "epoch": 0.48464668886422496,
+      "grad_norm": 1.5546875,
+      "learning_rate": 1.226656899238207e-05,
+      "loss": 0.5872,
+      "mean_token_accuracy": 0.8119950659322296,
+      "step": 1310
+    },
+    {
+      "epoch": 0.48649648538660745,
+      "grad_norm": 1.59375,
+      "learning_rate": 1.2203614429571475e-05,
+      "loss": 0.6542,
+      "mean_token_accuracy": 0.7913764966388273,
+      "step": 1315
+    },
+    {
+      "epoch": 0.48834628190899,
+      "grad_norm": 1.4140625,
+      "learning_rate": 1.214056793899394e-05,
+      "loss": 0.609,
+      "mean_token_accuracy": 0.8056629130019098,
+      "step": 1320
+    },
+    {
+      "epoch": 0.49019607843137253,
+      "grad_norm": 1.5078125,
+      "learning_rate": 1.207743215074801e-05,
+      "loss": 0.5935,
+      "mean_token_accuracy": 0.8097664075213468,
+      "step": 1325
+    },
+    {
+      "epoch": 0.49204587495375507,
+      "grad_norm": 1.4765625,
+      "learning_rate": 1.2014209698657424e-05,
+      "loss": 0.6321,
+      "mean_token_accuracy": 0.7995131589666756,
+      "step": 1330
+    },
+    {
+      "epoch": 0.4938956714761376,
+      "grad_norm": 1.5859375,
+      "learning_rate": 1.1950903220161286e-05,
+      "loss": 0.6236,
+      "mean_token_accuracy": 0.7996765940808375,
+      "step": 1335
+    },
+    {
+      "epoch": 0.49574546799852015,
+      "grad_norm": 1.546875,
+      "learning_rate": 1.1887515356203994e-05,
+      "loss": 0.6332,
+      "mean_token_accuracy": 0.7977098620882361,
+      "step": 1340
+    },
+    {
+      "epoch": 0.4975952645209027,
+      "grad_norm": 1.3125,
+      "learning_rate": 1.1824048751125101e-05,
+      "loss": 0.6403,
+      "mean_token_accuracy": 0.7962522755568017,
+      "step": 1345
+    },
+    {
+      "epoch": 0.49944506104328523,
+      "grad_norm": 1.5078125,
+      "learning_rate": 1.1760506052548984e-05,
+      "loss": 0.6329,
+      "mean_token_accuracy": 0.7970971517227641,
+      "step": 1350
+    },
+    {
+      "epoch": 0.5012948575656678,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.1696889911274394e-05,
+      "loss": 0.6022,
+      "mean_token_accuracy": 0.80603544612979,
+      "step": 1355
+    },
+    {
+      "epoch": 0.5031446540880503,
+      "grad_norm": 1.5,
+      "learning_rate": 1.1633202981163882e-05,
+      "loss": 0.6443,
+      "mean_token_accuracy": 0.79562622940442,
+      "step": 1360
+    },
+    {
+      "epoch": 0.5049944506104328,
+      "grad_norm": 1.484375,
+      "learning_rate": 1.1569447919033086e-05,
+      "loss": 0.6357,
+      "mean_token_accuracy": 0.7959378011357524,
+      "step": 1365
+    },
+    {
+      "epoch": 0.5068442471328154,
+      "grad_norm": 1.4765625,
+      "learning_rate": 1.1505627384539888e-05,
+      "loss": 0.6212,
+      "mean_token_accuracy": 0.801561536065542,
+      "step": 1370
+    },
+    {
+      "epoch": 0.5086940436551979,
+      "grad_norm": 1.7421875,
+      "learning_rate": 1.1441744040073469e-05,
+      "loss": 0.6396,
+      "mean_token_accuracy": 0.7960951091428123,
+      "step": 1375
+    },
+    {
+      "epoch": 0.5105438401775805,
+      "grad_norm": 1.4375,
+      "learning_rate": 1.1377800550643246e-05,
+      "loss": 0.6147,
+      "mean_token_accuracy": 0.802917483145199,
+      "step": 1380
+    },
+    {
+      "epoch": 0.512393636699963,
+      "grad_norm": 1.4140625,
+      "learning_rate": 1.1313799583767693e-05,
+      "loss": 0.6118,
+      "mean_token_accuracy": 0.8048380871561298,
+      "step": 1385
+    },
+    {
+      "epoch": 0.5142434332223456,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.124974380936305e-05,
+      "loss": 0.6182,
+      "mean_token_accuracy": 0.800282762150131,
+      "step": 1390
+    },
+    {
+      "epoch": 0.516093229744728,
+      "grad_norm": 1.4375,
+      "learning_rate": 1.1185635899631963e-05,
+      "loss": 0.6271,
+      "mean_token_accuracy": 0.8007738808738623,
+      "step": 1395
+    },
+    {
+      "epoch": 0.5179430262671106,
+      "grad_norm": 1.5546875,
+      "learning_rate": 1.1121478528951994e-05,
+      "loss": 0.61,
+      "mean_token_accuracy": 0.8071979776642234,
+      "step": 1400
+    },
+    {
+      "epoch": 0.5179430262671106,
+      "eval_loss": 0.6404861807823181,
+      "eval_mean_token_accuracy": 0.7941660008351938,
+      "eval_runtime": 89.2541,
+      "eval_samples_per_second": 1.445,
+      "eval_steps_per_second": 0.37,
+      "step": 1400
+    },
+    {
+      "epoch": 0.5197928227894931,
+      "grad_norm": 1.4765625,
+      "learning_rate": 1.1057274373764056e-05,
+      "loss": 0.6157,
+      "mean_token_accuracy": 0.8036740724033988,
+      "step": 1405
+    },
+    {
+      "epoch": 0.5216426193118757,
+      "grad_norm": 1.4453125,
+      "learning_rate": 1.0993026112460765e-05,
+      "loss": 0.616,
+      "mean_token_accuracy": 0.8046416038924604,
+      "step": 1410
+    },
+    {
+      "epoch": 0.5234924158342582,
+      "grad_norm": 1.4921875,
+      "learning_rate": 1.0928736425274702e-05,
+      "loss": 0.6139,
+      "mean_token_accuracy": 0.8030954723612285,
+      "step": 1415
+    },
+    {
+      "epoch": 0.5253422123566408,
+      "grad_norm": 1.4609375,
+      "learning_rate": 1.0864407994166609e-05,
+      "loss": 0.6197,
+      "mean_token_accuracy": 0.8030195823876982,
+      "step": 1420
+    },
+    {
+      "epoch": 0.5271920088790233,
+      "grad_norm": 1.296875,
+      "learning_rate": 1.0800043502713486e-05,
+      "loss": 0.6125,
+      "mean_token_accuracy": 0.8048843499557343,
+      "step": 1425
+    },
+    {
+      "epoch": 0.5290418054014059,
+      "grad_norm": 1.484375,
+      "learning_rate": 1.0735645635996676e-05,
+      "loss": 0.6389,
+      "mean_token_accuracy": 0.7974726569070282,
+      "step": 1430
+    },
+    {
+      "epoch": 0.5308916019237884,
+      "grad_norm": 1.5,
+      "learning_rate": 1.0671217080489816e-05,
+      "loss": 0.5986,
+      "mean_token_accuracy": 0.8085131258324557,
+      "step": 1435
+    },
+    {
+      "epoch": 0.532741398446171,
+      "grad_norm": 1.515625,
+      "learning_rate": 1.0606760523946789e-05,
+      "loss": 0.5726,
+      "mean_token_accuracy": 0.8163567403024615,
+      "step": 1440
+    },
+    {
+      "epoch": 0.5345911949685535,
+      "grad_norm": 1.375,
+      "learning_rate": 1.0542278655289588e-05,
+      "loss": 0.6126,
+      "mean_token_accuracy": 0.8059156301580199,
+      "step": 1445
+    },
+    {
+      "epoch": 0.536440991490936,
+      "grad_norm": 1.3046875,
+      "learning_rate": 1.047777416449615e-05,
+      "loss": 0.6058,
+      "mean_token_accuracy": 0.8067527785709215,
+      "step": 1450
+    },
+    {
+      "epoch": 0.5382907880133185,
+      "grad_norm": 1.59375,
+      "learning_rate": 1.0413249742488132e-05,
+      "loss": 0.6156,
+      "mean_token_accuracy": 0.8038016252838409,
+      "step": 1455
+    },
+    {
+      "epoch": 0.540140584535701,
+      "grad_norm": 1.6953125,
+      "learning_rate": 1.034870808101866e-05,
+      "loss": 0.6185,
+      "mean_token_accuracy": 0.8035043102600179,
+      "step": 1460
+    },
+    {
+      "epoch": 0.5419903810580836,
+      "grad_norm": 1.6484375,
+      "learning_rate": 1.0284151872560042e-05,
+      "loss": 0.6192,
+      "mean_token_accuracy": 0.8030750328145203,
+      "step": 1465
+    },
+    {
+      "epoch": 0.5438401775804661,
+      "grad_norm": 1.3125,
+      "learning_rate": 1.0219583810191422e-05,
+      "loss": 0.6049,
+      "mean_token_accuracy": 0.8064909436834569,
+      "step": 1470
+    },
+    {
+      "epoch": 0.5456899741028487,
+      "grad_norm": 1.3984375,
+      "learning_rate": 1.0155006587486468e-05,
+      "loss": 0.653,
+      "mean_token_accuracy": 0.7923704029235024,
+      "step": 1475
+    },
+    {
+      "epoch": 0.5475397706252312,
+      "grad_norm": 1.390625,
+      "learning_rate": 1.0090422898400985e-05,
+      "loss": 0.6046,
+      "mean_token_accuracy": 0.8049771548641909,
+      "step": 1480
+    },
+    {
+      "epoch": 0.5493895671476138,
+      "grad_norm": 1.4921875,
+      "learning_rate": 1.0025835437160523e-05,
+      "loss": 0.6313,
+      "mean_token_accuracy": 0.7971736355626806,
+      "step": 1485
+    },
+    {
+      "epoch": 0.5512393636699963,
+      "grad_norm": 1.4765625,
+      "learning_rate": 9.96124689814801e-06,
+      "loss": 0.6459,
+      "mean_token_accuracy": 0.7941516033793646,
+      "step": 1490
+    },
+    {
+      "epoch": 0.5530891601923789,
+      "grad_norm": 1.5859375,
+      "learning_rate": 9.89665997579133e-06,
+      "loss": 0.6259,
+      "mean_token_accuracy": 0.799138749944802,
+      "step": 1495
+    },
+    {
+      "epoch": 0.5549389567147613,
+      "grad_norm": 1.4296875,
+      "learning_rate": 9.832077364450915e-06,
+      "loss": 0.5586,
+      "mean_token_accuracy": 0.8199888914042646,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5549389567147613,
+      "eval_loss": 0.6393664479255676,
+      "eval_mean_token_accuracy": 0.7942906131232497,
+      "eval_runtime": 88.9262,
+      "eval_samples_per_second": 1.451,
+      "eval_steps_per_second": 0.371,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5567887532371439,
+      "grad_norm": 1.3203125,
+      "learning_rate": 9.767501758307376e-06,
+      "loss": 0.5924,
+      "mean_token_accuracy": 0.809022524594195,
+      "step": 1505
+    },
+    {
+      "epoch": 0.5586385497595264,
+      "grad_norm": 1.40625,
+      "learning_rate": 9.70293585124907e-06,
+      "loss": 0.5805,
+      "mean_token_accuracy": 0.8144158638567841,
+      "step": 1510
+    },
+    {
+      "epoch": 0.560488346281909,
+      "grad_norm": 1.5,
+      "learning_rate": 9.638382336759749e-06,
+      "loss": 0.6269,
+      "mean_token_accuracy": 0.8002326934751116,
+      "step": 1515
+    },
+    {
+      "epoch": 0.5623381428042915,
+      "grad_norm": 1.484375,
+      "learning_rate": 9.57384390780618e-06,
+      "loss": 0.608,
+      "mean_token_accuracy": 0.8048169125801135,
+      "step": 1520
+    },
+    {
+      "epoch": 0.5641879393266741,
+      "grad_norm": 1.4609375,
+      "learning_rate": 9.50932325672582e-06,
+      "loss": 0.6427,
+      "mean_token_accuracy": 0.794263757839041,
+      "step": 1525
+    },
+    {
+      "epoch": 0.5660377358490566,
+      "grad_norm": 1.4765625,
+      "learning_rate": 9.444823075114481e-06,
+      "loss": 0.5998,
+      "mean_token_accuracy": 0.8057361772156375,
+      "step": 1530
+    },
+    {
+      "epoch": 0.5678875323714392,
+      "grad_norm": 1.3984375,
+      "learning_rate": 9.380346053714055e-06,
+      "loss": 0.5727,
+      "mean_token_accuracy": 0.8141553185410195,
+      "step": 1535
+    },
+    {
+      "epoch": 0.5697373288938217,
+      "grad_norm": 1.453125,
+      "learning_rate": 9.315894882300269e-06,
+      "loss": 0.5535,
+      "mean_token_accuracy": 0.8231002426151157,
+      "step": 1540
+    },
+    {
+      "epoch": 0.5715871254162043,
+      "grad_norm": 1.4921875,
+      "learning_rate": 9.251472249570458e-06,
+      "loss": 0.6191,
+      "mean_token_accuracy": 0.8017774735014049,
+      "step": 1545
+    },
+    {
+      "epoch": 0.5734369219385868,
+      "grad_norm": 1.2734375,
+      "learning_rate": 9.187080843031436e-06,
+      "loss": 0.5935,
+      "mean_token_accuracy": 0.8102265850462294,
+      "step": 1550
+    },
+    {
+      "epoch": 0.5752867184609693,
+      "grad_norm": 1.59375,
+      "learning_rate": 9.12272334888734e-06,
+      "loss": 0.6363,
+      "mean_token_accuracy": 0.7959672921109016,
+      "step": 1555
+    },
+    {
+      "epoch": 0.5771365149833518,
+      "grad_norm": 1.2578125,
+      "learning_rate": 9.058402451927599e-06,
+      "loss": 0.5859,
+      "mean_token_accuracy": 0.81286955561369,
+      "step": 1560
+    },
+    {
+      "epoch": 0.5789863115057343,
+      "grad_norm": 1.5390625,
+      "learning_rate": 8.99412083541492e-06,
+      "loss": 0.5959,
+      "mean_token_accuracy": 0.8079413335296349,
+      "step": 1565
+    },
+    {
+      "epoch": 0.5808361080281169,
+      "grad_norm": 1.59375,
+      "learning_rate": 8.929881180973365e-06,
+      "loss": 0.6315,
+      "mean_token_accuracy": 0.7980688956447011,
+      "step": 1570
+    },
+    {
+      "epoch": 0.5826859045504994,
+      "grad_norm": 1.3671875,
+      "learning_rate": 8.865686168476458e-06,
+      "loss": 0.6153,
+      "mean_token_accuracy": 0.8025974547962257,
+      "step": 1575
+    },
+    {
+      "epoch": 0.584535701072882,
+      "grad_norm": 1.4453125,
+      "learning_rate": 8.801538475935413e-06,
+      "loss": 0.5894,
+      "mean_token_accuracy": 0.8110432721729401,
+      "step": 1580
+    },
+    {
+      "epoch": 0.5863854975952645,
+      "grad_norm": 1.375,
+      "learning_rate": 8.737440779387398e-06,
+      "loss": 0.6186,
+      "mean_token_accuracy": 0.8030020456850069,
+      "step": 1585
+    },
+    {
+      "epoch": 0.5882352941176471,
+      "grad_norm": 1.296875,
+      "learning_rate": 8.673395752783924e-06,
+      "loss": 0.6229,
+      "mean_token_accuracy": 0.80131154927805,
+      "step": 1590
+    },
+    {
+      "epoch": 0.5900850906400296,
+      "grad_norm": 1.6171875,
+      "learning_rate": 8.609406067879258e-06,
+      "loss": 0.6007,
+      "mean_token_accuracy": 0.808220151846878,
+      "step": 1595
+    },
+    {
+      "epoch": 0.5919348871624122,
+      "grad_norm": 1.453125,
+      "learning_rate": 8.54547439411901e-06,
+      "loss": 0.6334,
+      "mean_token_accuracy": 0.7980854227698776,
+      "step": 1600
+    },
+    {
+      "epoch": 0.5919348871624122,
+      "eval_loss": 0.6382184028625488,
+      "eval_mean_token_accuracy": 0.7946820543622822,
+      "eval_runtime": 88.9607,
+      "eval_samples_per_second": 1.45,
+      "eval_steps_per_second": 0.371,
+      "step": 1600
+    },
+    {
+      "epoch": 0.5937846836847946,
+      "grad_norm": 1.5546875,
+      "learning_rate": 8.481603398528727e-06,
+      "loss": 0.5717,
+      "mean_token_accuracy": 0.8171099271491468,
+      "step": 1605
+    },
+    {
+      "epoch": 0.5956344802071772,
+      "grad_norm": 1.453125,
+      "learning_rate": 8.417795745602679e-06,
+      "loss": 0.6231,
+      "mean_token_accuracy": 0.8005498304105656,
+      "step": 1610
+    },
+    {
+      "epoch": 0.5974842767295597,
+      "grad_norm": 1.5390625,
+      "learning_rate": 8.35405409719266e-06,
+      "loss": 0.6315,
+      "mean_token_accuracy": 0.7979197498648158,
+      "step": 1615
+    },
+    {
+      "epoch": 0.5993340732519423,
+      "grad_norm": 1.453125,
+      "learning_rate": 8.290381112396989e-06,
+      "loss": 0.5916,
+      "mean_token_accuracy": 0.8103999616712713,
+      "step": 1620
+    },
+    {
+      "epoch": 0.6011838697743248,
+      "grad_norm": 1.546875,
+      "learning_rate": 8.226779447449538e-06,
+      "loss": 0.5843,
+      "mean_token_accuracy": 0.8117410911132362,
+      "step": 1625
+    },
+    {
+      "epoch": 0.6030336662967074,
+      "grad_norm": 1.4609375,
+      "learning_rate": 8.163251755608954e-06,
+      "loss": 0.6095,
+      "mean_token_accuracy": 0.805165328621117,
+      "step": 1630
+    },
+    {
+      "epoch": 0.6048834628190899,
+      "grad_norm": 1.46875,
+      "learning_rate": 8.099800687047958e-06,
+      "loss": 0.5999,
+      "mean_token_accuracy": 0.8065257637450328,
+      "step": 1635
+    },
+    {
+      "epoch": 0.6067332593414725,
+      "grad_norm": 1.5,
+      "learning_rate": 8.036428888742789e-06,
+      "loss": 0.6052,
+      "mean_token_accuracy": 0.806758925851564,
+      "step": 1640
+    },
+    {
+      "epoch": 0.608583055863855,
+      "grad_norm": 1.5,
+      "learning_rate": 7.973139004362794e-06,
+      "loss": 0.6136,
+      "mean_token_accuracy": 0.8031655997888818,
+      "step": 1645
+    },
+    {
+      "epoch": 0.6104328523862376,
+      "grad_norm": 1.359375,
+      "learning_rate": 7.909933674160118e-06,
+      "loss": 0.633,
+      "mean_token_accuracy": 0.7979598173335406,
+      "step": 1650
+    },
+    {
+      "epoch": 0.61228264890862,
+      "grad_norm": 1.4296875,
+      "learning_rate": 7.846815534859592e-06,
+      "loss": 0.6303,
+      "mean_token_accuracy": 0.8010427396672002,
+      "step": 1655
+    },
+    {
+      "epoch": 0.6141324454310025,
+      "grad_norm": 1.40625,
+      "learning_rate": 7.7837872195487e-06,
+      "loss": 0.6117,
+      "mean_token_accuracy": 0.8041381692767452,
+      "step": 1660
+    },
+    {
+      "epoch": 0.6159822419533851,
+      "grad_norm": 1.375,
+      "learning_rate": 7.720851357567778e-06,
+      "loss": 0.5912,
+      "mean_token_accuracy": 0.810616065615282,
+      "step": 1665
+    },
+    {
+      "epoch": 0.6178320384757676,
+      "grad_norm": 1.5625,
+      "learning_rate": 7.658010574400285e-06,
+      "loss": 0.597,
+      "mean_token_accuracy": 0.8085118170263923,
+      "step": 1670
+    },
+    {
+      "epoch": 0.6196818349981502,
+      "grad_norm": 1.6484375,
+      "learning_rate": 7.595267491563311e-06,
+      "loss": 0.5951,
+      "mean_token_accuracy": 0.8085559229632636,
+      "step": 1675
+    },
+    {
+      "epoch": 0.6215316315205327,
+      "grad_norm": 1.5234375,
+      "learning_rate": 7.532624726498181e-06,
+      "loss": 0.6093,
+      "mean_token_accuracy": 0.8033529951931451,
+      "step": 1680
+    },
+    {
+      "epoch": 0.6233814280429153,
+      "grad_norm": 1.3828125,
+      "learning_rate": 7.470084892461305e-06,
+      "loss": 0.6293,
+      "mean_token_accuracy": 0.7978544418682161,
+      "step": 1685
+    },
+    {
+      "epoch": 0.6252312245652978,
+      "grad_norm": 1.3046875,
+      "learning_rate": 7.407650598415119e-06,
+      "loss": 0.6149,
+      "mean_token_accuracy": 0.8036706926329463,
+      "step": 1690
+    },
+    {
+      "epoch": 0.6270810210876804,
+      "grad_norm": 1.4375,
+      "learning_rate": 7.34532444891928e-06,
+      "loss": 0.6048,
+      "mean_token_accuracy": 0.806633769978057,
+      "step": 1695
+    },
+    {
+      "epoch": 0.6289308176100629,
+      "grad_norm": 1.4921875,
+      "learning_rate": 7.283109044021987e-06,
+      "loss": 0.6784,
+      "mean_token_accuracy": 0.7859875479641154,
+      "step": 1700
+    },
+    {
+      "epoch": 0.6289308176100629,
+      "eval_loss": 0.6376616954803467,
+      "eval_mean_token_accuracy": 0.7946947289578048,
+      "eval_runtime": 88.9729,
+      "eval_samples_per_second": 1.45,
+      "eval_steps_per_second": 0.371,
+      "step": 1700
+    },
+    {
+      "epoch": 0.6307806141324455,
+      "grad_norm": 1.5859375,
+      "learning_rate": 7.221006979151546e-06,
+      "loss": 0.6129,
+      "mean_token_accuracy": 0.8034805736848327,
+      "step": 1705
+    },
+    {
+      "epoch": 0.632630410654828,
+      "grad_norm": 1.5859375,
+      "learning_rate": 7.159020845008056e-06,
+      "loss": 0.6799,
+      "mean_token_accuracy": 0.7839307631204628,
+      "step": 1710
+    },
+    {
+      "epoch": 0.6344802071772105,
+      "grad_norm": 1.4375,
+      "learning_rate": 7.097153227455379e-06,
+      "loss": 0.609,
+      "mean_token_accuracy": 0.8053893334213564,
+      "step": 1715
+    },
+    {
+      "epoch": 0.636330003699593,
+      "grad_norm": 1.4765625,
+      "learning_rate": 7.035406707413221e-06,
+      "loss": 0.643,
+      "mean_token_accuracy": 0.7946525856421311,
+      "step": 1720
+    },
+    {
+      "epoch": 0.6381798002219756,
+      "grad_norm": 1.328125,
+      "learning_rate": 6.973783860749499e-06,
+      "loss": 0.6162,
+      "mean_token_accuracy": 0.8026447440977214,
+      "step": 1725
+    },
+    {
+      "epoch": 0.6400295967443581,
+      "grad_norm": 1.1171875,
+      "learning_rate": 6.912287258172869e-06,
+      "loss": 0.5537,
+      "mean_token_accuracy": 0.8240663879187713,
+      "step": 1730
+    },
+    {
+      "epoch": 0.6418793932667407,
+      "grad_norm": 1.34375,
+      "learning_rate": 6.8509194651254825e-06,
+      "loss": 0.6422,
+      "mean_token_accuracy": 0.7961569176861726,
+      "step": 1735
+    },
+    {
+      "epoch": 0.6437291897891232,
+      "grad_norm": 1.4609375,
+      "learning_rate": 6.789683041675977e-06,
+      "loss": 0.6294,
+      "mean_token_accuracy": 0.7987188532981321,
+      "step": 1740
+    },
+    {
+      "epoch": 0.6455789863115058,
+      "grad_norm": 1.3828125,
+      "learning_rate": 6.7285805424126585e-06,
+      "loss": 0.6318,
+      "mean_token_accuracy": 0.7974160403423387,
+      "step": 1745
+    },
+    {
+      "epoch": 0.6474287828338883,
+      "grad_norm": 1.4609375,
+      "learning_rate": 6.667614516336949e-06,
+      "loss": 0.6109,
+      "mean_token_accuracy": 0.8049147387865885,
+      "step": 1750
+    },
+    {
+      "epoch": 0.6492785793562708,
+      "grad_norm": 1.4375,
+      "learning_rate": 6.60678750675704e-06,
+      "loss": 0.6417,
+      "mean_token_accuracy": 0.7959335594022869,
+      "step": 1755
+    },
+    {
+      "epoch": 0.6511283758786534,
+      "grad_norm": 1.3515625,
+      "learning_rate": 6.546102051181808e-06,
+      "loss": 0.5734,
+      "mean_token_accuracy": 0.8164705619617376,
+      "step": 1760
+    },
+    {
+      "epoch": 0.6529781724010358,
+      "grad_norm": 1.46875,
+      "learning_rate": 6.485560681214933e-06,
+      "loss": 0.6234,
+      "mean_token_accuracy": 0.80105609720617,
+      "step": 1765
+    },
+    {
+      "epoch": 0.6548279689234184,
+      "grad_norm": 1.515625,
+      "learning_rate": 6.425165922449319e-06,
+      "loss": 0.644,
+      "mean_token_accuracy": 0.7950143347988867,
+      "step": 1770
+    },
+    {
+      "epoch": 0.6566777654458009,
+      "grad_norm": 1.3359375,
+      "learning_rate": 6.364920294361701e-06,
+      "loss": 0.6182,
+      "mean_token_accuracy": 0.8034045872177484,
+      "step": 1775
+    },
+    {
+      "epoch": 0.6585275619681835,
+      "grad_norm": 1.4609375,
+      "learning_rate": 6.3048263102075766e-06,
+      "loss": 0.6196,
+      "mean_token_accuracy": 0.8006927261393922,
+      "step": 1780
+    },
+    {
+      "epoch": 0.660377358490566,
+      "grad_norm": 1.3125,
+      "learning_rate": 6.244886476916325e-06,
+      "loss": 0.626,
+      "mean_token_accuracy": 0.8000366724494992,
+      "step": 1785
+    },
+    {
+      "epoch": 0.6622271550129486,
+      "grad_norm": 1.59375,
+      "learning_rate": 6.185103294986654e-06,
+      "loss": 0.6161,
+      "mean_token_accuracy": 0.8034170113383883,
+      "step": 1790
+    },
+    {
+      "epoch": 0.6640769515353311,
+      "grad_norm": 1.34375,
+      "learning_rate": 6.125479258382268e-06,
+      "loss": 0.6121,
+      "mean_token_accuracy": 0.8030250023686586,
+      "step": 1795
+    },
+    {
+      "epoch": 0.6659267480577137,
+      "grad_norm": 1.34375,
+      "learning_rate": 6.066016854427846e-06,
+      "loss": 0.5984,
+      "mean_token_accuracy": 0.8069742090958034,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6659267480577137,
+      "eval_loss": 0.6369358897209167,
+      "eval_mean_token_accuracy": 0.7949392639910889,
+      "eval_runtime": 88.9414,
+      "eval_samples_per_second": 1.45,
+      "eval_steps_per_second": 0.371,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6677765445800962,
+      "grad_norm": 1.578125,
+      "learning_rate": 6.006718563705258e-06,
+      "loss": 0.6417,
+      "mean_token_accuracy": 0.7950800883450222,
+      "step": 1805
+    },
+    {
+      "epoch": 0.6696263411024788,
+      "grad_norm": 1.328125,
+      "learning_rate": 5.947586859950103e-06,
+      "loss": 0.593,
+      "mean_token_accuracy": 0.8103679153162121,
+      "step": 1810
+    },
+    {
+      "epoch": 0.6714761376248612,
+      "grad_norm": 1.4453125,
+      "learning_rate": 5.888624209948495e-06,
+      "loss": 0.6088,
+      "mean_token_accuracy": 0.8045350157569725,
+      "step": 1815
+    },
+    {
+      "epoch": 0.6733259341472438,
+      "grad_norm": 1.40625,
+      "learning_rate": 5.82983307343416e-06,
+      "loss": 0.584,
+      "mean_token_accuracy": 0.812762890116858,
+      "step": 1820
+    },
+    {
+      "epoch": 0.6751757306696263,
+      "grad_norm": 1.3046875,
+      "learning_rate": 5.771215902985848e-06,
+      "loss": 0.5973,
+      "mean_token_accuracy": 0.8087440935559368,
+      "step": 1825
+    },
+    {
+      "epoch": 0.6770255271920089,
+      "grad_norm": 1.34375,
+      "learning_rate": 5.712775143924978e-06,
+      "loss": 0.5846,
+      "mean_token_accuracy": 0.8123533133451921,
+      "step": 1830
+    },
+    {
+      "epoch": 0.6788753237143914,
+      "grad_norm": 1.34375,
+      "learning_rate": 5.654513234213663e-06,
+      "loss": 0.6087,
+      "mean_token_accuracy": 0.8060109335265715,
+      "step": 1835
+    },
+    {
+      "epoch": 0.680725120236774,
+      "grad_norm": 1.3828125,
+      "learning_rate": 5.596432604352982e-06,
+      "loss": 0.5946,
+      "mean_token_accuracy": 0.8110177555862446,
+      "step": 1840
+    },
+    {
+      "epoch": 0.6825749167591565,
+      "grad_norm": 1.3046875,
+      "learning_rate": 5.538535677281608e-06,
+      "loss": 0.6212,
+      "mean_token_accuracy": 0.8015808232793761,
+      "step": 1845
+    },
+    {
+      "epoch": 0.684424713281539,
+      "grad_norm": 1.421875,
+      "learning_rate": 5.480824868274699e-06,
+      "loss": 0.6219,
+      "mean_token_accuracy": 0.8020745765107336,
+      "step": 1850
+    },
+    {
+      "epoch": 0.6862745098039216,
+      "grad_norm": 1.515625,
+      "learning_rate": 5.423302584843186e-06,
+      "loss": 0.6365,
+      "mean_token_accuracy": 0.7970287487813332,
+      "step": 1855
+    },
+    {
+      "epoch": 0.6881243063263041,
+      "grad_norm": 1.328125,
+      "learning_rate": 5.365971226633294e-06,
+      "loss": 0.5874,
+      "mean_token_accuracy": 0.8129541430751936,
+      "step": 1860
+    },
+    {
+      "epoch": 0.6899741028486867,
+      "grad_norm": 1.6015625,
+      "learning_rate": 5.308833185326472e-06,
+      "loss": 0.6182,
+      "mean_token_accuracy": 0.8022356779807035,
+      "step": 1865
+    },
+    {
+      "epoch": 0.6918238993710691,
+      "grad_norm": 1.515625,
+      "learning_rate": 5.25189084453959e-06,
+      "loss": 0.6118,
+      "mean_token_accuracy": 0.8028822441957664,
+      "step": 1870
+    },
+    {
+      "epoch": 0.6936736958934517,
+      "grad_norm": 1.390625,
+      "learning_rate": 5.19514657972553e-06,
+      "loss": 0.5685,
+      "mean_token_accuracy": 0.8190691074997625,
+      "step": 1875
+    },
+    {
+      "epoch": 0.6955234924158342,
+      "grad_norm": 1.3828125,
+      "learning_rate": 5.138602758074062e-06,
+      "loss": 0.569,
+      "mean_token_accuracy": 0.8185277743801059,
+      "step": 1880
+    },
+    {
+      "epoch": 0.6973732889382168,
+      "grad_norm": 1.6015625,
+      "learning_rate": 5.082261738413124e-06,
+      "loss": 0.647,
+      "mean_token_accuracy": 0.7909912780879745,
+      "step": 1885
+    },
+    {
+      "epoch": 0.6992230854605993,
+      "grad_norm": 1.3984375,
+      "learning_rate": 5.026125871110383e-06,
+      "loss": 0.6129,
+      "mean_token_accuracy": 0.8034293316832659,
+      "step": 1890
+    },
+    {
+      "epoch": 0.7010728819829819,
+      "grad_norm": 1.2890625,
+      "learning_rate": 4.970197497975216e-06,
+      "loss": 0.5871,
+      "mean_token_accuracy": 0.8136575444275269,
+      "step": 1895
+    },
+    {
+      "epoch": 0.7029226785053644,
+      "grad_norm": 1.3515625,
+      "learning_rate": 4.914478952160997e-06,
+      "loss": 0.6327,
+      "mean_token_accuracy": 0.7991577701812028,
+      "step": 1900
+    },
+    {
+      "epoch": 0.7029226785053644,
+      "eval_loss": 0.636552095413208,
+      "eval_mean_token_accuracy": 0.795276787314407,
+      "eval_runtime": 88.8952,
+      "eval_samples_per_second": 1.451,
+      "eval_steps_per_second": 0.371,
+      "step": 1900
+    },
+    {
+      "epoch": 0.704772475027747,
+      "grad_norm": 1.328125,
+      "learning_rate": 4.858972558067784e-06,
+      "loss": 0.5594,
+      "mean_token_accuracy": 0.8194452736455163,
+      "step": 1905
+    },
+    {
+      "epoch": 0.7066222715501295,
+      "grad_norm": 1.3359375,
+      "learning_rate": 4.803680631245333e-06,
+      "loss": 0.5789,
+      "mean_token_accuracy": 0.8144605539377286,
+      "step": 1910
+    },
+    {
+      "epoch": 0.7084720680725121,
+      "grad_norm": 1.1953125,
+      "learning_rate": 4.748605478296508e-06,
+      "loss": 0.5789,
+      "mean_token_accuracy": 0.8149056270159057,
+      "step": 1915
+    },
+    {
+      "epoch": 0.7103218645948945,
+      "grad_norm": 1.375,
+      "learning_rate": 4.6937493967810665e-06,
+      "loss": 0.5981,
+      "mean_token_accuracy": 0.8081419831484649,
+      "step": 1920
+    },
+    {
+      "epoch": 0.7121716611172771,
+      "grad_norm": 1.5546875,
+      "learning_rate": 4.639114675119797e-06,
+      "loss": 0.6364,
+      "mean_token_accuracy": 0.7951362587281494,
+      "step": 1925
+    },
+    {
+      "epoch": 0.7140214576396596,
+      "grad_norm": 1.453125,
+      "learning_rate": 4.584703592499068e-06,
+      "loss": 0.6156,
+      "mean_token_accuracy": 0.8046571229298831,
+      "step": 1930
+    },
+    {
+      "epoch": 0.7158712541620422,
+      "grad_norm": 1.390625,
+      "learning_rate": 4.530518418775734e-06,
+      "loss": 0.6079,
+      "mean_token_accuracy": 0.8058350059417514,
+      "step": 1935
+    },
+    {
+      "epoch": 0.7177210506844247,
+      "grad_norm": 1.28125,
+      "learning_rate": 4.476561414382451e-06,
+      "loss": 0.5565,
+      "mean_token_accuracy": 0.8217036726709779,
+      "step": 1940
+    },
+    {
+      "epoch": 0.7195708472068072,
+      "grad_norm": 1.3203125,
+      "learning_rate": 4.422834830233378e-06,
+      "loss": 0.6015,
+      "mean_token_accuracy": 0.8074042909270748,
+      "step": 1945
+    },
+    {
+      "epoch": 0.7214206437291898,
+      "grad_norm": 2.078125,
+      "learning_rate": 4.369340907630284e-06,
+      "loss": 0.642,
+      "mean_token_accuracy": 0.7949644403505403,
+      "step": 1950
+    },
+    {
+      "epoch": 0.7232704402515723,
+      "grad_norm": 1.453125,
+      "learning_rate": 4.3160818781690286e-06,
+      "loss": 0.6213,
+      "mean_token_accuracy": 0.8008243179695986,
+      "step": 1955
+    },
+    {
+      "epoch": 0.7251202367739549,
+      "grad_norm": 1.4453125,
+      "learning_rate": 4.26305996364649e-06,
+      "loss": 0.5994,
+      "mean_token_accuracy": 0.808540953538001,
+      "step": 1960
+    },
+    {
+      "epoch": 0.7269700332963374,
+      "grad_norm": 1.3984375,
+      "learning_rate": 4.210277375967855e-06,
+      "loss": 0.6174,
+      "mean_token_accuracy": 0.8023914874074023,
+      "step": 1965
+    },
+    {
+      "epoch": 0.72881982981872,
+      "grad_norm": 1.515625,
+      "learning_rate": 4.157736317054373e-06,
+      "loss": 0.6215,
+      "mean_token_accuracy": 0.8030253071351773,
+      "step": 1970
+    },
+    {
+      "epoch": 0.7306696263411024,
+      "grad_norm": 1.265625,
+      "learning_rate": 4.105438978751465e-06,
+      "loss": 0.5968,
+      "mean_token_accuracy": 0.8091518257150462,
+      "step": 1975
+    },
+    {
+      "epoch": 0.732519422863485,
+      "grad_norm": 1.4453125,
+      "learning_rate": 4.053387542737323e-06,
+      "loss": 0.6012,
+      "mean_token_accuracy": 0.808833440877461,
+      "step": 1980
+    },
+    {
+      "epoch": 0.7343692193858675,
+      "grad_norm": 1.5,
+      "learning_rate": 4.001584180431869e-06,
+      "loss": 0.5776,
+      "mean_token_accuracy": 0.8133157536554938,
+      "step": 1985
+    },
+    {
+      "epoch": 0.7362190159082501,
+      "grad_norm": 1.3828125,
+      "learning_rate": 3.950031052906183e-06,
+      "loss": 0.6029,
+      "mean_token_accuracy": 0.8072724870791322,
+      "step": 1990
+    },
+    {
+      "epoch": 0.7380688124306326,
+      "grad_norm": 1.3984375,
+      "learning_rate": 3.898730310792346e-06,
+      "loss": 0.6173,
+      "mean_token_accuracy": 0.8008440851413285,
+      "step": 1995
+    },
+    {
+      "epoch": 0.7399186089530152,
+      "grad_norm": 1.546875,
+      "learning_rate": 3.847684094193733e-06,
+      "loss": 0.6358,
+      "mean_token_accuracy": 0.7966047360693018,
+      "step": 2000
+    },
+    {
+      "epoch": 0.7399186089530152,
+      "eval_loss": 0.6362702250480652,
+      "eval_mean_token_accuracy": 0.7951077274225514,
+      "eval_runtime": 88.9029,
+      "eval_samples_per_second": 1.451,
+      "eval_steps_per_second": 0.371,
+      "step": 2000
+    },
+    {
+      "epoch": 0.7417684054753977,
+      "grad_norm": 1.359375,
+      "learning_rate": 3.7968945325957175e-06,
+      "loss": 0.6294,
+      "mean_token_accuracy": 0.7991666985360435,
+      "step": 2005
+    },
+    {
+      "epoch": 0.7436182019977803,
+      "grad_norm": 1.4609375,
+      "learning_rate": 3.746363744776844e-06,
+      "loss": 0.6314,
+      "mean_token_accuracy": 0.7993224783148821,
+      "step": 2010
+    },
+    {
+      "epoch": 0.7454679985201628,
+      "grad_norm": 1.34375,
+      "learning_rate": 3.69609383872045e-06,
+      "loss": 0.6065,
+      "mean_token_accuracy": 0.8058827700178032,
+      "step": 2015
+    },
+    {
+      "epoch": 0.7473177950425454,
+      "grad_norm": 1.375,
+      "learning_rate": 3.646086911526706e-06,
+      "loss": 0.5904,
+      "mean_token_accuracy": 0.8122150311847595,
+      "step": 2020
+    },
+    {
+      "epoch": 0.7491675915649278,
+      "grad_norm": 1.4453125,
+      "learning_rate": 3.5963450493251552e-06,
+      "loss": 0.628,
+      "mean_token_accuracy": 0.7990032442481549,
+      "step": 2025
+    },
+    {
+      "epoch": 0.7510173880873104,
+      "grad_norm": 1.421875,
+      "learning_rate": 3.5468703271876626e-06,
+      "loss": 0.5834,
+      "mean_token_accuracy": 0.8131149673012814,
+      "step": 2030
+    },
+    {
+      "epoch": 0.7528671846096929,
+      "grad_norm": 1.5234375,
+      "learning_rate": 3.4976648090418685e-06,
+      "loss": 0.609,
+      "mean_token_accuracy": 0.8072119662864601,
+      "step": 2035
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "grad_norm": 1.5,
+      "learning_rate": 3.4487305475850754e-06,
+      "loss": 0.6217,
+      "mean_token_accuracy": 0.8004756937561506,
+      "step": 2040
+    },
+    {
+      "epoch": 0.756566777654458,
+      "grad_norm": 1.296875,
+      "learning_rate": 3.400069584198633e-06,
+      "loss": 0.6092,
+      "mean_token_accuracy": 0.8051174523969028,
+      "step": 2045
+    },
+    {
+      "epoch": 0.7584165741768405,
+      "grad_norm": 1.421875,
+      "learning_rate": 3.351683948862753e-06,
+      "loss": 0.6299,
+      "mean_token_accuracy": 0.7972799460855309,
+      "step": 2050
+    },
+    {
+      "epoch": 0.7602663706992231,
+      "grad_norm": 1.375,
+      "learning_rate": 3.3035756600718515e-06,
+      "loss": 0.5962,
+      "mean_token_accuracy": 0.8083237856847505,
+      "step": 2055
+    },
+    {
+      "epoch": 0.7621161672216056,
+      "grad_norm": 1.4765625,
+      "learning_rate": 3.2557467247503205e-06,
+      "loss": 0.6123,
+      "mean_token_accuracy": 0.8035482977649073,
+      "step": 2060
+    },
+    {
+      "epoch": 0.7639659637439882,
+      "grad_norm": 1.375,
+      "learning_rate": 3.208199138168826e-06,
+      "loss": 0.6559,
+      "mean_token_accuracy": 0.7912612149917433,
+      "step": 2065
+    },
+    {
+      "epoch": 0.7658157602663707,
+      "grad_norm": 1.3203125,
+      "learning_rate": 3.160934883861053e-06,
+      "loss": 0.6261,
+      "mean_token_accuracy": 0.7983330563919967,
+      "step": 2070
+    },
+    {
+      "epoch": 0.7676655567887533,
+      "grad_norm": 1.3671875,
+      "learning_rate": 3.113955933540973e-06,
+      "loss": 0.5495,
+      "mean_token_accuracy": 0.8235621851919163,
+      "step": 2075
+    },
+    {
+      "epoch": 0.7695153533111357,
+      "grad_norm": 1.25,
+      "learning_rate": 3.0672642470205794e-06,
+      "loss": 0.6103,
+      "mean_token_accuracy": 0.8063489001242946,
+      "step": 2080
+    },
+    {
+      "epoch": 0.7713651498335183,
+      "grad_norm": 1.5546875,
+      "learning_rate": 3.020861772128145e-06,
+      "loss": 0.6577,
+      "mean_token_accuracy": 0.79064951743628,
+      "step": 2085
+    },
+    {
+      "epoch": 0.7732149463559008,
+      "grad_norm": 1.40625,
+      "learning_rate": 2.9747504446269447e-06,
+      "loss": 0.592,
+      "mean_token_accuracy": 0.8099496706195728,
+      "step": 2090
+    },
+    {
+      "epoch": 0.7750647428782834,
+      "grad_norm": 1.2734375,
+      "learning_rate": 2.9289321881345257e-06,
+      "loss": 0.6239,
+      "mean_token_accuracy": 0.7997651777082817,
+      "step": 2095
+    },
+    {
+      "epoch": 0.7769145394006659,
+      "grad_norm": 1.5625,
+      "learning_rate": 2.883408914042435e-06,
+      "loss": 0.6481,
+      "mean_token_accuracy": 0.7924718094783091,
+      "step": 2100
+    },
+    {
+      "epoch": 0.7769145394006659,
+      "eval_loss": 0.6361722946166992,
+      "eval_mean_token_accuracy": 0.7953076234449475,
+      "eval_runtime": 89.6084,
+      "eval_samples_per_second": 1.44,
+      "eval_steps_per_second": 0.368,
+      "step": 2100
+    },
+    {
+      "epoch": 0.7787643359230485,
+      "grad_norm": 1.4296875,
+      "learning_rate": 2.838182521436498e-06,
+      "loss": 0.6108,
+      "mean_token_accuracy": 0.8046374751525172,
+      "step": 2105
+    },
+    {
+      "epoch": 0.780614132445431,
+      "grad_norm": 1.53125,
+      "learning_rate": 2.7932548970175977e-06,
+      "loss": 0.644,
+      "mean_token_accuracy": 0.7951850078158065,
+      "step": 2110
+    },
+    {
+      "epoch": 0.7824639289678136,
+      "grad_norm": 1.3515625,
+      "learning_rate": 2.74862791502295e-06,
+      "loss": 0.6025,
+      "mean_token_accuracy": 0.8073302167098906,
+      "step": 2115
+    },
+    {
+      "epoch": 0.7843137254901961,
+      "grad_norm": 1.5078125,
+      "learning_rate": 2.7043034371479404e-06,
+      "loss": 0.6334,
+      "mean_token_accuracy": 0.7977907860989295,
+      "step": 2120
+    },
+    {
+      "epoch": 0.7861635220125787,
+      "grad_norm": 1.3515625,
+      "learning_rate": 2.660283312468438e-06,
+      "loss": 0.5621,
+      "mean_token_accuracy": 0.8203150652015333,
+      "step": 2125
+    },
+    {
+      "epoch": 0.7880133185349611,
+      "grad_norm": 1.3515625,
+      "learning_rate": 2.616569377363668e-06,
+      "loss": 0.6576,
+      "mean_token_accuracy": 0.7932708833063655,
+      "step": 2130
+    },
+    {
+      "epoch": 0.7898631150573437,
+      "grad_norm": 1.4375,
+      "learning_rate": 2.573163455439601e-06,
+      "loss": 0.6216,
+      "mean_token_accuracy": 0.8013081822930201,
+      "step": 2135
+    },
+    {
+      "epoch": 0.7917129115797262,
+      "grad_norm": 1.3828125,
+      "learning_rate": 2.530067357452892e-06,
+      "loss": 0.6123,
+      "mean_token_accuracy": 0.8031451747749095,
+      "step": 2140
+    },
+    {
+      "epoch": 0.7935627081021087,
+      "grad_norm": 1.46875,
+      "learning_rate": 2.4872828812353146e-06,
+      "loss": 0.5895,
+      "mean_token_accuracy": 0.8112052261886082,
+      "step": 2145
+    },
+    {
+      "epoch": 0.7954125046244913,
+      "grad_norm": 1.46875,
+      "learning_rate": 2.4448118116187847e-06,
+      "loss": 0.6399,
+      "mean_token_accuracy": 0.7946481315700542,
+      "step": 2150
+    },
+    {
+      "epoch": 0.7972623011468738,
+      "grad_norm": 1.5625,
+      "learning_rate": 2.402655920360889e-06,
+      "loss": 0.6106,
+      "mean_token_accuracy": 0.8067639166281928,
+      "step": 2155
+    },
+    {
+      "epoch": 0.7991120976692564,
+      "grad_norm": 1.4140625,
+      "learning_rate": 2.360816966070986e-06,
+      "loss": 0.5999,
+      "mean_token_accuracy": 0.8078720447074721,
+      "step": 2160
+    },
+    {
+      "epoch": 0.8009618941916389,
+      "grad_norm": 1.46875,
+      "learning_rate": 2.3192966941368247e-06,
+      "loss": 0.6241,
+      "mean_token_accuracy": 0.8013911659381087,
+      "step": 2165
+    },
+    {
+      "epoch": 0.8028116907140215,
+      "grad_norm": 1.3359375,
+      "learning_rate": 2.2780968366517476e-06,
+      "loss": 0.5688,
+      "mean_token_accuracy": 0.8170701631668347,
+      "step": 2170
+    },
+    {
+      "epoch": 0.804661487236404,
+      "grad_norm": 1.3359375,
+      "learning_rate": 2.237219112342426e-06,
+      "loss": 0.5977,
+      "mean_token_accuracy": 0.8071282343184721,
+      "step": 2175
+    },
+    {
+      "epoch": 0.8065112837587866,
+      "grad_norm": 1.453125,
+      "learning_rate": 2.196665226497169e-06,
+      "loss": 0.6198,
+      "mean_token_accuracy": 0.8036371622973151,
+      "step": 2180
+    },
+    {
+      "epoch": 0.808361080281169,
+      "grad_norm": 1.546875,
+      "learning_rate": 2.156436870894767e-06,
+      "loss": 0.6379,
+      "mean_token_accuracy": 0.7951281715593586,
+      "step": 2185
+    },
+    {
+      "epoch": 0.8102108768035516,
+      "grad_norm": 1.453125,
+      "learning_rate": 2.116535723733938e-06,
+      "loss": 0.58,
+      "mean_token_accuracy": 0.8144863232340553,
+      "step": 2190
+    },
+    {
+      "epoch": 0.8120606733259341,
+      "grad_norm": 1.4609375,
+      "learning_rate": 2.0769634495632986e-06,
+      "loss": 0.6233,
+      "mean_token_accuracy": 0.803235626972365,
+      "step": 2195
+    },
+    {
+      "epoch": 0.8139104698483167,
+      "grad_norm": 1.59375,
+      "learning_rate": 2.037721699211933e-06,
+      "loss": 0.6366,
+      "mean_token_accuracy": 0.796360484743256,
+      "step": 2200
+    },
+    {
+      "epoch": 0.8139104698483167,
+      "eval_loss": 0.6360675692558289,
+      "eval_mean_token_accuracy": 0.7950688748028012,
+      "eval_runtime": 89.3325,
+      "eval_samples_per_second": 1.444,
+      "eval_steps_per_second": 0.369,
+      "step": 2200
+    },
+    {
+      "epoch": 0.8157602663706992,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.998812109720535e-06,
+      "loss": 0.6263,
+      "mean_token_accuracy": 0.8000316439105708,
+      "step": 2205
+    },
+    {
+      "epoch": 0.8176100628930818,
+      "grad_norm": 1.484375,
+      "learning_rate": 1.9602363042730964e-06,
+      "loss": 0.6445,
+      "mean_token_accuracy": 0.7932918662176973,
+      "step": 2210
+    },
+    {
+      "epoch": 0.8194598594154643,
+      "grad_norm": 1.40625,
+      "learning_rate": 1.921995892129208e-06,
+      "loss": 0.6118,
+      "mean_token_accuracy": 0.8050862290508254,
+      "step": 2215
+    },
+    {
+      "epoch": 0.8213096559378469,
+      "grad_norm": 1.4609375,
+      "learning_rate": 1.8840924685569162e-06,
+      "loss": 0.5915,
+      "mean_token_accuracy": 0.8109227818382395,
+      "step": 2220
+    },
+    {
+      "epoch": 0.8231594524602294,
+      "grad_norm": 1.546875,
+      "learning_rate": 1.8465276147661905e-06,
+      "loss": 0.5725,
+      "mean_token_accuracy": 0.8164197439357548,
+      "step": 2225
+    },
+    {
+      "epoch": 0.825009248982612,
+      "grad_norm": 1.5078125,
+      "learning_rate": 1.809302897842934e-06,
+      "loss": 0.628,
+      "mean_token_accuracy": 0.800985687010994,
+      "step": 2230
+    },
+    {
+      "epoch": 0.8268590455049944,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.7724198706836372e-06,
+      "loss": 0.5861,
+      "mean_token_accuracy": 0.8139378673660802,
+      "step": 2235
+    },
+    {
+      "epoch": 0.8287088420273769,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.7358800719305735e-06,
+      "loss": 0.6317,
+      "mean_token_accuracy": 0.7970269865564765,
+      "step": 2240
+    },
+    {
+      "epoch": 0.8305586385497595,
+      "grad_norm": 1.4140625,
+      "learning_rate": 1.6996850259076303e-06,
+      "loss": 0.596,
+      "mean_token_accuracy": 0.8098391036162802,
+      "step": 2245
+    },
+    {
+      "epoch": 0.832408435072142,
+      "grad_norm": 1.453125,
+      "learning_rate": 1.6638362425567012e-06,
+      "loss": 0.6426,
+      "mean_token_accuracy": 0.7943611584189123,
+      "step": 2250
+    },
+    {
+      "epoch": 0.8342582315945246,
+      "grad_norm": 1.2890625,
+      "learning_rate": 1.6283352173747148e-06,
+      "loss": 0.6138,
+      "mean_token_accuracy": 0.8052465603074348,
+      "step": 2255
+    },
+    {
+      "epoch": 0.8361080281169071,
+      "grad_norm": 1.421875,
+      "learning_rate": 1.5931834313512307e-06,
+      "loss": 0.6341,
+      "mean_token_accuracy": 0.7974937414447355,
+      "step": 2260
+    },
+    {
+      "epoch": 0.8379578246392897,
+      "grad_norm": 1.359375,
+      "learning_rate": 1.5583823509066665e-06,
+      "loss": 0.6145,
+      "mean_token_accuracy": 0.8041653112534487,
+      "step": 2265
+    },
+    {
+      "epoch": 0.8398076211616722,
+      "grad_norm": 1.4453125,
+      "learning_rate": 1.5239334278311214e-06,
+      "loss": 0.6093,
+      "mean_token_accuracy": 0.8070861543640062,
+      "step": 2270
+    },
+    {
+      "epoch": 0.8416574176840548,
+      "grad_norm": 1.421875,
+      "learning_rate": 1.489838099223816e-06,
+      "loss": 0.6051,
+      "mean_token_accuracy": 0.8060529703011069,
+      "step": 2275
+    },
+    {
+      "epoch": 0.8435072142064373,
+      "grad_norm": 1.2421875,
+      "learning_rate": 1.4560977874331284e-06,
+      "loss": 0.5922,
+      "mean_token_accuracy": 0.8105152674053515,
+      "step": 2280
+    },
+    {
+      "epoch": 0.8453570107288199,
+      "grad_norm": 1.46875,
+      "learning_rate": 1.4227138999972801e-06,
+      "loss": 0.5945,
+      "mean_token_accuracy": 0.8086310246055654,
+      "step": 2285
+    },
+    {
+      "epoch": 0.8472068072512023,
+      "grad_norm": 1.4609375,
+      "learning_rate": 1.389687829585592e-06,
+      "loss": 0.6377,
+      "mean_token_accuracy": 0.796549113303902,
+      "step": 2290
+    },
+    {
+      "epoch": 0.8490566037735849,
+      "grad_norm": 1.609375,
+      "learning_rate": 1.3570209539404067e-06,
+      "loss": 0.5862,
+      "mean_token_accuracy": 0.8135113581326359,
+      "step": 2295
+    },
+    {
+      "epoch": 0.8509064002959674,
+      "grad_norm": 1.625,
+      "learning_rate": 1.3247146358196073e-06,
+      "loss": 0.6222,
+      "mean_token_accuracy": 0.8019241472487517,
+      "step": 2300
+    },
+    {
+      "epoch": 0.8509064002959674,
+      "eval_loss": 0.6358745694160461,
+      "eval_mean_token_accuracy": 0.7953131984576216,
+      "eval_runtime": 88.8837,
+      "eval_samples_per_second": 1.451,
+      "eval_steps_per_second": 0.371,
+      "step": 2300
+    },
+    {
+      "epoch": 0.85275619681835,
+      "grad_norm": 1.28125,
+      "learning_rate": 1.2927702229397633e-06,
+      "loss": 0.62,
+      "mean_token_accuracy": 0.8019399732926807,
+      "step": 2305
+    },
+    {
+      "epoch": 0.8546059933407325,
+      "grad_norm": 1.46875,
+      "learning_rate": 1.261189047919913e-06,
+      "loss": 0.6318,
+      "mean_token_accuracy": 0.7985092731169166,
+      "step": 2310
+    },
+    {
+      "epoch": 0.8564557898631151,
+      "grad_norm": 1.4140625,
+      "learning_rate": 1.2299724282259685e-06,
+      "loss": 0.6265,
+      "mean_token_accuracy": 0.7988450504701929,
+      "step": 2315
+    },
+    {
+      "epoch": 0.8583055863854976,
+      "grad_norm": 1.421875,
+      "learning_rate": 1.1991216661157611e-06,
+      "loss": 0.574,
+      "mean_token_accuracy": 0.8160490648005693,
+      "step": 2320
+    },
+    {
+      "epoch": 0.8601553829078802,
+      "grad_norm": 1.4375,
+      "learning_rate": 1.1686380485847027e-06,
+      "loss": 0.6255,
+      "mean_token_accuracy": 0.7995161635605181,
+      "step": 2325
+    },
+    {
+      "epoch": 0.8620051794302627,
+      "grad_norm": 1.3359375,
+      "learning_rate": 1.1385228473121113e-06,
+      "loss": 0.5667,
+      "mean_token_accuracy": 0.8174841242883628,
+      "step": 2330
+    },
+    {
+      "epoch": 0.8638549759526452,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.1087773186081474e-06,
+      "loss": 0.6364,
+      "mean_token_accuracy": 0.7976796929920809,
+      "step": 2335
+    },
+    {
+      "epoch": 0.8657047724750278,
+      "grad_norm": 1.34375,
+      "learning_rate": 1.0794027033614175e-06,
+      "loss": 0.5999,
+      "mean_token_accuracy": 0.8072352814910182,
+      "step": 2340
+    },
+    {
+      "epoch": 0.8675545689974102,
+      "grad_norm": 1.359375,
+      "learning_rate": 1.0504002269871927e-06,
+      "loss": 0.5886,
+      "mean_token_accuracy": 0.8102388589052836,
+      "step": 2345
+    },
+    {
+      "epoch": 0.8694043655197928,
+      "grad_norm": 1.4453125,
+      "learning_rate": 1.0217710993763086e-06,
+      "loss": 0.5729,
+      "mean_token_accuracy": 0.8176493356600355,
+      "step": 2350
+    },
+    {
+      "epoch": 0.8712541620421753,
+      "grad_norm": 1.3671875,
+      "learning_rate": 9.935165148446658e-07,
+      "loss": 0.644,
+      "mean_token_accuracy": 0.7930041816060179,
+      "step": 2355
+    },
+    {
+      "epoch": 0.8731039585645579,
+      "grad_norm": 1.3125,
+      "learning_rate": 9.656376520834365e-07,
+      "loss": 0.6099,
+      "mean_token_accuracy": 0.8048478285903217,
+      "step": 2360
+    },
+    {
+      "epoch": 0.8749537550869404,
+      "grad_norm": 1.453125,
+      "learning_rate": 9.381356741098702e-07,
+      "loss": 0.5778,
+      "mean_token_accuracy": 0.8172098544162718,
+      "step": 2365
+    },
+    {
+      "epoch": 0.876803551609323,
+      "grad_norm": 1.2890625,
+      "learning_rate": 9.110117282187892e-07,
+      "loss": 0.548,
+      "mean_token_accuracy": 0.8233513830708346,
+      "step": 2370
+    },
+    {
+      "epoch": 0.8786533481317055,
+      "grad_norm": 1.3203125,
+      "learning_rate": 8.842669459347186e-07,
+      "loss": 0.6211,
+      "mean_token_accuracy": 0.8008686938202082,
+      "step": 2375
+    },
+    {
+      "epoch": 0.8805031446540881,
+      "grad_norm": 1.4140625,
+      "learning_rate": 8.579024429646932e-07,
+      "loss": 0.6389,
+      "mean_token_accuracy": 0.7972190906843877,
+      "step": 2380
+    },
+    {
+      "epoch": 0.8823529411764706,
+      "grad_norm": 1.2734375,
+      "learning_rate": 8.319193191517016e-07,
+      "loss": 0.6001,
+      "mean_token_accuracy": 0.80808257825244,
+      "step": 2385
+    },
+    {
+      "epoch": 0.8842027376988532,
+      "grad_norm": 1.3828125,
+      "learning_rate": 8.063186584288108e-07,
+      "loss": 0.6026,
+      "mean_token_accuracy": 0.8078743640304683,
+      "step": 2390
+    },
+    {
+      "epoch": 0.8860525342212356,
+      "grad_norm": 1.296875,
+      "learning_rate": 7.81101528773951e-07,
+      "loss": 0.6371,
+      "mean_token_accuracy": 0.7978096037617607,
+      "step": 2395
+    },
+    {
+      "epoch": 0.8879023307436182,
+      "grad_norm": 1.4375,
+      "learning_rate": 7.562689821653568e-07,
+      "loss": 0.5872,
+      "mean_token_accuracy": 0.8119488109947964,
+      "step": 2400
+    },
+    {
+      "epoch": 0.8879023307436182,
+      "eval_loss": 0.6358762979507446,
+      "eval_mean_token_accuracy": 0.7952931032718785,
+      "eval_runtime": 88.5777,
+      "eval_samples_per_second": 1.456,
+      "eval_steps_per_second": 0.373,
+      "step": 2400
+    },
+    {
+      "epoch": 0.8897521272660007,
+      "grad_norm": 1.28125,
+      "learning_rate": 7.318220545376842e-07,
+      "loss": 0.5938,
+      "mean_token_accuracy": 0.8100592681592614,
+      "step": 2405
+    },
+    {
+      "epoch": 0.8916019237883833,
+      "grad_norm": 1.5234375,
+      "learning_rate": 7.077617657387936e-07,
+      "loss": 0.6156,
+      "mean_token_accuracy": 0.8047596377457686,
+      "step": 2410
+    },
+    {
+      "epoch": 0.8934517203107658,
+      "grad_norm": 1.4609375,
+      "learning_rate": 6.840891194872112e-07,
+      "loss": 0.6283,
+      "mean_token_accuracy": 0.799004966721237,
+      "step": 2415
+    },
+    {
+      "epoch": 0.8953015168331484,
+      "grad_norm": 1.4921875,
+      "learning_rate": 6.608051033302476e-07,
+      "loss": 0.6269,
+      "mean_token_accuracy": 0.8005153671912193,
+      "step": 2420
+    },
+    {
+      "epoch": 0.8971513133555309,
+      "grad_norm": 1.5546875,
+      "learning_rate": 6.379106886028086e-07,
+      "loss": 0.5744,
+      "mean_token_accuracy": 0.8168592590376427,
+      "step": 2425
+    },
+    {
+      "epoch": 0.8990011098779135,
+      "grad_norm": 1.6640625,
+      "learning_rate": 6.154068303868666e-07,
+      "loss": 0.5979,
+      "mean_token_accuracy": 0.8072424444091446,
+      "step": 2430
+    },
+    {
+      "epoch": 0.900850906400296,
+      "grad_norm": 1.5234375,
+      "learning_rate": 5.932944674716279e-07,
+      "loss": 0.6033,
+      "mean_token_accuracy": 0.8058459298057311,
+      "step": 2435
+    },
+    {
+      "epoch": 0.9027007029226785,
+      "grad_norm": 1.4453125,
+      "learning_rate": 5.715745223143566e-07,
+      "loss": 0.6209,
+      "mean_token_accuracy": 0.8017993150912781,
+      "step": 2440
+    },
+    {
+      "epoch": 0.904550499445061,
+      "grad_norm": 1.359375,
+      "learning_rate": 5.502479010019046e-07,
+      "loss": 0.5713,
+      "mean_token_accuracy": 0.8143789382182225,
+      "step": 2445
+    },
+    {
+      "epoch": 0.9064002959674435,
+      "grad_norm": 1.4609375,
+      "learning_rate": 5.293154932129019e-07,
+      "loss": 0.6181,
+      "mean_token_accuracy": 0.8036004302428553,
+      "step": 2450
+    },
+    {
+      "epoch": 0.9082500924898261,
+      "grad_norm": 1.4765625,
+      "learning_rate": 5.087781721806539e-07,
+      "loss": 0.57,
+      "mean_token_accuracy": 0.8162700858556553,
+      "step": 2455
+    },
+    {
+      "epoch": 0.9100998890122086,
+      "grad_norm": 1.484375,
+      "learning_rate": 4.886367946567005e-07,
+      "loss": 0.6254,
+      "mean_token_accuracy": 0.8003508330206619,
+      "step": 2460
+    },
+    {
+      "epoch": 0.9119496855345912,
+      "grad_norm": 1.3671875,
+      "learning_rate": 4.6889220087508514e-07,
+      "loss": 0.5784,
+      "mean_token_accuracy": 0.8151093103069682,
+      "step": 2465
+    },
+    {
+      "epoch": 0.9137994820569737,
+      "grad_norm": 1.2890625,
+      "learning_rate": 4.4954521451729626e-07,
+      "loss": 0.5269,
+      "mean_token_accuracy": 0.8300795790001747,
+      "step": 2470
+    },
+    {
+      "epoch": 0.9156492785793563,
+      "grad_norm": 1.5546875,
+      "learning_rate": 4.305966426779118e-07,
+      "loss": 0.6149,
+      "mean_token_accuracy": 0.8040454680377913,
+      "step": 2475
+    },
+    {
+      "epoch": 0.9174990751017388,
+      "grad_norm": 1.375,
+      "learning_rate": 4.120472758309213e-07,
+      "loss": 0.5649,
+      "mean_token_accuracy": 0.8189065511308661,
+      "step": 2480
+    },
+    {
+      "epoch": 0.9193488716241214,
+      "grad_norm": 1.5078125,
+      "learning_rate": 3.9389788779675806e-07,
+      "loss": 0.6408,
+      "mean_token_accuracy": 0.795292905080418,
+      "step": 2485
+    },
+    {
+      "epoch": 0.9211986681465039,
+      "grad_norm": 1.46875,
+      "learning_rate": 3.761492357100172e-07,
+      "loss": 0.622,
+      "mean_token_accuracy": 0.8013316607240542,
+      "step": 2490
+    },
+    {
+      "epoch": 0.9230484646688865,
+      "grad_norm": 1.28125,
+      "learning_rate": 3.588020599878639e-07,
+      "loss": 0.6174,
+      "mean_token_accuracy": 0.8037950615300791,
+      "step": 2495
+    },
+    {
+      "epoch": 0.9248982611912689,
+      "grad_norm": 1.6875,
+      "learning_rate": 3.41857084299152e-07,
+      "loss": 0.5847,
+      "mean_token_accuracy": 0.8127054211263489,
+      "step": 2500
+    },
+    {
+      "epoch": 0.9248982611912689,
+      "eval_loss": 0.6359613537788391,
+      "eval_mean_token_accuracy": 0.7952690594819041,
+      "eval_runtime": 88.2267,
+      "eval_samples_per_second": 1.462,
+      "eval_steps_per_second": 0.374,
+      "step": 2500
+    },
+    {
+      "epoch": 0.9267480577136515,
+      "grad_norm": 1.3828125,
+      "learning_rate": 3.2531501553422884e-07,
+      "loss": 0.5837,
+      "mean_token_accuracy": 0.8140720012712451,
+      "step": 2505
+    },
+    {
+      "epoch": 0.928597854236034,
+      "grad_norm": 1.578125,
+      "learning_rate": 3.0917654377545567e-07,
+      "loss": 0.5918,
+      "mean_token_accuracy": 0.811925036483165,
+      "step": 2510
+    },
+    {
+      "epoch": 0.9304476507584166,
+      "grad_norm": 1.328125,
+      "learning_rate": 2.9344234226840964e-07,
+      "loss": 0.6189,
+      "mean_token_accuracy": 0.8031204520377153,
+      "step": 2515
+    },
+    {
+      "epoch": 0.9322974472807991,
+      "grad_norm": 1.4765625,
+      "learning_rate": 2.781130673938037e-07,
+      "loss": 0.6063,
+      "mean_token_accuracy": 0.8058749114893903,
+      "step": 2520
+    },
+    {
+      "epoch": 0.9341472438031817,
+      "grad_norm": 1.5078125,
+      "learning_rate": 2.6318935864010133e-07,
+      "loss": 0.6161,
+      "mean_token_accuracy": 0.8024297524151034,
+      "step": 2525
+    },
+    {
+      "epoch": 0.9359970403255642,
+      "grad_norm": 1.4921875,
+      "learning_rate": 2.4867183857684385e-07,
+      "loss": 0.5831,
+      "mean_token_accuracy": 0.8129458009715069,
+      "step": 2530
+    },
+    {
+      "epoch": 0.9378468368479467,
+      "grad_norm": 1.3359375,
+      "learning_rate": 2.3456111282867178e-07,
+      "loss": 0.5966,
+      "mean_token_accuracy": 0.8081766834156456,
+      "step": 2535
+    },
+    {
+      "epoch": 0.9396966333703293,
+      "grad_norm": 1.4765625,
+      "learning_rate": 2.2085777005006692e-07,
+      "loss": 0.5975,
+      "mean_token_accuracy": 0.8093999744597842,
+      "step": 2540
+    },
+    {
+      "epoch": 0.9415464298927118,
+      "grad_norm": 1.390625,
+      "learning_rate": 2.0756238190078991e-07,
+      "loss": 0.5934,
+      "mean_token_accuracy": 0.8105008629388287,
+      "step": 2545
+    },
+    {
+      "epoch": 0.9433962264150944,
+      "grad_norm": 1.6171875,
+      "learning_rate": 1.946755030220393e-07,
+      "loss": 0.652,
+      "mean_token_accuracy": 0.7930431672722749,
+      "step": 2550
+    },
+    {
+      "epoch": 0.9452460229374768,
+      "grad_norm": 1.4375,
+      "learning_rate": 1.8219767101330442e-07,
+      "loss": 0.5952,
+      "mean_token_accuracy": 0.8096979071812503,
+      "step": 2555
+    },
+    {
+      "epoch": 0.9470958194598594,
+      "grad_norm": 1.4375,
+      "learning_rate": 1.7012940640995012e-07,
+      "loss": 0.612,
+      "mean_token_accuracy": 0.8058889193707606,
+      "step": 2560
+    },
+    {
+      "epoch": 0.9489456159822419,
+      "grad_norm": 1.3828125,
+      "learning_rate": 1.5847121266148847e-07,
+      "loss": 0.6112,
+      "mean_token_accuracy": 0.8028099208715872,
+      "step": 2565
+    },
+    {
+      "epoch": 0.9507954125046245,
+      "grad_norm": 1.3515625,
+      "learning_rate": 1.472235761105878e-07,
+      "loss": 0.5564,
+      "mean_token_accuracy": 0.8220496380344693,
+      "step": 2570
+    },
+    {
+      "epoch": 0.952645209027007,
+      "grad_norm": 1.546875,
+      "learning_rate": 1.3638696597277678e-07,
+      "loss": 0.6248,
+      "mean_token_accuracy": 0.8020731971208164,
+      "step": 2575
+    },
+    {
+      "epoch": 0.9544950055493896,
+      "grad_norm": 1.421875,
+      "learning_rate": 1.2596183431687336e-07,
+      "loss": 0.6032,
+      "mean_token_accuracy": 0.8067943763008809,
+      "step": 2580
+    },
+    {
+      "epoch": 0.9563448020717721,
+      "grad_norm": 1.5,
+      "learning_rate": 1.159486160461265e-07,
+      "loss": 0.6125,
+      "mean_token_accuracy": 0.8030868543118614,
+      "step": 2585
+    },
+    {
+      "epoch": 0.9581945985941547,
+      "grad_norm": 1.3359375,
+      "learning_rate": 1.0634772888007073e-07,
+      "loss": 0.5912,
+      "mean_token_accuracy": 0.8095937491756745,
+      "step": 2590
+    },
+    {
+      "epoch": 0.9600443951165372,
+      "grad_norm": 1.296875,
+      "learning_rate": 9.715957333710447e-08,
+      "loss": 0.5825,
+      "mean_token_accuracy": 0.8132522887021558,
+      "step": 2595
+    },
+    {
+      "epoch": 0.9618941916389198,
+      "grad_norm": 1.3828125,
+      "learning_rate": 8.838453271777347e-08,
+      "loss": 0.5724,
+      "mean_token_accuracy": 0.8153375880818068,
+      "step": 2600
+    },
+    {
+      "epoch": 0.9618941916389198,
+      "eval_loss": 0.6359190344810486,
+      "eval_mean_token_accuracy": 0.7952561057443887,
+      "eval_runtime": 88.6273,
+      "eval_samples_per_second": 1.456,
+      "eval_steps_per_second": 0.372,
+      "step": 2600
+    },
+    {
+      "epoch": 0.9637439881613022,
+      "grad_norm": 1.46875,
+      "learning_rate": 8.002297308879359e-08,
+      "loss": 0.6279,
+      "mean_token_accuracy": 0.8006996738236367,
+      "step": 2605
+    },
+    {
+      "epoch": 0.9655937846836848,
+      "grad_norm": 1.296875,
+      "learning_rate": 7.207524326776628e-08,
+      "loss": 0.5925,
+      "mean_token_accuracy": 0.8121808557190994,
+      "step": 2610
+    },
+    {
+      "epoch": 0.9674435812060673,
+      "grad_norm": 1.375,
+      "learning_rate": 6.454167480863694e-08,
+      "loss": 0.6473,
+      "mean_token_accuracy": 0.7943891499737403,
+      "step": 2615
+    },
+    {
+      "epoch": 0.9692933777284499,
+      "grad_norm": 1.46875,
+      "learning_rate": 5.7422581987856e-08,
+      "loss": 0.612,
+      "mean_token_accuracy": 0.8035113737871222,
+      "step": 2620
+    },
+    {
+      "epoch": 0.9711431742508324,
+      "grad_norm": 1.4609375,
+      "learning_rate": 5.0718261791274924e-08,
+      "loss": 0.6492,
+      "mean_token_accuracy": 0.7923496193001656,
+      "step": 2625
+    },
+    {
+      "epoch": 0.9729929707732149,
+      "grad_norm": 1.3046875,
+      "learning_rate": 4.4428993901753924e-08,
+      "loss": 0.587,
+      "mean_token_accuracy": 0.8118751615814832,
+      "step": 2630
+    },
+    {
+      "epoch": 0.9748427672955975,
+      "grad_norm": 1.515625,
+      "learning_rate": 3.8555040687493494e-08,
+      "loss": 0.5831,
+      "mean_token_accuracy": 0.8130001939594365,
+      "step": 2635
+    },
+    {
+      "epoch": 0.97669256381798,
+      "grad_norm": 1.3046875,
+      "learning_rate": 3.309664719108874e-08,
+      "loss": 0.5622,
+      "mean_token_accuracy": 0.8186328198713086,
+      "step": 2640
+    },
+    {
+      "epoch": 0.9785423603403626,
+      "grad_norm": 1.328125,
+      "learning_rate": 2.805404111931198e-08,
+      "loss": 0.6048,
+      "mean_token_accuracy": 0.8067804518895209,
+      "step": 2645
+    },
+    {
+      "epoch": 0.9803921568627451,
+      "grad_norm": 1.4609375,
+      "learning_rate": 2.3427432833607046e-08,
+      "loss": 0.6053,
+      "mean_token_accuracy": 0.8071034493802799,
+      "step": 2650
+    },
+    {
+      "epoch": 0.9822419533851277,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.9217015341318478e-08,
+      "loss": 0.6471,
+      "mean_token_accuracy": 0.7937907528956321,
+      "step": 2655
+    },
+    {
+      "epoch": 0.9840917499075101,
+      "grad_norm": 1.4140625,
+      "learning_rate": 1.542296428763801e-08,
+      "loss": 0.5779,
+      "mean_token_accuracy": 0.8143294357103997,
+      "step": 2660
+    },
+    {
+      "epoch": 0.9859415464298927,
+      "grad_norm": 1.4609375,
+      "learning_rate": 1.2045437948275952e-08,
+      "loss": 0.6241,
+      "mean_token_accuracy": 0.8003921085855472,
+      "step": 2665
+    },
+    {
+      "epoch": 0.9877913429522752,
+      "grad_norm": 1.3203125,
+      "learning_rate": 9.084577222860935e-09,
+      "loss": 0.575,
+      "mean_token_accuracy": 0.8146093108634356,
+      "step": 2670
+    },
+    {
+      "epoch": 0.9896411394746578,
+      "grad_norm": 1.4765625,
+      "learning_rate": 6.540505629061278e-09,
+      "loss": 0.6118,
+      "mean_token_accuracy": 0.805994879781133,
+      "step": 2675
+    },
+    {
+      "epoch": 0.9914909359970403,
+      "grad_norm": 1.5234375,
+      "learning_rate": 4.413329297430213e-09,
+      "loss": 0.5971,
+      "mean_token_accuracy": 0.8082195674382581,
+      "step": 2680
+    },
+    {
+      "epoch": 0.9933407325194229,
+      "grad_norm": 1.265625,
+      "learning_rate": 2.7031369669816566e-09,
+      "loss": 0.6609,
+      "mean_token_accuracy": 0.7891590740442234,
+      "step": 2685
+    },
+    {
+      "epoch": 0.9951905290418054,
+      "grad_norm": 1.4296875,
+      "learning_rate": 1.4099999814853883e-09,
+      "loss": 0.613,
+      "mean_token_accuracy": 0.803941472771221,
+      "step": 2690
+    },
+    {
+      "epoch": 0.997040325564188,
+      "grad_norm": 1.3828125,
+      "learning_rate": 5.339722864927677e-10,
+      "loss": 0.628,
+      "mean_token_accuracy": 0.7997587664831439,
+      "step": 2695
+    },
+    {
+      "epoch": 0.9988901220865705,
+      "grad_norm": 1.4921875,
+      "learning_rate": 7.509042708631065e-11,
+      "loss": 0.612,
+      "mean_token_accuracy": 0.8036164974277161,
+      "step": 2700
+    },
+    {
+      "epoch": 0.9988901220865705,
+      "eval_loss": 0.6359476447105408,
+      "eval_mean_token_accuracy": 0.795244707693059,
+      "eval_runtime": 88.5192,
+      "eval_samples_per_second": 1.457,
+      "eval_steps_per_second": 0.373,
+      "step": 2700
+    },
+    {
+      "epoch": 1.0,
+      "mean_token_accuracy": 0.8016683712475773,
+      "step": 2703,
+      "total_flos": 3.7573754284702433e+18,
+      "train_loss": 0.6375504625844726,
+      "train_runtime": 57055.9806,
+      "train_samples_per_second": 0.379,
+      "train_steps_per_second": 0.047
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 2703,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.7573754284702433e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}