End of training

all_results.json

@@ -1,20 +1,20 @@
 {
     "epoch": 99.98,
-    "eval_exact_match": 0.7570469798657719,
-    "eval_loss": 0.10638269782066345,
-    "eval_runtime": 97.9358,
+    "eval_exact_match": 0.7445190156599553,
+    "eval_loss": 0.12020513415336609,
+    "eval_runtime": 104.6036,
     "eval_samples": 2235,
-    "eval_samples_per_second": 22.821,
-    "eval_steps_per_second": 1.43,
-    "predict_exact_match": 0.7731418148654811,
-    "predict_loss": 0.09263205528259277,
-    "predict_runtime": 184.8274,
+    "eval_samples_per_second": 21.366,
+    "eval_steps_per_second": 1.338,
+    "predict_exact_match": 0.7578659370725034,
+    "predict_loss": 0.10240133106708527,
+    "predict_runtime": 205.505,
     "predict_samples": 4386,
-    "predict_samples_per_second": 23.73,
-    "predict_steps_per_second": 1.488,
-    "train_loss": 0.23219555624326071,
-    "train_runtime": 13657.0516,
+    "predict_samples_per_second": 21.343,
+    "predict_steps_per_second": 1.338,
+    "train_loss": 0.22847254665692648,
+    "train_runtime": 13959.7055,
     "train_samples": 15667,
-    "train_samples_per_second": 112.469,
-    "train_steps_per_second": 0.22
+    "train_samples_per_second": 110.031,
+    "train_steps_per_second": 0.215
 }

eval_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 99.98,
-    "eval_exact_match": 0.7570469798657719,
-    "eval_loss": 0.10638269782066345,
-    "eval_runtime": 97.9358,
+    "eval_exact_match": 0.7445190156599553,
+    "eval_loss": 0.12020513415336609,
+    "eval_runtime": 104.6036,
     "eval_samples": 2235,
-    "eval_samples_per_second": 22.821,
-    "eval_steps_per_second": 1.43
+    "eval_samples_per_second": 21.366,
+    "eval_steps_per_second": 1.338
 }

predict_results.json

@@ -1,8 +1,8 @@
 {
-    "predict_exact_match": 0.7731418148654811,
-    "predict_loss": 0.09263205528259277,
-    "predict_runtime": 184.8274,
+    "predict_exact_match": 0.7578659370725034,
+    "predict_loss": 0.10240133106708527,
+    "predict_runtime": 205.505,
     "predict_samples": 4386,
-    "predict_samples_per_second": 23.73,
-    "predict_steps_per_second": 1.488
+    "predict_samples_per_second": 21.343,
+    "predict_steps_per_second": 1.338
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 99.98,
-    "train_loss": 0.23219555624326071,
-    "train_runtime": 13657.0516,
+    "train_loss": 0.22847254665692648,
+    "train_runtime": 13959.7055,
     "train_samples": 15667,
-    "train_samples_per_second": 112.469,
-    "train_steps_per_second": 0.22
+    "train_samples_per_second": 110.031,
+    "train_steps_per_second": 0.215
 }

trainer_state.json

@@ -1,6 +1,6 @@
 {
-  "best_metric": 0.10638269782066345,
-  "best_model_checkpoint": "/data/wheld3/mt5-small-pointer-mtop/checkpoint-1400",
+  "best_metric": 0.12020513415336609,
+  "best_model_checkpoint": "/data/wheld3/mt5-small-pointer-mtop/checkpoint-1200",
   "epoch": 99.9795918367347,
   "global_step": 3000,
   "is_hyper_param_search": false,
@@ -10,236 +10,236 @@
     {
       "epoch": 6.65,
       "learning_rate": 0.0009333333333333333,
-      "loss": 2.0944,
+      "loss": 2.1451,
       "step": 200
     },
     {
       "epoch": 6.65,
-      "eval_exact_match": 0.0026845637583892616,
-      "eval_loss": 0.6548439264297485,
-      "eval_runtime": 44.8886,
-      "eval_samples_per_second": 49.79,
-      "eval_steps_per_second": 3.119,
+      "eval_exact_match": 0.013422818791946308,
+      "eval_loss": 0.5965703725814819,
+      "eval_runtime": 44.8708,
+      "eval_samples_per_second": 49.81,
+      "eval_steps_per_second": 3.12,
       "step": 200
     },
     {
       "epoch": 13.33,
       "learning_rate": 0.0008666666666666667,
-      "loss": 0.5307,
+      "loss": 0.4695,
       "step": 400
     },
     {
       "epoch": 13.33,
-      "eval_exact_match": 0.2778523489932886,
-      "eval_loss": 0.24562813341617584,
-      "eval_runtime": 45.1956,
-      "eval_samples_per_second": 49.452,
-      "eval_steps_per_second": 3.098,
+      "eval_exact_match": 0.29977628635346754,
+      "eval_loss": 0.22643320262432098,
+      "eval_runtime": 44.5288,
+      "eval_samples_per_second": 50.192,
+      "eval_steps_per_second": 3.144,
       "step": 400
     },
     {
       "epoch": 19.98,
       "learning_rate": 0.0008,
-      "loss": 0.2388,
+      "loss": 0.2229,
       "step": 600
     },
     {
       "epoch": 19.98,
-      "eval_exact_match": 0.4559284116331096,
-      "eval_loss": 0.14859849214553833,
-      "eval_runtime": 45.6558,
-      "eval_samples_per_second": 48.953,
-      "eval_steps_per_second": 3.066,
+      "eval_exact_match": 0.46487695749440716,
+      "eval_loss": 0.14458051323890686,
+      "eval_runtime": 47.7042,
+      "eval_samples_per_second": 46.851,
+      "eval_steps_per_second": 2.935,
       "step": 600
     },
     {
       "epoch": 26.65,
       "learning_rate": 0.0007333333333333333,
-      "loss": 0.1459,
+      "loss": 0.1389,
       "step": 800
     },
     {
       "epoch": 26.65,
-      "eval_exact_match": 0.5042505592841163,
-      "eval_loss": 0.11904484033584595,
-      "eval_runtime": 54.895,
-      "eval_samples_per_second": 40.714,
-      "eval_steps_per_second": 2.55,
+      "eval_exact_match": 0.5154362416107383,
+      "eval_loss": 0.12272302061319351,
+      "eval_runtime": 44.4647,
+      "eval_samples_per_second": 50.265,
+      "eval_steps_per_second": 3.149,
       "step": 800
     },
     {
       "epoch": 33.33,
       "learning_rate": 0.0006666666666666666,
-      "loss": 0.1011,
+      "loss": 0.097,
       "step": 1000
     },
     {
       "epoch": 33.33,
-      "eval_exact_match": 0.5230425055928412,
-      "eval_loss": 0.11172914505004883,
-      "eval_runtime": 53.9217,
-      "eval_samples_per_second": 41.449,
-      "eval_steps_per_second": 2.596,
+      "eval_exact_match": 0.5221476510067115,
+      "eval_loss": 0.12126713246107101,
+      "eval_runtime": 46.4733,
+      "eval_samples_per_second": 48.092,
+      "eval_steps_per_second": 3.012,
       "step": 1000
     },
     {
       "epoch": 39.98,
       "learning_rate": 0.0006,
-      "loss": 0.0774,
+      "loss": 0.0724,
       "step": 1200
     },
     {
       "epoch": 39.98,
-      "eval_exact_match": 0.5373601789709173,
-      "eval_loss": 0.10843723267316818,
-      "eval_runtime": 54.3522,
-      "eval_samples_per_second": 41.121,
-      "eval_steps_per_second": 2.576,
+      "eval_exact_match": 0.5364653243847874,
+      "eval_loss": 0.12020513415336609,
+      "eval_runtime": 44.6036,
+      "eval_samples_per_second": 50.108,
+      "eval_steps_per_second": 3.139,
       "step": 1200
     },
     {
       "epoch": 46.65,
       "learning_rate": 0.0005333333333333334,
-      "loss": 0.0598,
+      "loss": 0.0562,
       "step": 1400
     },
     {
       "epoch": 46.65,
-      "eval_exact_match": 0.5404921700223714,
-      "eval_loss": 0.10638269782066345,
-      "eval_runtime": 44.8128,
-      "eval_samples_per_second": 49.874,
+      "eval_exact_match": 0.5436241610738255,
+      "eval_loss": 0.12071848660707474,
+      "eval_runtime": 44.8086,
+      "eval_samples_per_second": 49.879,
       "eval_steps_per_second": 3.124,
       "step": 1400
     },
     {
       "epoch": 53.33,
       "learning_rate": 0.00046666666666666666,
-      "loss": 0.0478,
+      "loss": 0.0457,
       "step": 1600
     },
     {
       "epoch": 53.33,
-      "eval_exact_match": 0.545413870246085,
-      "eval_loss": 0.11469161510467529,
-      "eval_runtime": 45.8513,
-      "eval_samples_per_second": 48.745,
-      "eval_steps_per_second": 3.053,
+      "eval_exact_match": 0.5440715883668904,
+      "eval_loss": 0.12402455508708954,
+      "eval_runtime": 44.3425,
+      "eval_samples_per_second": 50.403,
+      "eval_steps_per_second": 3.157,
       "step": 1600
     },
     {
       "epoch": 59.98,
       "learning_rate": 0.0004,
-      "loss": 0.0397,
+      "loss": 0.0399,
       "step": 1800
     },
     {
       "epoch": 59.98,
-      "eval_exact_match": 0.5472035794183445,
-      "eval_loss": 0.11389175802469254,
-      "eval_runtime": 44.8546,
-      "eval_samples_per_second": 49.828,
-      "eval_steps_per_second": 3.121,
+      "eval_exact_match": 0.5440715883668904,
+      "eval_loss": 0.13486850261688232,
+      "eval_runtime": 44.9861,
+      "eval_samples_per_second": 49.682,
+      "eval_steps_per_second": 3.112,
       "step": 1800
     },
     {
       "epoch": 66.65,
       "learning_rate": 0.0003333333333333333,
-      "loss": 0.0337,
+      "loss": 0.0317,
       "step": 2000
     },
     {
       "epoch": 66.65,
-      "eval_exact_match": 0.5480984340044742,
-      "eval_loss": 0.117930106818676,
-      "eval_runtime": 47.7426,
-      "eval_samples_per_second": 46.814,
-      "eval_steps_per_second": 2.932,
+      "eval_exact_match": 0.5476510067114094,
+      "eval_loss": 0.13694295287132263,
+      "eval_runtime": 45.1401,
+      "eval_samples_per_second": 49.513,
+      "eval_steps_per_second": 3.101,
       "step": 2000
     },
     {
       "epoch": 73.33,
       "learning_rate": 0.0002666666666666667,
-      "loss": 0.0286,
+      "loss": 0.0271,
       "step": 2200
     },
     {
       "epoch": 73.33,
-      "eval_exact_match": 0.5498881431767337,
-      "eval_loss": 0.12433456629514694,
-      "eval_runtime": 45.3848,
-      "eval_samples_per_second": 49.246,
-      "eval_steps_per_second": 3.085,
+      "eval_exact_match": 0.548993288590604,
+      "eval_loss": 0.1409003585577011,
+      "eval_runtime": 44.2099,
+      "eval_samples_per_second": 50.554,
+      "eval_steps_per_second": 3.167,
       "step": 2200
     },
     {
       "epoch": 79.98,
       "learning_rate": 0.0002,
-      "loss": 0.0251,
+      "loss": 0.0237,
       "step": 2400
     },
     {
       "epoch": 79.98,
-      "eval_exact_match": 0.5480984340044742,
-      "eval_loss": 0.12593261897563934,
-      "eval_runtime": 45.7566,
-      "eval_samples_per_second": 48.845,
-      "eval_steps_per_second": 3.06,
+      "eval_exact_match": 0.5539149888143177,
+      "eval_loss": 0.14622744917869568,
+      "eval_runtime": 45.1454,
+      "eval_samples_per_second": 49.507,
+      "eval_steps_per_second": 3.101,
       "step": 2400
     },
     {
       "epoch": 86.65,
       "learning_rate": 0.00013333333333333334,
-      "loss": 0.0218,
+      "loss": 0.0207,
       "step": 2600
     },
     {
       "epoch": 86.65,
-      "eval_exact_match": 0.5503355704697986,
-      "eval_loss": 0.12759922444820404,
-      "eval_runtime": 45.5648,
-      "eval_samples_per_second": 49.051,
-      "eval_steps_per_second": 3.073,
+      "eval_exact_match": 0.5516778523489932,
+      "eval_loss": 0.14700742065906525,
+      "eval_runtime": 44.6922,
+      "eval_samples_per_second": 50.009,
+      "eval_steps_per_second": 3.133,
       "step": 2600
     },
     {
       "epoch": 93.33,
       "learning_rate": 6.666666666666667e-05,
-      "loss": 0.0197,
+      "loss": 0.0188,
       "step": 2800
     },
     {
       "epoch": 93.33,
-      "eval_exact_match": 0.5503355704697986,
-      "eval_loss": 0.13092631101608276,
-      "eval_runtime": 45.4393,
-      "eval_samples_per_second": 49.186,
-      "eval_steps_per_second": 3.081,
+      "eval_exact_match": 0.5507829977628635,
+      "eval_loss": 0.150468111038208,
+      "eval_runtime": 44.5533,
+      "eval_samples_per_second": 50.165,
+      "eval_steps_per_second": 3.142,
       "step": 2800
     },
     {
       "epoch": 99.98,
       "learning_rate": 0.0,
-      "loss": 0.0184,
+      "loss": 0.0174,
       "step": 3000
     },
     {
       "epoch": 99.98,
-      "eval_exact_match": 0.5503355704697986,
-      "eval_loss": 0.1317095011472702,
-      "eval_runtime": 45.2758,
-      "eval_samples_per_second": 49.364,
-      "eval_steps_per_second": 3.092,
+      "eval_exact_match": 0.5512304250559285,
+      "eval_loss": 0.15047426521778107,
+      "eval_runtime": 46.3824,
+      "eval_samples_per_second": 48.186,
+      "eval_steps_per_second": 3.018,
       "step": 3000
     },
     {
       "epoch": 99.98,
       "step": 3000,
       "total_flos": 3.831174875254272e+16,
-      "train_loss": 0.23219555624326071,
-      "train_runtime": 13657.0516,
-      "train_samples_per_second": 112.469,
-      "train_steps_per_second": 0.22
+      "train_loss": 0.22847254665692648,
+      "train_runtime": 13959.7055,
+      "train_samples_per_second": 110.031,
+      "train_steps_per_second": 0.215
     }
   ],
   "max_steps": 3000,