diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,12457 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 47.97773832941515,
+  "global_step": 1000000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98e-07,
+      "loss": 1.476,
+      "step": 500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998e-06,
+      "loss": 1.4485,
+      "step": 1000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9980000000000003e-06,
+      "loss": 1.429,
+      "step": 1500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.9980000000000005e-06,
+      "loss": 1.414,
+      "step": 2000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.998e-06,
+      "loss": 1.402,
+      "step": 2500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.998000000000001e-06,
+      "loss": 1.3921,
+      "step": 3000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.998000000000001e-06,
+      "loss": 1.3848,
+      "step": 3500
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 7.998e-06,
+      "loss": 1.3789,
+      "step": 4000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.996e-06,
+      "loss": 1.3727,
+      "step": 4500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.996e-06,
+      "loss": 1.3672,
+      "step": 5000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.0996e-05,
+      "loss": 1.3644,
+      "step": 5500
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.1996000000000002e-05,
+      "loss": 1.3584,
+      "step": 6000
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.2994000000000003e-05,
+      "loss": 1.3533,
+      "step": 6500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.3994e-05,
+      "loss": 1.3493,
+      "step": 7000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4994000000000002e-05,
+      "loss": 1.345,
+      "step": 7500
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.5994e-05,
+      "loss": 1.3424,
+      "step": 8000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6994e-05,
+      "loss": 1.3383,
+      "step": 8500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7990000000000002e-05,
+      "loss": 1.3371,
+      "step": 9000
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8990000000000003e-05,
+      "loss": 1.3366,
+      "step": 9500
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.9990000000000003e-05,
+      "loss": 1.3316,
+      "step": 10000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.999602594775125e-05,
+      "loss": 1.3259,
+      "step": 10500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.9992011753560593e-05,
+      "loss": 1.3241,
+      "step": 11000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.9987997559369934e-05,
+      "loss": 1.3218,
+      "step": 11500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9983983365179275e-05,
+      "loss": 1.3194,
+      "step": 12000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9979969170988617e-05,
+      "loss": 1.3155,
+      "step": 12500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.997596300518634e-05,
+      "loss": 1.3157,
+      "step": 13000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.9971948810995682e-05,
+      "loss": 1.3115,
+      "step": 13500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.9967934616805024e-05,
+      "loss": 1.3103,
+      "step": 14000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.996392042261437e-05,
+      "loss": 1.3067,
+      "step": 14500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.995990622842371e-05,
+      "loss": 1.3088,
+      "step": 15000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.995590006262143e-05,
+      "loss": 1.3048,
+      "step": 15500
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.9951885868430775e-05,
+      "loss": 1.2984,
+      "step": 16000
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.9947871674240116e-05,
+      "loss": 1.3001,
+      "step": 16500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.9943857480049455e-05,
+      "loss": 1.2977,
+      "step": 17000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9939843285858796e-05,
+      "loss": 1.2997,
+      "step": 17500
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.993582909166814e-05,
+      "loss": 1.2964,
+      "step": 18000
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.9931814897477482e-05,
+      "loss": 1.2931,
+      "step": 18500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9927800703286824e-05,
+      "loss": 1.2946,
+      "step": 19000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9923794537484547e-05,
+      "loss": 1.2931,
+      "step": 19500
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.991978034329389e-05,
+      "loss": 1.2905,
+      "step": 20000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.991576614910323e-05,
+      "loss": 1.2894,
+      "step": 20500
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.7384664361752822,
+      "eval_loss": 1.1405550241470337,
+      "eval_runtime": 50.7959,
+      "eval_samples_per_second": 894.796,
+      "eval_steps_per_second": 13.997,
+      "step": 20843
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.9911759983300954e-05,
+      "loss": 1.2902,
+      "step": 21000
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.9907745789110296e-05,
+      "loss": 1.2867,
+      "step": 21500
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.9903731594919637e-05,
+      "loss": 1.2833,
+      "step": 22000
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.989971740072898e-05,
+      "loss": 1.2828,
+      "step": 22500
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.989570320653832e-05,
+      "loss": 1.283,
+      "step": 23000
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.9891689012347665e-05,
+      "loss": 1.2803,
+      "step": 23500
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.9887674818157006e-05,
+      "loss": 1.2799,
+      "step": 24000
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.9883660623966348e-05,
+      "loss": 1.2783,
+      "step": 24500
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.987965445816407e-05,
+      "loss": 1.28,
+      "step": 25000
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.9875640263973413e-05,
+      "loss": 1.2769,
+      "step": 25500
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.987162606978275e-05,
+      "loss": 1.2756,
+      "step": 26000
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.9867619903980478e-05,
+      "loss": 1.2761,
+      "step": 26500
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.986360570978982e-05,
+      "loss": 1.2745,
+      "step": 27000
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.985959151559916e-05,
+      "loss": 1.2725,
+      "step": 27500
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.9855577321408502e-05,
+      "loss": 1.2745,
+      "step": 28000
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.9851563127217844e-05,
+      "loss": 1.2698,
+      "step": 28500
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.9847548933027185e-05,
+      "loss": 1.2697,
+      "step": 29000
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.9843534738836527e-05,
+      "loss": 1.2701,
+      "step": 29500
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.9839520544645868e-05,
+      "loss": 1.2679,
+      "step": 30000
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.9835514378843592e-05,
+      "loss": 1.2696,
+      "step": 30500
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.9831500184652933e-05,
+      "loss": 1.2663,
+      "step": 31000
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.9827485990462275e-05,
+      "loss": 1.268,
+      "step": 31500
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 1.982347179627162e-05,
+      "loss": 1.2667,
+      "step": 32000
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.981945760208096e-05,
+      "loss": 1.2663,
+      "step": 32500
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.9815443407890302e-05,
+      "loss": 1.2629,
+      "step": 33000
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9811429213699644e-05,
+      "loss": 1.2637,
+      "step": 33500
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.9807415019508985e-05,
+      "loss": 1.2631,
+      "step": 34000
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.9803400825318327e-05,
+      "loss": 1.2613,
+      "step": 34500
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.979939465951605e-05,
+      "loss": 1.261,
+      "step": 35000
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.9795380465325392e-05,
+      "loss": 1.26,
+      "step": 35500
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.9791366271134733e-05,
+      "loss": 1.2599,
+      "step": 36000
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.9787352076944075e-05,
+      "loss": 1.2596,
+      "step": 36500
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.9783337882753416e-05,
+      "loss": 1.2593,
+      "step": 37000
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.9779323688562758e-05,
+      "loss": 1.2566,
+      "step": 37500
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.977531752276048e-05,
+      "loss": 1.256,
+      "step": 38000
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.9771303328569823e-05,
+      "loss": 1.2581,
+      "step": 38500
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9767289134379168e-05,
+      "loss": 1.2557,
+      "step": 39000
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.976327494018851e-05,
+      "loss": 1.2551,
+      "step": 39500
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.975926074599785e-05,
+      "loss": 1.253,
+      "step": 40000
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.9755246551807192e-05,
+      "loss": 1.2538,
+      "step": 40500
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.9751232357616534e-05,
+      "loss": 1.2542,
+      "step": 41000
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.9747218163425875e-05,
+      "loss": 1.254,
+      "step": 41500
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.7441221615425662,
+      "eval_loss": 1.1075063943862915,
+      "eval_runtime": 50.7654,
+      "eval_samples_per_second": 895.334,
+      "eval_steps_per_second": 14.006,
+      "step": 41686
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 1.97432119976236e-05,
+      "loss": 1.2541,
+      "step": 42000
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 1.973919780343294e-05,
+      "loss": 1.2509,
+      "step": 42500
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 1.973518360924228e-05,
+      "loss": 1.2513,
+      "step": 43000
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 1.9731177443440005e-05,
+      "loss": 1.2515,
+      "step": 43500
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 1.9727163249249347e-05,
+      "loss": 1.2521,
+      "step": 44000
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 1.972314905505869e-05,
+      "loss": 1.251,
+      "step": 44500
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 1.971913486086803e-05,
+      "loss": 1.2474,
+      "step": 45000
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 1.9715128695065753e-05,
+      "loss": 1.2476,
+      "step": 45500
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 1.9711114500875098e-05,
+      "loss": 1.2475,
+      "step": 46000
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 1.9707100306684436e-05,
+      "loss": 1.2454,
+      "step": 46500
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 1.9703094140882163e-05,
+      "loss": 1.245,
+      "step": 47000
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 1.96990799466915e-05,
+      "loss": 1.2435,
+      "step": 47500
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 1.9695065752500843e-05,
+      "loss": 1.2451,
+      "step": 48000
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 1.9691051558310184e-05,
+      "loss": 1.244,
+      "step": 48500
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 1.968703736411953e-05,
+      "loss": 1.2461,
+      "step": 49000
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 1.968302316992887e-05,
+      "loss": 1.2424,
+      "step": 49500
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9679008975738212e-05,
+      "loss": 1.2443,
+      "step": 50000
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.9674994781547554e-05,
+      "loss": 1.2448,
+      "step": 50500
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.9670980587356895e-05,
+      "loss": 1.2422,
+      "step": 51000
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.966697442155462e-05,
+      "loss": 1.2429,
+      "step": 51500
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.966296022736396e-05,
+      "loss": 1.2397,
+      "step": 52000
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.96589460331733e-05,
+      "loss": 1.2433,
+      "step": 52500
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.9654931838982646e-05,
+      "loss": 1.2399,
+      "step": 53000
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.9650917644791988e-05,
+      "loss": 1.2439,
+      "step": 53500
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.9646911478989708e-05,
+      "loss": 1.2379,
+      "step": 54000
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.9642897284799053e-05,
+      "loss": 1.2375,
+      "step": 54500
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.9638883090608395e-05,
+      "loss": 1.238,
+      "step": 55000
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.9634868896417733e-05,
+      "loss": 1.24,
+      "step": 55500
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.9630854702227074e-05,
+      "loss": 1.2383,
+      "step": 56000
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.962684050803642e-05,
+      "loss": 1.2376,
+      "step": 56500
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.962282631384576e-05,
+      "loss": 1.2366,
+      "step": 57000
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.9618812119655102e-05,
+      "loss": 1.2346,
+      "step": 57500
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.9614805953852825e-05,
+      "loss": 1.2354,
+      "step": 58000
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9610791759662167e-05,
+      "loss": 1.2372,
+      "step": 58500
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.960677756547151e-05,
+      "loss": 1.2365,
+      "step": 59000
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.960276337128085e-05,
+      "loss": 1.2371,
+      "step": 59500
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 1.959874917709019e-05,
+      "loss": 1.2319,
+      "step": 60000
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 1.9594743011287915e-05,
+      "loss": 1.2307,
+      "step": 60500
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 1.9590728817097256e-05,
+      "loss": 1.2327,
+      "step": 61000
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.95867146229066e-05,
+      "loss": 1.2338,
+      "step": 61500
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.9582700428715943e-05,
+      "loss": 1.2325,
+      "step": 62000
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.9578686234525284e-05,
+      "loss": 1.2317,
+      "step": 62500
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.7469108220467837,
+      "eval_loss": 1.090468406677246,
+      "eval_runtime": 50.1635,
+      "eval_samples_per_second": 906.077,
+      "eval_steps_per_second": 14.174,
+      "step": 62529
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 1.9574680068723008e-05,
+      "loss": 1.2316,
+      "step": 63000
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 1.957066587453235e-05,
+      "loss": 1.2295,
+      "step": 63500
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 1.956665168034169e-05,
+      "loss": 1.2298,
+      "step": 64000
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 1.9562637486151032e-05,
+      "loss": 1.2294,
+      "step": 64500
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 1.9558623291960374e-05,
+      "loss": 1.2294,
+      "step": 65000
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 1.9554609097769715e-05,
+      "loss": 1.2268,
+      "step": 65500
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 1.9550594903579057e-05,
+      "loss": 1.2299,
+      "step": 66000
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 1.9546580709388398e-05,
+      "loss": 1.2301,
+      "step": 66500
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 1.9542574543586122e-05,
+      "loss": 1.229,
+      "step": 67000
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 1.9538568377783846e-05,
+      "loss": 1.2271,
+      "step": 67500
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 1.9534554183593187e-05,
+      "loss": 1.2263,
+      "step": 68000
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 1.953053998940253e-05,
+      "loss": 1.2279,
+      "step": 68500
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 1.952652579521187e-05,
+      "loss": 1.2259,
+      "step": 69000
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 1.9522519629409594e-05,
+      "loss": 1.227,
+      "step": 69500
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9518505435218935e-05,
+      "loss": 1.2235,
+      "step": 70000
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9514491241028276e-05,
+      "loss": 1.2277,
+      "step": 70500
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9510477046837618e-05,
+      "loss": 1.2255,
+      "step": 71000
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9506462852646963e-05,
+      "loss": 1.2255,
+      "step": 71500
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9502448658456304e-05,
+      "loss": 1.2226,
+      "step": 72000
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9498442492654025e-05,
+      "loss": 1.2242,
+      "step": 72500
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.949442829846337e-05,
+      "loss": 1.2241,
+      "step": 73000
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.949041410427271e-05,
+      "loss": 1.2254,
+      "step": 73500
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9486399910082052e-05,
+      "loss": 1.2266,
+      "step": 74000
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9482393744279776e-05,
+      "loss": 1.2233,
+      "step": 74500
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9478379550089117e-05,
+      "loss": 1.2234,
+      "step": 75000
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.947437338428684e-05,
+      "loss": 1.2256,
+      "step": 75500
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9470359190096183e-05,
+      "loss": 1.2197,
+      "step": 76000
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9466344995905524e-05,
+      "loss": 1.2225,
+      "step": 76500
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.9462330801714866e-05,
+      "loss": 1.2199,
+      "step": 77000
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.9458316607524207e-05,
+      "loss": 1.222,
+      "step": 77500
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.945430241333355e-05,
+      "loss": 1.2193,
+      "step": 78000
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.945028821914289e-05,
+      "loss": 1.2202,
+      "step": 78500
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.944627402495223e-05,
+      "loss": 1.2177,
+      "step": 79000
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.9442259830761573e-05,
+      "loss": 1.2204,
+      "step": 79500
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.9438245636570918e-05,
+      "loss": 1.2205,
+      "step": 80000
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.943423144238026e-05,
+      "loss": 1.2222,
+      "step": 80500
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.94302172481896e-05,
+      "loss": 1.2201,
+      "step": 81000
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.9426211082387324e-05,
+      "loss": 1.2184,
+      "step": 81500
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.9422204916585045e-05,
+      "loss": 1.2193,
+      "step": 82000
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.941819072239439e-05,
+      "loss": 1.2191,
+      "step": 82500
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.941417652820373e-05,
+      "loss": 1.2164,
+      "step": 83000
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.7502087859412128,
+      "eval_loss": 1.0713489055633545,
+      "eval_runtime": 50.7255,
+      "eval_samples_per_second": 896.038,
+      "eval_steps_per_second": 14.017,
+      "step": 83372
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.9410162334013072e-05,
+      "loss": 1.2183,
+      "step": 83500
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.9406148139822414e-05,
+      "loss": 1.2176,
+      "step": 84000
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.9402133945631755e-05,
+      "loss": 1.2153,
+      "step": 84500
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.939812777982948e-05,
+      "loss": 1.2137,
+      "step": 85000
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.939411358563882e-05,
+      "loss": 1.2172,
+      "step": 85500
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.9390099391448162e-05,
+      "loss": 1.2146,
+      "step": 86000
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.9386085197257503e-05,
+      "loss": 1.2169,
+      "step": 86500
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.9382071003066848e-05,
+      "loss": 1.2167,
+      "step": 87000
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.9378056808876186e-05,
+      "loss": 1.2157,
+      "step": 87500
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.9374042614685528e-05,
+      "loss": 1.2159,
+      "step": 88000
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.841478787878788e-05,
+      "loss": 1.2124,
+      "step": 88500
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.8404707070707074e-05,
+      "loss": 1.2114,
+      "step": 89000
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.8394606060606063e-05,
+      "loss": 1.2107,
+      "step": 89500
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.838450505050505e-05,
+      "loss": 1.2125,
+      "step": 90000
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.837440404040404e-05,
+      "loss": 1.2095,
+      "step": 90500
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.8364323232323234e-05,
+      "loss": 1.2119,
+      "step": 91000
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.8354242424242424e-05,
+      "loss": 1.2105,
+      "step": 91500
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.8344141414141416e-05,
+      "loss": 1.2122,
+      "step": 92000
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.8334040404040405e-05,
+      "loss": 1.2104,
+      "step": 92500
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.8323939393939397e-05,
+      "loss": 1.2087,
+      "step": 93000
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.8313858585858587e-05,
+      "loss": 1.2138,
+      "step": 93500
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.8303757575757576e-05,
+      "loss": 1.2115,
+      "step": 94000
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.8293656565656568e-05,
+      "loss": 1.2122,
+      "step": 94500
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.8283555555555557e-05,
+      "loss": 1.2108,
+      "step": 95000
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.827345454545455e-05,
+      "loss": 1.2087,
+      "step": 95500
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.8263353535353537e-05,
+      "loss": 1.2109,
+      "step": 96000
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.8253272727272728e-05,
+      "loss": 1.2094,
+      "step": 96500
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.824317171717172e-05,
+      "loss": 1.2102,
+      "step": 97000
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.823307070707071e-05,
+      "loss": 1.2094,
+      "step": 97500
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.82229696969697e-05,
+      "loss": 1.2112,
+      "step": 98000
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.821286868686869e-05,
+      "loss": 1.2091,
+      "step": 98500
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.8202767676767677e-05,
+      "loss": 1.2099,
+      "step": 99000
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.8192666666666666e-05,
+      "loss": 1.2068,
+      "step": 99500
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.8182565656565658e-05,
+      "loss": 1.2066,
+      "step": 100000
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.8172484848484852e-05,
+      "loss": 1.2057,
+      "step": 100500
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.816238383838384e-05,
+      "loss": 1.207,
+      "step": 101000
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.815228282828283e-05,
+      "loss": 1.2108,
+      "step": 101500
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.8142181818181818e-05,
+      "loss": 1.2084,
+      "step": 102000
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.8132101010101012e-05,
+      "loss": 1.2074,
+      "step": 102500
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.8122e-05,
+      "loss": 1.2085,
+      "step": 103000
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.811189898989899e-05,
+      "loss": 1.2064,
+      "step": 103500
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.810179797979798e-05,
+      "loss": 1.2046,
+      "step": 104000
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.809169696969697e-05,
+      "loss": 1.204,
+      "step": 104500
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.808159595959596e-05,
+      "loss": 1.2034,
+      "step": 105000
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.807149494949495e-05,
+      "loss": 1.2081,
+      "step": 105500
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.8061393939393942e-05,
+      "loss": 1.2057,
+      "step": 106000
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.8051313131313133e-05,
+      "loss": 1.2045,
+      "step": 106500
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.804121212121212e-05,
+      "loss": 1.2067,
+      "step": 107000
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.8031111111111113e-05,
+      "loss": 1.2034,
+      "step": 107500
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.8021010101010105e-05,
+      "loss": 1.2023,
+      "step": 108000
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.8010929292929292e-05,
+      "loss": 1.2039,
+      "step": 108500
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.7522697966340297,
+      "eval_loss": 1.059805154800415,
+      "eval_runtime": 50.5864,
+      "eval_samples_per_second": 898.503,
+      "eval_steps_per_second": 14.055,
+      "step": 108843
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.8000828282828284e-05,
+      "loss": 1.2035,
+      "step": 109000
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.7990727272727276e-05,
+      "loss": 1.2028,
+      "step": 109500
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.7980626262626265e-05,
+      "loss": 1.1997,
+      "step": 110000
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.7970525252525253e-05,
+      "loss": 1.2005,
+      "step": 110500
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.7960444444444447e-05,
+      "loss": 1.2001,
+      "step": 111000
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.7950343434343436e-05,
+      "loss": 1.2,
+      "step": 111500
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.7940242424242428e-05,
+      "loss": 1.204,
+      "step": 112000
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.7930141414141417e-05,
+      "loss": 1.1987,
+      "step": 112500
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.7920040404040405e-05,
+      "loss": 1.2023,
+      "step": 113000
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.79099595959596e-05,
+      "loss": 1.2011,
+      "step": 113500
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.7899858585858588e-05,
+      "loss": 1.2025,
+      "step": 114000
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.788975757575758e-05,
+      "loss": 1.2021,
+      "step": 114500
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.787967676767677e-05,
+      "loss": 1.1998,
+      "step": 115000
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.786957575757576e-05,
+      "loss": 1.1957,
+      "step": 115500
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.785947474747475e-05,
+      "loss": 1.2021,
+      "step": 116000
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.784937373737374e-05,
+      "loss": 1.1989,
+      "step": 116500
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.783929292929293e-05,
+      "loss": 1.1992,
+      "step": 117000
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.7829191919191922e-05,
+      "loss": 1.2002,
+      "step": 117500
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.781909090909091e-05,
+      "loss": 1.2009,
+      "step": 118000
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.7808989898989902e-05,
+      "loss": 1.1987,
+      "step": 118500
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.779888888888889e-05,
+      "loss": 1.1989,
+      "step": 119000
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.778878787878788e-05,
+      "loss": 1.1991,
+      "step": 119500
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.7778686868686868e-05,
+      "loss": 1.1983,
+      "step": 120000
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.776858585858586e-05,
+      "loss": 1.197,
+      "step": 120500
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.775848484848485e-05,
+      "loss": 1.1974,
+      "step": 121000
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.774838383838384e-05,
+      "loss": 1.1982,
+      "step": 121500
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.773830303030303e-05,
+      "loss": 1.1969,
+      "step": 122000
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.772820202020202e-05,
+      "loss": 1.1971,
+      "step": 122500
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.7718101010101012e-05,
+      "loss": 1.1969,
+      "step": 123000
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.7708e-05,
+      "loss": 1.2005,
+      "step": 123500
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.7697919191919194e-05,
+      "loss": 1.1947,
+      "step": 124000
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.7687818181818183e-05,
+      "loss": 1.1934,
+      "step": 124500
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.767771717171717e-05,
+      "loss": 1.1966,
+      "step": 125000
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.7667616161616164e-05,
+      "loss": 1.1939,
+      "step": 125500
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.7657515151515152e-05,
+      "loss": 1.1967,
+      "step": 126000
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.7647414141414144e-05,
+      "loss": 1.1985,
+      "step": 126500
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.7637333333333335e-05,
+      "loss": 1.1954,
+      "step": 127000
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.7627232323232323e-05,
+      "loss": 1.197,
+      "step": 127500
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.7617131313131315e-05,
+      "loss": 1.1936,
+      "step": 128000
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.7607030303030304e-05,
+      "loss": 1.1989,
+      "step": 128500
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.7596949494949494e-05,
+      "loss": 1.1959,
+      "step": 129000
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.7586848484848486e-05,
+      "loss": 1.195,
+      "step": 129500
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.753468044270212,
+      "eval_loss": 1.0520988702774048,
+      "eval_runtime": 50.6364,
+      "eval_samples_per_second": 897.615,
+      "eval_steps_per_second": 14.041,
+      "step": 129686
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.7576747474747475e-05,
+      "loss": 1.1925,
+      "step": 130000
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.7566646464646467e-05,
+      "loss": 1.1943,
+      "step": 130500
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.7556545454545456e-05,
+      "loss": 1.1923,
+      "step": 131000
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.7546444444444448e-05,
+      "loss": 1.1911,
+      "step": 131500
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.7536343434343436e-05,
+      "loss": 1.1934,
+      "step": 132000
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.7526242424242425e-05,
+      "loss": 1.1927,
+      "step": 132500
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.7516141414141417e-05,
+      "loss": 1.1943,
+      "step": 133000
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.7506040404040405e-05,
+      "loss": 1.192,
+      "step": 133500
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.7495959595959596e-05,
+      "loss": 1.1891,
+      "step": 134000
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.7485858585858588e-05,
+      "loss": 1.1918,
+      "step": 134500
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.7475757575757577e-05,
+      "loss": 1.1932,
+      "step": 135000
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.746565656565657e-05,
+      "loss": 1.1933,
+      "step": 135500
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.7455555555555557e-05,
+      "loss": 1.1918,
+      "step": 136000
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.744545454545455e-05,
+      "loss": 1.1914,
+      "step": 136500
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.743537373737374e-05,
+      "loss": 1.1904,
+      "step": 137000
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.7425272727272728e-05,
+      "loss": 1.189,
+      "step": 137500
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.741517171717172e-05,
+      "loss": 1.1894,
+      "step": 138000
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.740507070707071e-05,
+      "loss": 1.1917,
+      "step": 138500
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.7394969696969697e-05,
+      "loss": 1.191,
+      "step": 139000
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.738486868686869e-05,
+      "loss": 1.1908,
+      "step": 139500
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.7374767676767678e-05,
+      "loss": 1.19,
+      "step": 140000
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.7364666666666667e-05,
+      "loss": 1.1917,
+      "step": 140500
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.7354606060606062e-05,
+      "loss": 1.1894,
+      "step": 141000
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.734450505050505e-05,
+      "loss": 1.1918,
+      "step": 141500
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.7334404040404043e-05,
+      "loss": 1.1887,
+      "step": 142000
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.732430303030303e-05,
+      "loss": 1.1885,
+      "step": 142500
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.7314202020202024e-05,
+      "loss": 1.1885,
+      "step": 143000
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.7304101010101012e-05,
+      "loss": 1.1898,
+      "step": 143500
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 1.7294020202020203e-05,
+      "loss": 1.1878,
+      "step": 144000
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 1.7283919191919195e-05,
+      "loss": 1.188,
+      "step": 144500
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 1.7273838383838385e-05,
+      "loss": 1.1876,
+      "step": 145000
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 1.7263737373737374e-05,
+      "loss": 1.1899,
+      "step": 145500
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 1.7253636363636366e-05,
+      "loss": 1.1887,
+      "step": 146000
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 1.7243535353535354e-05,
+      "loss": 1.1872,
+      "step": 146500
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 1.7233434343434346e-05,
+      "loss": 1.1916,
+      "step": 147000
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 1.7223333333333335e-05,
+      "loss": 1.1894,
+      "step": 147500
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 1.7213232323232324e-05,
+      "loss": 1.1887,
+      "step": 148000
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 1.7203131313131312e-05,
+      "loss": 1.1874,
+      "step": 148500
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 1.7193030303030304e-05,
+      "loss": 1.1885,
+      "step": 149000
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 1.7182949494949498e-05,
+      "loss": 1.1873,
+      "step": 149500
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 1.7172848484848487e-05,
+      "loss": 1.1897,
+      "step": 150000
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 1.7162747474747475e-05,
+      "loss": 1.1863,
+      "step": 150500
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.7543280591911518,
+      "eval_loss": 1.0471174716949463,
+      "eval_runtime": 50.895,
+      "eval_samples_per_second": 893.054,
+      "eval_steps_per_second": 13.97,
+      "step": 150529
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 1.715266666666667e-05,
+      "loss": 1.1874,
+      "step": 151000
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 1.7142565656565658e-05,
+      "loss": 1.186,
+      "step": 151500
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 1.713246464646465e-05,
+      "loss": 1.1854,
+      "step": 152000
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 1.712236363636364e-05,
+      "loss": 1.186,
+      "step": 152500
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 1.7112262626262627e-05,
+      "loss": 1.1838,
+      "step": 153000
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 1.7102161616161616e-05,
+      "loss": 1.1857,
+      "step": 153500
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 1.7092060606060608e-05,
+      "loss": 1.1851,
+      "step": 154000
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 1.7081959595959596e-05,
+      "loss": 1.1859,
+      "step": 154500
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 1.7071878787878787e-05,
+      "loss": 1.1849,
+      "step": 155000
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 1.706177777777778e-05,
+      "loss": 1.1856,
+      "step": 155500
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 1.7051676767676767e-05,
+      "loss": 1.183,
+      "step": 156000
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 1.704157575757576e-05,
+      "loss": 1.1882,
+      "step": 156500
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 1.703147474747475e-05,
+      "loss": 1.1833,
+      "step": 157000
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 1.7021414141414144e-05,
+      "loss": 1.1847,
+      "step": 157500
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 1.7011313131313132e-05,
+      "loss": 1.1821,
+      "step": 158000
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 1.7001212121212124e-05,
+      "loss": 1.1847,
+      "step": 158500
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 1.6991111111111113e-05,
+      "loss": 1.1846,
+      "step": 159000
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 1.69810101010101e-05,
+      "loss": 1.1812,
+      "step": 159500
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 1.697090909090909e-05,
+      "loss": 1.1805,
+      "step": 160000
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 1.6960828282828284e-05,
+      "loss": 1.1834,
+      "step": 160500
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 1.6950727272727273e-05,
+      "loss": 1.1825,
+      "step": 161000
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 1.6940626262626265e-05,
+      "loss": 1.1827,
+      "step": 161500
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 1.6930525252525253e-05,
+      "loss": 1.1849,
+      "step": 162000
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 1.6920424242424245e-05,
+      "loss": 1.1843,
+      "step": 162500
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 1.6910323232323234e-05,
+      "loss": 1.1836,
+      "step": 163000
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 1.6900222222222226e-05,
+      "loss": 1.1824,
+      "step": 163500
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 1.6890121212121214e-05,
+      "loss": 1.1832,
+      "step": 164000
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 1.6880020202020203e-05,
+      "loss": 1.1806,
+      "step": 164500
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 1.686991919191919e-05,
+      "loss": 1.1829,
+      "step": 165000
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 1.6859818181818184e-05,
+      "loss": 1.1814,
+      "step": 165500
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 1.6849737373737377e-05,
+      "loss": 1.1819,
+      "step": 166000
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 1.6839636363636366e-05,
+      "loss": 1.1834,
+      "step": 166500
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 1.6829535353535355e-05,
+      "loss": 1.1805,
+      "step": 167000
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 1.6819434343434343e-05,
+      "loss": 1.1798,
+      "step": 167500
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 1.6809333333333335e-05,
+      "loss": 1.1804,
+      "step": 168000
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 1.6799232323232324e-05,
+      "loss": 1.1811,
+      "step": 168500
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 1.6789151515151514e-05,
+      "loss": 1.1794,
+      "step": 169000
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 1.6779050505050506e-05,
+      "loss": 1.1801,
+      "step": 169500
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 1.6768949494949495e-05,
+      "loss": 1.1818,
+      "step": 170000
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 1.6758848484848487e-05,
+      "loss": 1.1832,
+      "step": 170500
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 1.6748747474747476e-05,
+      "loss": 1.1799,
+      "step": 171000
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.7561843539819769,
+      "eval_loss": 1.0365961790084839,
+      "eval_runtime": 50.8197,
+      "eval_samples_per_second": 894.377,
+      "eval_steps_per_second": 13.991,
+      "step": 171372
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 1.6738646464646468e-05,
+      "loss": 1.1805,
+      "step": 171500
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 1.6728565656565658e-05,
+      "loss": 1.1782,
+      "step": 172000
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 1.6718464646464647e-05,
+      "loss": 1.1806,
+      "step": 172500
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 1.670836363636364e-05,
+      "loss": 1.177,
+      "step": 173000
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 1.6698262626262627e-05,
+      "loss": 1.1805,
+      "step": 173500
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 1.668816161616162e-05,
+      "loss": 1.1799,
+      "step": 174000
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 1.6678060606060608e-05,
+      "loss": 1.1791,
+      "step": 174500
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 1.6667959595959596e-05,
+      "loss": 1.1801,
+      "step": 175000
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 1.6657858585858585e-05,
+      "loss": 1.1785,
+      "step": 175500
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 1.664777777777778e-05,
+      "loss": 1.178,
+      "step": 176000
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 1.6637676767676768e-05,
+      "loss": 1.1758,
+      "step": 176500
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 1.662759595959596e-05,
+      "loss": 1.1773,
+      "step": 177000
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 1.661749494949495e-05,
+      "loss": 1.1783,
+      "step": 177500
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 1.6607393939393942e-05,
+      "loss": 1.1797,
+      "step": 178000
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 1.659729292929293e-05,
+      "loss": 1.1758,
+      "step": 178500
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 1.658719191919192e-05,
+      "loss": 1.1774,
+      "step": 179000
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 1.6577111111111113e-05,
+      "loss": 1.1753,
+      "step": 179500
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 1.6567010101010102e-05,
+      "loss": 1.1783,
+      "step": 180000
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 1.6556909090909094e-05,
+      "loss": 1.1793,
+      "step": 180500
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 1.6546808080808082e-05,
+      "loss": 1.177,
+      "step": 181000
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 1.6536727272727273e-05,
+      "loss": 1.1783,
+      "step": 181500
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 1.6526626262626265e-05,
+      "loss": 1.177,
+      "step": 182000
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 1.6516525252525253e-05,
+      "loss": 1.1777,
+      "step": 182500
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 1.6506424242424245e-05,
+      "loss": 1.1753,
+      "step": 183000
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 1.6496323232323234e-05,
+      "loss": 1.1775,
+      "step": 183500
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 1.6486222222222223e-05,
+      "loss": 1.1751,
+      "step": 184000
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 1.6476141414141417e-05,
+      "loss": 1.1772,
+      "step": 184500
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 1.6466040404040405e-05,
+      "loss": 1.1762,
+      "step": 185000
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 1.6455939393939394e-05,
+      "loss": 1.1742,
+      "step": 185500
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 1.6445858585858588e-05,
+      "loss": 1.1758,
+      "step": 186000
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 1.6435757575757576e-05,
+      "loss": 1.1748,
+      "step": 186500
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 1.6425656565656568e-05,
+      "loss": 1.1751,
+      "step": 187000
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 1.6415555555555557e-05,
+      "loss": 1.1777,
+      "step": 187500
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 1.6405454545454545e-05,
+      "loss": 1.1773,
+      "step": 188000
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 1.6395353535353537e-05,
+      "loss": 1.1745,
+      "step": 188500
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 1.6385252525252526e-05,
+      "loss": 1.1781,
+      "step": 189000
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 1.6375151515151518e-05,
+      "loss": 1.1718,
+      "step": 189500
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 1.636507070707071e-05,
+      "loss": 1.1753,
+      "step": 190000
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 1.6354969696969697e-05,
+      "loss": 1.1766,
+      "step": 190500
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 1.634486868686869e-05,
+      "loss": 1.1753,
+      "step": 191000
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 1.6334767676767678e-05,
+      "loss": 1.1762,
+      "step": 191500
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 1.632466666666667e-05,
+      "loss": 1.1747,
+      "step": 192000
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.7567075231320115,
+      "eval_loss": 1.0325676202774048,
+      "eval_runtime": 50.8835,
+      "eval_samples_per_second": 893.256,
+      "eval_steps_per_second": 13.973,
+      "step": 192215
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 1.631456565656566e-05,
+      "loss": 1.1755,
+      "step": 192500
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 1.6304464646464647e-05,
+      "loss": 1.1714,
+      "step": 193000
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 1.629438383838384e-05,
+      "loss": 1.1751,
+      "step": 193500
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 1.628428282828283e-05,
+      "loss": 1.1735,
+      "step": 194000
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 1.627418181818182e-05,
+      "loss": 1.1733,
+      "step": 194500
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 1.626408080808081e-05,
+      "loss": 1.1721,
+      "step": 195000
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 1.6254e-05,
+      "loss": 1.1737,
+      "step": 195500
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 1.6243898989898993e-05,
+      "loss": 1.1736,
+      "step": 196000
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 1.623379797979798e-05,
+      "loss": 1.1722,
+      "step": 196500
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 1.6223696969696973e-05,
+      "loss": 1.1731,
+      "step": 197000
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 1.6213595959595962e-05,
+      "loss": 1.1711,
+      "step": 197500
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 1.6203515151515152e-05,
+      "loss": 1.1731,
+      "step": 198000
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 1.6193434343434343e-05,
+      "loss": 1.1713,
+      "step": 198500
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 1.6183333333333335e-05,
+      "loss": 1.1741,
+      "step": 199000
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.6173232323232323e-05,
+      "loss": 1.1705,
+      "step": 199500
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.6163131313131315e-05,
+      "loss": 1.1713,
+      "step": 200000
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.6153030303030304e-05,
+      "loss": 1.1721,
+      "step": 200500
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.6142929292929296e-05,
+      "loss": 1.1729,
+      "step": 201000
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 1.6132828282828285e-05,
+      "loss": 1.1733,
+      "step": 201500
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 1.6122727272727273e-05,
+      "loss": 1.1718,
+      "step": 202000
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 1.6112626262626262e-05,
+      "loss": 1.1696,
+      "step": 202500
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 1.6102525252525254e-05,
+      "loss": 1.1725,
+      "step": 203000
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 1.6092424242424242e-05,
+      "loss": 1.1695,
+      "step": 203500
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 1.6082323232323234e-05,
+      "loss": 1.1719,
+      "step": 204000
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 1.6072242424242425e-05,
+      "loss": 1.1672,
+      "step": 204500
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 1.6062141414141413e-05,
+      "loss": 1.1717,
+      "step": 205000
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 1.6052040404040405e-05,
+      "loss": 1.1701,
+      "step": 205500
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 1.6041959595959596e-05,
+      "loss": 1.1727,
+      "step": 206000
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 1.6031858585858585e-05,
+      "loss": 1.1704,
+      "step": 206500
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 1.6021757575757577e-05,
+      "loss": 1.1709,
+      "step": 207000
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.6011656565656565e-05,
+      "loss": 1.1679,
+      "step": 207500
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 1.6001555555555557e-05,
+      "loss": 1.1704,
+      "step": 208000
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.7577379933925333,
+      "eval_loss": 1.027896523475647,
+      "eval_runtime": 50.1492,
+      "eval_samples_per_second": 906.335,
+      "eval_steps_per_second": 14.178,
+      "step": 208430
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 1.599145454545455e-05,
+      "loss": 1.1683,
+      "step": 208500
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 1.5981353535353538e-05,
+      "loss": 1.1713,
+      "step": 209000
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 1.5971252525252526e-05,
+      "loss": 1.1704,
+      "step": 209500
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 1.5961151515151515e-05,
+      "loss": 1.1691,
+      "step": 210000
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 1.595107070707071e-05,
+      "loss": 1.1674,
+      "step": 210500
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 1.59409696969697e-05,
+      "loss": 1.1688,
+      "step": 211000
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 1.593086868686869e-05,
+      "loss": 1.1714,
+      "step": 211500
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 1.5920767676767678e-05,
+      "loss": 1.1665,
+      "step": 212000
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 1.5910686868686872e-05,
+      "loss": 1.168,
+      "step": 212500
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 1.590058585858586e-05,
+      "loss": 1.1696,
+      "step": 213000
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 1.589048484848485e-05,
+      "loss": 1.1706,
+      "step": 213500
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 1.5880383838383838e-05,
+      "loss": 1.1684,
+      "step": 214000
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 1.587028282828283e-05,
+      "loss": 1.1664,
+      "step": 214500
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 1.5860181818181818e-05,
+      "loss": 1.1695,
+      "step": 215000
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 1.585008080808081e-05,
+      "loss": 1.167,
+      "step": 215500
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 1.58399797979798e-05,
+      "loss": 1.1673,
+      "step": 216000
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 1.582987878787879e-05,
+      "loss": 1.169,
+      "step": 216500
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 1.581979797979798e-05,
+      "loss": 1.1661,
+      "step": 217000
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 1.580969696969697e-05,
+      "loss": 1.166,
+      "step": 217500
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 1.5799616161616164e-05,
+      "loss": 1.1679,
+      "step": 218000
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 1.5789515151515153e-05,
+      "loss": 1.1683,
+      "step": 218500
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 1.577941414141414e-05,
+      "loss": 1.1699,
+      "step": 219000
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 1.5769313131313133e-05,
+      "loss": 1.1684,
+      "step": 219500
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 1.575921212121212e-05,
+      "loss": 1.1673,
+      "step": 220000
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 1.5749151515151518e-05,
+      "loss": 1.1674,
+      "step": 220500
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 1.5739050505050506e-05,
+      "loss": 1.1669,
+      "step": 221000
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 1.5728949494949498e-05,
+      "loss": 1.1651,
+      "step": 221500
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 1.5718848484848487e-05,
+      "loss": 1.1653,
+      "step": 222000
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 1.5708747474747475e-05,
+      "loss": 1.1662,
+      "step": 222500
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 1.5698646464646464e-05,
+      "loss": 1.1693,
+      "step": 223000
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 1.5688545454545456e-05,
+      "loss": 1.1661,
+      "step": 223500
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 1.5678444444444445e-05,
+      "loss": 1.1669,
+      "step": 224000
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 1.5668343434343436e-05,
+      "loss": 1.1665,
+      "step": 224500
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 1.5658262626262627e-05,
+      "loss": 1.1672,
+      "step": 225000
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 1.5648161616161616e-05,
+      "loss": 1.1665,
+      "step": 225500
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 1.5638060606060608e-05,
+      "loss": 1.1634,
+      "step": 226000
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 1.5627959595959596e-05,
+      "loss": 1.1672,
+      "step": 226500
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 1.5617858585858588e-05,
+      "loss": 1.1626,
+      "step": 227000
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 1.5607757575757577e-05,
+      "loss": 1.1666,
+      "step": 227500
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 1.5597656565656565e-05,
+      "loss": 1.1666,
+      "step": 228000
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 1.5587555555555557e-05,
+      "loss": 1.1672,
+      "step": 228500
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 1.5577474747474748e-05,
+      "loss": 1.1646,
+      "step": 229000
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.7584813338269704,
+      "eval_loss": 1.0239979028701782,
+      "eval_runtime": 50.2811,
+      "eval_samples_per_second": 903.957,
+      "eval_steps_per_second": 14.14,
+      "step": 229273
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 1.556737373737374e-05,
+      "loss": 1.1664,
+      "step": 229500
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 1.555727272727273e-05,
+      "loss": 1.162,
+      "step": 230000
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 1.5547171717171717e-05,
+      "loss": 1.1629,
+      "step": 230500
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 1.553707070707071e-05,
+      "loss": 1.1649,
+      "step": 231000
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 1.55269898989899e-05,
+      "loss": 1.1641,
+      "step": 231500
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 1.551688888888889e-05,
+      "loss": 1.1677,
+      "step": 232000
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 1.550678787878788e-05,
+      "loss": 1.1642,
+      "step": 232500
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 1.549668686868687e-05,
+      "loss": 1.165,
+      "step": 233000
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 1.5486606060606063e-05,
+      "loss": 1.1632,
+      "step": 233500
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 1.547650505050505e-05,
+      "loss": 1.1653,
+      "step": 234000
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 1.5466424242424242e-05,
+      "loss": 1.1642,
+      "step": 234500
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 1.5456323232323234e-05,
+      "loss": 1.1618,
+      "step": 235000
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 1.5446222222222222e-05,
+      "loss": 1.1618,
+      "step": 235500
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 1.5436121212121214e-05,
+      "loss": 1.1654,
+      "step": 236000
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 1.5426040404040405e-05,
+      "loss": 1.1617,
+      "step": 236500
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 1.5415939393939394e-05,
+      "loss": 1.1626,
+      "step": 237000
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 1.5405838383838386e-05,
+      "loss": 1.1617,
+      "step": 237500
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 1.5395737373737378e-05,
+      "loss": 1.1627,
+      "step": 238000
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 1.5385636363636366e-05,
+      "loss": 1.1611,
+      "step": 238500
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 1.5375555555555557e-05,
+      "loss": 1.1644,
+      "step": 239000
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 1.5365454545454545e-05,
+      "loss": 1.1622,
+      "step": 239500
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 1.5355353535353537e-05,
+      "loss": 1.1609,
+      "step": 240000
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 1.5345252525252526e-05,
+      "loss": 1.1645,
+      "step": 240500
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 1.5335171717171716e-05,
+      "loss": 1.161,
+      "step": 241000
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 1.532507070707071e-05,
+      "loss": 1.1637,
+      "step": 241500
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 1.53149696969697e-05,
+      "loss": 1.1616,
+      "step": 242000
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 1.530486868686869e-05,
+      "loss": 1.1652,
+      "step": 242500
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 1.5294767676767678e-05,
+      "loss": 1.162,
+      "step": 243000
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 1.5284666666666666e-05,
+      "loss": 1.1622,
+      "step": 243500
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 1.5274565656565658e-05,
+      "loss": 1.1604,
+      "step": 244000
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 1.5264464646464647e-05,
+      "loss": 1.1637,
+      "step": 244500
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 1.5254363636363639e-05,
+      "loss": 1.1612,
+      "step": 245000
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 1.5244282828282831e-05,
+      "loss": 1.1615,
+      "step": 245500
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 1.5234202020202023e-05,
+      "loss": 1.1614,
+      "step": 246000
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 1.5224101010101012e-05,
+      "loss": 1.1611,
+      "step": 246500
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 1.5214000000000002e-05,
+      "loss": 1.1606,
+      "step": 247000
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 1.520389898989899e-05,
+      "loss": 1.1629,
+      "step": 247500
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 1.5193797979797981e-05,
+      "loss": 1.1607,
+      "step": 248000
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 1.5183696969696971e-05,
+      "loss": 1.1622,
+      "step": 248500
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 1.5173595959595962e-05,
+      "loss": 1.1617,
+      "step": 249000
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 1.516349494949495e-05,
+      "loss": 1.1598,
+      "step": 249500
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 1.515339393939394e-05,
+      "loss": 1.1614,
+      "step": 250000
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.7591912982077108,
+      "eval_loss": 1.0198307037353516,
+      "eval_runtime": 50.8543,
+      "eval_samples_per_second": 893.769,
+      "eval_steps_per_second": 13.981,
+      "step": 250116
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 1.5143333333333335e-05,
+      "loss": 1.1648,
+      "step": 250500
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 1.5133252525252527e-05,
+      "loss": 1.158,
+      "step": 251000
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 1.5123151515151517e-05,
+      "loss": 1.1623,
+      "step": 251500
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 1.5113050505050506e-05,
+      "loss": 1.1612,
+      "step": 252000
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 1.5102949494949498e-05,
+      "loss": 1.1583,
+      "step": 252500
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 1.5092848484848486e-05,
+      "loss": 1.1575,
+      "step": 253000
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 1.5082747474747477e-05,
+      "loss": 1.1594,
+      "step": 253500
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 1.5072646464646465e-05,
+      "loss": 1.1587,
+      "step": 254000
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 1.5062545454545457e-05,
+      "loss": 1.1594,
+      "step": 254500
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 1.5052464646464648e-05,
+      "loss": 1.1593,
+      "step": 255000
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 1.5042363636363638e-05,
+      "loss": 1.16,
+      "step": 255500
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 1.5032262626262628e-05,
+      "loss": 1.1579,
+      "step": 256000
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 1.5022161616161617e-05,
+      "loss": 1.1572,
+      "step": 256500
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 1.5012060606060607e-05,
+      "loss": 1.1582,
+      "step": 257000
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 1.50019797979798e-05,
+      "loss": 1.1604,
+      "step": 257500
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 1.4991898989898992e-05,
+      "loss": 1.1582,
+      "step": 258000
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 1.498179797979798e-05,
+      "loss": 1.1586,
+      "step": 258500
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 1.4971696969696972e-05,
+      "loss": 1.1587,
+      "step": 259000
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 1.496159595959596e-05,
+      "loss": 1.1591,
+      "step": 259500
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 1.4951515151515153e-05,
+      "loss": 1.1577,
+      "step": 260000
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 1.4941414141414143e-05,
+      "loss": 1.1596,
+      "step": 260500
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 1.4931313131313132e-05,
+      "loss": 1.1576,
+      "step": 261000
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 1.4921212121212122e-05,
+      "loss": 1.1581,
+      "step": 261500
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 1.4911111111111113e-05,
+      "loss": 1.1581,
+      "step": 262000
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 1.4901010101010103e-05,
+      "loss": 1.1576,
+      "step": 262500
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 1.4890909090909091e-05,
+      "loss": 1.157,
+      "step": 263000
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 1.4880808080808082e-05,
+      "loss": 1.157,
+      "step": 263500
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 1.487070707070707e-05,
+      "loss": 1.1579,
+      "step": 264000
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 1.4860626262626263e-05,
+      "loss": 1.1576,
+      "step": 264500
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 1.4850525252525255e-05,
+      "loss": 1.1585,
+      "step": 265000
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 1.4840424242424243e-05,
+      "loss": 1.1594,
+      "step": 265500
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 1.4830323232323233e-05,
+      "loss": 1.1543,
+      "step": 266000
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 1.4820242424242426e-05,
+      "loss": 1.1573,
+      "step": 266500
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 1.4810141414141414e-05,
+      "loss": 1.1573,
+      "step": 267000
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 1.4800040404040406e-05,
+      "loss": 1.1572,
+      "step": 267500
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 1.4789939393939395e-05,
+      "loss": 1.1544,
+      "step": 268000
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 1.4779838383838385e-05,
+      "loss": 1.1571,
+      "step": 268500
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 1.4769737373737374e-05,
+      "loss": 1.1577,
+      "step": 269000
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 1.4759636363636366e-05,
+      "loss": 1.1554,
+      "step": 269500
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 1.4749555555555556e-05,
+      "loss": 1.1596,
+      "step": 270000
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 1.4739454545454547e-05,
+      "loss": 1.1577,
+      "step": 270500
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.7596913510542134,
+      "eval_loss": 1.0162992477416992,
+      "eval_runtime": 50.6857,
+      "eval_samples_per_second": 896.743,
+      "eval_steps_per_second": 14.028,
+      "step": 270959
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 1.4729353535353537e-05,
+      "loss": 1.1587,
+      "step": 271000
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 1.4719252525252525e-05,
+      "loss": 1.1558,
+      "step": 271500
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 1.4709171717171718e-05,
+      "loss": 1.1545,
+      "step": 272000
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 1.469909090909091e-05,
+      "loss": 1.1553,
+      "step": 272500
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 1.46889898989899e-05,
+      "loss": 1.1557,
+      "step": 273000
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 1.4678888888888889e-05,
+      "loss": 1.1554,
+      "step": 273500
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 1.466878787878788e-05,
+      "loss": 1.1562,
+      "step": 274000
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 1.465868686868687e-05,
+      "loss": 1.154,
+      "step": 274500
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 1.464858585858586e-05,
+      "loss": 1.1547,
+      "step": 275000
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 1.463848484848485e-05,
+      "loss": 1.1545,
+      "step": 275500
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 1.462838383838384e-05,
+      "loss": 1.1582,
+      "step": 276000
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 1.461828282828283e-05,
+      "loss": 1.1539,
+      "step": 276500
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 1.4608202020202023e-05,
+      "loss": 1.1569,
+      "step": 277000
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 1.4598101010101011e-05,
+      "loss": 1.1555,
+      "step": 277500
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 1.4588020202020204e-05,
+      "loss": 1.1535,
+      "step": 278000
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 1.4577919191919192e-05,
+      "loss": 1.1558,
+      "step": 278500
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 1.4567818181818182e-05,
+      "loss": 1.1535,
+      "step": 279000
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 1.4557717171717174e-05,
+      "loss": 1.1546,
+      "step": 279500
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 1.4547616161616163e-05,
+      "loss": 1.1575,
+      "step": 280000
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 1.4537515151515153e-05,
+      "loss": 1.1545,
+      "step": 280500
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 1.4527414141414142e-05,
+      "loss": 1.1564,
+      "step": 281000
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 1.4517313131313134e-05,
+      "loss": 1.154,
+      "step": 281500
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 1.4507212121212122e-05,
+      "loss": 1.1537,
+      "step": 282000
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 1.4497111111111113e-05,
+      "loss": 1.1562,
+      "step": 282500
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 1.4487010101010101e-05,
+      "loss": 1.1548,
+      "step": 283000
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 1.4476929292929294e-05,
+      "loss": 1.1555,
+      "step": 283500
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 1.4466828282828284e-05,
+      "loss": 1.1536,
+      "step": 284000
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 1.4456727272727274e-05,
+      "loss": 1.1551,
+      "step": 284500
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 1.4446626262626264e-05,
+      "loss": 1.1536,
+      "step": 285000
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 1.4436525252525253e-05,
+      "loss": 1.1543,
+      "step": 285500
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 1.4426424242424243e-05,
+      "loss": 1.1538,
+      "step": 286000
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 1.4416323232323232e-05,
+      "loss": 1.1549,
+      "step": 286500
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 1.4406222222222224e-05,
+      "loss": 1.153,
+      "step": 287000
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.4396121212121213e-05,
+      "loss": 1.1545,
+      "step": 287500
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 1.4386040404040405e-05,
+      "loss": 1.1545,
+      "step": 288000
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 1.4375939393939395e-05,
+      "loss": 1.1502,
+      "step": 288500
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.4365858585858587e-05,
+      "loss": 1.1533,
+      "step": 289000
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 1.4355757575757576e-05,
+      "loss": 1.1551,
+      "step": 289500
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 1.4345656565656568e-05,
+      "loss": 1.153,
+      "step": 290000
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 1.4335555555555556e-05,
+      "loss": 1.1507,
+      "step": 290500
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 1.4325454545454547e-05,
+      "loss": 1.1531,
+      "step": 291000
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 1.4315373737373739e-05,
+      "loss": 1.1523,
+      "step": 291500
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.7602065559563472,
+      "eval_loss": 1.0130006074905396,
+      "eval_runtime": 50.4699,
+      "eval_samples_per_second": 900.576,
+      "eval_steps_per_second": 14.088,
+      "step": 291802
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 1.4305272727272728e-05,
+      "loss": 1.1522,
+      "step": 292000
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.4295171717171718e-05,
+      "loss": 1.1504,
+      "step": 292500
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 1.4285070707070706e-05,
+      "loss": 1.1547,
+      "step": 293000
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.4274989898989899e-05,
+      "loss": 1.1501,
+      "step": 293500
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 1.426488888888889e-05,
+      "loss": 1.1509,
+      "step": 294000
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.425478787878788e-05,
+      "loss": 1.151,
+      "step": 294500
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 1.424468686868687e-05,
+      "loss": 1.1486,
+      "step": 295000
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.4234585858585858e-05,
+      "loss": 1.1523,
+      "step": 295500
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 1.422450505050505e-05,
+      "loss": 1.1507,
+      "step": 296000
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.4214404040404042e-05,
+      "loss": 1.1519,
+      "step": 296500
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 1.4204303030303031e-05,
+      "loss": 1.15,
+      "step": 297000
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 1.4194202020202021e-05,
+      "loss": 1.1494,
+      "step": 297500
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 1.4184101010101012e-05,
+      "loss": 1.1524,
+      "step": 298000
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 1.4174000000000002e-05,
+      "loss": 1.1488,
+      "step": 298500
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.4163898989898992e-05,
+      "loss": 1.1499,
+      "step": 299000
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.415379797979798e-05,
+      "loss": 1.1481,
+      "step": 299500
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.4143696969696971e-05,
+      "loss": 1.1543,
+      "step": 300000
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.4133616161616163e-05,
+      "loss": 1.1491,
+      "step": 300500
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.4123515151515152e-05,
+      "loss": 1.1507,
+      "step": 301000
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.4113414141414144e-05,
+      "loss": 1.1532,
+      "step": 301500
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.4103313131313132e-05,
+      "loss": 1.1534,
+      "step": 302000
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 1.4093212121212123e-05,
+      "loss": 1.1501,
+      "step": 302500
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 1.4083111111111111e-05,
+      "loss": 1.1501,
+      "step": 303000
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 1.4073030303030304e-05,
+      "loss": 1.1494,
+      "step": 303500
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 1.4062929292929296e-05,
+      "loss": 1.1489,
+      "step": 304000
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 1.4052828282828284e-05,
+      "loss": 1.1532,
+      "step": 304500
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.4042727272727274e-05,
+      "loss": 1.1518,
+      "step": 305000
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 1.4032626262626263e-05,
+      "loss": 1.1525,
+      "step": 305500
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 1.4022545454545455e-05,
+      "loss": 1.1511,
+      "step": 306000
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 1.4012444444444446e-05,
+      "loss": 1.1527,
+      "step": 306500
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 1.4002343434343436e-05,
+      "loss": 1.1493,
+      "step": 307000
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 1.3992242424242426e-05,
+      "loss": 1.1485,
+      "step": 307500
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 1.3982161616161618e-05,
+      "loss": 1.151,
+      "step": 308000
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 1.3972060606060607e-05,
+      "loss": 1.1474,
+      "step": 308500
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 1.3961959595959597e-05,
+      "loss": 1.1514,
+      "step": 309000
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 1.3951858585858586e-05,
+      "loss": 1.149,
+      "step": 309500
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 1.3941757575757578e-05,
+      "loss": 1.1535,
+      "step": 310000
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 1.3931656565656566e-05,
+      "loss": 1.15,
+      "step": 310500
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.3921555555555557e-05,
+      "loss": 1.1496,
+      "step": 311000
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 1.3911454545454545e-05,
+      "loss": 1.1498,
+      "step": 311500
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.3901353535353537e-05,
+      "loss": 1.1533,
+      "step": 312000
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 1.3891252525252526e-05,
+      "loss": 1.1471,
+      "step": 312500
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.7614573332336827,
+      "eval_loss": 1.0076284408569336,
+      "eval_runtime": 50.4131,
+      "eval_samples_per_second": 901.591,
+      "eval_steps_per_second": 14.103,
+      "step": 312645
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 1.3881171717171718e-05,
+      "loss": 1.1472,
+      "step": 313000
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 1.3871070707070708e-05,
+      "loss": 1.145,
+      "step": 313500
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.3860969696969697e-05,
+      "loss": 1.1484,
+      "step": 314000
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.3850868686868689e-05,
+      "loss": 1.1456,
+      "step": 314500
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.384078787878788e-05,
+      "loss": 1.1484,
+      "step": 315000
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.3830707070707072e-05,
+      "loss": 1.1462,
+      "step": 315500
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.382060606060606e-05,
+      "loss": 1.1478,
+      "step": 316000
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.3810505050505052e-05,
+      "loss": 1.1477,
+      "step": 316500
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.3800404040404041e-05,
+      "loss": 1.149,
+      "step": 317000
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.3790303030303031e-05,
+      "loss": 1.1463,
+      "step": 317500
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.378020202020202e-05,
+      "loss": 1.1458,
+      "step": 318000
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.3770101010101012e-05,
+      "loss": 1.147,
+      "step": 318500
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.3760020202020204e-05,
+      "loss": 1.1452,
+      "step": 319000
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.3749919191919193e-05,
+      "loss": 1.1471,
+      "step": 319500
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.3739818181818183e-05,
+      "loss": 1.1465,
+      "step": 320000
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.3729717171717172e-05,
+      "loss": 1.1492,
+      "step": 320500
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.3719616161616164e-05,
+      "loss": 1.1501,
+      "step": 321000
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.3709515151515154e-05,
+      "loss": 1.1458,
+      "step": 321500
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.3699414141414142e-05,
+      "loss": 1.1472,
+      "step": 322000
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.3689313131313133e-05,
+      "loss": 1.1445,
+      "step": 322500
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.3679232323232325e-05,
+      "loss": 1.1494,
+      "step": 323000
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.3669131313131314e-05,
+      "loss": 1.1477,
+      "step": 323500
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.3659030303030305e-05,
+      "loss": 1.1499,
+      "step": 324000
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.3648929292929294e-05,
+      "loss": 1.1474,
+      "step": 324500
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.3638848484848486e-05,
+      "loss": 1.1485,
+      "step": 325000
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.3628747474747477e-05,
+      "loss": 1.1458,
+      "step": 325500
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.3618646464646465e-05,
+      "loss": 1.1464,
+      "step": 326000
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.3608545454545457e-05,
+      "loss": 1.1459,
+      "step": 326500
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.3598444444444446e-05,
+      "loss": 1.1483,
+      "step": 327000
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.3588363636363638e-05,
+      "loss": 1.147,
+      "step": 327500
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.3578262626262628e-05,
+      "loss": 1.1456,
+      "step": 328000
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.3568161616161617e-05,
+      "loss": 1.1455,
+      "step": 328500
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.3558060606060607e-05,
+      "loss": 1.1478,
+      "step": 329000
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.3547959595959597e-05,
+      "loss": 1.1479,
+      "step": 329500
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.3537858585858588e-05,
+      "loss": 1.1472,
+      "step": 330000
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.3527757575757576e-05,
+      "loss": 1.1446,
+      "step": 330500
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.3517676767676769e-05,
+      "loss": 1.1453,
+      "step": 331000
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.3507575757575759e-05,
+      "loss": 1.148,
+      "step": 331500
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.3497474747474747e-05,
+      "loss": 1.15,
+      "step": 332000
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.348739393939394e-05,
+      "loss": 1.1443,
+      "step": 332500
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.3477292929292932e-05,
+      "loss": 1.1474,
+      "step": 333000
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.761879215337247,
+      "eval_loss": 1.0039067268371582,
+      "eval_runtime": 50.4617,
+      "eval_samples_per_second": 900.722,
+      "eval_steps_per_second": 14.09,
+      "step": 333488
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.346719191919192e-05,
+      "loss": 1.1441,
+      "step": 333500
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 1.345709090909091e-05,
+      "loss": 1.1461,
+      "step": 334000
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 1.34469898989899e-05,
+      "loss": 1.1447,
+      "step": 334500
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 1.3436888888888891e-05,
+      "loss": 1.1446,
+      "step": 335000
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 1.342678787878788e-05,
+      "loss": 1.1448,
+      "step": 335500
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 1.341668686868687e-05,
+      "loss": 1.1457,
+      "step": 336000
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 1.3406585858585859e-05,
+      "loss": 1.1453,
+      "step": 336500
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 1.3396505050505051e-05,
+      "loss": 1.1438,
+      "step": 337000
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 1.3386404040404041e-05,
+      "loss": 1.1437,
+      "step": 337500
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 1.337630303030303e-05,
+      "loss": 1.1448,
+      "step": 338000
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 1.3366202020202022e-05,
+      "loss": 1.1443,
+      "step": 338500
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 1.335610101010101e-05,
+      "loss": 1.1461,
+      "step": 339000
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 1.3346020202020203e-05,
+      "loss": 1.1475,
+      "step": 339500
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 1.3335919191919193e-05,
+      "loss": 1.147,
+      "step": 340000
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 1.3325818181818181e-05,
+      "loss": 1.1446,
+      "step": 340500
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 1.3315717171717173e-05,
+      "loss": 1.1428,
+      "step": 341000
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 1.3305636363636366e-05,
+      "loss": 1.143,
+      "step": 341500
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 1.3295535353535354e-05,
+      "loss": 1.1435,
+      "step": 342000
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 1.3285434343434345e-05,
+      "loss": 1.1439,
+      "step": 342500
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 1.3275333333333333e-05,
+      "loss": 1.1459,
+      "step": 343000
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 1.3265232323232325e-05,
+      "loss": 1.1455,
+      "step": 343500
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 1.3255131313131315e-05,
+      "loss": 1.1454,
+      "step": 344000
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 1.3245050505050506e-05,
+      "loss": 1.1423,
+      "step": 344500
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 1.3234949494949496e-05,
+      "loss": 1.1432,
+      "step": 345000
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 1.3224848484848485e-05,
+      "loss": 1.1431,
+      "step": 345500
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 1.3214747474747475e-05,
+      "loss": 1.1465,
+      "step": 346000
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 1.3204646464646467e-05,
+      "loss": 1.1439,
+      "step": 346500
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 1.3194545454545456e-05,
+      "loss": 1.144,
+      "step": 347000
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 1.3184444444444446e-05,
+      "loss": 1.1422,
+      "step": 347500
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 1.3174343434343435e-05,
+      "loss": 1.1453,
+      "step": 348000
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 1.3164242424242427e-05,
+      "loss": 1.145,
+      "step": 348500
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 1.3154161616161619e-05,
+      "loss": 1.146,
+      "step": 349000
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 1.3144060606060607e-05,
+      "loss": 1.145,
+      "step": 349500
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 1.3133959595959598e-05,
+      "loss": 1.1425,
+      "step": 350000
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 1.3123858585858586e-05,
+      "loss": 1.1453,
+      "step": 350500
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 1.3113777777777779e-05,
+      "loss": 1.1421,
+      "step": 351000
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 1.310369696969697e-05,
+      "loss": 1.1443,
+      "step": 351500
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 1.3093595959595961e-05,
+      "loss": 1.1407,
+      "step": 352000
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 1.308349494949495e-05,
+      "loss": 1.1431,
+      "step": 352500
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 1.3073393939393942e-05,
+      "loss": 1.1454,
+      "step": 353000
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 1.306329292929293e-05,
+      "loss": 1.1405,
+      "step": 353500
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 1.3053212121212123e-05,
+      "loss": 1.1415,
+      "step": 354000
+    },
+    {
+      "epoch": 17.0,
+      "eval_accuracy": 0.7617649545035797,
+      "eval_loss": 1.0038964748382568,
+      "eval_runtime": 50.6319,
+      "eval_samples_per_second": 897.695,
+      "eval_steps_per_second": 14.043,
+      "step": 354331
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 1.3043111111111113e-05,
+      "loss": 1.1429,
+      "step": 354500
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 1.3033010101010101e-05,
+      "loss": 1.1435,
+      "step": 355000
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 1.3022909090909093e-05,
+      "loss": 1.1408,
+      "step": 355500
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 1.3012828282828286e-05,
+      "loss": 1.1411,
+      "step": 356000
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 1.3002727272727274e-05,
+      "loss": 1.1442,
+      "step": 356500
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 1.2992626262626264e-05,
+      "loss": 1.1442,
+      "step": 357000
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 1.2982525252525253e-05,
+      "loss": 1.1435,
+      "step": 357500
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 1.2972424242424243e-05,
+      "loss": 1.1419,
+      "step": 358000
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 1.2962323232323234e-05,
+      "loss": 1.1435,
+      "step": 358500
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 1.2952242424242424e-05,
+      "loss": 1.1393,
+      "step": 359000
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 1.2942141414141416e-05,
+      "loss": 1.1433,
+      "step": 359500
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 1.2932040404040405e-05,
+      "loss": 1.1391,
+      "step": 360000
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 1.2921939393939395e-05,
+      "loss": 1.1416,
+      "step": 360500
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 1.2911838383838384e-05,
+      "loss": 1.1407,
+      "step": 361000
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 1.2901757575757576e-05,
+      "loss": 1.1417,
+      "step": 361500
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 1.2891656565656568e-05,
+      "loss": 1.1411,
+      "step": 362000
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 1.2881555555555556e-05,
+      "loss": 1.1413,
+      "step": 362500
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 1.2871474747474749e-05,
+      "loss": 1.1435,
+      "step": 363000
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 1.2861373737373739e-05,
+      "loss": 1.1414,
+      "step": 363500
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 1.2851272727272728e-05,
+      "loss": 1.1403,
+      "step": 364000
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 1.2841171717171718e-05,
+      "loss": 1.1387,
+      "step": 364500
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 1.2831070707070708e-05,
+      "loss": 1.1408,
+      "step": 365000
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 1.2820969696969698e-05,
+      "loss": 1.1396,
+      "step": 365500
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 1.2810868686868687e-05,
+      "loss": 1.1402,
+      "step": 366000
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 1.2800767676767677e-05,
+      "loss": 1.1399,
+      "step": 366500
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 1.2790666666666666e-05,
+      "loss": 1.1406,
+      "step": 367000
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 1.2780565656565658e-05,
+      "loss": 1.1395,
+      "step": 367500
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 1.277048484848485e-05,
+      "loss": 1.1428,
+      "step": 368000
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 1.2760383838383839e-05,
+      "loss": 1.1401,
+      "step": 368500
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 1.2750282828282829e-05,
+      "loss": 1.1417,
+      "step": 369000
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 1.2740181818181818e-05,
+      "loss": 1.1395,
+      "step": 369500
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 1.273008080808081e-05,
+      "loss": 1.1379,
+      "step": 370000
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 1.27199797979798e-05,
+      "loss": 1.1402,
+      "step": 370500
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 1.270989898989899e-05,
+      "loss": 1.1397,
+      "step": 371000
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 1.269979797979798e-05,
+      "loss": 1.1414,
+      "step": 371500
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 1.268969696969697e-05,
+      "loss": 1.141,
+      "step": 372000
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 1.2679595959595961e-05,
+      "loss": 1.1388,
+      "step": 372500
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 1.2669494949494952e-05,
+      "loss": 1.1391,
+      "step": 373000
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 1.265939393939394e-05,
+      "loss": 1.142,
+      "step": 373500
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 1.264929292929293e-05,
+      "loss": 1.139,
+      "step": 374000
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 1.2639191919191919e-05,
+      "loss": 1.1408,
+      "step": 374500
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 1.2629131313131313e-05,
+      "loss": 1.139,
+      "step": 375000
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.7623969718913042,
+      "eval_loss": 1.0014969110488892,
+      "eval_runtime": 50.2795,
+      "eval_samples_per_second": 903.986,
+      "eval_steps_per_second": 14.141,
+      "step": 375174
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 1.2619030303030304e-05,
+      "loss": 1.1401,
+      "step": 375500
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 1.2608929292929296e-05,
+      "loss": 1.1362,
+      "step": 376000
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 1.2598828282828284e-05,
+      "loss": 1.1402,
+      "step": 376500
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 1.2588727272727274e-05,
+      "loss": 1.1382,
+      "step": 377000
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 1.2578646464646467e-05,
+      "loss": 1.1399,
+      "step": 377500
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 1.2568545454545455e-05,
+      "loss": 1.1382,
+      "step": 378000
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 1.2558444444444447e-05,
+      "loss": 1.1388,
+      "step": 378500
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 1.2548363636363636e-05,
+      "loss": 1.1386,
+      "step": 379000
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 1.2538262626262626e-05,
+      "loss": 1.1364,
+      "step": 379500
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 1.2528161616161618e-05,
+      "loss": 1.1382,
+      "step": 380000
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 1.2518060606060607e-05,
+      "loss": 1.14,
+      "step": 380500
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 1.2507959595959597e-05,
+      "loss": 1.1371,
+      "step": 381000
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 1.2497858585858586e-05,
+      "loss": 1.1417,
+      "step": 381500
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 1.2487757575757578e-05,
+      "loss": 1.1382,
+      "step": 382000
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 1.247767676767677e-05,
+      "loss": 1.1384,
+      "step": 382500
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 1.2467575757575759e-05,
+      "loss": 1.1379,
+      "step": 383000
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 1.2457474747474749e-05,
+      "loss": 1.1384,
+      "step": 383500
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 1.2447373737373738e-05,
+      "loss": 1.1393,
+      "step": 384000
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 1.243727272727273e-05,
+      "loss": 1.1379,
+      "step": 384500
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 1.2427171717171718e-05,
+      "loss": 1.1393,
+      "step": 385000
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 1.2417070707070708e-05,
+      "loss": 1.1382,
+      "step": 385500
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 1.2406969696969697e-05,
+      "loss": 1.1347,
+      "step": 386000
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 1.2396868686868689e-05,
+      "loss": 1.1375,
+      "step": 386500
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 1.2386767676767678e-05,
+      "loss": 1.141,
+      "step": 387000
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 1.2376666666666668e-05,
+      "loss": 1.1371,
+      "step": 387500
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 1.236658585858586e-05,
+      "loss": 1.1403,
+      "step": 388000
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 1.2356484848484849e-05,
+      "loss": 1.1377,
+      "step": 388500
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 1.2346383838383839e-05,
+      "loss": 1.1361,
+      "step": 389000
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 1.2336282828282828e-05,
+      "loss": 1.1378,
+      "step": 389500
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 1.232618181818182e-05,
+      "loss": 1.1357,
+      "step": 390000
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.2316080808080808e-05,
+      "loss": 1.1388,
+      "step": 390500
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.2305979797979798e-05,
+      "loss": 1.1367,
+      "step": 391000
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.229589898989899e-05,
+      "loss": 1.1384,
+      "step": 391500
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.228579797979798e-05,
+      "loss": 1.1372,
+      "step": 392000
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.2275696969696971e-05,
+      "loss": 1.1391,
+      "step": 392500
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.2265616161616164e-05,
+      "loss": 1.1356,
+      "step": 393000
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.2255515151515152e-05,
+      "loss": 1.1363,
+      "step": 393500
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.2245414141414142e-05,
+      "loss": 1.1369,
+      "step": 394000
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.2235313131313131e-05,
+      "loss": 1.1382,
+      "step": 394500
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.2225212121212123e-05,
+      "loss": 1.138,
+      "step": 395000
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.2215111111111113e-05,
+      "loss": 1.1377,
+      "step": 395500
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.2205010101010102e-05,
+      "loss": 1.1371,
+      "step": 396000
+    },
+    {
+      "epoch": 19.0,
+      "eval_accuracy": 0.7630963198969265,
+      "eval_loss": 0.9979222416877747,
+      "eval_runtime": 50.6814,
+      "eval_samples_per_second": 896.818,
+      "eval_steps_per_second": 14.029,
+      "step": 396017
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.2194909090909092e-05,
+      "loss": 1.1386,
+      "step": 396500
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.218480808080808e-05,
+      "loss": 1.1363,
+      "step": 397000
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.2174727272727273e-05,
+      "loss": 1.1373,
+      "step": 397500
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.2164626262626265e-05,
+      "loss": 1.1369,
+      "step": 398000
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.2154525252525254e-05,
+      "loss": 1.1357,
+      "step": 398500
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.2144424242424244e-05,
+      "loss": 1.1374,
+      "step": 399000
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.2134323232323232e-05,
+      "loss": 1.1347,
+      "step": 399500
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.2124222222222224e-05,
+      "loss": 1.1358,
+      "step": 400000
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.2114121212121213e-05,
+      "loss": 1.1372,
+      "step": 400500
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.2104020202020203e-05,
+      "loss": 1.1379,
+      "step": 401000
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.2093939393939396e-05,
+      "loss": 1.1346,
+      "step": 401500
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.2083838383838384e-05,
+      "loss": 1.1337,
+      "step": 402000
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.2073737373737376e-05,
+      "loss": 1.1353,
+      "step": 402500
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.2063636363636365e-05,
+      "loss": 1.1355,
+      "step": 403000
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.2053555555555557e-05,
+      "loss": 1.136,
+      "step": 403500
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.2043454545454547e-05,
+      "loss": 1.1348,
+      "step": 404000
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.2033353535353536e-05,
+      "loss": 1.1352,
+      "step": 404500
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.2023252525252526e-05,
+      "loss": 1.1372,
+      "step": 405000
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.2013151515151515e-05,
+      "loss": 1.1365,
+      "step": 405500
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.2003070707070707e-05,
+      "loss": 1.1342,
+      "step": 406000
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.1992969696969699e-05,
+      "loss": 1.1345,
+      "step": 406500
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.1982868686868688e-05,
+      "loss": 1.1359,
+      "step": 407000
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.197278787878788e-05,
+      "loss": 1.1357,
+      "step": 407500
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.196268686868687e-05,
+      "loss": 1.135,
+      "step": 408000
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.1952585858585859e-05,
+      "loss": 1.1355,
+      "step": 408500
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 1.194248484848485e-05,
+      "loss": 1.1349,
+      "step": 409000
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 1.193238383838384e-05,
+      "loss": 1.1336,
+      "step": 409500
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 1.192228282828283e-05,
+      "loss": 1.1343,
+      "step": 410000
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 1.1912181818181818e-05,
+      "loss": 1.1382,
+      "step": 410500
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 1.1902080808080808e-05,
+      "loss": 1.1363,
+      "step": 411000
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 1.1892e-05,
+      "loss": 1.1349,
+      "step": 411500
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 1.188189898989899e-05,
+      "loss": 1.1329,
+      "step": 412000
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 1.1871797979797981e-05,
+      "loss": 1.1348,
+      "step": 412500
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 1.186169696969697e-05,
+      "loss": 1.1344,
+      "step": 413000
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 1.1851636363636366e-05,
+      "loss": 1.1328,
+      "step": 413500
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 1.1841535353535354e-05,
+      "loss": 1.1352,
+      "step": 414000
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.1831434343434345e-05,
+      "loss": 1.1336,
+      "step": 414500
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.1821333333333333e-05,
+      "loss": 1.1372,
+      "step": 415000
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.1811252525252525e-05,
+      "loss": 1.136,
+      "step": 415500
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.1801151515151516e-05,
+      "loss": 1.1337,
+      "step": 416000
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 1.1791050505050506e-05,
+      "loss": 1.1357,
+      "step": 416500
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": 0.7629886565345744,
+      "eval_loss": 0.9972801804542542,
+      "eval_runtime": 50.4083,
+      "eval_samples_per_second": 901.678,
+      "eval_steps_per_second": 14.105,
+      "step": 416860
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 1.1780949494949496e-05,
+      "loss": 1.1351,
+      "step": 417000
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 1.1770848484848485e-05,
+      "loss": 1.1349,
+      "step": 417500
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 1.1760747474747475e-05,
+      "loss": 1.1341,
+      "step": 418000
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 1.1750646464646465e-05,
+      "loss": 1.1356,
+      "step": 418500
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 1.1740545454545456e-05,
+      "loss": 1.1349,
+      "step": 419000
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 1.1730444444444444e-05,
+      "loss": 1.1306,
+      "step": 419500
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 1.1720363636363637e-05,
+      "loss": 1.1311,
+      "step": 420000
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 1.1710262626262627e-05,
+      "loss": 1.132,
+      "step": 420500
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.1700161616161615e-05,
+      "loss": 1.1317,
+      "step": 421000
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 1.1690060606060607e-05,
+      "loss": 1.1321,
+      "step": 421500
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 1.1679959595959598e-05,
+      "loss": 1.1329,
+      "step": 422000
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 1.1669858585858586e-05,
+      "loss": 1.1328,
+      "step": 422500
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 1.1659777777777779e-05,
+      "loss": 1.1335,
+      "step": 423000
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.1649676767676769e-05,
+      "loss": 1.1362,
+      "step": 423500
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.163957575757576e-05,
+      "loss": 1.1325,
+      "step": 424000
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 1.162947474747475e-05,
+      "loss": 1.135,
+      "step": 424500
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 1.1619373737373738e-05,
+      "loss": 1.1337,
+      "step": 425000
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 1.160929292929293e-05,
+      "loss": 1.1341,
+      "step": 425500
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 1.159919191919192e-05,
+      "loss": 1.1316,
+      "step": 426000
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 1.158909090909091e-05,
+      "loss": 1.1319,
+      "step": 426500
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 1.1578989898989901e-05,
+      "loss": 1.1326,
+      "step": 427000
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 1.156888888888889e-05,
+      "loss": 1.1345,
+      "step": 427500
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 1.1558828282828282e-05,
+      "loss": 1.1305,
+      "step": 428000
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 1.1548727272727274e-05,
+      "loss": 1.1328,
+      "step": 428500
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 1.1538626262626265e-05,
+      "loss": 1.1336,
+      "step": 429000
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 1.1528525252525253e-05,
+      "loss": 1.1319,
+      "step": 429500
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 1.1518424242424245e-05,
+      "loss": 1.1313,
+      "step": 430000
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 1.1508343434343436e-05,
+      "loss": 1.1332,
+      "step": 430500
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.1498242424242424e-05,
+      "loss": 1.1353,
+      "step": 431000
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 1.1488141414141416e-05,
+      "loss": 1.1342,
+      "step": 431500
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 1.1478040404040405e-05,
+      "loss": 1.132,
+      "step": 432000
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 1.1467939393939395e-05,
+      "loss": 1.1337,
+      "step": 432500
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 1.1457858585858587e-05,
+      "loss": 1.135,
+      "step": 433000
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.1447757575757576e-05,
+      "loss": 1.1322,
+      "step": 433500
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.1437656565656568e-05,
+      "loss": 1.1314,
+      "step": 434000
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 1.1427555555555557e-05,
+      "loss": 1.1317,
+      "step": 434500
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 1.1417454545454547e-05,
+      "loss": 1.1324,
+      "step": 435000
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.1407353535353535e-05,
+      "loss": 1.1317,
+      "step": 435500
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 1.1397252525252527e-05,
+      "loss": 1.1325,
+      "step": 436000
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 1.1387151515151516e-05,
+      "loss": 1.1314,
+      "step": 436500
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 1.1377050505050506e-05,
+      "loss": 1.1317,
+      "step": 437000
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 1.13669898989899e-05,
+      "loss": 1.1305,
+      "step": 437500
+    },
+    {
+      "epoch": 21.0,
+      "eval_accuracy": 0.7636198673729573,
+      "eval_loss": 0.994223415851593,
+      "eval_runtime": 49.999,
+      "eval_samples_per_second": 909.058,
+      "eval_steps_per_second": 14.22,
+      "step": 437703
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 1.135688888888889e-05,
+      "loss": 1.1328,
+      "step": 438000
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 1.134678787878788e-05,
+      "loss": 1.1297,
+      "step": 438500
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 1.133668686868687e-05,
+      "loss": 1.1278,
+      "step": 439000
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.1326585858585858e-05,
+      "loss": 1.1305,
+      "step": 439500
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.131648484848485e-05,
+      "loss": 1.1299,
+      "step": 440000
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 1.1306404040404042e-05,
+      "loss": 1.131,
+      "step": 440500
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.1296303030303031e-05,
+      "loss": 1.1278,
+      "step": 441000
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.1286202020202021e-05,
+      "loss": 1.1305,
+      "step": 441500
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 1.127610101010101e-05,
+      "loss": 1.1303,
+      "step": 442000
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.1266000000000002e-05,
+      "loss": 1.1308,
+      "step": 442500
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.125589898989899e-05,
+      "loss": 1.1307,
+      "step": 443000
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 1.124579797979798e-05,
+      "loss": 1.1315,
+      "step": 443500
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 1.123569696969697e-05,
+      "loss": 1.1315,
+      "step": 444000
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 1.1225616161616162e-05,
+      "loss": 1.1309,
+      "step": 444500
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 1.1215515151515154e-05,
+      "loss": 1.1312,
+      "step": 445000
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 1.1205414141414142e-05,
+      "loss": 1.1316,
+      "step": 445500
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 1.1195313131313132e-05,
+      "loss": 1.1316,
+      "step": 446000
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 1.1185212121212121e-05,
+      "loss": 1.1296,
+      "step": 446500
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.1175111111111111e-05,
+      "loss": 1.1304,
+      "step": 447000
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 1.1165010101010102e-05,
+      "loss": 1.1337,
+      "step": 447500
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 1.1154909090909092e-05,
+      "loss": 1.1293,
+      "step": 448000
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 1.1144808080808082e-05,
+      "loss": 1.1323,
+      "step": 448500
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 1.1134727272727273e-05,
+      "loss": 1.1307,
+      "step": 449000
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 1.1124646464646465e-05,
+      "loss": 1.1293,
+      "step": 449500
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 1.1114545454545455e-05,
+      "loss": 1.132,
+      "step": 450000
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.1104444444444444e-05,
+      "loss": 1.1285,
+      "step": 450500
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 1.1094343434343436e-05,
+      "loss": 1.1291,
+      "step": 451000
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 1.1084262626262628e-05,
+      "loss": 1.1314,
+      "step": 451500
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 1.1074161616161617e-05,
+      "loss": 1.1296,
+      "step": 452000
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 1.1064060606060607e-05,
+      "loss": 1.129,
+      "step": 452500
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 1.1053959595959596e-05,
+      "loss": 1.1296,
+      "step": 453000
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 1.1043858585858586e-05,
+      "loss": 1.1293,
+      "step": 453500
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 1.1033757575757578e-05,
+      "loss": 1.129,
+      "step": 454000
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 1.1023656565656566e-05,
+      "loss": 1.1315,
+      "step": 454500
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 1.1013555555555557e-05,
+      "loss": 1.1281,
+      "step": 455000
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 1.1003474747474749e-05,
+      "loss": 1.1337,
+      "step": 455500
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 1.0993373737373738e-05,
+      "loss": 1.1291,
+      "step": 456000
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 1.098327272727273e-05,
+      "loss": 1.1281,
+      "step": 456500
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.0973171717171718e-05,
+      "loss": 1.1283,
+      "step": 457000
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 1.096309090909091e-05,
+      "loss": 1.1294,
+      "step": 457500
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 1.09529898989899e-05,
+      "loss": 1.1255,
+      "step": 458000
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.094288888888889e-05,
+      "loss": 1.1277,
+      "step": 458500
+    },
+    {
+      "epoch": 22.0,
+      "eval_accuracy": 0.7642678862352545,
+      "eval_loss": 0.9911049008369446,
+      "eval_runtime": 50.5302,
+      "eval_samples_per_second": 899.502,
+      "eval_steps_per_second": 14.071,
+      "step": 458546
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 1.0932808080808082e-05,
+      "loss": 1.1285,
+      "step": 459000
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.0922707070707072e-05,
+      "loss": 1.1273,
+      "step": 459500
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.0912606060606062e-05,
+      "loss": 1.1273,
+      "step": 460000
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.0902505050505052e-05,
+      "loss": 1.1276,
+      "step": 460500
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 1.0892404040404041e-05,
+      "loss": 1.1321,
+      "step": 461000
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 1.0882303030303031e-05,
+      "loss": 1.1277,
+      "step": 461500
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 1.087220202020202e-05,
+      "loss": 1.1294,
+      "step": 462000
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 1.0862101010101012e-05,
+      "loss": 1.1291,
+      "step": 462500
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.0852e-05,
+      "loss": 1.1289,
+      "step": 463000
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 1.0841939393939396e-05,
+      "loss": 1.1293,
+      "step": 463500
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 1.0831838383838385e-05,
+      "loss": 1.1276,
+      "step": 464000
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 1.0821737373737375e-05,
+      "loss": 1.1296,
+      "step": 464500
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 1.0811636363636364e-05,
+      "loss": 1.1297,
+      "step": 465000
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 1.0801535353535356e-05,
+      "loss": 1.1304,
+      "step": 465500
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 1.0791434343434344e-05,
+      "loss": 1.1271,
+      "step": 466000
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 1.0781333333333335e-05,
+      "loss": 1.1257,
+      "step": 466500
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.0771252525252527e-05,
+      "loss": 1.1273,
+      "step": 467000
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.0761151515151516e-05,
+      "loss": 1.1295,
+      "step": 467500
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.0751070707070708e-05,
+      "loss": 1.1282,
+      "step": 468000
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 1.0740969696969698e-05,
+      "loss": 1.1273,
+      "step": 468500
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.0730868686868687e-05,
+      "loss": 1.1278,
+      "step": 469000
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 1.0720767676767679e-05,
+      "loss": 1.1285,
+      "step": 469500
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 1.0710666666666667e-05,
+      "loss": 1.1288,
+      "step": 470000
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 1.0700565656565658e-05,
+      "loss": 1.1287,
+      "step": 470500
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 1.0690464646464646e-05,
+      "loss": 1.1276,
+      "step": 471000
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 1.0680363636363638e-05,
+      "loss": 1.1275,
+      "step": 471500
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 1.0670262626262627e-05,
+      "loss": 1.127,
+      "step": 472000
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 1.0660181818181819e-05,
+      "loss": 1.1257,
+      "step": 472500
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 1.065008080808081e-05,
+      "loss": 1.1278,
+      "step": 473000
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 1.0639979797979798e-05,
+      "loss": 1.1285,
+      "step": 473500
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 1.062987878787879e-05,
+      "loss": 1.1288,
+      "step": 474000
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 1.0619777777777778e-05,
+      "loss": 1.1253,
+      "step": 474500
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 1.060969696969697e-05,
+      "loss": 1.1257,
+      "step": 475000
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 1.0599595959595961e-05,
+      "loss": 1.1284,
+      "step": 475500
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 1.058949494949495e-05,
+      "loss": 1.1275,
+      "step": 476000
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 1.057939393939394e-05,
+      "loss": 1.1255,
+      "step": 476500
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 1.0569292929292928e-05,
+      "loss": 1.1287,
+      "step": 477000
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 1.055919191919192e-05,
+      "loss": 1.1284,
+      "step": 477500
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 1.054909090909091e-05,
+      "loss": 1.1266,
+      "step": 478000
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 1.0539010101010101e-05,
+      "loss": 1.124,
+      "step": 478500
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 1.0528909090909091e-05,
+      "loss": 1.1306,
+      "step": 479000
+    },
+    {
+      "epoch": 23.0,
+      "eval_accuracy": 0.764223618849601,
+      "eval_loss": 0.991273045539856,
+      "eval_runtime": 50.2177,
+      "eval_samples_per_second": 905.099,
+      "eval_steps_per_second": 14.158,
+      "step": 479389
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 1.051880808080808e-05,
+      "loss": 1.1279,
+      "step": 479500
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 1.0508707070707072e-05,
+      "loss": 1.1263,
+      "step": 480000
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 1.0498606060606062e-05,
+      "loss": 1.1252,
+      "step": 480500
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 1.0488505050505051e-05,
+      "loss": 1.126,
+      "step": 481000
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 1.0478404040404043e-05,
+      "loss": 1.1268,
+      "step": 481500
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 1.0468323232323233e-05,
+      "loss": 1.1265,
+      "step": 482000
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 1.0458222222222224e-05,
+      "loss": 1.1257,
+      "step": 482500
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 1.0448121212121214e-05,
+      "loss": 1.1272,
+      "step": 483000
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 1.0438020202020203e-05,
+      "loss": 1.1275,
+      "step": 483500
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 1.0427919191919193e-05,
+      "loss": 1.1257,
+      "step": 484000
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 1.0417838383838385e-05,
+      "loss": 1.1277,
+      "step": 484500
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 1.0407737373737374e-05,
+      "loss": 1.1251,
+      "step": 485000
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 1.0397636363636366e-05,
+      "loss": 1.1245,
+      "step": 485500
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 1.0387535353535354e-05,
+      "loss": 1.1266,
+      "step": 486000
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 1.0377434343434345e-05,
+      "loss": 1.1257,
+      "step": 486500
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 1.0367333333333333e-05,
+      "loss": 1.1262,
+      "step": 487000
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 1.0357232323232325e-05,
+      "loss": 1.1277,
+      "step": 487500
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 1.0347131313131314e-05,
+      "loss": 1.1243,
+      "step": 488000
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 1.0337050505050506e-05,
+      "loss": 1.1238,
+      "step": 488500
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 1.0326949494949496e-05,
+      "loss": 1.1258,
+      "step": 489000
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 1.0316868686868689e-05,
+      "loss": 1.1268,
+      "step": 489500
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 1.0306767676767677e-05,
+      "loss": 1.1238,
+      "step": 490000
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 1.0296666666666667e-05,
+      "loss": 1.1266,
+      "step": 490500
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 1.0286565656565656e-05,
+      "loss": 1.1267,
+      "step": 491000
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 1.0276484848484848e-05,
+      "loss": 1.1247,
+      "step": 491500
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 1.026638383838384e-05,
+      "loss": 1.1245,
+      "step": 492000
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 1.0256282828282829e-05,
+      "loss": 1.1259,
+      "step": 492500
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 1.024618181818182e-05,
+      "loss": 1.1258,
+      "step": 493000
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 1.0236080808080808e-05,
+      "loss": 1.1265,
+      "step": 493500
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 1.02259797979798e-05,
+      "loss": 1.1263,
+      "step": 494000
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 1.0215878787878788e-05,
+      "loss": 1.1274,
+      "step": 494500
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 1.0205777777777779e-05,
+      "loss": 1.1246,
+      "step": 495000
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 1.0195676767676767e-05,
+      "loss": 1.1264,
+      "step": 495500
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 1.018557575757576e-05,
+      "loss": 1.1256,
+      "step": 496000
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 1.0175474747474748e-05,
+      "loss": 1.1258,
+      "step": 496500
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 1.0165373737373738e-05,
+      "loss": 1.1273,
+      "step": 497000
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 1.0155272727272728e-05,
+      "loss": 1.128,
+      "step": 497500
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 1.0145191919191919e-05,
+      "loss": 1.1238,
+      "step": 498000
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 1.0135111111111111e-05,
+      "loss": 1.1257,
+      "step": 498500
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 1.0125010101010101e-05,
+      "loss": 1.1247,
+      "step": 499000
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 1.011490909090909e-05,
+      "loss": 1.1258,
+      "step": 499500
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 1.0104808080808082e-05,
+      "loss": 1.1258,
+      "step": 500000
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy": 0.7651094778246644,
+      "eval_loss": 0.9861079454421997,
+      "eval_runtime": 50.6536,
+      "eval_samples_per_second": 897.31,
+      "eval_steps_per_second": 14.037,
+      "step": 500232
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 1.0094727272727274e-05,
+      "loss": 1.1272,
+      "step": 500500
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 1.0084626262626263e-05,
+      "loss": 1.1242,
+      "step": 501000
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 1.0074525252525253e-05,
+      "loss": 1.1241,
+      "step": 501500
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 1.0064424242424242e-05,
+      "loss": 1.1238,
+      "step": 502000
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 1.0054323232323234e-05,
+      "loss": 1.1215,
+      "step": 502500
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 1.0044222222222224e-05,
+      "loss": 1.1251,
+      "step": 503000
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 1.0034141414141415e-05,
+      "loss": 1.1224,
+      "step": 503500
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 1.0024040404040405e-05,
+      "loss": 1.1255,
+      "step": 504000
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 1.0013939393939395e-05,
+      "loss": 1.1222,
+      "step": 504500
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 1.0003838383838384e-05,
+      "loss": 1.1217,
+      "step": 505000
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 9.993757575757576e-06,
+      "loss": 1.1221,
+      "step": 505500
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 9.983656565656566e-06,
+      "loss": 1.1253,
+      "step": 506000
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 9.973555555555557e-06,
+      "loss": 1.1237,
+      "step": 506500
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 9.963454545454547e-06,
+      "loss": 1.1257,
+      "step": 507000
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 9.953353535353535e-06,
+      "loss": 1.1233,
+      "step": 507500
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 9.943252525252526e-06,
+      "loss": 1.1245,
+      "step": 508000
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 9.933171717171718e-06,
+      "loss": 1.1232,
+      "step": 508500
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 9.923070707070708e-06,
+      "loss": 1.1212,
+      "step": 509000
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 9.912969696969697e-06,
+      "loss": 1.1238,
+      "step": 509500
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 9.902868686868687e-06,
+      "loss": 1.1216,
+      "step": 510000
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 9.89278787878788e-06,
+      "loss": 1.1244,
+      "step": 510500
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 9.88268686868687e-06,
+      "loss": 1.1225,
+      "step": 511000
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 9.87258585858586e-06,
+      "loss": 1.1221,
+      "step": 511500
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 9.862484848484849e-06,
+      "loss": 1.1258,
+      "step": 512000
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 9.852383838383839e-06,
+      "loss": 1.1232,
+      "step": 512500
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 9.842282828282829e-06,
+      "loss": 1.1215,
+      "step": 513000
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 9.832181818181818e-06,
+      "loss": 1.1256,
+      "step": 513500
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 9.82210101010101e-06,
+      "loss": 1.1224,
+      "step": 514000
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 9.812e-06,
+      "loss": 1.1241,
+      "step": 514500
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 9.80189898989899e-06,
+      "loss": 1.1261,
+      "step": 515000
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 9.79179797979798e-06,
+      "loss": 1.1204,
+      "step": 515500
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 9.781696969696971e-06,
+      "loss": 1.1248,
+      "step": 516000
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 9.771595959595961e-06,
+      "loss": 1.1237,
+      "step": 516500
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 9.76149494949495e-06,
+      "loss": 1.1229,
+      "step": 517000
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 9.75139393939394e-06,
+      "loss": 1.1219,
+      "step": 517500
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 9.741313131313133e-06,
+      "loss": 1.124,
+      "step": 518000
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 9.731232323232323e-06,
+      "loss": 1.125,
+      "step": 518500
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 9.721131313131313e-06,
+      "loss": 1.1222,
+      "step": 519000
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 9.711050505050506e-06,
+      "loss": 1.1247,
+      "step": 519500
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 9.700949494949496e-06,
+      "loss": 1.1243,
+      "step": 520000
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 9.690848484848484e-06,
+      "loss": 1.1242,
+      "step": 520500
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 9.680747474747475e-06,
+      "loss": 1.1225,
+      "step": 521000
+    },
+    {
+      "epoch": 25.0,
+      "eval_accuracy": 0.7649492246185369,
+      "eval_loss": 0.9867243766784668,
+      "eval_runtime": 50.5963,
+      "eval_samples_per_second": 898.327,
+      "eval_steps_per_second": 14.052,
+      "step": 521075
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 9.670646464646467e-06,
+      "loss": 1.1219,
+      "step": 521500
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 9.660545454545455e-06,
+      "loss": 1.126,
+      "step": 522000
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 9.650444444444446e-06,
+      "loss": 1.1213,
+      "step": 522500
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 9.640343434343436e-06,
+      "loss": 1.1211,
+      "step": 523000
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 9.630262626262628e-06,
+      "loss": 1.1228,
+      "step": 523500
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 9.620161616161617e-06,
+      "loss": 1.1202,
+      "step": 524000
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 9.610060606060607e-06,
+      "loss": 1.1222,
+      "step": 524500
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 9.599959595959597e-06,
+      "loss": 1.1225,
+      "step": 525000
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 9.589858585858588e-06,
+      "loss": 1.122,
+      "step": 525500
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 9.579757575757576e-06,
+      "loss": 1.1247,
+      "step": 526000
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 9.569656565656567e-06,
+      "loss": 1.1201,
+      "step": 526500
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 9.559575757575759e-06,
+      "loss": 1.1202,
+      "step": 527000
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 9.549474747474749e-06,
+      "loss": 1.1231,
+      "step": 527500
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 9.539373737373738e-06,
+      "loss": 1.1223,
+      "step": 528000
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 9.529272727272728e-06,
+      "loss": 1.1227,
+      "step": 528500
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 9.519171717171718e-06,
+      "loss": 1.122,
+      "step": 529000
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 9.509070707070708e-06,
+      "loss": 1.1229,
+      "step": 529500
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 9.498969696969697e-06,
+      "loss": 1.1233,
+      "step": 530000
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 9.48888888888889e-06,
+      "loss": 1.1226,
+      "step": 530500
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 9.47878787878788e-06,
+      "loss": 1.1202,
+      "step": 531000
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 9.46868686868687e-06,
+      "loss": 1.1211,
+      "step": 531500
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 9.458585858585858e-06,
+      "loss": 1.1231,
+      "step": 532000
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 9.448484848484849e-06,
+      "loss": 1.1215,
+      "step": 532500
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 9.438383838383839e-06,
+      "loss": 1.1225,
+      "step": 533000
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 9.42828282828283e-06,
+      "loss": 1.1228,
+      "step": 533500
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 9.418202020202022e-06,
+      "loss": 1.1204,
+      "step": 534000
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 9.40810101010101e-06,
+      "loss": 1.121,
+      "step": 534500
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 9.398e-06,
+      "loss": 1.1223,
+      "step": 535000
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 9.38789898989899e-06,
+      "loss": 1.1221,
+      "step": 535500
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 9.37779797979798e-06,
+      "loss": 1.1208,
+      "step": 536000
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 9.367717171717172e-06,
+      "loss": 1.1218,
+      "step": 536500
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 9.357616161616162e-06,
+      "loss": 1.1213,
+      "step": 537000
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 9.347515151515152e-06,
+      "loss": 1.1211,
+      "step": 537500
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 9.337434343434344e-06,
+      "loss": 1.1204,
+      "step": 538000
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 9.327333333333333e-06,
+      "loss": 1.1195,
+      "step": 538500
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 9.317232323232323e-06,
+      "loss": 1.1214,
+      "step": 539000
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 9.307131313131314e-06,
+      "loss": 1.1211,
+      "step": 539500
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 9.297030303030304e-06,
+      "loss": 1.1186,
+      "step": 540000
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 9.286929292929294e-06,
+      "loss": 1.1198,
+      "step": 540500
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 9.276828282828284e-06,
+      "loss": 1.124,
+      "step": 541000
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 9.266747474747475e-06,
+      "loss": 1.1187,
+      "step": 541500
+    },
+    {
+      "epoch": 26.0,
+      "eval_accuracy": 0.7656655092426662,
+      "eval_loss": 0.985273540019989,
+      "eval_runtime": 50.0306,
+      "eval_samples_per_second": 908.484,
+      "eval_steps_per_second": 14.211,
+      "step": 541918
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 9.256646464646465e-06,
+      "loss": 1.1194,
+      "step": 542000
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 9.246545454545456e-06,
+      "loss": 1.1197,
+      "step": 542500
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 9.236444444444446e-06,
+      "loss": 1.1183,
+      "step": 543000
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 9.226343434343436e-06,
+      "loss": 1.1219,
+      "step": 543500
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 9.216262626262627e-06,
+      "loss": 1.1193,
+      "step": 544000
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 9.206161616161617e-06,
+      "loss": 1.1199,
+      "step": 544500
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 9.196060606060607e-06,
+      "loss": 1.1219,
+      "step": 545000
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 9.185959595959598e-06,
+      "loss": 1.1172,
+      "step": 545500
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 9.175858585858586e-06,
+      "loss": 1.1226,
+      "step": 546000
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 9.165757575757576e-06,
+      "loss": 1.1221,
+      "step": 546500
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 9.155656565656567e-06,
+      "loss": 1.1225,
+      "step": 547000
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 9.145555555555557e-06,
+      "loss": 1.1184,
+      "step": 547500
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 9.135454545454546e-06,
+      "loss": 1.1201,
+      "step": 548000
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 9.125353535353536e-06,
+      "loss": 1.1193,
+      "step": 548500
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 9.115272727272728e-06,
+      "loss": 1.1198,
+      "step": 549000
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 9.105171717171718e-06,
+      "loss": 1.1189,
+      "step": 549500
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 9.095070707070707e-06,
+      "loss": 1.1185,
+      "step": 550000
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 9.084969696969697e-06,
+      "loss": 1.1196,
+      "step": 550500
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 9.074868686868688e-06,
+      "loss": 1.1212,
+      "step": 551000
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 9.06478787878788e-06,
+      "loss": 1.1187,
+      "step": 551500
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 9.05468686868687e-06,
+      "loss": 1.1194,
+      "step": 552000
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 9.04460606060606e-06,
+      "loss": 1.1209,
+      "step": 552500
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 9.034505050505051e-06,
+      "loss": 1.1181,
+      "step": 553000
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 9.024404040404041e-06,
+      "loss": 1.1176,
+      "step": 553500
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 9.014303030303032e-06,
+      "loss": 1.1198,
+      "step": 554000
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 9.00420202020202e-06,
+      "loss": 1.1179,
+      "step": 554500
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 8.99410101010101e-06,
+      "loss": 1.1205,
+      "step": 555000
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 8.984e-06,
+      "loss": 1.1193,
+      "step": 555500
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 8.973898989898991e-06,
+      "loss": 1.121,
+      "step": 556000
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 8.96379797979798e-06,
+      "loss": 1.1202,
+      "step": 556500
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 8.95369696969697e-06,
+      "loss": 1.1182,
+      "step": 557000
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 8.94359595959596e-06,
+      "loss": 1.1209,
+      "step": 557500
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 8.93349494949495e-06,
+      "loss": 1.1209,
+      "step": 558000
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 8.92339393939394e-06,
+      "loss": 1.1221,
+      "step": 558500
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 8.913313131313131e-06,
+      "loss": 1.1177,
+      "step": 559000
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 8.903232323232324e-06,
+      "loss": 1.1215,
+      "step": 559500
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 8.893131313131314e-06,
+      "loss": 1.1196,
+      "step": 560000
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 8.883030303030304e-06,
+      "loss": 1.1166,
+      "step": 560500
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 8.872929292929293e-06,
+      "loss": 1.122,
+      "step": 561000
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 8.862848484848485e-06,
+      "loss": 1.1191,
+      "step": 561500
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 8.852747474747475e-06,
+      "loss": 1.1192,
+      "step": 562000
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 8.842646464646466e-06,
+      "loss": 1.1187,
+      "step": 562500
+    },
+    {
+      "epoch": 27.0,
+      "eval_accuracy": 0.7652794822906015,
+      "eval_loss": 0.9845146536827087,
+      "eval_runtime": 50.1388,
+      "eval_samples_per_second": 906.524,
+      "eval_steps_per_second": 14.181,
+      "step": 562761
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 8.832545454545454e-06,
+      "loss": 1.1179,
+      "step": 563000
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 8.822444444444446e-06,
+      "loss": 1.1174,
+      "step": 563500
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 8.812343434343435e-06,
+      "loss": 1.1185,
+      "step": 564000
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 8.802242424242425e-06,
+      "loss": 1.1171,
+      "step": 564500
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 8.792161616161616e-06,
+      "loss": 1.1177,
+      "step": 565000
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 8.782060606060608e-06,
+      "loss": 1.118,
+      "step": 565500
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 8.771959595959598e-06,
+      "loss": 1.1191,
+      "step": 566000
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 8.761858585858586e-06,
+      "loss": 1.1206,
+      "step": 566500
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 8.751777777777779e-06,
+      "loss": 1.119,
+      "step": 567000
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 8.741676767676769e-06,
+      "loss": 1.1171,
+      "step": 567500
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 8.73157575757576e-06,
+      "loss": 1.119,
+      "step": 568000
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 8.721474747474748e-06,
+      "loss": 1.1196,
+      "step": 568500
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 8.711373737373738e-06,
+      "loss": 1.1181,
+      "step": 569000
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 8.701272727272728e-06,
+      "loss": 1.1184,
+      "step": 569500
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 8.691171717171719e-06,
+      "loss": 1.1229,
+      "step": 570000
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 8.681090909090911e-06,
+      "loss": 1.1205,
+      "step": 570500
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 8.6709898989899e-06,
+      "loss": 1.1174,
+      "step": 571000
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 8.66088888888889e-06,
+      "loss": 1.1184,
+      "step": 571500
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 8.65078787878788e-06,
+      "loss": 1.1191,
+      "step": 572000
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 8.640707070707072e-06,
+      "loss": 1.1182,
+      "step": 572500
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 8.630606060606061e-06,
+      "loss": 1.1174,
+      "step": 573000
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 8.620505050505051e-06,
+      "loss": 1.1185,
+      "step": 573500
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 8.610404040404042e-06,
+      "loss": 1.1163,
+      "step": 574000
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 8.600303030303032e-06,
+      "loss": 1.1171,
+      "step": 574500
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 8.59020202020202e-06,
+      "loss": 1.1183,
+      "step": 575000
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 8.58010101010101e-06,
+      "loss": 1.1178,
+      "step": 575500
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 8.570020202020203e-06,
+      "loss": 1.1186,
+      "step": 576000
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 8.559919191919193e-06,
+      "loss": 1.119,
+      "step": 576500
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 8.549818181818182e-06,
+      "loss": 1.1163,
+      "step": 577000
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 8.539717171717172e-06,
+      "loss": 1.1164,
+      "step": 577500
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 8.529616161616162e-06,
+      "loss": 1.1169,
+      "step": 578000
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 8.519515151515153e-06,
+      "loss": 1.1152,
+      "step": 578500
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 8.509434343434343e-06,
+      "loss": 1.1207,
+      "step": 579000
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 8.499333333333333e-06,
+      "loss": 1.1214,
+      "step": 579500
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 8.489232323232324e-06,
+      "loss": 1.1178,
+      "step": 580000
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 8.479131313131314e-06,
+      "loss": 1.1166,
+      "step": 580500
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 8.469030303030303e-06,
+      "loss": 1.1176,
+      "step": 581000
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 8.458929292929293e-06,
+      "loss": 1.118,
+      "step": 581500
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 8.448828282828283e-06,
+      "loss": 1.119,
+      "step": 582000
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 8.438727272727274e-06,
+      "loss": 1.1185,
+      "step": 582500
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 8.428646464646466e-06,
+      "loss": 1.1158,
+      "step": 583000
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 8.418545454545454e-06,
+      "loss": 1.119,
+      "step": 583500
+    },
+    {
+      "epoch": 28.0,
+      "eval_accuracy": 0.7658769146335664,
+      "eval_loss": 0.9811434149742126,
+      "eval_runtime": 51.1021,
+      "eval_samples_per_second": 889.436,
+      "eval_steps_per_second": 13.913,
+      "step": 583604
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 8.408464646464647e-06,
+      "loss": 1.1161,
+      "step": 584000
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 8.398363636363637e-06,
+      "loss": 1.119,
+      "step": 584500
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 8.388262626262627e-06,
+      "loss": 1.1171,
+      "step": 585000
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 8.378161616161616e-06,
+      "loss": 1.1187,
+      "step": 585500
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 8.368060606060606e-06,
+      "loss": 1.119,
+      "step": 586000
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 8.357959595959596e-06,
+      "loss": 1.118,
+      "step": 586500
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 8.347878787878789e-06,
+      "loss": 1.1148,
+      "step": 587000
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 8.337777777777777e-06,
+      "loss": 1.1143,
+      "step": 587500
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 8.32769696969697e-06,
+      "loss": 1.1178,
+      "step": 588000
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 8.31759595959596e-06,
+      "loss": 1.1131,
+      "step": 588500
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 8.30749494949495e-06,
+      "loss": 1.1186,
+      "step": 589000
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 8.29739393939394e-06,
+      "loss": 1.1166,
+      "step": 589500
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 8.28729292929293e-06,
+      "loss": 1.1173,
+      "step": 590000
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 8.277191919191921e-06,
+      "loss": 1.1163,
+      "step": 590500
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 8.26709090909091e-06,
+      "loss": 1.1175,
+      "step": 591000
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 8.2569898989899e-06,
+      "loss": 1.1163,
+      "step": 591500
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 8.24688888888889e-06,
+      "loss": 1.1158,
+      "step": 592000
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 8.23678787878788e-06,
+      "loss": 1.1171,
+      "step": 592500
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 8.226686868686869e-06,
+      "loss": 1.1169,
+      "step": 593000
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 8.216606060606061e-06,
+      "loss": 1.1166,
+      "step": 593500
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 8.206505050505051e-06,
+      "loss": 1.1177,
+      "step": 594000
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 8.196404040404042e-06,
+      "loss": 1.1148,
+      "step": 594500
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 8.186323232323234e-06,
+      "loss": 1.1163,
+      "step": 595000
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 8.176222222222223e-06,
+      "loss": 1.1166,
+      "step": 595500
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 8.166121212121213e-06,
+      "loss": 1.1165,
+      "step": 596000
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 8.156020202020203e-06,
+      "loss": 1.1167,
+      "step": 596500
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 8.145919191919193e-06,
+      "loss": 1.1151,
+      "step": 597000
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 8.135818181818182e-06,
+      "loss": 1.1145,
+      "step": 597500
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 8.125717171717172e-06,
+      "loss": 1.1182,
+      "step": 598000
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 8.115616161616163e-06,
+      "loss": 1.1164,
+      "step": 598500
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 8.105515151515151e-06,
+      "loss": 1.1181,
+      "step": 599000
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 8.095414141414141e-06,
+      "loss": 1.1171,
+      "step": 599500
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 8.085313131313132e-06,
+      "loss": 1.1155,
+      "step": 600000
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 8.075232323232324e-06,
+      "loss": 1.117,
+      "step": 600500
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 8.065131313131314e-06,
+      "loss": 1.1155,
+      "step": 601000
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 8.055030303030303e-06,
+      "loss": 1.1144,
+      "step": 601500
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 8.044929292929293e-06,
+      "loss": 1.1159,
+      "step": 602000
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 8.034828282828283e-06,
+      "loss": 1.1158,
+      "step": 602500
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 8.024747474747476e-06,
+      "loss": 1.118,
+      "step": 603000
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 8.014646464646464e-06,
+      "loss": 1.1179,
+      "step": 603500
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 8.004545454545455e-06,
+      "loss": 1.1152,
+      "step": 604000
+    },
+    {
+      "epoch": 29.0,
+      "eval_accuracy": 0.7660435184632822,
+      "eval_loss": 0.9821996092796326,
+      "eval_runtime": 50.6607,
+      "eval_samples_per_second": 897.184,
+      "eval_steps_per_second": 14.035,
+      "step": 604447
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 7.994444444444445e-06,
+      "loss": 1.1175,
+      "step": 604500
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 7.984343434343435e-06,
+      "loss": 1.1173,
+      "step": 605000
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 7.974242424242424e-06,
+      "loss": 1.1181,
+      "step": 605500
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 7.964161616161616e-06,
+      "loss": 1.1148,
+      "step": 606000
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 7.954060606060606e-06,
+      "loss": 1.1147,
+      "step": 606500
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 7.943959595959597e-06,
+      "loss": 1.1145,
+      "step": 607000
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 7.933858585858587e-06,
+      "loss": 1.1119,
+      "step": 607500
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 7.923757575757577e-06,
+      "loss": 1.1143,
+      "step": 608000
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 7.913656565656567e-06,
+      "loss": 1.1138,
+      "step": 608500
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 7.903555555555556e-06,
+      "loss": 1.1142,
+      "step": 609000
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 7.893474747474748e-06,
+      "loss": 1.1162,
+      "step": 609500
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 7.883373737373739e-06,
+      "loss": 1.1134,
+      "step": 610000
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 7.873272727272729e-06,
+      "loss": 1.1141,
+      "step": 610500
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 7.863171717171717e-06,
+      "loss": 1.1124,
+      "step": 611000
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 7.853070707070708e-06,
+      "loss": 1.1146,
+      "step": 611500
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 7.842969696969698e-06,
+      "loss": 1.1122,
+      "step": 612000
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 7.83288888888889e-06,
+      "loss": 1.1163,
+      "step": 612500
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 7.822787878787879e-06,
+      "loss": 1.1132,
+      "step": 613000
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 7.81268686868687e-06,
+      "loss": 1.1143,
+      "step": 613500
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 7.802606060606061e-06,
+      "loss": 1.1154,
+      "step": 614000
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 7.792505050505052e-06,
+      "loss": 1.117,
+      "step": 614500
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 7.782404040404042e-06,
+      "loss": 1.1125,
+      "step": 615000
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 7.77230303030303e-06,
+      "loss": 1.1173,
+      "step": 615500
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 7.762202020202021e-06,
+      "loss": 1.1141,
+      "step": 616000
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 7.752101010101011e-06,
+      "loss": 1.1137,
+      "step": 616500
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 7.742000000000001e-06,
+      "loss": 1.118,
+      "step": 617000
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 7.73189898989899e-06,
+      "loss": 1.1124,
+      "step": 617500
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 7.72179797979798e-06,
+      "loss": 1.1157,
+      "step": 618000
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 7.71169696969697e-06,
+      "loss": 1.1167,
+      "step": 618500
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 7.701616161616163e-06,
+      "loss": 1.1128,
+      "step": 619000
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 7.691515151515151e-06,
+      "loss": 1.1163,
+      "step": 619500
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 7.681414141414142e-06,
+      "loss": 1.1182,
+      "step": 620000
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 7.671313131313132e-06,
+      "loss": 1.1151,
+      "step": 620500
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 7.661212121212122e-06,
+      "loss": 1.1152,
+      "step": 621000
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 7.651131313131313e-06,
+      "loss": 1.1124,
+      "step": 621500
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 7.641030303030303e-06,
+      "loss": 1.1155,
+      "step": 622000
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 7.630949494949495e-06,
+      "loss": 1.1147,
+      "step": 622500
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 7.620848484848486e-06,
+      "loss": 1.1145,
+      "step": 623000
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 7.610747474747475e-06,
+      "loss": 1.1148,
+      "step": 623500
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 7.600646464646465e-06,
+      "loss": 1.1133,
+      "step": 624000
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 7.590545454545455e-06,
+      "loss": 1.1133,
+      "step": 624500
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 7.580444444444445e-06,
+      "loss": 1.1142,
+      "step": 625000
+    },
+    {
+      "epoch": 30.0,
+      "eval_accuracy": 0.7664225183785052,
+      "eval_loss": 0.977989137172699,
+      "eval_runtime": 50.7055,
+      "eval_samples_per_second": 896.392,
+      "eval_steps_per_second": 14.022,
+      "step": 625290
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 7.5703434343434346e-06,
+      "loss": 1.1163,
+      "step": 625500
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 7.560242424242424e-06,
+      "loss": 1.1139,
+      "step": 626000
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 7.550161616161616e-06,
+      "loss": 1.1135,
+      "step": 626500
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 7.5400606060606065e-06,
+      "loss": 1.1145,
+      "step": 627000
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 7.529959595959596e-06,
+      "loss": 1.1133,
+      "step": 627500
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 7.519858585858586e-06,
+      "loss": 1.1138,
+      "step": 628000
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 7.5097575757575766e-06,
+      "loss": 1.1143,
+      "step": 628500
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 7.499656565656567e-06,
+      "loss": 1.1166,
+      "step": 629000
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 7.489555555555556e-06,
+      "loss": 1.1131,
+      "step": 629500
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 7.4794545454545466e-06,
+      "loss": 1.1109,
+      "step": 630000
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 7.469373737373739e-06,
+      "loss": 1.1104,
+      "step": 630500
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 7.459272727272728e-06,
+      "loss": 1.1123,
+      "step": 631000
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 7.4491717171717185e-06,
+      "loss": 1.1129,
+      "step": 631500
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 7.43909090909091e-06,
+      "loss": 1.1122,
+      "step": 632000
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 7.4289898989899e-06,
+      "loss": 1.1114,
+      "step": 632500
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 7.41888888888889e-06,
+      "loss": 1.1106,
+      "step": 633000
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 7.40878787878788e-06,
+      "loss": 1.114,
+      "step": 633500
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 7.398686868686869e-06,
+      "loss": 1.115,
+      "step": 634000
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 7.38858585858586e-06,
+      "loss": 1.1127,
+      "step": 634500
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 7.378484848484849e-06,
+      "loss": 1.1133,
+      "step": 635000
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 7.368383838383839e-06,
+      "loss": 1.1126,
+      "step": 635500
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 7.358282828282829e-06,
+      "loss": 1.1132,
+      "step": 636000
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 7.348202020202021e-06,
+      "loss": 1.1143,
+      "step": 636500
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 7.3381010101010105e-06,
+      "loss": 1.1146,
+      "step": 637000
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 7.328000000000001e-06,
+      "loss": 1.1122,
+      "step": 637500
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 7.31789898989899e-06,
+      "loss": 1.1141,
+      "step": 638000
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 7.3077979797979805e-06,
+      "loss": 1.1141,
+      "step": 638500
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 7.29769696969697e-06,
+      "loss": 1.1158,
+      "step": 639000
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 7.28759595959596e-06,
+      "loss": 1.1125,
+      "step": 639500
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 7.27749494949495e-06,
+      "loss": 1.1135,
+      "step": 640000
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 7.267434343434344e-06,
+      "loss": 1.116,
+      "step": 640500
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 7.257333333333334e-06,
+      "loss": 1.1141,
+      "step": 641000
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 7.247232323232324e-06,
+      "loss": 1.112,
+      "step": 641500
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 7.237131313131314e-06,
+      "loss": 1.1121,
+      "step": 642000
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 7.227030303030303e-06,
+      "loss": 1.1128,
+      "step": 642500
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 7.216929292929294e-06,
+      "loss": 1.1115,
+      "step": 643000
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 7.206828282828283e-06,
+      "loss": 1.1125,
+      "step": 643500
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 7.196747474747475e-06,
+      "loss": 1.1129,
+      "step": 644000
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 7.186646464646465e-06,
+      "loss": 1.1115,
+      "step": 644500
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 7.176545454545455e-06,
+      "loss": 1.1123,
+      "step": 645000
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 7.1664444444444445e-06,
+      "loss": 1.1143,
+      "step": 645500
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 7.156343434343435e-06,
+      "loss": 1.1112,
+      "step": 646000
+    },
+    {
+      "epoch": 31.0,
+      "eval_accuracy": 0.766590941311319,
+      "eval_loss": 0.9784865975379944,
+      "eval_runtime": 50.2474,
+      "eval_samples_per_second": 904.565,
+      "eval_steps_per_second": 14.15,
+      "step": 646133
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 7.146242424242424e-06,
+      "loss": 1.1127,
+      "step": 646500
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 7.1361414141414145e-06,
+      "loss": 1.1133,
+      "step": 647000
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 7.126040404040404e-06,
+      "loss": 1.1113,
+      "step": 647500
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 7.115979797979798e-06,
+      "loss": 1.1121,
+      "step": 648000
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 7.1058787878787885e-06,
+      "loss": 1.11,
+      "step": 648500
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 7.095797979797981e-06,
+      "loss": 1.1134,
+      "step": 649000
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 7.08569696969697e-06,
+      "loss": 1.1105,
+      "step": 649500
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 7.0755959595959604e-06,
+      "loss": 1.1122,
+      "step": 650000
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 7.06549494949495e-06,
+      "loss": 1.1118,
+      "step": 650500
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 7.055393939393939e-06,
+      "loss": 1.1095,
+      "step": 651000
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 7.04529292929293e-06,
+      "loss": 1.1103,
+      "step": 651500
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 7.035191919191919e-06,
+      "loss": 1.1129,
+      "step": 652000
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 7.025090909090909e-06,
+      "loss": 1.1131,
+      "step": 652500
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 7.014989898989899e-06,
+      "loss": 1.11,
+      "step": 653000
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 7.004909090909091e-06,
+      "loss": 1.1124,
+      "step": 653500
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 6.994808080808081e-06,
+      "loss": 1.1118,
+      "step": 654000
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 6.984707070707071e-06,
+      "loss": 1.1123,
+      "step": 654500
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 6.974606060606062e-06,
+      "loss": 1.1131,
+      "step": 655000
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 6.9645252525252524e-06,
+      "loss": 1.1101,
+      "step": 655500
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 6.954424242424243e-06,
+      "loss": 1.1126,
+      "step": 656000
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 6.944323232323232e-06,
+      "loss": 1.1114,
+      "step": 656500
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 6.934222222222223e-06,
+      "loss": 1.1121,
+      "step": 657000
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 6.924121212121213e-06,
+      "loss": 1.1124,
+      "step": 657500
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 6.914020202020203e-06,
+      "loss": 1.1111,
+      "step": 658000
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 6.9039191919191925e-06,
+      "loss": 1.1118,
+      "step": 658500
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 6.893818181818183e-06,
+      "loss": 1.1109,
+      "step": 659000
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 6.883717171717172e-06,
+      "loss": 1.1112,
+      "step": 659500
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 6.8736161616161625e-06,
+      "loss": 1.1113,
+      "step": 660000
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 6.863515151515152e-06,
+      "loss": 1.1132,
+      "step": 660500
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 6.853414141414142e-06,
+      "loss": 1.1098,
+      "step": 661000
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 6.843313131313132e-06,
+      "loss": 1.1109,
+      "step": 661500
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 6.833252525252526e-06,
+      "loss": 1.1112,
+      "step": 662000
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 6.823151515151516e-06,
+      "loss": 1.1094,
+      "step": 662500
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 6.8130505050505056e-06,
+      "loss": 1.1128,
+      "step": 663000
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 6.802949494949496e-06,
+      "loss": 1.113,
+      "step": 663500
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 6.792848484848485e-06,
+      "loss": 1.1133,
+      "step": 664000
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 6.782747474747476e-06,
+      "loss": 1.1118,
+      "step": 664500
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 6.772646464646465e-06,
+      "loss": 1.1089,
+      "step": 665000
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 6.762545454545455e-06,
+      "loss": 1.1096,
+      "step": 665500
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 6.752464646464647e-06,
+      "loss": 1.1098,
+      "step": 666000
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 6.742363636363637e-06,
+      "loss": 1.113,
+      "step": 666500
+    },
+    {
+      "epoch": 32.0,
+      "eval_accuracy": 0.7673372197904346,
+      "eval_loss": 0.972804605960846,
+      "eval_runtime": 50.5938,
+      "eval_samples_per_second": 898.37,
+      "eval_steps_per_second": 14.053,
+      "step": 666976
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 6.732282828282829e-06,
+      "loss": 1.1093,
+      "step": 667000
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 6.722181818181819e-06,
+      "loss": 1.1096,
+      "step": 667500
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 6.712080808080809e-06,
+      "loss": 1.1117,
+      "step": 668000
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 6.701979797979798e-06,
+      "loss": 1.1104,
+      "step": 668500
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 6.691878787878789e-06,
+      "loss": 1.1114,
+      "step": 669000
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 6.681777777777778e-06,
+      "loss": 1.1105,
+      "step": 669500
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 6.671676767676768e-06,
+      "loss": 1.1089,
+      "step": 670000
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 6.661575757575758e-06,
+      "loss": 1.1109,
+      "step": 670500
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 6.651474747474747e-06,
+      "loss": 1.1086,
+      "step": 671000
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 6.641414141414142e-06,
+      "loss": 1.1094,
+      "step": 671500
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 6.631313131313132e-06,
+      "loss": 1.1121,
+      "step": 672000
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 6.621212121212121e-06,
+      "loss": 1.1111,
+      "step": 672500
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 6.6111111111111115e-06,
+      "loss": 1.1088,
+      "step": 673000
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 6.601010101010101e-06,
+      "loss": 1.1087,
+      "step": 673500
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 6.590909090909091e-06,
+      "loss": 1.112,
+      "step": 674000
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 6.580808080808081e-06,
+      "loss": 1.111,
+      "step": 674500
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 6.570707070707071e-06,
+      "loss": 1.1088,
+      "step": 675000
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 6.560626262626263e-06,
+      "loss": 1.1122,
+      "step": 675500
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 6.550525252525253e-06,
+      "loss": 1.1112,
+      "step": 676000
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 6.540444444444445e-06,
+      "loss": 1.1106,
+      "step": 676500
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 6.530343434343434e-06,
+      "loss": 1.1128,
+      "step": 677000
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 6.520242424242425e-06,
+      "loss": 1.1121,
+      "step": 677500
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 6.510141414141414e-06,
+      "loss": 1.111,
+      "step": 678000
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 6.500040404040404e-06,
+      "loss": 1.111,
+      "step": 678500
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 6.489939393939394e-06,
+      "loss": 1.1061,
+      "step": 679000
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 6.479838383838384e-06,
+      "loss": 1.1074,
+      "step": 679500
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 6.4697575757575755e-06,
+      "loss": 1.1097,
+      "step": 680000
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 6.459656565656566e-06,
+      "loss": 1.1099,
+      "step": 680500
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 6.449555555555555e-06,
+      "loss": 1.1107,
+      "step": 681000
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 6.439454545454546e-06,
+      "loss": 1.1107,
+      "step": 681500
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 6.429353535353537e-06,
+      "loss": 1.1105,
+      "step": 682000
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 6.419252525252526e-06,
+      "loss": 1.1106,
+      "step": 682500
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 6.409151515151516e-06,
+      "loss": 1.112,
+      "step": 683000
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 6.399050505050506e-06,
+      "loss": 1.1102,
+      "step": 683500
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 6.388969696969698e-06,
+      "loss": 1.1101,
+      "step": 684000
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 6.378888888888889e-06,
+      "loss": 1.1124,
+      "step": 684500
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 6.36878787878788e-06,
+      "loss": 1.1099,
+      "step": 685000
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 6.358707070707071e-06,
+      "loss": 1.1107,
+      "step": 685500
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 6.348606060606061e-06,
+      "loss": 1.112,
+      "step": 686000
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 6.338505050505051e-06,
+      "loss": 1.1103,
+      "step": 686500
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 6.328404040404041e-06,
+      "loss": 1.1069,
+      "step": 687000
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 6.3183030303030314e-06,
+      "loss": 1.1115,
+      "step": 687500
+    },
+    {
+      "epoch": 33.0,
+      "eval_accuracy": 0.7668397649537687,
+      "eval_loss": 0.9765338897705078,
+      "eval_runtime": 51.0181,
+      "eval_samples_per_second": 890.899,
+      "eval_steps_per_second": 13.936,
+      "step": 687819
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 6.308202020202021e-06,
+      "loss": 1.1079,
+      "step": 688000
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 6.298101010101011e-06,
+      "loss": 1.11,
+      "step": 688500
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 6.288000000000001e-06,
+      "loss": 1.1121,
+      "step": 689000
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 6.277919191919193e-06,
+      "loss": 1.109,
+      "step": 689500
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 6.267818181818182e-06,
+      "loss": 1.1087,
+      "step": 690000
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 6.257717171717173e-06,
+      "loss": 1.109,
+      "step": 690500
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 6.247616161616162e-06,
+      "loss": 1.1103,
+      "step": 691000
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 6.237515151515152e-06,
+      "loss": 1.1067,
+      "step": 691500
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 6.227414141414142e-06,
+      "loss": 1.1128,
+      "step": 692000
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 6.217333333333334e-06,
+      "loss": 1.1068,
+      "step": 692500
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 6.207232323232324e-06,
+      "loss": 1.1089,
+      "step": 693000
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 6.197151515151516e-06,
+      "loss": 1.1075,
+      "step": 693500
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 6.187050505050506e-06,
+      "loss": 1.108,
+      "step": 694000
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 6.176949494949495e-06,
+      "loss": 1.1056,
+      "step": 694500
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 6.166848484848486e-06,
+      "loss": 1.1081,
+      "step": 695000
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 6.156747474747475e-06,
+      "loss": 1.1103,
+      "step": 695500
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 6.1466464646464654e-06,
+      "loss": 1.1092,
+      "step": 696000
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 6.136545454545455e-06,
+      "loss": 1.1094,
+      "step": 696500
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 6.126444444444445e-06,
+      "loss": 1.1089,
+      "step": 697000
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 6.1163636363636366e-06,
+      "loss": 1.1088,
+      "step": 697500
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 6.106262626262627e-06,
+      "loss": 1.1055,
+      "step": 698000
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 6.096161616161616e-06,
+      "loss": 1.1077,
+      "step": 698500
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 6.0860606060606066e-06,
+      "loss": 1.1094,
+      "step": 699000
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 6.075959595959596e-06,
+      "loss": 1.1114,
+      "step": 699500
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 6.065858585858586e-06,
+      "loss": 1.1111,
+      "step": 700000
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 6.055757575757576e-06,
+      "loss": 1.1098,
+      "step": 700500
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 6.045656565656566e-06,
+      "loss": 1.1108,
+      "step": 701000
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 6.0355555555555555e-06,
+      "loss": 1.1091,
+      "step": 701500
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 6.025474747474748e-06,
+      "loss": 1.107,
+      "step": 702000
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 6.015373737373737e-06,
+      "loss": 1.1078,
+      "step": 702500
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 6.0052727272727274e-06,
+      "loss": 1.1094,
+      "step": 703000
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 5.99519191919192e-06,
+      "loss": 1.1087,
+      "step": 703500
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 5.985090909090909e-06,
+      "loss": 1.1077,
+      "step": 704000
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 5.974989898989899e-06,
+      "loss": 1.1094,
+      "step": 704500
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 5.964888888888889e-06,
+      "loss": 1.1092,
+      "step": 705000
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 5.954787878787879e-06,
+      "loss": 1.1083,
+      "step": 705500
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 5.944686868686869e-06,
+      "loss": 1.112,
+      "step": 706000
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 5.93458585858586e-06,
+      "loss": 1.1109,
+      "step": 706500
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 5.92450505050505e-06,
+      "loss": 1.1071,
+      "step": 707000
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 5.9144040404040406e-06,
+      "loss": 1.1089,
+      "step": 707500
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 5.904303030303032e-06,
+      "loss": 1.1074,
+      "step": 708000
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 5.894202020202021e-06,
+      "loss": 1.1058,
+      "step": 708500
+    },
+    {
+      "epoch": 34.0,
+      "eval_accuracy": 0.7673810843282054,
+      "eval_loss": 0.9737652540206909,
+      "eval_runtime": 50.9425,
+      "eval_samples_per_second": 892.221,
+      "eval_steps_per_second": 13.957,
+      "step": 708662
+    },
+    {
+      "epoch": 34.02,
+      "learning_rate": 5.8841212121212125e-06,
+      "loss": 1.1074,
+      "step": 709000
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 5.874020202020202e-06,
+      "loss": 1.1065,
+      "step": 709500
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 5.863919191919193e-06,
+      "loss": 1.1079,
+      "step": 710000
+    },
+    {
+      "epoch": 34.09,
+      "learning_rate": 5.8538181818181825e-06,
+      "loss": 1.1061,
+      "step": 710500
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 5.843717171717173e-06,
+      "loss": 1.1097,
+      "step": 711000
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 5.833616161616162e-06,
+      "loss": 1.1057,
+      "step": 711500
+    },
+    {
+      "epoch": 34.16,
+      "learning_rate": 5.8235151515151526e-06,
+      "loss": 1.1053,
+      "step": 712000
+    },
+    {
+      "epoch": 34.18,
+      "learning_rate": 5.813414141414142e-06,
+      "loss": 1.107,
+      "step": 712500
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 5.803313131313132e-06,
+      "loss": 1.1086,
+      "step": 713000
+    },
+    {
+      "epoch": 34.23,
+      "learning_rate": 5.7932525252525265e-06,
+      "loss": 1.1113,
+      "step": 713500
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 5.783171717171717e-06,
+      "loss": 1.1094,
+      "step": 714000
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 5.773070707070707e-06,
+      "loss": 1.1083,
+      "step": 714500
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 5.762969696969698e-06,
+      "loss": 1.1086,
+      "step": 715000
+    },
+    {
+      "epoch": 34.33,
+      "learning_rate": 5.752868686868688e-06,
+      "loss": 1.1085,
+      "step": 715500
+    },
+    {
+      "epoch": 34.35,
+      "learning_rate": 5.742767676767677e-06,
+      "loss": 1.1099,
+      "step": 716000
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 5.732666666666668e-06,
+      "loss": 1.1096,
+      "step": 716500
+    },
+    {
+      "epoch": 34.4,
+      "learning_rate": 5.722565656565657e-06,
+      "loss": 1.1067,
+      "step": 717000
+    },
+    {
+      "epoch": 34.42,
+      "learning_rate": 5.712464646464647e-06,
+      "loss": 1.1037,
+      "step": 717500
+    },
+    {
+      "epoch": 34.45,
+      "learning_rate": 5.702363636363637e-06,
+      "loss": 1.1084,
+      "step": 718000
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 5.692282828282829e-06,
+      "loss": 1.1088,
+      "step": 718500
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 5.682181818181819e-06,
+      "loss": 1.106,
+      "step": 719000
+    },
+    {
+      "epoch": 34.52,
+      "learning_rate": 5.672080808080809e-06,
+      "loss": 1.1084,
+      "step": 719500
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 5.661979797979798e-06,
+      "loss": 1.1097,
+      "step": 720000
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 5.6518787878787885e-06,
+      "loss": 1.1076,
+      "step": 720500
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 5.641797979797981e-06,
+      "loss": 1.1085,
+      "step": 721000
+    },
+    {
+      "epoch": 34.62,
+      "learning_rate": 5.63169696969697e-06,
+      "loss": 1.1077,
+      "step": 721500
+    },
+    {
+      "epoch": 34.64,
+      "learning_rate": 5.6215959595959605e-06,
+      "loss": 1.1055,
+      "step": 722000
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 5.61149494949495e-06,
+      "loss": 1.1065,
+      "step": 722500
+    },
+    {
+      "epoch": 34.69,
+      "learning_rate": 5.601414141414142e-06,
+      "loss": 1.1073,
+      "step": 723000
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 5.591313131313132e-06,
+      "loss": 1.1076,
+      "step": 723500
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 5.581212121212122e-06,
+      "loss": 1.1102,
+      "step": 724000
+    },
+    {
+      "epoch": 34.76,
+      "learning_rate": 5.571111111111111e-06,
+      "loss": 1.1092,
+      "step": 724500
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 5.561010101010102e-06,
+      "loss": 1.1047,
+      "step": 725000
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 5.550909090909091e-06,
+      "loss": 1.1072,
+      "step": 725500
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 5.540828282828283e-06,
+      "loss": 1.1081,
+      "step": 726000
+    },
+    {
+      "epoch": 34.86,
+      "learning_rate": 5.530727272727274e-06,
+      "loss": 1.1074,
+      "step": 726500
+    },
+    {
+      "epoch": 34.88,
+      "learning_rate": 5.520626262626263e-06,
+      "loss": 1.1067,
+      "step": 727000
+    },
+    {
+      "epoch": 34.9,
+      "learning_rate": 5.5105252525252525e-06,
+      "loss": 1.1082,
+      "step": 727500
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 5.500424242424243e-06,
+      "loss": 1.1089,
+      "step": 728000
+    },
+    {
+      "epoch": 34.95,
+      "learning_rate": 5.490323232323232e-06,
+      "loss": 1.1072,
+      "step": 728500
+    },
+    {
+      "epoch": 34.98,
+      "learning_rate": 5.4802222222222225e-06,
+      "loss": 1.1047,
+      "step": 729000
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 5.470121212121212e-06,
+      "loss": 1.1058,
+      "step": 729500
+    },
+    {
+      "epoch": 35.0,
+      "eval_accuracy": 0.7672661505696224,
+      "eval_loss": 0.9742304682731628,
+      "eval_runtime": 51.0173,
+      "eval_samples_per_second": 890.913,
+      "eval_steps_per_second": 13.936,
+      "step": 729505
+    },
+    {
+      "epoch": 35.02,
+      "learning_rate": 5.460020202020202e-06,
+      "loss": 1.1072,
+      "step": 730000
+    },
+    {
+      "epoch": 35.05,
+      "learning_rate": 5.449919191919192e-06,
+      "loss": 1.1063,
+      "step": 730500
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 5.439818181818183e-06,
+      "loss": 1.1088,
+      "step": 731000
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 5.429737373737373e-06,
+      "loss": 1.1077,
+      "step": 731500
+    },
+    {
+      "epoch": 35.12,
+      "learning_rate": 5.419636363636364e-06,
+      "loss": 1.1051,
+      "step": 732000
+    },
+    {
+      "epoch": 35.14,
+      "learning_rate": 5.409535353535355e-06,
+      "loss": 1.1067,
+      "step": 732500
+    },
+    {
+      "epoch": 35.17,
+      "learning_rate": 5.399434343434344e-06,
+      "loss": 1.1056,
+      "step": 733000
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 5.3893737373737376e-06,
+      "loss": 1.1074,
+      "step": 733500
+    },
+    {
+      "epoch": 35.22,
+      "learning_rate": 5.379272727272728e-06,
+      "loss": 1.1069,
+      "step": 734000
+    },
+    {
+      "epoch": 35.24,
+      "learning_rate": 5.369171717171717e-06,
+      "loss": 1.1083,
+      "step": 734500
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 5.359070707070707e-06,
+      "loss": 1.1049,
+      "step": 735000
+    },
+    {
+      "epoch": 35.29,
+      "learning_rate": 5.348969696969697e-06,
+      "loss": 1.1061,
+      "step": 735500
+    },
+    {
+      "epoch": 35.31,
+      "learning_rate": 5.3388686868686864e-06,
+      "loss": 1.1056,
+      "step": 736000
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 5.328767676767678e-06,
+      "loss": 1.1056,
+      "step": 736500
+    },
+    {
+      "epoch": 35.36,
+      "learning_rate": 5.318666666666668e-06,
+      "loss": 1.1073,
+      "step": 737000
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 5.308565656565657e-06,
+      "loss": 1.1051,
+      "step": 737500
+    },
+    {
+      "epoch": 35.41,
+      "learning_rate": 5.298464646464648e-06,
+      "loss": 1.1044,
+      "step": 738000
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 5.288363636363637e-06,
+      "loss": 1.1051,
+      "step": 738500
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 5.278282828282829e-06,
+      "loss": 1.1063,
+      "step": 739000
+    },
+    {
+      "epoch": 35.48,
+      "learning_rate": 5.268181818181819e-06,
+      "loss": 1.1071,
+      "step": 739500
+    },
+    {
+      "epoch": 35.5,
+      "learning_rate": 5.258080808080809e-06,
+      "loss": 1.1051,
+      "step": 740000
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 5.2479797979797984e-06,
+      "loss": 1.108,
+      "step": 740500
+    },
+    {
+      "epoch": 35.55,
+      "learning_rate": 5.237878787878789e-06,
+      "loss": 1.1039,
+      "step": 741000
+    },
+    {
+      "epoch": 35.58,
+      "learning_rate": 5.227777777777778e-06,
+      "loss": 1.1062,
+      "step": 741500
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 5.2176767676767685e-06,
+      "loss": 1.1048,
+      "step": 742000
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 5.20759595959596e-06,
+      "loss": 1.108,
+      "step": 742500
+    },
+    {
+      "epoch": 35.65,
+      "learning_rate": 5.19749494949495e-06,
+      "loss": 1.1056,
+      "step": 743000
+    },
+    {
+      "epoch": 35.67,
+      "learning_rate": 5.18739393939394e-06,
+      "loss": 1.1086,
+      "step": 743500
+    },
+    {
+      "epoch": 35.7,
+      "learning_rate": 5.17729292929293e-06,
+      "loss": 1.1061,
+      "step": 744000
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 5.167191919191919e-06,
+      "loss": 1.1064,
+      "step": 744500
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 5.1571111111111116e-06,
+      "loss": 1.106,
+      "step": 745000
+    },
+    {
+      "epoch": 35.77,
+      "learning_rate": 5.147010101010102e-06,
+      "loss": 1.1075,
+      "step": 745500
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 5.136909090909091e-06,
+      "loss": 1.1055,
+      "step": 746000
+    },
+    {
+      "epoch": 35.82,
+      "learning_rate": 5.126808080808081e-06,
+      "loss": 1.1083,
+      "step": 746500
+    },
+    {
+      "epoch": 35.84,
+      "learning_rate": 5.116707070707071e-06,
+      "loss": 1.1053,
+      "step": 747000
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 5.106626262626263e-06,
+      "loss": 1.1091,
+      "step": 747500
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 5.096525252525253e-06,
+      "loss": 1.1057,
+      "step": 748000
+    },
+    {
+      "epoch": 35.91,
+      "learning_rate": 5.086424242424243e-06,
+      "loss": 1.1054,
+      "step": 748500
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 5.0763232323232324e-06,
+      "loss": 1.1055,
+      "step": 749000
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 5.066242424242425e-06,
+      "loss": 1.1065,
+      "step": 749500
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 5.056141414141414e-06,
+      "loss": 1.1044,
+      "step": 750000
+    },
+    {
+      "epoch": 36.0,
+      "eval_accuracy": 0.766882168484176,
+      "eval_loss": 0.9757182598114014,
+      "eval_runtime": 50.5986,
+      "eval_samples_per_second": 898.285,
+      "eval_steps_per_second": 14.052,
+      "step": 750348
+    },
+    {
+      "epoch": 36.01,
+      "learning_rate": 5.046040404040404e-06,
+      "loss": 1.1085,
+      "step": 750500
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 5.035939393939394e-06,
+      "loss": 1.1058,
+      "step": 751000
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 5.025838383838384e-06,
+      "loss": 1.1058,
+      "step": 751500
+    },
+    {
+      "epoch": 36.08,
+      "learning_rate": 5.0157373737373736e-06,
+      "loss": 1.1066,
+      "step": 752000
+    },
+    {
+      "epoch": 36.1,
+      "learning_rate": 5.005636363636364e-06,
+      "loss": 1.1053,
+      "step": 752500
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 4.995555555555556e-06,
+      "loss": 1.1062,
+      "step": 753000
+    },
+    {
+      "epoch": 36.15,
+      "learning_rate": 4.9854747474747475e-06,
+      "loss": 1.1042,
+      "step": 753500
+    },
+    {
+      "epoch": 36.18,
+      "learning_rate": 4.975373737373738e-06,
+      "loss": 1.1049,
+      "step": 754000
+    },
+    {
+      "epoch": 36.2,
+      "learning_rate": 4.965272727272727e-06,
+      "loss": 1.1055,
+      "step": 754500
+    },
+    {
+      "epoch": 36.22,
+      "learning_rate": 4.9551717171717175e-06,
+      "loss": 1.1036,
+      "step": 755000
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 4.945070707070708e-06,
+      "loss": 1.1052,
+      "step": 755500
+    },
+    {
+      "epoch": 36.27,
+      "learning_rate": 4.934969696969697e-06,
+      "loss": 1.1037,
+      "step": 756000
+    },
+    {
+      "epoch": 36.3,
+      "learning_rate": 4.9248686868686875e-06,
+      "loss": 1.1033,
+      "step": 756500
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 4.91478787878788e-06,
+      "loss": 1.105,
+      "step": 757000
+    },
+    {
+      "epoch": 36.34,
+      "learning_rate": 4.904686868686869e-06,
+      "loss": 1.1048,
+      "step": 757500
+    },
+    {
+      "epoch": 36.37,
+      "learning_rate": 4.8945858585858595e-06,
+      "loss": 1.1056,
+      "step": 758000
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 4.884484848484849e-06,
+      "loss": 1.1059,
+      "step": 758500
+    },
+    {
+      "epoch": 36.42,
+      "learning_rate": 4.874383838383838e-06,
+      "loss": 1.106,
+      "step": 759000
+    },
+    {
+      "epoch": 36.44,
+      "learning_rate": 4.864282828282829e-06,
+      "loss": 1.1056,
+      "step": 759500
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 4.854181818181818e-06,
+      "loss": 1.102,
+      "step": 760000
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 4.844080808080808e-06,
+      "loss": 1.1059,
+      "step": 760500
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 4.833979797979798e-06,
+      "loss": 1.1043,
+      "step": 761000
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 4.823878787878788e-06,
+      "loss": 1.108,
+      "step": 761500
+    },
+    {
+      "epoch": 36.56,
+      "learning_rate": 4.81379797979798e-06,
+      "loss": 1.1018,
+      "step": 762000
+    },
+    {
+      "epoch": 36.58,
+      "learning_rate": 4.80369696969697e-06,
+      "loss": 1.1068,
+      "step": 762500
+    },
+    {
+      "epoch": 36.61,
+      "learning_rate": 4.79359595959596e-06,
+      "loss": 1.1035,
+      "step": 763000
+    },
+    {
+      "epoch": 36.63,
+      "learning_rate": 4.78349494949495e-06,
+      "loss": 1.1053,
+      "step": 763500
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 4.77339393939394e-06,
+      "loss": 1.105,
+      "step": 764000
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 4.76329292929293e-06,
+      "loss": 1.1057,
+      "step": 764500
+    },
+    {
+      "epoch": 36.7,
+      "learning_rate": 4.7531919191919195e-06,
+      "loss": 1.1063,
+      "step": 765000
+    },
+    {
+      "epoch": 36.73,
+      "learning_rate": 4.743111111111112e-06,
+      "loss": 1.1045,
+      "step": 765500
+    },
+    {
+      "epoch": 36.75,
+      "learning_rate": 4.733010101010101e-06,
+      "loss": 1.1043,
+      "step": 766000
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 4.7229090909090915e-06,
+      "loss": 1.1058,
+      "step": 766500
+    },
+    {
+      "epoch": 36.8,
+      "learning_rate": 4.712808080808081e-06,
+      "loss": 1.1062,
+      "step": 767000
+    },
+    {
+      "epoch": 36.82,
+      "learning_rate": 4.702707070707071e-06,
+      "loss": 1.1031,
+      "step": 767500
+    },
+    {
+      "epoch": 36.85,
+      "learning_rate": 4.692606060606061e-06,
+      "loss": 1.1049,
+      "step": 768000
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 4.682505050505051e-06,
+      "loss": 1.1038,
+      "step": 768500
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 4.672404040404041e-06,
+      "loss": 1.1034,
+      "step": 769000
+    },
+    {
+      "epoch": 36.92,
+      "learning_rate": 4.662303030303031e-06,
+      "loss": 1.1043,
+      "step": 769500
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 4.652222222222222e-06,
+      "loss": 1.1046,
+      "step": 770000
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 4.642141414141414e-06,
+      "loss": 1.1046,
+      "step": 770500
+    },
+    {
+      "epoch": 36.99,
+      "learning_rate": 4.632040404040405e-06,
+      "loss": 1.1015,
+      "step": 771000
+    },
+    {
+      "epoch": 37.0,
+      "eval_accuracy": 0.7671923426823506,
+      "eval_loss": 0.9729419350624084,
+      "eval_runtime": 50.781,
+      "eval_samples_per_second": 895.059,
+      "eval_steps_per_second": 14.001,
+      "step": 771191
+    },
+    {
+      "epoch": 37.01,
+      "learning_rate": 4.621939393939394e-06,
+      "loss": 1.1031,
+      "step": 771500
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 4.611838383838384e-06,
+      "loss": 1.104,
+      "step": 772000
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 4.601737373737374e-06,
+      "loss": 1.1035,
+      "step": 772500
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 4.591636363636364e-06,
+      "loss": 1.1045,
+      "step": 773000
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 4.5815555555555555e-06,
+      "loss": 1.1051,
+      "step": 773500
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 4.571454545454546e-06,
+      "loss": 1.1043,
+      "step": 774000
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 4.561353535353536e-06,
+      "loss": 1.1031,
+      "step": 774500
+    },
+    {
+      "epoch": 37.18,
+      "learning_rate": 4.5512525252525255e-06,
+      "loss": 1.1029,
+      "step": 775000
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 4.541151515151516e-06,
+      "loss": 1.1044,
+      "step": 775500
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 4.531050505050505e-06,
+      "loss": 1.1036,
+      "step": 776000
+    },
+    {
+      "epoch": 37.25,
+      "learning_rate": 4.5209696969696975e-06,
+      "loss": 1.1037,
+      "step": 776500
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 4.510868686868688e-06,
+      "loss": 1.1019,
+      "step": 777000
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 4.500767676767677e-06,
+      "loss": 1.1017,
+      "step": 777500
+    },
+    {
+      "epoch": 37.33,
+      "learning_rate": 4.490666666666667e-06,
+      "loss": 1.104,
+      "step": 778000
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 4.480565656565657e-06,
+      "loss": 1.1039,
+      "step": 778500
+    },
+    {
+      "epoch": 37.37,
+      "learning_rate": 4.470464646464646e-06,
+      "loss": 1.1051,
+      "step": 779000
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 4.460363636363637e-06,
+      "loss": 1.1039,
+      "step": 779500
+    },
+    {
+      "epoch": 37.42,
+      "learning_rate": 4.450262626262626e-06,
+      "loss": 1.1062,
+      "step": 780000
+    },
+    {
+      "epoch": 37.45,
+      "learning_rate": 4.440161616161616e-06,
+      "loss": 1.1048,
+      "step": 780500
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 4.430080808080809e-06,
+      "loss": 1.1059,
+      "step": 781000
+    },
+    {
+      "epoch": 37.49,
+      "learning_rate": 4.419979797979798e-06,
+      "loss": 1.1024,
+      "step": 781500
+    },
+    {
+      "epoch": 37.52,
+      "learning_rate": 4.409878787878788e-06,
+      "loss": 1.1054,
+      "step": 782000
+    },
+    {
+      "epoch": 37.54,
+      "learning_rate": 4.39979797979798e-06,
+      "loss": 1.1015,
+      "step": 782500
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 4.38969696969697e-06,
+      "loss": 1.1048,
+      "step": 783000
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 4.3795959595959595e-06,
+      "loss": 1.1026,
+      "step": 783500
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 4.36949494949495e-06,
+      "loss": 1.108,
+      "step": 784000
+    },
+    {
+      "epoch": 37.64,
+      "learning_rate": 4.35939393939394e-06,
+      "loss": 1.1038,
+      "step": 784500
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 4.3493131313131315e-06,
+      "loss": 1.104,
+      "step": 785000
+    },
+    {
+      "epoch": 37.69,
+      "learning_rate": 4.339212121212121e-06,
+      "loss": 1.1061,
+      "step": 785500
+    },
+    {
+      "epoch": 37.71,
+      "learning_rate": 4.329111111111112e-06,
+      "loss": 1.1046,
+      "step": 786000
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 4.3190303030303034e-06,
+      "loss": 1.1034,
+      "step": 786500
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 4.308929292929293e-06,
+      "loss": 1.1045,
+      "step": 787000
+    },
+    {
+      "epoch": 37.78,
+      "learning_rate": 4.298828282828283e-06,
+      "loss": 1.1022,
+      "step": 787500
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 4.2887272727272734e-06,
+      "loss": 1.1027,
+      "step": 788000
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 4.278626262626263e-06,
+      "loss": 1.1047,
+      "step": 788500
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 4.268525252525253e-06,
+      "loss": 1.1074,
+      "step": 789000
+    },
+    {
+      "epoch": 37.88,
+      "learning_rate": 4.258424242424243e-06,
+      "loss": 1.1041,
+      "step": 789500
+    },
+    {
+      "epoch": 37.9,
+      "learning_rate": 4.248323232323233e-06,
+      "loss": 1.102,
+      "step": 790000
+    },
+    {
+      "epoch": 37.93,
+      "learning_rate": 4.238222222222222e-06,
+      "loss": 1.1026,
+      "step": 790500
+    },
+    {
+      "epoch": 37.95,
+      "learning_rate": 4.228121212121213e-06,
+      "loss": 1.1043,
+      "step": 791000
+    },
+    {
+      "epoch": 37.97,
+      "learning_rate": 4.218020202020202e-06,
+      "loss": 1.1003,
+      "step": 791500
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 4.207919191919192e-06,
+      "loss": 1.1029,
+      "step": 792000
+    },
+    {
+      "epoch": 38.0,
+      "eval_accuracy": 0.7684662755933331,
+      "eval_loss": 0.9669340252876282,
+      "eval_runtime": 50.5571,
+      "eval_samples_per_second": 899.023,
+      "eval_steps_per_second": 14.063,
+      "step": 792034
+    },
+    {
+      "epoch": 38.02,
+      "learning_rate": 4.197838383838384e-06,
+      "loss": 1.104,
+      "step": 792500
+    },
+    {
+      "epoch": 38.05,
+      "learning_rate": 4.187737373737374e-06,
+      "loss": 1.0995,
+      "step": 793000
+    },
+    {
+      "epoch": 38.07,
+      "learning_rate": 4.177656565656566e-06,
+      "loss": 1.102,
+      "step": 793500
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 4.167555555555556e-06,
+      "loss": 1.1008,
+      "step": 794000
+    },
+    {
+      "epoch": 38.12,
+      "learning_rate": 4.157454545454546e-06,
+      "loss": 1.1021,
+      "step": 794500
+    },
+    {
+      "epoch": 38.14,
+      "learning_rate": 4.1473535353535355e-06,
+      "loss": 1.1018,
+      "step": 795000
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 4.137252525252526e-06,
+      "loss": 1.1028,
+      "step": 795500
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 4.127151515151516e-06,
+      "loss": 1.1056,
+      "step": 796000
+    },
+    {
+      "epoch": 38.21,
+      "learning_rate": 4.1170707070707074e-06,
+      "loss": 1.1035,
+      "step": 796500
+    },
+    {
+      "epoch": 38.24,
+      "learning_rate": 4.106969696969697e-06,
+      "loss": 1.1025,
+      "step": 797000
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 4.096868686868687e-06,
+      "loss": 1.1021,
+      "step": 797500
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 4.0867676767676774e-06,
+      "loss": 1.1033,
+      "step": 798000
+    },
+    {
+      "epoch": 38.31,
+      "learning_rate": 4.076666666666667e-06,
+      "loss": 1.1028,
+      "step": 798500
+    },
+    {
+      "epoch": 38.33,
+      "learning_rate": 4.066585858585859e-06,
+      "loss": 1.1022,
+      "step": 799000
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 4.056484848484849e-06,
+      "loss": 1.1024,
+      "step": 799500
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 4.046383838383839e-06,
+      "loss": 1.1055,
+      "step": 800000
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 4.036282828282828e-06,
+      "loss": 1.1006,
+      "step": 800500
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 4.026181818181819e-06,
+      "loss": 1.1047,
+      "step": 801000
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 4.016080808080808e-06,
+      "loss": 1.1034,
+      "step": 801500
+    },
+    {
+      "epoch": 38.48,
+      "learning_rate": 4.005979797979798e-06,
+      "loss": 1.1015,
+      "step": 802000
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 3.995878787878788e-06,
+      "loss": 1.1011,
+      "step": 802500
+    },
+    {
+      "epoch": 38.53,
+      "learning_rate": 3.985777777777778e-06,
+      "loss": 1.1019,
+      "step": 803000
+    },
+    {
+      "epoch": 38.55,
+      "learning_rate": 3.97569696969697e-06,
+      "loss": 1.1008,
+      "step": 803500
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 3.96559595959596e-06,
+      "loss": 1.1044,
+      "step": 804000
+    },
+    {
+      "epoch": 38.6,
+      "learning_rate": 3.955515151515152e-06,
+      "loss": 1.1031,
+      "step": 804500
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 3.945414141414141e-06,
+      "loss": 1.105,
+      "step": 805000
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 3.935333333333334e-06,
+      "loss": 1.1026,
+      "step": 805500
+    },
+    {
+      "epoch": 38.67,
+      "learning_rate": 3.925232323232324e-06,
+      "loss": 1.1023,
+      "step": 806000
+    },
+    {
+      "epoch": 38.69,
+      "learning_rate": 3.915131313131313e-06,
+      "loss": 1.1024,
+      "step": 806500
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 3.905030303030304e-06,
+      "loss": 1.1033,
+      "step": 807000
+    },
+    {
+      "epoch": 38.74,
+      "learning_rate": 3.894929292929293e-06,
+      "loss": 1.1009,
+      "step": 807500
+    },
+    {
+      "epoch": 38.77,
+      "learning_rate": 3.8848282828282826e-06,
+      "loss": 1.1013,
+      "step": 808000
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 3.874727272727273e-06,
+      "loss": 1.1042,
+      "step": 808500
+    },
+    {
+      "epoch": 38.81,
+      "learning_rate": 3.864626262626263e-06,
+      "loss": 1.1034,
+      "step": 809000
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 3.854525252525253e-06,
+      "loss": 1.1021,
+      "step": 809500
+    },
+    {
+      "epoch": 38.86,
+      "learning_rate": 3.844444444444445e-06,
+      "loss": 1.1038,
+      "step": 810000
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 3.834343434343435e-06,
+      "loss": 1.1019,
+      "step": 810500
+    },
+    {
+      "epoch": 38.91,
+      "learning_rate": 3.8242424242424245e-06,
+      "loss": 1.0994,
+      "step": 811000
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 3.8141616161616164e-06,
+      "loss": 1.1025,
+      "step": 811500
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 3.8040606060606062e-06,
+      "loss": 1.1026,
+      "step": 812000
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 3.7939595959595965e-06,
+      "loss": 1.1034,
+      "step": 812500
+    },
+    {
+      "epoch": 39.0,
+      "eval_accuracy": 0.768050772184178,
+      "eval_loss": 0.9690855145454407,
+      "eval_runtime": 50.8414,
+      "eval_samples_per_second": 893.996,
+      "eval_steps_per_second": 13.985,
+      "step": 812877
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 3.7838585858585864e-06,
+      "loss": 1.1041,
+      "step": 813000
+    },
+    {
+      "epoch": 39.03,
+      "learning_rate": 3.7737575757575762e-06,
+      "loss": 1.1015,
+      "step": 813500
+    },
+    {
+      "epoch": 39.05,
+      "learning_rate": 3.763656565656566e-06,
+      "loss": 1.1018,
+      "step": 814000
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 3.753555555555556e-06,
+      "loss": 1.1039,
+      "step": 814500
+    },
+    {
+      "epoch": 39.1,
+      "learning_rate": 3.743454545454546e-06,
+      "loss": 1.0995,
+      "step": 815000
+    },
+    {
+      "epoch": 39.13,
+      "learning_rate": 3.7333535353535357e-06,
+      "loss": 1.1006,
+      "step": 815500
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 3.723252525252525e-06,
+      "loss": 1.1017,
+      "step": 816000
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 3.7131717171717174e-06,
+      "loss": 1.1069,
+      "step": 816500
+    },
+    {
+      "epoch": 39.2,
+      "learning_rate": 3.7030707070707072e-06,
+      "loss": 1.1032,
+      "step": 817000
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 3.692969696969697e-06,
+      "loss": 1.1027,
+      "step": 817500
+    },
+    {
+      "epoch": 39.25,
+      "learning_rate": 3.682868686868687e-06,
+      "loss": 1.103,
+      "step": 818000
+    },
+    {
+      "epoch": 39.27,
+      "learning_rate": 3.6727676767676773e-06,
+      "loss": 1.101,
+      "step": 818500
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 3.662666666666667e-06,
+      "loss": 1.1044,
+      "step": 819000
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 3.652565656565657e-06,
+      "loss": 1.1011,
+      "step": 819500
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 3.642464646464647e-06,
+      "loss": 1.1013,
+      "step": 820000
+    },
+    {
+      "epoch": 39.37,
+      "learning_rate": 3.632383838383839e-06,
+      "loss": 1.1027,
+      "step": 820500
+    },
+    {
+      "epoch": 39.39,
+      "learning_rate": 3.6222828282828285e-06,
+      "loss": 1.1041,
+      "step": 821000
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 3.6121818181818184e-06,
+      "loss": 1.1043,
+      "step": 821500
+    },
+    {
+      "epoch": 39.44,
+      "learning_rate": 3.6020808080808083e-06,
+      "loss": 1.1035,
+      "step": 822000
+    },
+    {
+      "epoch": 39.46,
+      "learning_rate": 3.591979797979798e-06,
+      "loss": 1.1018,
+      "step": 822500
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 3.5818989898989904e-06,
+      "loss": 1.1007,
+      "step": 823000
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 3.5718181818181818e-06,
+      "loss": 1.1017,
+      "step": 823500
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 3.561717171717172e-06,
+      "loss": 1.103,
+      "step": 824000
+    },
+    {
+      "epoch": 39.56,
+      "learning_rate": 3.551616161616162e-06,
+      "loss": 1.1034,
+      "step": 824500
+    },
+    {
+      "epoch": 39.58,
+      "learning_rate": 3.5415151515151518e-06,
+      "loss": 1.1018,
+      "step": 825000
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 3.531434343434344e-06,
+      "loss": 1.1005,
+      "step": 825500
+    },
+    {
+      "epoch": 39.63,
+      "learning_rate": 3.521333333333334e-06,
+      "loss": 1.102,
+      "step": 826000
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 3.5112323232323238e-06,
+      "loss": 1.1012,
+      "step": 826500
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 3.5011313131313136e-06,
+      "loss": 1.1017,
+      "step": 827000
+    },
+    {
+      "epoch": 39.7,
+      "learning_rate": 3.4910303030303035e-06,
+      "loss": 1.1041,
+      "step": 827500
+    },
+    {
+      "epoch": 39.73,
+      "learning_rate": 3.4809292929292933e-06,
+      "loss": 1.1006,
+      "step": 828000
+    },
+    {
+      "epoch": 39.75,
+      "learning_rate": 3.4708282828282828e-06,
+      "loss": 1.1021,
+      "step": 828500
+    },
+    {
+      "epoch": 39.77,
+      "learning_rate": 3.4607272727272727e-06,
+      "loss": 1.1016,
+      "step": 829000
+    },
+    {
+      "epoch": 39.8,
+      "learning_rate": 3.4506262626262625e-06,
+      "loss": 1.1025,
+      "step": 829500
+    },
+    {
+      "epoch": 39.82,
+      "learning_rate": 3.4405454545454548e-06,
+      "loss": 1.1039,
+      "step": 830000
+    },
+    {
+      "epoch": 39.85,
+      "learning_rate": 3.4304444444444446e-06,
+      "loss": 1.0982,
+      "step": 830500
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 3.4203434343434345e-06,
+      "loss": 1.1017,
+      "step": 831000
+    },
+    {
+      "epoch": 39.89,
+      "learning_rate": 3.4102424242424248e-06,
+      "loss": 1.102,
+      "step": 831500
+    },
+    {
+      "epoch": 39.92,
+      "learning_rate": 3.4001414141414146e-06,
+      "loss": 1.1015,
+      "step": 832000
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 3.3900404040404045e-06,
+      "loss": 1.1021,
+      "step": 832500
+    },
+    {
+      "epoch": 39.97,
+      "learning_rate": 3.379959595959596e-06,
+      "loss": 1.1,
+      "step": 833000
+    },
+    {
+      "epoch": 39.99,
+      "learning_rate": 3.369858585858586e-06,
+      "loss": 1.1001,
+      "step": 833500
+    },
+    {
+      "epoch": 40.0,
+      "eval_accuracy": 0.7678935195953602,
+      "eval_loss": 0.9703824520111084,
+      "eval_runtime": 50.9782,
+      "eval_samples_per_second": 891.597,
+      "eval_steps_per_second": 13.947,
+      "step": 833720
+    },
+    {
+      "epoch": 40.01,
+      "learning_rate": 3.359757575757576e-06,
+      "loss": 1.1006,
+      "step": 834000
+    },
+    {
+      "epoch": 40.04,
+      "learning_rate": 3.349656565656566e-06,
+      "loss": 1.1025,
+      "step": 834500
+    },
+    {
+      "epoch": 40.06,
+      "learning_rate": 3.3395555555555558e-06,
+      "loss": 1.1003,
+      "step": 835000
+    },
+    {
+      "epoch": 40.09,
+      "learning_rate": 3.3294545454545456e-06,
+      "loss": 1.1019,
+      "step": 835500
+    },
+    {
+      "epoch": 40.11,
+      "learning_rate": 3.3193535353535355e-06,
+      "loss": 1.1037,
+      "step": 836000
+    },
+    {
+      "epoch": 40.13,
+      "learning_rate": 3.3092727272727278e-06,
+      "loss": 1.1024,
+      "step": 836500
+    },
+    {
+      "epoch": 40.16,
+      "learning_rate": 3.2991717171717176e-06,
+      "loss": 1.1012,
+      "step": 837000
+    },
+    {
+      "epoch": 40.18,
+      "learning_rate": 3.2890707070707075e-06,
+      "loss": 1.1015,
+      "step": 837500
+    },
+    {
+      "epoch": 40.21,
+      "learning_rate": 3.2789696969696973e-06,
+      "loss": 1.1011,
+      "step": 838000
+    },
+    {
+      "epoch": 40.23,
+      "learning_rate": 3.2688686868686868e-06,
+      "loss": 1.1009,
+      "step": 838500
+    },
+    {
+      "epoch": 40.25,
+      "learning_rate": 3.2587676767676766e-06,
+      "loss": 1.1005,
+      "step": 839000
+    },
+    {
+      "epoch": 40.28,
+      "learning_rate": 3.2486666666666674e-06,
+      "loss": 1.1027,
+      "step": 839500
+    },
+    {
+      "epoch": 40.3,
+      "learning_rate": 3.2385656565656572e-06,
+      "loss": 1.0999,
+      "step": 840000
+    },
+    {
+      "epoch": 40.33,
+      "learning_rate": 3.2284646464646467e-06,
+      "loss": 1.1024,
+      "step": 840500
+    },
+    {
+      "epoch": 40.35,
+      "learning_rate": 3.2183838383838385e-06,
+      "loss": 1.1016,
+      "step": 841000
+    },
+    {
+      "epoch": 40.37,
+      "learning_rate": 3.2082828282828288e-06,
+      "loss": 1.1018,
+      "step": 841500
+    },
+    {
+      "epoch": 40.4,
+      "learning_rate": 3.19820202020202e-06,
+      "loss": 1.1004,
+      "step": 842000
+    },
+    {
+      "epoch": 40.42,
+      "learning_rate": 3.18810101010101e-06,
+      "loss": 1.1021,
+      "step": 842500
+    },
+    {
+      "epoch": 40.45,
+      "learning_rate": 3.1780000000000003e-06,
+      "loss": 1.1022,
+      "step": 843000
+    },
+    {
+      "epoch": 40.47,
+      "learning_rate": 3.16789898989899e-06,
+      "loss": 1.1004,
+      "step": 843500
+    },
+    {
+      "epoch": 40.49,
+      "learning_rate": 3.15779797979798e-06,
+      "loss": 1.1038,
+      "step": 844000
+    },
+    {
+      "epoch": 40.52,
+      "learning_rate": 3.14769696969697e-06,
+      "loss": 1.0992,
+      "step": 844500
+    },
+    {
+      "epoch": 40.54,
+      "learning_rate": 3.1375959595959598e-06,
+      "loss": 1.1028,
+      "step": 845000
+    },
+    {
+      "epoch": 40.57,
+      "learning_rate": 3.1274949494949496e-06,
+      "loss": 1.1016,
+      "step": 845500
+    },
+    {
+      "epoch": 40.59,
+      "learning_rate": 3.117414141414142e-06,
+      "loss": 1.0997,
+      "step": 846000
+    },
+    {
+      "epoch": 40.61,
+      "learning_rate": 3.1073131313131317e-06,
+      "loss": 1.1014,
+      "step": 846500
+    },
+    {
+      "epoch": 40.64,
+      "learning_rate": 3.0972121212121216e-06,
+      "loss": 1.0996,
+      "step": 847000
+    },
+    {
+      "epoch": 40.66,
+      "learning_rate": 3.0871111111111115e-06,
+      "loss": 1.1008,
+      "step": 847500
+    },
+    {
+      "epoch": 40.69,
+      "learning_rate": 3.0770303030303033e-06,
+      "loss": 1.1006,
+      "step": 848000
+    },
+    {
+      "epoch": 40.71,
+      "learning_rate": 3.066929292929293e-06,
+      "loss": 1.098,
+      "step": 848500
+    },
+    {
+      "epoch": 40.73,
+      "learning_rate": 3.056828282828283e-06,
+      "loss": 1.1,
+      "step": 849000
+    },
+    {
+      "epoch": 40.76,
+      "learning_rate": 3.046727272727273e-06,
+      "loss": 1.1025,
+      "step": 849500
+    },
+    {
+      "epoch": 40.78,
+      "learning_rate": 3.0366262626262628e-06,
+      "loss": 1.0997,
+      "step": 850000
+    },
+    {
+      "epoch": 40.81,
+      "learning_rate": 3.0265252525252526e-06,
+      "loss": 1.0983,
+      "step": 850500
+    },
+    {
+      "epoch": 40.83,
+      "learning_rate": 3.0164444444444444e-06,
+      "loss": 1.1014,
+      "step": 851000
+    },
+    {
+      "epoch": 40.85,
+      "learning_rate": 3.0063434343434343e-06,
+      "loss": 1.1006,
+      "step": 851500
+    },
+    {
+      "epoch": 40.88,
+      "learning_rate": 2.996242424242424e-06,
+      "loss": 1.0997,
+      "step": 852000
+    },
+    {
+      "epoch": 40.9,
+      "learning_rate": 2.986141414141414e-06,
+      "loss": 1.1014,
+      "step": 852500
+    },
+    {
+      "epoch": 40.93,
+      "learning_rate": 2.9760404040404043e-06,
+      "loss": 1.1005,
+      "step": 853000
+    },
+    {
+      "epoch": 40.95,
+      "learning_rate": 2.965959595959596e-06,
+      "loss": 1.1028,
+      "step": 853500
+    },
+    {
+      "epoch": 40.97,
+      "learning_rate": 2.955858585858586e-06,
+      "loss": 1.0978,
+      "step": 854000
+    },
+    {
+      "epoch": 41.0,
+      "learning_rate": 2.9457575757575763e-06,
+      "loss": 1.1007,
+      "step": 854500
+    },
+    {
+      "epoch": 41.0,
+      "eval_accuracy": 0.768079747725581,
+      "eval_loss": 0.9693915843963623,
+      "eval_runtime": 51.3143,
+      "eval_samples_per_second": 885.757,
+      "eval_steps_per_second": 13.856,
+      "step": 854563
+    },
+    {
+      "epoch": 41.02,
+      "learning_rate": 2.935656565656566e-06,
+      "loss": 1.1008,
+      "step": 855000
+    },
+    {
+      "epoch": 41.04,
+      "learning_rate": 2.925555555555556e-06,
+      "loss": 1.0994,
+      "step": 855500
+    },
+    {
+      "epoch": 41.07,
+      "learning_rate": 2.915454545454546e-06,
+      "loss": 1.1008,
+      "step": 856000
+    },
+    {
+      "epoch": 41.09,
+      "learning_rate": 2.9053737373737377e-06,
+      "loss": 1.101,
+      "step": 856500
+    },
+    {
+      "epoch": 41.12,
+      "learning_rate": 2.8952929292929295e-06,
+      "loss": 1.0995,
+      "step": 857000
+    },
+    {
+      "epoch": 41.14,
+      "learning_rate": 2.8851919191919194e-06,
+      "loss": 1.1004,
+      "step": 857500
+    },
+    {
+      "epoch": 41.16,
+      "learning_rate": 2.8750909090909097e-06,
+      "loss": 1.1012,
+      "step": 858000
+    },
+    {
+      "epoch": 41.19,
+      "learning_rate": 2.8649898989898995e-06,
+      "loss": 1.0994,
+      "step": 858500
+    },
+    {
+      "epoch": 41.21,
+      "learning_rate": 2.8548888888888894e-06,
+      "loss": 1.0996,
+      "step": 859000
+    },
+    {
+      "epoch": 41.24,
+      "learning_rate": 2.8447878787878793e-06,
+      "loss": 1.1025,
+      "step": 859500
+    },
+    {
+      "epoch": 41.26,
+      "learning_rate": 2.834686868686869e-06,
+      "loss": 1.0996,
+      "step": 860000
+    },
+    {
+      "epoch": 41.28,
+      "learning_rate": 2.8245858585858586e-06,
+      "loss": 1.0992,
+      "step": 860500
+    },
+    {
+      "epoch": 41.31,
+      "learning_rate": 2.8144848484848484e-06,
+      "loss": 1.1009,
+      "step": 861000
+    },
+    {
+      "epoch": 41.33,
+      "learning_rate": 2.8044040404040407e-06,
+      "loss": 1.0985,
+      "step": 861500
+    },
+    {
+      "epoch": 41.36,
+      "learning_rate": 2.7943030303030305e-06,
+      "loss": 1.0994,
+      "step": 862000
+    },
+    {
+      "epoch": 41.38,
+      "learning_rate": 2.7842020202020204e-06,
+      "loss": 1.101,
+      "step": 862500
+    },
+    {
+      "epoch": 41.4,
+      "learning_rate": 2.7741010101010103e-06,
+      "loss": 1.0993,
+      "step": 863000
+    },
+    {
+      "epoch": 41.43,
+      "learning_rate": 2.764e-06,
+      "loss": 1.0999,
+      "step": 863500
+    },
+    {
+      "epoch": 41.45,
+      "learning_rate": 2.753919191919192e-06,
+      "loss": 1.1017,
+      "step": 864000
+    },
+    {
+      "epoch": 41.48,
+      "learning_rate": 2.743818181818182e-06,
+      "loss": 1.1004,
+      "step": 864500
+    },
+    {
+      "epoch": 41.5,
+      "learning_rate": 2.7337171717171717e-06,
+      "loss": 1.0988,
+      "step": 865000
+    },
+    {
+      "epoch": 41.52,
+      "learning_rate": 2.7236161616161615e-06,
+      "loss": 1.1008,
+      "step": 865500
+    },
+    {
+      "epoch": 41.55,
+      "learning_rate": 2.713515151515152e-06,
+      "loss": 1.1017,
+      "step": 866000
+    },
+    {
+      "epoch": 41.57,
+      "learning_rate": 2.7034141414141417e-06,
+      "loss": 1.0998,
+      "step": 866500
+    },
+    {
+      "epoch": 41.6,
+      "learning_rate": 2.6933333333333335e-06,
+      "loss": 1.101,
+      "step": 867000
+    },
+    {
+      "epoch": 41.62,
+      "learning_rate": 2.6832323232323234e-06,
+      "loss": 1.0979,
+      "step": 867500
+    },
+    {
+      "epoch": 41.64,
+      "learning_rate": 2.6731313131313137e-06,
+      "loss": 1.0981,
+      "step": 868000
+    },
+    {
+      "epoch": 41.67,
+      "learning_rate": 2.6630303030303035e-06,
+      "loss": 1.0989,
+      "step": 868500
+    },
+    {
+      "epoch": 41.69,
+      "learning_rate": 2.6529292929292934e-06,
+      "loss": 1.0983,
+      "step": 869000
+    },
+    {
+      "epoch": 41.72,
+      "learning_rate": 2.6428282828282833e-06,
+      "loss": 1.099,
+      "step": 869500
+    },
+    {
+      "epoch": 41.74,
+      "learning_rate": 2.6327272727272727e-06,
+      "loss": 1.101,
+      "step": 870000
+    },
+    {
+      "epoch": 41.76,
+      "learning_rate": 2.6226262626262626e-06,
+      "loss": 1.1013,
+      "step": 870500
+    },
+    {
+      "epoch": 41.79,
+      "learning_rate": 2.612545454545455e-06,
+      "loss": 1.1014,
+      "step": 871000
+    },
+    {
+      "epoch": 41.81,
+      "learning_rate": 2.6024444444444447e-06,
+      "loss": 1.0979,
+      "step": 871500
+    },
+    {
+      "epoch": 41.84,
+      "learning_rate": 2.5923434343434345e-06,
+      "loss": 1.1002,
+      "step": 872000
+    },
+    {
+      "epoch": 41.86,
+      "learning_rate": 2.5822424242424244e-06,
+      "loss": 1.0998,
+      "step": 872500
+    },
+    {
+      "epoch": 41.88,
+      "learning_rate": 2.5721414141414143e-06,
+      "loss": 1.1002,
+      "step": 873000
+    },
+    {
+      "epoch": 41.91,
+      "learning_rate": 2.562040404040404e-06,
+      "loss": 1.1003,
+      "step": 873500
+    },
+    {
+      "epoch": 41.93,
+      "learning_rate": 2.551959595959596e-06,
+      "loss": 1.0956,
+      "step": 874000
+    },
+    {
+      "epoch": 41.96,
+      "learning_rate": 2.541878787878788e-06,
+      "loss": 1.0978,
+      "step": 874500
+    },
+    {
+      "epoch": 41.98,
+      "learning_rate": 2.531777777777778e-06,
+      "loss": 1.1024,
+      "step": 875000
+    },
+    {
+      "epoch": 42.0,
+      "eval_accuracy": 0.7688446251390236,
+      "eval_loss": 0.9656499624252319,
+      "eval_runtime": 51.3866,
+      "eval_samples_per_second": 884.511,
+      "eval_steps_per_second": 13.836,
+      "step": 875406
+    },
+    {
+      "epoch": 42.0,
+      "learning_rate": 2.521676767676768e-06,
+      "loss": 1.1015,
+      "step": 875500
+    },
+    {
+      "epoch": 42.03,
+      "learning_rate": 2.511575757575758e-06,
+      "loss": 1.1005,
+      "step": 876000
+    },
+    {
+      "epoch": 42.05,
+      "learning_rate": 2.5014747474747477e-06,
+      "loss": 1.0999,
+      "step": 876500
+    },
+    {
+      "epoch": 42.08,
+      "learning_rate": 2.4913737373737375e-06,
+      "loss": 1.0978,
+      "step": 877000
+    },
+    {
+      "epoch": 42.1,
+      "learning_rate": 2.4812727272727274e-06,
+      "loss": 1.0988,
+      "step": 877500
+    },
+    {
+      "epoch": 42.12,
+      "learning_rate": 2.4711919191919196e-06,
+      "loss": 1.0996,
+      "step": 878000
+    },
+    {
+      "epoch": 42.15,
+      "learning_rate": 2.4610909090909095e-06,
+      "loss": 1.0998,
+      "step": 878500
+    },
+    {
+      "epoch": 42.17,
+      "learning_rate": 2.4509898989898994e-06,
+      "loss": 1.1002,
+      "step": 879000
+    },
+    {
+      "epoch": 42.2,
+      "learning_rate": 2.4408888888888892e-06,
+      "loss": 1.0961,
+      "step": 879500
+    },
+    {
+      "epoch": 42.22,
+      "learning_rate": 2.4307878787878787e-06,
+      "loss": 1.1008,
+      "step": 880000
+    },
+    {
+      "epoch": 42.24,
+      "learning_rate": 2.420707070707071e-06,
+      "loss": 1.1016,
+      "step": 880500
+    },
+    {
+      "epoch": 42.27,
+      "learning_rate": 2.4106060606060608e-06,
+      "loss": 1.0985,
+      "step": 881000
+    },
+    {
+      "epoch": 42.29,
+      "learning_rate": 2.4005252525252526e-06,
+      "loss": 1.0997,
+      "step": 881500
+    },
+    {
+      "epoch": 42.32,
+      "learning_rate": 2.390424242424243e-06,
+      "loss": 1.0993,
+      "step": 882000
+    },
+    {
+      "epoch": 42.34,
+      "learning_rate": 2.3803232323232327e-06,
+      "loss": 1.0986,
+      "step": 882500
+    },
+    {
+      "epoch": 42.36,
+      "learning_rate": 2.370222222222222e-06,
+      "loss": 1.0998,
+      "step": 883000
+    },
+    {
+      "epoch": 42.39,
+      "learning_rate": 2.360121212121212e-06,
+      "loss": 1.1,
+      "step": 883500
+    },
+    {
+      "epoch": 42.41,
+      "learning_rate": 2.3500202020202023e-06,
+      "loss": 1.1007,
+      "step": 884000
+    },
+    {
+      "epoch": 42.44,
+      "learning_rate": 2.339919191919192e-06,
+      "loss": 1.0993,
+      "step": 884500
+    },
+    {
+      "epoch": 42.46,
+      "learning_rate": 2.329818181818182e-06,
+      "loss": 1.1003,
+      "step": 885000
+    },
+    {
+      "epoch": 42.48,
+      "learning_rate": 2.319717171717172e-06,
+      "loss": 1.0975,
+      "step": 885500
+    },
+    {
+      "epoch": 42.51,
+      "learning_rate": 2.3096161616161618e-06,
+      "loss": 1.0996,
+      "step": 886000
+    },
+    {
+      "epoch": 42.53,
+      "learning_rate": 2.2995151515151516e-06,
+      "loss": 1.0995,
+      "step": 886500
+    },
+    {
+      "epoch": 42.56,
+      "learning_rate": 2.2894141414141415e-06,
+      "loss": 1.0987,
+      "step": 887000
+    },
+    {
+      "epoch": 42.58,
+      "learning_rate": 2.2793333333333333e-06,
+      "loss": 1.0962,
+      "step": 887500
+    },
+    {
+      "epoch": 42.6,
+      "learning_rate": 2.2692323232323236e-06,
+      "loss": 1.0972,
+      "step": 888000
+    },
+    {
+      "epoch": 42.63,
+      "learning_rate": 2.2591313131313135e-06,
+      "loss": 1.0995,
+      "step": 888500
+    },
+    {
+      "epoch": 42.65,
+      "learning_rate": 2.2490303030303033e-06,
+      "loss": 1.0967,
+      "step": 889000
+    },
+    {
+      "epoch": 42.68,
+      "learning_rate": 2.238929292929293e-06,
+      "loss": 1.1017,
+      "step": 889500
+    },
+    {
+      "epoch": 42.7,
+      "learning_rate": 2.228848484848485e-06,
+      "loss": 1.098,
+      "step": 890000
+    },
+    {
+      "epoch": 42.72,
+      "learning_rate": 2.218747474747475e-06,
+      "loss": 1.0998,
+      "step": 890500
+    },
+    {
+      "epoch": 42.75,
+      "learning_rate": 2.2086666666666667e-06,
+      "loss": 1.0981,
+      "step": 891000
+    },
+    {
+      "epoch": 42.77,
+      "learning_rate": 2.198565656565657e-06,
+      "loss": 1.0979,
+      "step": 891500
+    },
+    {
+      "epoch": 42.8,
+      "learning_rate": 2.188464646464647e-06,
+      "loss": 1.1021,
+      "step": 892000
+    },
+    {
+      "epoch": 42.82,
+      "learning_rate": 2.1783636363636363e-06,
+      "loss": 1.0972,
+      "step": 892500
+    },
+    {
+      "epoch": 42.84,
+      "learning_rate": 2.168262626262626e-06,
+      "loss": 1.1008,
+      "step": 893000
+    },
+    {
+      "epoch": 42.87,
+      "learning_rate": 2.1581818181818184e-06,
+      "loss": 1.0989,
+      "step": 893500
+    },
+    {
+      "epoch": 42.89,
+      "learning_rate": 2.1480808080808083e-06,
+      "loss": 1.1022,
+      "step": 894000
+    },
+    {
+      "epoch": 42.92,
+      "learning_rate": 2.137979797979798e-06,
+      "loss": 1.1021,
+      "step": 894500
+    },
+    {
+      "epoch": 42.94,
+      "learning_rate": 2.127878787878788e-06,
+      "loss": 1.0994,
+      "step": 895000
+    },
+    {
+      "epoch": 42.96,
+      "learning_rate": 2.117777777777778e-06,
+      "loss": 1.1022,
+      "step": 895500
+    },
+    {
+      "epoch": 42.99,
+      "learning_rate": 2.1076969696969697e-06,
+      "loss": 1.1014,
+      "step": 896000
+    },
+    {
+      "epoch": 43.0,
+      "eval_accuracy": 0.7690242645334654,
+      "eval_loss": 0.9640320539474487,
+      "eval_runtime": 52.2442,
+      "eval_samples_per_second": 869.992,
+      "eval_steps_per_second": 13.609,
+      "step": 896249
+    },
+    {
+      "epoch": 43.01,
+      "learning_rate": 2.0975959595959596e-06,
+      "loss": 1.0983,
+      "step": 896500
+    },
+    {
+      "epoch": 43.04,
+      "learning_rate": 2.08749494949495e-06,
+      "loss": 1.0992,
+      "step": 897000
+    },
+    {
+      "epoch": 43.06,
+      "learning_rate": 2.0773939393939397e-06,
+      "loss": 1.1004,
+      "step": 897500
+    },
+    {
+      "epoch": 43.08,
+      "learning_rate": 2.0672929292929296e-06,
+      "loss": 1.0975,
+      "step": 898000
+    },
+    {
+      "epoch": 43.11,
+      "learning_rate": 2.0571919191919194e-06,
+      "loss": 1.0955,
+      "step": 898500
+    },
+    {
+      "epoch": 43.13,
+      "learning_rate": 2.0471111111111113e-06,
+      "loss": 1.0947,
+      "step": 899000
+    },
+    {
+      "epoch": 43.16,
+      "learning_rate": 2.037010101010101e-06,
+      "loss": 1.0982,
+      "step": 899500
+    },
+    {
+      "epoch": 43.18,
+      "learning_rate": 2.026909090909091e-06,
+      "loss": 1.0982,
+      "step": 900000
+    },
+    {
+      "epoch": 43.2,
+      "learning_rate": 2.016808080808081e-06,
+      "loss": 1.0978,
+      "step": 900500
+    },
+    {
+      "epoch": 43.23,
+      "learning_rate": 2.006727272727273e-06,
+      "loss": 1.0986,
+      "step": 901000
+    },
+    {
+      "epoch": 43.25,
+      "learning_rate": 1.996626262626263e-06,
+      "loss": 1.0981,
+      "step": 901500
+    },
+    {
+      "epoch": 43.28,
+      "learning_rate": 1.986525252525253e-06,
+      "loss": 1.0974,
+      "step": 902000
+    },
+    {
+      "epoch": 43.3,
+      "learning_rate": 1.9764242424242423e-06,
+      "loss": 1.099,
+      "step": 902500
+    },
+    {
+      "epoch": 43.32,
+      "learning_rate": 1.9663232323232326e-06,
+      "loss": 1.0997,
+      "step": 903000
+    },
+    {
+      "epoch": 43.35,
+      "learning_rate": 1.9562222222222224e-06,
+      "loss": 1.098,
+      "step": 903500
+    },
+    {
+      "epoch": 43.37,
+      "learning_rate": 1.9461212121212123e-06,
+      "loss": 1.0996,
+      "step": 904000
+    },
+    {
+      "epoch": 43.4,
+      "learning_rate": 1.936020202020202e-06,
+      "loss": 1.098,
+      "step": 904500
+    },
+    {
+      "epoch": 43.42,
+      "learning_rate": 1.925939393939394e-06,
+      "loss": 1.0993,
+      "step": 905000
+    },
+    {
+      "epoch": 43.44,
+      "learning_rate": 1.915838383838384e-06,
+      "loss": 1.099,
+      "step": 905500
+    },
+    {
+      "epoch": 43.47,
+      "learning_rate": 1.9057575757575759e-06,
+      "loss": 1.0967,
+      "step": 906000
+    },
+    {
+      "epoch": 43.49,
+      "learning_rate": 1.895656565656566e-06,
+      "loss": 1.0977,
+      "step": 906500
+    },
+    {
+      "epoch": 43.52,
+      "learning_rate": 1.8855555555555558e-06,
+      "loss": 1.0994,
+      "step": 907000
+    },
+    {
+      "epoch": 43.54,
+      "learning_rate": 1.8754545454545457e-06,
+      "loss": 1.0999,
+      "step": 907500
+    },
+    {
+      "epoch": 43.56,
+      "learning_rate": 1.8653535353535353e-06,
+      "loss": 1.1009,
+      "step": 908000
+    },
+    {
+      "epoch": 43.59,
+      "learning_rate": 1.8552727272727276e-06,
+      "loss": 1.1008,
+      "step": 908500
+    },
+    {
+      "epoch": 43.61,
+      "learning_rate": 1.8451717171717174e-06,
+      "loss": 1.0971,
+      "step": 909000
+    },
+    {
+      "epoch": 43.64,
+      "learning_rate": 1.835070707070707e-06,
+      "loss": 1.1018,
+      "step": 909500
+    },
+    {
+      "epoch": 43.66,
+      "learning_rate": 1.824969696969697e-06,
+      "loss": 1.0992,
+      "step": 910000
+    },
+    {
+      "epoch": 43.68,
+      "learning_rate": 1.814868686868687e-06,
+      "loss": 1.0982,
+      "step": 910500
+    },
+    {
+      "epoch": 43.71,
+      "learning_rate": 1.8047676767676769e-06,
+      "loss": 1.0979,
+      "step": 911000
+    },
+    {
+      "epoch": 43.73,
+      "learning_rate": 1.7946666666666667e-06,
+      "loss": 1.1003,
+      "step": 911500
+    },
+    {
+      "epoch": 43.76,
+      "learning_rate": 1.7845656565656566e-06,
+      "loss": 1.1016,
+      "step": 912000
+    },
+    {
+      "epoch": 43.78,
+      "learning_rate": 1.7744646464646467e-06,
+      "loss": 1.0979,
+      "step": 912500
+    },
+    {
+      "epoch": 43.8,
+      "learning_rate": 1.7643838383838385e-06,
+      "loss": 1.0984,
+      "step": 913000
+    },
+    {
+      "epoch": 43.83,
+      "learning_rate": 1.7543030303030303e-06,
+      "loss": 1.097,
+      "step": 913500
+    },
+    {
+      "epoch": 43.85,
+      "learning_rate": 1.7442020202020204e-06,
+      "loss": 1.099,
+      "step": 914000
+    },
+    {
+      "epoch": 43.88,
+      "learning_rate": 1.7341010101010103e-06,
+      "loss": 1.0977,
+      "step": 914500
+    },
+    {
+      "epoch": 43.9,
+      "learning_rate": 1.7240000000000001e-06,
+      "loss": 1.0966,
+      "step": 915000
+    },
+    {
+      "epoch": 43.92,
+      "learning_rate": 1.71389898989899e-06,
+      "loss": 1.1018,
+      "step": 915500
+    },
+    {
+      "epoch": 43.95,
+      "learning_rate": 1.703818181818182e-06,
+      "loss": 1.0974,
+      "step": 916000
+    },
+    {
+      "epoch": 43.97,
+      "learning_rate": 1.693717171717172e-06,
+      "loss": 1.0972,
+      "step": 916500
+    },
+    {
+      "epoch": 44.0,
+      "learning_rate": 1.6836161616161618e-06,
+      "loss": 1.0987,
+      "step": 917000
+    },
+    {
+      "epoch": 44.0,
+      "eval_accuracy": 0.7686325749145947,
+      "eval_loss": 0.9663205146789551,
+      "eval_runtime": 52.2411,
+      "eval_samples_per_second": 870.043,
+      "eval_steps_per_second": 13.61,
+      "step": 917092
+    },
+    {
+      "epoch": 44.02,
+      "learning_rate": 1.6735151515151516e-06,
+      "loss": 1.0989,
+      "step": 917500
+    },
+    {
+      "epoch": 44.04,
+      "learning_rate": 1.6634141414141417e-06,
+      "loss": 1.0999,
+      "step": 918000
+    },
+    {
+      "epoch": 44.07,
+      "learning_rate": 1.6533131313131316e-06,
+      "loss": 1.0967,
+      "step": 918500
+    },
+    {
+      "epoch": 44.09,
+      "learning_rate": 1.6432121212121212e-06,
+      "loss": 1.0969,
+      "step": 919000
+    },
+    {
+      "epoch": 44.12,
+      "learning_rate": 1.6331313131313135e-06,
+      "loss": 1.0973,
+      "step": 919500
+    },
+    {
+      "epoch": 44.14,
+      "learning_rate": 1.6230303030303033e-06,
+      "loss": 1.0989,
+      "step": 920000
+    },
+    {
+      "epoch": 44.16,
+      "learning_rate": 1.612929292929293e-06,
+      "loss": 1.098,
+      "step": 920500
+    },
+    {
+      "epoch": 44.19,
+      "learning_rate": 1.6028282828282828e-06,
+      "loss": 1.0951,
+      "step": 921000
+    },
+    {
+      "epoch": 44.21,
+      "learning_rate": 1.5927272727272727e-06,
+      "loss": 1.0951,
+      "step": 921500
+    },
+    {
+      "epoch": 44.24,
+      "learning_rate": 1.5826262626262628e-06,
+      "loss": 1.0987,
+      "step": 922000
+    },
+    {
+      "epoch": 44.26,
+      "learning_rate": 1.5725252525252526e-06,
+      "loss": 1.0993,
+      "step": 922500
+    },
+    {
+      "epoch": 44.28,
+      "learning_rate": 1.5624242424242425e-06,
+      "loss": 1.093,
+      "step": 923000
+    },
+    {
+      "epoch": 44.31,
+      "learning_rate": 1.5523434343434345e-06,
+      "loss": 1.0995,
+      "step": 923500
+    },
+    {
+      "epoch": 44.33,
+      "learning_rate": 1.5422424242424244e-06,
+      "loss": 1.1001,
+      "step": 924000
+    },
+    {
+      "epoch": 44.36,
+      "learning_rate": 1.5321414141414143e-06,
+      "loss": 1.0982,
+      "step": 924500
+    },
+    {
+      "epoch": 44.38,
+      "learning_rate": 1.5220404040404041e-06,
+      "loss": 1.0955,
+      "step": 925000
+    },
+    {
+      "epoch": 44.4,
+      "learning_rate": 1.511939393939394e-06,
+      "loss": 1.0974,
+      "step": 925500
+    },
+    {
+      "epoch": 44.43,
+      "learning_rate": 1.501838383838384e-06,
+      "loss": 1.0981,
+      "step": 926000
+    },
+    {
+      "epoch": 44.45,
+      "learning_rate": 1.491737373737374e-06,
+      "loss": 1.0998,
+      "step": 926500
+    },
+    {
+      "epoch": 44.48,
+      "learning_rate": 1.4816363636363638e-06,
+      "loss": 1.1008,
+      "step": 927000
+    },
+    {
+      "epoch": 44.5,
+      "learning_rate": 1.4715555555555558e-06,
+      "loss": 1.096,
+      "step": 927500
+    },
+    {
+      "epoch": 44.52,
+      "learning_rate": 1.4614747474747477e-06,
+      "loss": 1.0996,
+      "step": 928000
+    },
+    {
+      "epoch": 44.55,
+      "learning_rate": 1.4513737373737375e-06,
+      "loss": 1.0969,
+      "step": 928500
+    },
+    {
+      "epoch": 44.57,
+      "learning_rate": 1.4412727272727272e-06,
+      "loss": 1.0962,
+      "step": 929000
+    },
+    {
+      "epoch": 44.6,
+      "learning_rate": 1.4311717171717175e-06,
+      "loss": 1.096,
+      "step": 929500
+    },
+    {
+      "epoch": 44.62,
+      "learning_rate": 1.4210707070707071e-06,
+      "loss": 1.0981,
+      "step": 930000
+    },
+    {
+      "epoch": 44.64,
+      "learning_rate": 1.410989898989899e-06,
+      "loss": 1.0998,
+      "step": 930500
+    },
+    {
+      "epoch": 44.67,
+      "learning_rate": 1.4008888888888892e-06,
+      "loss": 1.0983,
+      "step": 931000
+    },
+    {
+      "epoch": 44.69,
+      "learning_rate": 1.3907878787878789e-06,
+      "loss": 1.0983,
+      "step": 931500
+    },
+    {
+      "epoch": 44.72,
+      "learning_rate": 1.3806868686868687e-06,
+      "loss": 1.0982,
+      "step": 932000
+    },
+    {
+      "epoch": 44.74,
+      "learning_rate": 1.3705858585858586e-06,
+      "loss": 1.0977,
+      "step": 932500
+    },
+    {
+      "epoch": 44.76,
+      "learning_rate": 1.3605050505050506e-06,
+      "loss": 1.0996,
+      "step": 933000
+    },
+    {
+      "epoch": 44.79,
+      "learning_rate": 1.3504040404040405e-06,
+      "loss": 1.1003,
+      "step": 933500
+    },
+    {
+      "epoch": 44.81,
+      "learning_rate": 1.3403030303030304e-06,
+      "loss": 1.0986,
+      "step": 934000
+    },
+    {
+      "epoch": 44.84,
+      "learning_rate": 1.3302020202020202e-06,
+      "loss": 1.0972,
+      "step": 934500
+    },
+    {
+      "epoch": 44.86,
+      "learning_rate": 1.3201010101010103e-06,
+      "loss": 1.0985,
+      "step": 935000
+    },
+    {
+      "epoch": 44.88,
+      "learning_rate": 1.3100000000000002e-06,
+      "loss": 1.0954,
+      "step": 935500
+    },
+    {
+      "epoch": 44.91,
+      "learning_rate": 1.29989898989899e-06,
+      "loss": 1.0975,
+      "step": 936000
+    },
+    {
+      "epoch": 44.93,
+      "learning_rate": 1.2898181818181819e-06,
+      "loss": 1.0967,
+      "step": 936500
+    },
+    {
+      "epoch": 44.96,
+      "learning_rate": 1.279717171717172e-06,
+      "loss": 1.0997,
+      "step": 937000
+    },
+    {
+      "epoch": 44.98,
+      "learning_rate": 1.2696161616161618e-06,
+      "loss": 1.0979,
+      "step": 937500
+    },
+    {
+      "epoch": 45.0,
+      "eval_accuracy": 0.7690747624983905,
+      "eval_loss": 0.9645295739173889,
+      "eval_runtime": 50.915,
+      "eval_samples_per_second": 892.704,
+      "eval_steps_per_second": 13.964,
+      "step": 937935
+    },
+    {
+      "epoch": 45.0,
+      "learning_rate": 1.2595151515151517e-06,
+      "loss": 1.0975,
+      "step": 938000
+    },
+    {
+      "epoch": 45.03,
+      "learning_rate": 1.2494141414141415e-06,
+      "loss": 1.0951,
+      "step": 938500
+    },
+    {
+      "epoch": 45.05,
+      "learning_rate": 1.2393333333333336e-06,
+      "loss": 1.0977,
+      "step": 939000
+    },
+    {
+      "epoch": 45.08,
+      "learning_rate": 1.2292323232323234e-06,
+      "loss": 1.0978,
+      "step": 939500
+    },
+    {
+      "epoch": 45.1,
+      "learning_rate": 1.219131313131313e-06,
+      "loss": 1.096,
+      "step": 940000
+    },
+    {
+      "epoch": 45.12,
+      "learning_rate": 1.2090303030303031e-06,
+      "loss": 1.0971,
+      "step": 940500
+    },
+    {
+      "epoch": 45.15,
+      "learning_rate": 1.198929292929293e-06,
+      "loss": 1.0992,
+      "step": 941000
+    },
+    {
+      "epoch": 45.17,
+      "learning_rate": 1.1888282828282829e-06,
+      "loss": 1.0975,
+      "step": 941500
+    },
+    {
+      "epoch": 45.2,
+      "learning_rate": 1.178747474747475e-06,
+      "loss": 1.0981,
+      "step": 942000
+    },
+    {
+      "epoch": 45.22,
+      "learning_rate": 1.1686464646464648e-06,
+      "loss": 1.0967,
+      "step": 942500
+    },
+    {
+      "epoch": 45.24,
+      "learning_rate": 1.1585454545454546e-06,
+      "loss": 1.0977,
+      "step": 943000
+    },
+    {
+      "epoch": 45.27,
+      "learning_rate": 1.1484444444444445e-06,
+      "loss": 1.0966,
+      "step": 943500
+    },
+    {
+      "epoch": 45.29,
+      "learning_rate": 1.1383434343434346e-06,
+      "loss": 1.0983,
+      "step": 944000
+    },
+    {
+      "epoch": 45.31,
+      "learning_rate": 1.1282424242424244e-06,
+      "loss": 1.0996,
+      "step": 944500
+    },
+    {
+      "epoch": 45.34,
+      "learning_rate": 1.1181616161616163e-06,
+      "loss": 1.0986,
+      "step": 945000
+    },
+    {
+      "epoch": 45.36,
+      "learning_rate": 1.1080606060606061e-06,
+      "loss": 1.0992,
+      "step": 945500
+    },
+    {
+      "epoch": 45.39,
+      "learning_rate": 1.097959595959596e-06,
+      "loss": 1.0975,
+      "step": 946000
+    },
+    {
+      "epoch": 45.41,
+      "learning_rate": 1.0878585858585858e-06,
+      "loss": 1.0965,
+      "step": 946500
+    },
+    {
+      "epoch": 45.43,
+      "learning_rate": 1.077757575757576e-06,
+      "loss": 1.0971,
+      "step": 947000
+    },
+    {
+      "epoch": 45.46,
+      "learning_rate": 1.0676767676767677e-06,
+      "loss": 1.0966,
+      "step": 947500
+    },
+    {
+      "epoch": 45.48,
+      "learning_rate": 1.0575757575757576e-06,
+      "loss": 1.0952,
+      "step": 948000
+    },
+    {
+      "epoch": 45.51,
+      "learning_rate": 1.0474747474747475e-06,
+      "loss": 1.0978,
+      "step": 948500
+    },
+    {
+      "epoch": 45.53,
+      "learning_rate": 1.0373737373737375e-06,
+      "loss": 1.0973,
+      "step": 949000
+    },
+    {
+      "epoch": 45.55,
+      "learning_rate": 1.0272929292929294e-06,
+      "loss": 1.0953,
+      "step": 949500
+    },
+    {
+      "epoch": 45.58,
+      "learning_rate": 1.0171919191919192e-06,
+      "loss": 1.0943,
+      "step": 950000
+    },
+    {
+      "epoch": 45.6,
+      "learning_rate": 1.0070909090909093e-06,
+      "loss": 1.0973,
+      "step": 950500
+    },
+    {
+      "epoch": 45.63,
+      "learning_rate": 9.96989898989899e-07,
+      "loss": 1.0961,
+      "step": 951000
+    },
+    {
+      "epoch": 45.65,
+      "learning_rate": 9.86909090909091e-07,
+      "loss": 1.0983,
+      "step": 951500
+    },
+    {
+      "epoch": 45.67,
+      "learning_rate": 9.768080808080809e-07,
+      "loss": 1.0965,
+      "step": 952000
+    },
+    {
+      "epoch": 45.7,
+      "learning_rate": 9.667070707070707e-07,
+      "loss": 1.0978,
+      "step": 952500
+    },
+    {
+      "epoch": 45.72,
+      "learning_rate": 9.566060606060606e-07,
+      "loss": 1.0971,
+      "step": 953000
+    },
+    {
+      "epoch": 45.75,
+      "learning_rate": 9.465050505050507e-07,
+      "loss": 1.098,
+      "step": 953500
+    },
+    {
+      "epoch": 45.77,
+      "learning_rate": 9.364242424242426e-07,
+      "loss": 1.096,
+      "step": 954000
+    },
+    {
+      "epoch": 45.79,
+      "learning_rate": 9.263232323232325e-07,
+      "loss": 1.1006,
+      "step": 954500
+    },
+    {
+      "epoch": 45.82,
+      "learning_rate": 9.162222222222222e-07,
+      "loss": 1.0959,
+      "step": 955000
+    },
+    {
+      "epoch": 45.84,
+      "learning_rate": 9.061212121212122e-07,
+      "loss": 1.0954,
+      "step": 955500
+    },
+    {
+      "epoch": 45.87,
+      "learning_rate": 8.960404040404041e-07,
+      "loss": 1.0966,
+      "step": 956000
+    },
+    {
+      "epoch": 45.89,
+      "learning_rate": 8.85939393939394e-07,
+      "loss": 1.0926,
+      "step": 956500
+    },
+    {
+      "epoch": 45.91,
+      "learning_rate": 8.758383838383839e-07,
+      "loss": 1.0973,
+      "step": 957000
+    },
+    {
+      "epoch": 45.94,
+      "learning_rate": 8.657373737373738e-07,
+      "loss": 1.0974,
+      "step": 957500
+    },
+    {
+      "epoch": 45.96,
+      "learning_rate": 8.556363636363638e-07,
+      "loss": 1.096,
+      "step": 958000
+    },
+    {
+      "epoch": 45.99,
+      "learning_rate": 8.455353535353536e-07,
+      "loss": 1.0992,
+      "step": 958500
+    },
+    {
+      "epoch": 46.0,
+      "eval_accuracy": 0.7690765342380963,
+      "eval_loss": 0.9633356928825378,
+      "eval_runtime": 53.1296,
+      "eval_samples_per_second": 855.494,
+      "eval_steps_per_second": 13.382,
+      "step": 958778
+    },
+    {
+      "epoch": 46.01,
+      "learning_rate": 8.354343434343436e-07,
+      "loss": 1.0996,
+      "step": 959000
+    },
+    {
+      "epoch": 46.03,
+      "learning_rate": 8.253333333333334e-07,
+      "loss": 1.1001,
+      "step": 959500
+    },
+    {
+      "epoch": 46.06,
+      "learning_rate": 8.152323232323232e-07,
+      "loss": 1.0948,
+      "step": 960000
+    },
+    {
+      "epoch": 46.08,
+      "learning_rate": 8.051717171717173e-07,
+      "loss": 1.0946,
+      "step": 960500
+    },
+    {
+      "epoch": 46.11,
+      "learning_rate": 7.950707070707071e-07,
+      "loss": 1.0973,
+      "step": 961000
+    },
+    {
+      "epoch": 46.13,
+      "learning_rate": 7.84969696969697e-07,
+      "loss": 1.0977,
+      "step": 961500
+    },
+    {
+      "epoch": 46.15,
+      "learning_rate": 7.748686868686869e-07,
+      "loss": 1.0989,
+      "step": 962000
+    },
+    {
+      "epoch": 46.18,
+      "learning_rate": 7.647676767676768e-07,
+      "loss": 1.0953,
+      "step": 962500
+    },
+    {
+      "epoch": 46.2,
+      "learning_rate": 7.546666666666668e-07,
+      "loss": 1.0942,
+      "step": 963000
+    },
+    {
+      "epoch": 46.23,
+      "learning_rate": 7.445656565656566e-07,
+      "loss": 1.0966,
+      "step": 963500
+    },
+    {
+      "epoch": 46.25,
+      "learning_rate": 7.344646464646466e-07,
+      "loss": 1.0997,
+      "step": 964000
+    },
+    {
+      "epoch": 46.27,
+      "learning_rate": 7.243636363636363e-07,
+      "loss": 1.0962,
+      "step": 964500
+    },
+    {
+      "epoch": 46.3,
+      "learning_rate": 7.142626262626263e-07,
+      "loss": 1.0948,
+      "step": 965000
+    },
+    {
+      "epoch": 46.32,
+      "learning_rate": 7.041818181818183e-07,
+      "loss": 1.0967,
+      "step": 965500
+    },
+    {
+      "epoch": 46.35,
+      "learning_rate": 6.940808080808081e-07,
+      "loss": 1.0967,
+      "step": 966000
+    },
+    {
+      "epoch": 46.37,
+      "learning_rate": 6.839797979797981e-07,
+      "loss": 1.095,
+      "step": 966500
+    },
+    {
+      "epoch": 46.39,
+      "learning_rate": 6.738989898989899e-07,
+      "loss": 1.0971,
+      "step": 967000
+    },
+    {
+      "epoch": 46.42,
+      "learning_rate": 6.637979797979799e-07,
+      "loss": 1.0993,
+      "step": 967500
+    },
+    {
+      "epoch": 46.44,
+      "learning_rate": 6.536969696969697e-07,
+      "loss": 1.096,
+      "step": 968000
+    },
+    {
+      "epoch": 46.47,
+      "learning_rate": 6.435959595959597e-07,
+      "loss": 1.0969,
+      "step": 968500
+    },
+    {
+      "epoch": 46.49,
+      "learning_rate": 6.334949494949496e-07,
+      "loss": 1.096,
+      "step": 969000
+    },
+    {
+      "epoch": 46.51,
+      "learning_rate": 6.234343434343436e-07,
+      "loss": 1.0944,
+      "step": 969500
+    },
+    {
+      "epoch": 46.54,
+      "learning_rate": 6.133333333333333e-07,
+      "loss": 1.0963,
+      "step": 970000
+    },
+    {
+      "epoch": 46.56,
+      "learning_rate": 6.032323232323233e-07,
+      "loss": 1.0965,
+      "step": 970500
+    },
+    {
+      "epoch": 46.59,
+      "learning_rate": 5.931313131313132e-07,
+      "loss": 1.098,
+      "step": 971000
+    },
+    {
+      "epoch": 46.61,
+      "learning_rate": 5.83030303030303e-07,
+      "loss": 1.0961,
+      "step": 971500
+    },
+    {
+      "epoch": 46.63,
+      "learning_rate": 5.72949494949495e-07,
+      "loss": 1.0958,
+      "step": 972000
+    },
+    {
+      "epoch": 46.66,
+      "learning_rate": 5.628484848484849e-07,
+      "loss": 1.0975,
+      "step": 972500
+    },
+    {
+      "epoch": 46.68,
+      "learning_rate": 5.527474747474748e-07,
+      "loss": 1.0969,
+      "step": 973000
+    },
+    {
+      "epoch": 46.71,
+      "learning_rate": 5.426464646464647e-07,
+      "loss": 1.0954,
+      "step": 973500
+    },
+    {
+      "epoch": 46.73,
+      "learning_rate": 5.325454545454546e-07,
+      "loss": 1.0968,
+      "step": 974000
+    },
+    {
+      "epoch": 46.75,
+      "learning_rate": 5.224444444444445e-07,
+      "loss": 1.0967,
+      "step": 974500
+    },
+    {
+      "epoch": 46.78,
+      "learning_rate": 5.123434343434343e-07,
+      "loss": 1.0942,
+      "step": 975000
+    },
+    {
+      "epoch": 46.8,
+      "learning_rate": 5.022424242424243e-07,
+      "loss": 1.0974,
+      "step": 975500
+    },
+    {
+      "epoch": 46.83,
+      "learning_rate": 4.921414141414142e-07,
+      "loss": 1.0951,
+      "step": 976000
+    },
+    {
+      "epoch": 46.85,
+      "learning_rate": 4.82040404040404e-07,
+      "loss": 1.0967,
+      "step": 976500
+    },
+    {
+      "epoch": 46.87,
+      "learning_rate": 4.7193939393939395e-07,
+      "loss": 1.0959,
+      "step": 977000
+    },
+    {
+      "epoch": 46.9,
+      "learning_rate": 4.6183838383838386e-07,
+      "loss": 1.0961,
+      "step": 977500
+    },
+    {
+      "epoch": 46.92,
+      "learning_rate": 4.517575757575758e-07,
+      "loss": 1.0974,
+      "step": 978000
+    },
+    {
+      "epoch": 46.95,
+      "learning_rate": 4.416565656565657e-07,
+      "loss": 1.0962,
+      "step": 978500
+    },
+    {
+      "epoch": 46.97,
+      "learning_rate": 4.315555555555556e-07,
+      "loss": 1.0963,
+      "step": 979000
+    },
+    {
+      "epoch": 46.99,
+      "learning_rate": 4.2145454545454554e-07,
+      "loss": 1.0947,
+      "step": 979500
+    },
+    {
+      "epoch": 47.0,
+      "eval_accuracy": 0.7692219494003684,
+      "eval_loss": 0.9634374976158142,
+      "eval_runtime": 50.3267,
+      "eval_samples_per_second": 903.138,
+      "eval_steps_per_second": 14.128,
+      "step": 979621
+    },
+    {
+      "epoch": 47.02,
+      "learning_rate": 4.113535353535354e-07,
+      "loss": 1.0968,
+      "step": 980000
+    },
+    {
+      "epoch": 47.04,
+      "learning_rate": 4.0125252525252527e-07,
+      "loss": 1.0952,
+      "step": 980500
+    },
+    {
+      "epoch": 47.07,
+      "learning_rate": 3.9115151515151513e-07,
+      "loss": 1.0978,
+      "step": 981000
+    },
+    {
+      "epoch": 47.09,
+      "learning_rate": 3.8105050505050505e-07,
+      "loss": 1.0974,
+      "step": 981500
+    },
+    {
+      "epoch": 47.11,
+      "learning_rate": 3.7094949494949496e-07,
+      "loss": 1.0961,
+      "step": 982000
+    },
+    {
+      "epoch": 47.14,
+      "learning_rate": 3.608686868686869e-07,
+      "loss": 1.0946,
+      "step": 982500
+    },
+    {
+      "epoch": 47.16,
+      "learning_rate": 3.507676767676768e-07,
+      "loss": 1.0972,
+      "step": 983000
+    },
+    {
+      "epoch": 47.19,
+      "learning_rate": 3.406666666666667e-07,
+      "loss": 1.0969,
+      "step": 983500
+    },
+    {
+      "epoch": 47.21,
+      "learning_rate": 3.3056565656565664e-07,
+      "loss": 1.0946,
+      "step": 984000
+    },
+    {
+      "epoch": 47.23,
+      "learning_rate": 3.204848484848485e-07,
+      "loss": 1.097,
+      "step": 984500
+    },
+    {
+      "epoch": 47.26,
+      "learning_rate": 3.103838383838384e-07,
+      "loss": 1.0973,
+      "step": 985000
+    },
+    {
+      "epoch": 47.28,
+      "learning_rate": 3.002828282828283e-07,
+      "loss": 1.0949,
+      "step": 985500
+    },
+    {
+      "epoch": 47.31,
+      "learning_rate": 2.901818181818182e-07,
+      "loss": 1.0965,
+      "step": 986000
+    },
+    {
+      "epoch": 47.33,
+      "learning_rate": 2.8008080808080813e-07,
+      "loss": 1.0969,
+      "step": 986500
+    },
+    {
+      "epoch": 47.35,
+      "learning_rate": 2.69979797979798e-07,
+      "loss": 1.096,
+      "step": 987000
+    },
+    {
+      "epoch": 47.38,
+      "learning_rate": 2.598787878787879e-07,
+      "loss": 1.0962,
+      "step": 987500
+    },
+    {
+      "epoch": 47.4,
+      "learning_rate": 2.497777777777778e-07,
+      "loss": 1.0966,
+      "step": 988000
+    },
+    {
+      "epoch": 47.43,
+      "learning_rate": 2.396969696969697e-07,
+      "loss": 1.0982,
+      "step": 988500
+    },
+    {
+      "epoch": 47.45,
+      "learning_rate": 2.2959595959595962e-07,
+      "loss": 1.0969,
+      "step": 989000
+    },
+    {
+      "epoch": 47.47,
+      "learning_rate": 2.194949494949495e-07,
+      "loss": 1.0964,
+      "step": 989500
+    },
+    {
+      "epoch": 47.5,
+      "learning_rate": 2.093939393939394e-07,
+      "loss": 1.0937,
+      "step": 990000
+    },
+    {
+      "epoch": 47.52,
+      "learning_rate": 1.993131313131313e-07,
+      "loss": 1.0963,
+      "step": 990500
+    },
+    {
+      "epoch": 47.55,
+      "learning_rate": 1.8921212121212122e-07,
+      "loss": 1.0964,
+      "step": 991000
+    },
+    {
+      "epoch": 47.57,
+      "learning_rate": 1.791111111111111e-07,
+      "loss": 1.096,
+      "step": 991500
+    },
+    {
+      "epoch": 47.59,
+      "learning_rate": 1.6901010101010103e-07,
+      "loss": 1.0992,
+      "step": 992000
+    },
+    {
+      "epoch": 47.62,
+      "learning_rate": 1.5892929292929293e-07,
+      "loss": 1.0971,
+      "step": 992500
+    },
+    {
+      "epoch": 47.64,
+      "learning_rate": 1.4882828282828285e-07,
+      "loss": 1.0942,
+      "step": 993000
+    },
+    {
+      "epoch": 47.67,
+      "learning_rate": 1.3872727272727274e-07,
+      "loss": 1.0957,
+      "step": 993500
+    },
+    {
+      "epoch": 47.69,
+      "learning_rate": 1.2862626262626265e-07,
+      "loss": 1.0948,
+      "step": 994000
+    },
+    {
+      "epoch": 47.71,
+      "learning_rate": 1.1852525252525254e-07,
+      "loss": 1.0953,
+      "step": 994500
+    },
+    {
+      "epoch": 47.74,
+      "learning_rate": 1.0842424242424243e-07,
+      "loss": 1.0955,
+      "step": 995000
+    },
+    {
+      "epoch": 47.76,
+      "learning_rate": 9.832323232323233e-08,
+      "loss": 1.0986,
+      "step": 995500
+    },
+    {
+      "epoch": 47.79,
+      "learning_rate": 8.822222222222224e-08,
+      "loss": 1.0963,
+      "step": 996000
+    },
+    {
+      "epoch": 47.81,
+      "learning_rate": 7.814141414141414e-08,
+      "loss": 1.0951,
+      "step": 996500
+    },
+    {
+      "epoch": 47.83,
+      "learning_rate": 6.804040404040404e-08,
+      "loss": 1.0978,
+      "step": 997000
+    },
+    {
+      "epoch": 47.86,
+      "learning_rate": 5.7939393939393946e-08,
+      "loss": 1.0963,
+      "step": 997500
+    },
+    {
+      "epoch": 47.88,
+      "learning_rate": 4.783838383838385e-08,
+      "loss": 1.0969,
+      "step": 998000
+    },
+    {
+      "epoch": 47.91,
+      "learning_rate": 3.773737373737374e-08,
+      "loss": 1.0947,
+      "step": 998500
+    },
+    {
+      "epoch": 47.93,
+      "learning_rate": 2.7636363636363637e-08,
+      "loss": 1.0944,
+      "step": 999000
+    },
+    {
+      "epoch": 47.95,
+      "learning_rate": 1.7535353535353537e-08,
+      "loss": 1.0965,
+      "step": 999500
+    },
+    {
+      "epoch": 47.98,
+      "learning_rate": 7.434343434343435e-09,
+      "loss": 1.095,
+      "step": 1000000
+    },
+    {
+      "epoch": 47.98,
+      "eval_accuracy": 0.7698673283515727,
+      "eval_loss": 0.9603139162063599,
+      "eval_runtime": 51.1794,
+      "eval_samples_per_second": 888.092,
+      "eval_steps_per_second": 13.892,
+      "step": 1000000
+    },
+    {
+      "epoch": 47.98,
+      "step": 1000000,
+      "total_flos": 6.738058032353536e+19,
+      "train_loss": 0.032887611633300784,
+      "train_runtime": 14645.775,
+      "train_samples_per_second": 17479.444,
+      "train_steps_per_second": 68.279
+    }
+  ],
+  "max_steps": 1000000,
+  "num_train_epochs": 48,
+  "total_flos": 6.738058032353536e+19,
+  "trial_name": null,
+  "trial_params": null
+}