End of training

all_results.json

@@ -38,8 +38,8 @@
     "predict_steps_per_second": 6.081,
     "total_flos": 6.747257278287053e+16,
     "train_loss": 0.6624447870913739,
-    "train_runtime": 485.797,
+    "train_runtime": 488.5119,
     "train_samples": 247,
-    "train_samples_per_second": 10.169,
-    "train_steps_per_second": 2.553
+    "train_samples_per_second": 10.112,
+    "train_steps_per_second": 2.538
 }

train_results.json

@@ -2,8 +2,8 @@
     "epoch": 14.0,
     "total_flos": 6.747257278287053e+16,
     "train_loss": 0.6624447870913739,
-    "train_runtime": 485.797,
+    "train_runtime": 488.5119,
     "train_samples": 247,
-    "train_samples_per_second": 10.169,
-    "train_steps_per_second": 2.553
+    "train_samples_per_second": 10.112,
+    "train_steps_per_second": 2.538
 }

trainer_state.json

@@ -18,9 +18,9 @@
       "eval_precision-micro": 0.6509204584925321,
       "eval_recall-macro": 0.24249037703341708,
       "eval_recall-micro": 0.6509204584925321,
-      "eval_runtime": 1.8199,
-      "eval_samples_per_second": 16.484,
-      "eval_steps_per_second": 4.396,
+      "eval_runtime": 1.5286,
+      "eval_samples_per_second": 19.625,
+      "eval_steps_per_second": 5.233,
       "step": 62
     },
     {
@@ -33,9 +33,9 @@
       "eval_precision-micro": 0.7349774227162209,
       "eval_recall-macro": 0.5291052629047126,
       "eval_recall-micro": 0.7349774227162209,
-      "eval_runtime": 4.0287,
-      "eval_samples_per_second": 7.447,
-      "eval_steps_per_second": 1.986,
+      "eval_runtime": 1.6436,
+      "eval_samples_per_second": 18.252,
+      "eval_steps_per_second": 4.867,
       "step": 124
     },
     {
@@ -48,9 +48,9 @@
       "eval_precision-micro": 0.7342827370614797,
       "eval_recall-macro": 0.5381948279596481,
       "eval_recall-micro": 0.7342827370614797,
-      "eval_runtime": 1.4616,
-      "eval_samples_per_second": 20.525,
-      "eval_steps_per_second": 5.473,
+      "eval_runtime": 1.5882,
+      "eval_samples_per_second": 18.889,
+      "eval_steps_per_second": 5.037,
       "step": 186
     },
     {
@@ -63,9 +63,9 @@
       "eval_precision-micro": 0.7502605071205279,
       "eval_recall-macro": 0.5200955397553431,
       "eval_recall-micro": 0.7502605071205279,
-      "eval_runtime": 2.1073,
-      "eval_samples_per_second": 14.236,
-      "eval_steps_per_second": 3.796,
+      "eval_runtime": 1.601,
+      "eval_samples_per_second": 18.738,
+      "eval_steps_per_second": 4.997,
       "step": 248
     },
     {
@@ -78,9 +78,9 @@
       "eval_precision-micro": 0.7641542202153525,
       "eval_recall-macro": 0.5462553107739512,
       "eval_recall-micro": 0.7641542202153525,
-      "eval_runtime": 1.4721,
-      "eval_samples_per_second": 20.379,
-      "eval_steps_per_second": 5.434,
+      "eval_runtime": 2.1379,
+      "eval_samples_per_second": 14.032,
+      "eval_steps_per_second": 3.742,
       "step": 310
     },
     {
@@ -93,9 +93,9 @@
       "eval_precision-micro": 0.774574505036471,
       "eval_recall-macro": 0.5431283891127849,
       "eval_recall-micro": 0.774574505036471,
-      "eval_runtime": 1.6723,
-      "eval_samples_per_second": 17.939,
-      "eval_steps_per_second": 4.784,
+      "eval_runtime": 2.1932,
+      "eval_samples_per_second": 13.679,
+      "eval_steps_per_second": 3.648,
       "step": 372
     },
     {
@@ -108,9 +108,9 @@
       "eval_precision-micro": 0.7811740187565127,
       "eval_recall-macro": 0.5125569147899907,
       "eval_recall-micro": 0.7811740187565127,
-      "eval_runtime": 1.4649,
-      "eval_samples_per_second": 20.479,
-      "eval_steps_per_second": 5.461,
+      "eval_runtime": 1.6371,
+      "eval_samples_per_second": 18.326,
+      "eval_steps_per_second": 4.887,
       "step": 434
     },
     {
@@ -123,9 +123,9 @@
       "eval_precision-micro": 0.7825633900659952,
       "eval_recall-macro": 0.5470775441802167,
       "eval_recall-micro": 0.7825633900659952,
-      "eval_runtime": 1.473,
-      "eval_samples_per_second": 20.367,
-      "eval_steps_per_second": 5.431,
+      "eval_runtime": 1.6786,
+      "eval_samples_per_second": 17.872,
+      "eval_steps_per_second": 4.766,
       "step": 496
     },
     {
@@ -145,9 +145,9 @@
       "eval_precision-micro": 0.7735324765543592,
       "eval_recall-macro": 0.5818475586367124,
       "eval_recall-micro": 0.7735324765543592,
-      "eval_runtime": 1.9237,
-      "eval_samples_per_second": 15.595,
-      "eval_steps_per_second": 4.159,
+      "eval_runtime": 1.5468,
+      "eval_samples_per_second": 19.395,
+      "eval_steps_per_second": 5.172,
       "step": 558
     },
     {
@@ -160,9 +160,9 @@
       "eval_precision-micro": 0.7783952761375478,
       "eval_recall-macro": 0.5655629578179421,
       "eval_recall-micro": 0.7783952761375478,
-      "eval_runtime": 1.4668,
-      "eval_samples_per_second": 20.453,
-      "eval_steps_per_second": 5.454,
+      "eval_runtime": 2.2548,
+      "eval_samples_per_second": 13.305,
+      "eval_steps_per_second": 3.548,
       "step": 620
     },
     {
@@ -175,9 +175,9 @@
       "eval_precision-micro": 0.7829107328933658,
       "eval_recall-macro": 0.5834711775606751,
       "eval_recall-micro": 0.7829107328933658,
-      "eval_runtime": 1.4639,
-      "eval_samples_per_second": 20.493,
-      "eval_steps_per_second": 5.465,
+      "eval_runtime": 2.2566,
+      "eval_samples_per_second": 13.295,
+      "eval_steps_per_second": 3.545,
       "step": 682
     },
     {
@@ -190,9 +190,9 @@
       "eval_precision-micro": 0.7738798193817298,
       "eval_recall-macro": 0.5885123710730829,
       "eval_recall-micro": 0.7738798193817298,
-      "eval_runtime": 1.4843,
-      "eval_samples_per_second": 20.211,
-      "eval_steps_per_second": 5.39,
+      "eval_runtime": 1.5545,
+      "eval_samples_per_second": 19.299,
+      "eval_steps_per_second": 5.146,
       "step": 744
     },
     {
@@ -205,9 +205,9 @@
       "eval_precision-micro": 0.7780479333101772,
       "eval_recall-macro": 0.5748658781079373,
       "eval_recall-micro": 0.7780479333101772,
-      "eval_runtime": 3.173,
-      "eval_samples_per_second": 9.455,
-      "eval_steps_per_second": 2.521,
+      "eval_runtime": 1.5795,
+      "eval_samples_per_second": 18.993,
+      "eval_steps_per_second": 5.065,
       "step": 806
     },
     {
@@ -220,9 +220,9 @@
       "eval_precision-micro": 0.7797846474470302,
       "eval_recall-macro": 0.5868004304340952,
       "eval_recall-micro": 0.7797846474470302,
-      "eval_runtime": 2.1474,
-      "eval_samples_per_second": 13.97,
-      "eval_steps_per_second": 3.725,
+      "eval_runtime": 2.326,
+      "eval_samples_per_second": 12.898,
+      "eval_steps_per_second": 3.439,
       "step": 868
     },
     {
@@ -230,9 +230,9 @@
       "step": 868,
       "total_flos": 6.747257278287053e+16,
       "train_loss": 0.6624447870913739,
-      "train_runtime": 485.797,
-      "train_samples_per_second": 10.169,
-      "train_steps_per_second": 2.553
+      "train_runtime": 488.5119,
+      "train_samples_per_second": 10.112,
+      "train_steps_per_second": 2.538
     }
   ],
   "logging_steps": 500,