End of training

all_results.json

@@ -1,14 +1,14 @@
 {
     "epoch": 20.0,
-    "eval_loss": 1.7561472654342651,
-    "eval_runtime": 63.6624,
+    "eval_loss": 1.754138469696045,
+    "eval_runtime": 63.5156,
     "eval_samples": 1680,
-    "eval_samples_per_second": 26.389,
-    "eval_steps_per_second": 26.389,
-    "eval_wer": 0.7971194266418579,
-    "train_loss": 2.3102814756590746,
-    "train_runtime": 4074.7961,
+    "eval_samples_per_second": 26.45,
+    "eval_steps_per_second": 26.45,
+    "eval_wer": 0.8060781476121563,
+    "train_loss": 2.314238773214406,
+    "train_runtime": 4067.0318,
     "train_samples": 4620,
-    "train_samples_per_second": 22.676,
-    "train_steps_per_second": 0.712
+    "train_samples_per_second": 22.719,
+    "train_steps_per_second": 0.713
 }

emissions.csv

@@ -1,2 +1,3 @@
 timestamp,experiment_id,project_name,duration,emissions,energy_consumed,country_name,country_iso_code,region,on_cloud,cloud_provider,cloud_region
 2021-10-27T14:36:34,4a240274-9a5c-450e-8973-cba63eaca4a6,codecarbon,4071.9189558029175,0.01008591898710101,0.04791958355479716,United States,USA,new york,N,,
+2021-10-27T17:13:43,f2a4ada6-7daa-4826-8b1b-8690dde510d9,codecarbon,4064.1793417930603,0.010069911942805982,0.04784353189331327,United States,USA,new york,N,,

eval_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 20.0,
-    "eval_loss": 1.7561472654342651,
-    "eval_runtime": 63.6624,
+    "eval_loss": 1.754138469696045,
+    "eval_runtime": 63.5156,
     "eval_samples": 1680,
-    "eval_samples_per_second": 26.389,
-    "eval_steps_per_second": 26.389,
-    "eval_wer": 0.7971194266418579
+    "eval_samples_per_second": 26.45,
+    "eval_steps_per_second": 26.45,
+    "eval_wer": 0.8060781476121563
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9070fd8b49d5d6c5f2707c22213ba0d06e76a4228ad327299c6dd3c7b09f56a5
+oid sha256:4024a23c429f41f5ad32173a51c26d61d24d79cd90beb516baaa28df171f942f
 size 164046635

runs/Oct27_16-04-41_brutasse/events.out.tfevents.1635350753.brutasse.21935.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cb2a528c85d0f2af4277e3ed6f7a1f5bd8a20c2f845d73ba9fd0efd0980e2d98
-size 57201
+oid sha256:e621c09fa484301709f06d1e9032e6cfa30cb2a906fcb556e8abf82c42b4b605
+size 59443

runs/Oct27_16-04-41_brutasse/events.out.tfevents.1635354889.brutasse.21935.2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:478071f6a026d89dec737724c8e0dc46a385e1b4d3892b736ef8f7395ceba675
+size 358

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 20.0,
-    "train_loss": 2.3102814756590746,
-    "train_runtime": 4074.7961,
+    "train_loss": 2.314238773214406,
+    "train_runtime": 4067.0318,
     "train_samples": 4620,
-    "train_samples_per_second": 22.676,
-    "train_steps_per_second": 0.712
+    "train_samples_per_second": 22.719,
+    "train_steps_per_second": 0.713
 }

trainer_state.json

@@ -69,10 +69,10 @@
     },
     {
       "epoch": 0.69,
-      "eval_loss": 4.080197811126709,
-      "eval_runtime": 62.227,
-      "eval_samples_per_second": 26.998,
-      "eval_steps_per_second": 26.998,
+      "eval_loss": 4.080199241638184,
+      "eval_runtime": 62.2219,
+      "eval_samples_per_second": 27.0,
+      "eval_steps_per_second": 27.0,
       "eval_wer": 1.0,
       "step": 100
     },
@@ -133,15 +133,15 @@
     {
       "epoch": 1.38,
       "learning_rate": 1.97e-05,
-      "loss": 2.9806,
+      "loss": 2.9805,
       "step": 200
     },
     {
       "epoch": 1.38,
-      "eval_loss": 2.979236125946045,
-      "eval_runtime": 62.498,
-      "eval_samples_per_second": 26.881,
-      "eval_steps_per_second": 26.881,
+      "eval_loss": 2.9792351722717285,
+      "eval_runtime": 62.532,
+      "eval_samples_per_second": 26.866,
+      "eval_steps_per_second": 26.866,
       "eval_wer": 1.0,
       "step": 200
     },
@@ -207,10 +207,10 @@
     },
     {
       "epoch": 2.07,
-      "eval_loss": 2.9407615661621094,
-      "eval_runtime": 62.3432,
-      "eval_samples_per_second": 26.948,
-      "eval_steps_per_second": 26.948,
+      "eval_loss": 2.940765857696533,
+      "eval_runtime": 62.8838,
+      "eval_samples_per_second": 26.716,
+      "eval_steps_per_second": 26.716,
       "eval_wer": 1.0,
       "step": 300
     },
@@ -276,10 +276,10 @@
     },
     {
       "epoch": 2.76,
-      "eval_loss": 2.914332151412964,
-      "eval_runtime": 62.583,
-      "eval_samples_per_second": 26.844,
-      "eval_steps_per_second": 26.844,
+      "eval_loss": 2.9143340587615967,
+      "eval_runtime": 63.136,
+      "eval_samples_per_second": 26.609,
+      "eval_steps_per_second": 26.609,
       "eval_wer": 1.0,
       "step": 400
     },
@@ -316,19 +316,19 @@
     {
       "epoch": 3.17,
       "learning_rate": 4.5700000000000006e-05,
-      "loss": 2.8883,
+      "loss": 2.8882,
       "step": 460
     },
     {
       "epoch": 3.24,
       "learning_rate": 4.6700000000000003e-05,
-      "loss": 2.8962,
+      "loss": 2.8961,
       "step": 470
     },
     {
       "epoch": 3.31,
       "learning_rate": 4.77e-05,
-      "loss": 2.9122,
+      "loss": 2.9121,
       "step": 480
     },
     {
@@ -345,1677 +345,1677 @@
     },
     {
       "epoch": 3.45,
-      "eval_loss": 2.877377986907959,
-      "eval_runtime": 62.5594,
-      "eval_samples_per_second": 26.854,
-      "eval_steps_per_second": 26.854,
+      "eval_loss": 2.8774502277374268,
+      "eval_runtime": 63.3794,
+      "eval_samples_per_second": 26.507,
+      "eval_steps_per_second": 26.507,
       "eval_wer": 1.0,
       "step": 500
     },
     {
       "epoch": 3.52,
       "learning_rate": 5.0700000000000006e-05,
-      "loss": 2.9153,
+      "loss": 2.9154,
       "step": 510
     },
     {
       "epoch": 3.59,
       "learning_rate": 5.17e-05,
-      "loss": 2.8849,
+      "loss": 2.885,
       "step": 520
     },
     {
       "epoch": 3.66,
       "learning_rate": 5.270000000000001e-05,
-      "loss": 2.8636,
+      "loss": 2.8637,
       "step": 530
     },
     {
       "epoch": 3.72,
       "learning_rate": 5.3700000000000004e-05,
-      "loss": 2.8881,
+      "loss": 2.888,
       "step": 540
     },
     {
       "epoch": 3.79,
       "learning_rate": 5.470000000000001e-05,
-      "loss": 2.9056,
+      "loss": 2.9059,
       "step": 550
     },
     {
       "epoch": 3.86,
       "learning_rate": 5.5700000000000005e-05,
-      "loss": 2.8631,
+      "loss": 2.863,
       "step": 560
     },
     {
       "epoch": 3.93,
       "learning_rate": 5.6699999999999996e-05,
-      "loss": 2.844,
+      "loss": 2.8439,
       "step": 570
     },
     {
       "epoch": 4.0,
       "learning_rate": 5.77e-05,
-      "loss": 2.8526,
+      "loss": 2.8525,
       "step": 580
     },
     {
       "epoch": 4.07,
       "learning_rate": 5.87e-05,
-      "loss": 2.8439,
+      "loss": 2.8441,
       "step": 590
     },
     {
       "epoch": 4.14,
       "learning_rate": 5.97e-05,
-      "loss": 2.7712,
+      "loss": 2.7718,
       "step": 600
     },
     {
       "epoch": 4.14,
-      "eval_loss": 2.7769241333007812,
-      "eval_runtime": 62.7853,
-      "eval_samples_per_second": 26.758,
-      "eval_steps_per_second": 26.758,
-      "eval_wer": 0.9999310867617669,
+      "eval_loss": 2.7786543369293213,
+      "eval_runtime": 63.096,
+      "eval_samples_per_second": 26.626,
+      "eval_steps_per_second": 26.626,
+      "eval_wer": 1.0,
       "step": 600
     },
     {
       "epoch": 4.21,
       "learning_rate": 6.07e-05,
-      "loss": 2.7714,
+      "loss": 2.772,
       "step": 610
     },
     {
       "epoch": 4.28,
       "learning_rate": 6.170000000000001e-05,
-      "loss": 2.8165,
+      "loss": 2.8153,
       "step": 620
     },
     {
       "epoch": 4.34,
       "learning_rate": 6.27e-05,
-      "loss": 2.7616,
+      "loss": 2.7641,
       "step": 630
     },
     {
       "epoch": 4.41,
       "learning_rate": 6.37e-05,
-      "loss": 2.7563,
+      "loss": 2.7527,
       "step": 640
     },
     {
       "epoch": 4.48,
       "learning_rate": 6.47e-05,
-      "loss": 2.7404,
+      "loss": 2.746,
       "step": 650
     },
     {
       "epoch": 4.55,
       "learning_rate": 6.570000000000001e-05,
-      "loss": 2.7597,
+      "loss": 2.7824,
       "step": 660
     },
     {
       "epoch": 4.62,
       "learning_rate": 6.670000000000001e-05,
-      "loss": 2.69,
+      "loss": 2.6965,
       "step": 670
     },
     {
       "epoch": 4.69,
       "learning_rate": 6.77e-05,
-      "loss": 2.6965,
+      "loss": 2.701,
       "step": 680
     },
     {
       "epoch": 4.76,
       "learning_rate": 6.87e-05,
-      "loss": 2.7619,
+      "loss": 2.7603,
       "step": 690
     },
     {
       "epoch": 4.83,
       "learning_rate": 6.97e-05,
-      "loss": 2.6662,
+      "loss": 2.6711,
       "step": 700
     },
     {
       "epoch": 4.83,
-      "eval_loss": 2.642500162124634,
-      "eval_runtime": 62.8037,
-      "eval_samples_per_second": 26.75,
-      "eval_steps_per_second": 26.75,
-      "eval_wer": 0.9789125491006823,
+      "eval_loss": 2.640113115310669,
+      "eval_runtime": 63.3201,
+      "eval_samples_per_second": 26.532,
+      "eval_steps_per_second": 26.532,
+      "eval_wer": 0.978567982909517,
       "step": 700
     },
     {
       "epoch": 4.9,
       "learning_rate": 7.07e-05,
-      "loss": 2.6053,
+      "loss": 2.6076,
       "step": 710
     },
     {
       "epoch": 4.97,
       "learning_rate": 7.17e-05,
-      "loss": 2.6388,
+      "loss": 2.6269,
       "step": 720
     },
     {
       "epoch": 5.03,
       "learning_rate": 7.27e-05,
-      "loss": 2.7087,
+      "loss": 2.7102,
       "step": 730
     },
     {
       "epoch": 5.1,
       "learning_rate": 7.37e-05,
-      "loss": 2.5896,
+      "loss": 2.6061,
       "step": 740
     },
     {
       "epoch": 5.17,
       "learning_rate": 7.47e-05,
-      "loss": 2.5602,
+      "loss": 2.5717,
       "step": 750
     },
     {
       "epoch": 5.24,
       "learning_rate": 7.570000000000001e-05,
-      "loss": 2.6433,
+      "loss": 2.6457,
       "step": 760
     },
     {
       "epoch": 5.31,
       "learning_rate": 7.670000000000001e-05,
-      "loss": 2.6369,
+      "loss": 2.6255,
       "step": 770
     },
     {
       "epoch": 5.38,
       "learning_rate": 7.77e-05,
-      "loss": 2.5393,
+      "loss": 2.5464,
       "step": 780
     },
     {
       "epoch": 5.45,
       "learning_rate": 7.87e-05,
-      "loss": 2.5482,
+      "loss": 2.5584,
       "step": 790
     },
     {
       "epoch": 5.52,
       "learning_rate": 7.970000000000001e-05,
-      "loss": 2.632,
+      "loss": 2.6403,
       "step": 800
     },
     {
       "epoch": 5.52,
-      "eval_loss": 2.51424503326416,
-      "eval_runtime": 62.9233,
-      "eval_samples_per_second": 26.699,
-      "eval_steps_per_second": 26.699,
-      "eval_wer": 1.0317690028254427,
+      "eval_loss": 2.5434978008270264,
+      "eval_runtime": 63.2107,
+      "eval_samples_per_second": 26.578,
+      "eval_steps_per_second": 26.578,
+      "eval_wer": 1.0392116325546137,
       "step": 800
     },
     {
       "epoch": 5.59,
       "learning_rate": 8.070000000000001e-05,
-      "loss": 2.5368,
+      "loss": 2.5656,
       "step": 810
     },
     {
       "epoch": 5.66,
       "learning_rate": 8.17e-05,
-      "loss": 2.4602,
+      "loss": 2.4717,
       "step": 820
     },
     {
       "epoch": 5.72,
       "learning_rate": 8.27e-05,
-      "loss": 2.5246,
+      "loss": 2.533,
       "step": 830
     },
     {
       "epoch": 5.79,
       "learning_rate": 8.37e-05,
-      "loss": 2.6324,
+      "loss": 2.6475,
       "step": 840
     },
     {
       "epoch": 5.86,
       "learning_rate": 8.47e-05,
-      "loss": 2.4211,
+      "loss": 2.4591,
       "step": 850
     },
     {
       "epoch": 5.93,
       "learning_rate": 8.57e-05,
-      "loss": 2.4506,
+      "loss": 2.4935,
       "step": 860
     },
     {
       "epoch": 6.0,
       "learning_rate": 8.67e-05,
-      "loss": 2.5508,
+      "loss": 2.5621,
       "step": 870
     },
     {
       "epoch": 6.07,
       "learning_rate": 8.77e-05,
-      "loss": 2.4846,
+      "loss": 2.5034,
       "step": 880
     },
     {
       "epoch": 6.14,
       "learning_rate": 8.87e-05,
-      "loss": 2.3699,
+      "loss": 2.4035,
       "step": 890
     },
     {
       "epoch": 6.21,
       "learning_rate": 8.970000000000001e-05,
-      "loss": 2.3794,
+      "loss": 2.4052,
       "step": 900
     },
     {
       "epoch": 6.21,
-      "eval_loss": 2.436049222946167,
-      "eval_runtime": 62.8458,
-      "eval_samples_per_second": 26.732,
-      "eval_steps_per_second": 26.732,
-      "eval_wer": 1.1474743298187582,
+      "eval_loss": 2.4579782485961914,
+      "eval_runtime": 63.5786,
+      "eval_samples_per_second": 26.424,
+      "eval_steps_per_second": 26.424,
+      "eval_wer": 1.0706360691888912,
       "step": 900
     },
     {
       "epoch": 6.28,
       "learning_rate": 9.070000000000001e-05,
-      "loss": 2.565,
+      "loss": 2.5836,
       "step": 910
     },
     {
       "epoch": 6.34,
       "learning_rate": 9.17e-05,
-      "loss": 2.4147,
+      "loss": 2.419,
       "step": 920
     },
     {
       "epoch": 6.41,
       "learning_rate": 9.27e-05,
-      "loss": 2.3703,
+      "loss": 2.391,
       "step": 930
     },
     {
       "epoch": 6.48,
       "learning_rate": 9.370000000000001e-05,
-      "loss": 2.4564,
+      "loss": 2.4647,
       "step": 940
     },
     {
       "epoch": 6.55,
       "learning_rate": 9.47e-05,
-      "loss": 2.5176,
+      "loss": 2.5118,
       "step": 950
     },
     {
       "epoch": 6.62,
       "learning_rate": 9.57e-05,
-      "loss": 2.2307,
+      "loss": 2.2727,
       "step": 960
     },
     {
       "epoch": 6.69,
       "learning_rate": 9.67e-05,
-      "loss": 2.2363,
+      "loss": 2.267,
       "step": 970
     },
     {
       "epoch": 6.76,
       "learning_rate": 9.77e-05,
-      "loss": 2.5199,
+      "loss": 2.5176,
       "step": 980
     },
     {
       "epoch": 6.83,
       "learning_rate": 9.87e-05,
-      "loss": 2.3646,
+      "loss": 2.4048,
       "step": 990
     },
     {
       "epoch": 6.9,
       "learning_rate": 9.970000000000001e-05,
-      "loss": 2.1406,
+      "loss": 2.1708,
       "step": 1000
     },
     {
       "epoch": 6.9,
-      "eval_loss": 2.293168783187866,
-      "eval_runtime": 62.6106,
-      "eval_samples_per_second": 26.833,
-      "eval_steps_per_second": 26.833,
-      "eval_wer": 0.9962097718971814,
+      "eval_loss": 2.2800345420837402,
+      "eval_runtime": 62.9699,
+      "eval_samples_per_second": 26.679,
+      "eval_steps_per_second": 26.679,
+      "eval_wer": 1.0089587209702984,
       "step": 1000
     },
     {
       "epoch": 6.97,
       "learning_rate": 9.963157894736843e-05,
-      "loss": 2.3291,
+      "loss": 2.3227,
       "step": 1010
     },
     {
       "epoch": 7.03,
       "learning_rate": 9.910526315789475e-05,
-      "loss": 2.5112,
+      "loss": 2.5246,
       "step": 1020
     },
     {
       "epoch": 7.1,
       "learning_rate": 9.857894736842106e-05,
-      "loss": 2.1736,
+      "loss": 2.2273,
       "step": 1030
     },
     {
       "epoch": 7.17,
       "learning_rate": 9.805263157894737e-05,
-      "loss": 2.1442,
+      "loss": 2.181,
       "step": 1040
     },
     {
       "epoch": 7.24,
       "learning_rate": 9.752631578947369e-05,
-      "loss": 2.3392,
+      "loss": 2.3324,
       "step": 1050
     },
     {
       "epoch": 7.31,
       "learning_rate": 9.7e-05,
-      "loss": 2.3714,
+      "loss": 2.3922,
       "step": 1060
     },
     {
       "epoch": 7.38,
       "learning_rate": 9.647368421052631e-05,
-      "loss": 1.978,
+      "loss": 2.0177,
       "step": 1070
     },
     {
       "epoch": 7.45,
       "learning_rate": 9.594736842105264e-05,
-      "loss": 2.1326,
+      "loss": 2.1559,
       "step": 1080
     },
     {
       "epoch": 7.52,
       "learning_rate": 9.542105263157895e-05,
-      "loss": 2.445,
+      "loss": 2.458,
       "step": 1090
     },
     {
       "epoch": 7.59,
       "learning_rate": 9.489473684210527e-05,
-      "loss": 2.223,
+      "loss": 2.2555,
       "step": 1100
     },
     {
       "epoch": 7.59,
-      "eval_loss": 2.159000873565674,
-      "eval_runtime": 63.0762,
-      "eval_samples_per_second": 26.634,
-      "eval_steps_per_second": 26.634,
-      "eval_wer": 0.9280545792846806,
+      "eval_loss": 2.1493031978607178,
+      "eval_runtime": 63.4,
+      "eval_samples_per_second": 26.498,
+      "eval_steps_per_second": 26.498,
+      "eval_wer": 0.9578940114395975,
       "step": 1100
     },
     {
       "epoch": 7.66,
       "learning_rate": 9.436842105263158e-05,
-      "loss": 1.969,
+      "loss": 1.9823,
       "step": 1110
     },
     {
       "epoch": 7.72,
       "learning_rate": 9.384210526315789e-05,
-      "loss": 2.1386,
+      "loss": 2.1439,
       "step": 1120
     },
     {
       "epoch": 7.79,
       "learning_rate": 9.331578947368422e-05,
-      "loss": 2.3359,
+      "loss": 2.4031,
       "step": 1130
     },
     {
       "epoch": 7.86,
       "learning_rate": 9.278947368421053e-05,
-      "loss": 1.9546,
+      "loss": 2.0473,
       "step": 1140
     },
     {
       "epoch": 7.93,
       "learning_rate": 9.226315789473686e-05,
-      "loss": 2.0345,
+      "loss": 2.0511,
       "step": 1150
     },
     {
       "epoch": 8.0,
       "learning_rate": 9.173684210526317e-05,
-      "loss": 2.3499,
+      "loss": 2.379,
       "step": 1160
     },
     {
       "epoch": 8.07,
       "learning_rate": 9.121052631578948e-05,
-      "loss": 2.2055,
+      "loss": 2.2715,
       "step": 1170
     },
     {
       "epoch": 8.14,
       "learning_rate": 9.06842105263158e-05,
-      "loss": 1.9114,
+      "loss": 1.9654,
       "step": 1180
     },
     {
       "epoch": 8.21,
       "learning_rate": 9.015789473684211e-05,
-      "loss": 2.0313,
+      "loss": 2.0454,
       "step": 1190
     },
     {
       "epoch": 8.28,
       "learning_rate": 8.963157894736842e-05,
-      "loss": 2.3607,
+      "loss": 2.3673,
       "step": 1200
     },
     {
       "epoch": 8.28,
-      "eval_loss": 2.0553243160247803,
-      "eval_runtime": 62.862,
-      "eval_samples_per_second": 26.725,
-      "eval_steps_per_second": 26.725,
-      "eval_wer": 0.8681689752601475,
+      "eval_loss": 2.0709428787231445,
+      "eval_runtime": 63.9178,
+      "eval_samples_per_second": 26.284,
+      "eval_steps_per_second": 26.284,
+      "eval_wer": 0.9051064709530701,
       "step": 1200
     },
     {
       "epoch": 8.34,
       "learning_rate": 8.910526315789474e-05,
-      "loss": 1.8724,
+      "loss": 1.9115,
       "step": 1210
     },
     {
       "epoch": 8.41,
       "learning_rate": 8.857894736842106e-05,
-      "loss": 1.875,
+      "loss": 1.9225,
       "step": 1220
     },
     {
       "epoch": 8.48,
       "learning_rate": 8.805263157894737e-05,
-      "loss": 2.187,
+      "loss": 2.2256,
       "step": 1230
     },
     {
       "epoch": 8.55,
       "learning_rate": 8.752631578947369e-05,
-      "loss": 2.2535,
+      "loss": 2.3292,
       "step": 1240
     },
     {
       "epoch": 8.62,
       "learning_rate": 8.7e-05,
-      "loss": 1.8278,
+      "loss": 1.8711,
       "step": 1250
     },
     {
       "epoch": 8.69,
       "learning_rate": 8.647368421052631e-05,
-      "loss": 1.9544,
+      "loss": 1.9655,
       "step": 1260
     },
     {
       "epoch": 8.76,
       "learning_rate": 8.594736842105263e-05,
-      "loss": 2.3521,
+      "loss": 2.3567,
       "step": 1270
     },
     {
       "epoch": 8.83,
       "learning_rate": 8.542105263157894e-05,
-      "loss": 2.0843,
+      "loss": 2.0859,
       "step": 1280
     },
     {
       "epoch": 8.9,
       "learning_rate": 8.489473684210527e-05,
-      "loss": 1.8205,
+      "loss": 1.8144,
       "step": 1290
     },
     {
       "epoch": 8.97,
       "learning_rate": 8.436842105263158e-05,
-      "loss": 2.1058,
+      "loss": 2.091,
       "step": 1300
     },
     {
       "epoch": 8.97,
-      "eval_loss": 2.0442521572113037,
-      "eval_runtime": 62.644,
-      "eval_samples_per_second": 26.818,
-      "eval_steps_per_second": 26.818,
-      "eval_wer": 0.8901522982564951,
+      "eval_loss": 2.025775909423828,
+      "eval_runtime": 63.4247,
+      "eval_samples_per_second": 26.488,
+      "eval_steps_per_second": 26.488,
+      "eval_wer": 0.8926331748328854,
       "step": 1300
     },
     {
       "epoch": 9.03,
       "learning_rate": 8.38421052631579e-05,
-      "loss": 2.3144,
+      "loss": 2.3111,
       "step": 1310
     },
     {
       "epoch": 9.1,
       "learning_rate": 8.331578947368422e-05,
-      "loss": 1.7695,
+      "loss": 1.7863,
       "step": 1320
     },
     {
       "epoch": 9.17,
       "learning_rate": 8.278947368421053e-05,
-      "loss": 1.8144,
+      "loss": 1.8556,
       "step": 1330
     },
     {
       "epoch": 9.24,
       "learning_rate": 8.226315789473684e-05,
-      "loss": 2.174,
+      "loss": 2.217,
       "step": 1340
     },
     {
       "epoch": 9.31,
       "learning_rate": 8.173684210526317e-05,
-      "loss": 2.1989,
+      "loss": 2.2324,
       "step": 1350
     },
     {
       "epoch": 9.38,
       "learning_rate": 8.121052631578948e-05,
-      "loss": 1.8499,
+      "loss": 1.8721,
       "step": 1360
     },
     {
       "epoch": 9.45,
       "learning_rate": 8.06842105263158e-05,
-      "loss": 1.9672,
+      "loss": 1.9822,
       "step": 1370
     },
     {
       "epoch": 9.52,
       "learning_rate": 8.015789473684211e-05,
-      "loss": 2.2523,
+      "loss": 2.2722,
       "step": 1380
     },
     {
       "epoch": 9.59,
       "learning_rate": 7.963157894736842e-05,
-      "loss": 1.8925,
+      "loss": 1.8985,
       "step": 1390
     },
     {
       "epoch": 9.66,
       "learning_rate": 7.910526315789474e-05,
-      "loss": 1.8191,
+      "loss": 1.8433,
       "step": 1400
     },
     {
       "epoch": 9.66,
-      "eval_loss": 1.9585875272750854,
-      "eval_runtime": 62.9578,
-      "eval_samples_per_second": 26.685,
-      "eval_steps_per_second": 26.685,
-      "eval_wer": 0.8237199365998208,
+      "eval_loss": 1.9644908905029297,
+      "eval_runtime": 63.8196,
+      "eval_samples_per_second": 26.324,
+      "eval_steps_per_second": 26.324,
+      "eval_wer": 0.8242712425056854,
       "step": 1400
     },
     {
       "epoch": 9.72,
       "learning_rate": 7.857894736842105e-05,
-      "loss": 2.1013,
+      "loss": 2.1261,
       "step": 1410
     },
     {
       "epoch": 9.79,
       "learning_rate": 7.805263157894738e-05,
-      "loss": 2.2354,
+      "loss": 2.2486,
       "step": 1420
     },
     {
       "epoch": 9.86,
       "learning_rate": 7.752631578947369e-05,
-      "loss": 1.7551,
+      "loss": 1.7644,
       "step": 1430
     },
     {
       "epoch": 9.93,
       "learning_rate": 7.7e-05,
-      "loss": 1.8477,
+      "loss": 1.845,
       "step": 1440
     },
     {
       "epoch": 10.0,
       "learning_rate": 7.647368421052631e-05,
-      "loss": 2.1596,
+      "loss": 2.1699,
       "step": 1450
     },
     {
       "epoch": 10.07,
       "learning_rate": 7.594736842105263e-05,
-      "loss": 2.0151,
+      "loss": 2.0384,
       "step": 1460
     },
     {
       "epoch": 10.14,
       "learning_rate": 7.542105263157895e-05,
-      "loss": 1.7181,
+      "loss": 1.7081,
       "step": 1470
     },
     {
       "epoch": 10.21,
       "learning_rate": 7.489473684210527e-05,
-      "loss": 1.8951,
+      "loss": 1.8563,
       "step": 1480
     },
     {
       "epoch": 10.28,
       "learning_rate": 7.43684210526316e-05,
-      "loss": 2.3172,
+      "loss": 2.2783,
       "step": 1490
     },
     {
       "epoch": 10.34,
       "learning_rate": 7.38421052631579e-05,
-      "loss": 1.7013,
+      "loss": 1.6824,
       "step": 1500
     },
     {
       "epoch": 10.34,
-      "eval_loss": 1.9585844278335571,
-      "eval_runtime": 62.8863,
-      "eval_samples_per_second": 26.715,
-      "eval_steps_per_second": 26.715,
-      "eval_wer": 0.8689270208807112,
+      "eval_loss": 1.9211441278457642,
+      "eval_runtime": 63.3051,
+      "eval_samples_per_second": 26.538,
+      "eval_steps_per_second": 26.538,
+      "eval_wer": 0.8707187650747709,
       "step": 1500
     },
     {
       "epoch": 10.41,
       "learning_rate": 7.331578947368422e-05,
-      "loss": 1.7497,
+      "loss": 1.7436,
       "step": 1510
     },
     {
       "epoch": 10.48,
       "learning_rate": 7.278947368421053e-05,
-      "loss": 2.0471,
+      "loss": 2.0382,
       "step": 1520
     },
     {
       "epoch": 10.55,
       "learning_rate": 7.226315789473685e-05,
-      "loss": 2.2056,
+      "loss": 2.2084,
       "step": 1530
     },
     {
       "epoch": 10.62,
       "learning_rate": 7.173684210526316e-05,
-      "loss": 1.7579,
+      "loss": 1.7541,
       "step": 1540
     },
     {
       "epoch": 10.69,
       "learning_rate": 7.121052631578947e-05,
-      "loss": 1.9813,
+      "loss": 1.9633,
       "step": 1550
     },
     {
       "epoch": 10.76,
       "learning_rate": 7.06842105263158e-05,
-      "loss": 2.2292,
+      "loss": 2.2264,
       "step": 1560
     },
     {
       "epoch": 10.83,
       "learning_rate": 7.015789473684211e-05,
-      "loss": 1.9037,
+      "loss": 1.8931,
       "step": 1570
     },
     {
       "epoch": 10.9,
       "learning_rate": 6.963157894736842e-05,
-      "loss": 1.736,
+      "loss": 1.7337,
       "step": 1580
     },
     {
       "epoch": 10.97,
       "learning_rate": 6.910526315789474e-05,
-      "loss": 1.8723,
+      "loss": 1.8795,
       "step": 1590
     },
     {
       "epoch": 11.03,
       "learning_rate": 6.857894736842105e-05,
-      "loss": 2.2289,
+      "loss": 2.2282,
       "step": 1600
     },
     {
       "epoch": 11.03,
-      "eval_loss": 1.908236026763916,
-      "eval_runtime": 62.9372,
-      "eval_samples_per_second": 26.693,
-      "eval_steps_per_second": 26.693,
-      "eval_wer": 0.8610709117221418,
+      "eval_loss": 1.8913612365722656,
+      "eval_runtime": 63.4568,
+      "eval_samples_per_second": 26.475,
+      "eval_steps_per_second": 26.475,
+      "eval_wer": 0.8695472400248088,
       "step": 1600
     },
     {
       "epoch": 11.1,
       "learning_rate": 6.805263157894736e-05,
-      "loss": 1.7338,
+      "loss": 1.7207,
       "step": 1610
     },
     {
       "epoch": 11.17,
       "learning_rate": 6.752631578947368e-05,
-      "loss": 1.7262,
+      "loss": 1.714,
       "step": 1620
     },
     {
       "epoch": 11.24,
       "learning_rate": 6.7e-05,
-      "loss": 2.0395,
+      "loss": 2.034,
       "step": 1630
     },
     {
       "epoch": 11.31,
       "learning_rate": 6.647368421052632e-05,
-      "loss": 2.0661,
+      "loss": 2.0639,
       "step": 1640
     },
     {
       "epoch": 11.38,
       "learning_rate": 6.594736842105264e-05,
-      "loss": 1.7218,
+      "loss": 1.7275,
       "step": 1650
     },
     {
       "epoch": 11.45,
       "learning_rate": 6.542105263157895e-05,
-      "loss": 1.8517,
+      "loss": 1.8704,
       "step": 1660
     },
     {
       "epoch": 11.52,
       "learning_rate": 6.489473684210527e-05,
-      "loss": 2.2154,
+      "loss": 2.2262,
       "step": 1670
     },
     {
       "epoch": 11.59,
       "learning_rate": 6.436842105263158e-05,
-      "loss": 1.7295,
+      "loss": 1.7371,
       "step": 1680
     },
     {
       "epoch": 11.66,
       "learning_rate": 6.384210526315791e-05,
-      "loss": 1.6468,
+      "loss": 1.6343,
       "step": 1690
     },
     {
       "epoch": 11.72,
       "learning_rate": 6.331578947368422e-05,
-      "loss": 1.9125,
+      "loss": 1.9027,
       "step": 1700
     },
     {
       "epoch": 11.72,
-      "eval_loss": 1.8771533966064453,
-      "eval_runtime": 62.6761,
-      "eval_samples_per_second": 26.804,
-      "eval_steps_per_second": 26.804,
-      "eval_wer": 0.8149679553442216,
+      "eval_loss": 1.871816873550415,
+      "eval_runtime": 63.6857,
+      "eval_samples_per_second": 26.38,
+      "eval_steps_per_second": 26.38,
+      "eval_wer": 0.8342636620494797,
       "step": 1700
     },
     {
       "epoch": 11.79,
       "learning_rate": 6.278947368421053e-05,
-      "loss": 2.1522,
+      "loss": 2.1587,
       "step": 1710
     },
     {
       "epoch": 11.86,
       "learning_rate": 6.226315789473685e-05,
-      "loss": 1.6395,
+      "loss": 1.643,
       "step": 1720
     },
     {
       "epoch": 11.93,
       "learning_rate": 6.173684210526316e-05,
-      "loss": 1.7794,
+      "loss": 1.7811,
       "step": 1730
     },
     {
       "epoch": 12.0,
       "learning_rate": 6.121052631578947e-05,
-      "loss": 2.1512,
+      "loss": 2.1443,
       "step": 1740
     },
     {
       "epoch": 12.07,
       "learning_rate": 6.0684210526315785e-05,
-      "loss": 1.9098,
+      "loss": 1.9129,
       "step": 1750
     },
     {
       "epoch": 12.14,
       "learning_rate": 6.015789473684211e-05,
-      "loss": 1.6335,
+      "loss": 1.629,
       "step": 1760
     },
     {
       "epoch": 12.21,
       "learning_rate": 5.9631578947368425e-05,
-      "loss": 1.8464,
+      "loss": 1.8447,
       "step": 1770
     },
     {
       "epoch": 12.28,
       "learning_rate": 5.9105263157894744e-05,
-      "loss": 2.15,
+      "loss": 2.1536,
       "step": 1780
     },
     {
       "epoch": 12.34,
       "learning_rate": 5.857894736842106e-05,
-      "loss": 1.7132,
+      "loss": 1.7021,
       "step": 1790
     },
     {
       "epoch": 12.41,
       "learning_rate": 5.805263157894737e-05,
-      "loss": 1.6424,
+      "loss": 1.6303,
       "step": 1800
     },
     {
       "epoch": 12.41,
-      "eval_loss": 1.8670605421066284,
-      "eval_runtime": 62.8262,
-      "eval_samples_per_second": 26.74,
-      "eval_steps_per_second": 26.74,
-      "eval_wer": 0.7870580938598305,
+      "eval_loss": 1.86460280418396,
+      "eval_runtime": 63.8627,
+      "eval_samples_per_second": 26.306,
+      "eval_steps_per_second": 26.306,
+      "eval_wer": 0.8231686306939563,
       "step": 1800
     },
     {
       "epoch": 12.48,
       "learning_rate": 5.752631578947368e-05,
-      "loss": 2.0029,
+      "loss": 1.9938,
       "step": 1810
     },
     {
       "epoch": 12.55,
       "learning_rate": 5.6999999999999996e-05,
-      "loss": 2.0917,
+      "loss": 2.0685,
       "step": 1820
     },
     {
       "epoch": 12.62,
       "learning_rate": 5.647368421052632e-05,
-      "loss": 1.6332,
+      "loss": 1.624,
       "step": 1830
     },
     {
       "epoch": 12.69,
       "learning_rate": 5.5947368421052636e-05,
-      "loss": 1.7216,
+      "loss": 1.7289,
       "step": 1840
     },
     {
       "epoch": 12.76,
       "learning_rate": 5.542105263157895e-05,
-      "loss": 2.0909,
+      "loss": 2.1195,
       "step": 1850
     },
     {
       "epoch": 12.83,
       "learning_rate": 5.489473684210527e-05,
-      "loss": 1.8317,
+      "loss": 1.8338,
       "step": 1860
     },
     {
       "epoch": 12.9,
       "learning_rate": 5.436842105263158e-05,
-      "loss": 1.6383,
+      "loss": 1.6352,
       "step": 1870
     },
     {
       "epoch": 12.97,
       "learning_rate": 5.3842105263157895e-05,
-      "loss": 1.8577,
+      "loss": 1.8484,
       "step": 1880
     },
     {
       "epoch": 13.03,
       "learning_rate": 5.331578947368421e-05,
-      "loss": 2.1335,
+      "loss": 2.1297,
       "step": 1890
     },
     {
       "epoch": 13.1,
       "learning_rate": 5.2789473684210534e-05,
-      "loss": 1.6553,
+      "loss": 1.648,
       "step": 1900
     },
     {
       "epoch": 13.1,
-      "eval_loss": 1.819168210029602,
-      "eval_runtime": 62.7381,
-      "eval_samples_per_second": 26.778,
-      "eval_steps_per_second": 26.778,
-      "eval_wer": 0.8120735993384329,
+      "eval_loss": 1.8297438621520996,
+      "eval_runtime": 63.5984,
+      "eval_samples_per_second": 26.416,
+      "eval_steps_per_second": 26.416,
+      "eval_wer": 0.8176555716353111,
       "step": 1900
     },
     {
       "epoch": 13.17,
       "learning_rate": 5.226315789473685e-05,
-      "loss": 1.7138,
+      "loss": 1.7123,
       "step": 1910
     },
     {
       "epoch": 13.24,
       "learning_rate": 5.173684210526316e-05,
-      "loss": 1.9229,
+      "loss": 1.9177,
       "step": 1920
     },
     {
       "epoch": 13.31,
       "learning_rate": 5.121052631578947e-05,
-      "loss": 1.9276,
+      "loss": 1.9108,
       "step": 1930
     },
     {
       "epoch": 13.38,
       "learning_rate": 5.068421052631579e-05,
-      "loss": 1.6501,
+      "loss": 1.6338,
       "step": 1940
     },
     {
       "epoch": 13.45,
       "learning_rate": 5.0157894736842106e-05,
-      "loss": 1.7051,
+      "loss": 1.6991,
       "step": 1950
     },
     {
       "epoch": 13.52,
       "learning_rate": 4.9631578947368426e-05,
-      "loss": 2.1649,
+      "loss": 2.1586,
       "step": 1960
     },
     {
       "epoch": 13.59,
       "learning_rate": 4.910526315789474e-05,
-      "loss": 1.7654,
+      "loss": 1.747,
       "step": 1970
     },
     {
       "epoch": 13.66,
       "learning_rate": 4.857894736842106e-05,
-      "loss": 1.6051,
+      "loss": 1.5962,
       "step": 1980
     },
     {
       "epoch": 13.72,
       "learning_rate": 4.805263157894737e-05,
-      "loss": 1.9606,
+      "loss": 1.9579,
       "step": 1990
     },
     {
       "epoch": 13.79,
       "learning_rate": 4.7526315789473684e-05,
-      "loss": 2.0382,
+      "loss": 2.0429,
       "step": 2000
     },
     {
       "epoch": 13.79,
-      "eval_loss": 1.814583659172058,
-      "eval_runtime": 62.7874,
-      "eval_samples_per_second": 26.757,
-      "eval_steps_per_second": 26.757,
-      "eval_wer": 0.8440493418785748,
+      "eval_loss": 1.812680721282959,
+      "eval_runtime": 63.4889,
+      "eval_samples_per_second": 26.461,
+      "eval_steps_per_second": 26.461,
+      "eval_wer": 0.8641720074426297,
       "step": 2000
     },
     {
       "epoch": 13.86,
       "learning_rate": 4.7e-05,
-      "loss": 1.5853,
+      "loss": 1.5885,
       "step": 2010
     },
     {
       "epoch": 13.93,
       "learning_rate": 4.647368421052632e-05,
-      "loss": 1.6782,
+      "loss": 1.6708,
       "step": 2020
     },
     {
       "epoch": 14.0,
       "learning_rate": 4.594736842105264e-05,
-      "loss": 1.9267,
+      "loss": 1.9297,
       "step": 2030
     },
     {
       "epoch": 14.07,
       "learning_rate": 4.542105263157895e-05,
-      "loss": 1.854,
+      "loss": 1.8333,
       "step": 2040
     },
     {
       "epoch": 14.14,
       "learning_rate": 4.489473684210527e-05,
-      "loss": 1.5179,
+      "loss": 1.5051,
       "step": 2050
     },
     {
       "epoch": 14.21,
       "learning_rate": 4.436842105263158e-05,
-      "loss": 1.776,
+      "loss": 1.7685,
       "step": 2060
     },
     {
       "epoch": 14.28,
       "learning_rate": 4.3842105263157895e-05,
-      "loss": 2.1206,
+      "loss": 2.1308,
       "step": 2070
     },
     {
       "epoch": 14.34,
       "learning_rate": 4.3315789473684215e-05,
-      "loss": 1.679,
+      "loss": 1.671,
       "step": 2080
     },
     {
       "epoch": 14.41,
       "learning_rate": 4.278947368421053e-05,
-      "loss": 1.6648,
+      "loss": 1.6557,
       "step": 2090
     },
     {
       "epoch": 14.48,
       "learning_rate": 4.226315789473684e-05,
-      "loss": 1.8785,
+      "loss": 1.8833,
       "step": 2100
     },
     {
       "epoch": 14.48,
-      "eval_loss": 1.8094158172607422,
-      "eval_runtime": 62.7821,
-      "eval_samples_per_second": 26.759,
-      "eval_steps_per_second": 26.759,
-      "eval_wer": 0.8202053614499345,
+      "eval_loss": 1.80045747756958,
+      "eval_runtime": 63.6753,
+      "eval_samples_per_second": 26.384,
+      "eval_steps_per_second": 26.384,
+      "eval_wer": 0.8307490868995934,
       "step": 2100
     },
     {
       "epoch": 14.55,
       "learning_rate": 4.1736842105263154e-05,
-      "loss": 1.9647,
+      "loss": 1.9599,
       "step": 2110
     },
     {
       "epoch": 14.62,
       "learning_rate": 4.1210526315789474e-05,
-      "loss": 1.5669,
+      "loss": 1.5676,
       "step": 2120
     },
     {
       "epoch": 14.69,
       "learning_rate": 4.0684210526315794e-05,
-      "loss": 1.7738,
+      "loss": 1.7685,
       "step": 2130
     },
     {
       "epoch": 14.76,
       "learning_rate": 4.015789473684211e-05,
-      "loss": 2.025,
+      "loss": 2.0334,
       "step": 2140
     },
     {
       "epoch": 14.83,
       "learning_rate": 3.9631578947368426e-05,
-      "loss": 1.7119,
+      "loss": 1.7082,
       "step": 2150
     },
     {
       "epoch": 14.9,
       "learning_rate": 3.910526315789474e-05,
-      "loss": 1.5006,
+      "loss": 1.4858,
       "step": 2160
     },
     {
       "epoch": 14.97,
       "learning_rate": 3.857894736842105e-05,
-      "loss": 1.7046,
+      "loss": 1.6838,
       "step": 2170
     },
     {
       "epoch": 15.03,
       "learning_rate": 3.8052631578947365e-05,
-      "loss": 2.0324,
+      "loss": 2.0267,
       "step": 2180
     },
     {
       "epoch": 15.1,
       "learning_rate": 3.7526315789473685e-05,
-      "loss": 1.574,
+      "loss": 1.5655,
       "step": 2190
     },
     {
       "epoch": 15.17,
       "learning_rate": 3.7e-05,
-      "loss": 1.6148,
+      "loss": 1.5996,
       "step": 2200
     },
     {
       "epoch": 15.17,
-      "eval_loss": 1.8131372928619385,
-      "eval_runtime": 63.0605,
-      "eval_samples_per_second": 26.641,
-      "eval_steps_per_second": 26.641,
-      "eval_wer": 0.8233753704086555,
+      "eval_loss": 1.7925808429718018,
+      "eval_runtime": 63.6696,
+      "eval_samples_per_second": 26.386,
+      "eval_steps_per_second": 26.386,
+      "eval_wer": 0.8467369581696644,
       "step": 2200
     },
     {
       "epoch": 15.24,
       "learning_rate": 3.647368421052632e-05,
-      "loss": 1.8876,
+      "loss": 1.8772,
       "step": 2210
     },
     {
       "epoch": 15.31,
       "learning_rate": 3.594736842105264e-05,
-      "loss": 1.8269,
+      "loss": 1.803,
       "step": 2220
     },
     {
       "epoch": 15.38,
       "learning_rate": 3.542105263157895e-05,
-      "loss": 1.5885,
+      "loss": 1.5784,
       "step": 2230
     },
     {
       "epoch": 15.45,
       "learning_rate": 3.4894736842105264e-05,
-      "loss": 1.6395,
+      "loss": 1.6237,
       "step": 2240
     },
     {
       "epoch": 15.52,
       "learning_rate": 3.436842105263158e-05,
-      "loss": 2.0412,
+      "loss": 2.0538,
       "step": 2250
     },
     {
       "epoch": 15.59,
       "learning_rate": 3.3842105263157896e-05,
-      "loss": 1.5959,
+      "loss": 1.5915,
       "step": 2260
     },
     {
       "epoch": 15.66,
       "learning_rate": 3.331578947368421e-05,
-      "loss": 1.5854,
+      "loss": 1.5799,
       "step": 2270
     },
     {
       "epoch": 15.72,
       "learning_rate": 3.278947368421052e-05,
-      "loss": 1.8598,
+      "loss": 1.8569,
       "step": 2280
     },
     {
       "epoch": 15.79,
       "learning_rate": 3.226315789473684e-05,
-      "loss": 2.0513,
+      "loss": 2.054,
       "step": 2290
     },
     {
       "epoch": 15.86,
       "learning_rate": 3.173684210526316e-05,
-      "loss": 1.4948,
+      "loss": 1.4876,
       "step": 2300
     },
     {
       "epoch": 15.86,
-      "eval_loss": 1.7968575954437256,
-      "eval_runtime": 62.8797,
-      "eval_samples_per_second": 26.718,
-      "eval_steps_per_second": 26.718,
-      "eval_wer": 0.8255805940321136,
+      "eval_loss": 1.779452919960022,
+      "eval_runtime": 63.3412,
+      "eval_samples_per_second": 26.523,
+      "eval_steps_per_second": 26.523,
+      "eval_wer": 0.8341258355730136,
       "step": 2300
     },
     {
       "epoch": 15.93,
       "learning_rate": 3.1210526315789475e-05,
-      "loss": 1.609,
+      "loss": 1.602,
       "step": 2310
     },
     {
       "epoch": 16.0,
       "learning_rate": 3.0684210526315795e-05,
-      "loss": 1.9741,
+      "loss": 1.989,
       "step": 2320
     },
     {
       "epoch": 16.07,
       "learning_rate": 3.0157894736842108e-05,
-      "loss": 1.805,
+      "loss": 1.8007,
       "step": 2330
     },
     {
       "epoch": 16.14,
       "learning_rate": 2.963157894736842e-05,
-      "loss": 1.489,
+      "loss": 1.4876,
       "step": 2340
     },
     {
       "epoch": 16.21,
       "learning_rate": 2.910526315789474e-05,
-      "loss": 1.6549,
+      "loss": 1.666,
       "step": 2350
     },
     {
       "epoch": 16.28,
       "learning_rate": 2.8578947368421057e-05,
-      "loss": 2.0304,
+      "loss": 2.0479,
       "step": 2360
     },
     {
       "epoch": 16.34,
       "learning_rate": 2.805263157894737e-05,
-      "loss": 1.5466,
+      "loss": 1.5446,
       "step": 2370
     },
     {
       "epoch": 16.41,
       "learning_rate": 2.7526315789473683e-05,
-      "loss": 1.5875,
+      "loss": 1.5795,
       "step": 2380
     },
     {
       "epoch": 16.48,
       "learning_rate": 2.7000000000000002e-05,
-      "loss": 1.7809,
+      "loss": 1.7922,
       "step": 2390
     },
     {
       "epoch": 16.55,
       "learning_rate": 2.647368421052632e-05,
-      "loss": 1.8844,
+      "loss": 1.8925,
       "step": 2400
     },
     {
       "epoch": 16.55,
-      "eval_loss": 1.7790474891662598,
-      "eval_runtime": 62.917,
-      "eval_samples_per_second": 26.702,
-      "eval_steps_per_second": 26.702,
-      "eval_wer": 0.8066983667562538,
+      "eval_loss": 1.7715719938278198,
+      "eval_runtime": 63.5709,
+      "eval_samples_per_second": 26.427,
+      "eval_steps_per_second": 26.427,
+      "eval_wer": 0.8198607952587692,
       "step": 2400
     },
     {
       "epoch": 16.62,
       "learning_rate": 2.5947368421052632e-05,
-      "loss": 1.4775,
+      "loss": 1.4745,
       "step": 2410
     },
     {
       "epoch": 16.69,
       "learning_rate": 2.542105263157895e-05,
-      "loss": 1.6398,
+      "loss": 1.6365,
       "step": 2420
     },
     {
       "epoch": 16.76,
       "learning_rate": 2.4894736842105264e-05,
-      "loss": 1.9707,
+      "loss": 1.9693,
       "step": 2430
     },
     {
       "epoch": 16.83,
       "learning_rate": 2.436842105263158e-05,
-      "loss": 1.6675,
+      "loss": 1.6706,
       "step": 2440
     },
     {
       "epoch": 16.9,
       "learning_rate": 2.3842105263157897e-05,
-      "loss": 1.5405,
+      "loss": 1.5347,
       "step": 2450
     },
     {
       "epoch": 16.97,
       "learning_rate": 2.331578947368421e-05,
-      "loss": 1.6781,
+      "loss": 1.68,
       "step": 2460
     },
     {
       "epoch": 17.03,
       "learning_rate": 2.2789473684210527e-05,
-      "loss": 1.993,
+      "loss": 1.9789,
       "step": 2470
     },
     {
       "epoch": 17.1,
       "learning_rate": 2.2263157894736843e-05,
-      "loss": 1.4809,
+      "loss": 1.4876,
       "step": 2480
     },
     {
       "epoch": 17.17,
       "learning_rate": 2.173684210526316e-05,
-      "loss": 1.5379,
+      "loss": 1.5282,
       "step": 2490
     },
     {
       "epoch": 17.24,
       "learning_rate": 2.1210526315789476e-05,
-      "loss": 1.8099,
+      "loss": 1.814,
       "step": 2500
     },
     {
       "epoch": 17.24,
-      "eval_loss": 1.7782752513885498,
-      "eval_runtime": 62.8161,
-      "eval_samples_per_second": 26.745,
-      "eval_steps_per_second": 26.745,
-      "eval_wer": 0.8073185859003514,
+      "eval_loss": 1.7846348285675049,
+      "eval_runtime": 63.8548,
+      "eval_samples_per_second": 26.31,
+      "eval_steps_per_second": 26.31,
+      "eval_wer": 0.8085590241885466,
       "step": 2500
     },
     {
       "epoch": 17.31,
       "learning_rate": 2.068421052631579e-05,
-      "loss": 1.8612,
+      "loss": 1.8587,
       "step": 2510
     },
     {
       "epoch": 17.38,
       "learning_rate": 2.0157894736842105e-05,
-      "loss": 1.4152,
+      "loss": 1.4217,
       "step": 2520
     },
     {
       "epoch": 17.45,
       "learning_rate": 1.963157894736842e-05,
-      "loss": 1.6037,
+      "loss": 1.6074,
       "step": 2530
     },
     {
       "epoch": 17.52,
       "learning_rate": 1.9105263157894738e-05,
-      "loss": 1.9064,
+      "loss": 1.9219,
       "step": 2540
     },
     {
       "epoch": 17.59,
       "learning_rate": 1.8578947368421054e-05,
-      "loss": 1.6072,
+      "loss": 1.6039,
       "step": 2550
     },
     {
       "epoch": 17.66,
       "learning_rate": 1.8052631578947367e-05,
-      "loss": 1.5197,
+      "loss": 1.506,
       "step": 2560
     },
     {
       "epoch": 17.72,
       "learning_rate": 1.7526315789473683e-05,
-      "loss": 1.745,
+      "loss": 1.7399,
       "step": 2570
     },
     {
       "epoch": 17.79,
       "learning_rate": 1.7000000000000003e-05,
-      "loss": 1.89,
+      "loss": 1.8959,
       "step": 2580
     },
     {
       "epoch": 17.86,
       "learning_rate": 1.6473684210526316e-05,
-      "loss": 1.4768,
+      "loss": 1.4699,
       "step": 2590
     },
     {
       "epoch": 17.93,
       "learning_rate": 1.5947368421052633e-05,
-      "loss": 1.5488,
+      "loss": 1.536,
       "step": 2600
     },
     {
       "epoch": 17.93,
-      "eval_loss": 1.766753911972046,
-      "eval_runtime": 62.8076,
-      "eval_samples_per_second": 26.748,
-      "eval_steps_per_second": 26.748,
-      "eval_wer": 0.7971194266418579,
+      "eval_loss": 1.7655223608016968,
+      "eval_runtime": 63.3968,
+      "eval_samples_per_second": 26.5,
+      "eval_steps_per_second": 26.5,
+      "eval_wer": 0.8018744400799394,
       "step": 2600
     },
     {
       "epoch": 18.0,
       "learning_rate": 1.5421052631578946e-05,
-      "loss": 1.9043,
+      "loss": 1.9096,
       "step": 2610
     },
     {
       "epoch": 18.07,
       "learning_rate": 1.4894736842105264e-05,
-      "loss": 1.6967,
+      "loss": 1.6969,
       "step": 2620
     },
     {
       "epoch": 18.14,
       "learning_rate": 1.4368421052631582e-05,
-      "loss": 1.4113,
+      "loss": 1.4026,
       "step": 2630
     },
     {
       "epoch": 18.21,
       "learning_rate": 1.3842105263157895e-05,
-      "loss": 1.616,
+      "loss": 1.611,
       "step": 2640
     },
     {
       "epoch": 18.28,
       "learning_rate": 1.3315789473684213e-05,
-      "loss": 1.9563,
+      "loss": 1.947,
       "step": 2650
     },
     {
       "epoch": 18.34,
       "learning_rate": 1.2789473684210526e-05,
-      "loss": 1.4982,
+      "loss": 1.5006,
       "step": 2660
     },
     {
       "epoch": 18.41,
       "learning_rate": 1.2263157894736844e-05,
-      "loss": 1.4753,
+      "loss": 1.4805,
       "step": 2670
     },
     {
       "epoch": 18.48,
       "learning_rate": 1.1736842105263158e-05,
-      "loss": 1.777,
+      "loss": 1.7787,
       "step": 2680
     },
     {
       "epoch": 18.55,
       "learning_rate": 1.1210526315789475e-05,
-      "loss": 1.801,
+      "loss": 1.8082,
       "step": 2690
     },
     {
       "epoch": 18.62,
       "learning_rate": 1.068421052631579e-05,
-      "loss": 1.458,
+      "loss": 1.4476,
       "step": 2700
     },
     {
       "epoch": 18.62,
-      "eval_loss": 1.7622854709625244,
-      "eval_runtime": 63.2501,
-      "eval_samples_per_second": 26.561,
-      "eval_steps_per_second": 26.561,
-      "eval_wer": 0.7973261663565571,
+      "eval_loss": 1.7598618268966675,
+      "eval_runtime": 63.6208,
+      "eval_samples_per_second": 26.406,
+      "eval_steps_per_second": 26.406,
+      "eval_wer": 0.8070429329474192,
       "step": 2700
     },
     {
       "epoch": 18.69,
       "learning_rate": 1.0157894736842106e-05,
-      "loss": 1.5744,
+      "loss": 1.5683,
       "step": 2710
     },
     {
       "epoch": 18.76,
       "learning_rate": 9.631578947368422e-06,
-      "loss": 1.9091,
+      "loss": 1.9225,
       "step": 2720
     },
     {
       "epoch": 18.83,
       "learning_rate": 9.105263157894737e-06,
-      "loss": 1.5814,
+      "loss": 1.5732,
       "step": 2730
     },
     {
       "epoch": 18.9,
       "learning_rate": 8.578947368421053e-06,
-      "loss": 1.5344,
+      "loss": 1.5404,
       "step": 2740
     },
     {
       "epoch": 18.97,
       "learning_rate": 8.052631578947368e-06,
-      "loss": 1.6318,
+      "loss": 1.6251,
       "step": 2750
     },
     {
       "epoch": 19.03,
       "learning_rate": 7.526315789473684e-06,
-      "loss": 1.8497,
+      "loss": 1.8527,
       "step": 2760
     },
     {
       "epoch": 19.1,
       "learning_rate": 7.000000000000001e-06,
-      "loss": 1.4525,
+      "loss": 1.4474,
       "step": 2770
     },
     {
       "epoch": 19.17,
       "learning_rate": 6.473684210526316e-06,
-      "loss": 1.467,
+      "loss": 1.475,
       "step": 2780
     },
     {
       "epoch": 19.24,
       "learning_rate": 5.947368421052632e-06,
-      "loss": 1.8153,
+      "loss": 1.8146,
       "step": 2790
     },
     {
       "epoch": 19.31,
       "learning_rate": 5.421052631578947e-06,
-      "loss": 1.7656,
+      "loss": 1.7629,
       "step": 2800
     },
     {
       "epoch": 19.31,
-      "eval_loss": 1.7574056386947632,
-      "eval_runtime": 62.7899,
-      "eval_samples_per_second": 26.756,
-      "eval_steps_per_second": 26.756,
-      "eval_wer": 0.8012542209358418,
+      "eval_loss": 1.7588989734649658,
+      "eval_runtime": 63.459,
+      "eval_samples_per_second": 26.474,
+      "eval_steps_per_second": 26.474,
+      "eval_wer": 0.8119357728619668,
       "step": 2800
     },
     {
       "epoch": 19.38,
       "learning_rate": 4.894736842105263e-06,
-      "loss": 1.4587,
+      "loss": 1.4493,
       "step": 2810
     },
     {
       "epoch": 19.45,
       "learning_rate": 4.368421052631579e-06,
-      "loss": 1.5571,
+      "loss": 1.5521,
       "step": 2820
     },
     {
       "epoch": 19.52,
       "learning_rate": 3.842105263157895e-06,
-      "loss": 1.9397,
+      "loss": 1.9465,
       "step": 2830
     },
     {
       "epoch": 19.59,
       "learning_rate": 3.315789473684211e-06,
-      "loss": 1.498,
+      "loss": 1.4912,
       "step": 2840
     },
     {
       "epoch": 19.66,
       "learning_rate": 2.7894736842105266e-06,
-      "loss": 1.4464,
+      "loss": 1.4484,
       "step": 2850
     },
     {
       "epoch": 19.72,
       "learning_rate": 2.2631578947368426e-06,
-      "loss": 1.6899,
+      "loss": 1.6934,
       "step": 2860
     },
     {
       "epoch": 19.79,
       "learning_rate": 1.7368421052631579e-06,
-      "loss": 1.8521,
+      "loss": 1.8646,
       "step": 2870
     },
     {
       "epoch": 19.86,
       "learning_rate": 1.2105263157894738e-06,
-      "loss": 1.4235,
+      "loss": 1.4162,
       "step": 2880
     },
     {
       "epoch": 19.93,
       "learning_rate": 6.842105263157896e-07,
-      "loss": 1.5813,
+      "loss": 1.5734,
       "step": 2890
     },
     {
       "epoch": 20.0,
       "learning_rate": 1.5789473684210527e-07,
-      "loss": 1.7583,
+      "loss": 1.7646,
       "step": 2900
     },
     {
       "epoch": 20.0,
-      "eval_loss": 1.7561472654342651,
-      "eval_runtime": 63.1867,
-      "eval_samples_per_second": 26.588,
-      "eval_steps_per_second": 26.588,
-      "eval_wer": 0.7971194266418579,
+      "eval_loss": 1.754138469696045,
+      "eval_runtime": 63.415,
+      "eval_samples_per_second": 26.492,
+      "eval_steps_per_second": 26.492,
+      "eval_wer": 0.8060781476121563,
       "step": 2900
     },
     {
       "epoch": 20.0,
       "step": 2900,
       "total_flos": 1.1463619274186412e+18,
-      "train_loss": 2.3102814756590746,
-      "train_runtime": 4074.7961,
-      "train_samples_per_second": 22.676,
-      "train_steps_per_second": 0.712
+      "train_loss": 2.314238773214406,
+      "train_runtime": 4067.0318,
+      "train_samples_per_second": 22.719,
+      "train_steps_per_second": 0.713
     }
   ],
   "max_steps": 2900,