Training in progress, epoch 1, checkpoint

last-checkpoint/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:021b25b4ae02b4c67260af9aaa828a04ad34d01af99d0da4a565af18e7741c65
+oid sha256:f805d53d7dddec9940b8dda0a4fc6f84c5e194e588072b3124f4b98d4dba6b2d
 size 37789864

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1f1f2c45e7d776d5a5ad5cc885d98e86d0067fec7d8d3681868cb22bdd658413
+oid sha256:138083fcab266a12c8bdf751a269df328c6c9ed84e6b49b0dd7314b5b256a7c2
 size 2622266

last-checkpoint/trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 3.0806901454925537,
+  "best_metric": 3.016300916671753,
   "best_model_checkpoint": "dq158/pingusPongus/checkpoint-68803",
   "epoch": 1.0,
   "eval_steps": 500,
@@ -11,823 +11,823 @@
     {
       "epoch": 0.01,
       "learning_rate": 0.0001,
-      "loss": 4.4427,
+      "loss": 4.5647,
       "step": 500
     },
     {
       "epoch": 0.01,
       "learning_rate": 9.999998551451928e-05,
-      "loss": 3.7748,
+      "loss": 3.7798,
       "step": 1000
     },
     {
       "epoch": 0.02,
       "learning_rate": 9.999994205808551e-05,
-      "loss": 3.6469,
+      "loss": 3.6154,
       "step": 1500
     },
     {
       "epoch": 0.03,
       "learning_rate": 9.999986963072388e-05,
-      "loss": 3.6827,
+      "loss": 3.629,
       "step": 2000
     },
     {
       "epoch": 0.04,
       "learning_rate": 9.999976823247632e-05,
-      "loss": 3.6186,
+      "loss": 3.5877,
       "step": 2500
     },
     {
       "epoch": 0.04,
       "learning_rate": 9.999963786340163e-05,
-      "loss": 3.5199,
+      "loss": 3.5401,
       "step": 3000
     },
     {
       "epoch": 0.05,
       "learning_rate": 9.999947852357531e-05,
-      "loss": 3.5221,
+      "loss": 3.5804,
       "step": 3500
     },
     {
       "epoch": 0.06,
       "learning_rate": 9.999929021308971e-05,
-      "loss": 3.469,
+      "loss": 3.4367,
       "step": 4000
     },
     {
       "epoch": 0.07,
       "learning_rate": 9.999907293205393e-05,
-      "loss": 3.5388,
+      "loss": 3.4906,
       "step": 4500
     },
     {
       "epoch": 0.07,
       "learning_rate": 9.999882668059387e-05,
-      "loss": 3.4063,
+      "loss": 3.4193,
       "step": 5000
     },
     {
       "epoch": 0.08,
       "learning_rate": 9.99985514588522e-05,
-      "loss": 3.4314,
+      "loss": 3.4256,
       "step": 5500
     },
     {
       "epoch": 0.09,
       "learning_rate": 9.99982472669884e-05,
-      "loss": 3.4252,
+      "loss": 3.4774,
       "step": 6000
     },
     {
       "epoch": 0.09,
       "learning_rate": 9.999791410517874e-05,
-      "loss": 3.4736,
+      "loss": 3.4629,
       "step": 6500
     },
     {
       "epoch": 0.1,
       "learning_rate": 9.999755197361624e-05,
-      "loss": 3.4011,
+      "loss": 3.3818,
       "step": 7000
     },
     {
       "epoch": 0.11,
       "learning_rate": 9.999716087251072e-05,
-      "loss": 3.4709,
+      "loss": 3.4648,
       "step": 7500
     },
     {
       "epoch": 0.12,
       "learning_rate": 9.99967408020888e-05,
-      "loss": 3.401,
+      "loss": 3.4705,
       "step": 8000
     },
     {
       "epoch": 0.12,
       "learning_rate": 9.999629176259391e-05,
-      "loss": 3.4339,
+      "loss": 3.4356,
       "step": 8500
     },
     {
       "epoch": 0.13,
       "learning_rate": 9.999581375428617e-05,
-      "loss": 3.3573,
+      "loss": 3.4238,
       "step": 9000
     },
     {
       "epoch": 0.14,
       "learning_rate": 9.999530677744258e-05,
-      "loss": 3.364,
+      "loss": 3.4732,
       "step": 9500
     },
     {
       "epoch": 0.15,
       "learning_rate": 9.999477083235691e-05,
-      "loss": 3.4358,
+      "loss": 3.3216,
       "step": 10000
     },
     {
       "epoch": 0.15,
       "learning_rate": 9.999420591933965e-05,
-      "loss": 3.369,
+      "loss": 3.465,
       "step": 10500
     },
     {
       "epoch": 0.16,
       "learning_rate": 9.999361203871817e-05,
-      "loss": 3.3874,
+      "loss": 3.3641,
       "step": 11000
     },
     {
       "epoch": 0.17,
       "learning_rate": 9.999298919083656e-05,
-      "loss": 3.3604,
+      "loss": 3.4407,
       "step": 11500
     },
     {
       "epoch": 0.17,
       "learning_rate": 9.99923373760557e-05,
-      "loss": 3.3134,
+      "loss": 3.3962,
       "step": 12000
     },
     {
       "epoch": 0.18,
       "learning_rate": 9.999165659475324e-05,
-      "loss": 3.3988,
+      "loss": 3.3776,
       "step": 12500
     },
     {
       "epoch": 0.19,
       "learning_rate": 9.999094684732369e-05,
-      "loss": 3.3562,
+      "loss": 3.3157,
       "step": 13000
     },
     {
       "epoch": 0.2,
       "learning_rate": 9.999020813417826e-05,
-      "loss": 3.4156,
+      "loss": 3.2517,
       "step": 13500
     },
     {
       "epoch": 0.2,
       "learning_rate": 9.998944045574499e-05,
-      "loss": 3.2524,
+      "loss": 3.4232,
       "step": 14000
     },
     {
       "epoch": 0.21,
       "learning_rate": 9.998864381246869e-05,
-      "loss": 3.4463,
+      "loss": 3.3539,
       "step": 14500
     },
     {
       "epoch": 0.22,
       "learning_rate": 9.998781820481091e-05,
-      "loss": 3.3492,
+      "loss": 3.2431,
       "step": 15000
     },
     {
       "epoch": 0.23,
       "learning_rate": 9.998696363325009e-05,
-      "loss": 3.4512,
+      "loss": 3.2796,
       "step": 15500
     },
     {
       "epoch": 0.23,
       "learning_rate": 9.998608009828132e-05,
-      "loss": 3.3218,
+      "loss": 3.3468,
       "step": 16000
     },
     {
       "epoch": 0.24,
       "learning_rate": 9.998516760041659e-05,
-      "loss": 3.2985,
+      "loss": 3.3159,
       "step": 16500
     },
     {
       "epoch": 0.25,
       "learning_rate": 9.998422614018456e-05,
-      "loss": 3.3771,
+      "loss": 3.3635,
       "step": 17000
     },
     {
       "epoch": 0.25,
       "learning_rate": 9.998325571813079e-05,
-      "loss": 3.3023,
+      "loss": 3.3708,
       "step": 17500
     },
     {
       "epoch": 0.26,
       "learning_rate": 9.998225633481753e-05,
-      "loss": 3.2226,
+      "loss": 3.3435,
       "step": 18000
     },
     {
       "epoch": 0.27,
       "learning_rate": 9.998122799082386e-05,
-      "loss": 3.3422,
+      "loss": 3.3649,
       "step": 18500
     },
     {
       "epoch": 0.28,
       "learning_rate": 9.998017068674558e-05,
-      "loss": 3.3089,
+      "loss": 3.3519,
       "step": 19000
     },
     {
       "epoch": 0.28,
       "learning_rate": 9.997908442319536e-05,
-      "loss": 3.2337,
+      "loss": 3.3408,
       "step": 19500
     },
     {
       "epoch": 0.29,
       "learning_rate": 9.99779692008026e-05,
-      "loss": 3.3586,
+      "loss": 3.2678,
       "step": 20000
     },
     {
       "epoch": 0.3,
       "learning_rate": 9.997682502021345e-05,
-      "loss": 3.2019,
+      "loss": 3.2453,
       "step": 20500
     },
     {
       "epoch": 0.31,
       "learning_rate": 9.997565188209089e-05,
-      "loss": 3.2937,
+      "loss": 3.2948,
       "step": 21000
     },
     {
       "epoch": 0.31,
       "learning_rate": 9.997444978711465e-05,
-      "loss": 3.4064,
+      "loss": 3.2849,
       "step": 21500
     },
     {
       "epoch": 0.32,
       "learning_rate": 9.997321873598125e-05,
-      "loss": 3.339,
+      "loss": 3.3274,
       "step": 22000
     },
     {
       "epoch": 0.33,
       "learning_rate": 9.9971958729404e-05,
-      "loss": 3.3237,
+      "loss": 3.3084,
       "step": 22500
     },
     {
       "epoch": 0.33,
       "learning_rate": 9.997066976811294e-05,
-      "loss": 3.3782,
+      "loss": 3.3019,
       "step": 23000
     },
     {
       "epoch": 0.34,
       "learning_rate": 9.996935185285495e-05,
-      "loss": 3.336,
+      "loss": 3.2998,
       "step": 23500
     },
     {
       "epoch": 0.35,
       "learning_rate": 9.996800498439362e-05,
-      "loss": 3.1749,
+      "loss": 3.268,
       "step": 24000
     },
     {
       "epoch": 0.36,
       "learning_rate": 9.99666291635094e-05,
-      "loss": 3.252,
+      "loss": 3.248,
       "step": 24500
     },
     {
       "epoch": 0.36,
       "learning_rate": 9.996522439099943e-05,
-      "loss": 3.3544,
+      "loss": 3.2204,
       "step": 25000
     },
     {
       "epoch": 0.37,
       "learning_rate": 9.99637906676777e-05,
-      "loss": 3.2898,
+      "loss": 3.2922,
       "step": 25500
     },
     {
       "epoch": 0.38,
       "learning_rate": 9.996232799437487e-05,
-      "loss": 3.2753,
+      "loss": 3.2716,
       "step": 26000
     },
     {
       "epoch": 0.39,
       "learning_rate": 9.996083637193849e-05,
-      "loss": 3.3539,
+      "loss": 3.3181,
       "step": 26500
     },
     {
       "epoch": 0.39,
       "learning_rate": 9.995931580123284e-05,
-      "loss": 3.2567,
+      "loss": 3.2899,
       "step": 27000
     },
     {
       "epoch": 0.4,
       "learning_rate": 9.995776628313896e-05,
-      "loss": 3.1842,
+      "loss": 3.266,
       "step": 27500
     },
     {
       "epoch": 0.41,
       "learning_rate": 9.995618781855464e-05,
-      "loss": 3.2446,
+      "loss": 3.3644,
       "step": 28000
     },
     {
       "epoch": 0.41,
       "learning_rate": 9.995458040839452e-05,
-      "loss": 3.2132,
+      "loss": 3.2413,
       "step": 28500
     },
     {
       "epoch": 0.42,
       "learning_rate": 9.995294405358993e-05,
-      "loss": 3.2992,
+      "loss": 3.2759,
       "step": 29000
     },
     {
       "epoch": 0.43,
       "learning_rate": 9.995127875508903e-05,
-      "loss": 3.2555,
+      "loss": 3.3065,
       "step": 29500
     },
     {
       "epoch": 0.44,
       "learning_rate": 9.99495845138567e-05,
-      "loss": 3.3704,
+      "loss": 3.3028,
       "step": 30000
     },
     {
       "epoch": 0.44,
       "learning_rate": 9.994786133087464e-05,
-      "loss": 3.2629,
+      "loss": 3.2613,
       "step": 30500
     },
     {
       "epoch": 0.45,
       "learning_rate": 9.994610920714126e-05,
-      "loss": 3.224,
+      "loss": 3.2137,
       "step": 31000
     },
     {
       "epoch": 0.46,
       "learning_rate": 9.994432814367183e-05,
-      "loss": 3.31,
+      "loss": 3.2899,
       "step": 31500
     },
     {
       "epoch": 0.47,
       "learning_rate": 9.99425181414983e-05,
-      "loss": 3.2763,
+      "loss": 3.2877,
       "step": 32000
     },
     {
       "epoch": 0.47,
       "learning_rate": 9.994067920166939e-05,
-      "loss": 3.2862,
+      "loss": 3.2139,
       "step": 32500
     },
     {
       "epoch": 0.48,
       "learning_rate": 9.993881132525067e-05,
-      "loss": 3.3125,
+      "loss": 3.2407,
       "step": 33000
     },
     {
       "epoch": 0.49,
       "learning_rate": 9.993691451332439e-05,
-      "loss": 3.1288,
+      "loss": 3.3226,
       "step": 33500
     },
     {
       "epoch": 0.49,
       "learning_rate": 9.993498876698963e-05,
-      "loss": 3.2958,
+      "loss": 3.2004,
       "step": 34000
     },
     {
       "epoch": 0.5,
       "learning_rate": 9.993303408736217e-05,
-      "loss": 3.1933,
+      "loss": 3.2773,
       "step": 34500
     },
     {
       "epoch": 0.51,
       "learning_rate": 9.993105047557461e-05,
-      "loss": 3.2504,
+      "loss": 3.2183,
       "step": 35000
     },
     {
       "epoch": 0.52,
       "learning_rate": 9.992903793277628e-05,
-      "loss": 3.3293,
+      "loss": 3.2648,
       "step": 35500
     },
     {
       "epoch": 0.52,
       "learning_rate": 9.99269964601333e-05,
-      "loss": 3.1969,
+      "loss": 3.1963,
       "step": 36000
     },
     {
       "epoch": 0.53,
       "learning_rate": 9.992492605882853e-05,
-      "loss": 3.2087,
+      "loss": 3.2825,
       "step": 36500
     },
     {
       "epoch": 0.54,
       "learning_rate": 9.99228267300616e-05,
-      "loss": 3.3125,
+      "loss": 3.2658,
       "step": 37000
     },
     {
       "epoch": 0.55,
       "learning_rate": 9.992069847504891e-05,
-      "loss": 3.2677,
+      "loss": 3.1058,
       "step": 37500
     },
     {
       "epoch": 0.55,
       "learning_rate": 9.99185412950236e-05,
-      "loss": 3.2132,
+      "loss": 3.1375,
       "step": 38000
     },
     {
       "epoch": 0.56,
       "learning_rate": 9.991635519123559e-05,
-      "loss": 3.2534,
+      "loss": 3.2075,
       "step": 38500
     },
     {
       "epoch": 0.57,
       "learning_rate": 9.991414016495155e-05,
-      "loss": 3.1735,
+      "loss": 3.2228,
       "step": 39000
     },
     {
       "epoch": 0.57,
       "learning_rate": 9.99118962174549e-05,
-      "loss": 3.2422,
+      "loss": 3.2004,
       "step": 39500
     },
     {
       "epoch": 0.58,
       "learning_rate": 9.990962335004584e-05,
-      "loss": 3.1625,
+      "loss": 3.1924,
       "step": 40000
     },
     {
       "epoch": 0.59,
       "learning_rate": 9.99073215640413e-05,
-      "loss": 3.1163,
+      "loss": 3.2956,
       "step": 40500
     },
     {
       "epoch": 0.6,
       "learning_rate": 9.990499086077498e-05,
-      "loss": 3.1521,
+      "loss": 3.2698,
       "step": 41000
     },
     {
       "epoch": 0.6,
       "learning_rate": 9.990263124159736e-05,
-      "loss": 3.2036,
+      "loss": 3.2863,
       "step": 41500
     },
     {
       "epoch": 0.61,
       "learning_rate": 9.990024270787561e-05,
-      "loss": 3.181,
+      "loss": 3.3611,
       "step": 42000
     },
     {
       "epoch": 0.62,
       "learning_rate": 9.989782526099372e-05,
-      "loss": 3.1672,
+      "loss": 3.1691,
       "step": 42500
     },
     {
       "epoch": 0.62,
       "learning_rate": 9.989537890235238e-05,
-      "loss": 3.2336,
+      "loss": 3.1085,
       "step": 43000
     },
     {
       "epoch": 0.63,
       "learning_rate": 9.989290363336908e-05,
-      "loss": 3.1455,
+      "loss": 3.1825,
       "step": 43500
     },
     {
       "epoch": 0.64,
       "learning_rate": 9.989039945547803e-05,
-      "loss": 3.1859,
+      "loss": 3.2333,
       "step": 44000
     },
     {
       "epoch": 0.65,
       "learning_rate": 9.98878663701302e-05,
-      "loss": 3.1396,
+      "loss": 3.2635,
       "step": 44500
     },
     {
       "epoch": 0.65,
       "learning_rate": 9.988530437879333e-05,
-      "loss": 3.2585,
+      "loss": 3.2907,
       "step": 45000
     },
     {
       "epoch": 0.66,
       "learning_rate": 9.988271348295184e-05,
-      "loss": 3.2201,
+      "loss": 3.1334,
       "step": 45500
     },
     {
       "epoch": 0.67,
       "learning_rate": 9.988009368410698e-05,
-      "loss": 3.2758,
+      "loss": 3.2239,
       "step": 46000
     },
     {
       "epoch": 0.68,
       "learning_rate": 9.98774449837767e-05,
-      "loss": 3.1742,
+      "loss": 3.1904,
       "step": 46500
     },
     {
       "epoch": 0.68,
       "learning_rate": 9.987476738349571e-05,
-      "loss": 3.3212,
+      "loss": 3.2781,
       "step": 47000
     },
     {
       "epoch": 0.69,
       "learning_rate": 9.987206088481545e-05,
-      "loss": 3.1915,
+      "loss": 3.19,
       "step": 47500
     },
     {
       "epoch": 0.7,
       "learning_rate": 9.986932548930414e-05,
-      "loss": 3.1608,
+      "loss": 3.2235,
       "step": 48000
     },
     {
       "epoch": 0.7,
       "learning_rate": 9.986656119854672e-05,
-      "loss": 3.217,
+      "loss": 3.2302,
       "step": 48500
     },
     {
       "epoch": 0.71,
       "learning_rate": 9.986376801414485e-05,
-      "loss": 3.1989,
+      "loss": 3.289,
       "step": 49000
     },
     {
       "epoch": 0.72,
       "learning_rate": 9.986094593771699e-05,
-      "loss": 3.3067,
+      "loss": 3.2874,
       "step": 49500
     },
     {
       "epoch": 0.73,
       "learning_rate": 9.985809497089827e-05,
-      "loss": 3.2195,
+      "loss": 3.144,
       "step": 50000
     },
     {
       "epoch": 0.73,
       "learning_rate": 9.985521511534062e-05,
-      "loss": 3.148,
+      "loss": 3.1967,
       "step": 50500
     },
     {
       "epoch": 0.74,
       "learning_rate": 9.985230637271266e-05,
-      "loss": 3.1987,
+      "loss": 3.248,
       "step": 51000
     },
     {
       "epoch": 0.75,
       "learning_rate": 9.984936874469979e-05,
-      "loss": 3.1153,
+      "loss": 3.1933,
       "step": 51500
     },
     {
       "epoch": 0.76,
       "learning_rate": 9.984640223300413e-05,
-      "loss": 3.2841,
+      "loss": 3.2108,
       "step": 52000
     },
     {
       "epoch": 0.76,
       "learning_rate": 9.98434068393445e-05,
-      "loss": 3.2033,
+      "loss": 3.0986,
       "step": 52500
     },
     {
       "epoch": 0.77,
       "learning_rate": 9.984038256545653e-05,
-      "loss": 3.2102,
+      "loss": 3.2016,
       "step": 53000
     },
     {
       "epoch": 0.78,
       "learning_rate": 9.983732941309253e-05,
-      "loss": 3.1817,
+      "loss": 3.1967,
       "step": 53500
     },
     {
       "epoch": 0.78,
       "learning_rate": 9.983424738402156e-05,
-      "loss": 3.1485,
+      "loss": 3.2803,
       "step": 54000
     },
     {
       "epoch": 0.79,
       "learning_rate": 9.98311364800294e-05,
-      "loss": 3.2417,
+      "loss": 3.2952,
       "step": 54500
     },
     {
       "epoch": 0.8,
       "learning_rate": 9.982799670291857e-05,
-      "loss": 3.2174,
+      "loss": 3.18,
       "step": 55000
     },
     {
       "epoch": 0.81,
       "learning_rate": 9.98248280545083e-05,
-      "loss": 3.2862,
+      "loss": 3.1768,
       "step": 55500
     },
     {
       "epoch": 0.81,
       "learning_rate": 9.982163053663459e-05,
-      "loss": 3.201,
+      "loss": 3.1986,
       "step": 56000
     },
     {
       "epoch": 0.82,
       "learning_rate": 9.981840415115014e-05,
-      "loss": 3.3873,
+      "loss": 3.2241,
       "step": 56500
     },
     {
       "epoch": 0.83,
       "learning_rate": 9.981514889992436e-05,
-      "loss": 3.1844,
+      "loss": 3.2754,
       "step": 57000
     },
     {
       "epoch": 0.84,
       "learning_rate": 9.981186478484344e-05,
-      "loss": 3.1807,
+      "loss": 3.2162,
       "step": 57500
     },
     {
       "epoch": 0.84,
       "learning_rate": 9.980855180781021e-05,
-      "loss": 3.2758,
+      "loss": 3.2659,
       "step": 58000
     },
     {
       "epoch": 0.85,
       "learning_rate": 9.980520997074432e-05,
-      "loss": 3.1406,
+      "loss": 3.172,
       "step": 58500
     },
     {
       "epoch": 0.86,
       "learning_rate": 9.980183927558207e-05,
-      "loss": 3.1607,
+      "loss": 3.3035,
       "step": 59000
     },
     {
       "epoch": 0.86,
       "learning_rate": 9.97984397242765e-05,
-      "loss": 3.1829,
+      "loss": 3.1746,
       "step": 59500
     },
     {
       "epoch": 0.87,
       "learning_rate": 9.979501131879741e-05,
-      "loss": 3.2238,
+      "loss": 3.1678,
       "step": 60000
     },
     {
       "epoch": 0.88,
       "learning_rate": 9.979155406113124e-05,
-      "loss": 3.2348,
+      "loss": 3.162,
       "step": 60500
     },
     {
       "epoch": 0.89,
       "learning_rate": 9.978806795328121e-05,
-      "loss": 3.2933,
+      "loss": 3.3041,
       "step": 61000
     },
     {
       "epoch": 0.89,
       "learning_rate": 9.978455299726726e-05,
-      "loss": 3.2051,
+      "loss": 3.2086,
       "step": 61500
     },
     {
       "epoch": 0.9,
       "learning_rate": 9.978100919512598e-05,
-      "loss": 3.1736,
+      "loss": 3.1832,
       "step": 62000
     },
     {
       "epoch": 0.91,
       "learning_rate": 9.977743654891077e-05,
-      "loss": 3.173,
+      "loss": 3.1839,
       "step": 62500
     },
     {
       "epoch": 0.92,
       "learning_rate": 9.977383506069164e-05,
-      "loss": 3.2732,
+      "loss": 3.123,
       "step": 63000
     },
     {
       "epoch": 0.92,
       "learning_rate": 9.977020473255539e-05,
-      "loss": 3.2447,
+      "loss": 3.2868,
       "step": 63500
     },
     {
       "epoch": 0.93,
       "learning_rate": 9.976654556660548e-05,
-      "loss": 3.2526,
+      "loss": 3.1488,
       "step": 64000
     },
     {
       "epoch": 0.94,
       "learning_rate": 9.976285756496211e-05,
-      "loss": 3.1814,
+      "loss": 3.2116,
       "step": 64500
     },
     {
       "epoch": 0.94,
       "learning_rate": 9.97591407297622e-05,
-      "loss": 3.1533,
+      "loss": 3.2233,
       "step": 65000
     },
     {
       "epoch": 0.95,
       "learning_rate": 9.975539506315933e-05,
-      "loss": 3.0491,
+      "loss": 3.1018,
       "step": 65500
     },
     {
       "epoch": 0.96,
       "learning_rate": 9.975162056732385e-05,
-      "loss": 3.1918,
+      "loss": 3.1404,
       "step": 66000
     },
     {
       "epoch": 0.97,
       "learning_rate": 9.974781724444272e-05,
-      "loss": 3.1918,
+      "loss": 3.1984,
       "step": 66500
     },
     {
       "epoch": 0.97,
       "learning_rate": 9.974398509671969e-05,
-      "loss": 3.1769,
+      "loss": 3.2421,
       "step": 67000
     },
     {
       "epoch": 0.98,
       "learning_rate": 9.974012412637517e-05,
-      "loss": 3.196,
+      "loss": 3.1499,
       "step": 67500
     },
     {
       "epoch": 0.99,
       "learning_rate": 9.97362343356463e-05,
-      "loss": 3.1799,
+      "loss": 3.174,
       "step": 68000
     },
     {
       "epoch": 1.0,
       "learning_rate": 9.973231572678686e-05,
-      "loss": 3.2516,
+      "loss": 3.1816,
       "step": 68500
     },
     {
@@ -835,7 +835,7 @@
       "eval_bleu": 1.0,
       "eval_brevity_penalty": 1.0,
       "eval_length_ratio": 1.0,
-      "eval_loss": 3.0806901454925537,
+      "eval_loss": 3.016300916671753,
       "eval_precisions": [
         1.0,
         1.0,
@@ -843,9 +843,9 @@
         1.0
       ],
       "eval_reference_length": 7828480,
-      "eval_runtime": 21022.3759,
-      "eval_samples_per_second": 0.727,
-      "eval_steps_per_second": 0.364,
+      "eval_runtime": 20103.3013,
+      "eval_samples_per_second": 0.761,
+      "eval_steps_per_second": 0.38,
       "eval_translation_length": 7828480,
       "step": 68803
     }