End of training

all_results.json

@@ -2,13 +2,13 @@
     "epoch": 15.0,
     "eval_accuracy": 0.9067641496658325,
     "eval_loss": 0.4963526129722595,
-    "eval_runtime": 2.1559,
+    "eval_runtime": 2.1572,
     "eval_samples": 1094,
-    "eval_samples_per_second": 507.439,
-    "eval_steps_per_second": 63.546,
+    "eval_samples_per_second": 507.137,
+    "eval_steps_per_second": 63.508,
     "train_loss": 0.7935866661746093,
-    "train_runtime": 1992.9944,
+    "train_runtime": 2196.4523,
     "train_samples": 10556,
-    "train_samples_per_second": 79.448,
-    "train_steps_per_second": 9.935
+    "train_samples_per_second": 72.089,
+    "train_steps_per_second": 9.015
 }

eval_results.json

@@ -2,8 +2,8 @@
     "epoch": 15.0,
     "eval_accuracy": 0.9067641496658325,
     "eval_loss": 0.4963526129722595,
-    "eval_runtime": 2.1559,
+    "eval_runtime": 2.1572,
     "eval_samples": 1094,
-    "eval_samples_per_second": 507.439,
-    "eval_steps_per_second": 63.546
+    "eval_samples_per_second": 507.137,
+    "eval_steps_per_second": 63.508
 }

runs/Dec15_13-02-52_jbuhmann/events.out.tfevents.1639573387.jbuhmann.27644.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:de261f2d58890161dd734031b5f018dc6b21a500e7a400ae386c561eb3eba708
-size 27441
+oid sha256:81dc8d880fb7bfe8bf848cf48003ee2dfcbe1431c695df973adbccb0603994f2
+size 27801

runs/Dec15_13-02-52_jbuhmann/events.out.tfevents.1639575587.jbuhmann.27644.2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b1002843e24880cba7c24beedfff50921e1c6eec0957b6ad3bc3f049f6dc8e1
+size 369

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 15.0,
     "train_loss": 0.7935866661746093,
-    "train_runtime": 1992.9944,
+    "train_runtime": 2196.4523,
     "train_samples": 10556,
-    "train_samples_per_second": 79.448,
-    "train_steps_per_second": 9.935
+    "train_samples_per_second": 72.089,
+    "train_steps_per_second": 9.015
 }

trainer_state.json

@@ -23,9 +23,9 @@
       "epoch": 1.0,
       "eval_accuracy": 0.4232175648212433,
       "eval_loss": 3.169081211090088,
-      "eval_runtime": 2.1608,
-      "eval_samples_per_second": 506.291,
-      "eval_steps_per_second": 63.402,
+      "eval_runtime": 2.1415,
+      "eval_samples_per_second": 510.858,
+      "eval_steps_per_second": 63.974,
       "step": 1320
     },
     {
@@ -50,9 +50,9 @@
       "epoch": 2.0,
       "eval_accuracy": 0.6435100436210632,
       "eval_loss": 2.04946231842041,
-      "eval_runtime": 2.0332,
-      "eval_samples_per_second": 538.061,
-      "eval_steps_per_second": 67.381,
+      "eval_runtime": 2.1238,
+      "eval_samples_per_second": 515.114,
+      "eval_steps_per_second": 64.507,
       "step": 2640
     },
     {
@@ -71,9 +71,9 @@
       "epoch": 3.0,
       "eval_accuracy": 0.7861060500144958,
       "eval_loss": 1.3566279411315918,
-      "eval_runtime": 2.6595,
-      "eval_samples_per_second": 411.352,
-      "eval_steps_per_second": 51.513,
+      "eval_runtime": 2.0996,
+      "eval_samples_per_second": 521.061,
+      "eval_steps_per_second": 65.252,
       "step": 3960
     },
     {
@@ -98,9 +98,9 @@
       "epoch": 4.0,
       "eval_accuracy": 0.8555758595466614,
       "eval_loss": 0.9564884305000305,
-      "eval_runtime": 2.2538,
-      "eval_samples_per_second": 485.4,
-      "eval_steps_per_second": 60.786,
+      "eval_runtime": 2.1078,
+      "eval_samples_per_second": 519.033,
+      "eval_steps_per_second": 64.998,
       "step": 5280
     },
     {
@@ -125,9 +125,9 @@
       "epoch": 5.0,
       "eval_accuracy": 0.880255937576294,
       "eval_loss": 0.7118942141532898,
-      "eval_runtime": 2.3159,
-      "eval_samples_per_second": 472.394,
-      "eval_steps_per_second": 59.157,
+      "eval_runtime": 2.0744,
+      "eval_samples_per_second": 527.383,
+      "eval_steps_per_second": 66.043,
       "step": 6600
     },
     {
@@ -146,9 +146,9 @@
       "epoch": 6.0,
       "eval_accuracy": 0.8912248611450195,
       "eval_loss": 0.5804843902587891,
-      "eval_runtime": 2.0423,
-      "eval_samples_per_second": 535.658,
-      "eval_steps_per_second": 67.08,
+      "eval_runtime": 2.0875,
+      "eval_samples_per_second": 524.077,
+      "eval_steps_per_second": 65.629,
       "step": 7920
     },
     {
@@ -173,9 +173,9 @@
       "epoch": 7.0,
       "eval_accuracy": 0.8994515538215637,
       "eval_loss": 0.5123510956764221,
-      "eval_runtime": 2.0346,
-      "eval_samples_per_second": 537.708,
-      "eval_steps_per_second": 67.336,
+      "eval_runtime": 2.1263,
+      "eval_samples_per_second": 514.504,
+      "eval_steps_per_second": 64.431,
       "step": 9240
     },
     {
@@ -200,9 +200,9 @@
       "epoch": 8.0,
       "eval_accuracy": 0.9021937847137451,
       "eval_loss": 0.49525973200798035,
-      "eval_runtime": 2.04,
-      "eval_samples_per_second": 536.263,
-      "eval_steps_per_second": 67.155,
+      "eval_runtime": 2.0716,
+      "eval_samples_per_second": 528.088,
+      "eval_steps_per_second": 66.132,
       "step": 10560
     },
     {
@@ -221,9 +221,9 @@
       "epoch": 9.0,
       "eval_accuracy": 0.9021937847137451,
       "eval_loss": 0.486285001039505,
-      "eval_runtime": 2.0348,
-      "eval_samples_per_second": 537.647,
-      "eval_steps_per_second": 67.329,
+      "eval_runtime": 2.0803,
+      "eval_samples_per_second": 525.893,
+      "eval_steps_per_second": 65.857,
       "step": 11880
     },
     {
@@ -248,9 +248,9 @@
       "epoch": 10.0,
       "eval_accuracy": 0.9049360156059265,
       "eval_loss": 0.48929038643836975,
-      "eval_runtime": 2.0542,
-      "eval_samples_per_second": 532.578,
-      "eval_steps_per_second": 66.694,
+      "eval_runtime": 2.0752,
+      "eval_samples_per_second": 527.187,
+      "eval_steps_per_second": 66.019,
       "step": 13200
     },
     {
@@ -275,9 +275,9 @@
       "epoch": 11.0,
       "eval_accuracy": 0.9076782464981079,
       "eval_loss": 0.485994428396225,
-      "eval_runtime": 2.0424,
-      "eval_samples_per_second": 535.632,
-      "eval_steps_per_second": 67.076,
+      "eval_runtime": 2.1075,
+      "eval_samples_per_second": 519.1,
+      "eval_steps_per_second": 65.006,
       "step": 14520
     },
     {
@@ -296,9 +296,9 @@
       "epoch": 12.0,
       "eval_accuracy": 0.9058501124382019,
       "eval_loss": 0.4888700246810913,
-      "eval_runtime": 2.0363,
-      "eval_samples_per_second": 537.252,
-      "eval_steps_per_second": 67.279,
+      "eval_runtime": 2.0826,
+      "eval_samples_per_second": 525.296,
+      "eval_steps_per_second": 65.782,
       "step": 15840
     },
     {
@@ -323,9 +323,9 @@
       "epoch": 13.0,
       "eval_accuracy": 0.9067641496658325,
       "eval_loss": 0.4926171898841858,
-      "eval_runtime": 2.0427,
-      "eval_samples_per_second": 535.562,
-      "eval_steps_per_second": 67.068,
+      "eval_runtime": 2.3107,
+      "eval_samples_per_second": 473.444,
+      "eval_steps_per_second": 59.289,
       "step": 17160
     },
     {
@@ -344,9 +344,9 @@
       "epoch": 14.0,
       "eval_accuracy": 0.9067641496658325,
       "eval_loss": 0.495064914226532,
-      "eval_runtime": 2.0497,
-      "eval_samples_per_second": 533.739,
-      "eval_steps_per_second": 66.839,
+      "eval_runtime": 2.1748,
+      "eval_samples_per_second": 503.039,
+      "eval_steps_per_second": 62.995,
       "step": 18480
     },
     {
@@ -371,9 +371,9 @@
       "epoch": 15.0,
       "eval_accuracy": 0.9067641496658325,
       "eval_loss": 0.4963526129722595,
-      "eval_runtime": 2.0474,
-      "eval_samples_per_second": 534.328,
-      "eval_steps_per_second": 66.913,
+      "eval_runtime": 2.0713,
+      "eval_samples_per_second": 528.178,
+      "eval_steps_per_second": 66.143,
       "step": 19800
     },
     {
@@ -381,9 +381,9 @@
       "step": 19800,
       "total_flos": 5215995096399360.0,
       "train_loss": 0.7935866661746093,
-      "train_runtime": 1992.9944,
-      "train_samples_per_second": 79.448,
-      "train_steps_per_second": 9.935
+      "train_runtime": 2196.4523,
+      "train_samples_per_second": 72.089,
+      "train_steps_per_second": 9.015
     }
   ],
   "max_steps": 19800,