End of training

all_results.json

@@ -1,14 +1,14 @@
 {
     "epoch": 3.0,
     "eval_loss": 18.79399871826172,
-    "eval_runtime": 1070.4999,
+    "eval_runtime": 1047.771,
     "eval_samples": 2642,
-    "eval_samples_per_second": 2.468,
-    "eval_steps_per_second": 0.309,
+    "eval_samples_per_second": 2.522,
+    "eval_steps_per_second": 0.316,
     "eval_wer": 1.0,
     "train_loss": 5.609615023672491,
-    "train_runtime": 24845.7878,
+    "train_runtime": 24667.9377,
     "train_samples": 28538,
-    "train_samples_per_second": 3.446,
-    "train_steps_per_second": 0.215
+    "train_samples_per_second": 3.471,
+    "train_steps_per_second": 0.217
 }

eval_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 3.0,
     "eval_loss": 18.79399871826172,
-    "eval_runtime": 1070.4999,
+    "eval_runtime": 1047.771,
     "eval_samples": 2642,
-    "eval_samples_per_second": 2.468,
-    "eval_steps_per_second": 0.309,
+    "eval_samples_per_second": 2.522,
+    "eval_steps_per_second": 0.316,
     "eval_wer": 1.0
 }

runs/Feb21_17-57-39_sanchit--v100/events.out.tfevents.1645492095.sanchit--v100.90840.2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f42088e23202217784c98094e222584ea0d3338f6c6bf0987d271ac2cc3db5da
+size 358

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 3.0,
     "train_loss": 5.609615023672491,
-    "train_runtime": 24845.7878,
+    "train_runtime": 24667.9377,
     "train_samples": 28538,
-    "train_samples_per_second": 3.446,
-    "train_steps_per_second": 0.215
+    "train_samples_per_second": 3.471,
+    "train_steps_per_second": 0.217
 }

trainer_state.json

@@ -3010,9 +3010,9 @@
     {
       "epoch": 0.28,
       "eval_loss": 9.04012393951416,
-      "eval_runtime": 866.6614,
-      "eval_samples_per_second": 3.048,
-      "eval_steps_per_second": 0.382,
+      "eval_runtime": 872.8535,
+      "eval_samples_per_second": 3.027,
+      "eval_steps_per_second": 0.379,
       "eval_wer": 1.0119595398651329,
       "step": 500
     },
@@ -6019,9 +6019,9 @@
     {
       "epoch": 0.56,
       "eval_loss": 9.319939613342285,
-      "eval_runtime": 865.2484,
-      "eval_samples_per_second": 3.053,
-      "eval_steps_per_second": 0.383,
+      "eval_runtime": 873.1949,
+      "eval_samples_per_second": 3.026,
+      "eval_steps_per_second": 0.379,
       "eval_wer": 1.0,
       "step": 1000
     },
@@ -9028,9 +9028,9 @@
     {
       "epoch": 0.84,
       "eval_loss": 8.443438529968262,
-      "eval_runtime": 869.1873,
-      "eval_samples_per_second": 3.04,
-      "eval_steps_per_second": 0.381,
+      "eval_runtime": 873.0258,
+      "eval_samples_per_second": 3.026,
+      "eval_steps_per_second": 0.379,
       "eval_wer": 1.4562673542245141,
       "step": 1500
     },
@@ -12037,9 +12037,9 @@
     {
       "epoch": 1.12,
       "eval_loss": 15.180464744567871,
-      "eval_runtime": 876.5269,
-      "eval_samples_per_second": 3.014,
-      "eval_steps_per_second": 0.378,
+      "eval_runtime": 877.3652,
+      "eval_samples_per_second": 3.011,
+      "eval_steps_per_second": 0.377,
       "eval_wer": 0.9999801666005553,
       "step": 2000
     },
@@ -15046,9 +15046,9 @@
     {
       "epoch": 1.4,
       "eval_loss": 15.79842472076416,
-      "eval_runtime": 873.0125,
-      "eval_samples_per_second": 3.026,
-      "eval_steps_per_second": 0.379,
+      "eval_runtime": 879.1328,
+      "eval_samples_per_second": 3.005,
+      "eval_steps_per_second": 0.377,
       "eval_wer": 1.0020825069416899,
       "step": 2500
     },
@@ -18055,9 +18055,9 @@
     {
       "epoch": 1.68,
       "eval_loss": 18.492834091186523,
-      "eval_runtime": 868.8561,
-      "eval_samples_per_second": 3.041,
-      "eval_steps_per_second": 0.381,
+      "eval_runtime": 877.1195,
+      "eval_samples_per_second": 3.012,
+      "eval_steps_per_second": 0.377,
       "eval_wer": 1.0,
       "step": 3000
     },
@@ -21064,9 +21064,9 @@
     {
       "epoch": 1.96,
       "eval_loss": 17.48856544494629,
-      "eval_runtime": 871.6792,
-      "eval_samples_per_second": 3.031,
-      "eval_steps_per_second": 0.38,
+      "eval_runtime": 876.0525,
+      "eval_samples_per_second": 3.016,
+      "eval_steps_per_second": 0.378,
       "eval_wer": 0.9999801666005553,
       "step": 3500
     },
@@ -24073,9 +24073,9 @@
     {
       "epoch": 2.24,
       "eval_loss": 18.945817947387695,
-      "eval_runtime": 896.893,
-      "eval_samples_per_second": 2.946,
-      "eval_steps_per_second": 0.369,
+      "eval_runtime": 874.0614,
+      "eval_samples_per_second": 3.023,
+      "eval_steps_per_second": 0.379,
       "eval_wer": 1.0,
       "step": 4000
     },
@@ -27082,9 +27082,9 @@
     {
       "epoch": 2.52,
       "eval_loss": 18.279401779174805,
-      "eval_runtime": 900.0015,
-      "eval_samples_per_second": 2.936,
-      "eval_steps_per_second": 0.368,
+      "eval_runtime": 876.9552,
+      "eval_samples_per_second": 3.013,
+      "eval_steps_per_second": 0.377,
       "eval_wer": 1.0005553351844507,
       "step": 4500
     },
@@ -30091,9 +30091,9 @@
     {
       "epoch": 2.8,
       "eval_loss": 18.789825439453125,
-      "eval_runtime": 902.298,
-      "eval_samples_per_second": 2.928,
-      "eval_steps_per_second": 0.367,
+      "eval_runtime": 881.3583,
+      "eval_samples_per_second": 2.998,
+      "eval_steps_per_second": 0.376,
       "eval_wer": 1.0,
       "step": 5000
     },
@@ -32214,9 +32214,9 @@
       "step": 5352,
       "total_flos": 0.0,
       "train_loss": 5.609615023672491,
-      "train_runtime": 24845.7878,
-      "train_samples_per_second": 3.446,
-      "train_steps_per_second": 0.215
+      "train_runtime": 24667.9377,
+      "train_samples_per_second": 3.471,
+      "train_steps_per_second": 0.217
     }
   ],
   "max_steps": 5352,