diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,8187 @@
+{
+  "best_metric": 0.012130416929721832,
+  "best_model_checkpoint": "./outputs/models/flan-t5-large-summ/checkpoint-97000",
+  "epoch": 20.0,
+  "eval_steps": 1000,
+  "global_step": 104660,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01910949742021785,
+      "grad_norm": 2.4572737216949463,
+      "learning_rate": 1e-05,
+      "loss": 1.3075,
+      "step": 100
+    },
+    {
+      "epoch": 0.0382189948404357,
+      "grad_norm": 1.3809181451797485,
+      "learning_rate": 2e-05,
+      "loss": 1.1747,
+      "step": 200
+    },
+    {
+      "epoch": 0.05732849226065354,
+      "grad_norm": 1.076543927192688,
+      "learning_rate": 3e-05,
+      "loss": 1.1598,
+      "step": 300
+    },
+    {
+      "epoch": 0.0764379896808714,
+      "grad_norm": 1.3211930990219116,
+      "learning_rate": 4e-05,
+      "loss": 1.1265,
+      "step": 400
+    },
+    {
+      "epoch": 0.09554748710108923,
+      "grad_norm": 1.7730636596679688,
+      "learning_rate": 5e-05,
+      "loss": 1.0228,
+      "step": 500
+    },
+    {
+      "epoch": 0.11465698452130708,
+      "grad_norm": 1.103384256362915,
+      "learning_rate": 4.995199692780338e-05,
+      "loss": 0.9792,
+      "step": 600
+    },
+    {
+      "epoch": 0.13376648194152493,
+      "grad_norm": 0.9394871592521667,
+      "learning_rate": 4.990399385560676e-05,
+      "loss": 0.9821,
+      "step": 700
+    },
+    {
+      "epoch": 0.1528759793617428,
+      "grad_norm": 0.9111909866333008,
+      "learning_rate": 4.985599078341014e-05,
+      "loss": 0.9793,
+      "step": 800
+    },
+    {
+      "epoch": 0.17198547678196063,
+      "grad_norm": 1.2519441843032837,
+      "learning_rate": 4.9807987711213524e-05,
+      "loss": 0.9411,
+      "step": 900
+    },
+    {
+      "epoch": 0.19109497420217847,
+      "grad_norm": 0.7635079622268677,
+      "learning_rate": 4.9759984639016896e-05,
+      "loss": 0.922,
+      "step": 1000
+    },
+    {
+      "epoch": 0.19109497420217847,
+      "eval_events-synergy/entsum_processed_loss": 0.7808578610420227,
+      "eval_events-synergy/entsum_processed_runtime": 86.3527,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.936,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.74,
+      "step": 1000
+    },
+    {
+      "epoch": 0.21020447162239633,
+      "grad_norm": 1.012012004852295,
+      "learning_rate": 4.971198156682028e-05,
+      "loss": 0.908,
+      "step": 1100
+    },
+    {
+      "epoch": 0.22931396904261417,
+      "grad_norm": 1.2864316701889038,
+      "learning_rate": 4.966397849462366e-05,
+      "loss": 0.8234,
+      "step": 1200
+    },
+    {
+      "epoch": 0.24842346646283203,
+      "grad_norm": 1.4988912343978882,
+      "learning_rate": 4.961597542242703e-05,
+      "loss": 0.8118,
+      "step": 1300
+    },
+    {
+      "epoch": 0.26753296388304987,
+      "grad_norm": 1.2037246227264404,
+      "learning_rate": 4.956797235023042e-05,
+      "loss": 0.8052,
+      "step": 1400
+    },
+    {
+      "epoch": 0.2866424613032677,
+      "grad_norm": 1.3416869640350342,
+      "learning_rate": 4.9519969278033797e-05,
+      "loss": 0.6814,
+      "step": 1500
+    },
+    {
+      "epoch": 0.3057519587234856,
+      "grad_norm": 0.7591760158538818,
+      "learning_rate": 4.9471966205837175e-05,
+      "loss": 0.835,
+      "step": 1600
+    },
+    {
+      "epoch": 0.32486145614370343,
+      "grad_norm": 1.0816859006881714,
+      "learning_rate": 4.9423963133640554e-05,
+      "loss": 0.7875,
+      "step": 1700
+    },
+    {
+      "epoch": 0.34397095356392127,
+      "grad_norm": 1.381515383720398,
+      "learning_rate": 4.937596006144393e-05,
+      "loss": 0.7727,
+      "step": 1800
+    },
+    {
+      "epoch": 0.3630804509841391,
+      "grad_norm": 1.031540870666504,
+      "learning_rate": 4.932795698924731e-05,
+      "loss": 0.7416,
+      "step": 1900
+    },
+    {
+      "epoch": 0.38218994840435694,
+      "grad_norm": 1.013592004776001,
+      "learning_rate": 4.927995391705069e-05,
+      "loss": 0.7544,
+      "step": 2000
+    },
+    {
+      "epoch": 0.38218994840435694,
+      "eval_events-synergy/entsum_processed_loss": 0.5898063778877258,
+      "eval_events-synergy/entsum_processed_runtime": 86.5907,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.862,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.721,
+      "step": 2000
+    },
+    {
+      "epoch": 0.40129944582457483,
+      "grad_norm": 0.993009626865387,
+      "learning_rate": 4.9231950844854076e-05,
+      "loss": 0.6104,
+      "step": 2100
+    },
+    {
+      "epoch": 0.42040894324479267,
+      "grad_norm": 1.2820205688476562,
+      "learning_rate": 4.9183947772657455e-05,
+      "loss": 0.6266,
+      "step": 2200
+    },
+    {
+      "epoch": 0.4395184406650105,
+      "grad_norm": 1.0375840663909912,
+      "learning_rate": 4.913594470046083e-05,
+      "loss": 0.6959,
+      "step": 2300
+    },
+    {
+      "epoch": 0.45862793808522834,
+      "grad_norm": 1.4805232286453247,
+      "learning_rate": 4.908794162826421e-05,
+      "loss": 0.7472,
+      "step": 2400
+    },
+    {
+      "epoch": 0.47773743550544623,
+      "grad_norm": 1.1135224103927612,
+      "learning_rate": 4.903993855606759e-05,
+      "loss": 0.6271,
+      "step": 2500
+    },
+    {
+      "epoch": 0.49684693292566406,
+      "grad_norm": 0.6857364773750305,
+      "learning_rate": 4.899193548387097e-05,
+      "loss": 0.5972,
+      "step": 2600
+    },
+    {
+      "epoch": 0.5159564303458819,
+      "grad_norm": 1.0915530920028687,
+      "learning_rate": 4.894393241167435e-05,
+      "loss": 0.6082,
+      "step": 2700
+    },
+    {
+      "epoch": 0.5350659277660997,
+      "grad_norm": 0.6362367272377014,
+      "learning_rate": 4.8895929339477734e-05,
+      "loss": 0.6479,
+      "step": 2800
+    },
+    {
+      "epoch": 0.5541754251863176,
+      "grad_norm": 1.4901950359344482,
+      "learning_rate": 4.8847926267281106e-05,
+      "loss": 0.6431,
+      "step": 2900
+    },
+    {
+      "epoch": 0.5732849226065354,
+      "grad_norm": 1.0960026979446411,
+      "learning_rate": 4.879992319508449e-05,
+      "loss": 0.5337,
+      "step": 3000
+    },
+    {
+      "epoch": 0.5732849226065354,
+      "eval_events-synergy/entsum_processed_loss": 0.4623138904571533,
+      "eval_events-synergy/entsum_processed_runtime": 86.5166,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.885,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.727,
+      "step": 3000
+    },
+    {
+      "epoch": 0.5923944200267532,
+      "grad_norm": 1.622821569442749,
+      "learning_rate": 4.875192012288787e-05,
+      "loss": 0.5383,
+      "step": 3100
+    },
+    {
+      "epoch": 0.6115039174469712,
+      "grad_norm": 0.9446718096733093,
+      "learning_rate": 4.870391705069124e-05,
+      "loss": 0.5512,
+      "step": 3200
+    },
+    {
+      "epoch": 0.630613414867189,
+      "grad_norm": 1.555004358291626,
+      "learning_rate": 4.865591397849463e-05,
+      "loss": 0.5534,
+      "step": 3300
+    },
+    {
+      "epoch": 0.6497229122874069,
+      "grad_norm": 1.4712913036346436,
+      "learning_rate": 4.8607910906298006e-05,
+      "loss": 0.571,
+      "step": 3400
+    },
+    {
+      "epoch": 0.6688324097076247,
+      "grad_norm": 0.7977001070976257,
+      "learning_rate": 4.8559907834101385e-05,
+      "loss": 0.5197,
+      "step": 3500
+    },
+    {
+      "epoch": 0.6879419071278425,
+      "grad_norm": 0.742138683795929,
+      "learning_rate": 4.8511904761904764e-05,
+      "loss": 0.5445,
+      "step": 3600
+    },
+    {
+      "epoch": 0.7070514045480604,
+      "grad_norm": 2.7435715198516846,
+      "learning_rate": 4.846390168970814e-05,
+      "loss": 0.5571,
+      "step": 3700
+    },
+    {
+      "epoch": 0.7261609019682782,
+      "grad_norm": 2.0406129360198975,
+      "learning_rate": 4.841589861751152e-05,
+      "loss": 0.5156,
+      "step": 3800
+    },
+    {
+      "epoch": 0.745270399388496,
+      "grad_norm": 2.4138622283935547,
+      "learning_rate": 4.83678955453149e-05,
+      "loss": 0.4924,
+      "step": 3900
+    },
+    {
+      "epoch": 0.7643798968087139,
+      "grad_norm": 1.1996113061904907,
+      "learning_rate": 4.8319892473118285e-05,
+      "loss": 0.4759,
+      "step": 4000
+    },
+    {
+      "epoch": 0.7643798968087139,
+      "eval_events-synergy/entsum_processed_loss": 0.37834471464157104,
+      "eval_events-synergy/entsum_processed_runtime": 86.4756,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.898,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.73,
+      "step": 4000
+    },
+    {
+      "epoch": 0.7834893942289318,
+      "grad_norm": 1.4116686582565308,
+      "learning_rate": 4.827188940092166e-05,
+      "loss": 0.4346,
+      "step": 4100
+    },
+    {
+      "epoch": 0.8025988916491497,
+      "grad_norm": 1.2636014223098755,
+      "learning_rate": 4.822388632872504e-05,
+      "loss": 0.4435,
+      "step": 4200
+    },
+    {
+      "epoch": 0.8217083890693675,
+      "grad_norm": 1.1817740201950073,
+      "learning_rate": 4.817588325652842e-05,
+      "loss": 0.4403,
+      "step": 4300
+    },
+    {
+      "epoch": 0.8408178864895853,
+      "grad_norm": 1.4977765083312988,
+      "learning_rate": 4.81278801843318e-05,
+      "loss": 0.4201,
+      "step": 4400
+    },
+    {
+      "epoch": 0.8599273839098032,
+      "grad_norm": 1.4559648036956787,
+      "learning_rate": 4.807987711213518e-05,
+      "loss": 0.4928,
+      "step": 4500
+    },
+    {
+      "epoch": 0.879036881330021,
+      "grad_norm": 1.160703182220459,
+      "learning_rate": 4.803187403993856e-05,
+      "loss": 0.4822,
+      "step": 4600
+    },
+    {
+      "epoch": 0.8981463787502388,
+      "grad_norm": 1.7978248596191406,
+      "learning_rate": 4.7983870967741937e-05,
+      "loss": 0.4342,
+      "step": 4700
+    },
+    {
+      "epoch": 0.9172558761704567,
+      "grad_norm": 0.9042201638221741,
+      "learning_rate": 4.7935867895545315e-05,
+      "loss": 0.5015,
+      "step": 4800
+    },
+    {
+      "epoch": 0.9363653735906745,
+      "grad_norm": 0.9743239879608154,
+      "learning_rate": 4.78878648233487e-05,
+      "loss": 0.4664,
+      "step": 4900
+    },
+    {
+      "epoch": 0.9554748710108925,
+      "grad_norm": 2.9479408264160156,
+      "learning_rate": 4.783986175115208e-05,
+      "loss": 0.4586,
+      "step": 5000
+    },
+    {
+      "epoch": 0.9554748710108925,
+      "eval_events-synergy/entsum_processed_loss": 0.30807018280029297,
+      "eval_events-synergy/entsum_processed_runtime": 86.3824,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.927,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.737,
+      "step": 5000
+    },
+    {
+      "epoch": 0.9745843684311103,
+      "grad_norm": 0.7617493271827698,
+      "learning_rate": 4.779185867895545e-05,
+      "loss": 0.4103,
+      "step": 5100
+    },
+    {
+      "epoch": 0.9936938658513281,
+      "grad_norm": 1.3358664512634277,
+      "learning_rate": 4.774385560675884e-05,
+      "loss": 0.4177,
+      "step": 5200
+    },
+    {
+      "epoch": 1.0128033632715459,
+      "grad_norm": 1.880685567855835,
+      "learning_rate": 4.7695852534562216e-05,
+      "loss": 0.3909,
+      "step": 5300
+    },
+    {
+      "epoch": 1.0319128606917638,
+      "grad_norm": 1.12627375125885,
+      "learning_rate": 4.7647849462365594e-05,
+      "loss": 0.3738,
+      "step": 5400
+    },
+    {
+      "epoch": 1.0510223581119817,
+      "grad_norm": 1.3265856504440308,
+      "learning_rate": 4.759984639016897e-05,
+      "loss": 0.3564,
+      "step": 5500
+    },
+    {
+      "epoch": 1.0701318555321995,
+      "grad_norm": 2.0652267932891846,
+      "learning_rate": 4.755184331797235e-05,
+      "loss": 0.3301,
+      "step": 5600
+    },
+    {
+      "epoch": 1.0892413529524174,
+      "grad_norm": 1.5408031940460205,
+      "learning_rate": 4.750384024577573e-05,
+      "loss": 0.3282,
+      "step": 5700
+    },
+    {
+      "epoch": 1.1083508503726351,
+      "grad_norm": 2.1643123626708984,
+      "learning_rate": 4.745583717357911e-05,
+      "loss": 0.4158,
+      "step": 5800
+    },
+    {
+      "epoch": 1.127460347792853,
+      "grad_norm": 2.325970411300659,
+      "learning_rate": 4.7407834101382495e-05,
+      "loss": 0.3348,
+      "step": 5900
+    },
+    {
+      "epoch": 1.1465698452130708,
+      "grad_norm": 1.4004697799682617,
+      "learning_rate": 4.735983102918587e-05,
+      "loss": 0.3563,
+      "step": 6000
+    },
+    {
+      "epoch": 1.1465698452130708,
+      "eval_events-synergy/entsum_processed_loss": 0.25603175163269043,
+      "eval_events-synergy/entsum_processed_runtime": 86.5218,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.883,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.727,
+      "step": 6000
+    },
+    {
+      "epoch": 1.1656793426332888,
+      "grad_norm": 1.0601403713226318,
+      "learning_rate": 4.731182795698925e-05,
+      "loss": 0.3606,
+      "step": 6100
+    },
+    {
+      "epoch": 1.1847888400535065,
+      "grad_norm": 1.7814487218856812,
+      "learning_rate": 4.726382488479263e-05,
+      "loss": 0.3574,
+      "step": 6200
+    },
+    {
+      "epoch": 1.2038983374737244,
+      "grad_norm": 0.9902318716049194,
+      "learning_rate": 4.7215821812596e-05,
+      "loss": 0.2871,
+      "step": 6300
+    },
+    {
+      "epoch": 1.2230078348939424,
+      "grad_norm": 1.6244627237319946,
+      "learning_rate": 4.716781874039939e-05,
+      "loss": 0.3045,
+      "step": 6400
+    },
+    {
+      "epoch": 1.24211733231416,
+      "grad_norm": 0.4082527458667755,
+      "learning_rate": 4.711981566820277e-05,
+      "loss": 0.3104,
+      "step": 6500
+    },
+    {
+      "epoch": 1.261226829734378,
+      "grad_norm": 2.545142889022827,
+      "learning_rate": 4.7071812596006146e-05,
+      "loss": 0.3443,
+      "step": 6600
+    },
+    {
+      "epoch": 1.2803363271545958,
+      "grad_norm": 1.6806279420852661,
+      "learning_rate": 4.7023809523809525e-05,
+      "loss": 0.3004,
+      "step": 6700
+    },
+    {
+      "epoch": 1.2994458245748137,
+      "grad_norm": 1.1669882535934448,
+      "learning_rate": 4.697580645161291e-05,
+      "loss": 0.3168,
+      "step": 6800
+    },
+    {
+      "epoch": 1.3185553219950314,
+      "grad_norm": 1.3955861330032349,
+      "learning_rate": 4.692780337941628e-05,
+      "loss": 0.2662,
+      "step": 6900
+    },
+    {
+      "epoch": 1.3376648194152494,
+      "grad_norm": 1.6163365840911865,
+      "learning_rate": 4.687980030721966e-05,
+      "loss": 0.3172,
+      "step": 7000
+    },
+    {
+      "epoch": 1.3376648194152494,
+      "eval_events-synergy/entsum_processed_loss": 0.21874384582042694,
+      "eval_events-synergy/entsum_processed_runtime": 86.5224,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.883,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.727,
+      "step": 7000
+    },
+    {
+      "epoch": 1.3567743168354673,
+      "grad_norm": 0.675295352935791,
+      "learning_rate": 4.6831797235023047e-05,
+      "loss": 0.2594,
+      "step": 7100
+    },
+    {
+      "epoch": 1.375883814255685,
+      "grad_norm": 1.194022297859192,
+      "learning_rate": 4.6783794162826425e-05,
+      "loss": 0.329,
+      "step": 7200
+    },
+    {
+      "epoch": 1.3949933116759028,
+      "grad_norm": 2.0018792152404785,
+      "learning_rate": 4.6735791090629804e-05,
+      "loss": 0.2772,
+      "step": 7300
+    },
+    {
+      "epoch": 1.4141028090961207,
+      "grad_norm": 1.3288326263427734,
+      "learning_rate": 4.668778801843318e-05,
+      "loss": 0.2779,
+      "step": 7400
+    },
+    {
+      "epoch": 1.4332123065163387,
+      "grad_norm": 0.8486484885215759,
+      "learning_rate": 4.663978494623656e-05,
+      "loss": 0.3124,
+      "step": 7500
+    },
+    {
+      "epoch": 1.4523218039365564,
+      "grad_norm": 1.3498897552490234,
+      "learning_rate": 4.659178187403994e-05,
+      "loss": 0.3085,
+      "step": 7600
+    },
+    {
+      "epoch": 1.4714313013567744,
+      "grad_norm": 1.0366922616958618,
+      "learning_rate": 4.654377880184332e-05,
+      "loss": 0.2847,
+      "step": 7700
+    },
+    {
+      "epoch": 1.490540798776992,
+      "grad_norm": 1.5916386842727661,
+      "learning_rate": 4.6495775729646705e-05,
+      "loss": 0.2589,
+      "step": 7800
+    },
+    {
+      "epoch": 1.50965029619721,
+      "grad_norm": 2.4099597930908203,
+      "learning_rate": 4.6447772657450076e-05,
+      "loss": 0.3322,
+      "step": 7900
+    },
+    {
+      "epoch": 1.5287597936174278,
+      "grad_norm": 2.6880438327789307,
+      "learning_rate": 4.639976958525346e-05,
+      "loss": 0.2397,
+      "step": 8000
+    },
+    {
+      "epoch": 1.5287597936174278,
+      "eval_events-synergy/entsum_processed_loss": 0.18345867097377777,
+      "eval_events-synergy/entsum_processed_runtime": 86.5562,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.873,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.724,
+      "step": 8000
+    },
+    {
+      "epoch": 1.5478692910376457,
+      "grad_norm": 0.704814612865448,
+      "learning_rate": 4.635176651305684e-05,
+      "loss": 0.2491,
+      "step": 8100
+    },
+    {
+      "epoch": 1.5669787884578636,
+      "grad_norm": 0.7323247790336609,
+      "learning_rate": 4.630376344086021e-05,
+      "loss": 0.227,
+      "step": 8200
+    },
+    {
+      "epoch": 1.5860882858780814,
+      "grad_norm": 1.2306073904037476,
+      "learning_rate": 4.62557603686636e-05,
+      "loss": 0.2801,
+      "step": 8300
+    },
+    {
+      "epoch": 1.605197783298299,
+      "grad_norm": 2.0440738201141357,
+      "learning_rate": 4.620775729646698e-05,
+      "loss": 0.2542,
+      "step": 8400
+    },
+    {
+      "epoch": 1.6243072807185173,
+      "grad_norm": 2.2261791229248047,
+      "learning_rate": 4.6159754224270356e-05,
+      "loss": 0.2575,
+      "step": 8500
+    },
+    {
+      "epoch": 1.643416778138735,
+      "grad_norm": 1.3056387901306152,
+      "learning_rate": 4.6111751152073734e-05,
+      "loss": 0.2149,
+      "step": 8600
+    },
+    {
+      "epoch": 1.6625262755589527,
+      "grad_norm": 1.1693623065948486,
+      "learning_rate": 4.606374807987712e-05,
+      "loss": 0.2344,
+      "step": 8700
+    },
+    {
+      "epoch": 1.6816357729791707,
+      "grad_norm": 1.912260890007019,
+      "learning_rate": 4.601574500768049e-05,
+      "loss": 0.2575,
+      "step": 8800
+    },
+    {
+      "epoch": 1.7007452703993886,
+      "grad_norm": 0.8953251838684082,
+      "learning_rate": 4.596774193548387e-05,
+      "loss": 0.2718,
+      "step": 8900
+    },
+    {
+      "epoch": 1.7198547678196063,
+      "grad_norm": 1.2062451839447021,
+      "learning_rate": 4.5919738863287256e-05,
+      "loss": 0.2208,
+      "step": 9000
+    },
+    {
+      "epoch": 1.7198547678196063,
+      "eval_events-synergy/entsum_processed_loss": 0.15974336862564087,
+      "eval_events-synergy/entsum_processed_runtime": 86.5109,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.887,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.727,
+      "step": 9000
+    },
+    {
+      "epoch": 1.738964265239824,
+      "grad_norm": 1.1229045391082764,
+      "learning_rate": 4.587173579109063e-05,
+      "loss": 0.225,
+      "step": 9100
+    },
+    {
+      "epoch": 1.758073762660042,
+      "grad_norm": 1.2543933391571045,
+      "learning_rate": 4.5823732718894014e-05,
+      "loss": 0.2131,
+      "step": 9200
+    },
+    {
+      "epoch": 1.77718326008026,
+      "grad_norm": 0.8586951494216919,
+      "learning_rate": 4.577572964669739e-05,
+      "loss": 0.2323,
+      "step": 9300
+    },
+    {
+      "epoch": 1.7962927575004777,
+      "grad_norm": 1.023917555809021,
+      "learning_rate": 4.5727726574500764e-05,
+      "loss": 0.2316,
+      "step": 9400
+    },
+    {
+      "epoch": 1.8154022549206956,
+      "grad_norm": 1.0598609447479248,
+      "learning_rate": 4.567972350230415e-05,
+      "loss": 0.2307,
+      "step": 9500
+    },
+    {
+      "epoch": 1.8345117523409136,
+      "grad_norm": 1.7368364334106445,
+      "learning_rate": 4.563172043010753e-05,
+      "loss": 0.2052,
+      "step": 9600
+    },
+    {
+      "epoch": 1.8536212497611313,
+      "grad_norm": 1.440076470375061,
+      "learning_rate": 4.558371735791091e-05,
+      "loss": 0.2044,
+      "step": 9700
+    },
+    {
+      "epoch": 1.872730747181349,
+      "grad_norm": 0.9440079927444458,
+      "learning_rate": 4.5535714285714286e-05,
+      "loss": 0.2446,
+      "step": 9800
+    },
+    {
+      "epoch": 1.891840244601567,
+      "grad_norm": 0.9267650246620178,
+      "learning_rate": 4.548771121351767e-05,
+      "loss": 0.1726,
+      "step": 9900
+    },
+    {
+      "epoch": 1.910949742021785,
+      "grad_norm": 1.5710424184799194,
+      "learning_rate": 4.5439708141321044e-05,
+      "loss": 0.2366,
+      "step": 10000
+    },
+    {
+      "epoch": 1.910949742021785,
+      "eval_events-synergy/entsum_processed_loss": 0.13830891251564026,
+      "eval_events-synergy/entsum_processed_runtime": 86.4674,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.9,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.731,
+      "step": 10000
+    },
+    {
+      "epoch": 1.9300592394420026,
+      "grad_norm": 0.9097877740859985,
+      "learning_rate": 4.539170506912442e-05,
+      "loss": 0.1884,
+      "step": 10100
+    },
+    {
+      "epoch": 1.9491687368622204,
+      "grad_norm": 1.7585983276367188,
+      "learning_rate": 4.534370199692781e-05,
+      "loss": 0.2067,
+      "step": 10200
+    },
+    {
+      "epoch": 1.9682782342824385,
+      "grad_norm": 1.4785964488983154,
+      "learning_rate": 4.5295698924731187e-05,
+      "loss": 0.2044,
+      "step": 10300
+    },
+    {
+      "epoch": 1.9873877317026563,
+      "grad_norm": 2.9566831588745117,
+      "learning_rate": 4.5247695852534565e-05,
+      "loss": 0.1765,
+      "step": 10400
+    },
+    {
+      "epoch": 2.006497229122874,
+      "grad_norm": 0.8678722381591797,
+      "learning_rate": 4.5199692780337944e-05,
+      "loss": 0.1899,
+      "step": 10500
+    },
+    {
+      "epoch": 2.0256067265430917,
+      "grad_norm": 1.2664000988006592,
+      "learning_rate": 4.515168970814132e-05,
+      "loss": 0.1667,
+      "step": 10600
+    },
+    {
+      "epoch": 2.04471622396331,
+      "grad_norm": 1.3981448411941528,
+      "learning_rate": 4.51036866359447e-05,
+      "loss": 0.1902,
+      "step": 10700
+    },
+    {
+      "epoch": 2.0638257213835276,
+      "grad_norm": 0.7927885055541992,
+      "learning_rate": 4.505568356374808e-05,
+      "loss": 0.1743,
+      "step": 10800
+    },
+    {
+      "epoch": 2.0829352188037453,
+      "grad_norm": 1.3861682415008545,
+      "learning_rate": 4.5007680491551466e-05,
+      "loss": 0.1807,
+      "step": 10900
+    },
+    {
+      "epoch": 2.1020447162239635,
+      "grad_norm": 1.4484479427337646,
+      "learning_rate": 4.495967741935484e-05,
+      "loss": 0.1485,
+      "step": 11000
+    },
+    {
+      "epoch": 2.1020447162239635,
+      "eval_events-synergy/entsum_processed_loss": 0.1214805543422699,
+      "eval_events-synergy/entsum_processed_runtime": 86.1466,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.0,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.756,
+      "step": 11000
+    },
+    {
+      "epoch": 2.121154213644181,
+      "grad_norm": 0.9914194941520691,
+      "learning_rate": 4.491167434715822e-05,
+      "loss": 0.1889,
+      "step": 11100
+    },
+    {
+      "epoch": 2.140263711064399,
+      "grad_norm": 0.7929261326789856,
+      "learning_rate": 4.48636712749616e-05,
+      "loss": 0.1562,
+      "step": 11200
+    },
+    {
+      "epoch": 2.1593732084846167,
+      "grad_norm": 1.4530657529830933,
+      "learning_rate": 4.4815668202764974e-05,
+      "loss": 0.1641,
+      "step": 11300
+    },
+    {
+      "epoch": 2.178482705904835,
+      "grad_norm": 1.6586015224456787,
+      "learning_rate": 4.476766513056836e-05,
+      "loss": 0.1759,
+      "step": 11400
+    },
+    {
+      "epoch": 2.1975922033250526,
+      "grad_norm": 1.327807068824768,
+      "learning_rate": 4.471966205837174e-05,
+      "loss": 0.1919,
+      "step": 11500
+    },
+    {
+      "epoch": 2.2167017007452703,
+      "grad_norm": 2.2015275955200195,
+      "learning_rate": 4.467165898617512e-05,
+      "loss": 0.1933,
+      "step": 11600
+    },
+    {
+      "epoch": 2.2358111981654885,
+      "grad_norm": 1.014055609703064,
+      "learning_rate": 4.4623655913978496e-05,
+      "loss": 0.1655,
+      "step": 11700
+    },
+    {
+      "epoch": 2.254920695585706,
+      "grad_norm": 1.8720685243606567,
+      "learning_rate": 4.457565284178188e-05,
+      "loss": 0.1782,
+      "step": 11800
+    },
+    {
+      "epoch": 2.274030193005924,
+      "grad_norm": 1.1258978843688965,
+      "learning_rate": 4.452764976958525e-05,
+      "loss": 0.1533,
+      "step": 11900
+    },
+    {
+      "epoch": 2.2931396904261416,
+      "grad_norm": 1.5024638175964355,
+      "learning_rate": 4.447964669738863e-05,
+      "loss": 0.1114,
+      "step": 12000
+    },
+    {
+      "epoch": 2.2931396904261416,
+      "eval_events-synergy/entsum_processed_loss": 0.10810433328151703,
+      "eval_events-synergy/entsum_processed_runtime": 86.4359,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.91,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.733,
+      "step": 12000
+    },
+    {
+      "epoch": 2.31224918784636,
+      "grad_norm": 1.3650909662246704,
+      "learning_rate": 4.443164362519202e-05,
+      "loss": 0.1627,
+      "step": 12100
+    },
+    {
+      "epoch": 2.3313586852665775,
+      "grad_norm": 1.9314254522323608,
+      "learning_rate": 4.438364055299539e-05,
+      "loss": 0.1495,
+      "step": 12200
+    },
+    {
+      "epoch": 2.3504681826867952,
+      "grad_norm": 0.8482472896575928,
+      "learning_rate": 4.4335637480798775e-05,
+      "loss": 0.1428,
+      "step": 12300
+    },
+    {
+      "epoch": 2.369577680107013,
+      "grad_norm": 0.9307497143745422,
+      "learning_rate": 4.4287634408602154e-05,
+      "loss": 0.1649,
+      "step": 12400
+    },
+    {
+      "epoch": 2.388687177527231,
+      "grad_norm": 1.1554982662200928,
+      "learning_rate": 4.423963133640553e-05,
+      "loss": 0.1257,
+      "step": 12500
+    },
+    {
+      "epoch": 2.407796674947449,
+      "grad_norm": 3.0138888359069824,
+      "learning_rate": 4.419162826420891e-05,
+      "loss": 0.1481,
+      "step": 12600
+    },
+    {
+      "epoch": 2.4269061723676666,
+      "grad_norm": 0.6184584498405457,
+      "learning_rate": 4.414362519201229e-05,
+      "loss": 0.1619,
+      "step": 12700
+    },
+    {
+      "epoch": 2.4460156697878848,
+      "grad_norm": 0.8029180765151978,
+      "learning_rate": 4.409562211981567e-05,
+      "loss": 0.1383,
+      "step": 12800
+    },
+    {
+      "epoch": 2.4651251672081025,
+      "grad_norm": 0.27215951681137085,
+      "learning_rate": 4.404761904761905e-05,
+      "loss": 0.1604,
+      "step": 12900
+    },
+    {
+      "epoch": 2.48423466462832,
+      "grad_norm": 0.5887466073036194,
+      "learning_rate": 4.399961597542243e-05,
+      "loss": 0.1126,
+      "step": 13000
+    },
+    {
+      "epoch": 2.48423466462832,
+      "eval_events-synergy/entsum_processed_loss": 0.09368874132633209,
+      "eval_events-synergy/entsum_processed_runtime": 86.3411,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.94,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.741,
+      "step": 13000
+    },
+    {
+      "epoch": 2.5033441620485384,
+      "grad_norm": 2.6320173740386963,
+      "learning_rate": 4.395161290322581e-05,
+      "loss": 0.1223,
+      "step": 13100
+    },
+    {
+      "epoch": 2.522453659468756,
+      "grad_norm": 1.1739882230758667,
+      "learning_rate": 4.3903609831029184e-05,
+      "loss": 0.1801,
+      "step": 13200
+    },
+    {
+      "epoch": 2.541563156888974,
+      "grad_norm": 0.972703218460083,
+      "learning_rate": 4.385560675883257e-05,
+      "loss": 0.1248,
+      "step": 13300
+    },
+    {
+      "epoch": 2.5606726543091916,
+      "grad_norm": 0.37191712856292725,
+      "learning_rate": 4.380760368663595e-05,
+      "loss": 0.1428,
+      "step": 13400
+    },
+    {
+      "epoch": 2.5797821517294093,
+      "grad_norm": 1.420135498046875,
+      "learning_rate": 4.3759600614439327e-05,
+      "loss": 0.1234,
+      "step": 13500
+    },
+    {
+      "epoch": 2.5988916491496274,
+      "grad_norm": 0.970924437046051,
+      "learning_rate": 4.3711597542242705e-05,
+      "loss": 0.1447,
+      "step": 13600
+    },
+    {
+      "epoch": 2.618001146569845,
+      "grad_norm": 1.2393814325332642,
+      "learning_rate": 4.366359447004609e-05,
+      "loss": 0.1264,
+      "step": 13700
+    },
+    {
+      "epoch": 2.637110643990063,
+      "grad_norm": 1.179800271987915,
+      "learning_rate": 4.361559139784946e-05,
+      "loss": 0.1418,
+      "step": 13800
+    },
+    {
+      "epoch": 2.656220141410281,
+      "grad_norm": 0.8592195510864258,
+      "learning_rate": 4.356758832565284e-05,
+      "loss": 0.1326,
+      "step": 13900
+    },
+    {
+      "epoch": 2.675329638830499,
+      "grad_norm": 2.3403942584991455,
+      "learning_rate": 4.351958525345623e-05,
+      "loss": 0.1579,
+      "step": 14000
+    },
+    {
+      "epoch": 2.675329638830499,
+      "eval_events-synergy/entsum_processed_loss": 0.08366435021162033,
+      "eval_events-synergy/entsum_processed_runtime": 85.9027,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.077,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.775,
+      "step": 14000
+    },
+    {
+      "epoch": 2.6944391362507165,
+      "grad_norm": 1.2683295011520386,
+      "learning_rate": 4.34715821812596e-05,
+      "loss": 0.1452,
+      "step": 14100
+    },
+    {
+      "epoch": 2.7135486336709347,
+      "grad_norm": 0.9251212477684021,
+      "learning_rate": 4.3423579109062984e-05,
+      "loss": 0.1261,
+      "step": 14200
+    },
+    {
+      "epoch": 2.7326581310911524,
+      "grad_norm": 0.2922651171684265,
+      "learning_rate": 4.337557603686636e-05,
+      "loss": 0.1131,
+      "step": 14300
+    },
+    {
+      "epoch": 2.75176762851137,
+      "grad_norm": 0.5204638838768005,
+      "learning_rate": 4.332757296466974e-05,
+      "loss": 0.109,
+      "step": 14400
+    },
+    {
+      "epoch": 2.770877125931588,
+      "grad_norm": 1.6479909420013428,
+      "learning_rate": 4.327956989247312e-05,
+      "loss": 0.1217,
+      "step": 14500
+    },
+    {
+      "epoch": 2.7899866233518056,
+      "grad_norm": 0.693134605884552,
+      "learning_rate": 4.32315668202765e-05,
+      "loss": 0.1259,
+      "step": 14600
+    },
+    {
+      "epoch": 2.8090961207720238,
+      "grad_norm": 0.1304488480091095,
+      "learning_rate": 4.318356374807988e-05,
+      "loss": 0.1271,
+      "step": 14700
+    },
+    {
+      "epoch": 2.8282056181922415,
+      "grad_norm": 1.7225552797317505,
+      "learning_rate": 4.313556067588326e-05,
+      "loss": 0.1205,
+      "step": 14800
+    },
+    {
+      "epoch": 2.847315115612459,
+      "grad_norm": 1.0020724534988403,
+      "learning_rate": 4.308755760368664e-05,
+      "loss": 0.1177,
+      "step": 14900
+    },
+    {
+      "epoch": 2.8664246130326774,
+      "grad_norm": 1.1193674802780151,
+      "learning_rate": 4.3039554531490014e-05,
+      "loss": 0.1099,
+      "step": 15000
+    },
+    {
+      "epoch": 2.8664246130326774,
+      "eval_events-synergy/entsum_processed_loss": 0.07461204379796982,
+      "eval_events-synergy/entsum_processed_runtime": 86.2815,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.958,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.745,
+      "step": 15000
+    },
+    {
+      "epoch": 2.885534110452895,
+      "grad_norm": 1.0606135129928589,
+      "learning_rate": 4.299155145929339e-05,
+      "loss": 0.1193,
+      "step": 15100
+    },
+    {
+      "epoch": 2.904643607873113,
+      "grad_norm": 0.7155654430389404,
+      "learning_rate": 4.294354838709678e-05,
+      "loss": 0.1064,
+      "step": 15200
+    },
+    {
+      "epoch": 2.923753105293331,
+      "grad_norm": 0.7445634007453918,
+      "learning_rate": 4.289554531490016e-05,
+      "loss": 0.1343,
+      "step": 15300
+    },
+    {
+      "epoch": 2.9428626027135487,
+      "grad_norm": 2.339707612991333,
+      "learning_rate": 4.2847542242703536e-05,
+      "loss": 0.1443,
+      "step": 15400
+    },
+    {
+      "epoch": 2.9619721001337664,
+      "grad_norm": 1.8104724884033203,
+      "learning_rate": 4.2799539170506915e-05,
+      "loss": 0.101,
+      "step": 15500
+    },
+    {
+      "epoch": 2.981081597553984,
+      "grad_norm": 1.3034611940383911,
+      "learning_rate": 4.2751536098310294e-05,
+      "loss": 0.1061,
+      "step": 15600
+    },
+    {
+      "epoch": 3.0001910949742023,
+      "grad_norm": 0.9755160212516785,
+      "learning_rate": 4.270353302611367e-05,
+      "loss": 0.1145,
+      "step": 15700
+    },
+    {
+      "epoch": 3.01930059239442,
+      "grad_norm": 1.1601845026016235,
+      "learning_rate": 4.265552995391705e-05,
+      "loss": 0.0902,
+      "step": 15800
+    },
+    {
+      "epoch": 3.038410089814638,
+      "grad_norm": 0.6864597201347351,
+      "learning_rate": 4.260752688172043e-05,
+      "loss": 0.1238,
+      "step": 15900
+    },
+    {
+      "epoch": 3.0575195872348555,
+      "grad_norm": 1.6638556718826294,
+      "learning_rate": 4.255952380952381e-05,
+      "loss": 0.0891,
+      "step": 16000
+    },
+    {
+      "epoch": 3.0575195872348555,
+      "eval_events-synergy/entsum_processed_loss": 0.06794700771570206,
+      "eval_events-synergy/entsum_processed_runtime": 86.5192,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.884,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.727,
+      "step": 16000
+    },
+    {
+      "epoch": 3.0766290846550737,
+      "grad_norm": 0.6231021285057068,
+      "learning_rate": 4.2511520737327194e-05,
+      "loss": 0.1146,
+      "step": 16100
+    },
+    {
+      "epoch": 3.0957385820752914,
+      "grad_norm": 0.8148559927940369,
+      "learning_rate": 4.246351766513057e-05,
+      "loss": 0.0907,
+      "step": 16200
+    },
+    {
+      "epoch": 3.114848079495509,
+      "grad_norm": 0.6905009150505066,
+      "learning_rate": 4.241551459293395e-05,
+      "loss": 0.0984,
+      "step": 16300
+    },
+    {
+      "epoch": 3.1339575769157273,
+      "grad_norm": 0.5867490768432617,
+      "learning_rate": 4.236751152073733e-05,
+      "loss": 0.105,
+      "step": 16400
+    },
+    {
+      "epoch": 3.153067074335945,
+      "grad_norm": 1.3492326736450195,
+      "learning_rate": 4.231950844854071e-05,
+      "loss": 0.0838,
+      "step": 16500
+    },
+    {
+      "epoch": 3.1721765717561627,
+      "grad_norm": 1.4845542907714844,
+      "learning_rate": 4.227150537634409e-05,
+      "loss": 0.0988,
+      "step": 16600
+    },
+    {
+      "epoch": 3.1912860691763805,
+      "grad_norm": 0.6312690377235413,
+      "learning_rate": 4.2223502304147466e-05,
+      "loss": 0.0831,
+      "step": 16700
+    },
+    {
+      "epoch": 3.2103955665965986,
+      "grad_norm": 1.1302180290222168,
+      "learning_rate": 4.217549923195085e-05,
+      "loss": 0.0818,
+      "step": 16800
+    },
+    {
+      "epoch": 3.2295050640168164,
+      "grad_norm": 0.8319332003593445,
+      "learning_rate": 4.2127496159754224e-05,
+      "loss": 0.0874,
+      "step": 16900
+    },
+    {
+      "epoch": 3.248614561437034,
+      "grad_norm": 1.3741902112960815,
+      "learning_rate": 4.20794930875576e-05,
+      "loss": 0.081,
+      "step": 17000
+    },
+    {
+      "epoch": 3.248614561437034,
+      "eval_events-synergy/entsum_processed_loss": 0.06280920654535294,
+      "eval_events-synergy/entsum_processed_runtime": 86.681,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.834,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.714,
+      "step": 17000
+    },
+    {
+      "epoch": 3.267724058857252,
+      "grad_norm": 1.1995327472686768,
+      "learning_rate": 4.203149001536099e-05,
+      "loss": 0.097,
+      "step": 17100
+    },
+    {
+      "epoch": 3.28683355627747,
+      "grad_norm": 0.8017995357513428,
+      "learning_rate": 4.198348694316436e-05,
+      "loss": 0.0873,
+      "step": 17200
+    },
+    {
+      "epoch": 3.3059430536976877,
+      "grad_norm": 0.2565690875053406,
+      "learning_rate": 4.1935483870967746e-05,
+      "loss": 0.1096,
+      "step": 17300
+    },
+    {
+      "epoch": 3.3250525511179054,
+      "grad_norm": 0.5259280204772949,
+      "learning_rate": 4.1887480798771124e-05,
+      "loss": 0.0853,
+      "step": 17400
+    },
+    {
+      "epoch": 3.3441620485381236,
+      "grad_norm": 1.4056413173675537,
+      "learning_rate": 4.18394777265745e-05,
+      "loss": 0.0977,
+      "step": 17500
+    },
+    {
+      "epoch": 3.3632715459583413,
+      "grad_norm": 0.9580443501472473,
+      "learning_rate": 4.179147465437788e-05,
+      "loss": 0.0895,
+      "step": 17600
+    },
+    {
+      "epoch": 3.382381043378559,
+      "grad_norm": 0.5705116391181946,
+      "learning_rate": 4.174347158218126e-05,
+      "loss": 0.0937,
+      "step": 17700
+    },
+    {
+      "epoch": 3.401490540798777,
+      "grad_norm": 0.7241528630256653,
+      "learning_rate": 4.169546850998464e-05,
+      "loss": 0.0952,
+      "step": 17800
+    },
+    {
+      "epoch": 3.420600038218995,
+      "grad_norm": 0.5498321652412415,
+      "learning_rate": 4.164746543778802e-05,
+      "loss": 0.0827,
+      "step": 17900
+    },
+    {
+      "epoch": 3.4397095356392127,
+      "grad_norm": 0.7606751918792725,
+      "learning_rate": 4.1599462365591404e-05,
+      "loss": 0.0755,
+      "step": 18000
+    },
+    {
+      "epoch": 3.4397095356392127,
+      "eval_events-synergy/entsum_processed_loss": 0.05818604305386543,
+      "eval_events-synergy/entsum_processed_runtime": 86.615,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.854,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.719,
+      "step": 18000
+    },
+    {
+      "epoch": 3.4588190330594304,
+      "grad_norm": 1.3093830347061157,
+      "learning_rate": 4.1551459293394776e-05,
+      "loss": 0.0863,
+      "step": 18100
+    },
+    {
+      "epoch": 3.4779285304796486,
+      "grad_norm": 0.3217012286186218,
+      "learning_rate": 4.1503456221198154e-05,
+      "loss": 0.093,
+      "step": 18200
+    },
+    {
+      "epoch": 3.4970380278998663,
+      "grad_norm": 2.5801539421081543,
+      "learning_rate": 4.145545314900154e-05,
+      "loss": 0.0735,
+      "step": 18300
+    },
+    {
+      "epoch": 3.516147525320084,
+      "grad_norm": 1.4098963737487793,
+      "learning_rate": 4.140745007680492e-05,
+      "loss": 0.0632,
+      "step": 18400
+    },
+    {
+      "epoch": 3.5352570227403017,
+      "grad_norm": 2.017245054244995,
+      "learning_rate": 4.13594470046083e-05,
+      "loss": 0.0771,
+      "step": 18500
+    },
+    {
+      "epoch": 3.55436652016052,
+      "grad_norm": 0.4717109203338623,
+      "learning_rate": 4.1311443932411676e-05,
+      "loss": 0.0806,
+      "step": 18600
+    },
+    {
+      "epoch": 3.5734760175807376,
+      "grad_norm": 1.616327166557312,
+      "learning_rate": 4.1263440860215055e-05,
+      "loss": 0.0702,
+      "step": 18700
+    },
+    {
+      "epoch": 3.5925855150009554,
+      "grad_norm": 0.4756339192390442,
+      "learning_rate": 4.1215437788018434e-05,
+      "loss": 0.0893,
+      "step": 18800
+    },
+    {
+      "epoch": 3.6116950124211735,
+      "grad_norm": 0.8106458783149719,
+      "learning_rate": 4.116743471582181e-05,
+      "loss": 0.0608,
+      "step": 18900
+    },
+    {
+      "epoch": 3.6308045098413912,
+      "grad_norm": 0.7149679660797119,
+      "learning_rate": 4.11194316436252e-05,
+      "loss": 0.083,
+      "step": 19000
+    },
+    {
+      "epoch": 3.6308045098413912,
+      "eval_events-synergy/entsum_processed_loss": 0.053829826414585114,
+      "eval_events-synergy/entsum_processed_runtime": 86.5846,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.864,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.722,
+      "step": 19000
+    },
+    {
+      "epoch": 3.649914007261609,
+      "grad_norm": 0.7926931381225586,
+      "learning_rate": 4.107142857142857e-05,
+      "loss": 0.0849,
+      "step": 19100
+    },
+    {
+      "epoch": 3.669023504681827,
+      "grad_norm": 0.5450174808502197,
+      "learning_rate": 4.1023425499231955e-05,
+      "loss": 0.0651,
+      "step": 19200
+    },
+    {
+      "epoch": 3.688133002102045,
+      "grad_norm": 1.2390661239624023,
+      "learning_rate": 4.0975422427035334e-05,
+      "loss": 0.0768,
+      "step": 19300
+    },
+    {
+      "epoch": 3.7072424995222626,
+      "grad_norm": 1.3439677953720093,
+      "learning_rate": 4.092741935483871e-05,
+      "loss": 0.1014,
+      "step": 19400
+    },
+    {
+      "epoch": 3.7263519969424803,
+      "grad_norm": 0.7183663845062256,
+      "learning_rate": 4.087941628264209e-05,
+      "loss": 0.082,
+      "step": 19500
+    },
+    {
+      "epoch": 3.745461494362698,
+      "grad_norm": 1.6611928939819336,
+      "learning_rate": 4.083141321044547e-05,
+      "loss": 0.0732,
+      "step": 19600
+    },
+    {
+      "epoch": 3.764570991782916,
+      "grad_norm": 1.2751322984695435,
+      "learning_rate": 4.078341013824885e-05,
+      "loss": 0.0768,
+      "step": 19700
+    },
+    {
+      "epoch": 3.783680489203134,
+      "grad_norm": 1.3333556652069092,
+      "learning_rate": 4.073540706605223e-05,
+      "loss": 0.0794,
+      "step": 19800
+    },
+    {
+      "epoch": 3.8027899866233517,
+      "grad_norm": 0.6069827675819397,
+      "learning_rate": 4.068740399385561e-05,
+      "loss": 0.0846,
+      "step": 19900
+    },
+    {
+      "epoch": 3.82189948404357,
+      "grad_norm": 0.38667362928390503,
+      "learning_rate": 4.0639400921658985e-05,
+      "loss": 0.0894,
+      "step": 20000
+    },
+    {
+      "epoch": 3.82189948404357,
+      "eval_events-synergy/entsum_processed_loss": 0.04852017015218735,
+      "eval_events-synergy/entsum_processed_runtime": 86.5692,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.869,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.723,
+      "step": 20000
+    },
+    {
+      "epoch": 3.8410089814637876,
+      "grad_norm": 1.1866488456726074,
+      "learning_rate": 4.0591397849462364e-05,
+      "loss": 0.055,
+      "step": 20100
+    },
+    {
+      "epoch": 3.8601184788840053,
+      "grad_norm": 1.6856223344802856,
+      "learning_rate": 4.054339477726575e-05,
+      "loss": 0.0713,
+      "step": 20200
+    },
+    {
+      "epoch": 3.8792279763042234,
+      "grad_norm": 0.5651134252548218,
+      "learning_rate": 4.049539170506912e-05,
+      "loss": 0.075,
+      "step": 20300
+    },
+    {
+      "epoch": 3.898337473724441,
+      "grad_norm": 0.9011269807815552,
+      "learning_rate": 4.044738863287251e-05,
+      "loss": 0.061,
+      "step": 20400
+    },
+    {
+      "epoch": 3.917446971144659,
+      "grad_norm": 1.215333104133606,
+      "learning_rate": 4.0399385560675886e-05,
+      "loss": 0.0704,
+      "step": 20500
+    },
+    {
+      "epoch": 3.9365564685648766,
+      "grad_norm": 0.8905503749847412,
+      "learning_rate": 4.0351382488479264e-05,
+      "loss": 0.0804,
+      "step": 20600
+    },
+    {
+      "epoch": 3.9556659659850943,
+      "grad_norm": 0.3945884108543396,
+      "learning_rate": 4.030337941628264e-05,
+      "loss": 0.0731,
+      "step": 20700
+    },
+    {
+      "epoch": 3.9747754634053125,
+      "grad_norm": 1.3491140604019165,
+      "learning_rate": 4.025537634408602e-05,
+      "loss": 0.0811,
+      "step": 20800
+    },
+    {
+      "epoch": 3.9938849608255302,
+      "grad_norm": 0.6310815811157227,
+      "learning_rate": 4.02073732718894e-05,
+      "loss": 0.0609,
+      "step": 20900
+    },
+    {
+      "epoch": 4.012994458245748,
+      "grad_norm": 0.8954095840454102,
+      "learning_rate": 4.015937019969278e-05,
+      "loss": 0.0605,
+      "step": 21000
+    },
+    {
+      "epoch": 4.012994458245748,
+      "eval_events-synergy/entsum_processed_loss": 0.04423968493938446,
+      "eval_events-synergy/entsum_processed_runtime": 85.8578,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.091,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.779,
+      "step": 21000
+    },
+    {
+      "epoch": 4.032103955665966,
+      "grad_norm": 0.7362313866615295,
+      "learning_rate": 4.0111367127496165e-05,
+      "loss": 0.0473,
+      "step": 21100
+    },
+    {
+      "epoch": 4.051213453086183,
+      "grad_norm": 0.45443108677864075,
+      "learning_rate": 4.0063364055299544e-05,
+      "loss": 0.0526,
+      "step": 21200
+    },
+    {
+      "epoch": 4.070322950506402,
+      "grad_norm": 0.5123448371887207,
+      "learning_rate": 4.001536098310292e-05,
+      "loss": 0.0604,
+      "step": 21300
+    },
+    {
+      "epoch": 4.08943244792662,
+      "grad_norm": 1.018021583557129,
+      "learning_rate": 3.99673579109063e-05,
+      "loss": 0.0629,
+      "step": 21400
+    },
+    {
+      "epoch": 4.108541945346837,
+      "grad_norm": 1.2121423482894897,
+      "learning_rate": 3.991935483870968e-05,
+      "loss": 0.0554,
+      "step": 21500
+    },
+    {
+      "epoch": 4.127651442767055,
+      "grad_norm": 1.4117833375930786,
+      "learning_rate": 3.987135176651306e-05,
+      "loss": 0.0562,
+      "step": 21600
+    },
+    {
+      "epoch": 4.146760940187273,
+      "grad_norm": 0.7586542367935181,
+      "learning_rate": 3.982334869431644e-05,
+      "loss": 0.0604,
+      "step": 21700
+    },
+    {
+      "epoch": 4.165870437607491,
+      "grad_norm": 1.4312876462936401,
+      "learning_rate": 3.977534562211982e-05,
+      "loss": 0.0683,
+      "step": 21800
+    },
+    {
+      "epoch": 4.184979935027709,
+      "grad_norm": 1.278427243232727,
+      "learning_rate": 3.9727342549923195e-05,
+      "loss": 0.0607,
+      "step": 21900
+    },
+    {
+      "epoch": 4.204089432447927,
+      "grad_norm": 2.7680790424346924,
+      "learning_rate": 3.9679339477726574e-05,
+      "loss": 0.0535,
+      "step": 22000
+    },
+    {
+      "epoch": 4.204089432447927,
+      "eval_events-synergy/entsum_processed_loss": 0.04234563931822777,
+      "eval_events-synergy/entsum_processed_runtime": 86.4674,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.9,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.731,
+      "step": 22000
+    },
+    {
+      "epoch": 4.223198929868144,
+      "grad_norm": 1.4007564783096313,
+      "learning_rate": 3.963133640552996e-05,
+      "loss": 0.0715,
+      "step": 22100
+    },
+    {
+      "epoch": 4.242308427288362,
+      "grad_norm": 0.6837427020072937,
+      "learning_rate": 3.958333333333333e-05,
+      "loss": 0.0611,
+      "step": 22200
+    },
+    {
+      "epoch": 4.26141792470858,
+      "grad_norm": 1.1406750679016113,
+      "learning_rate": 3.9535330261136717e-05,
+      "loss": 0.055,
+      "step": 22300
+    },
+    {
+      "epoch": 4.280527422128798,
+      "grad_norm": 0.38464096188545227,
+      "learning_rate": 3.9487327188940095e-05,
+      "loss": 0.0562,
+      "step": 22400
+    },
+    {
+      "epoch": 4.299636919549016,
+      "grad_norm": 1.4123332500457764,
+      "learning_rate": 3.9439324116743474e-05,
+      "loss": 0.0505,
+      "step": 22500
+    },
+    {
+      "epoch": 4.318746416969233,
+      "grad_norm": 0.24340485036373138,
+      "learning_rate": 3.939132104454685e-05,
+      "loss": 0.0642,
+      "step": 22600
+    },
+    {
+      "epoch": 4.3378559143894515,
+      "grad_norm": 0.541613757610321,
+      "learning_rate": 3.934331797235023e-05,
+      "loss": 0.059,
+      "step": 22700
+    },
+    {
+      "epoch": 4.35696541180967,
+      "grad_norm": 1.4583444595336914,
+      "learning_rate": 3.929531490015361e-05,
+      "loss": 0.0515,
+      "step": 22800
+    },
+    {
+      "epoch": 4.376074909229887,
+      "grad_norm": 1.1488178968429565,
+      "learning_rate": 3.924731182795699e-05,
+      "loss": 0.0532,
+      "step": 22900
+    },
+    {
+      "epoch": 4.395184406650105,
+      "grad_norm": 1.302866816520691,
+      "learning_rate": 3.9199308755760374e-05,
+      "loss": 0.0515,
+      "step": 23000
+    },
+    {
+      "epoch": 4.395184406650105,
+      "eval_events-synergy/entsum_processed_loss": 0.04016019403934479,
+      "eval_events-synergy/entsum_processed_runtime": 86.5369,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.879,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.725,
+      "step": 23000
+    },
+    {
+      "epoch": 4.414293904070323,
+      "grad_norm": 0.5283303260803223,
+      "learning_rate": 3.9151305683563746e-05,
+      "loss": 0.0534,
+      "step": 23100
+    },
+    {
+      "epoch": 4.433403401490541,
+      "grad_norm": 0.9314780831336975,
+      "learning_rate": 3.910330261136713e-05,
+      "loss": 0.05,
+      "step": 23200
+    },
+    {
+      "epoch": 4.452512898910759,
+      "grad_norm": 0.44866007566452026,
+      "learning_rate": 3.905529953917051e-05,
+      "loss": 0.0584,
+      "step": 23300
+    },
+    {
+      "epoch": 4.471622396330977,
+      "grad_norm": 0.8532125353813171,
+      "learning_rate": 3.900729646697389e-05,
+      "loss": 0.0651,
+      "step": 23400
+    },
+    {
+      "epoch": 4.490731893751194,
+      "grad_norm": 2.1083550453186035,
+      "learning_rate": 3.895929339477727e-05,
+      "loss": 0.0498,
+      "step": 23500
+    },
+    {
+      "epoch": 4.509841391171412,
+      "grad_norm": 0.4969152808189392,
+      "learning_rate": 3.891129032258065e-05,
+      "loss": 0.0538,
+      "step": 23600
+    },
+    {
+      "epoch": 4.5289508885916305,
+      "grad_norm": 0.577306866645813,
+      "learning_rate": 3.8863287250384026e-05,
+      "loss": 0.0388,
+      "step": 23700
+    },
+    {
+      "epoch": 4.548060386011848,
+      "grad_norm": 0.08934194594621658,
+      "learning_rate": 3.8815284178187404e-05,
+      "loss": 0.0487,
+      "step": 23800
+    },
+    {
+      "epoch": 4.567169883432066,
+      "grad_norm": 1.1038001775741577,
+      "learning_rate": 3.876728110599078e-05,
+      "loss": 0.0558,
+      "step": 23900
+    },
+    {
+      "epoch": 4.586279380852283,
+      "grad_norm": 1.019621729850769,
+      "learning_rate": 3.871927803379416e-05,
+      "loss": 0.0659,
+      "step": 24000
+    },
+    {
+      "epoch": 4.586279380852283,
+      "eval_events-synergy/entsum_processed_loss": 0.03685642033815384,
+      "eval_events-synergy/entsum_processed_runtime": 84.6028,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.493,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.879,
+      "step": 24000
+    },
+    {
+      "epoch": 4.605388878272501,
+      "grad_norm": 0.8093920350074768,
+      "learning_rate": 3.867127496159754e-05,
+      "loss": 0.0452,
+      "step": 24100
+    },
+    {
+      "epoch": 4.62449837569272,
+      "grad_norm": 1.2581048011779785,
+      "learning_rate": 3.8623271889400926e-05,
+      "loss": 0.0445,
+      "step": 24200
+    },
+    {
+      "epoch": 4.643607873112937,
+      "grad_norm": 0.908857524394989,
+      "learning_rate": 3.8575268817204305e-05,
+      "loss": 0.0469,
+      "step": 24300
+    },
+    {
+      "epoch": 4.662717370533155,
+      "grad_norm": 1.6877511739730835,
+      "learning_rate": 3.8527265745007684e-05,
+      "loss": 0.062,
+      "step": 24400
+    },
+    {
+      "epoch": 4.681826867953372,
+      "grad_norm": 0.47951963543891907,
+      "learning_rate": 3.847926267281106e-05,
+      "loss": 0.0761,
+      "step": 24500
+    },
+    {
+      "epoch": 4.7009363653735905,
+      "grad_norm": 0.7642994523048401,
+      "learning_rate": 3.843125960061444e-05,
+      "loss": 0.0534,
+      "step": 24600
+    },
+    {
+      "epoch": 4.720045862793809,
+      "grad_norm": 0.6277188062667847,
+      "learning_rate": 3.838325652841782e-05,
+      "loss": 0.0521,
+      "step": 24700
+    },
+    {
+      "epoch": 4.739155360214026,
+      "grad_norm": 0.9474277496337891,
+      "learning_rate": 3.83352534562212e-05,
+      "loss": 0.0625,
+      "step": 24800
+    },
+    {
+      "epoch": 4.758264857634244,
+      "grad_norm": 0.6333776116371155,
+      "learning_rate": 3.8287250384024584e-05,
+      "loss": 0.0384,
+      "step": 24900
+    },
+    {
+      "epoch": 4.777374355054462,
+      "grad_norm": 0.5701904892921448,
+      "learning_rate": 3.8239247311827956e-05,
+      "loss": 0.0574,
+      "step": 25000
+    },
+    {
+      "epoch": 4.777374355054462,
+      "eval_events-synergy/entsum_processed_loss": 0.03454362228512764,
+      "eval_events-synergy/entsum_processed_runtime": 86.3766,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.929,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.738,
+      "step": 25000
+    },
+    {
+      "epoch": 4.79648385247468,
+      "grad_norm": 1.0204541683197021,
+      "learning_rate": 3.819124423963134e-05,
+      "loss": 0.0551,
+      "step": 25100
+    },
+    {
+      "epoch": 4.815593349894898,
+      "grad_norm": 1.5141241550445557,
+      "learning_rate": 3.814324116743472e-05,
+      "loss": 0.0529,
+      "step": 25200
+    },
+    {
+      "epoch": 4.834702847315116,
+      "grad_norm": 1.2822602987289429,
+      "learning_rate": 3.809523809523809e-05,
+      "loss": 0.0627,
+      "step": 25300
+    },
+    {
+      "epoch": 4.853812344735333,
+      "grad_norm": 0.21899272501468658,
+      "learning_rate": 3.804723502304148e-05,
+      "loss": 0.0383,
+      "step": 25400
+    },
+    {
+      "epoch": 4.872921842155551,
+      "grad_norm": 0.30676454305648804,
+      "learning_rate": 3.7999231950844856e-05,
+      "loss": 0.0453,
+      "step": 25500
+    },
+    {
+      "epoch": 4.8920313395757695,
+      "grad_norm": 0.33828213810920715,
+      "learning_rate": 3.7951228878648235e-05,
+      "loss": 0.0402,
+      "step": 25600
+    },
+    {
+      "epoch": 4.911140836995987,
+      "grad_norm": 0.7716867327690125,
+      "learning_rate": 3.7903225806451614e-05,
+      "loss": 0.0523,
+      "step": 25700
+    },
+    {
+      "epoch": 4.930250334416205,
+      "grad_norm": 0.31507009267807007,
+      "learning_rate": 3.785522273425499e-05,
+      "loss": 0.0415,
+      "step": 25800
+    },
+    {
+      "epoch": 4.949359831836423,
+      "grad_norm": 0.7771691083908081,
+      "learning_rate": 3.780721966205837e-05,
+      "loss": 0.0581,
+      "step": 25900
+    },
+    {
+      "epoch": 4.96846932925664,
+      "grad_norm": 1.4588698148727417,
+      "learning_rate": 3.775921658986175e-05,
+      "loss": 0.0501,
+      "step": 26000
+    },
+    {
+      "epoch": 4.96846932925664,
+      "eval_events-synergy/entsum_processed_loss": 0.03175923973321915,
+      "eval_events-synergy/entsum_processed_runtime": 85.1461,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.318,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.835,
+      "step": 26000
+    },
+    {
+      "epoch": 4.987578826676859,
+      "grad_norm": 0.24702726304531097,
+      "learning_rate": 3.7711213517665136e-05,
+      "loss": 0.0488,
+      "step": 26100
+    },
+    {
+      "epoch": 5.006688324097076,
+      "grad_norm": 0.3701643645763397,
+      "learning_rate": 3.766321044546851e-05,
+      "loss": 0.0422,
+      "step": 26200
+    },
+    {
+      "epoch": 5.025797821517294,
+      "grad_norm": 0.2334865927696228,
+      "learning_rate": 3.761520737327189e-05,
+      "loss": 0.0369,
+      "step": 26300
+    },
+    {
+      "epoch": 5.044907318937512,
+      "grad_norm": 0.4860515892505646,
+      "learning_rate": 3.756720430107527e-05,
+      "loss": 0.0467,
+      "step": 26400
+    },
+    {
+      "epoch": 5.0640168163577295,
+      "grad_norm": 1.0469186305999756,
+      "learning_rate": 3.751920122887865e-05,
+      "loss": 0.0388,
+      "step": 26500
+    },
+    {
+      "epoch": 5.083126313777948,
+      "grad_norm": 0.10929598659276962,
+      "learning_rate": 3.747119815668203e-05,
+      "loss": 0.039,
+      "step": 26600
+    },
+    {
+      "epoch": 5.102235811198166,
+      "grad_norm": 0.9259900450706482,
+      "learning_rate": 3.742319508448541e-05,
+      "loss": 0.0537,
+      "step": 26700
+    },
+    {
+      "epoch": 5.121345308618383,
+      "grad_norm": 0.6803724765777588,
+      "learning_rate": 3.737519201228879e-05,
+      "loss": 0.0429,
+      "step": 26800
+    },
+    {
+      "epoch": 5.140454806038601,
+      "grad_norm": 0.35482415556907654,
+      "learning_rate": 3.7327188940092166e-05,
+      "loss": 0.0454,
+      "step": 26900
+    },
+    {
+      "epoch": 5.1595643034588194,
+      "grad_norm": 0.4932660162448883,
+      "learning_rate": 3.727918586789555e-05,
+      "loss": 0.0389,
+      "step": 27000
+    },
+    {
+      "epoch": 5.1595643034588194,
+      "eval_events-synergy/entsum_processed_loss": 0.0308841560035944,
+      "eval_events-synergy/entsum_processed_runtime": 86.4654,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.901,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.731,
+      "step": 27000
+    },
+    {
+      "epoch": 5.178673800879037,
+      "grad_norm": 0.7944111824035645,
+      "learning_rate": 3.723118279569893e-05,
+      "loss": 0.0367,
+      "step": 27100
+    },
+    {
+      "epoch": 5.197783298299255,
+      "grad_norm": 0.3408133387565613,
+      "learning_rate": 3.71831797235023e-05,
+      "loss": 0.0412,
+      "step": 27200
+    },
+    {
+      "epoch": 5.216892795719472,
+      "grad_norm": 1.3057624101638794,
+      "learning_rate": 3.713517665130569e-05,
+      "loss": 0.0313,
+      "step": 27300
+    },
+    {
+      "epoch": 5.23600229313969,
+      "grad_norm": 0.28078633546829224,
+      "learning_rate": 3.7087173579109066e-05,
+      "loss": 0.0432,
+      "step": 27400
+    },
+    {
+      "epoch": 5.2551117905599085,
+      "grad_norm": 0.3789531886577606,
+      "learning_rate": 3.7039170506912445e-05,
+      "loss": 0.0415,
+      "step": 27500
+    },
+    {
+      "epoch": 5.274221287980126,
+      "grad_norm": 0.5109480619430542,
+      "learning_rate": 3.6991167434715824e-05,
+      "loss": 0.0328,
+      "step": 27600
+    },
+    {
+      "epoch": 5.293330785400344,
+      "grad_norm": 0.010233950801193714,
+      "learning_rate": 3.69431643625192e-05,
+      "loss": 0.0399,
+      "step": 27700
+    },
+    {
+      "epoch": 5.312440282820562,
+      "grad_norm": 0.4560386538505554,
+      "learning_rate": 3.689516129032258e-05,
+      "loss": 0.0387,
+      "step": 27800
+    },
+    {
+      "epoch": 5.331549780240779,
+      "grad_norm": 0.1968875378370285,
+      "learning_rate": 3.684715821812596e-05,
+      "loss": 0.0371,
+      "step": 27900
+    },
+    {
+      "epoch": 5.350659277660998,
+      "grad_norm": 0.8004895448684692,
+      "learning_rate": 3.6799155145929345e-05,
+      "loss": 0.0437,
+      "step": 28000
+    },
+    {
+      "epoch": 5.350659277660998,
+      "eval_events-synergy/entsum_processed_loss": 0.029284106567502022,
+      "eval_events-synergy/entsum_processed_runtime": 85.8087,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.107,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.783,
+      "step": 28000
+    },
+    {
+      "epoch": 5.369768775081216,
+      "grad_norm": 0.9914371371269226,
+      "learning_rate": 3.675115207373272e-05,
+      "loss": 0.0352,
+      "step": 28100
+    },
+    {
+      "epoch": 5.388878272501433,
+      "grad_norm": 1.3968762159347534,
+      "learning_rate": 3.67031490015361e-05,
+      "loss": 0.0305,
+      "step": 28200
+    },
+    {
+      "epoch": 5.407987769921651,
+      "grad_norm": 0.677922785282135,
+      "learning_rate": 3.665514592933948e-05,
+      "loss": 0.0321,
+      "step": 28300
+    },
+    {
+      "epoch": 5.427097267341869,
+      "grad_norm": 0.5966954231262207,
+      "learning_rate": 3.6607142857142853e-05,
+      "loss": 0.0356,
+      "step": 28400
+    },
+    {
+      "epoch": 5.446206764762087,
+      "grad_norm": 0.47409358620643616,
+      "learning_rate": 3.655913978494624e-05,
+      "loss": 0.0281,
+      "step": 28500
+    },
+    {
+      "epoch": 5.465316262182305,
+      "grad_norm": 0.5637366771697998,
+      "learning_rate": 3.651113671274962e-05,
+      "loss": 0.0374,
+      "step": 28600
+    },
+    {
+      "epoch": 5.484425759602522,
+      "grad_norm": 0.5510061383247375,
+      "learning_rate": 3.6463133640552996e-05,
+      "loss": 0.0321,
+      "step": 28700
+    },
+    {
+      "epoch": 5.50353525702274,
+      "grad_norm": 0.7010207176208496,
+      "learning_rate": 3.6415130568356375e-05,
+      "loss": 0.0368,
+      "step": 28800
+    },
+    {
+      "epoch": 5.522644754442958,
+      "grad_norm": 1.2387152910232544,
+      "learning_rate": 3.636712749615976e-05,
+      "loss": 0.0397,
+      "step": 28900
+    },
+    {
+      "epoch": 5.541754251863176,
+      "grad_norm": 1.7459216117858887,
+      "learning_rate": 3.631912442396313e-05,
+      "loss": 0.0386,
+      "step": 29000
+    },
+    {
+      "epoch": 5.541754251863176,
+      "eval_events-synergy/entsum_processed_loss": 0.02899482287466526,
+      "eval_events-synergy/entsum_processed_runtime": 84.3392,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.579,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.901,
+      "step": 29000
+    },
+    {
+      "epoch": 5.560863749283394,
+      "grad_norm": 0.6622141003608704,
+      "learning_rate": 3.627112135176651e-05,
+      "loss": 0.0368,
+      "step": 29100
+    },
+    {
+      "epoch": 5.579973246703612,
+      "grad_norm": 1.0212815999984741,
+      "learning_rate": 3.62231182795699e-05,
+      "loss": 0.0373,
+      "step": 29200
+    },
+    {
+      "epoch": 5.599082744123829,
+      "grad_norm": 0.4521411657333374,
+      "learning_rate": 3.6175115207373276e-05,
+      "loss": 0.0365,
+      "step": 29300
+    },
+    {
+      "epoch": 5.6181922415440475,
+      "grad_norm": 0.8316215872764587,
+      "learning_rate": 3.6127112135176654e-05,
+      "loss": 0.0401,
+      "step": 29400
+    },
+    {
+      "epoch": 5.637301738964265,
+      "grad_norm": 0.3077566623687744,
+      "learning_rate": 3.607910906298003e-05,
+      "loss": 0.044,
+      "step": 29500
+    },
+    {
+      "epoch": 5.656411236384483,
+      "grad_norm": 1.1517149209976196,
+      "learning_rate": 3.603110599078341e-05,
+      "loss": 0.0355,
+      "step": 29600
+    },
+    {
+      "epoch": 5.675520733804701,
+      "grad_norm": 1.3990229368209839,
+      "learning_rate": 3.598310291858679e-05,
+      "loss": 0.0349,
+      "step": 29700
+    },
+    {
+      "epoch": 5.694630231224918,
+      "grad_norm": 1.5779931545257568,
+      "learning_rate": 3.593509984639017e-05,
+      "loss": 0.0393,
+      "step": 29800
+    },
+    {
+      "epoch": 5.713739728645137,
+      "grad_norm": 1.200478434562683,
+      "learning_rate": 3.5887096774193555e-05,
+      "loss": 0.0333,
+      "step": 29900
+    },
+    {
+      "epoch": 5.732849226065355,
+      "grad_norm": 1.5373812913894653,
+      "learning_rate": 3.583909370199693e-05,
+      "loss": 0.0403,
+      "step": 30000
+    },
+    {
+      "epoch": 5.732849226065355,
+      "eval_events-synergy/entsum_processed_loss": 0.02680058218538761,
+      "eval_events-synergy/entsum_processed_runtime": 86.269,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.962,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.746,
+      "step": 30000
+    },
+    {
+      "epoch": 5.751958723485572,
+      "grad_norm": 0.24801170825958252,
+      "learning_rate": 3.579109062980031e-05,
+      "loss": 0.0383,
+      "step": 30100
+    },
+    {
+      "epoch": 5.77106822090579,
+      "grad_norm": 1.638857126235962,
+      "learning_rate": 3.574308755760369e-05,
+      "loss": 0.0466,
+      "step": 30200
+    },
+    {
+      "epoch": 5.790177718326008,
+      "grad_norm": 0.6621894836425781,
+      "learning_rate": 3.569508448540706e-05,
+      "loss": 0.0332,
+      "step": 30300
+    },
+    {
+      "epoch": 5.809287215746226,
+      "grad_norm": 0.7041098475456238,
+      "learning_rate": 3.564708141321045e-05,
+      "loss": 0.0252,
+      "step": 30400
+    },
+    {
+      "epoch": 5.828396713166444,
+      "grad_norm": 0.1067354753613472,
+      "learning_rate": 3.559907834101383e-05,
+      "loss": 0.029,
+      "step": 30500
+    },
+    {
+      "epoch": 5.847506210586662,
+      "grad_norm": 0.8987241387367249,
+      "learning_rate": 3.5551075268817206e-05,
+      "loss": 0.039,
+      "step": 30600
+    },
+    {
+      "epoch": 5.866615708006879,
+      "grad_norm": 2.081352710723877,
+      "learning_rate": 3.5503072196620585e-05,
+      "loss": 0.033,
+      "step": 30700
+    },
+    {
+      "epoch": 5.885725205427097,
+      "grad_norm": 1.5779916048049927,
+      "learning_rate": 3.545506912442397e-05,
+      "loss": 0.0337,
+      "step": 30800
+    },
+    {
+      "epoch": 5.904834702847316,
+      "grad_norm": 0.8174372911453247,
+      "learning_rate": 3.540706605222734e-05,
+      "loss": 0.0355,
+      "step": 30900
+    },
+    {
+      "epoch": 5.923944200267533,
+      "grad_norm": 0.35382506251335144,
+      "learning_rate": 3.535906298003072e-05,
+      "loss": 0.0477,
+      "step": 31000
+    },
+    {
+      "epoch": 5.923944200267533,
+      "eval_events-synergy/entsum_processed_loss": 0.025150079280138016,
+      "eval_events-synergy/entsum_processed_runtime": 86.0706,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.024,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.762,
+      "step": 31000
+    },
+    {
+      "epoch": 5.943053697687751,
+      "grad_norm": 2.253109931945801,
+      "learning_rate": 3.5311059907834107e-05,
+      "loss": 0.0349,
+      "step": 31100
+    },
+    {
+      "epoch": 5.962163195107968,
+      "grad_norm": 0.9015364646911621,
+      "learning_rate": 3.526305683563748e-05,
+      "loss": 0.0284,
+      "step": 31200
+    },
+    {
+      "epoch": 5.9812726925281865,
+      "grad_norm": 0.24150536954402924,
+      "learning_rate": 3.5215053763440864e-05,
+      "loss": 0.0379,
+      "step": 31300
+    },
+    {
+      "epoch": 6.000382189948405,
+      "grad_norm": 0.7606573700904846,
+      "learning_rate": 3.516705069124424e-05,
+      "loss": 0.0285,
+      "step": 31400
+    },
+    {
+      "epoch": 6.019491687368622,
+      "grad_norm": 0.6848638653755188,
+      "learning_rate": 3.511904761904762e-05,
+      "loss": 0.0301,
+      "step": 31500
+    },
+    {
+      "epoch": 6.03860118478884,
+      "grad_norm": 1.088733196258545,
+      "learning_rate": 3.5071044546851e-05,
+      "loss": 0.0265,
+      "step": 31600
+    },
+    {
+      "epoch": 6.057710682209058,
+      "grad_norm": 1.046855092048645,
+      "learning_rate": 3.502304147465438e-05,
+      "loss": 0.0269,
+      "step": 31700
+    },
+    {
+      "epoch": 6.076820179629276,
+      "grad_norm": 0.9664869904518127,
+      "learning_rate": 3.497503840245776e-05,
+      "loss": 0.0234,
+      "step": 31800
+    },
+    {
+      "epoch": 6.095929677049494,
+      "grad_norm": 0.5576639175415039,
+      "learning_rate": 3.4927035330261136e-05,
+      "loss": 0.0332,
+      "step": 31900
+    },
+    {
+      "epoch": 6.115039174469711,
+      "grad_norm": 0.839173436164856,
+      "learning_rate": 3.487903225806452e-05,
+      "loss": 0.0301,
+      "step": 32000
+    },
+    {
+      "epoch": 6.115039174469711,
+      "eval_events-synergy/entsum_processed_loss": 0.02411143109202385,
+      "eval_events-synergy/entsum_processed_runtime": 86.4163,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.916,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.735,
+      "step": 32000
+    },
+    {
+      "epoch": 6.134148671889929,
+      "grad_norm": 0.783420741558075,
+      "learning_rate": 3.4831029185867894e-05,
+      "loss": 0.0326,
+      "step": 32100
+    },
+    {
+      "epoch": 6.153258169310147,
+      "grad_norm": 0.7571488618850708,
+      "learning_rate": 3.478302611367127e-05,
+      "loss": 0.0266,
+      "step": 32200
+    },
+    {
+      "epoch": 6.172367666730365,
+      "grad_norm": 0.08053430169820786,
+      "learning_rate": 3.473502304147466e-05,
+      "loss": 0.0278,
+      "step": 32300
+    },
+    {
+      "epoch": 6.191477164150583,
+      "grad_norm": 1.4743053913116455,
+      "learning_rate": 3.468701996927804e-05,
+      "loss": 0.0256,
+      "step": 32400
+    },
+    {
+      "epoch": 6.210586661570801,
+      "grad_norm": 0.20271699130535126,
+      "learning_rate": 3.4639016897081416e-05,
+      "loss": 0.0298,
+      "step": 32500
+    },
+    {
+      "epoch": 6.229696158991018,
+      "grad_norm": 1.0754730701446533,
+      "learning_rate": 3.4591013824884794e-05,
+      "loss": 0.0253,
+      "step": 32600
+    },
+    {
+      "epoch": 6.248805656411236,
+      "grad_norm": 1.5914433002471924,
+      "learning_rate": 3.454301075268817e-05,
+      "loss": 0.0354,
+      "step": 32700
+    },
+    {
+      "epoch": 6.267915153831455,
+      "grad_norm": 0.37541651725769043,
+      "learning_rate": 3.449500768049155e-05,
+      "loss": 0.0274,
+      "step": 32800
+    },
+    {
+      "epoch": 6.287024651251672,
+      "grad_norm": 0.6610062718391418,
+      "learning_rate": 3.444700460829493e-05,
+      "loss": 0.0308,
+      "step": 32900
+    },
+    {
+      "epoch": 6.30613414867189,
+      "grad_norm": 0.7302411198616028,
+      "learning_rate": 3.4399001536098316e-05,
+      "loss": 0.0281,
+      "step": 33000
+    },
+    {
+      "epoch": 6.30613414867189,
+      "eval_events-synergy/entsum_processed_loss": 0.023598061874508858,
+      "eval_events-synergy/entsum_processed_runtime": 86.4223,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.914,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.734,
+      "step": 33000
+    },
+    {
+      "epoch": 6.325243646092108,
+      "grad_norm": 1.2536879777908325,
+      "learning_rate": 3.435099846390169e-05,
+      "loss": 0.0292,
+      "step": 33100
+    },
+    {
+      "epoch": 6.3443531435123255,
+      "grad_norm": 0.050880610942840576,
+      "learning_rate": 3.4302995391705074e-05,
+      "loss": 0.023,
+      "step": 33200
+    },
+    {
+      "epoch": 6.363462640932544,
+      "grad_norm": 0.07612816989421844,
+      "learning_rate": 3.425499231950845e-05,
+      "loss": 0.027,
+      "step": 33300
+    },
+    {
+      "epoch": 6.382572138352761,
+      "grad_norm": 0.2280048131942749,
+      "learning_rate": 3.4206989247311824e-05,
+      "loss": 0.031,
+      "step": 33400
+    },
+    {
+      "epoch": 6.401681635772979,
+      "grad_norm": 0.9767467975616455,
+      "learning_rate": 3.415898617511521e-05,
+      "loss": 0.029,
+      "step": 33500
+    },
+    {
+      "epoch": 6.420791133193197,
+      "grad_norm": 1.0423424243927002,
+      "learning_rate": 3.411098310291859e-05,
+      "loss": 0.0244,
+      "step": 33600
+    },
+    {
+      "epoch": 6.439900630613415,
+      "grad_norm": 1.2598810195922852,
+      "learning_rate": 3.406298003072197e-05,
+      "loss": 0.0293,
+      "step": 33700
+    },
+    {
+      "epoch": 6.459010128033633,
+      "grad_norm": 0.1793285310268402,
+      "learning_rate": 3.4014976958525346e-05,
+      "loss": 0.0294,
+      "step": 33800
+    },
+    {
+      "epoch": 6.478119625453851,
+      "grad_norm": 0.011976735666394234,
+      "learning_rate": 3.396697388632873e-05,
+      "loss": 0.0276,
+      "step": 33900
+    },
+    {
+      "epoch": 6.497229122874068,
+      "grad_norm": 1.0099623203277588,
+      "learning_rate": 3.3918970814132103e-05,
+      "loss": 0.0351,
+      "step": 34000
+    },
+    {
+      "epoch": 6.497229122874068,
+      "eval_events-synergy/entsum_processed_loss": 0.02258702553808689,
+      "eval_events-synergy/entsum_processed_runtime": 86.2714,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.961,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.746,
+      "step": 34000
+    },
+    {
+      "epoch": 6.516338620294286,
+      "grad_norm": 0.6294183135032654,
+      "learning_rate": 3.387096774193548e-05,
+      "loss": 0.0236,
+      "step": 34100
+    },
+    {
+      "epoch": 6.535448117714504,
+      "grad_norm": 0.5773456692695618,
+      "learning_rate": 3.382296466973887e-05,
+      "loss": 0.024,
+      "step": 34200
+    },
+    {
+      "epoch": 6.554557615134722,
+      "grad_norm": 0.7001476883888245,
+      "learning_rate": 3.377496159754224e-05,
+      "loss": 0.03,
+      "step": 34300
+    },
+    {
+      "epoch": 6.57366711255494,
+      "grad_norm": 1.5043959617614746,
+      "learning_rate": 3.3726958525345625e-05,
+      "loss": 0.023,
+      "step": 34400
+    },
+    {
+      "epoch": 6.592776609975157,
+      "grad_norm": 0.37344950437545776,
+      "learning_rate": 3.3678955453149004e-05,
+      "loss": 0.0383,
+      "step": 34500
+    },
+    {
+      "epoch": 6.611886107395375,
+      "grad_norm": 0.7408236265182495,
+      "learning_rate": 3.363095238095238e-05,
+      "loss": 0.0253,
+      "step": 34600
+    },
+    {
+      "epoch": 6.630995604815594,
+      "grad_norm": 0.08738729357719421,
+      "learning_rate": 3.358294930875576e-05,
+      "loss": 0.023,
+      "step": 34700
+    },
+    {
+      "epoch": 6.650105102235811,
+      "grad_norm": 0.45454615354537964,
+      "learning_rate": 3.353494623655914e-05,
+      "loss": 0.025,
+      "step": 34800
+    },
+    {
+      "epoch": 6.669214599656029,
+      "grad_norm": 0.4266088008880615,
+      "learning_rate": 3.348694316436252e-05,
+      "loss": 0.0251,
+      "step": 34900
+    },
+    {
+      "epoch": 6.688324097076247,
+      "grad_norm": 0.44192788004875183,
+      "learning_rate": 3.34389400921659e-05,
+      "loss": 0.0349,
+      "step": 35000
+    },
+    {
+      "epoch": 6.688324097076247,
+      "eval_events-synergy/entsum_processed_loss": 0.022197816520929337,
+      "eval_events-synergy/entsum_processed_runtime": 86.47,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.899,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.731,
+      "step": 35000
+    },
+    {
+      "epoch": 6.7074335944964645,
+      "grad_norm": 1.1022039651870728,
+      "learning_rate": 3.339093701996928e-05,
+      "loss": 0.0291,
+      "step": 35100
+    },
+    {
+      "epoch": 6.726543091916683,
+      "grad_norm": 0.7178447842597961,
+      "learning_rate": 3.334293394777266e-05,
+      "loss": 0.0252,
+      "step": 35200
+    },
+    {
+      "epoch": 6.745652589336901,
+      "grad_norm": 0.20827040076255798,
+      "learning_rate": 3.3294930875576034e-05,
+      "loss": 0.026,
+      "step": 35300
+    },
+    {
+      "epoch": 6.764762086757118,
+      "grad_norm": 0.044019654393196106,
+      "learning_rate": 3.324692780337942e-05,
+      "loss": 0.018,
+      "step": 35400
+    },
+    {
+      "epoch": 6.783871584177336,
+      "grad_norm": 0.4710674285888672,
+      "learning_rate": 3.31989247311828e-05,
+      "loss": 0.0275,
+      "step": 35500
+    },
+    {
+      "epoch": 6.802981081597554,
+      "grad_norm": 1.3574426174163818,
+      "learning_rate": 3.315092165898618e-05,
+      "loss": 0.0245,
+      "step": 35600
+    },
+    {
+      "epoch": 6.822090579017772,
+      "grad_norm": 0.2804085612297058,
+      "learning_rate": 3.3102918586789556e-05,
+      "loss": 0.0207,
+      "step": 35700
+    },
+    {
+      "epoch": 6.84120007643799,
+      "grad_norm": 0.16950590908527374,
+      "learning_rate": 3.305491551459294e-05,
+      "loss": 0.028,
+      "step": 35800
+    },
+    {
+      "epoch": 6.860309573858207,
+      "grad_norm": 0.38544100522994995,
+      "learning_rate": 3.300691244239631e-05,
+      "loss": 0.0369,
+      "step": 35900
+    },
+    {
+      "epoch": 6.879419071278425,
+      "grad_norm": 0.36500322818756104,
+      "learning_rate": 3.295890937019969e-05,
+      "loss": 0.0255,
+      "step": 36000
+    },
+    {
+      "epoch": 6.879419071278425,
+      "eval_events-synergy/entsum_processed_loss": 0.021442364901304245,
+      "eval_events-synergy/entsum_processed_runtime": 85.1296,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.323,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.837,
+      "step": 36000
+    },
+    {
+      "epoch": 6.8985285686986435,
+      "grad_norm": 0.8930972218513489,
+      "learning_rate": 3.291090629800308e-05,
+      "loss": 0.023,
+      "step": 36100
+    },
+    {
+      "epoch": 6.917638066118861,
+      "grad_norm": 0.02050691284239292,
+      "learning_rate": 3.286290322580645e-05,
+      "loss": 0.0213,
+      "step": 36200
+    },
+    {
+      "epoch": 6.936747563539079,
+      "grad_norm": 0.268373042345047,
+      "learning_rate": 3.2814900153609835e-05,
+      "loss": 0.0211,
+      "step": 36300
+    },
+    {
+      "epoch": 6.955857060959297,
+      "grad_norm": 0.4497075080871582,
+      "learning_rate": 3.2766897081413214e-05,
+      "loss": 0.0275,
+      "step": 36400
+    },
+    {
+      "epoch": 6.974966558379514,
+      "grad_norm": 0.024803461506962776,
+      "learning_rate": 3.271889400921659e-05,
+      "loss": 0.0259,
+      "step": 36500
+    },
+    {
+      "epoch": 6.994076055799733,
+      "grad_norm": 0.030653996393084526,
+      "learning_rate": 3.267089093701997e-05,
+      "loss": 0.0225,
+      "step": 36600
+    },
+    {
+      "epoch": 7.013185553219951,
+      "grad_norm": 0.3798658549785614,
+      "learning_rate": 3.262288786482335e-05,
+      "loss": 0.0246,
+      "step": 36700
+    },
+    {
+      "epoch": 7.032295050640168,
+      "grad_norm": 0.9082667231559753,
+      "learning_rate": 3.257488479262673e-05,
+      "loss": 0.0217,
+      "step": 36800
+    },
+    {
+      "epoch": 7.051404548060386,
+      "grad_norm": 0.7902773022651672,
+      "learning_rate": 3.252688172043011e-05,
+      "loss": 0.02,
+      "step": 36900
+    },
+    {
+      "epoch": 7.0705140454806035,
+      "grad_norm": 0.6506838202476501,
+      "learning_rate": 3.247887864823349e-05,
+      "loss": 0.0185,
+      "step": 37000
+    },
+    {
+      "epoch": 7.0705140454806035,
+      "eval_events-synergy/entsum_processed_loss": 0.021531905978918076,
+      "eval_events-synergy/entsum_processed_runtime": 85.174,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.309,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.833,
+      "step": 37000
+    },
+    {
+      "epoch": 7.089623542900822,
+      "grad_norm": 0.624746561050415,
+      "learning_rate": 3.2430875576036865e-05,
+      "loss": 0.0192,
+      "step": 37100
+    },
+    {
+      "epoch": 7.10873304032104,
+      "grad_norm": 0.5149529576301575,
+      "learning_rate": 3.2382872503840243e-05,
+      "loss": 0.0276,
+      "step": 37200
+    },
+    {
+      "epoch": 7.127842537741257,
+      "grad_norm": 3.0162253379821777,
+      "learning_rate": 3.233486943164363e-05,
+      "loss": 0.0209,
+      "step": 37300
+    },
+    {
+      "epoch": 7.146952035161475,
+      "grad_norm": 0.3581557273864746,
+      "learning_rate": 3.228686635944701e-05,
+      "loss": 0.017,
+      "step": 37400
+    },
+    {
+      "epoch": 7.166061532581693,
+      "grad_norm": 0.534199595451355,
+      "learning_rate": 3.2238863287250386e-05,
+      "loss": 0.0225,
+      "step": 37500
+    },
+    {
+      "epoch": 7.185171030001911,
+      "grad_norm": 0.584403395652771,
+      "learning_rate": 3.2190860215053765e-05,
+      "loss": 0.0246,
+      "step": 37600
+    },
+    {
+      "epoch": 7.204280527422129,
+      "grad_norm": 0.33575019240379333,
+      "learning_rate": 3.2142857142857144e-05,
+      "loss": 0.0175,
+      "step": 37700
+    },
+    {
+      "epoch": 7.223390024842347,
+      "grad_norm": 0.10965080559253693,
+      "learning_rate": 3.209485407066052e-05,
+      "loss": 0.0226,
+      "step": 37800
+    },
+    {
+      "epoch": 7.242499522262564,
+      "grad_norm": 0.35581934452056885,
+      "learning_rate": 3.20468509984639e-05,
+      "loss": 0.0238,
+      "step": 37900
+    },
+    {
+      "epoch": 7.2616090196827825,
+      "grad_norm": 0.34397947788238525,
+      "learning_rate": 3.199884792626729e-05,
+      "loss": 0.0143,
+      "step": 38000
+    },
+    {
+      "epoch": 7.2616090196827825,
+      "eval_events-synergy/entsum_processed_loss": 0.020761175081133842,
+      "eval_events-synergy/entsum_processed_runtime": 85.5012,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.204,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.807,
+      "step": 38000
+    },
+    {
+      "epoch": 7.280718517103,
+      "grad_norm": 0.42107951641082764,
+      "learning_rate": 3.195084485407066e-05,
+      "loss": 0.0247,
+      "step": 38100
+    },
+    {
+      "epoch": 7.299828014523218,
+      "grad_norm": 0.02115056663751602,
+      "learning_rate": 3.1902841781874044e-05,
+      "loss": 0.0195,
+      "step": 38200
+    },
+    {
+      "epoch": 7.318937511943436,
+      "grad_norm": 0.5838721990585327,
+      "learning_rate": 3.185483870967742e-05,
+      "loss": 0.0219,
+      "step": 38300
+    },
+    {
+      "epoch": 7.338047009363653,
+      "grad_norm": 0.1621849685907364,
+      "learning_rate": 3.18068356374808e-05,
+      "loss": 0.0205,
+      "step": 38400
+    },
+    {
+      "epoch": 7.357156506783872,
+      "grad_norm": 0.9421815872192383,
+      "learning_rate": 3.175883256528418e-05,
+      "loss": 0.0255,
+      "step": 38500
+    },
+    {
+      "epoch": 7.37626600420409,
+      "grad_norm": 0.534527063369751,
+      "learning_rate": 3.171082949308756e-05,
+      "loss": 0.0193,
+      "step": 38600
+    },
+    {
+      "epoch": 7.395375501624307,
+      "grad_norm": 0.25326958298683167,
+      "learning_rate": 3.166282642089094e-05,
+      "loss": 0.0173,
+      "step": 38700
+    },
+    {
+      "epoch": 7.414484999044525,
+      "grad_norm": 0.45906540751457214,
+      "learning_rate": 3.161482334869432e-05,
+      "loss": 0.0217,
+      "step": 38800
+    },
+    {
+      "epoch": 7.433594496464743,
+      "grad_norm": 0.20368924736976624,
+      "learning_rate": 3.15668202764977e-05,
+      "loss": 0.0204,
+      "step": 38900
+    },
+    {
+      "epoch": 7.452703993884961,
+      "grad_norm": 0.16196204721927643,
+      "learning_rate": 3.1518817204301074e-05,
+      "loss": 0.0269,
+      "step": 39000
+    },
+    {
+      "epoch": 7.452703993884961,
+      "eval_events-synergy/entsum_processed_loss": 0.01936463639140129,
+      "eval_events-synergy/entsum_processed_runtime": 85.1337,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.322,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.836,
+      "step": 39000
+    },
+    {
+      "epoch": 7.471813491305179,
+      "grad_norm": 0.9692901372909546,
+      "learning_rate": 3.147081413210445e-05,
+      "loss": 0.0233,
+      "step": 39100
+    },
+    {
+      "epoch": 7.490922988725396,
+      "grad_norm": 0.772438645362854,
+      "learning_rate": 3.142281105990784e-05,
+      "loss": 0.0207,
+      "step": 39200
+    },
+    {
+      "epoch": 7.510032486145614,
+      "grad_norm": 0.39141640067100525,
+      "learning_rate": 3.137480798771121e-05,
+      "loss": 0.0191,
+      "step": 39300
+    },
+    {
+      "epoch": 7.529141983565832,
+      "grad_norm": 0.19886603951454163,
+      "learning_rate": 3.1326804915514596e-05,
+      "loss": 0.0189,
+      "step": 39400
+    },
+    {
+      "epoch": 7.54825148098605,
+      "grad_norm": 0.8104422092437744,
+      "learning_rate": 3.1278801843317975e-05,
+      "loss": 0.0247,
+      "step": 39500
+    },
+    {
+      "epoch": 7.567360978406268,
+      "grad_norm": 1.5498480796813965,
+      "learning_rate": 3.1230798771121354e-05,
+      "loss": 0.0227,
+      "step": 39600
+    },
+    {
+      "epoch": 7.586470475826486,
+      "grad_norm": 0.7283300757408142,
+      "learning_rate": 3.118279569892473e-05,
+      "loss": 0.0226,
+      "step": 39700
+    },
+    {
+      "epoch": 7.605579973246703,
+      "grad_norm": 1.3604527711868286,
+      "learning_rate": 3.113479262672811e-05,
+      "loss": 0.0167,
+      "step": 39800
+    },
+    {
+      "epoch": 7.6246894706669215,
+      "grad_norm": 1.0663504600524902,
+      "learning_rate": 3.108678955453149e-05,
+      "loss": 0.0251,
+      "step": 39900
+    },
+    {
+      "epoch": 7.64379896808714,
+      "grad_norm": 0.7986360192298889,
+      "learning_rate": 3.103878648233487e-05,
+      "loss": 0.0178,
+      "step": 40000
+    },
+    {
+      "epoch": 7.64379896808714,
+      "eval_events-synergy/entsum_processed_loss": 0.01874077506363392,
+      "eval_events-synergy/entsum_processed_runtime": 86.5004,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.89,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.728,
+      "step": 40000
+    },
+    {
+      "epoch": 7.662908465507357,
+      "grad_norm": 0.20992030203342438,
+      "learning_rate": 3.0990783410138254e-05,
+      "loss": 0.0169,
+      "step": 40100
+    },
+    {
+      "epoch": 7.682017962927575,
+      "grad_norm": 0.19716721773147583,
+      "learning_rate": 3.0942780337941626e-05,
+      "loss": 0.0176,
+      "step": 40200
+    },
+    {
+      "epoch": 7.701127460347793,
+      "grad_norm": 0.015498577617108822,
+      "learning_rate": 3.089477726574501e-05,
+      "loss": 0.0211,
+      "step": 40300
+    },
+    {
+      "epoch": 7.7202369577680106,
+      "grad_norm": 0.6486418843269348,
+      "learning_rate": 3.084677419354839e-05,
+      "loss": 0.0155,
+      "step": 40400
+    },
+    {
+      "epoch": 7.739346455188229,
+      "grad_norm": 0.6476672887802124,
+      "learning_rate": 3.079877112135177e-05,
+      "loss": 0.0201,
+      "step": 40500
+    },
+    {
+      "epoch": 7.758455952608447,
+      "grad_norm": 1.1004291772842407,
+      "learning_rate": 3.075076804915515e-05,
+      "loss": 0.0242,
+      "step": 40600
+    },
+    {
+      "epoch": 7.777565450028664,
+      "grad_norm": 0.46351030468940735,
+      "learning_rate": 3.0702764976958526e-05,
+      "loss": 0.0185,
+      "step": 40700
+    },
+    {
+      "epoch": 7.796674947448882,
+      "grad_norm": 1.0854315757751465,
+      "learning_rate": 3.0654761904761905e-05,
+      "loss": 0.0201,
+      "step": 40800
+    },
+    {
+      "epoch": 7.8157844448691,
+      "grad_norm": 0.8841759562492371,
+      "learning_rate": 3.0606758832565284e-05,
+      "loss": 0.0184,
+      "step": 40900
+    },
+    {
+      "epoch": 7.834893942289318,
+      "grad_norm": 0.9825415015220642,
+      "learning_rate": 3.055875576036866e-05,
+      "loss": 0.0187,
+      "step": 41000
+    },
+    {
+      "epoch": 7.834893942289318,
+      "eval_events-synergy/entsum_processed_loss": 0.018535524606704712,
+      "eval_events-synergy/entsum_processed_runtime": 85.5215,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.198,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.805,
+      "step": 41000
+    },
+    {
+      "epoch": 7.854003439709536,
+      "grad_norm": 2.4730401039123535,
+      "learning_rate": 3.0510752688172045e-05,
+      "loss": 0.0191,
+      "step": 41100
+    },
+    {
+      "epoch": 7.873112937129753,
+      "grad_norm": 2.2196543216705322,
+      "learning_rate": 3.0462749615975423e-05,
+      "loss": 0.0283,
+      "step": 41200
+    },
+    {
+      "epoch": 7.892222434549971,
+      "grad_norm": 0.16002187132835388,
+      "learning_rate": 3.0414746543778806e-05,
+      "loss": 0.0199,
+      "step": 41300
+    },
+    {
+      "epoch": 7.911331931970189,
+      "grad_norm": 1.5348676443099976,
+      "learning_rate": 3.036674347158218e-05,
+      "loss": 0.0161,
+      "step": 41400
+    },
+    {
+      "epoch": 7.930441429390407,
+      "grad_norm": 0.6840288043022156,
+      "learning_rate": 3.0318740399385563e-05,
+      "loss": 0.014,
+      "step": 41500
+    },
+    {
+      "epoch": 7.949550926810625,
+      "grad_norm": 0.7691094875335693,
+      "learning_rate": 3.0270737327188942e-05,
+      "loss": 0.0171,
+      "step": 41600
+    },
+    {
+      "epoch": 7.968660424230842,
+      "grad_norm": 0.9019795060157776,
+      "learning_rate": 3.0222734254992317e-05,
+      "loss": 0.0208,
+      "step": 41700
+    },
+    {
+      "epoch": 7.9877699216510605,
+      "grad_norm": 0.40712770819664,
+      "learning_rate": 3.0174731182795703e-05,
+      "loss": 0.0216,
+      "step": 41800
+    },
+    {
+      "epoch": 8.006879419071279,
+      "grad_norm": 0.9297962784767151,
+      "learning_rate": 3.0126728110599078e-05,
+      "loss": 0.0206,
+      "step": 41900
+    },
+    {
+      "epoch": 8.025988916491496,
+      "grad_norm": 0.9476808309555054,
+      "learning_rate": 3.007872503840246e-05,
+      "loss": 0.0179,
+      "step": 42000
+    },
+    {
+      "epoch": 8.025988916491496,
+      "eval_events-synergy/entsum_processed_loss": 0.017886348068714142,
+      "eval_events-synergy/entsum_processed_runtime": 86.4493,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.906,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.732,
+      "step": 42000
+    },
+    {
+      "epoch": 8.045098413911715,
+      "grad_norm": 0.2303045094013214,
+      "learning_rate": 3.003072196620584e-05,
+      "loss": 0.0166,
+      "step": 42100
+    },
+    {
+      "epoch": 8.064207911331932,
+      "grad_norm": 0.2901061773300171,
+      "learning_rate": 2.998271889400922e-05,
+      "loss": 0.0192,
+      "step": 42200
+    },
+    {
+      "epoch": 8.08331740875215,
+      "grad_norm": 0.051759280264377594,
+      "learning_rate": 2.9934715821812596e-05,
+      "loss": 0.0161,
+      "step": 42300
+    },
+    {
+      "epoch": 8.102426906172367,
+      "grad_norm": 0.4911729097366333,
+      "learning_rate": 2.9886712749615975e-05,
+      "loss": 0.018,
+      "step": 42400
+    },
+    {
+      "epoch": 8.121536403592586,
+      "grad_norm": 0.4533239006996155,
+      "learning_rate": 2.9838709677419357e-05,
+      "loss": 0.0177,
+      "step": 42500
+    },
+    {
+      "epoch": 8.140645901012803,
+      "grad_norm": 0.03935829922556877,
+      "learning_rate": 2.9790706605222736e-05,
+      "loss": 0.016,
+      "step": 42600
+    },
+    {
+      "epoch": 8.15975539843302,
+      "grad_norm": 0.5652256011962891,
+      "learning_rate": 2.9742703533026118e-05,
+      "loss": 0.0125,
+      "step": 42700
+    },
+    {
+      "epoch": 8.17886489585324,
+      "grad_norm": 0.5128132104873657,
+      "learning_rate": 2.9694700460829493e-05,
+      "loss": 0.0123,
+      "step": 42800
+    },
+    {
+      "epoch": 8.197974393273457,
+      "grad_norm": 0.27111709117889404,
+      "learning_rate": 2.9646697388632872e-05,
+      "loss": 0.0173,
+      "step": 42900
+    },
+    {
+      "epoch": 8.217083890693674,
+      "grad_norm": 0.1685011088848114,
+      "learning_rate": 2.9598694316436254e-05,
+      "loss": 0.0161,
+      "step": 43000
+    },
+    {
+      "epoch": 8.217083890693674,
+      "eval_events-synergy/entsum_processed_loss": 0.017874937504529953,
+      "eval_events-synergy/entsum_processed_runtime": 86.3239,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.945,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.742,
+      "step": 43000
+    },
+    {
+      "epoch": 8.236193388113893,
+      "grad_norm": 0.3956407606601715,
+      "learning_rate": 2.955069124423963e-05,
+      "loss": 0.0143,
+      "step": 43100
+    },
+    {
+      "epoch": 8.25530288553411,
+      "grad_norm": 0.46672216057777405,
+      "learning_rate": 2.9502688172043015e-05,
+      "loss": 0.0136,
+      "step": 43200
+    },
+    {
+      "epoch": 8.274412382954328,
+      "grad_norm": 0.27463072538375854,
+      "learning_rate": 2.945468509984639e-05,
+      "loss": 0.0152,
+      "step": 43300
+    },
+    {
+      "epoch": 8.293521880374547,
+      "grad_norm": 0.11778465658426285,
+      "learning_rate": 2.9406682027649773e-05,
+      "loss": 0.0227,
+      "step": 43400
+    },
+    {
+      "epoch": 8.312631377794764,
+      "grad_norm": 0.966484785079956,
+      "learning_rate": 2.935867895545315e-05,
+      "loss": 0.0146,
+      "step": 43500
+    },
+    {
+      "epoch": 8.331740875214981,
+      "grad_norm": 0.10549327731132507,
+      "learning_rate": 2.9310675883256527e-05,
+      "loss": 0.0164,
+      "step": 43600
+    },
+    {
+      "epoch": 8.3508503726352,
+      "grad_norm": 0.19580169022083282,
+      "learning_rate": 2.926267281105991e-05,
+      "loss": 0.0192,
+      "step": 43700
+    },
+    {
+      "epoch": 8.369959870055418,
+      "grad_norm": 0.2825284004211426,
+      "learning_rate": 2.9214669738863288e-05,
+      "loss": 0.0183,
+      "step": 43800
+    },
+    {
+      "epoch": 8.389069367475635,
+      "grad_norm": 0.4127114713191986,
+      "learning_rate": 2.916666666666667e-05,
+      "loss": 0.0167,
+      "step": 43900
+    },
+    {
+      "epoch": 8.408178864895854,
+      "grad_norm": 0.2957821488380432,
+      "learning_rate": 2.911866359447005e-05,
+      "loss": 0.0152,
+      "step": 44000
+    },
+    {
+      "epoch": 8.408178864895854,
+      "eval_events-synergy/entsum_processed_loss": 0.018097488209605217,
+      "eval_events-synergy/entsum_processed_runtime": 86.2713,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.961,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.746,
+      "step": 44000
+    },
+    {
+      "epoch": 8.427288362316071,
+      "grad_norm": 1.0461006164550781,
+      "learning_rate": 2.907066052227343e-05,
+      "loss": 0.0168,
+      "step": 44100
+    },
+    {
+      "epoch": 8.446397859736289,
+      "grad_norm": 0.4242873191833496,
+      "learning_rate": 2.9022657450076806e-05,
+      "loss": 0.0184,
+      "step": 44200
+    },
+    {
+      "epoch": 8.465507357156508,
+      "grad_norm": 0.4693470001220703,
+      "learning_rate": 2.8974654377880185e-05,
+      "loss": 0.0196,
+      "step": 44300
+    },
+    {
+      "epoch": 8.484616854576725,
+      "grad_norm": 0.018570706248283386,
+      "learning_rate": 2.8926651305683567e-05,
+      "loss": 0.0157,
+      "step": 44400
+    },
+    {
+      "epoch": 8.503726351996942,
+      "grad_norm": 0.5416761040687561,
+      "learning_rate": 2.8878648233486942e-05,
+      "loss": 0.0178,
+      "step": 44500
+    },
+    {
+      "epoch": 8.52283584941716,
+      "grad_norm": 0.5635196566581726,
+      "learning_rate": 2.8830645161290328e-05,
+      "loss": 0.0122,
+      "step": 44600
+    },
+    {
+      "epoch": 8.541945346837378,
+      "grad_norm": 0.42741984128952026,
+      "learning_rate": 2.8782642089093703e-05,
+      "loss": 0.0145,
+      "step": 44700
+    },
+    {
+      "epoch": 8.561054844257596,
+      "grad_norm": 1.090361475944519,
+      "learning_rate": 2.8734639016897082e-05,
+      "loss": 0.0107,
+      "step": 44800
+    },
+    {
+      "epoch": 8.580164341677813,
+      "grad_norm": 0.792757511138916,
+      "learning_rate": 2.8686635944700464e-05,
+      "loss": 0.017,
+      "step": 44900
+    },
+    {
+      "epoch": 8.599273839098032,
+      "grad_norm": 0.07932941615581512,
+      "learning_rate": 2.863863287250384e-05,
+      "loss": 0.0141,
+      "step": 45000
+    },
+    {
+      "epoch": 8.599273839098032,
+      "eval_events-synergy/entsum_processed_loss": 0.017112286761403084,
+      "eval_events-synergy/entsum_processed_runtime": 86.6905,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.831,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.714,
+      "step": 45000
+    },
+    {
+      "epoch": 8.61838333651825,
+      "grad_norm": 0.42730385065078735,
+      "learning_rate": 2.859062980030722e-05,
+      "loss": 0.0168,
+      "step": 45100
+    },
+    {
+      "epoch": 8.637492833938467,
+      "grad_norm": 0.8624597787857056,
+      "learning_rate": 2.85426267281106e-05,
+      "loss": 0.021,
+      "step": 45200
+    },
+    {
+      "epoch": 8.656602331358686,
+      "grad_norm": 0.34436488151550293,
+      "learning_rate": 2.8494623655913982e-05,
+      "loss": 0.0152,
+      "step": 45300
+    },
+    {
+      "epoch": 8.675711828778903,
+      "grad_norm": 1.8274009227752686,
+      "learning_rate": 2.844662058371736e-05,
+      "loss": 0.0114,
+      "step": 45400
+    },
+    {
+      "epoch": 8.69482132619912,
+      "grad_norm": 0.1283120959997177,
+      "learning_rate": 2.8398617511520736e-05,
+      "loss": 0.0203,
+      "step": 45500
+    },
+    {
+      "epoch": 8.71393082361934,
+      "grad_norm": 1.9227547645568848,
+      "learning_rate": 2.835061443932412e-05,
+      "loss": 0.0161,
+      "step": 45600
+    },
+    {
+      "epoch": 8.733040321039557,
+      "grad_norm": 1.129739761352539,
+      "learning_rate": 2.8302611367127497e-05,
+      "loss": 0.0149,
+      "step": 45700
+    },
+    {
+      "epoch": 8.752149818459774,
+      "grad_norm": 0.807822048664093,
+      "learning_rate": 2.825460829493088e-05,
+      "loss": 0.0123,
+      "step": 45800
+    },
+    {
+      "epoch": 8.771259315879993,
+      "grad_norm": 0.24655993282794952,
+      "learning_rate": 2.8206605222734255e-05,
+      "loss": 0.0165,
+      "step": 45900
+    },
+    {
+      "epoch": 8.79036881330021,
+      "grad_norm": 0.19655448198318481,
+      "learning_rate": 2.8158602150537637e-05,
+      "loss": 0.0174,
+      "step": 46000
+    },
+    {
+      "epoch": 8.79036881330021,
+      "eval_events-synergy/entsum_processed_loss": 0.01653146930038929,
+      "eval_events-synergy/entsum_processed_runtime": 86.3181,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.947,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.743,
+      "step": 46000
+    },
+    {
+      "epoch": 8.809478310720428,
+      "grad_norm": 0.37631919980049133,
+      "learning_rate": 2.8110599078341016e-05,
+      "loss": 0.0186,
+      "step": 46100
+    },
+    {
+      "epoch": 8.828587808140647,
+      "grad_norm": 0.25859326124191284,
+      "learning_rate": 2.8062596006144394e-05,
+      "loss": 0.0194,
+      "step": 46200
+    },
+    {
+      "epoch": 8.847697305560864,
+      "grad_norm": 0.9864445924758911,
+      "learning_rate": 2.8014592933947776e-05,
+      "loss": 0.0163,
+      "step": 46300
+    },
+    {
+      "epoch": 8.866806802981081,
+      "grad_norm": 1.691253900527954,
+      "learning_rate": 2.7966589861751152e-05,
+      "loss": 0.017,
+      "step": 46400
+    },
+    {
+      "epoch": 8.8859163004013,
+      "grad_norm": 0.9325816035270691,
+      "learning_rate": 2.7918586789554534e-05,
+      "loss": 0.0126,
+      "step": 46500
+    },
+    {
+      "epoch": 8.905025797821517,
+      "grad_norm": 0.8106985688209534,
+      "learning_rate": 2.7870583717357913e-05,
+      "loss": 0.0155,
+      "step": 46600
+    },
+    {
+      "epoch": 8.924135295241735,
+      "grad_norm": 0.5398417711257935,
+      "learning_rate": 2.7822580645161288e-05,
+      "loss": 0.0141,
+      "step": 46700
+    },
+    {
+      "epoch": 8.943244792661954,
+      "grad_norm": 0.5932244658470154,
+      "learning_rate": 2.777457757296467e-05,
+      "loss": 0.016,
+      "step": 46800
+    },
+    {
+      "epoch": 8.962354290082171,
+      "grad_norm": 0.2489900141954422,
+      "learning_rate": 2.772657450076805e-05,
+      "loss": 0.0157,
+      "step": 46900
+    },
+    {
+      "epoch": 8.981463787502388,
+      "grad_norm": 0.5513284802436829,
+      "learning_rate": 2.767857142857143e-05,
+      "loss": 0.0133,
+      "step": 47000
+    },
+    {
+      "epoch": 8.981463787502388,
+      "eval_events-synergy/entsum_processed_loss": 0.016639988869428635,
+      "eval_events-synergy/entsum_processed_runtime": 86.6238,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.852,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.719,
+      "step": 47000
+    },
+    {
+      "epoch": 9.000573284922606,
+      "grad_norm": 0.7625744342803955,
+      "learning_rate": 2.763056835637481e-05,
+      "loss": 0.013,
+      "step": 47100
+    },
+    {
+      "epoch": 9.019682782342825,
+      "grad_norm": 0.5866988897323608,
+      "learning_rate": 2.7582565284178192e-05,
+      "loss": 0.0132,
+      "step": 47200
+    },
+    {
+      "epoch": 9.038792279763042,
+      "grad_norm": 0.21316476166248322,
+      "learning_rate": 2.7534562211981567e-05,
+      "loss": 0.0126,
+      "step": 47300
+    },
+    {
+      "epoch": 9.05790177718326,
+      "grad_norm": 0.2244674414396286,
+      "learning_rate": 2.7486559139784946e-05,
+      "loss": 0.0125,
+      "step": 47400
+    },
+    {
+      "epoch": 9.077011274603478,
+      "grad_norm": 0.48919981718063354,
+      "learning_rate": 2.7438556067588328e-05,
+      "loss": 0.0145,
+      "step": 47500
+    },
+    {
+      "epoch": 9.096120772023696,
+      "grad_norm": 1.7857321500778198,
+      "learning_rate": 2.7390552995391703e-05,
+      "loss": 0.0136,
+      "step": 47600
+    },
+    {
+      "epoch": 9.115230269443913,
+      "grad_norm": 1.4337810277938843,
+      "learning_rate": 2.734254992319509e-05,
+      "loss": 0.0124,
+      "step": 47700
+    },
+    {
+      "epoch": 9.134339766864132,
+      "grad_norm": 0.1657884567975998,
+      "learning_rate": 2.7294546850998464e-05,
+      "loss": 0.0131,
+      "step": 47800
+    },
+    {
+      "epoch": 9.15344926428435,
+      "grad_norm": 0.32333338260650635,
+      "learning_rate": 2.7246543778801846e-05,
+      "loss": 0.0113,
+      "step": 47900
+    },
+    {
+      "epoch": 9.172558761704567,
+      "grad_norm": 0.05588236078619957,
+      "learning_rate": 2.7198540706605225e-05,
+      "loss": 0.0114,
+      "step": 48000
+    },
+    {
+      "epoch": 9.172558761704567,
+      "eval_events-synergy/entsum_processed_loss": 0.01658570021390915,
+      "eval_events-synergy/entsum_processed_runtime": 86.4495,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.906,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.732,
+      "step": 48000
+    },
+    {
+      "epoch": 9.191668259124786,
+      "grad_norm": 0.12975239753723145,
+      "learning_rate": 2.71505376344086e-05,
+      "loss": 0.0157,
+      "step": 48100
+    },
+    {
+      "epoch": 9.210777756545003,
+      "grad_norm": 0.21814003586769104,
+      "learning_rate": 2.7102534562211983e-05,
+      "loss": 0.011,
+      "step": 48200
+    },
+    {
+      "epoch": 9.22988725396522,
+      "grad_norm": 0.04857701435685158,
+      "learning_rate": 2.705453149001536e-05,
+      "loss": 0.0113,
+      "step": 48300
+    },
+    {
+      "epoch": 9.24899675138544,
+      "grad_norm": 0.17843018472194672,
+      "learning_rate": 2.7006528417818743e-05,
+      "loss": 0.0114,
+      "step": 48400
+    },
+    {
+      "epoch": 9.268106248805656,
+      "grad_norm": 0.4526139497756958,
+      "learning_rate": 2.6958525345622122e-05,
+      "loss": 0.0164,
+      "step": 48500
+    },
+    {
+      "epoch": 9.287215746225874,
+      "grad_norm": 0.6078909635543823,
+      "learning_rate": 2.6910522273425498e-05,
+      "loss": 0.0105,
+      "step": 48600
+    },
+    {
+      "epoch": 9.306325243646093,
+      "grad_norm": 0.8294253349304199,
+      "learning_rate": 2.686251920122888e-05,
+      "loss": 0.0129,
+      "step": 48700
+    },
+    {
+      "epoch": 9.32543474106631,
+      "grad_norm": 0.6839629411697388,
+      "learning_rate": 2.681451612903226e-05,
+      "loss": 0.0156,
+      "step": 48800
+    },
+    {
+      "epoch": 9.344544238486527,
+      "grad_norm": 0.05435250699520111,
+      "learning_rate": 2.676651305683564e-05,
+      "loss": 0.0131,
+      "step": 48900
+    },
+    {
+      "epoch": 9.363653735906746,
+      "grad_norm": 1.4614830017089844,
+      "learning_rate": 2.6718509984639016e-05,
+      "loss": 0.0114,
+      "step": 49000
+    },
+    {
+      "epoch": 9.363653735906746,
+      "eval_events-synergy/entsum_processed_loss": 0.015850717201828957,
+      "eval_events-synergy/entsum_processed_runtime": 86.4368,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.91,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.733,
+      "step": 49000
+    },
+    {
+      "epoch": 9.382763233326964,
+      "grad_norm": 0.2169920802116394,
+      "learning_rate": 2.66705069124424e-05,
+      "loss": 0.0128,
+      "step": 49100
+    },
+    {
+      "epoch": 9.401872730747181,
+      "grad_norm": 0.2624705135822296,
+      "learning_rate": 2.6622503840245777e-05,
+      "loss": 0.0109,
+      "step": 49200
+    },
+    {
+      "epoch": 9.4209822281674,
+      "grad_norm": 0.3865724802017212,
+      "learning_rate": 2.6574500768049156e-05,
+      "loss": 0.0151,
+      "step": 49300
+    },
+    {
+      "epoch": 9.440091725587617,
+      "grad_norm": 0.6372859477996826,
+      "learning_rate": 2.6526497695852538e-05,
+      "loss": 0.0127,
+      "step": 49400
+    },
+    {
+      "epoch": 9.459201223007835,
+      "grad_norm": 0.4839339256286621,
+      "learning_rate": 2.6478494623655913e-05,
+      "loss": 0.0132,
+      "step": 49500
+    },
+    {
+      "epoch": 9.478310720428052,
+      "grad_norm": 0.5721094608306885,
+      "learning_rate": 2.6430491551459295e-05,
+      "loss": 0.0142,
+      "step": 49600
+    },
+    {
+      "epoch": 9.497420217848271,
+      "grad_norm": 0.03586877882480621,
+      "learning_rate": 2.6382488479262674e-05,
+      "loss": 0.0123,
+      "step": 49700
+    },
+    {
+      "epoch": 9.516529715268488,
+      "grad_norm": 0.38037604093551636,
+      "learning_rate": 2.633448540706605e-05,
+      "loss": 0.0235,
+      "step": 49800
+    },
+    {
+      "epoch": 9.535639212688706,
+      "grad_norm": 0.2859026789665222,
+      "learning_rate": 2.6286482334869435e-05,
+      "loss": 0.0124,
+      "step": 49900
+    },
+    {
+      "epoch": 9.554748710108925,
+      "grad_norm": 0.27045369148254395,
+      "learning_rate": 2.623847926267281e-05,
+      "loss": 0.017,
+      "step": 50000
+    },
+    {
+      "epoch": 9.554748710108925,
+      "eval_events-synergy/entsum_processed_loss": 0.01477838959544897,
+      "eval_events-synergy/entsum_processed_runtime": 85.8776,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.085,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.777,
+      "step": 50000
+    },
+    {
+      "epoch": 9.573858207529142,
+      "grad_norm": 0.1873704493045807,
+      "learning_rate": 2.6190476190476192e-05,
+      "loss": 0.0115,
+      "step": 50100
+    },
+    {
+      "epoch": 9.59296770494936,
+      "grad_norm": 0.1264422982931137,
+      "learning_rate": 2.614247311827957e-05,
+      "loss": 0.0099,
+      "step": 50200
+    },
+    {
+      "epoch": 9.612077202369578,
+      "grad_norm": 0.8306263089179993,
+      "learning_rate": 2.6094470046082953e-05,
+      "loss": 0.0136,
+      "step": 50300
+    },
+    {
+      "epoch": 9.631186699789795,
+      "grad_norm": 0.5657772421836853,
+      "learning_rate": 2.604646697388633e-05,
+      "loss": 0.0113,
+      "step": 50400
+    },
+    {
+      "epoch": 9.650296197210013,
+      "grad_norm": 0.7657700181007385,
+      "learning_rate": 2.5998463901689707e-05,
+      "loss": 0.0127,
+      "step": 50500
+    },
+    {
+      "epoch": 9.669405694630232,
+      "grad_norm": 0.4362422227859497,
+      "learning_rate": 2.595046082949309e-05,
+      "loss": 0.01,
+      "step": 50600
+    },
+    {
+      "epoch": 9.688515192050449,
+      "grad_norm": 1.3221834897994995,
+      "learning_rate": 2.5902457757296468e-05,
+      "loss": 0.013,
+      "step": 50700
+    },
+    {
+      "epoch": 9.707624689470666,
+      "grad_norm": 0.10563758760690689,
+      "learning_rate": 2.585445468509985e-05,
+      "loss": 0.0139,
+      "step": 50800
+    },
+    {
+      "epoch": 9.726734186890885,
+      "grad_norm": 0.40996453166007996,
+      "learning_rate": 2.5806451612903226e-05,
+      "loss": 0.0097,
+      "step": 50900
+    },
+    {
+      "epoch": 9.745843684311103,
+      "grad_norm": 0.14993074536323547,
+      "learning_rate": 2.5758448540706608e-05,
+      "loss": 0.0117,
+      "step": 51000
+    },
+    {
+      "epoch": 9.745843684311103,
+      "eval_events-synergy/entsum_processed_loss": 0.015352104790508747,
+      "eval_events-synergy/entsum_processed_runtime": 86.4339,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.911,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.733,
+      "step": 51000
+    },
+    {
+      "epoch": 9.76495318173132,
+      "grad_norm": 0.034810420125722885,
+      "learning_rate": 2.5710445468509986e-05,
+      "loss": 0.0117,
+      "step": 51100
+    },
+    {
+      "epoch": 9.784062679151539,
+      "grad_norm": 0.5618245005607605,
+      "learning_rate": 2.5662442396313362e-05,
+      "loss": 0.0113,
+      "step": 51200
+    },
+    {
+      "epoch": 9.803172176571756,
+      "grad_norm": 0.26905587315559387,
+      "learning_rate": 2.5614439324116747e-05,
+      "loss": 0.0135,
+      "step": 51300
+    },
+    {
+      "epoch": 9.822281673991974,
+      "grad_norm": 0.4496970772743225,
+      "learning_rate": 2.5566436251920123e-05,
+      "loss": 0.0136,
+      "step": 51400
+    },
+    {
+      "epoch": 9.841391171412193,
+      "grad_norm": 0.7193441987037659,
+      "learning_rate": 2.5518433179723505e-05,
+      "loss": 0.0109,
+      "step": 51500
+    },
+    {
+      "epoch": 9.86050066883241,
+      "grad_norm": 1.0633965730667114,
+      "learning_rate": 2.5470430107526883e-05,
+      "loss": 0.0169,
+      "step": 51600
+    },
+    {
+      "epoch": 9.879610166252627,
+      "grad_norm": 0.03440142795443535,
+      "learning_rate": 2.542242703533026e-05,
+      "loss": 0.012,
+      "step": 51700
+    },
+    {
+      "epoch": 9.898719663672846,
+      "grad_norm": 0.7566103339195251,
+      "learning_rate": 2.537442396313364e-05,
+      "loss": 0.0145,
+      "step": 51800
+    },
+    {
+      "epoch": 9.917829161093064,
+      "grad_norm": 0.39200007915496826,
+      "learning_rate": 2.532642089093702e-05,
+      "loss": 0.0129,
+      "step": 51900
+    },
+    {
+      "epoch": 9.93693865851328,
+      "grad_norm": 1.0183656215667725,
+      "learning_rate": 2.5278417818740402e-05,
+      "loss": 0.0129,
+      "step": 52000
+    },
+    {
+      "epoch": 9.93693865851328,
+      "eval_events-synergy/entsum_processed_loss": 0.014359832741320133,
+      "eval_events-synergy/entsum_processed_runtime": 86.5634,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.87,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.723,
+      "step": 52000
+    },
+    {
+      "epoch": 9.956048155933498,
+      "grad_norm": 0.4568009376525879,
+      "learning_rate": 2.523041474654378e-05,
+      "loss": 0.0092,
+      "step": 52100
+    },
+    {
+      "epoch": 9.975157653353717,
+      "grad_norm": 0.4945560097694397,
+      "learning_rate": 2.5182411674347163e-05,
+      "loss": 0.0107,
+      "step": 52200
+    },
+    {
+      "epoch": 9.994267150773934,
+      "grad_norm": 0.4079298675060272,
+      "learning_rate": 2.5134408602150538e-05,
+      "loss": 0.0145,
+      "step": 52300
+    },
+    {
+      "epoch": 10.013376648194152,
+      "grad_norm": 0.7107488512992859,
+      "learning_rate": 2.5086405529953917e-05,
+      "loss": 0.0091,
+      "step": 52400
+    },
+    {
+      "epoch": 10.03248614561437,
+      "grad_norm": 0.5608364939689636,
+      "learning_rate": 2.50384024577573e-05,
+      "loss": 0.0105,
+      "step": 52500
+    },
+    {
+      "epoch": 10.051595643034588,
+      "grad_norm": 0.15459409356117249,
+      "learning_rate": 2.4990399385560678e-05,
+      "loss": 0.0079,
+      "step": 52600
+    },
+    {
+      "epoch": 10.070705140454805,
+      "grad_norm": 0.5406343340873718,
+      "learning_rate": 2.4942396313364056e-05,
+      "loss": 0.011,
+      "step": 52700
+    },
+    {
+      "epoch": 10.089814637875024,
+      "grad_norm": 0.45881524682044983,
+      "learning_rate": 2.4894393241167435e-05,
+      "loss": 0.0097,
+      "step": 52800
+    },
+    {
+      "epoch": 10.108924135295242,
+      "grad_norm": 1.0400211811065674,
+      "learning_rate": 2.4846390168970814e-05,
+      "loss": 0.0091,
+      "step": 52900
+    },
+    {
+      "epoch": 10.128033632715459,
+      "grad_norm": 0.4115142524242401,
+      "learning_rate": 2.4798387096774196e-05,
+      "loss": 0.0121,
+      "step": 53000
+    },
+    {
+      "epoch": 10.128033632715459,
+      "eval_events-synergy/entsum_processed_loss": 0.01460443064570427,
+      "eval_events-synergy/entsum_processed_runtime": 86.3665,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.932,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.739,
+      "step": 53000
+    },
+    {
+      "epoch": 10.147143130135678,
+      "grad_norm": 0.7973828911781311,
+      "learning_rate": 2.4750384024577575e-05,
+      "loss": 0.0109,
+      "step": 53100
+    },
+    {
+      "epoch": 10.166252627555895,
+      "grad_norm": 0.21007944643497467,
+      "learning_rate": 2.4702380952380953e-05,
+      "loss": 0.01,
+      "step": 53200
+    },
+    {
+      "epoch": 10.185362124976113,
+      "grad_norm": 0.14576242864131927,
+      "learning_rate": 2.4654377880184332e-05,
+      "loss": 0.01,
+      "step": 53300
+    },
+    {
+      "epoch": 10.204471622396332,
+      "grad_norm": 0.08291052281856537,
+      "learning_rate": 2.460637480798771e-05,
+      "loss": 0.0097,
+      "step": 53400
+    },
+    {
+      "epoch": 10.223581119816549,
+      "grad_norm": 0.06554676592350006,
+      "learning_rate": 2.4558371735791093e-05,
+      "loss": 0.0095,
+      "step": 53500
+    },
+    {
+      "epoch": 10.242690617236766,
+      "grad_norm": 0.7229769229888916,
+      "learning_rate": 2.4510368663594472e-05,
+      "loss": 0.0098,
+      "step": 53600
+    },
+    {
+      "epoch": 10.261800114656985,
+      "grad_norm": 1.9222382307052612,
+      "learning_rate": 2.446236559139785e-05,
+      "loss": 0.0161,
+      "step": 53700
+    },
+    {
+      "epoch": 10.280909612077203,
+      "grad_norm": 0.05762649327516556,
+      "learning_rate": 2.4414362519201233e-05,
+      "loss": 0.0092,
+      "step": 53800
+    },
+    {
+      "epoch": 10.30001910949742,
+      "grad_norm": 1.128400206565857,
+      "learning_rate": 2.436635944700461e-05,
+      "loss": 0.0114,
+      "step": 53900
+    },
+    {
+      "epoch": 10.319128606917639,
+      "grad_norm": 0.5140525102615356,
+      "learning_rate": 2.4318356374807987e-05,
+      "loss": 0.017,
+      "step": 54000
+    },
+    {
+      "epoch": 10.319128606917639,
+      "eval_events-synergy/entsum_processed_loss": 0.014034281484782696,
+      "eval_events-synergy/entsum_processed_runtime": 86.3783,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.928,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.738,
+      "step": 54000
+    },
+    {
+      "epoch": 10.338238104337856,
+      "grad_norm": 0.011483533307909966,
+      "learning_rate": 2.427035330261137e-05,
+      "loss": 0.0077,
+      "step": 54100
+    },
+    {
+      "epoch": 10.357347601758073,
+      "grad_norm": 0.1507517546415329,
+      "learning_rate": 2.4222350230414748e-05,
+      "loss": 0.0082,
+      "step": 54200
+    },
+    {
+      "epoch": 10.376457099178293,
+      "grad_norm": 0.3378009498119354,
+      "learning_rate": 2.4174347158218126e-05,
+      "loss": 0.0123,
+      "step": 54300
+    },
+    {
+      "epoch": 10.39556659659851,
+      "grad_norm": 0.491153359413147,
+      "learning_rate": 2.412634408602151e-05,
+      "loss": 0.0105,
+      "step": 54400
+    },
+    {
+      "epoch": 10.414676094018727,
+      "grad_norm": 0.7116023898124695,
+      "learning_rate": 2.4078341013824887e-05,
+      "loss": 0.0098,
+      "step": 54500
+    },
+    {
+      "epoch": 10.433785591438944,
+      "grad_norm": 0.05201850086450577,
+      "learning_rate": 2.4030337941628263e-05,
+      "loss": 0.0108,
+      "step": 54600
+    },
+    {
+      "epoch": 10.452895088859163,
+      "grad_norm": 1.5810301303863525,
+      "learning_rate": 2.3982334869431645e-05,
+      "loss": 0.009,
+      "step": 54700
+    },
+    {
+      "epoch": 10.47200458627938,
+      "grad_norm": 0.7053477764129639,
+      "learning_rate": 2.3934331797235023e-05,
+      "loss": 0.0096,
+      "step": 54800
+    },
+    {
+      "epoch": 10.491114083699598,
+      "grad_norm": 0.5703982710838318,
+      "learning_rate": 2.3886328725038402e-05,
+      "loss": 0.0084,
+      "step": 54900
+    },
+    {
+      "epoch": 10.510223581119817,
+      "grad_norm": 0.3266231119632721,
+      "learning_rate": 2.3838325652841784e-05,
+      "loss": 0.0119,
+      "step": 55000
+    },
+    {
+      "epoch": 10.510223581119817,
+      "eval_events-synergy/entsum_processed_loss": 0.014698924496769905,
+      "eval_events-synergy/entsum_processed_runtime": 86.2536,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.967,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.748,
+      "step": 55000
+    },
+    {
+      "epoch": 10.529333078540034,
+      "grad_norm": 0.17007987201213837,
+      "learning_rate": 2.3790322580645163e-05,
+      "loss": 0.0099,
+      "step": 55100
+    },
+    {
+      "epoch": 10.548442575960252,
+      "grad_norm": 0.9838082194328308,
+      "learning_rate": 2.3742319508448542e-05,
+      "loss": 0.0091,
+      "step": 55200
+    },
+    {
+      "epoch": 10.56755207338047,
+      "grad_norm": 0.32569730281829834,
+      "learning_rate": 2.369431643625192e-05,
+      "loss": 0.0105,
+      "step": 55300
+    },
+    {
+      "epoch": 10.586661570800688,
+      "grad_norm": 0.3832113742828369,
+      "learning_rate": 2.36463133640553e-05,
+      "loss": 0.01,
+      "step": 55400
+    },
+    {
+      "epoch": 10.605771068220905,
+      "grad_norm": 0.43005886673927307,
+      "learning_rate": 2.359831029185868e-05,
+      "loss": 0.0101,
+      "step": 55500
+    },
+    {
+      "epoch": 10.624880565641124,
+      "grad_norm": 1.174752950668335,
+      "learning_rate": 2.355030721966206e-05,
+      "loss": 0.0103,
+      "step": 55600
+    },
+    {
+      "epoch": 10.643990063061342,
+      "grad_norm": 0.16343650221824646,
+      "learning_rate": 2.350230414746544e-05,
+      "loss": 0.0092,
+      "step": 55700
+    },
+    {
+      "epoch": 10.663099560481559,
+      "grad_norm": 0.3868729770183563,
+      "learning_rate": 2.345430107526882e-05,
+      "loss": 0.0155,
+      "step": 55800
+    },
+    {
+      "epoch": 10.682209057901778,
+      "grad_norm": 0.18367870151996613,
+      "learning_rate": 2.3406298003072196e-05,
+      "loss": 0.0093,
+      "step": 55900
+    },
+    {
+      "epoch": 10.701318555321995,
+      "grad_norm": 0.7105177640914917,
+      "learning_rate": 2.3358294930875575e-05,
+      "loss": 0.0086,
+      "step": 56000
+    },
+    {
+      "epoch": 10.701318555321995,
+      "eval_events-synergy/entsum_processed_loss": 0.014125595800578594,
+      "eval_events-synergy/entsum_processed_runtime": 86.2929,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.955,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.744,
+      "step": 56000
+    },
+    {
+      "epoch": 10.720428052742212,
+      "grad_norm": 0.1564301997423172,
+      "learning_rate": 2.3310291858678957e-05,
+      "loss": 0.0068,
+      "step": 56100
+    },
+    {
+      "epoch": 10.739537550162432,
+      "grad_norm": 0.13135480880737305,
+      "learning_rate": 2.3262288786482336e-05,
+      "loss": 0.0105,
+      "step": 56200
+    },
+    {
+      "epoch": 10.758647047582649,
+      "grad_norm": 0.15838299691677094,
+      "learning_rate": 2.3214285714285715e-05,
+      "loss": 0.008,
+      "step": 56300
+    },
+    {
+      "epoch": 10.777756545002866,
+      "grad_norm": 0.42811164259910583,
+      "learning_rate": 2.3166282642089097e-05,
+      "loss": 0.0106,
+      "step": 56400
+    },
+    {
+      "epoch": 10.796866042423083,
+      "grad_norm": 0.1568647176027298,
+      "learning_rate": 2.3118279569892472e-05,
+      "loss": 0.0113,
+      "step": 56500
+    },
+    {
+      "epoch": 10.815975539843302,
+      "grad_norm": 0.41632843017578125,
+      "learning_rate": 2.3070276497695854e-05,
+      "loss": 0.0087,
+      "step": 56600
+    },
+    {
+      "epoch": 10.83508503726352,
+      "grad_norm": 0.3496435582637787,
+      "learning_rate": 2.3022273425499233e-05,
+      "loss": 0.0085,
+      "step": 56700
+    },
+    {
+      "epoch": 10.854194534683739,
+      "grad_norm": 0.04413243755698204,
+      "learning_rate": 2.2974270353302612e-05,
+      "loss": 0.0095,
+      "step": 56800
+    },
+    {
+      "epoch": 10.873304032103956,
+      "grad_norm": 0.2369256615638733,
+      "learning_rate": 2.2926267281105994e-05,
+      "loss": 0.0088,
+      "step": 56900
+    },
+    {
+      "epoch": 10.892413529524173,
+      "grad_norm": 0.07838968932628632,
+      "learning_rate": 2.2878264208909373e-05,
+      "loss": 0.0095,
+      "step": 57000
+    },
+    {
+      "epoch": 10.892413529524173,
+      "eval_events-synergy/entsum_processed_loss": 0.01357998140156269,
+      "eval_events-synergy/entsum_processed_runtime": 86.3828,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.927,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.737,
+      "step": 57000
+    },
+    {
+      "epoch": 10.91152302694439,
+      "grad_norm": 0.3263333737850189,
+      "learning_rate": 2.2830261136712748e-05,
+      "loss": 0.0103,
+      "step": 57100
+    },
+    {
+      "epoch": 10.93063252436461,
+      "grad_norm": 0.7420934438705444,
+      "learning_rate": 2.278225806451613e-05,
+      "loss": 0.0142,
+      "step": 57200
+    },
+    {
+      "epoch": 10.949742021784827,
+      "grad_norm": 0.4374001622200012,
+      "learning_rate": 2.273425499231951e-05,
+      "loss": 0.0101,
+      "step": 57300
+    },
+    {
+      "epoch": 10.968851519205044,
+      "grad_norm": 1.4210389852523804,
+      "learning_rate": 2.2686251920122888e-05,
+      "loss": 0.0102,
+      "step": 57400
+    },
+    {
+      "epoch": 10.987961016625263,
+      "grad_norm": 0.6043274402618408,
+      "learning_rate": 2.263824884792627e-05,
+      "loss": 0.014,
+      "step": 57500
+    },
+    {
+      "epoch": 11.00707051404548,
+      "grad_norm": 0.18480466306209564,
+      "learning_rate": 2.259024577572965e-05,
+      "loss": 0.0097,
+      "step": 57600
+    },
+    {
+      "epoch": 11.026180011465698,
+      "grad_norm": 0.013819389045238495,
+      "learning_rate": 2.2542242703533027e-05,
+      "loss": 0.0073,
+      "step": 57700
+    },
+    {
+      "epoch": 11.045289508885917,
+      "grad_norm": 1.205317497253418,
+      "learning_rate": 2.2494239631336406e-05,
+      "loss": 0.0092,
+      "step": 57800
+    },
+    {
+      "epoch": 11.064399006306134,
+      "grad_norm": 0.15929065644741058,
+      "learning_rate": 2.2446236559139785e-05,
+      "loss": 0.0086,
+      "step": 57900
+    },
+    {
+      "epoch": 11.083508503726351,
+      "grad_norm": 1.1358189582824707,
+      "learning_rate": 2.2398233486943167e-05,
+      "loss": 0.0074,
+      "step": 58000
+    },
+    {
+      "epoch": 11.083508503726351,
+      "eval_events-synergy/entsum_processed_loss": 0.013762025162577629,
+      "eval_events-synergy/entsum_processed_runtime": 86.4105,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.918,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.735,
+      "step": 58000
+    },
+    {
+      "epoch": 11.10261800114657,
+      "grad_norm": 0.684076189994812,
+      "learning_rate": 2.2350230414746546e-05,
+      "loss": 0.0085,
+      "step": 58100
+    },
+    {
+      "epoch": 11.121727498566788,
+      "grad_norm": 1.0887881517410278,
+      "learning_rate": 2.2302227342549924e-05,
+      "loss": 0.0083,
+      "step": 58200
+    },
+    {
+      "epoch": 11.140836995987005,
+      "grad_norm": 0.6392953395843506,
+      "learning_rate": 2.2254224270353306e-05,
+      "loss": 0.0079,
+      "step": 58300
+    },
+    {
+      "epoch": 11.159946493407224,
+      "grad_norm": 0.858275830745697,
+      "learning_rate": 2.2206221198156682e-05,
+      "loss": 0.0084,
+      "step": 58400
+    },
+    {
+      "epoch": 11.179055990827441,
+      "grad_norm": 0.0732496976852417,
+      "learning_rate": 2.215821812596006e-05,
+      "loss": 0.0074,
+      "step": 58500
+    },
+    {
+      "epoch": 11.198165488247659,
+      "grad_norm": 0.38673388957977295,
+      "learning_rate": 2.2110215053763443e-05,
+      "loss": 0.0081,
+      "step": 58600
+    },
+    {
+      "epoch": 11.217274985667878,
+      "grad_norm": 0.09929162263870239,
+      "learning_rate": 2.206221198156682e-05,
+      "loss": 0.0094,
+      "step": 58700
+    },
+    {
+      "epoch": 11.236384483088095,
+      "grad_norm": 0.504121720790863,
+      "learning_rate": 2.20142089093702e-05,
+      "loss": 0.0069,
+      "step": 58800
+    },
+    {
+      "epoch": 11.255493980508312,
+      "grad_norm": 0.11551874130964279,
+      "learning_rate": 2.1966205837173582e-05,
+      "loss": 0.0087,
+      "step": 58900
+    },
+    {
+      "epoch": 11.27460347792853,
+      "grad_norm": 0.8204399943351746,
+      "learning_rate": 2.1918202764976958e-05,
+      "loss": 0.0072,
+      "step": 59000
+    },
+    {
+      "epoch": 11.27460347792853,
+      "eval_events-synergy/entsum_processed_loss": 0.013163074851036072,
+      "eval_events-synergy/entsum_processed_runtime": 86.5869,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.863,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.722,
+      "step": 59000
+    },
+    {
+      "epoch": 11.293712975348749,
+      "grad_norm": 0.2919471859931946,
+      "learning_rate": 2.187019969278034e-05,
+      "loss": 0.0085,
+      "step": 59100
+    },
+    {
+      "epoch": 11.312822472768966,
+      "grad_norm": 0.399214506149292,
+      "learning_rate": 2.182219662058372e-05,
+      "loss": 0.0154,
+      "step": 59200
+    },
+    {
+      "epoch": 11.331931970189183,
+      "grad_norm": 0.9353770613670349,
+      "learning_rate": 2.1774193548387097e-05,
+      "loss": 0.0063,
+      "step": 59300
+    },
+    {
+      "epoch": 11.351041467609402,
+      "grad_norm": 0.24659943580627441,
+      "learning_rate": 2.172619047619048e-05,
+      "loss": 0.0073,
+      "step": 59400
+    },
+    {
+      "epoch": 11.37015096502962,
+      "grad_norm": 0.8196508288383484,
+      "learning_rate": 2.1678187403993858e-05,
+      "loss": 0.0066,
+      "step": 59500
+    },
+    {
+      "epoch": 11.389260462449837,
+      "grad_norm": 0.05071408301591873,
+      "learning_rate": 2.1630184331797233e-05,
+      "loss": 0.0067,
+      "step": 59600
+    },
+    {
+      "epoch": 11.408369959870056,
+      "grad_norm": 0.6439305543899536,
+      "learning_rate": 2.1582181259600616e-05,
+      "loss": 0.0092,
+      "step": 59700
+    },
+    {
+      "epoch": 11.427479457290273,
+      "grad_norm": 0.009084120392799377,
+      "learning_rate": 2.1534178187403994e-05,
+      "loss": 0.0095,
+      "step": 59800
+    },
+    {
+      "epoch": 11.44658895471049,
+      "grad_norm": 0.04028179123997688,
+      "learning_rate": 2.1486175115207373e-05,
+      "loss": 0.0081,
+      "step": 59900
+    },
+    {
+      "epoch": 11.46569845213071,
+      "grad_norm": 0.4081525206565857,
+      "learning_rate": 2.1438172043010755e-05,
+      "loss": 0.0087,
+      "step": 60000
+    },
+    {
+      "epoch": 11.46569845213071,
+      "eval_events-synergy/entsum_processed_loss": 0.013131758198142052,
+      "eval_events-synergy/entsum_processed_runtime": 86.1545,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.998,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.755,
+      "step": 60000
+    },
+    {
+      "epoch": 11.484807949550927,
+      "grad_norm": 0.0647350549697876,
+      "learning_rate": 2.1390168970814134e-05,
+      "loss": 0.008,
+      "step": 60100
+    },
+    {
+      "epoch": 11.503917446971144,
+      "grad_norm": 0.2612893581390381,
+      "learning_rate": 2.1342165898617513e-05,
+      "loss": 0.0084,
+      "step": 60200
+    },
+    {
+      "epoch": 11.523026944391363,
+      "grad_norm": 0.21366460621356964,
+      "learning_rate": 2.129416282642089e-05,
+      "loss": 0.0106,
+      "step": 60300
+    },
+    {
+      "epoch": 11.54213644181158,
+      "grad_norm": 1.9202722311019897,
+      "learning_rate": 2.124615975422427e-05,
+      "loss": 0.0085,
+      "step": 60400
+    },
+    {
+      "epoch": 11.561245939231798,
+      "grad_norm": 0.7038975358009338,
+      "learning_rate": 2.1198156682027652e-05,
+      "loss": 0.0082,
+      "step": 60500
+    },
+    {
+      "epoch": 11.580355436652017,
+      "grad_norm": 0.5983741879463196,
+      "learning_rate": 2.115015360983103e-05,
+      "loss": 0.0074,
+      "step": 60600
+    },
+    {
+      "epoch": 11.599464934072234,
+      "grad_norm": 0.09934480488300323,
+      "learning_rate": 2.110215053763441e-05,
+      "loss": 0.0076,
+      "step": 60700
+    },
+    {
+      "epoch": 11.618574431492451,
+      "grad_norm": 0.021143363788723946,
+      "learning_rate": 2.1054147465437792e-05,
+      "loss": 0.0075,
+      "step": 60800
+    },
+    {
+      "epoch": 11.63768392891267,
+      "grad_norm": 0.14107128977775574,
+      "learning_rate": 2.1006144393241167e-05,
+      "loss": 0.0092,
+      "step": 60900
+    },
+    {
+      "epoch": 11.656793426332888,
+      "grad_norm": 0.695513129234314,
+      "learning_rate": 2.0958141321044546e-05,
+      "loss": 0.0084,
+      "step": 61000
+    },
+    {
+      "epoch": 11.656793426332888,
+      "eval_events-synergy/entsum_processed_loss": 0.013082274235785007,
+      "eval_events-synergy/entsum_processed_runtime": 85.9161,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.073,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.774,
+      "step": 61000
+    },
+    {
+      "epoch": 11.675902923753105,
+      "grad_norm": 0.14702828228473663,
+      "learning_rate": 2.0910138248847928e-05,
+      "loss": 0.008,
+      "step": 61100
+    },
+    {
+      "epoch": 11.695012421173324,
+      "grad_norm": 0.8319016695022583,
+      "learning_rate": 2.0862135176651307e-05,
+      "loss": 0.0077,
+      "step": 61200
+    },
+    {
+      "epoch": 11.714121918593541,
+      "grad_norm": 1.6200295686721802,
+      "learning_rate": 2.0814132104454685e-05,
+      "loss": 0.0096,
+      "step": 61300
+    },
+    {
+      "epoch": 11.733231416013759,
+      "grad_norm": 0.11744115501642227,
+      "learning_rate": 2.0766129032258068e-05,
+      "loss": 0.0069,
+      "step": 61400
+    },
+    {
+      "epoch": 11.752340913433976,
+      "grad_norm": 0.008127202279865742,
+      "learning_rate": 2.0718125960061443e-05,
+      "loss": 0.0082,
+      "step": 61500
+    },
+    {
+      "epoch": 11.771450410854195,
+      "grad_norm": 0.28937405347824097,
+      "learning_rate": 2.0670122887864825e-05,
+      "loss": 0.019,
+      "step": 61600
+    },
+    {
+      "epoch": 11.790559908274412,
+      "grad_norm": 0.5325780510902405,
+      "learning_rate": 2.0622119815668204e-05,
+      "loss": 0.0085,
+      "step": 61700
+    },
+    {
+      "epoch": 11.80966940569463,
+      "grad_norm": 0.3603396415710449,
+      "learning_rate": 2.0574116743471583e-05,
+      "loss": 0.0087,
+      "step": 61800
+    },
+    {
+      "epoch": 11.828778903114848,
+      "grad_norm": 0.3811150789260864,
+      "learning_rate": 2.052611367127496e-05,
+      "loss": 0.007,
+      "step": 61900
+    },
+    {
+      "epoch": 11.847888400535066,
+      "grad_norm": 0.017322426661849022,
+      "learning_rate": 2.0478110599078343e-05,
+      "loss": 0.0077,
+      "step": 62000
+    },
+    {
+      "epoch": 11.847888400535066,
+      "eval_events-synergy/entsum_processed_loss": 0.01322674285620451,
+      "eval_events-synergy/entsum_processed_runtime": 85.465,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.216,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.81,
+      "step": 62000
+    },
+    {
+      "epoch": 11.866997897955283,
+      "grad_norm": 0.5657706260681152,
+      "learning_rate": 2.0430107526881722e-05,
+      "loss": 0.0087,
+      "step": 62100
+    },
+    {
+      "epoch": 11.886107395375502,
+      "grad_norm": 0.7124789953231812,
+      "learning_rate": 2.03821044546851e-05,
+      "loss": 0.008,
+      "step": 62200
+    },
+    {
+      "epoch": 11.90521689279572,
+      "grad_norm": 0.5974131226539612,
+      "learning_rate": 2.033410138248848e-05,
+      "loss": 0.0079,
+      "step": 62300
+    },
+    {
+      "epoch": 11.924326390215937,
+      "grad_norm": 0.722249448299408,
+      "learning_rate": 2.028609831029186e-05,
+      "loss": 0.0086,
+      "step": 62400
+    },
+    {
+      "epoch": 11.943435887636156,
+      "grad_norm": 0.6730612516403198,
+      "learning_rate": 2.023809523809524e-05,
+      "loss": 0.0086,
+      "step": 62500
+    },
+    {
+      "epoch": 11.962545385056373,
+      "grad_norm": 0.4014939069747925,
+      "learning_rate": 2.019009216589862e-05,
+      "loss": 0.0078,
+      "step": 62600
+    },
+    {
+      "epoch": 11.98165488247659,
+      "grad_norm": 0.7014623880386353,
+      "learning_rate": 2.0142089093701998e-05,
+      "loss": 0.0111,
+      "step": 62700
+    },
+    {
+      "epoch": 12.00076437989681,
+      "grad_norm": 0.8713180422782898,
+      "learning_rate": 2.0094086021505377e-05,
+      "loss": 0.0098,
+      "step": 62800
+    },
+    {
+      "epoch": 12.019873877317027,
+      "grad_norm": 0.3292623460292816,
+      "learning_rate": 2.0046082949308755e-05,
+      "loss": 0.0068,
+      "step": 62900
+    },
+    {
+      "epoch": 12.038983374737244,
+      "grad_norm": 0.6158481240272522,
+      "learning_rate": 1.9998079877112134e-05,
+      "loss": 0.007,
+      "step": 63000
+    },
+    {
+      "epoch": 12.038983374737244,
+      "eval_events-synergy/entsum_processed_loss": 0.013169880956411362,
+      "eval_events-synergy/entsum_processed_runtime": 86.3306,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.943,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.742,
+      "step": 63000
+    },
+    {
+      "epoch": 12.058092872157463,
+      "grad_norm": 0.025973064824938774,
+      "learning_rate": 1.9950076804915516e-05,
+      "loss": 0.0071,
+      "step": 63100
+    },
+    {
+      "epoch": 12.07720236957768,
+      "grad_norm": 0.4744926989078522,
+      "learning_rate": 1.9902073732718895e-05,
+      "loss": 0.01,
+      "step": 63200
+    },
+    {
+      "epoch": 12.096311866997898,
+      "grad_norm": 0.7474972009658813,
+      "learning_rate": 1.9854070660522274e-05,
+      "loss": 0.0091,
+      "step": 63300
+    },
+    {
+      "epoch": 12.115421364418117,
+      "grad_norm": 0.13019229471683502,
+      "learning_rate": 1.9806067588325653e-05,
+      "loss": 0.0049,
+      "step": 63400
+    },
+    {
+      "epoch": 12.134530861838334,
+      "grad_norm": 0.8603349328041077,
+      "learning_rate": 1.975806451612903e-05,
+      "loss": 0.0074,
+      "step": 63500
+    },
+    {
+      "epoch": 12.153640359258551,
+      "grad_norm": 0.3122252821922302,
+      "learning_rate": 1.9710061443932413e-05,
+      "loss": 0.0073,
+      "step": 63600
+    },
+    {
+      "epoch": 12.17274985667877,
+      "grad_norm": 0.048645805567502975,
+      "learning_rate": 1.9662058371735792e-05,
+      "loss": 0.0057,
+      "step": 63700
+    },
+    {
+      "epoch": 12.191859354098987,
+      "grad_norm": 0.5081863403320312,
+      "learning_rate": 1.961405529953917e-05,
+      "loss": 0.0088,
+      "step": 63800
+    },
+    {
+      "epoch": 12.210968851519205,
+      "grad_norm": 0.28996503353118896,
+      "learning_rate": 1.9566052227342553e-05,
+      "loss": 0.0065,
+      "step": 63900
+    },
+    {
+      "epoch": 12.230078348939422,
+      "grad_norm": 0.4138163626194,
+      "learning_rate": 1.9518049155145932e-05,
+      "loss": 0.0049,
+      "step": 64000
+    },
+    {
+      "epoch": 12.230078348939422,
+      "eval_events-synergy/entsum_processed_loss": 0.013458210043609142,
+      "eval_events-synergy/entsum_processed_runtime": 86.2248,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.976,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.75,
+      "step": 64000
+    },
+    {
+      "epoch": 12.249187846359641,
+      "grad_norm": 0.057438481599092484,
+      "learning_rate": 1.9470046082949307e-05,
+      "loss": 0.0065,
+      "step": 64100
+    },
+    {
+      "epoch": 12.268297343779858,
+      "grad_norm": 0.40838849544525146,
+      "learning_rate": 1.942204301075269e-05,
+      "loss": 0.0067,
+      "step": 64200
+    },
+    {
+      "epoch": 12.287406841200076,
+      "grad_norm": 0.2802123427391052,
+      "learning_rate": 1.9374039938556068e-05,
+      "loss": 0.0075,
+      "step": 64300
+    },
+    {
+      "epoch": 12.306516338620295,
+      "grad_norm": 0.009569051675498486,
+      "learning_rate": 1.9326036866359447e-05,
+      "loss": 0.0071,
+      "step": 64400
+    },
+    {
+      "epoch": 12.325625836040512,
+      "grad_norm": 0.03632039576768875,
+      "learning_rate": 1.927803379416283e-05,
+      "loss": 0.0068,
+      "step": 64500
+    },
+    {
+      "epoch": 12.34473533346073,
+      "grad_norm": 0.46252092719078064,
+      "learning_rate": 1.9230030721966208e-05,
+      "loss": 0.0062,
+      "step": 64600
+    },
+    {
+      "epoch": 12.363844830880948,
+      "grad_norm": 0.04237058386206627,
+      "learning_rate": 1.9182027649769586e-05,
+      "loss": 0.0078,
+      "step": 64700
+    },
+    {
+      "epoch": 12.382954328301166,
+      "grad_norm": 0.050396714359521866,
+      "learning_rate": 1.9134024577572965e-05,
+      "loss": 0.0089,
+      "step": 64800
+    },
+    {
+      "epoch": 12.402063825721383,
+      "grad_norm": 0.8509294986724854,
+      "learning_rate": 1.9086021505376344e-05,
+      "loss": 0.006,
+      "step": 64900
+    },
+    {
+      "epoch": 12.421173323141602,
+      "grad_norm": 0.07883996516466141,
+      "learning_rate": 1.9038018433179726e-05,
+      "loss": 0.0068,
+      "step": 65000
+    },
+    {
+      "epoch": 12.421173323141602,
+      "eval_events-synergy/entsum_processed_loss": 0.013203460723161697,
+      "eval_events-synergy/entsum_processed_runtime": 86.1892,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.987,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.753,
+      "step": 65000
+    },
+    {
+      "epoch": 12.44028282056182,
+      "grad_norm": 0.0476163849234581,
+      "learning_rate": 1.8990015360983105e-05,
+      "loss": 0.0079,
+      "step": 65100
+    },
+    {
+      "epoch": 12.459392317982037,
+      "grad_norm": 0.02199789695441723,
+      "learning_rate": 1.8942012288786483e-05,
+      "loss": 0.0067,
+      "step": 65200
+    },
+    {
+      "epoch": 12.478501815402256,
+      "grad_norm": 0.05000906065106392,
+      "learning_rate": 1.8894009216589862e-05,
+      "loss": 0.0082,
+      "step": 65300
+    },
+    {
+      "epoch": 12.497611312822473,
+      "grad_norm": 0.8533111810684204,
+      "learning_rate": 1.884600614439324e-05,
+      "loss": 0.0074,
+      "step": 65400
+    },
+    {
+      "epoch": 12.51672081024269,
+      "grad_norm": 1.0276228189468384,
+      "learning_rate": 1.879800307219662e-05,
+      "loss": 0.0055,
+      "step": 65500
+    },
+    {
+      "epoch": 12.53583030766291,
+      "grad_norm": 0.05433342233300209,
+      "learning_rate": 1.8750000000000002e-05,
+      "loss": 0.0089,
+      "step": 65600
+    },
+    {
+      "epoch": 12.554939805083126,
+      "grad_norm": 0.15978392958641052,
+      "learning_rate": 1.870199692780338e-05,
+      "loss": 0.0064,
+      "step": 65700
+    },
+    {
+      "epoch": 12.574049302503344,
+      "grad_norm": 0.3056204319000244,
+      "learning_rate": 1.865399385560676e-05,
+      "loss": 0.0064,
+      "step": 65800
+    },
+    {
+      "epoch": 12.593158799923563,
+      "grad_norm": 0.009515652433037758,
+      "learning_rate": 1.860599078341014e-05,
+      "loss": 0.0069,
+      "step": 65900
+    },
+    {
+      "epoch": 12.61226829734378,
+      "grad_norm": 0.026574306190013885,
+      "learning_rate": 1.8557987711213517e-05,
+      "loss": 0.0061,
+      "step": 66000
+    },
+    {
+      "epoch": 12.61226829734378,
+      "eval_events-synergy/entsum_processed_loss": 0.013461146503686905,
+      "eval_events-synergy/entsum_processed_runtime": 86.4904,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.893,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.729,
+      "step": 66000
+    },
+    {
+      "epoch": 12.631377794763997,
+      "grad_norm": 0.21460069715976715,
+      "learning_rate": 1.85099846390169e-05,
+      "loss": 0.0078,
+      "step": 66100
+    },
+    {
+      "epoch": 12.650487292184216,
+      "grad_norm": 0.29758745431900024,
+      "learning_rate": 1.8461981566820278e-05,
+      "loss": 0.0073,
+      "step": 66200
+    },
+    {
+      "epoch": 12.669596789604434,
+      "grad_norm": 0.4597108066082001,
+      "learning_rate": 1.8413978494623656e-05,
+      "loss": 0.0046,
+      "step": 66300
+    },
+    {
+      "epoch": 12.688706287024651,
+      "grad_norm": 0.07014264166355133,
+      "learning_rate": 1.836597542242704e-05,
+      "loss": 0.0073,
+      "step": 66400
+    },
+    {
+      "epoch": 12.707815784444868,
+      "grad_norm": 0.6235468983650208,
+      "learning_rate": 1.8317972350230417e-05,
+      "loss": 0.0085,
+      "step": 66500
+    },
+    {
+      "epoch": 12.726925281865087,
+      "grad_norm": 0.49677661061286926,
+      "learning_rate": 1.8269969278033793e-05,
+      "loss": 0.0066,
+      "step": 66600
+    },
+    {
+      "epoch": 12.746034779285305,
+      "grad_norm": 0.023319154977798462,
+      "learning_rate": 1.8221966205837175e-05,
+      "loss": 0.0061,
+      "step": 66700
+    },
+    {
+      "epoch": 12.765144276705522,
+      "grad_norm": 0.5906485319137573,
+      "learning_rate": 1.8173963133640553e-05,
+      "loss": 0.0078,
+      "step": 66800
+    },
+    {
+      "epoch": 12.784253774125741,
+      "grad_norm": 0.4909743666648865,
+      "learning_rate": 1.8125960061443932e-05,
+      "loss": 0.008,
+      "step": 66900
+    },
+    {
+      "epoch": 12.803363271545958,
+      "grad_norm": 1.3858819007873535,
+      "learning_rate": 1.8077956989247314e-05,
+      "loss": 0.0066,
+      "step": 67000
+    },
+    {
+      "epoch": 12.803363271545958,
+      "eval_events-synergy/entsum_processed_loss": 0.01290189754217863,
+      "eval_events-synergy/entsum_processed_runtime": 86.4251,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.913,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.734,
+      "step": 67000
+    },
+    {
+      "epoch": 12.822472768966175,
+      "grad_norm": 0.8135608434677124,
+      "learning_rate": 1.8029953917050693e-05,
+      "loss": 0.0087,
+      "step": 67100
+    },
+    {
+      "epoch": 12.841582266386395,
+      "grad_norm": 0.3143382966518402,
+      "learning_rate": 1.7981950844854072e-05,
+      "loss": 0.0074,
+      "step": 67200
+    },
+    {
+      "epoch": 12.860691763806612,
+      "grad_norm": 0.1856617033481598,
+      "learning_rate": 1.793394777265745e-05,
+      "loss": 0.0073,
+      "step": 67300
+    },
+    {
+      "epoch": 12.87980126122683,
+      "grad_norm": 0.0038214472588151693,
+      "learning_rate": 1.788594470046083e-05,
+      "loss": 0.007,
+      "step": 67400
+    },
+    {
+      "epoch": 12.898910758647048,
+      "grad_norm": 0.12916350364685059,
+      "learning_rate": 1.783794162826421e-05,
+      "loss": 0.006,
+      "step": 67500
+    },
+    {
+      "epoch": 12.918020256067265,
+      "grad_norm": 0.3312636613845825,
+      "learning_rate": 1.778993855606759e-05,
+      "loss": 0.0054,
+      "step": 67600
+    },
+    {
+      "epoch": 12.937129753487483,
+      "grad_norm": 0.008426926098763943,
+      "learning_rate": 1.774193548387097e-05,
+      "loss": 0.0055,
+      "step": 67700
+    },
+    {
+      "epoch": 12.956239250907702,
+      "grad_norm": 0.02026149071753025,
+      "learning_rate": 1.769393241167435e-05,
+      "loss": 0.0055,
+      "step": 67800
+    },
+    {
+      "epoch": 12.975348748327919,
+      "grad_norm": 0.13877490162849426,
+      "learning_rate": 1.7645929339477726e-05,
+      "loss": 0.0053,
+      "step": 67900
+    },
+    {
+      "epoch": 12.994458245748136,
+      "grad_norm": 0.4303775429725647,
+      "learning_rate": 1.7597926267281105e-05,
+      "loss": 0.0073,
+      "step": 68000
+    },
+    {
+      "epoch": 12.994458245748136,
+      "eval_events-synergy/entsum_processed_loss": 0.013035867363214493,
+      "eval_events-synergy/entsum_processed_runtime": 86.4767,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.897,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.73,
+      "step": 68000
+    },
+    {
+      "epoch": 13.013567743168355,
+      "grad_norm": 0.43861302733421326,
+      "learning_rate": 1.7549923195084487e-05,
+      "loss": 0.006,
+      "step": 68100
+    },
+    {
+      "epoch": 13.032677240588573,
+      "grad_norm": 0.5167204141616821,
+      "learning_rate": 1.7501920122887866e-05,
+      "loss": 0.0058,
+      "step": 68200
+    },
+    {
+      "epoch": 13.05178673800879,
+      "grad_norm": 0.9433317184448242,
+      "learning_rate": 1.7453917050691245e-05,
+      "loss": 0.0057,
+      "step": 68300
+    },
+    {
+      "epoch": 13.070896235429009,
+      "grad_norm": 0.01894897036254406,
+      "learning_rate": 1.7405913978494627e-05,
+      "loss": 0.0051,
+      "step": 68400
+    },
+    {
+      "epoch": 13.090005732849226,
+      "grad_norm": 0.9866701364517212,
+      "learning_rate": 1.7357910906298002e-05,
+      "loss": 0.006,
+      "step": 68500
+    },
+    {
+      "epoch": 13.109115230269444,
+      "grad_norm": 0.014975917525589466,
+      "learning_rate": 1.7309907834101384e-05,
+      "loss": 0.0045,
+      "step": 68600
+    },
+    {
+      "epoch": 13.128224727689663,
+      "grad_norm": 0.016140179708600044,
+      "learning_rate": 1.7261904761904763e-05,
+      "loss": 0.0046,
+      "step": 68700
+    },
+    {
+      "epoch": 13.14733422510988,
+      "grad_norm": 0.42392489314079285,
+      "learning_rate": 1.7213901689708142e-05,
+      "loss": 0.0073,
+      "step": 68800
+    },
+    {
+      "epoch": 13.166443722530097,
+      "grad_norm": 0.44540801644325256,
+      "learning_rate": 1.7165898617511524e-05,
+      "loss": 0.0061,
+      "step": 68900
+    },
+    {
+      "epoch": 13.185553219950314,
+      "grad_norm": 0.3063792884349823,
+      "learning_rate": 1.7117895545314903e-05,
+      "loss": 0.0062,
+      "step": 69000
+    },
+    {
+      "epoch": 13.185553219950314,
+      "eval_events-synergy/entsum_processed_loss": 0.013066727668046951,
+      "eval_events-synergy/entsum_processed_runtime": 84.3501,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.576,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.9,
+      "step": 69000
+    },
+    {
+      "epoch": 13.204662717370534,
+      "grad_norm": 0.09294837713241577,
+      "learning_rate": 1.7069892473118278e-05,
+      "loss": 0.0056,
+      "step": 69100
+    },
+    {
+      "epoch": 13.22377221479075,
+      "grad_norm": 0.07464176416397095,
+      "learning_rate": 1.702188940092166e-05,
+      "loss": 0.0058,
+      "step": 69200
+    },
+    {
+      "epoch": 13.242881712210968,
+      "grad_norm": 0.06809643656015396,
+      "learning_rate": 1.697388632872504e-05,
+      "loss": 0.005,
+      "step": 69300
+    },
+    {
+      "epoch": 13.261991209631187,
+      "grad_norm": 0.007226106245070696,
+      "learning_rate": 1.6925883256528418e-05,
+      "loss": 0.0076,
+      "step": 69400
+    },
+    {
+      "epoch": 13.281100707051404,
+      "grad_norm": 0.2170906960964203,
+      "learning_rate": 1.68778801843318e-05,
+      "loss": 0.0063,
+      "step": 69500
+    },
+    {
+      "epoch": 13.300210204471622,
+      "grad_norm": 0.01999014802277088,
+      "learning_rate": 1.682987711213518e-05,
+      "loss": 0.0065,
+      "step": 69600
+    },
+    {
+      "epoch": 13.31931970189184,
+      "grad_norm": 0.0101222088560462,
+      "learning_rate": 1.6781874039938557e-05,
+      "loss": 0.0054,
+      "step": 69700
+    },
+    {
+      "epoch": 13.338429199312058,
+      "grad_norm": 0.7783103585243225,
+      "learning_rate": 1.6733870967741936e-05,
+      "loss": 0.0057,
+      "step": 69800
+    },
+    {
+      "epoch": 13.357538696732275,
+      "grad_norm": 0.17000514268875122,
+      "learning_rate": 1.6685867895545315e-05,
+      "loss": 0.0058,
+      "step": 69900
+    },
+    {
+      "epoch": 13.376648194152494,
+      "grad_norm": 0.2570934295654297,
+      "learning_rate": 1.6637864823348693e-05,
+      "loss": 0.006,
+      "step": 70000
+    },
+    {
+      "epoch": 13.376648194152494,
+      "eval_events-synergy/entsum_processed_loss": 0.01279551349580288,
+      "eval_events-synergy/entsum_processed_runtime": 86.2214,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.977,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.75,
+      "step": 70000
+    },
+    {
+      "epoch": 13.395757691572712,
+      "grad_norm": 0.8089013695716858,
+      "learning_rate": 1.6589861751152075e-05,
+      "loss": 0.0061,
+      "step": 70100
+    },
+    {
+      "epoch": 13.414867188992929,
+      "grad_norm": 0.6852768063545227,
+      "learning_rate": 1.6541858678955454e-05,
+      "loss": 0.0044,
+      "step": 70200
+    },
+    {
+      "epoch": 13.433976686413148,
+      "grad_norm": 0.15175899863243103,
+      "learning_rate": 1.6493855606758833e-05,
+      "loss": 0.0061,
+      "step": 70300
+    },
+    {
+      "epoch": 13.453086183833365,
+      "grad_norm": 0.05197775736451149,
+      "learning_rate": 1.6445852534562212e-05,
+      "loss": 0.0063,
+      "step": 70400
+    },
+    {
+      "epoch": 13.472195681253583,
+      "grad_norm": 0.5456545948982239,
+      "learning_rate": 1.639784946236559e-05,
+      "loss": 0.0058,
+      "step": 70500
+    },
+    {
+      "epoch": 13.491305178673802,
+      "grad_norm": 0.16188669204711914,
+      "learning_rate": 1.6349846390168973e-05,
+      "loss": 0.0064,
+      "step": 70600
+    },
+    {
+      "epoch": 13.510414676094019,
+      "grad_norm": 0.8403982520103455,
+      "learning_rate": 1.630184331797235e-05,
+      "loss": 0.0053,
+      "step": 70700
+    },
+    {
+      "epoch": 13.529524173514236,
+      "grad_norm": 0.00283870380371809,
+      "learning_rate": 1.625384024577573e-05,
+      "loss": 0.0064,
+      "step": 70800
+    },
+    {
+      "epoch": 13.548633670934455,
+      "grad_norm": 0.05162491649389267,
+      "learning_rate": 1.6205837173579112e-05,
+      "loss": 0.0054,
+      "step": 70900
+    },
+    {
+      "epoch": 13.567743168354673,
+      "grad_norm": 0.5664875507354736,
+      "learning_rate": 1.6157834101382488e-05,
+      "loss": 0.0054,
+      "step": 71000
+    },
+    {
+      "epoch": 13.567743168354673,
+      "eval_events-synergy/entsum_processed_loss": 0.012747672386467457,
+      "eval_events-synergy/entsum_processed_runtime": 86.3771,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.928,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.738,
+      "step": 71000
+    },
+    {
+      "epoch": 13.58685266577489,
+      "grad_norm": 0.2854113280773163,
+      "learning_rate": 1.6109831029185866e-05,
+      "loss": 0.0047,
+      "step": 71100
+    },
+    {
+      "epoch": 13.605962163195109,
+      "grad_norm": 0.2796891927719116,
+      "learning_rate": 1.606182795698925e-05,
+      "loss": 0.0059,
+      "step": 71200
+    },
+    {
+      "epoch": 13.625071660615326,
+      "grad_norm": 0.4434114098548889,
+      "learning_rate": 1.6013824884792627e-05,
+      "loss": 0.0088,
+      "step": 71300
+    },
+    {
+      "epoch": 13.644181158035543,
+      "grad_norm": 0.11509229242801666,
+      "learning_rate": 1.5965821812596006e-05,
+      "loss": 0.0082,
+      "step": 71400
+    },
+    {
+      "epoch": 13.66329065545576,
+      "grad_norm": 0.034501317888498306,
+      "learning_rate": 1.5917818740399388e-05,
+      "loss": 0.0094,
+      "step": 71500
+    },
+    {
+      "epoch": 13.68240015287598,
+      "grad_norm": 0.483247846364975,
+      "learning_rate": 1.5869815668202767e-05,
+      "loss": 0.005,
+      "step": 71600
+    },
+    {
+      "epoch": 13.701509650296197,
+      "grad_norm": 0.17208580672740936,
+      "learning_rate": 1.5821812596006145e-05,
+      "loss": 0.0051,
+      "step": 71700
+    },
+    {
+      "epoch": 13.720619147716414,
+      "grad_norm": 0.42245280742645264,
+      "learning_rate": 1.5773809523809524e-05,
+      "loss": 0.0052,
+      "step": 71800
+    },
+    {
+      "epoch": 13.739728645136633,
+      "grad_norm": 0.5766732096672058,
+      "learning_rate": 1.5725806451612903e-05,
+      "loss": 0.0056,
+      "step": 71900
+    },
+    {
+      "epoch": 13.75883814255685,
+      "grad_norm": 0.1782081127166748,
+      "learning_rate": 1.5677803379416285e-05,
+      "loss": 0.0067,
+      "step": 72000
+    },
+    {
+      "epoch": 13.75883814255685,
+      "eval_events-synergy/entsum_processed_loss": 0.012470792047679424,
+      "eval_events-synergy/entsum_processed_runtime": 86.3378,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.941,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.741,
+      "step": 72000
+    },
+    {
+      "epoch": 13.777947639977068,
+      "grad_norm": 0.27173900604248047,
+      "learning_rate": 1.5629800307219664e-05,
+      "loss": 0.0042,
+      "step": 72100
+    },
+    {
+      "epoch": 13.797057137397287,
+      "grad_norm": 0.9423706531524658,
+      "learning_rate": 1.5581797235023043e-05,
+      "loss": 0.0061,
+      "step": 72200
+    },
+    {
+      "epoch": 13.816166634817504,
+      "grad_norm": 0.27326226234436035,
+      "learning_rate": 1.553379416282642e-05,
+      "loss": 0.005,
+      "step": 72300
+    },
+    {
+      "epoch": 13.835276132237722,
+      "grad_norm": 0.2667993903160095,
+      "learning_rate": 1.54857910906298e-05,
+      "loss": 0.0055,
+      "step": 72400
+    },
+    {
+      "epoch": 13.85438562965794,
+      "grad_norm": 0.4120780825614929,
+      "learning_rate": 1.543778801843318e-05,
+      "loss": 0.0043,
+      "step": 72500
+    },
+    {
+      "epoch": 13.873495127078158,
+      "grad_norm": 0.6277773976325989,
+      "learning_rate": 1.538978494623656e-05,
+      "loss": 0.0077,
+      "step": 72600
+    },
+    {
+      "epoch": 13.892604624498375,
+      "grad_norm": 0.4718801975250244,
+      "learning_rate": 1.534178187403994e-05,
+      "loss": 0.0056,
+      "step": 72700
+    },
+    {
+      "epoch": 13.911714121918594,
+      "grad_norm": 0.18229953944683075,
+      "learning_rate": 1.529377880184332e-05,
+      "loss": 0.0081,
+      "step": 72800
+    },
+    {
+      "epoch": 13.930823619338812,
+      "grad_norm": 0.08315181732177734,
+      "learning_rate": 1.5245775729646697e-05,
+      "loss": 0.0073,
+      "step": 72900
+    },
+    {
+      "epoch": 13.949933116759029,
+      "grad_norm": 0.006552177015691996,
+      "learning_rate": 1.5197772657450078e-05,
+      "loss": 0.0055,
+      "step": 73000
+    },
+    {
+      "epoch": 13.949933116759029,
+      "eval_events-synergy/entsum_processed_loss": 0.012344392947852612,
+      "eval_events-synergy/entsum_processed_runtime": 86.3295,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.943,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.742,
+      "step": 73000
+    },
+    {
+      "epoch": 13.969042614179248,
+      "grad_norm": 0.150691419839859,
+      "learning_rate": 1.5149769585253456e-05,
+      "loss": 0.0067,
+      "step": 73100
+    },
+    {
+      "epoch": 13.988152111599465,
+      "grad_norm": 0.7237725257873535,
+      "learning_rate": 1.5101766513056837e-05,
+      "loss": 0.0067,
+      "step": 73200
+    },
+    {
+      "epoch": 14.007261609019682,
+      "grad_norm": 0.07672421634197235,
+      "learning_rate": 1.5053763440860215e-05,
+      "loss": 0.0053,
+      "step": 73300
+    },
+    {
+      "epoch": 14.026371106439901,
+      "grad_norm": 0.2431519776582718,
+      "learning_rate": 1.5005760368663596e-05,
+      "loss": 0.0063,
+      "step": 73400
+    },
+    {
+      "epoch": 14.045480603860119,
+      "grad_norm": 0.048979759216308594,
+      "learning_rate": 1.4957757296466976e-05,
+      "loss": 0.005,
+      "step": 73500
+    },
+    {
+      "epoch": 14.064590101280336,
+      "grad_norm": 0.9289604425430298,
+      "learning_rate": 1.4909754224270353e-05,
+      "loss": 0.0036,
+      "step": 73600
+    },
+    {
+      "epoch": 14.083699598700553,
+      "grad_norm": 0.06590162217617035,
+      "learning_rate": 1.4861751152073732e-05,
+      "loss": 0.004,
+      "step": 73700
+    },
+    {
+      "epoch": 14.102809096120772,
+      "grad_norm": 0.3335033357143402,
+      "learning_rate": 1.4813748079877113e-05,
+      "loss": 0.005,
+      "step": 73800
+    },
+    {
+      "epoch": 14.12191859354099,
+      "grad_norm": 0.08950705081224442,
+      "learning_rate": 1.4765745007680493e-05,
+      "loss": 0.0045,
+      "step": 73900
+    },
+    {
+      "epoch": 14.141028090961207,
+      "grad_norm": 0.7303321957588196,
+      "learning_rate": 1.4717741935483872e-05,
+      "loss": 0.0071,
+      "step": 74000
+    },
+    {
+      "epoch": 14.141028090961207,
+      "eval_events-synergy/entsum_processed_loss": 0.012597361579537392,
+      "eval_events-synergy/entsum_processed_runtime": 86.4655,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.901,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.731,
+      "step": 74000
+    },
+    {
+      "epoch": 14.160137588381426,
+      "grad_norm": 0.31180980801582336,
+      "learning_rate": 1.4669738863287252e-05,
+      "loss": 0.0055,
+      "step": 74100
+    },
+    {
+      "epoch": 14.179247085801643,
+      "grad_norm": 0.019986744970083237,
+      "learning_rate": 1.462173579109063e-05,
+      "loss": 0.005,
+      "step": 74200
+    },
+    {
+      "epoch": 14.19835658322186,
+      "grad_norm": 0.6268077492713928,
+      "learning_rate": 1.457373271889401e-05,
+      "loss": 0.0046,
+      "step": 74300
+    },
+    {
+      "epoch": 14.21746608064208,
+      "grad_norm": 0.18772102892398834,
+      "learning_rate": 1.4525729646697388e-05,
+      "loss": 0.0052,
+      "step": 74400
+    },
+    {
+      "epoch": 14.236575578062297,
+      "grad_norm": 0.014273080974817276,
+      "learning_rate": 1.4477726574500769e-05,
+      "loss": 0.0052,
+      "step": 74500
+    },
+    {
+      "epoch": 14.255685075482514,
+      "grad_norm": 0.04534445330500603,
+      "learning_rate": 1.442972350230415e-05,
+      "loss": 0.0078,
+      "step": 74600
+    },
+    {
+      "epoch": 14.274794572902733,
+      "grad_norm": 0.06626222282648087,
+      "learning_rate": 1.4381720430107528e-05,
+      "loss": 0.0049,
+      "step": 74700
+    },
+    {
+      "epoch": 14.29390407032295,
+      "grad_norm": 0.05039865896105766,
+      "learning_rate": 1.4333717357910905e-05,
+      "loss": 0.0062,
+      "step": 74800
+    },
+    {
+      "epoch": 14.313013567743168,
+      "grad_norm": 0.0062214708887040615,
+      "learning_rate": 1.4285714285714285e-05,
+      "loss": 0.0043,
+      "step": 74900
+    },
+    {
+      "epoch": 14.332123065163387,
+      "grad_norm": 0.08620163798332214,
+      "learning_rate": 1.4237711213517666e-05,
+      "loss": 0.0057,
+      "step": 75000
+    },
+    {
+      "epoch": 14.332123065163387,
+      "eval_events-synergy/entsum_processed_loss": 0.012711184099316597,
+      "eval_events-synergy/entsum_processed_runtime": 86.4444,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.907,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.733,
+      "step": 75000
+    },
+    {
+      "epoch": 14.351232562583604,
+      "grad_norm": 0.04931287094950676,
+      "learning_rate": 1.4189708141321045e-05,
+      "loss": 0.0083,
+      "step": 75100
+    },
+    {
+      "epoch": 14.370342060003821,
+      "grad_norm": 0.01421100553125143,
+      "learning_rate": 1.4141705069124425e-05,
+      "loss": 0.004,
+      "step": 75200
+    },
+    {
+      "epoch": 14.38945155742404,
+      "grad_norm": 0.011329934932291508,
+      "learning_rate": 1.4093701996927805e-05,
+      "loss": 0.0049,
+      "step": 75300
+    },
+    {
+      "epoch": 14.408561054844258,
+      "grad_norm": 0.15495194494724274,
+      "learning_rate": 1.4045698924731183e-05,
+      "loss": 0.0032,
+      "step": 75400
+    },
+    {
+      "epoch": 14.427670552264475,
+      "grad_norm": 0.05016593262553215,
+      "learning_rate": 1.3997695852534561e-05,
+      "loss": 0.0055,
+      "step": 75500
+    },
+    {
+      "epoch": 14.446780049684694,
+      "grad_norm": 1.6086093187332153,
+      "learning_rate": 1.3949692780337942e-05,
+      "loss": 0.005,
+      "step": 75600
+    },
+    {
+      "epoch": 14.465889547104911,
+      "grad_norm": 0.4085695743560791,
+      "learning_rate": 1.3901689708141322e-05,
+      "loss": 0.0054,
+      "step": 75700
+    },
+    {
+      "epoch": 14.484999044525129,
+      "grad_norm": 0.1665886640548706,
+      "learning_rate": 1.3853686635944701e-05,
+      "loss": 0.0047,
+      "step": 75800
+    },
+    {
+      "epoch": 14.504108541945346,
+      "grad_norm": 0.1481260061264038,
+      "learning_rate": 1.3805683563748081e-05,
+      "loss": 0.0043,
+      "step": 75900
+    },
+    {
+      "epoch": 14.523218039365565,
+      "grad_norm": 0.22774682939052582,
+      "learning_rate": 1.3757680491551462e-05,
+      "loss": 0.0037,
+      "step": 76000
+    },
+    {
+      "epoch": 14.523218039365565,
+      "eval_events-synergy/entsum_processed_loss": 0.012892031110823154,
+      "eval_events-synergy/entsum_processed_runtime": 86.4183,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.916,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.735,
+      "step": 76000
+    },
+    {
+      "epoch": 14.542327536785782,
+      "grad_norm": 0.34367671608924866,
+      "learning_rate": 1.3709677419354839e-05,
+      "loss": 0.0054,
+      "step": 76100
+    },
+    {
+      "epoch": 14.561437034206,
+      "grad_norm": 0.7251562476158142,
+      "learning_rate": 1.3661674347158218e-05,
+      "loss": 0.0062,
+      "step": 76200
+    },
+    {
+      "epoch": 14.580546531626219,
+      "grad_norm": 0.1083359494805336,
+      "learning_rate": 1.3613671274961598e-05,
+      "loss": 0.005,
+      "step": 76300
+    },
+    {
+      "epoch": 14.599656029046436,
+      "grad_norm": 0.08868356049060822,
+      "learning_rate": 1.3565668202764978e-05,
+      "loss": 0.004,
+      "step": 76400
+    },
+    {
+      "epoch": 14.618765526466653,
+      "grad_norm": 0.4599528908729553,
+      "learning_rate": 1.3517665130568357e-05,
+      "loss": 0.0045,
+      "step": 76500
+    },
+    {
+      "epoch": 14.637875023886872,
+      "grad_norm": 0.5329752564430237,
+      "learning_rate": 1.3469662058371738e-05,
+      "loss": 0.0052,
+      "step": 76600
+    },
+    {
+      "epoch": 14.65698452130709,
+      "grad_norm": 0.050466131418943405,
+      "learning_rate": 1.3421658986175115e-05,
+      "loss": 0.0049,
+      "step": 76700
+    },
+    {
+      "epoch": 14.676094018727307,
+      "grad_norm": 0.2970987856388092,
+      "learning_rate": 1.3373655913978495e-05,
+      "loss": 0.0042,
+      "step": 76800
+    },
+    {
+      "epoch": 14.695203516147526,
+      "grad_norm": 0.10357856750488281,
+      "learning_rate": 1.3325652841781874e-05,
+      "loss": 0.005,
+      "step": 76900
+    },
+    {
+      "epoch": 14.714313013567743,
+      "grad_norm": 0.05402665585279465,
+      "learning_rate": 1.3277649769585254e-05,
+      "loss": 0.0047,
+      "step": 77000
+    },
+    {
+      "epoch": 14.714313013567743,
+      "eval_events-synergy/entsum_processed_loss": 0.012962962500751019,
+      "eval_events-synergy/entsum_processed_runtime": 86.4889,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.894,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.729,
+      "step": 77000
+    },
+    {
+      "epoch": 14.73342251098796,
+      "grad_norm": 0.15317828953266144,
+      "learning_rate": 1.3229646697388635e-05,
+      "loss": 0.0054,
+      "step": 77100
+    },
+    {
+      "epoch": 14.75253200840818,
+      "grad_norm": 0.1756182163953781,
+      "learning_rate": 1.3181643625192013e-05,
+      "loss": 0.0055,
+      "step": 77200
+    },
+    {
+      "epoch": 14.771641505828397,
+      "grad_norm": 1.027187466621399,
+      "learning_rate": 1.313364055299539e-05,
+      "loss": 0.0034,
+      "step": 77300
+    },
+    {
+      "epoch": 14.790751003248614,
+      "grad_norm": 0.005386768840253353,
+      "learning_rate": 1.308563748079877e-05,
+      "loss": 0.0043,
+      "step": 77400
+    },
+    {
+      "epoch": 14.809860500668833,
+      "grad_norm": 0.005857468582689762,
+      "learning_rate": 1.3037634408602151e-05,
+      "loss": 0.0045,
+      "step": 77500
+    },
+    {
+      "epoch": 14.82896999808905,
+      "grad_norm": 0.19254671037197113,
+      "learning_rate": 1.298963133640553e-05,
+      "loss": 0.0041,
+      "step": 77600
+    },
+    {
+      "epoch": 14.848079495509268,
+      "grad_norm": 0.00358038698323071,
+      "learning_rate": 1.294162826420891e-05,
+      "loss": 0.0045,
+      "step": 77700
+    },
+    {
+      "epoch": 14.867188992929487,
+      "grad_norm": 0.019394252449274063,
+      "learning_rate": 1.2893625192012291e-05,
+      "loss": 0.0069,
+      "step": 77800
+    },
+    {
+      "epoch": 14.886298490349704,
+      "grad_norm": 0.03039630688726902,
+      "learning_rate": 1.284562211981567e-05,
+      "loss": 0.0039,
+      "step": 77900
+    },
+    {
+      "epoch": 14.905407987769921,
+      "grad_norm": 0.20222346484661102,
+      "learning_rate": 1.2797619047619047e-05,
+      "loss": 0.005,
+      "step": 78000
+    },
+    {
+      "epoch": 14.905407987769921,
+      "eval_events-synergy/entsum_processed_loss": 0.012617943808436394,
+      "eval_events-synergy/entsum_processed_runtime": 86.5195,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.884,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.727,
+      "step": 78000
+    },
+    {
+      "epoch": 14.92451748519014,
+      "grad_norm": 0.049767039716243744,
+      "learning_rate": 1.2749615975422427e-05,
+      "loss": 0.0049,
+      "step": 78100
+    },
+    {
+      "epoch": 14.943626982610358,
+      "grad_norm": 0.0789194330573082,
+      "learning_rate": 1.2701612903225808e-05,
+      "loss": 0.0049,
+      "step": 78200
+    },
+    {
+      "epoch": 14.962736480030575,
+      "grad_norm": 0.013039722107350826,
+      "learning_rate": 1.2653609831029186e-05,
+      "loss": 0.0049,
+      "step": 78300
+    },
+    {
+      "epoch": 14.981845977450792,
+      "grad_norm": 0.00856608897447586,
+      "learning_rate": 1.2605606758832567e-05,
+      "loss": 0.0045,
+      "step": 78400
+    },
+    {
+      "epoch": 15.000955474871011,
+      "grad_norm": 0.22470723092556,
+      "learning_rate": 1.2557603686635947e-05,
+      "loss": 0.0042,
+      "step": 78500
+    },
+    {
+      "epoch": 15.020064972291229,
+      "grad_norm": 0.03667416051030159,
+      "learning_rate": 1.2509600614439324e-05,
+      "loss": 0.0044,
+      "step": 78600
+    },
+    {
+      "epoch": 15.039174469711446,
+      "grad_norm": 0.2480105310678482,
+      "learning_rate": 1.2461597542242705e-05,
+      "loss": 0.004,
+      "step": 78700
+    },
+    {
+      "epoch": 15.058283967131665,
+      "grad_norm": 0.24030384421348572,
+      "learning_rate": 1.2413594470046083e-05,
+      "loss": 0.0048,
+      "step": 78800
+    },
+    {
+      "epoch": 15.077393464551882,
+      "grad_norm": 0.891026496887207,
+      "learning_rate": 1.2365591397849464e-05,
+      "loss": 0.0047,
+      "step": 78900
+    },
+    {
+      "epoch": 15.0965029619721,
+      "grad_norm": 0.666159987449646,
+      "learning_rate": 1.2317588325652843e-05,
+      "loss": 0.0042,
+      "step": 79000
+    },
+    {
+      "epoch": 15.0965029619721,
+      "eval_events-synergy/entsum_processed_loss": 0.013020485639572144,
+      "eval_events-synergy/entsum_processed_runtime": 86.4768,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.897,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.73,
+      "step": 79000
+    },
+    {
+      "epoch": 15.115612459392318,
+      "grad_norm": 0.4374198913574219,
+      "learning_rate": 1.2269585253456221e-05,
+      "loss": 0.0046,
+      "step": 79100
+    },
+    {
+      "epoch": 15.134721956812536,
+      "grad_norm": 0.13781531155109406,
+      "learning_rate": 1.2221582181259602e-05,
+      "loss": 0.0035,
+      "step": 79200
+    },
+    {
+      "epoch": 15.153831454232753,
+      "grad_norm": 0.008838070556521416,
+      "learning_rate": 1.217357910906298e-05,
+      "loss": 0.0046,
+      "step": 79300
+    },
+    {
+      "epoch": 15.172940951652972,
+      "grad_norm": 0.04405039921402931,
+      "learning_rate": 1.212557603686636e-05,
+      "loss": 0.0035,
+      "step": 79400
+    },
+    {
+      "epoch": 15.19205044907319,
+      "grad_norm": 0.386900395154953,
+      "learning_rate": 1.207757296466974e-05,
+      "loss": 0.0042,
+      "step": 79500
+    },
+    {
+      "epoch": 15.211159946493407,
+      "grad_norm": 0.06320745497941971,
+      "learning_rate": 1.202956989247312e-05,
+      "loss": 0.0049,
+      "step": 79600
+    },
+    {
+      "epoch": 15.230269443913626,
+      "grad_norm": 0.6713384985923767,
+      "learning_rate": 1.1981566820276497e-05,
+      "loss": 0.0047,
+      "step": 79700
+    },
+    {
+      "epoch": 15.249378941333843,
+      "grad_norm": 0.23006173968315125,
+      "learning_rate": 1.1933563748079878e-05,
+      "loss": 0.0042,
+      "step": 79800
+    },
+    {
+      "epoch": 15.26848843875406,
+      "grad_norm": 0.24073895812034607,
+      "learning_rate": 1.1885560675883258e-05,
+      "loss": 0.0055,
+      "step": 79900
+    },
+    {
+      "epoch": 15.28759793617428,
+      "grad_norm": 0.2813238799571991,
+      "learning_rate": 1.1837557603686637e-05,
+      "loss": 0.0035,
+      "step": 80000
+    },
+    {
+      "epoch": 15.28759793617428,
+      "eval_events-synergy/entsum_processed_loss": 0.012684877961874008,
+      "eval_events-synergy/entsum_processed_runtime": 86.376,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.929,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.738,
+      "step": 80000
+    },
+    {
+      "epoch": 15.306707433594497,
+      "grad_norm": 0.05870668590068817,
+      "learning_rate": 1.1789554531490015e-05,
+      "loss": 0.0052,
+      "step": 80100
+    },
+    {
+      "epoch": 15.325816931014714,
+      "grad_norm": 0.383779913187027,
+      "learning_rate": 1.1741551459293396e-05,
+      "loss": 0.0056,
+      "step": 80200
+    },
+    {
+      "epoch": 15.344926428434933,
+      "grad_norm": 0.18271517753601074,
+      "learning_rate": 1.1693548387096775e-05,
+      "loss": 0.0042,
+      "step": 80300
+    },
+    {
+      "epoch": 15.36403592585515,
+      "grad_norm": 0.049872301518917084,
+      "learning_rate": 1.1645545314900153e-05,
+      "loss": 0.0041,
+      "step": 80400
+    },
+    {
+      "epoch": 15.383145423275367,
+      "grad_norm": 0.16262871026992798,
+      "learning_rate": 1.1597542242703534e-05,
+      "loss": 0.0045,
+      "step": 80500
+    },
+    {
+      "epoch": 15.402254920695587,
+      "grad_norm": 0.12631197273731232,
+      "learning_rate": 1.1549539170506914e-05,
+      "loss": 0.0086,
+      "step": 80600
+    },
+    {
+      "epoch": 15.421364418115804,
+      "grad_norm": 0.02645757794380188,
+      "learning_rate": 1.1501536098310293e-05,
+      "loss": 0.0044,
+      "step": 80700
+    },
+    {
+      "epoch": 15.440473915536021,
+      "grad_norm": 0.4429394006729126,
+      "learning_rate": 1.1453533026113672e-05,
+      "loss": 0.0041,
+      "step": 80800
+    },
+    {
+      "epoch": 15.459583412956238,
+      "grad_norm": 0.4982936382293701,
+      "learning_rate": 1.1405529953917052e-05,
+      "loss": 0.0041,
+      "step": 80900
+    },
+    {
+      "epoch": 15.478692910376457,
+      "grad_norm": 0.03018922545015812,
+      "learning_rate": 1.135752688172043e-05,
+      "loss": 0.0041,
+      "step": 81000
+    },
+    {
+      "epoch": 15.478692910376457,
+      "eval_events-synergy/entsum_processed_loss": 0.012665621004998684,
+      "eval_events-synergy/entsum_processed_runtime": 86.4282,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.913,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.734,
+      "step": 81000
+    },
+    {
+      "epoch": 15.497802407796675,
+      "grad_norm": 0.04739225283265114,
+      "learning_rate": 1.130952380952381e-05,
+      "loss": 0.0035,
+      "step": 81100
+    },
+    {
+      "epoch": 15.516911905216892,
+      "grad_norm": 0.16332654654979706,
+      "learning_rate": 1.126152073732719e-05,
+      "loss": 0.0042,
+      "step": 81200
+    },
+    {
+      "epoch": 15.536021402637111,
+      "grad_norm": 0.24561795592308044,
+      "learning_rate": 1.1213517665130569e-05,
+      "loss": 0.0038,
+      "step": 81300
+    },
+    {
+      "epoch": 15.555130900057328,
+      "grad_norm": 0.6066825985908508,
+      "learning_rate": 1.1165514592933947e-05,
+      "loss": 0.0066,
+      "step": 81400
+    },
+    {
+      "epoch": 15.574240397477546,
+      "grad_norm": 0.12053509801626205,
+      "learning_rate": 1.1117511520737328e-05,
+      "loss": 0.0028,
+      "step": 81500
+    },
+    {
+      "epoch": 15.593349894897765,
+      "grad_norm": 0.21273471415042877,
+      "learning_rate": 1.1069508448540707e-05,
+      "loss": 0.004,
+      "step": 81600
+    },
+    {
+      "epoch": 15.612459392317982,
+      "grad_norm": 0.05035385861992836,
+      "learning_rate": 1.1021505376344087e-05,
+      "loss": 0.0029,
+      "step": 81700
+    },
+    {
+      "epoch": 15.6315688897382,
+      "grad_norm": 0.4434753656387329,
+      "learning_rate": 1.0973502304147466e-05,
+      "loss": 0.0037,
+      "step": 81800
+    },
+    {
+      "epoch": 15.650678387158418,
+      "grad_norm": 0.05798697844147682,
+      "learning_rate": 1.0925499231950845e-05,
+      "loss": 0.0037,
+      "step": 81900
+    },
+    {
+      "epoch": 15.669787884578636,
+      "grad_norm": 0.04873861372470856,
+      "learning_rate": 1.0877496159754225e-05,
+      "loss": 0.0041,
+      "step": 82000
+    },
+    {
+      "epoch": 15.669787884578636,
+      "eval_events-synergy/entsum_processed_loss": 0.012837884947657585,
+      "eval_events-synergy/entsum_processed_runtime": 86.4918,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.893,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.729,
+      "step": 82000
+    },
+    {
+      "epoch": 15.688897381998853,
+      "grad_norm": 0.0296151302754879,
+      "learning_rate": 1.0829493087557604e-05,
+      "loss": 0.0049,
+      "step": 82100
+    },
+    {
+      "epoch": 15.708006879419072,
+      "grad_norm": 0.43933552503585815,
+      "learning_rate": 1.0781490015360982e-05,
+      "loss": 0.0044,
+      "step": 82200
+    },
+    {
+      "epoch": 15.72711637683929,
+      "grad_norm": 0.016396446153521538,
+      "learning_rate": 1.0733486943164363e-05,
+      "loss": 0.0048,
+      "step": 82300
+    },
+    {
+      "epoch": 15.746225874259506,
+      "grad_norm": 0.027049915865063667,
+      "learning_rate": 1.0685483870967743e-05,
+      "loss": 0.0054,
+      "step": 82400
+    },
+    {
+      "epoch": 15.765335371679726,
+      "grad_norm": 0.16051483154296875,
+      "learning_rate": 1.0637480798771122e-05,
+      "loss": 0.0036,
+      "step": 82500
+    },
+    {
+      "epoch": 15.784444869099943,
+      "grad_norm": 0.24747684597969055,
+      "learning_rate": 1.05894777265745e-05,
+      "loss": 0.0042,
+      "step": 82600
+    },
+    {
+      "epoch": 15.80355436652016,
+      "grad_norm": 0.14130574464797974,
+      "learning_rate": 1.0541474654377881e-05,
+      "loss": 0.0037,
+      "step": 82700
+    },
+    {
+      "epoch": 15.82266386394038,
+      "grad_norm": 1.0581270456314087,
+      "learning_rate": 1.049347158218126e-05,
+      "loss": 0.0044,
+      "step": 82800
+    },
+    {
+      "epoch": 15.841773361360596,
+      "grad_norm": 0.25455254316329956,
+      "learning_rate": 1.0445468509984639e-05,
+      "loss": 0.004,
+      "step": 82900
+    },
+    {
+      "epoch": 15.860882858780814,
+      "grad_norm": 0.01451896596699953,
+      "learning_rate": 1.039746543778802e-05,
+      "loss": 0.0038,
+      "step": 83000
+    },
+    {
+      "epoch": 15.860882858780814,
+      "eval_events-synergy/entsum_processed_loss": 0.012506287544965744,
+      "eval_events-synergy/entsum_processed_runtime": 86.5196,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.884,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.727,
+      "step": 83000
+    },
+    {
+      "epoch": 15.879992356201033,
+      "grad_norm": 0.1098129004240036,
+      "learning_rate": 1.03494623655914e-05,
+      "loss": 0.0038,
+      "step": 83100
+    },
+    {
+      "epoch": 15.89910185362125,
+      "grad_norm": 0.017855707556009293,
+      "learning_rate": 1.0301459293394777e-05,
+      "loss": 0.0039,
+      "step": 83200
+    },
+    {
+      "epoch": 15.918211351041467,
+      "grad_norm": 0.3735440671443939,
+      "learning_rate": 1.0253456221198157e-05,
+      "loss": 0.005,
+      "step": 83300
+    },
+    {
+      "epoch": 15.937320848461685,
+      "grad_norm": 0.3725346028804779,
+      "learning_rate": 1.0205453149001538e-05,
+      "loss": 0.0043,
+      "step": 83400
+    },
+    {
+      "epoch": 15.956430345881904,
+      "grad_norm": 0.717900812625885,
+      "learning_rate": 1.0157450076804916e-05,
+      "loss": 0.0043,
+      "step": 83500
+    },
+    {
+      "epoch": 15.975539843302121,
+      "grad_norm": 0.006927446462213993,
+      "learning_rate": 1.0109447004608295e-05,
+      "loss": 0.0028,
+      "step": 83600
+    },
+    {
+      "epoch": 15.994649340722338,
+      "grad_norm": 0.23666119575500488,
+      "learning_rate": 1.0061443932411675e-05,
+      "loss": 0.0036,
+      "step": 83700
+    },
+    {
+      "epoch": 16.013758838142557,
+      "grad_norm": 0.07709264755249023,
+      "learning_rate": 1.0013440860215054e-05,
+      "loss": 0.0029,
+      "step": 83800
+    },
+    {
+      "epoch": 16.032868335562775,
+      "grad_norm": 0.2303076684474945,
+      "learning_rate": 9.965437788018433e-06,
+      "loss": 0.0036,
+      "step": 83900
+    },
+    {
+      "epoch": 16.051977832982992,
+      "grad_norm": 0.4032735228538513,
+      "learning_rate": 9.917434715821813e-06,
+      "loss": 0.0035,
+      "step": 84000
+    },
+    {
+      "epoch": 16.051977832982992,
+      "eval_events-synergy/entsum_processed_loss": 0.012511693872511387,
+      "eval_events-synergy/entsum_processed_runtime": 86.3726,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.93,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.738,
+      "step": 84000
+    },
+    {
+      "epoch": 16.07108733040321,
+      "grad_norm": 0.020075472071766853,
+      "learning_rate": 9.869431643625192e-06,
+      "loss": 0.0035,
+      "step": 84100
+    },
+    {
+      "epoch": 16.09019682782343,
+      "grad_norm": 0.008200431242585182,
+      "learning_rate": 9.821428571428573e-06,
+      "loss": 0.0033,
+      "step": 84200
+    },
+    {
+      "epoch": 16.109306325243647,
+      "grad_norm": 0.5671067237854004,
+      "learning_rate": 9.773425499231951e-06,
+      "loss": 0.0031,
+      "step": 84300
+    },
+    {
+      "epoch": 16.128415822663865,
+      "grad_norm": 0.5184391140937805,
+      "learning_rate": 9.725422427035332e-06,
+      "loss": 0.0034,
+      "step": 84400
+    },
+    {
+      "epoch": 16.147525320084082,
+      "grad_norm": 0.8578518033027649,
+      "learning_rate": 9.67741935483871e-06,
+      "loss": 0.004,
+      "step": 84500
+    },
+    {
+      "epoch": 16.1666348175043,
+      "grad_norm": 0.0992913693189621,
+      "learning_rate": 9.62941628264209e-06,
+      "loss": 0.0037,
+      "step": 84600
+    },
+    {
+      "epoch": 16.185744314924516,
+      "grad_norm": 0.2536514699459076,
+      "learning_rate": 9.58141321044547e-06,
+      "loss": 0.0041,
+      "step": 84700
+    },
+    {
+      "epoch": 16.204853812344734,
+      "grad_norm": 0.03721817210316658,
+      "learning_rate": 9.533410138248848e-06,
+      "loss": 0.0026,
+      "step": 84800
+    },
+    {
+      "epoch": 16.223963309764954,
+      "grad_norm": 0.008737379685044289,
+      "learning_rate": 9.485407066052227e-06,
+      "loss": 0.0034,
+      "step": 84900
+    },
+    {
+      "epoch": 16.24307280718517,
+      "grad_norm": 0.04447716847062111,
+      "learning_rate": 9.437403993855607e-06,
+      "loss": 0.0034,
+      "step": 85000
+    },
+    {
+      "epoch": 16.24307280718517,
+      "eval_events-synergy/entsum_processed_loss": 0.012916664592921734,
+      "eval_events-synergy/entsum_processed_runtime": 85.6566,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.155,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.795,
+      "step": 85000
+    },
+    {
+      "epoch": 16.26218230460539,
+      "grad_norm": 0.015641506761312485,
+      "learning_rate": 9.389400921658986e-06,
+      "loss": 0.0027,
+      "step": 85100
+    },
+    {
+      "epoch": 16.281291802025606,
+      "grad_norm": 0.8110576272010803,
+      "learning_rate": 9.341397849462367e-06,
+      "loss": 0.0041,
+      "step": 85200
+    },
+    {
+      "epoch": 16.300401299445824,
+      "grad_norm": 0.005773237440735102,
+      "learning_rate": 9.293394777265745e-06,
+      "loss": 0.0037,
+      "step": 85300
+    },
+    {
+      "epoch": 16.31951079686604,
+      "grad_norm": 0.003712354227900505,
+      "learning_rate": 9.245391705069124e-06,
+      "loss": 0.0036,
+      "step": 85400
+    },
+    {
+      "epoch": 16.33862029428626,
+      "grad_norm": 0.31228092312812805,
+      "learning_rate": 9.197388632872505e-06,
+      "loss": 0.0037,
+      "step": 85500
+    },
+    {
+      "epoch": 16.35772979170648,
+      "grad_norm": 0.2536914646625519,
+      "learning_rate": 9.149385560675883e-06,
+      "loss": 0.004,
+      "step": 85600
+    },
+    {
+      "epoch": 16.376839289126696,
+      "grad_norm": 0.18308839201927185,
+      "learning_rate": 9.101382488479262e-06,
+      "loss": 0.0032,
+      "step": 85700
+    },
+    {
+      "epoch": 16.395948786546914,
+      "grad_norm": 0.008483047597110271,
+      "learning_rate": 9.053379416282642e-06,
+      "loss": 0.0031,
+      "step": 85800
+    },
+    {
+      "epoch": 16.41505828396713,
+      "grad_norm": 0.4126696288585663,
+      "learning_rate": 9.005376344086023e-06,
+      "loss": 0.0029,
+      "step": 85900
+    },
+    {
+      "epoch": 16.434167781387348,
+      "grad_norm": 0.24156266450881958,
+      "learning_rate": 8.9573732718894e-06,
+      "loss": 0.0057,
+      "step": 86000
+    },
+    {
+      "epoch": 16.434167781387348,
+      "eval_events-synergy/entsum_processed_loss": 0.012442585080862045,
+      "eval_events-synergy/entsum_processed_runtime": 86.8846,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.771,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.699,
+      "step": 86000
+    },
+    {
+      "epoch": 16.45327727880757,
+      "grad_norm": 0.011875503696501255,
+      "learning_rate": 8.90937019969278e-06,
+      "loss": 0.004,
+      "step": 86100
+    },
+    {
+      "epoch": 16.472386776227786,
+      "grad_norm": 0.02928953431546688,
+      "learning_rate": 8.86136712749616e-06,
+      "loss": 0.0049,
+      "step": 86200
+    },
+    {
+      "epoch": 16.491496273648004,
+      "grad_norm": 0.04371361806988716,
+      "learning_rate": 8.81336405529954e-06,
+      "loss": 0.0039,
+      "step": 86300
+    },
+    {
+      "epoch": 16.51060577106822,
+      "grad_norm": 0.023907937109470367,
+      "learning_rate": 8.765360983102918e-06,
+      "loss": 0.0025,
+      "step": 86400
+    },
+    {
+      "epoch": 16.529715268488438,
+      "grad_norm": 1.1203492879867554,
+      "learning_rate": 8.717357910906299e-06,
+      "loss": 0.0038,
+      "step": 86500
+    },
+    {
+      "epoch": 16.548824765908655,
+      "grad_norm": 0.01931983418762684,
+      "learning_rate": 8.66935483870968e-06,
+      "loss": 0.0046,
+      "step": 86600
+    },
+    {
+      "epoch": 16.567934263328873,
+      "grad_norm": 0.05801007151603699,
+      "learning_rate": 8.621351766513056e-06,
+      "loss": 0.0039,
+      "step": 86700
+    },
+    {
+      "epoch": 16.587043760749093,
+      "grad_norm": 0.3823702335357666,
+      "learning_rate": 8.573348694316437e-06,
+      "loss": 0.0031,
+      "step": 86800
+    },
+    {
+      "epoch": 16.60615325816931,
+      "grad_norm": 0.1722560077905655,
+      "learning_rate": 8.525345622119817e-06,
+      "loss": 0.0032,
+      "step": 86900
+    },
+    {
+      "epoch": 16.625262755589528,
+      "grad_norm": 1.4988058805465698,
+      "learning_rate": 8.477342549923196e-06,
+      "loss": 0.0039,
+      "step": 87000
+    },
+    {
+      "epoch": 16.625262755589528,
+      "eval_events-synergy/entsum_processed_loss": 0.012560435570776463,
+      "eval_events-synergy/entsum_processed_runtime": 86.7389,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.816,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.71,
+      "step": 87000
+    },
+    {
+      "epoch": 16.644372253009745,
+      "grad_norm": 0.15429021418094635,
+      "learning_rate": 8.429339477726575e-06,
+      "loss": 0.0034,
+      "step": 87100
+    },
+    {
+      "epoch": 16.663481750429963,
+      "grad_norm": 0.11306668817996979,
+      "learning_rate": 8.381336405529955e-06,
+      "loss": 0.0034,
+      "step": 87200
+    },
+    {
+      "epoch": 16.68259124785018,
+      "grad_norm": 0.7127687335014343,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 0.0046,
+      "step": 87300
+    },
+    {
+      "epoch": 16.7017007452704,
+      "grad_norm": 0.12723377346992493,
+      "learning_rate": 8.285330261136712e-06,
+      "loss": 0.0037,
+      "step": 87400
+    },
+    {
+      "epoch": 16.720810242690618,
+      "grad_norm": 0.08388692140579224,
+      "learning_rate": 8.237327188940093e-06,
+      "loss": 0.0024,
+      "step": 87500
+    },
+    {
+      "epoch": 16.739919740110835,
+      "grad_norm": 0.024774732068181038,
+      "learning_rate": 8.189324116743472e-06,
+      "loss": 0.0042,
+      "step": 87600
+    },
+    {
+      "epoch": 16.759029237531053,
+      "grad_norm": 0.021856360137462616,
+      "learning_rate": 8.141321044546852e-06,
+      "loss": 0.0033,
+      "step": 87700
+    },
+    {
+      "epoch": 16.77813873495127,
+      "grad_norm": 0.09436067193746567,
+      "learning_rate": 8.09331797235023e-06,
+      "loss": 0.0041,
+      "step": 87800
+    },
+    {
+      "epoch": 16.797248232371487,
+      "grad_norm": 1.1164593696594238,
+      "learning_rate": 8.04531490015361e-06,
+      "loss": 0.0034,
+      "step": 87900
+    },
+    {
+      "epoch": 16.816357729791708,
+      "grad_norm": 0.2703384757041931,
+      "learning_rate": 7.99731182795699e-06,
+      "loss": 0.003,
+      "step": 88000
+    },
+    {
+      "epoch": 16.816357729791708,
+      "eval_events-synergy/entsum_processed_loss": 0.012502993457019329,
+      "eval_events-synergy/entsum_processed_runtime": 86.9088,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.764,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.697,
+      "step": 88000
+    },
+    {
+      "epoch": 16.835467227211925,
+      "grad_norm": 0.09698379784822464,
+      "learning_rate": 7.949308755760369e-06,
+      "loss": 0.0035,
+      "step": 88100
+    },
+    {
+      "epoch": 16.854576724632143,
+      "grad_norm": 0.10032069683074951,
+      "learning_rate": 7.901305683563747e-06,
+      "loss": 0.0034,
+      "step": 88200
+    },
+    {
+      "epoch": 16.87368622205236,
+      "grad_norm": 0.2772189974784851,
+      "learning_rate": 7.853302611367128e-06,
+      "loss": 0.0032,
+      "step": 88300
+    },
+    {
+      "epoch": 16.892795719472577,
+      "grad_norm": 1.200387716293335,
+      "learning_rate": 7.805299539170508e-06,
+      "loss": 0.004,
+      "step": 88400
+    },
+    {
+      "epoch": 16.911905216892794,
+      "grad_norm": 0.41922566294670105,
+      "learning_rate": 7.757296466973887e-06,
+      "loss": 0.0037,
+      "step": 88500
+    },
+    {
+      "epoch": 16.931014714313015,
+      "grad_norm": 0.030310144647955894,
+      "learning_rate": 7.709293394777266e-06,
+      "loss": 0.0038,
+      "step": 88600
+    },
+    {
+      "epoch": 16.950124211733232,
+      "grad_norm": 0.7613725066184998,
+      "learning_rate": 7.661290322580646e-06,
+      "loss": 0.0027,
+      "step": 88700
+    },
+    {
+      "epoch": 16.96923370915345,
+      "grad_norm": 0.3432716727256775,
+      "learning_rate": 7.613287250384026e-06,
+      "loss": 0.0038,
+      "step": 88800
+    },
+    {
+      "epoch": 16.988343206573667,
+      "grad_norm": 0.07921448349952698,
+      "learning_rate": 7.565284178187404e-06,
+      "loss": 0.0031,
+      "step": 88900
+    },
+    {
+      "epoch": 17.007452703993884,
+      "grad_norm": 1.4262216091156006,
+      "learning_rate": 7.517281105990784e-06,
+      "loss": 0.0031,
+      "step": 89000
+    },
+    {
+      "epoch": 17.007452703993884,
+      "eval_events-synergy/entsum_processed_loss": 0.01249308604747057,
+      "eval_events-synergy/entsum_processed_runtime": 84.734,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.451,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.869,
+      "step": 89000
+    },
+    {
+      "epoch": 17.0265622014141,
+      "grad_norm": 0.5929850339889526,
+      "learning_rate": 7.469278033794164e-06,
+      "loss": 0.0033,
+      "step": 89100
+    },
+    {
+      "epoch": 17.045671698834322,
+      "grad_norm": 0.01337860431522131,
+      "learning_rate": 7.4212749615975425e-06,
+      "loss": 0.0031,
+      "step": 89200
+    },
+    {
+      "epoch": 17.06478119625454,
+      "grad_norm": 0.436662882566452,
+      "learning_rate": 7.373271889400922e-06,
+      "loss": 0.0033,
+      "step": 89300
+    },
+    {
+      "epoch": 17.083890693674757,
+      "grad_norm": 0.031468212604522705,
+      "learning_rate": 7.325268817204302e-06,
+      "loss": 0.0025,
+      "step": 89400
+    },
+    {
+      "epoch": 17.103000191094974,
+      "grad_norm": 0.12954148650169373,
+      "learning_rate": 7.27726574500768e-06,
+      "loss": 0.0029,
+      "step": 89500
+    },
+    {
+      "epoch": 17.12210968851519,
+      "grad_norm": 0.04875854402780533,
+      "learning_rate": 7.22926267281106e-06,
+      "loss": 0.0031,
+      "step": 89600
+    },
+    {
+      "epoch": 17.14121918593541,
+      "grad_norm": 0.20842742919921875,
+      "learning_rate": 7.18125960061444e-06,
+      "loss": 0.0048,
+      "step": 89700
+    },
+    {
+      "epoch": 17.160328683355626,
+      "grad_norm": 0.022663576528429985,
+      "learning_rate": 7.133256528417818e-06,
+      "loss": 0.0052,
+      "step": 89800
+    },
+    {
+      "epoch": 17.179438180775847,
+      "grad_norm": 0.00754079595208168,
+      "learning_rate": 7.085253456221199e-06,
+      "loss": 0.0033,
+      "step": 89900
+    },
+    {
+      "epoch": 17.198547678196064,
+      "grad_norm": 0.06156223267316818,
+      "learning_rate": 7.037250384024578e-06,
+      "loss": 0.0031,
+      "step": 90000
+    },
+    {
+      "epoch": 17.198547678196064,
+      "eval_events-synergy/entsum_processed_loss": 0.012280634604394436,
+      "eval_events-synergy/entsum_processed_runtime": 85.4755,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.212,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.809,
+      "step": 90000
+    },
+    {
+      "epoch": 17.21765717561628,
+      "grad_norm": 0.011858302168548107,
+      "learning_rate": 6.989247311827957e-06,
+      "loss": 0.0032,
+      "step": 90100
+    },
+    {
+      "epoch": 17.2367666730365,
+      "grad_norm": 0.23627856373786926,
+      "learning_rate": 6.941244239631337e-06,
+      "loss": 0.0037,
+      "step": 90200
+    },
+    {
+      "epoch": 17.255876170456716,
+      "grad_norm": 0.6709181666374207,
+      "learning_rate": 6.893241167434716e-06,
+      "loss": 0.0038,
+      "step": 90300
+    },
+    {
+      "epoch": 17.274985667876933,
+      "grad_norm": 0.13345150649547577,
+      "learning_rate": 6.845238095238096e-06,
+      "loss": 0.0032,
+      "step": 90400
+    },
+    {
+      "epoch": 17.294095165297154,
+      "grad_norm": 0.00865347869694233,
+      "learning_rate": 6.7972350230414745e-06,
+      "loss": 0.0033,
+      "step": 90500
+    },
+    {
+      "epoch": 17.31320466271737,
+      "grad_norm": 0.2757847309112549,
+      "learning_rate": 6.749231950844854e-06,
+      "loss": 0.003,
+      "step": 90600
+    },
+    {
+      "epoch": 17.33231416013759,
+      "grad_norm": 0.2280185967683792,
+      "learning_rate": 6.7012288786482346e-06,
+      "loss": 0.0026,
+      "step": 90700
+    },
+    {
+      "epoch": 17.351423657557806,
+      "grad_norm": 0.3780621886253357,
+      "learning_rate": 6.6532258064516125e-06,
+      "loss": 0.0038,
+      "step": 90800
+    },
+    {
+      "epoch": 17.370533154978023,
+      "grad_norm": 0.2748869061470032,
+      "learning_rate": 6.605222734254993e-06,
+      "loss": 0.0029,
+      "step": 90900
+    },
+    {
+      "epoch": 17.38964265239824,
+      "grad_norm": 0.7025009393692017,
+      "learning_rate": 6.5572196620583725e-06,
+      "loss": 0.0033,
+      "step": 91000
+    },
+    {
+      "epoch": 17.38964265239824,
+      "eval_events-synergy/entsum_processed_loss": 0.012524567544460297,
+      "eval_events-synergy/entsum_processed_runtime": 85.033,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.354,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.844,
+      "step": 91000
+    },
+    {
+      "epoch": 17.40875214981846,
+      "grad_norm": 0.07058654725551605,
+      "learning_rate": 6.509216589861751e-06,
+      "loss": 0.0024,
+      "step": 91100
+    },
+    {
+      "epoch": 17.42786164723868,
+      "grad_norm": 0.016350701451301575,
+      "learning_rate": 6.461213517665131e-06,
+      "loss": 0.0032,
+      "step": 91200
+    },
+    {
+      "epoch": 17.446971144658896,
+      "grad_norm": 1.0580590963363647,
+      "learning_rate": 6.41321044546851e-06,
+      "loss": 0.0033,
+      "step": 91300
+    },
+    {
+      "epoch": 17.466080642079113,
+      "grad_norm": 0.029303809627890587,
+      "learning_rate": 6.365207373271889e-06,
+      "loss": 0.0028,
+      "step": 91400
+    },
+    {
+      "epoch": 17.48519013949933,
+      "grad_norm": 0.45363864302635193,
+      "learning_rate": 6.317204301075269e-06,
+      "loss": 0.0046,
+      "step": 91500
+    },
+    {
+      "epoch": 17.504299636919548,
+      "grad_norm": 0.15257449448108673,
+      "learning_rate": 6.269201228878649e-06,
+      "loss": 0.0024,
+      "step": 91600
+    },
+    {
+      "epoch": 17.523409134339765,
+      "grad_norm": 0.4897935688495636,
+      "learning_rate": 6.221198156682028e-06,
+      "loss": 0.0026,
+      "step": 91700
+    },
+    {
+      "epoch": 17.542518631759986,
+      "grad_norm": 0.09276042878627777,
+      "learning_rate": 6.1731950844854075e-06,
+      "loss": 0.0042,
+      "step": 91800
+    },
+    {
+      "epoch": 17.561628129180203,
+      "grad_norm": 0.053013380616903305,
+      "learning_rate": 6.125192012288787e-06,
+      "loss": 0.0034,
+      "step": 91900
+    },
+    {
+      "epoch": 17.58073762660042,
+      "grad_norm": 0.004806125536561012,
+      "learning_rate": 6.077188940092167e-06,
+      "loss": 0.0032,
+      "step": 92000
+    },
+    {
+      "epoch": 17.58073762660042,
+      "eval_events-synergy/entsum_processed_loss": 0.012403319589793682,
+      "eval_events-synergy/entsum_processed_runtime": 84.7299,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.452,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.869,
+      "step": 92000
+    },
+    {
+      "epoch": 17.599847124020638,
+      "grad_norm": 0.5843932032585144,
+      "learning_rate": 6.029185867895545e-06,
+      "loss": 0.004,
+      "step": 92100
+    },
+    {
+      "epoch": 17.618956621440855,
+      "grad_norm": 0.1348479837179184,
+      "learning_rate": 5.981182795698925e-06,
+      "loss": 0.0024,
+      "step": 92200
+    },
+    {
+      "epoch": 17.638066118861072,
+      "grad_norm": 0.7288301587104797,
+      "learning_rate": 5.9331797235023045e-06,
+      "loss": 0.0039,
+      "step": 92300
+    },
+    {
+      "epoch": 17.657175616281293,
+      "grad_norm": 0.018965117633342743,
+      "learning_rate": 5.885176651305683e-06,
+      "loss": 0.0034,
+      "step": 92400
+    },
+    {
+      "epoch": 17.67628511370151,
+      "grad_norm": 0.21183770895004272,
+      "learning_rate": 5.837173579109064e-06,
+      "loss": 0.0029,
+      "step": 92500
+    },
+    {
+      "epoch": 17.695394611121728,
+      "grad_norm": 0.28580865263938904,
+      "learning_rate": 5.7891705069124425e-06,
+      "loss": 0.0036,
+      "step": 92600
+    },
+    {
+      "epoch": 17.714504108541945,
+      "grad_norm": 0.09748672693967819,
+      "learning_rate": 5.741167434715822e-06,
+      "loss": 0.0038,
+      "step": 92700
+    },
+    {
+      "epoch": 17.733613605962162,
+      "grad_norm": 0.03202127292752266,
+      "learning_rate": 5.693164362519202e-06,
+      "loss": 0.0028,
+      "step": 92800
+    },
+    {
+      "epoch": 17.75272310338238,
+      "grad_norm": 0.5783613920211792,
+      "learning_rate": 5.64516129032258e-06,
+      "loss": 0.0025,
+      "step": 92900
+    },
+    {
+      "epoch": 17.7718326008026,
+      "grad_norm": 0.1477188616991043,
+      "learning_rate": 5.597158218125961e-06,
+      "loss": 0.0029,
+      "step": 93000
+    },
+    {
+      "epoch": 17.7718326008026,
+      "eval_events-synergy/entsum_processed_loss": 0.012467662803828716,
+      "eval_events-synergy/entsum_processed_runtime": 86.2702,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.962,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.746,
+      "step": 93000
+    },
+    {
+      "epoch": 17.790942098222818,
+      "grad_norm": 0.029744425788521767,
+      "learning_rate": 5.5491551459293395e-06,
+      "loss": 0.0019,
+      "step": 93100
+    },
+    {
+      "epoch": 17.810051595643035,
+      "grad_norm": 0.006620775908231735,
+      "learning_rate": 5.501152073732719e-06,
+      "loss": 0.004,
+      "step": 93200
+    },
+    {
+      "epoch": 17.829161093063252,
+      "grad_norm": 0.03536098822951317,
+      "learning_rate": 5.453149001536099e-06,
+      "loss": 0.0029,
+      "step": 93300
+    },
+    {
+      "epoch": 17.84827059048347,
+      "grad_norm": 0.18313711881637573,
+      "learning_rate": 5.405145929339478e-06,
+      "loss": 0.0034,
+      "step": 93400
+    },
+    {
+      "epoch": 17.867380087903687,
+      "grad_norm": 0.11048083007335663,
+      "learning_rate": 5.357142857142857e-06,
+      "loss": 0.0032,
+      "step": 93500
+    },
+    {
+      "epoch": 17.886489585323908,
+      "grad_norm": 0.02983025275170803,
+      "learning_rate": 5.309139784946237e-06,
+      "loss": 0.0028,
+      "step": 93600
+    },
+    {
+      "epoch": 17.905599082744125,
+      "grad_norm": 0.050138670951128006,
+      "learning_rate": 5.261136712749616e-06,
+      "loss": 0.0028,
+      "step": 93700
+    },
+    {
+      "epoch": 17.924708580164342,
+      "grad_norm": 0.015463109128177166,
+      "learning_rate": 5.213133640552996e-06,
+      "loss": 0.0023,
+      "step": 93800
+    },
+    {
+      "epoch": 17.94381807758456,
+      "grad_norm": 0.3205445408821106,
+      "learning_rate": 5.165130568356375e-06,
+      "loss": 0.0047,
+      "step": 93900
+    },
+    {
+      "epoch": 17.962927575004777,
+      "grad_norm": 0.02310132421553135,
+      "learning_rate": 5.117127496159754e-06,
+      "loss": 0.0019,
+      "step": 94000
+    },
+    {
+      "epoch": 17.962927575004777,
+      "eval_events-synergy/entsum_processed_loss": 0.012523534707725048,
+      "eval_events-synergy/entsum_processed_runtime": 84.462,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.539,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.891,
+      "step": 94000
+    },
+    {
+      "epoch": 17.982037072424994,
+      "grad_norm": 0.09872489422559738,
+      "learning_rate": 5.0691244239631346e-06,
+      "loss": 0.0027,
+      "step": 94100
+    },
+    {
+      "epoch": 18.00114656984521,
+      "grad_norm": 0.009674746543169022,
+      "learning_rate": 5.021121351766513e-06,
+      "loss": 0.0025,
+      "step": 94200
+    },
+    {
+      "epoch": 18.020256067265432,
+      "grad_norm": 0.1077992394566536,
+      "learning_rate": 4.973118279569893e-06,
+      "loss": 0.003,
+      "step": 94300
+    },
+    {
+      "epoch": 18.03936556468565,
+      "grad_norm": 0.15195026993751526,
+      "learning_rate": 4.9251152073732725e-06,
+      "loss": 0.0035,
+      "step": 94400
+    },
+    {
+      "epoch": 18.058475062105867,
+      "grad_norm": 0.048992399126291275,
+      "learning_rate": 4.877112135176651e-06,
+      "loss": 0.0026,
+      "step": 94500
+    },
+    {
+      "epoch": 18.077584559526084,
+      "grad_norm": 0.0054066977463662624,
+      "learning_rate": 4.829109062980031e-06,
+      "loss": 0.0027,
+      "step": 94600
+    },
+    {
+      "epoch": 18.0966940569463,
+      "grad_norm": 0.07554273307323456,
+      "learning_rate": 4.78110599078341e-06,
+      "loss": 0.0031,
+      "step": 94700
+    },
+    {
+      "epoch": 18.11580355436652,
+      "grad_norm": 0.184183269739151,
+      "learning_rate": 4.73310291858679e-06,
+      "loss": 0.0031,
+      "step": 94800
+    },
+    {
+      "epoch": 18.13491305178674,
+      "grad_norm": 0.5222034454345703,
+      "learning_rate": 4.6850998463901695e-06,
+      "loss": 0.0032,
+      "step": 94900
+    },
+    {
+      "epoch": 18.154022549206957,
+      "grad_norm": 0.07612292468547821,
+      "learning_rate": 4.637096774193548e-06,
+      "loss": 0.0025,
+      "step": 95000
+    },
+    {
+      "epoch": 18.154022549206957,
+      "eval_events-synergy/entsum_processed_loss": 0.012190896086394787,
+      "eval_events-synergy/entsum_processed_runtime": 84.6592,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.475,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.875,
+      "step": 95000
+    },
+    {
+      "epoch": 18.173132046627174,
+      "grad_norm": 0.3161191940307617,
+      "learning_rate": 4.589093701996928e-06,
+      "loss": 0.0028,
+      "step": 95100
+    },
+    {
+      "epoch": 18.19224154404739,
+      "grad_norm": 0.0276503823697567,
+      "learning_rate": 4.5410906298003075e-06,
+      "loss": 0.0026,
+      "step": 95200
+    },
+    {
+      "epoch": 18.21135104146761,
+      "grad_norm": 0.03638790175318718,
+      "learning_rate": 4.493087557603687e-06,
+      "loss": 0.003,
+      "step": 95300
+    },
+    {
+      "epoch": 18.230460538887826,
+      "grad_norm": 0.11877056211233139,
+      "learning_rate": 4.445084485407066e-06,
+      "loss": 0.0036,
+      "step": 95400
+    },
+    {
+      "epoch": 18.249570036308047,
+      "grad_norm": 0.033486805856227875,
+      "learning_rate": 4.397081413210446e-06,
+      "loss": 0.0032,
+      "step": 95500
+    },
+    {
+      "epoch": 18.268679533728264,
+      "grad_norm": 0.07359365373849869,
+      "learning_rate": 4.349078341013825e-06,
+      "loss": 0.0023,
+      "step": 95600
+    },
+    {
+      "epoch": 18.28778903114848,
+      "grad_norm": 0.07991454005241394,
+      "learning_rate": 4.3010752688172045e-06,
+      "loss": 0.0029,
+      "step": 95700
+    },
+    {
+      "epoch": 18.3068985285687,
+      "grad_norm": 0.0038367731031030416,
+      "learning_rate": 4.253072196620584e-06,
+      "loss": 0.0022,
+      "step": 95800
+    },
+    {
+      "epoch": 18.326008025988916,
+      "grad_norm": 0.33883294463157654,
+      "learning_rate": 4.205069124423963e-06,
+      "loss": 0.0035,
+      "step": 95900
+    },
+    {
+      "epoch": 18.345117523409133,
+      "grad_norm": 0.02871253900229931,
+      "learning_rate": 4.157066052227343e-06,
+      "loss": 0.0033,
+      "step": 96000
+    },
+    {
+      "epoch": 18.345117523409133,
+      "eval_events-synergy/entsum_processed_loss": 0.012162734754383564,
+      "eval_events-synergy/entsum_processed_runtime": 86.128,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.006,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.757,
+      "step": 96000
+    },
+    {
+      "epoch": 18.364227020829354,
+      "grad_norm": 0.3818151354789734,
+      "learning_rate": 4.109062980030722e-06,
+      "loss": 0.0024,
+      "step": 96100
+    },
+    {
+      "epoch": 18.38333651824957,
+      "grad_norm": 0.4807966649532318,
+      "learning_rate": 4.061059907834102e-06,
+      "loss": 0.0024,
+      "step": 96200
+    },
+    {
+      "epoch": 18.40244601566979,
+      "grad_norm": 0.5500592589378357,
+      "learning_rate": 4.013056835637481e-06,
+      "loss": 0.004,
+      "step": 96300
+    },
+    {
+      "epoch": 18.421555513090006,
+      "grad_norm": 0.03873921185731888,
+      "learning_rate": 3.96505376344086e-06,
+      "loss": 0.0033,
+      "step": 96400
+    },
+    {
+      "epoch": 18.440665010510223,
+      "grad_norm": 0.003130296478047967,
+      "learning_rate": 3.9170506912442395e-06,
+      "loss": 0.0031,
+      "step": 96500
+    },
+    {
+      "epoch": 18.45977450793044,
+      "grad_norm": 0.04637152701616287,
+      "learning_rate": 3.869047619047619e-06,
+      "loss": 0.0043,
+      "step": 96600
+    },
+    {
+      "epoch": 18.478884005350658,
+      "grad_norm": 0.1211615651845932,
+      "learning_rate": 3.821044546850999e-06,
+      "loss": 0.0032,
+      "step": 96700
+    },
+    {
+      "epoch": 18.49799350277088,
+      "grad_norm": 0.17123758792877197,
+      "learning_rate": 3.7730414746543783e-06,
+      "loss": 0.002,
+      "step": 96800
+    },
+    {
+      "epoch": 18.517103000191096,
+      "grad_norm": 0.2270338535308838,
+      "learning_rate": 3.7250384024577574e-06,
+      "loss": 0.0029,
+      "step": 96900
+    },
+    {
+      "epoch": 18.536212497611313,
+      "grad_norm": 0.07671944797039032,
+      "learning_rate": 3.6770353302611366e-06,
+      "loss": 0.003,
+      "step": 97000
+    },
+    {
+      "epoch": 18.536212497611313,
+      "eval_events-synergy/entsum_processed_loss": 0.012130416929721832,
+      "eval_events-synergy/entsum_processed_runtime": 86.1559,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.998,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.755,
+      "step": 97000
+    },
+    {
+      "epoch": 18.55532199503153,
+      "grad_norm": 0.5913010835647583,
+      "learning_rate": 3.6290322580645166e-06,
+      "loss": 0.0028,
+      "step": 97100
+    },
+    {
+      "epoch": 18.574431492451748,
+      "grad_norm": 1.1253652572631836,
+      "learning_rate": 3.5810291858678958e-06,
+      "loss": 0.0036,
+      "step": 97200
+    },
+    {
+      "epoch": 18.593540989871965,
+      "grad_norm": 0.029553644359111786,
+      "learning_rate": 3.533026113671275e-06,
+      "loss": 0.0034,
+      "step": 97300
+    },
+    {
+      "epoch": 18.612650487292186,
+      "grad_norm": 0.057064156979322433,
+      "learning_rate": 3.4850230414746545e-06,
+      "loss": 0.003,
+      "step": 97400
+    },
+    {
+      "epoch": 18.631759984712403,
+      "grad_norm": 0.08024504035711288,
+      "learning_rate": 3.4370199692780337e-06,
+      "loss": 0.0028,
+      "step": 97500
+    },
+    {
+      "epoch": 18.65086948213262,
+      "grad_norm": 0.17282864451408386,
+      "learning_rate": 3.389016897081413e-06,
+      "loss": 0.0026,
+      "step": 97600
+    },
+    {
+      "epoch": 18.669978979552837,
+      "grad_norm": 0.046736475080251694,
+      "learning_rate": 3.341013824884793e-06,
+      "loss": 0.0031,
+      "step": 97700
+    },
+    {
+      "epoch": 18.689088476973055,
+      "grad_norm": 0.01335865817964077,
+      "learning_rate": 3.293010752688172e-06,
+      "loss": 0.0033,
+      "step": 97800
+    },
+    {
+      "epoch": 18.708197974393272,
+      "grad_norm": 0.24637210369110107,
+      "learning_rate": 3.245007680491552e-06,
+      "loss": 0.0021,
+      "step": 97900
+    },
+    {
+      "epoch": 18.727307471813493,
+      "grad_norm": 0.11811818182468414,
+      "learning_rate": 3.197004608294931e-06,
+      "loss": 0.0022,
+      "step": 98000
+    },
+    {
+      "epoch": 18.727307471813493,
+      "eval_events-synergy/entsum_processed_loss": 0.012472053989768028,
+      "eval_events-synergy/entsum_processed_runtime": 85.9829,
+      "eval_events-synergy/entsum_processed_samples_per_second": 27.052,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.769,
+      "step": 98000
+    },
+    {
+      "epoch": 18.74641696923371,
+      "grad_norm": 0.2450024038553238,
+      "learning_rate": 3.1490015360983104e-06,
+      "loss": 0.003,
+      "step": 98100
+    },
+    {
+      "epoch": 18.765526466653927,
+      "grad_norm": 0.06988529115915298,
+      "learning_rate": 3.1009984639016895e-06,
+      "loss": 0.0029,
+      "step": 98200
+    },
+    {
+      "epoch": 18.784635964074145,
+      "grad_norm": 0.05243882164359093,
+      "learning_rate": 3.052995391705069e-06,
+      "loss": 0.0028,
+      "step": 98300
+    },
+    {
+      "epoch": 18.803745461494362,
+      "grad_norm": 0.005393981467932463,
+      "learning_rate": 3.0049923195084487e-06,
+      "loss": 0.0033,
+      "step": 98400
+    },
+    {
+      "epoch": 18.82285495891458,
+      "grad_norm": 1.3344509601593018,
+      "learning_rate": 2.9569892473118283e-06,
+      "loss": 0.0029,
+      "step": 98500
+    },
+    {
+      "epoch": 18.8419644563348,
+      "grad_norm": 0.0016993408789858222,
+      "learning_rate": 2.9089861751152074e-06,
+      "loss": 0.0017,
+      "step": 98600
+    },
+    {
+      "epoch": 18.861073953755017,
+      "grad_norm": 0.010616108775138855,
+      "learning_rate": 2.860983102918587e-06,
+      "loss": 0.0042,
+      "step": 98700
+    },
+    {
+      "epoch": 18.880183451175235,
+      "grad_norm": 0.01788509450852871,
+      "learning_rate": 2.8129800307219666e-06,
+      "loss": 0.0028,
+      "step": 98800
+    },
+    {
+      "epoch": 18.899292948595452,
+      "grad_norm": 0.07264339178800583,
+      "learning_rate": 2.7649769585253458e-06,
+      "loss": 0.0028,
+      "step": 98900
+    },
+    {
+      "epoch": 18.91840244601567,
+      "grad_norm": 0.6919617652893066,
+      "learning_rate": 2.716973886328725e-06,
+      "loss": 0.0022,
+      "step": 99000
+    },
+    {
+      "epoch": 18.91840244601567,
+      "eval_events-synergy/entsum_processed_loss": 0.012552082538604736,
+      "eval_events-synergy/entsum_processed_runtime": 86.508,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.888,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.728,
+      "step": 99000
+    },
+    {
+      "epoch": 18.937511943435887,
+      "grad_norm": 0.13486118614673615,
+      "learning_rate": 2.6689708141321045e-06,
+      "loss": 0.0025,
+      "step": 99100
+    },
+    {
+      "epoch": 18.956621440856104,
+      "grad_norm": 0.36750730872154236,
+      "learning_rate": 2.620967741935484e-06,
+      "loss": 0.0029,
+      "step": 99200
+    },
+    {
+      "epoch": 18.975730938276325,
+      "grad_norm": 0.01782979629933834,
+      "learning_rate": 2.5729646697388633e-06,
+      "loss": 0.002,
+      "step": 99300
+    },
+    {
+      "epoch": 18.994840435696542,
+      "grad_norm": 0.012732760980725288,
+      "learning_rate": 2.524961597542243e-06,
+      "loss": 0.0019,
+      "step": 99400
+    },
+    {
+      "epoch": 19.01394993311676,
+      "grad_norm": 0.8101866245269775,
+      "learning_rate": 2.4769585253456224e-06,
+      "loss": 0.003,
+      "step": 99500
+    },
+    {
+      "epoch": 19.033059430536976,
+      "grad_norm": 0.06456371396780014,
+      "learning_rate": 2.4289554531490016e-06,
+      "loss": 0.0021,
+      "step": 99600
+    },
+    {
+      "epoch": 19.052168927957194,
+      "grad_norm": 0.18568381667137146,
+      "learning_rate": 2.3809523809523808e-06,
+      "loss": 0.003,
+      "step": 99700
+    },
+    {
+      "epoch": 19.07127842537741,
+      "grad_norm": 0.025632772594690323,
+      "learning_rate": 2.3329493087557603e-06,
+      "loss": 0.0033,
+      "step": 99800
+    },
+    {
+      "epoch": 19.090387922797632,
+      "grad_norm": 0.4020284414291382,
+      "learning_rate": 2.28494623655914e-06,
+      "loss": 0.0024,
+      "step": 99900
+    },
+    {
+      "epoch": 19.10949742021785,
+      "grad_norm": 0.02960309013724327,
+      "learning_rate": 2.2369431643625195e-06,
+      "loss": 0.002,
+      "step": 100000
+    },
+    {
+      "epoch": 19.10949742021785,
+      "eval_events-synergy/entsum_processed_loss": 0.01250072568655014,
+      "eval_events-synergy/entsum_processed_runtime": 86.4178,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.916,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.735,
+      "step": 100000
+    },
+    {
+      "epoch": 19.128606917638066,
+      "grad_norm": 0.0692773312330246,
+      "learning_rate": 2.1889400921658987e-06,
+      "loss": 0.0019,
+      "step": 100100
+    },
+    {
+      "epoch": 19.147716415058284,
+      "grad_norm": 1.2192124128341675,
+      "learning_rate": 2.1409370199692783e-06,
+      "loss": 0.003,
+      "step": 100200
+    },
+    {
+      "epoch": 19.1668259124785,
+      "grad_norm": 0.027012521401047707,
+      "learning_rate": 2.0929339477726574e-06,
+      "loss": 0.0028,
+      "step": 100300
+    },
+    {
+      "epoch": 19.18593540989872,
+      "grad_norm": 1.0921802520751953,
+      "learning_rate": 2.044930875576037e-06,
+      "loss": 0.003,
+      "step": 100400
+    },
+    {
+      "epoch": 19.20504490731894,
+      "grad_norm": 0.047722671180963516,
+      "learning_rate": 1.996927803379416e-06,
+      "loss": 0.0034,
+      "step": 100500
+    },
+    {
+      "epoch": 19.224154404739156,
+      "grad_norm": 0.14935222268104553,
+      "learning_rate": 1.9489247311827958e-06,
+      "loss": 0.0036,
+      "step": 100600
+    },
+    {
+      "epoch": 19.243263902159374,
+      "grad_norm": 0.35229289531707764,
+      "learning_rate": 1.9009216589861754e-06,
+      "loss": 0.0026,
+      "step": 100700
+    },
+    {
+      "epoch": 19.26237339957959,
+      "grad_norm": 0.09162531048059464,
+      "learning_rate": 1.8529185867895545e-06,
+      "loss": 0.0029,
+      "step": 100800
+    },
+    {
+      "epoch": 19.281482896999808,
+      "grad_norm": 0.10979507118463516,
+      "learning_rate": 1.8049155145929339e-06,
+      "loss": 0.0027,
+      "step": 100900
+    },
+    {
+      "epoch": 19.300592394420025,
+      "grad_norm": 0.11148755252361298,
+      "learning_rate": 1.7569124423963135e-06,
+      "loss": 0.0029,
+      "step": 101000
+    },
+    {
+      "epoch": 19.300592394420025,
+      "eval_events-synergy/entsum_processed_loss": 0.012415879406034946,
+      "eval_events-synergy/entsum_processed_runtime": 86.4215,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.915,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.734,
+      "step": 101000
+    },
+    {
+      "epoch": 19.319701891840246,
+      "grad_norm": 0.345984548330307,
+      "learning_rate": 1.708909370199693e-06,
+      "loss": 0.0025,
+      "step": 101100
+    },
+    {
+      "epoch": 19.338811389260464,
+      "grad_norm": 0.08527397364377975,
+      "learning_rate": 1.6609062980030722e-06,
+      "loss": 0.0021,
+      "step": 101200
+    },
+    {
+      "epoch": 19.35792088668068,
+      "grad_norm": 0.4920309782028198,
+      "learning_rate": 1.6129032258064516e-06,
+      "loss": 0.0025,
+      "step": 101300
+    },
+    {
+      "epoch": 19.377030384100898,
+      "grad_norm": 0.28533655405044556,
+      "learning_rate": 1.5649001536098312e-06,
+      "loss": 0.0024,
+      "step": 101400
+    },
+    {
+      "epoch": 19.396139881521115,
+      "grad_norm": 0.008158371783792973,
+      "learning_rate": 1.5168970814132106e-06,
+      "loss": 0.0019,
+      "step": 101500
+    },
+    {
+      "epoch": 19.415249378941333,
+      "grad_norm": 0.00896093063056469,
+      "learning_rate": 1.46889400921659e-06,
+      "loss": 0.0028,
+      "step": 101600
+    },
+    {
+      "epoch": 19.43435887636155,
+      "grad_norm": 0.11068089306354523,
+      "learning_rate": 1.4208909370199693e-06,
+      "loss": 0.0036,
+      "step": 101700
+    },
+    {
+      "epoch": 19.45346837378177,
+      "grad_norm": 0.029200438410043716,
+      "learning_rate": 1.3728878648233489e-06,
+      "loss": 0.0029,
+      "step": 101800
+    },
+    {
+      "epoch": 19.472577871201988,
+      "grad_norm": 0.06840227544307709,
+      "learning_rate": 1.3248847926267283e-06,
+      "loss": 0.0029,
+      "step": 101900
+    },
+    {
+      "epoch": 19.491687368622205,
+      "grad_norm": 0.23425918817520142,
+      "learning_rate": 1.2768817204301076e-06,
+      "loss": 0.0022,
+      "step": 102000
+    },
+    {
+      "epoch": 19.491687368622205,
+      "eval_events-synergy/entsum_processed_loss": 0.012387688271701336,
+      "eval_events-synergy/entsum_processed_runtime": 86.7065,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.826,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.712,
+      "step": 102000
+    },
+    {
+      "epoch": 19.510796866042423,
+      "grad_norm": 0.08164483308792114,
+      "learning_rate": 1.228878648233487e-06,
+      "loss": 0.0026,
+      "step": 102100
+    },
+    {
+      "epoch": 19.52990636346264,
+      "grad_norm": 0.4924757778644562,
+      "learning_rate": 1.1808755760368664e-06,
+      "loss": 0.003,
+      "step": 102200
+    },
+    {
+      "epoch": 19.549015860882857,
+      "grad_norm": 0.2224261462688446,
+      "learning_rate": 1.1328725038402458e-06,
+      "loss": 0.0022,
+      "step": 102300
+    },
+    {
+      "epoch": 19.568125358303078,
+      "grad_norm": 0.08271826058626175,
+      "learning_rate": 1.0848694316436251e-06,
+      "loss": 0.0035,
+      "step": 102400
+    },
+    {
+      "epoch": 19.587234855723295,
+      "grad_norm": 0.09882861375808716,
+      "learning_rate": 1.0368663594470047e-06,
+      "loss": 0.0029,
+      "step": 102500
+    },
+    {
+      "epoch": 19.606344353143513,
+      "grad_norm": 0.054669950157403946,
+      "learning_rate": 9.88863287250384e-07,
+      "loss": 0.0027,
+      "step": 102600
+    },
+    {
+      "epoch": 19.62545385056373,
+      "grad_norm": 1.1425254344940186,
+      "learning_rate": 9.408602150537635e-07,
+      "loss": 0.0034,
+      "step": 102700
+    },
+    {
+      "epoch": 19.644563347983947,
+      "grad_norm": 0.0010029702680185437,
+      "learning_rate": 8.928571428571428e-07,
+      "loss": 0.0023,
+      "step": 102800
+    },
+    {
+      "epoch": 19.663672845404164,
+      "grad_norm": 0.025402860715985298,
+      "learning_rate": 8.448540706605223e-07,
+      "loss": 0.0029,
+      "step": 102900
+    },
+    {
+      "epoch": 19.682782342824385,
+      "grad_norm": 0.2910292446613312,
+      "learning_rate": 7.968509984639017e-07,
+      "loss": 0.0039,
+      "step": 103000
+    },
+    {
+      "epoch": 19.682782342824385,
+      "eval_events-synergy/entsum_processed_loss": 0.012337877415120602,
+      "eval_events-synergy/entsum_processed_runtime": 86.2586,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.965,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.747,
+      "step": 103000
+    },
+    {
+      "epoch": 19.701891840244603,
+      "grad_norm": 0.0035727862268686295,
+      "learning_rate": 7.488479262672811e-07,
+      "loss": 0.0029,
+      "step": 103100
+    },
+    {
+      "epoch": 19.72100133766482,
+      "grad_norm": 0.25086501240730286,
+      "learning_rate": 7.008448540706605e-07,
+      "loss": 0.0028,
+      "step": 103200
+    },
+    {
+      "epoch": 19.740110835085037,
+      "grad_norm": 0.00701789790764451,
+      "learning_rate": 6.528417818740399e-07,
+      "loss": 0.0031,
+      "step": 103300
+    },
+    {
+      "epoch": 19.759220332505254,
+      "grad_norm": 0.8931819200515747,
+      "learning_rate": 6.048387096774194e-07,
+      "loss": 0.0037,
+      "step": 103400
+    },
+    {
+      "epoch": 19.77832982992547,
+      "grad_norm": 0.021326079964637756,
+      "learning_rate": 5.568356374807988e-07,
+      "loss": 0.0023,
+      "step": 103500
+    },
+    {
+      "epoch": 19.797439327345693,
+      "grad_norm": 0.01552259735763073,
+      "learning_rate": 5.088325652841783e-07,
+      "loss": 0.0022,
+      "step": 103600
+    },
+    {
+      "epoch": 19.81654882476591,
+      "grad_norm": 0.026935463771224022,
+      "learning_rate": 4.6082949308755763e-07,
+      "loss": 0.0023,
+      "step": 103700
+    },
+    {
+      "epoch": 19.835658322186127,
+      "grad_norm": 0.013973666355013847,
+      "learning_rate": 4.1282642089093706e-07,
+      "loss": 0.0028,
+      "step": 103800
+    },
+    {
+      "epoch": 19.854767819606344,
+      "grad_norm": 0.001906028133817017,
+      "learning_rate": 3.648233486943165e-07,
+      "loss": 0.0025,
+      "step": 103900
+    },
+    {
+      "epoch": 19.87387731702656,
+      "grad_norm": 0.6687213778495789,
+      "learning_rate": 3.168202764976959e-07,
+      "loss": 0.0028,
+      "step": 104000
+    },
+    {
+      "epoch": 19.87387731702656,
+      "eval_events-synergy/entsum_processed_loss": 0.012370930053293705,
+      "eval_events-synergy/entsum_processed_runtime": 86.4336,
+      "eval_events-synergy/entsum_processed_samples_per_second": 26.911,
+      "eval_events-synergy/entsum_processed_steps_per_second": 6.733,
+      "step": 104000
+    },
+    {
+      "epoch": 19.89298681444678,
+      "grad_norm": 0.019737957045435905,
+      "learning_rate": 2.688172043010753e-07,
+      "loss": 0.0023,
+      "step": 104100
+    },
+    {
+      "epoch": 19.912096311866996,
+      "grad_norm": 0.14768967032432556,
+      "learning_rate": 2.2081413210445468e-07,
+      "loss": 0.0031,
+      "step": 104200
+    },
+    {
+      "epoch": 19.931205809287217,
+      "grad_norm": 0.057144857943058014,
+      "learning_rate": 1.728110599078341e-07,
+      "loss": 0.0029,
+      "step": 104300
+    },
+    {
+      "epoch": 19.950315306707434,
+      "grad_norm": 0.320926696062088,
+      "learning_rate": 1.248079877112135e-07,
+      "loss": 0.0022,
+      "step": 104400
+    },
+    {
+      "epoch": 19.96942480412765,
+      "grad_norm": 0.519444465637207,
+      "learning_rate": 7.680491551459294e-08,
+      "loss": 0.0025,
+      "step": 104500
+    },
+    {
+      "epoch": 19.98853430154787,
+      "grad_norm": 0.07837866246700287,
+      "learning_rate": 2.8801843317972352e-08,
+      "loss": 0.0024,
+      "step": 104600
+    }
+  ],
+  "logging_steps": 100,
+  "max_steps": 104660,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 20,
+  "save_steps": 1000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.647758642249728e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}