Training in progress, step 15462, checkpoint

last-checkpoint/trainer_state.json

@@ -25,9 +25,9 @@
     {
       "epoch": 0.19402405898331393,
       "eval_loss": 1.1814582347869873,
-      "eval_runtime": 49.4809,
-      "eval_samples_per_second": 104.747,
-      "eval_steps_per_second": 3.274,
+      "eval_runtime": 49.6691,
+      "eval_samples_per_second": 104.351,
+      "eval_steps_per_second": 3.262,
       "step": 1000
     },
     {
@@ -47,9 +47,9 @@
     {
       "epoch": 0.38804811796662786,
       "eval_loss": 1.1394085884094238,
-      "eval_runtime": 49.4813,
-      "eval_samples_per_second": 104.747,
-      "eval_steps_per_second": 3.274,
+      "eval_runtime": 49.6277,
+      "eval_samples_per_second": 104.438,
+      "eval_steps_per_second": 3.264,
       "step": 2000
     },
     {
@@ -69,9 +69,9 @@
     {
       "epoch": 0.5820721769499418,
       "eval_loss": 1.1180071830749512,
-      "eval_runtime": 49.4875,
-      "eval_samples_per_second": 104.733,
-      "eval_steps_per_second": 3.274,
+      "eval_runtime": 49.6817,
+      "eval_samples_per_second": 104.324,
+      "eval_steps_per_second": 3.261,
       "step": 3000
     },
     {
@@ -91,9 +91,9 @@
     {
       "epoch": 0.7760962359332557,
       "eval_loss": 1.1015874147415161,
-      "eval_runtime": 49.4673,
-      "eval_samples_per_second": 104.776,
-      "eval_steps_per_second": 3.275,
+      "eval_runtime": 49.6378,
+      "eval_samples_per_second": 104.416,
+      "eval_steps_per_second": 3.264,
       "step": 4000
     },
     {
@@ -113,9 +113,9 @@
     {
       "epoch": 0.9701202949165697,
       "eval_loss": 1.091620922088623,
-      "eval_runtime": 49.4696,
-      "eval_samples_per_second": 104.771,
-      "eval_steps_per_second": 3.275,
+      "eval_runtime": 49.6213,
+      "eval_samples_per_second": 104.451,
+      "eval_steps_per_second": 3.265,
       "step": 5000
     },
     {
@@ -135,9 +135,9 @@
     {
       "epoch": 1.1641443538998835,
       "eval_loss": 1.0831941366195679,
-      "eval_runtime": 49.4837,
-      "eval_samples_per_second": 104.742,
-      "eval_steps_per_second": 3.274,
+      "eval_runtime": 49.6774,
+      "eval_samples_per_second": 104.333,
+      "eval_steps_per_second": 3.261,
       "step": 6000
     },
     {
@@ -157,9 +157,9 @@
     {
       "epoch": 1.3581684128831975,
       "eval_loss": 1.0764732360839844,
-      "eval_runtime": 49.4861,
-      "eval_samples_per_second": 104.736,
-      "eval_steps_per_second": 3.274,
+      "eval_runtime": 49.6981,
+      "eval_samples_per_second": 104.29,
+      "eval_steps_per_second": 3.26,
       "step": 7000
     },
     {
@@ -179,9 +179,9 @@
     {
       "epoch": 1.5521924718665114,
       "eval_loss": 1.07025945186615,
-      "eval_runtime": 49.4895,
-      "eval_samples_per_second": 104.729,
-      "eval_steps_per_second": 3.273,
+      "eval_runtime": 49.6881,
+      "eval_samples_per_second": 104.311,
+      "eval_steps_per_second": 3.26,
       "step": 8000
     },
     {
@@ -201,9 +201,9 @@
     {
       "epoch": 1.7462165308498254,
       "eval_loss": 1.0663487911224365,
-      "eval_runtime": 49.4879,
-      "eval_samples_per_second": 104.733,
-      "eval_steps_per_second": 3.274,
+      "eval_runtime": 49.6731,
+      "eval_samples_per_second": 104.342,
+      "eval_steps_per_second": 3.261,
       "step": 9000
     },
     {
@@ -223,9 +223,9 @@
     {
       "epoch": 1.9402405898331394,
       "eval_loss": 1.0637460947036743,
-      "eval_runtime": 49.4829,
-      "eval_samples_per_second": 104.743,
-      "eval_steps_per_second": 3.274,
+      "eval_runtime": 49.672,
+      "eval_samples_per_second": 104.344,
+      "eval_steps_per_second": 3.261,
       "step": 10000
     },
     {
@@ -245,9 +245,9 @@
     {
       "epoch": 2.1342646488164534,
       "eval_loss": 1.0619136095046997,
-      "eval_runtime": 49.4906,
-      "eval_samples_per_second": 104.727,
-      "eval_steps_per_second": 3.273,
+      "eval_runtime": 49.6813,
+      "eval_samples_per_second": 104.325,
+      "eval_steps_per_second": 3.261,
       "step": 11000
     },
     {
@@ -267,9 +267,9 @@
     {
       "epoch": 2.328288707799767,
       "eval_loss": 1.0580228567123413,
-      "eval_runtime": 49.4724,
-      "eval_samples_per_second": 104.765,
-      "eval_steps_per_second": 3.275,
+      "eval_runtime": 49.6253,
+      "eval_samples_per_second": 104.443,
+      "eval_steps_per_second": 3.264,
       "step": 12000
     },
     {
@@ -289,9 +289,9 @@
     {
       "epoch": 2.522312766783081,
       "eval_loss": 1.0561449527740479,
-      "eval_runtime": 49.4628,
-      "eval_samples_per_second": 104.786,
-      "eval_steps_per_second": 3.275,
+      "eval_runtime": 49.663,
+      "eval_samples_per_second": 104.363,
+      "eval_steps_per_second": 3.262,
       "step": 13000
     },
     {
@@ -311,9 +311,9 @@
     {
       "epoch": 2.716336825766395,
       "eval_loss": 1.0557653903961182,
-      "eval_runtime": 49.4704,
-      "eval_samples_per_second": 104.77,
-      "eval_steps_per_second": 3.275,
+      "eval_runtime": 49.6516,
+      "eval_samples_per_second": 104.387,
+      "eval_steps_per_second": 3.263,
       "step": 14000
     },
     {
@@ -333,9 +333,9 @@
     {
       "epoch": 2.910360884749709,
       "eval_loss": 1.054747223854065,
-      "eval_runtime": 49.4705,
-      "eval_samples_per_second": 104.769,
-      "eval_steps_per_second": 3.275,
+      "eval_runtime": 49.5431,
+      "eval_samples_per_second": 104.616,
+      "eval_steps_per_second": 3.27,
       "step": 15000
     }
   ],