End of training

all_results.json

@@ -1,14 +1,14 @@
 {
     "epoch": 50.0,
-    "eval_loss": 0.7062973976135254,
-    "eval_runtime": 74.685,
+    "eval_loss": 0.5196508765220642,
+    "eval_runtime": 80.8909,
     "eval_samples": 1700,
-    "eval_samples_per_second": 22.762,
-    "eval_steps_per_second": 1.433,
-    "eval_wer": 0.5557036754903828,
-    "train_loss": 0.7598760600297347,
-    "train_runtime": 13000.5685,
+    "eval_samples_per_second": 21.016,
+    "eval_steps_per_second": 1.323,
+    "eval_wer": 0.4689265536723164,
+    "train_loss": 1.123646092788033,
+    "train_runtime": 13730.8326,
     "train_samples": 3676,
-    "train_samples_per_second": 14.138,
-    "train_steps_per_second": 0.442
+    "train_samples_per_second": 13.386,
+    "train_steps_per_second": 0.419
 }

eval_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 50.0,
-    "eval_loss": 0.7062973976135254,
-    "eval_runtime": 74.685,
+    "eval_loss": 0.5196508765220642,
+    "eval_runtime": 80.8909,
     "eval_samples": 1700,
-    "eval_samples_per_second": 22.762,
-    "eval_steps_per_second": 1.433,
-    "eval_wer": 0.5557036754903828
+    "eval_samples_per_second": 21.016,
+    "eval_steps_per_second": 1.323,
+    "eval_wer": 0.4689265536723164
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 50.0,
-    "train_loss": 0.7598760600297347,
-    "train_runtime": 13000.5685,
+    "train_loss": 1.123646092788033,
+    "train_runtime": 13730.8326,
     "train_samples": 3676,
-    "train_samples_per_second": 14.138,
-    "train_steps_per_second": 0.442
+    "train_samples_per_second": 13.386,
+    "train_steps_per_second": 0.419
 }

trainer_state.json

@@ -9,525 +9,525 @@
   "log_history": [
     {
       "epoch": 0.87,
-      "learning_rate": 5.7166666666666664e-06,
-      "loss": 13.1691,
+      "learning_rate": 3.3949999999999997e-06,
+      "loss": 13.6823,
       "step": 100
     },
     {
       "epoch": 1.74,
-      "learning_rate": 1.155e-05,
-      "loss": 5.7916,
+      "learning_rate": 6.895e-06,
+      "loss": 7.5854,
       "step": 200
     },
     {
       "epoch": 2.61,
-      "learning_rate": 1.738333333333333e-05,
-      "loss": 3.9554,
+      "learning_rate": 1.0394999999999998e-05,
+      "loss": 4.3711,
       "step": 300
     },
     {
       "epoch": 2.61,
-      "eval_loss": 3.8166096210479736,
-      "eval_runtime": 76.9069,
-      "eval_samples_per_second": 22.105,
-      "eval_steps_per_second": 1.391,
+      "eval_loss": 4.312221050262451,
+      "eval_runtime": 80.91,
+      "eval_samples_per_second": 21.011,
+      "eval_steps_per_second": 1.322,
       "eval_wer": 1.0,
       "step": 300
     },
     {
       "epoch": 3.48,
-      "learning_rate": 2.3216666666666664e-05,
-      "loss": 3.4177,
+      "learning_rate": 1.3895e-05,
+      "loss": 3.8129,
       "step": 400
     },
     {
       "epoch": 4.35,
-      "learning_rate": 2.9049999999999995e-05,
-      "loss": 3.1088,
+      "learning_rate": 1.7395e-05,
+      "loss": 3.4258,
       "step": 500
     },
     {
       "epoch": 5.22,
-      "learning_rate": 3.488333333333333e-05,
-      "loss": 2.9853,
+      "learning_rate": 2.0894999999999996e-05,
+      "loss": 3.1653,
       "step": 600
     },
     {
       "epoch": 5.22,
-      "eval_loss": 2.979321002960205,
-      "eval_runtime": 76.5995,
-      "eval_samples_per_second": 22.193,
-      "eval_steps_per_second": 1.397,
+      "eval_loss": 3.115588426589966,
+      "eval_runtime": 81.1622,
+      "eval_samples_per_second": 20.946,
+      "eval_steps_per_second": 1.318,
       "eval_wer": 1.0,
       "step": 600
     },
     {
       "epoch": 6.09,
-      "learning_rate": 4.071666666666666e-05,
-      "loss": 2.6633,
+      "learning_rate": 2.4394999999999996e-05,
+      "loss": 3.0356,
       "step": 700
     },
     {
       "epoch": 6.96,
-      "learning_rate": 4.655e-05,
-      "loss": 1.3761,
+      "learning_rate": 2.7895e-05,
+      "loss": 2.9791,
       "step": 800
     },
     {
       "epoch": 7.83,
-      "learning_rate": 5.2383333333333324e-05,
-      "loss": 0.6939,
+      "learning_rate": 3.1395e-05,
+      "loss": 2.8904,
       "step": 900
     },
     {
       "epoch": 7.83,
-      "eval_loss": 0.972253143787384,
-      "eval_runtime": 77.2431,
-      "eval_samples_per_second": 22.008,
-      "eval_steps_per_second": 1.385,
-      "eval_wer": 0.9149368374277915,
+      "eval_loss": 2.842055320739746,
+      "eval_runtime": 84.1755,
+      "eval_samples_per_second": 20.196,
+      "eval_steps_per_second": 1.271,
+      "eval_wer": 0.9918110836031232,
       "step": 900
     },
     {
       "epoch": 8.7,
-      "learning_rate": 5.821666666666666e-05,
-      "loss": 0.5141,
+      "learning_rate": 3.4895e-05,
+      "loss": 2.1422,
       "step": 1000
     },
     {
       "epoch": 9.57,
-      "learning_rate": 6.405e-05,
-      "loss": 0.4189,
+      "learning_rate": 3.8394999999999994e-05,
+      "loss": 1.2257,
       "step": 1100
     },
     {
       "epoch": 10.43,
-      "learning_rate": 6.988333333333333e-05,
-      "loss": 0.352,
+      "learning_rate": 4.1895e-05,
+      "loss": 0.9207,
       "step": 1200
     },
     {
       "epoch": 10.43,
-      "eval_loss": 0.6556435823440552,
-      "eval_runtime": 76.3866,
-      "eval_samples_per_second": 22.255,
-      "eval_steps_per_second": 1.401,
-      "eval_wer": 0.7296387989589285,
+      "eval_loss": 0.9894591569900513,
+      "eval_runtime": 82.3044,
+      "eval_samples_per_second": 20.655,
+      "eval_steps_per_second": 1.3,
+      "eval_wer": 0.8688503777058338,
       "step": 1200
     },
     {
       "epoch": 11.3,
-      "learning_rate": 6.849230769230769e-05,
-      "loss": 0.307,
+      "learning_rate": 4.5394999999999995e-05,
+      "loss": 0.7881,
       "step": 1300
     },
     {
       "epoch": 12.17,
-      "learning_rate": 6.695384615384615e-05,
-      "loss": 0.2658,
+      "learning_rate": 4.8895e-05,
+      "loss": 0.7047,
       "step": 1400
     },
     {
       "epoch": 13.04,
-      "learning_rate": 6.541538461538461e-05,
-      "loss": 0.2369,
+      "learning_rate": 5.2395e-05,
+      "loss": 0.6384,
       "step": 1500
     },
     {
       "epoch": 13.04,
-      "eval_loss": 0.6314664483070374,
-      "eval_runtime": 76.2318,
-      "eval_samples_per_second": 22.3,
-      "eval_steps_per_second": 1.404,
-      "eval_wer": 0.6927569351869485,
+      "eval_loss": 0.6993927359580994,
+      "eval_runtime": 82.4631,
+      "eval_samples_per_second": 20.615,
+      "eval_steps_per_second": 1.298,
+      "eval_wer": 0.7700120611946931,
       "step": 1500
     },
     {
       "epoch": 13.91,
-      "learning_rate": 6.387692307692307e-05,
-      "loss": 0.2084,
+      "learning_rate": 5.589499999999999e-05,
+      "loss": 0.5989,
       "step": 1600
     },
     {
       "epoch": 14.78,
-      "learning_rate": 6.233846153846153e-05,
-      "loss": 0.1904,
+      "learning_rate": 5.9394999999999996e-05,
+      "loss": 0.5601,
       "step": 1700
     },
     {
       "epoch": 15.65,
-      "learning_rate": 6.0799999999999994e-05,
-      "loss": 0.1757,
+      "learning_rate": 6.289499999999999e-05,
+      "loss": 0.5215,
       "step": 1800
     },
     {
       "epoch": 15.65,
-      "eval_loss": 0.6387060284614563,
-      "eval_runtime": 76.1573,
-      "eval_samples_per_second": 22.322,
-      "eval_steps_per_second": 1.405,
-      "eval_wer": 0.6723798641528598,
+      "eval_loss": 0.5627515912055969,
+      "eval_runtime": 81.3944,
+      "eval_samples_per_second": 20.886,
+      "eval_steps_per_second": 1.315,
+      "eval_wer": 0.6443217164984447,
       "step": 1800
     },
     {
       "epoch": 16.52,
-      "learning_rate": 5.9261538461538453e-05,
-      "loss": 0.1693,
+      "learning_rate": 6.639499999999999e-05,
+      "loss": 0.4949,
       "step": 1900
     },
     {
       "epoch": 17.39,
-      "learning_rate": 5.772307692307692e-05,
-      "loss": 0.1572,
+      "learning_rate": 6.9895e-05,
+      "loss": 0.4869,
       "step": 2000
     },
     {
       "epoch": 18.26,
-      "learning_rate": 5.618461538461538e-05,
-      "loss": 0.1517,
+      "learning_rate": 6.818933333333333e-05,
+      "loss": 0.4573,
       "step": 2100
     },
     {
       "epoch": 18.26,
-      "eval_loss": 0.6607237458229065,
-      "eval_runtime": 77.027,
-      "eval_samples_per_second": 22.07,
-      "eval_steps_per_second": 1.389,
-      "eval_wer": 0.6265473243191773,
+      "eval_loss": 0.5316212177276611,
+      "eval_runtime": 82.2602,
+      "eval_samples_per_second": 20.666,
+      "eval_steps_per_second": 1.301,
+      "eval_wer": 0.6174062083412684,
       "step": 2100
     },
     {
       "epoch": 19.13,
-      "learning_rate": 5.464615384615384e-05,
-      "loss": 0.1403,
+      "learning_rate": 6.632266666666666e-05,
+      "loss": 0.438,
       "step": 2200
     },
     {
       "epoch": 20.0,
-      "learning_rate": 5.3107692307692305e-05,
-      "loss": 0.1382,
+      "learning_rate": 6.4456e-05,
+      "loss": 0.4153,
       "step": 2300
     },
     {
       "epoch": 20.87,
-      "learning_rate": 5.1569230769230765e-05,
-      "loss": 0.1229,
+      "learning_rate": 6.258933333333333e-05,
+      "loss": 0.3875,
       "step": 2400
     },
     {
       "epoch": 20.87,
-      "eval_loss": 0.6519985198974609,
-      "eval_runtime": 76.1295,
-      "eval_samples_per_second": 22.33,
-      "eval_steps_per_second": 1.405,
-      "eval_wer": 0.6182949279502317,
+      "eval_loss": 0.4931696653366089,
+      "eval_runtime": 80.8801,
+      "eval_samples_per_second": 21.019,
+      "eval_steps_per_second": 1.323,
+      "eval_wer": 0.5778581857423982,
       "step": 2400
     },
     {
       "epoch": 21.74,
-      "learning_rate": 5.0030769230769225e-05,
-      "loss": 0.129,
+      "learning_rate": 6.072266666666667e-05,
+      "loss": 0.3807,
       "step": 2500
     },
     {
       "epoch": 22.61,
-      "learning_rate": 4.849230769230769e-05,
-      "loss": 0.1172,
+      "learning_rate": 5.8855999999999993e-05,
+      "loss": 0.3715,
       "step": 2600
     },
     {
       "epoch": 23.48,
-      "learning_rate": 4.695384615384615e-05,
-      "loss": 0.1201,
+      "learning_rate": 5.6989333333333333e-05,
+      "loss": 0.3562,
       "step": 2700
     },
     {
       "epoch": 23.48,
-      "eval_loss": 0.67499840259552,
-      "eval_runtime": 76.4187,
-      "eval_samples_per_second": 22.246,
-      "eval_steps_per_second": 1.4,
-      "eval_wer": 0.6115660509109376,
+      "eval_loss": 0.4971640110015869,
+      "eval_runtime": 82.278,
+      "eval_samples_per_second": 20.662,
+      "eval_steps_per_second": 1.3,
+      "eval_wer": 0.547514759093506,
       "step": 2700
     },
     {
       "epoch": 24.35,
-      "learning_rate": 4.541538461538461e-05,
-      "loss": 0.1112,
+      "learning_rate": 5.512266666666666e-05,
+      "loss": 0.3457,
       "step": 2800
     },
     {
       "epoch": 25.22,
-      "learning_rate": 4.387692307692307e-05,
-      "loss": 0.1131,
+      "learning_rate": 5.3256e-05,
+      "loss": 0.3356,
       "step": 2900
     },
     {
       "epoch": 26.09,
-      "learning_rate": 4.2338461538461536e-05,
-      "loss": 0.1076,
+      "learning_rate": 5.1389333333333326e-05,
+      "loss": 0.3218,
       "step": 3000
     },
     {
       "epoch": 26.09,
-      "eval_loss": 0.6698060035705566,
-      "eval_runtime": 75.169,
-      "eval_samples_per_second": 22.616,
-      "eval_steps_per_second": 1.423,
-      "eval_wer": 0.6006474957151019,
+      "eval_loss": 0.4894775450229645,
+      "eval_runtime": 81.4053,
+      "eval_samples_per_second": 20.883,
+      "eval_steps_per_second": 1.314,
+      "eval_wer": 0.5219323303497746,
       "step": 3000
     },
     {
       "epoch": 26.96,
-      "learning_rate": 4.0799999999999996e-05,
-      "loss": 0.104,
+      "learning_rate": 4.9522666666666666e-05,
+      "loss": 0.3072,
       "step": 3100
     },
     {
       "epoch": 27.83,
-      "learning_rate": 3.9261538461538455e-05,
-      "loss": 0.1014,
+      "learning_rate": 4.765599999999999e-05,
+      "loss": 0.3006,
       "step": 3200
     },
     {
       "epoch": 28.7,
-      "learning_rate": 3.772307692307692e-05,
-      "loss": 0.1006,
+      "learning_rate": 4.578933333333333e-05,
+      "loss": 0.2954,
       "step": 3300
     },
     {
       "epoch": 28.7,
-      "eval_loss": 0.680385947227478,
-      "eval_runtime": 75.8528,
-      "eval_samples_per_second": 22.412,
-      "eval_steps_per_second": 1.411,
-      "eval_wer": 0.5871897416365137,
+      "eval_loss": 0.5226009488105774,
+      "eval_runtime": 82.2645,
+      "eval_samples_per_second": 20.665,
+      "eval_steps_per_second": 1.301,
+      "eval_wer": 0.5192026915508157,
       "step": 3300
     },
     {
       "epoch": 29.57,
-      "learning_rate": 3.618461538461538e-05,
-      "loss": 0.1013,
+      "learning_rate": 4.392266666666666e-05,
+      "loss": 0.2965,
       "step": 3400
     },
     {
       "epoch": 30.43,
-      "learning_rate": 3.464615384615384e-05,
-      "loss": 0.0972,
+      "learning_rate": 4.2056e-05,
+      "loss": 0.286,
       "step": 3500
     },
     {
       "epoch": 31.3,
-      "learning_rate": 3.310769230769231e-05,
-      "loss": 0.0952,
+      "learning_rate": 4.018933333333333e-05,
+      "loss": 0.287,
       "step": 3600
     },
     {
       "epoch": 31.3,
-      "eval_loss": 0.7009897828102112,
-      "eval_runtime": 74.9836,
-      "eval_samples_per_second": 22.672,
-      "eval_steps_per_second": 1.427,
-      "eval_wer": 0.600266615882689,
+      "eval_loss": 0.495715469121933,
+      "eval_runtime": 79.9357,
+      "eval_samples_per_second": 21.267,
+      "eval_steps_per_second": 1.339,
+      "eval_wer": 0.5145686535897924,
       "step": 3600
     },
     {
       "epoch": 32.17,
-      "learning_rate": 3.156923076923077e-05,
-      "loss": 0.091,
+      "learning_rate": 3.8322666666666665e-05,
+      "loss": 0.2768,
       "step": 3700
     },
     {
       "epoch": 33.04,
-      "learning_rate": 3.0030769230769226e-05,
-      "loss": 0.0924,
+      "learning_rate": 3.6456e-05,
+      "loss": 0.2731,
       "step": 3800
     },
     {
       "epoch": 33.91,
-      "learning_rate": 2.849230769230769e-05,
-      "loss": 0.0894,
+      "learning_rate": 3.458933333333333e-05,
+      "loss": 0.2587,
       "step": 3900
     },
     {
       "epoch": 33.91,
-      "eval_loss": 0.7068695425987244,
-      "eval_runtime": 74.6323,
-      "eval_samples_per_second": 22.778,
-      "eval_steps_per_second": 1.434,
-      "eval_wer": 0.584015743033073,
+      "eval_loss": 0.49437007308006287,
+      "eval_runtime": 82.64,
+      "eval_samples_per_second": 20.571,
+      "eval_steps_per_second": 1.295,
+      "eval_wer": 0.48930362470640515,
       "step": 3900
     },
     {
       "epoch": 34.78,
-      "learning_rate": 2.695384615384615e-05,
-      "loss": 0.0869,
+      "learning_rate": 3.2722666666666664e-05,
+      "loss": 0.2549,
       "step": 4000
     },
     {
       "epoch": 35.65,
-      "learning_rate": 2.5415384615384612e-05,
-      "loss": 0.0848,
+      "learning_rate": 3.0856e-05,
+      "loss": 0.2538,
       "step": 4100
     },
     {
       "epoch": 36.52,
-      "learning_rate": 2.3876923076923075e-05,
-      "loss": 0.0873,
+      "learning_rate": 2.8989333333333334e-05,
+      "loss": 0.2496,
       "step": 4200
     },
     {
       "epoch": 36.52,
-      "eval_loss": 0.6764819622039795,
-      "eval_runtime": 75.5212,
-      "eval_samples_per_second": 22.51,
-      "eval_steps_per_second": 1.417,
-      "eval_wer": 0.5781121056306735,
+      "eval_loss": 0.4975605010986328,
+      "eval_runtime": 82.1209,
+      "eval_samples_per_second": 20.701,
+      "eval_steps_per_second": 1.303,
+      "eval_wer": 0.4894940646226116,
       "step": 4200
     },
     {
       "epoch": 37.39,
-      "learning_rate": 2.2338461538461534e-05,
-      "loss": 0.0845,
+      "learning_rate": 2.7122666666666667e-05,
+      "loss": 0.2461,
       "step": 4300
     },
     {
       "epoch": 38.26,
-      "learning_rate": 2.0799999999999997e-05,
-      "loss": 0.0813,
+      "learning_rate": 2.5256e-05,
+      "loss": 0.2359,
       "step": 4400
     },
     {
       "epoch": 39.13,
-      "learning_rate": 1.9261538461538457e-05,
-      "loss": 0.0798,
+      "learning_rate": 2.3389333333333333e-05,
+      "loss": 0.2365,
       "step": 4500
     },
     {
       "epoch": 39.13,
-      "eval_loss": 0.6821776032447815,
-      "eval_runtime": 76.5806,
-      "eval_samples_per_second": 22.199,
-      "eval_steps_per_second": 1.397,
-      "eval_wer": 0.5617342728369199,
+      "eval_loss": 0.5185123085975647,
+      "eval_runtime": 79.6732,
+      "eval_samples_per_second": 21.337,
+      "eval_steps_per_second": 1.343,
+      "eval_wer": 0.4818764679743541,
       "step": 4500
     },
     {
       "epoch": 40.0,
-      "learning_rate": 1.772307692307692e-05,
-      "loss": 0.0783,
+      "learning_rate": 2.1522666666666666e-05,
+      "loss": 0.2357,
       "step": 4600
     },
     {
       "epoch": 40.87,
-      "learning_rate": 1.6199999999999997e-05,
-      "loss": 0.0753,
+      "learning_rate": 1.9656e-05,
+      "loss": 0.2289,
       "step": 4700
     },
     {
       "epoch": 41.74,
-      "learning_rate": 1.466153846153846e-05,
-      "loss": 0.0767,
+      "learning_rate": 1.7789333333333333e-05,
+      "loss": 0.2264,
       "step": 4800
     },
     {
       "epoch": 41.74,
-      "eval_loss": 0.7003459334373474,
-      "eval_runtime": 74.7077,
-      "eval_samples_per_second": 22.755,
-      "eval_steps_per_second": 1.432,
-      "eval_wer": 0.5637656319431219,
+      "eval_loss": 0.5152125954627991,
+      "eval_runtime": 79.201,
+      "eval_samples_per_second": 21.464,
+      "eval_steps_per_second": 1.351,
+      "eval_wer": 0.47755982987367485,
       "step": 4800
     },
     {
       "epoch": 42.61,
-      "learning_rate": 1.3123076923076922e-05,
-      "loss": 0.0739,
+      "learning_rate": 1.5922666666666666e-05,
+      "loss": 0.2211,
       "step": 4900
     },
     {
       "epoch": 43.48,
-      "learning_rate": 1.1584615384615385e-05,
-      "loss": 0.0766,
+      "learning_rate": 1.4055999999999999e-05,
+      "loss": 0.2186,
       "step": 5000
     },
     {
       "epoch": 44.35,
-      "learning_rate": 1.0046153846153846e-05,
-      "loss": 0.0717,
+      "learning_rate": 1.2189333333333332e-05,
+      "loss": 0.2224,
       "step": 5100
     },
     {
       "epoch": 44.35,
-      "eval_loss": 0.6969868540763855,
-      "eval_runtime": 74.6644,
-      "eval_samples_per_second": 22.769,
-      "eval_steps_per_second": 1.433,
-      "eval_wer": 0.5590681140100299,
+      "eval_loss": 0.5030579566955566,
+      "eval_runtime": 80.9089,
+      "eval_samples_per_second": 21.011,
+      "eval_steps_per_second": 1.322,
+      "eval_wer": 0.4745762711864407,
       "step": 5100
     },
     {
       "epoch": 45.22,
-      "learning_rate": 8.507692307692307e-06,
-      "loss": 0.0732,
+      "learning_rate": 1.0322666666666665e-05,
+      "loss": 0.2162,
       "step": 5200
     },
     {
       "epoch": 46.09,
-      "learning_rate": 6.9692307692307684e-06,
-      "loss": 0.0713,
+      "learning_rate": 8.456e-06,
+      "loss": 0.2159,
       "step": 5300
     },
     {
       "epoch": 46.96,
-      "learning_rate": 5.4307692307692306e-06,
-      "loss": 0.0687,
+      "learning_rate": 6.589333333333332e-06,
+      "loss": 0.2096,
       "step": 5400
     },
     {
       "epoch": 46.96,
-      "eval_loss": 0.7072671055793762,
-      "eval_runtime": 74.5796,
-      "eval_samples_per_second": 22.794,
-      "eval_steps_per_second": 1.435,
-      "eval_wer": 0.560337713451406,
+      "eval_loss": 0.5061585307121277,
+      "eval_runtime": 81.3005,
+      "eval_samples_per_second": 20.91,
+      "eval_steps_per_second": 1.316,
+      "eval_wer": 0.47076747286231196,
       "step": 5400
     },
     {
       "epoch": 47.83,
-      "learning_rate": 3.892307692307692e-06,
-      "loss": 0.0674,
+      "learning_rate": 4.7226666666666654e-06,
+      "loss": 0.205,
       "step": 5500
     },
     {
       "epoch": 48.7,
-      "learning_rate": 2.3538461538461536e-06,
-      "loss": 0.0702,
+      "learning_rate": 2.856e-06,
+      "loss": 0.2038,
       "step": 5600
     },
     {
       "epoch": 49.57,
-      "learning_rate": 8.153846153846153e-07,
-      "loss": 0.0669,
+      "learning_rate": 9.893333333333332e-07,
+      "loss": 0.2038,
       "step": 5700
     },
     {
       "epoch": 49.57,
-      "eval_loss": 0.7059715390205383,
-      "eval_runtime": 74.624,
-      "eval_samples_per_second": 22.781,
-      "eval_steps_per_second": 1.434,
-      "eval_wer": 0.5558941154065892,
+      "eval_loss": 0.5217297077178955,
+      "eval_runtime": 83.7172,
+      "eval_samples_per_second": 20.306,
+      "eval_steps_per_second": 1.278,
+      "eval_wer": 0.46981527328127975,
       "step": 5700
     },
     {
       "epoch": 50.0,
       "step": 5750,
       "total_flos": 2.9609940258263142e+19,
-      "train_loss": 0.7598760600297347,
-      "train_runtime": 13000.5685,
-      "train_samples_per_second": 14.138,
-      "train_steps_per_second": 0.442
+      "train_loss": 1.123646092788033,
+      "train_runtime": 13730.8326,
+      "train_samples_per_second": 13.386,
+      "train_steps_per_second": 0.419
     }
   ],
   "max_steps": 5750,