diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,8 +1,8 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 100.0,
-  "global_step": 6900,
+  "epoch": 200.0,
+  "global_step": 13800,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -3717,566 +3717,4832 @@
     },
     {
       "epoch": 87.1,
-      "learning_rate": 1.5135593220338984e-05,
-      "loss": 0.5942,
+      "learning_rate": 6.088281250000001e-05,
+      "loss": 0.595,
       "step": 6010
     },
     {
       "epoch": 87.25,
-      "learning_rate": 1.4966101694915256e-05,
-      "loss": 0.6446,
+      "learning_rate": 6.08046875e-05,
+      "loss": 0.6517,
       "step": 6020
     },
     {
       "epoch": 87.39,
-      "learning_rate": 1.4796610169491525e-05,
-      "loss": 0.5991,
+      "learning_rate": 6.07265625e-05,
+      "loss": 0.6079,
       "step": 6030
     },
     {
       "epoch": 87.54,
-      "learning_rate": 1.4627118644067797e-05,
-      "loss": 0.6077,
+      "learning_rate": 6.06484375e-05,
+      "loss": 0.6203,
       "step": 6040
     },
     {
       "epoch": 87.68,
-      "learning_rate": 1.4457627118644068e-05,
-      "loss": 0.5856,
+      "learning_rate": 6.05703125e-05,
+      "loss": 0.6005,
       "step": 6050
     },
     {
       "epoch": 87.83,
-      "learning_rate": 1.428813559322034e-05,
-      "loss": 0.6133,
+      "learning_rate": 6.0492187500000006e-05,
+      "loss": 0.6321,
       "step": 6060
     },
     {
       "epoch": 87.97,
-      "learning_rate": 1.411864406779661e-05,
-      "loss": 0.5992,
+      "learning_rate": 6.04140625e-05,
+      "loss": 0.6156,
       "step": 6070
     },
     {
       "epoch": 88.12,
-      "learning_rate": 1.3949152542372881e-05,
-      "loss": 0.6122,
+      "learning_rate": 6.0335937500000005e-05,
+      "loss": 0.6329,
       "step": 6080
     },
     {
       "epoch": 88.26,
-      "learning_rate": 1.3779661016949153e-05,
-      "loss": 0.6178,
+      "learning_rate": 6.02578125e-05,
+      "loss": 0.6311,
       "step": 6090
     },
     {
       "epoch": 88.41,
-      "learning_rate": 1.3610169491525424e-05,
-      "loss": 0.6505,
+      "learning_rate": 6.0179687500000005e-05,
+      "loss": 0.6689,
       "step": 6100
     },
     {
       "epoch": 88.55,
-      "learning_rate": 1.3440677966101695e-05,
-      "loss": 0.5641,
+      "learning_rate": 6.010156250000001e-05,
+      "loss": 0.5824,
       "step": 6110
     },
     {
       "epoch": 88.7,
-      "learning_rate": 1.3271186440677965e-05,
-      "loss": 0.6329,
+      "learning_rate": 6.0023437500000005e-05,
+      "loss": 0.6491,
       "step": 6120
     },
     {
       "epoch": 88.84,
-      "learning_rate": 1.3101694915254237e-05,
-      "loss": 0.6119,
+      "learning_rate": 5.994531250000001e-05,
+      "loss": 0.6326,
       "step": 6130
     },
     {
       "epoch": 88.99,
-      "learning_rate": 1.2932203389830508e-05,
-      "loss": 0.6043,
+      "learning_rate": 5.98671875e-05,
+      "loss": 0.6272,
       "step": 6140
     },
     {
       "epoch": 89.13,
-      "learning_rate": 1.276271186440678e-05,
-      "loss": 0.6661,
+      "learning_rate": 5.97890625e-05,
+      "loss": 0.6865,
       "step": 6150
     },
     {
       "epoch": 89.28,
-      "learning_rate": 1.2593220338983053e-05,
-      "loss": 0.5938,
+      "learning_rate": 5.971093750000001e-05,
+      "loss": 0.6149,
       "step": 6160
     },
     {
       "epoch": 89.42,
-      "learning_rate": 1.2423728813559323e-05,
-      "loss": 0.6135,
+      "learning_rate": 5.96328125e-05,
+      "loss": 0.6385,
       "step": 6170
     },
     {
       "epoch": 89.57,
-      "learning_rate": 1.2254237288135594e-05,
-      "loss": 0.6354,
+      "learning_rate": 5.9554687500000003e-05,
+      "loss": 0.6621,
       "step": 6180
     },
     {
       "epoch": 89.71,
-      "learning_rate": 1.2084745762711865e-05,
-      "loss": 0.5597,
+      "learning_rate": 5.94765625e-05,
+      "loss": 0.5817,
       "step": 6190
     },
     {
       "epoch": 89.86,
-      "learning_rate": 1.1915254237288135e-05,
-      "loss": 0.6065,
+      "learning_rate": 5.93984375e-05,
+      "loss": 0.6333,
       "step": 6200
     },
     {
       "epoch": 90.0,
-      "learning_rate": 1.1745762711864407e-05,
-      "loss": 0.5579,
+      "learning_rate": 5.9320312500000006e-05,
+      "loss": 0.5794,
       "step": 6210
     },
     {
       "epoch": 90.14,
-      "learning_rate": 1.157627118644068e-05,
-      "loss": 0.5914,
+      "learning_rate": 5.92421875e-05,
+      "loss": 0.6153,
       "step": 6220
     },
     {
       "epoch": 90.29,
-      "learning_rate": 1.140677966101695e-05,
-      "loss": 0.6552,
+      "learning_rate": 5.9164062500000006e-05,
+      "loss": 0.6794,
       "step": 6230
     },
     {
       "epoch": 90.43,
-      "learning_rate": 1.1237288135593221e-05,
-      "loss": 0.5762,
+      "learning_rate": 5.90859375e-05,
+      "loss": 0.601,
       "step": 6240
     },
     {
       "epoch": 90.58,
-      "learning_rate": 1.1067796610169492e-05,
-      "loss": 0.5902,
+      "learning_rate": 5.9007812500000005e-05,
+      "loss": 0.6176,
       "step": 6250
     },
     {
       "epoch": 90.72,
-      "learning_rate": 1.0898305084745764e-05,
-      "loss": 0.5821,
+      "learning_rate": 5.892968750000001e-05,
+      "loss": 0.6055,
       "step": 6260
     },
     {
       "epoch": 90.87,
-      "learning_rate": 1.0728813559322035e-05,
-      "loss": 0.6163,
+      "learning_rate": 5.88515625e-05,
+      "loss": 0.6389,
       "step": 6270
     },
     {
       "epoch": 91.01,
-      "learning_rate": 1.0559322033898305e-05,
-      "loss": 0.6451,
+      "learning_rate": 5.877343750000001e-05,
+      "loss": 0.6727,
       "step": 6280
     },
     {
       "epoch": 91.16,
-      "learning_rate": 1.0389830508474577e-05,
-      "loss": 0.6399,
+      "learning_rate": 5.86953125e-05,
+      "loss": 0.6592,
       "step": 6290
     },
     {
       "epoch": 91.3,
-      "learning_rate": 1.0220338983050848e-05,
-      "loss": 0.6132,
+      "learning_rate": 5.86171875e-05,
+      "loss": 0.6367,
       "step": 6300
     },
     {
       "epoch": 91.45,
-      "learning_rate": 1.005084745762712e-05,
-      "loss": 0.612,
+      "learning_rate": 5.853906250000001e-05,
+      "loss": 0.6366,
       "step": 6310
     },
     {
       "epoch": 91.59,
-      "learning_rate": 9.88135593220339e-06,
-      "loss": 0.5939,
+      "learning_rate": 5.84609375e-05,
+      "loss": 0.6171,
       "step": 6320
     },
     {
       "epoch": 91.74,
-      "learning_rate": 9.71186440677966e-06,
-      "loss": 0.6379,
+      "learning_rate": 5.8382812500000004e-05,
+      "loss": 0.6642,
       "step": 6330
     },
     {
       "epoch": 91.88,
-      "learning_rate": 9.542372881355932e-06,
-      "loss": 0.6063,
+      "learning_rate": 5.83046875e-05,
+      "loss": 0.6345,
       "step": 6340
     },
     {
       "epoch": 92.03,
-      "learning_rate": 9.372881355932204e-06,
-      "loss": 0.6005,
+      "learning_rate": 5.82265625e-05,
+      "loss": 0.6265,
       "step": 6350
     },
     {
       "epoch": 92.17,
-      "learning_rate": 9.203389830508475e-06,
-      "loss": 0.6076,
+      "learning_rate": 5.8148437500000006e-05,
+      "loss": 0.6302,
       "step": 6360
     },
     {
       "epoch": 92.32,
-      "learning_rate": 9.033898305084747e-06,
-      "loss": 0.6082,
+      "learning_rate": 5.80703125e-05,
+      "loss": 0.6347,
       "step": 6370
     },
     {
       "epoch": 92.46,
-      "learning_rate": 8.864406779661018e-06,
-      "loss": 0.6571,
+      "learning_rate": 5.7992187500000006e-05,
+      "loss": 0.6858,
       "step": 6380
     },
     {
       "epoch": 92.61,
-      "learning_rate": 8.69491525423729e-06,
-      "loss": 0.5666,
+      "learning_rate": 5.79140625e-05,
+      "loss": 0.5876,
       "step": 6390
     },
     {
       "epoch": 92.75,
-      "learning_rate": 8.52542372881356e-06,
-      "loss": 0.5932,
+      "learning_rate": 5.7835937500000006e-05,
+      "loss": 0.6195,
       "step": 6400
     },
     {
       "epoch": 92.9,
-      "learning_rate": 8.35593220338983e-06,
-      "loss": 0.6317,
+      "learning_rate": 5.775781250000001e-05,
+      "loss": 0.6559,
       "step": 6410
     },
     {
       "epoch": 93.04,
-      "learning_rate": 8.186440677966102e-06,
-      "loss": 0.6665,
+      "learning_rate": 5.76796875e-05,
+      "loss": 0.6908,
       "step": 6420
     },
     {
       "epoch": 93.19,
-      "learning_rate": 8.016949152542374e-06,
-      "loss": 0.5687,
+      "learning_rate": 5.760156250000001e-05,
+      "loss": 0.5887,
       "step": 6430
     },
     {
       "epoch": 93.33,
-      "learning_rate": 7.847457627118643e-06,
-      "loss": 0.5904,
+      "learning_rate": 5.75234375e-05,
+      "loss": 0.6118,
       "step": 6440
     },
     {
       "epoch": 93.48,
-      "learning_rate": 7.677966101694915e-06,
-      "loss": 0.7073,
+      "learning_rate": 5.74453125e-05,
+      "loss": 0.7352,
       "step": 6450
     },
     {
       "epoch": 93.62,
-      "learning_rate": 7.508474576271186e-06,
-      "loss": 0.619,
+      "learning_rate": 5.736718750000001e-05,
+      "loss": 0.6462,
       "step": 6460
     },
     {
       "epoch": 93.77,
-      "learning_rate": 7.338983050847458e-06,
-      "loss": 0.6203,
+      "learning_rate": 5.72890625e-05,
+      "loss": 0.6448,
       "step": 6470
     },
     {
       "epoch": 93.91,
-      "learning_rate": 7.1694915254237284e-06,
-      "loss": 0.6576,
+      "learning_rate": 5.7210937500000004e-05,
+      "loss": 0.6806,
       "step": 6480
     },
     {
       "epoch": 94.06,
-      "learning_rate": 7.000000000000001e-06,
-      "loss": 0.6249,
+      "learning_rate": 5.71328125e-05,
+      "loss": 0.6465,
       "step": 6490
     },
     {
       "epoch": 94.2,
-      "learning_rate": 6.830508474576272e-06,
-      "loss": 0.588,
+      "learning_rate": 5.7062500000000005e-05,
+      "loss": 0.608,
       "step": 6500
     },
     {
       "epoch": 94.2,
-      "eval_loss": 0.2143363356590271,
-      "eval_runtime": 573.5555,
-      "eval_samples_per_second": 5.917,
-      "eval_steps_per_second": 0.741,
-      "eval_wer": 0.15602067565679725,
+      "eval_loss": 0.22552849352359772,
+      "eval_runtime": 596.3798,
+      "eval_samples_per_second": 5.691,
+      "eval_steps_per_second": 0.713,
+      "eval_wer": 0.1638849854958131,
       "step": 6500
     },
     {
       "epoch": 94.35,
-      "learning_rate": 6.661016949152543e-06,
-      "loss": 0.669,
+      "learning_rate": 5.6984375e-05,
+      "loss": 0.6931,
       "step": 6510
     },
     {
       "epoch": 94.49,
-      "learning_rate": 6.491525423728814e-06,
-      "loss": 0.5974,
+      "learning_rate": 5.6906250000000004e-05,
+      "loss": 0.6172,
       "step": 6520
     },
     {
       "epoch": 94.64,
-      "learning_rate": 6.322033898305085e-06,
-      "loss": 0.6671,
+      "learning_rate": 5.6828125e-05,
+      "loss": 0.686,
       "step": 6530
     },
     {
       "epoch": 94.78,
-      "learning_rate": 6.152542372881356e-06,
-      "loss": 0.6735,
+      "learning_rate": 5.6750000000000004e-05,
+      "loss": 0.6993,
       "step": 6540
     },
     {
       "epoch": 94.93,
-      "learning_rate": 5.983050847457628e-06,
-      "loss": 0.6356,
+      "learning_rate": 5.667187500000001e-05,
+      "loss": 0.6582,
       "step": 6550
     },
     {
       "epoch": 95.07,
-      "learning_rate": 5.813559322033898e-06,
-      "loss": 0.6135,
+      "learning_rate": 5.6593750000000003e-05,
+      "loss": 0.6366,
       "step": 6560
     },
     {
       "epoch": 95.22,
-      "learning_rate": 5.64406779661017e-06,
-      "loss": 0.6403,
+      "learning_rate": 5.6515625000000007e-05,
+      "loss": 0.6609,
       "step": 6570
     },
     {
       "epoch": 95.36,
-      "learning_rate": 5.4745762711864405e-06,
-      "loss": 0.5574,
+      "learning_rate": 5.6437499999999996e-05,
+      "loss": 0.5802,
       "step": 6580
     },
     {
       "epoch": 95.51,
-      "learning_rate": 5.305084745762713e-06,
-      "loss": 0.552,
+      "learning_rate": 5.6359375000000006e-05,
+      "loss": 0.5748,
       "step": 6590
     },
     {
       "epoch": 95.65,
-      "learning_rate": 5.135593220338983e-06,
-      "loss": 0.5979,
+      "learning_rate": 5.628125000000001e-05,
+      "loss": 0.6195,
       "step": 6600
     },
     {
       "epoch": 95.8,
-      "learning_rate": 4.966101694915255e-06,
-      "loss": 0.6342,
+      "learning_rate": 5.6203125e-05,
+      "loss": 0.6524,
       "step": 6610
     },
     {
       "epoch": 95.94,
-      "learning_rate": 4.7966101694915255e-06,
-      "loss": 0.6397,
+      "learning_rate": 5.6125e-05,
+      "loss": 0.6621,
       "step": 6620
     },
     {
       "epoch": 96.09,
-      "learning_rate": 4.627118644067797e-06,
-      "loss": 0.5847,
+      "learning_rate": 5.6046875e-05,
+      "loss": 0.6074,
       "step": 6630
     },
     {
       "epoch": 96.23,
-      "learning_rate": 4.4576271186440676e-06,
-      "loss": 0.5736,
+      "learning_rate": 5.596875e-05,
+      "loss": 0.598,
       "step": 6640
     },
     {
       "epoch": 96.38,
-      "learning_rate": 4.288135593220339e-06,
-      "loss": 0.6148,
+      "learning_rate": 5.5890625000000005e-05,
+      "loss": 0.6296,
       "step": 6650
     },
     {
       "epoch": 96.52,
-      "learning_rate": 4.1186440677966105e-06,
-      "loss": 0.6084,
+      "learning_rate": 5.58125e-05,
+      "loss": 0.6353,
       "step": 6660
     },
     {
       "epoch": 96.67,
-      "learning_rate": 3.949152542372882e-06,
-      "loss": 0.5835,
+      "learning_rate": 5.5734375000000005e-05,
+      "loss": 0.6013,
       "step": 6670
     },
     {
       "epoch": 96.81,
-      "learning_rate": 3.779661016949153e-06,
-      "loss": 0.5791,
+      "learning_rate": 5.565625e-05,
+      "loss": 0.606,
       "step": 6680
     },
     {
       "epoch": 96.96,
-      "learning_rate": 3.610169491525424e-06,
-      "loss": 0.7132,
+      "learning_rate": 5.5578125000000004e-05,
+      "loss": 0.742,
       "step": 6690
     },
     {
       "epoch": 97.1,
-      "learning_rate": 3.440677966101695e-06,
-      "loss": 0.5572,
+      "learning_rate": 5.550000000000001e-05,
+      "loss": 0.5763,
       "step": 6700
     },
     {
       "epoch": 97.25,
-      "learning_rate": 3.271186440677966e-06,
-      "loss": 0.6607,
+      "learning_rate": 5.5421875000000004e-05,
+      "loss": 0.6839,
       "step": 6710
     },
     {
       "epoch": 97.39,
-      "learning_rate": 3.1016949152542375e-06,
-      "loss": 0.6343,
+      "learning_rate": 5.534375000000001e-05,
+      "loss": 0.6565,
       "step": 6720
     },
     {
       "epoch": 97.54,
-      "learning_rate": 2.9322033898305086e-06,
-      "loss": 0.6126,
+      "learning_rate": 5.5265624999999997e-05,
+      "loss": 0.6341,
       "step": 6730
     },
     {
       "epoch": 97.68,
-      "learning_rate": 2.76271186440678e-06,
-      "loss": 0.6536,
+      "learning_rate": 5.51875e-05,
+      "loss": 0.6721,
       "step": 6740
     },
     {
       "epoch": 97.83,
-      "learning_rate": 2.593220338983051e-06,
-      "loss": 0.5843,
+      "learning_rate": 5.510937500000001e-05,
+      "loss": 0.6134,
       "step": 6750
     },
     {
       "epoch": 97.97,
-      "learning_rate": 2.423728813559322e-06,
-      "loss": 0.5874,
+      "learning_rate": 5.503125e-05,
+      "loss": 0.6161,
       "step": 6760
     },
     {
       "epoch": 98.12,
-      "learning_rate": 2.254237288135593e-06,
-      "loss": 0.6067,
+      "learning_rate": 5.4953125e-05,
+      "loss": 0.6273,
       "step": 6770
     },
     {
       "epoch": 98.26,
-      "learning_rate": 2.0847457627118646e-06,
-      "loss": 0.5626,
+      "learning_rate": 5.4875e-05,
+      "loss": 0.5778,
       "step": 6780
     },
     {
       "epoch": 98.41,
-      "learning_rate": 1.9152542372881356e-06,
-      "loss": 0.6659,
+      "learning_rate": 5.4796875e-05,
+      "loss": 0.6879,
       "step": 6790
     },
     {
       "epoch": 98.55,
-      "learning_rate": 1.7457627118644067e-06,
-      "loss": 0.6462,
+      "learning_rate": 5.4718750000000005e-05,
+      "loss": 0.6652,
       "step": 6800
     },
     {
       "epoch": 98.7,
-      "learning_rate": 1.5762711864406781e-06,
-      "loss": 0.67,
+      "learning_rate": 5.4640625e-05,
+      "loss": 0.6894,
       "step": 6810
     },
     {
       "epoch": 98.84,
-      "learning_rate": 1.4067796610169492e-06,
-      "loss": 0.5734,
+      "learning_rate": 5.4562500000000005e-05,
+      "loss": 0.5951,
       "step": 6820
     },
     {
       "epoch": 98.99,
-      "learning_rate": 1.2372881355932204e-06,
-      "loss": 0.6133,
+      "learning_rate": 5.4484375e-05,
+      "loss": 0.6397,
       "step": 6830
     },
     {
       "epoch": 99.13,
-      "learning_rate": 1.0677966101694917e-06,
-      "loss": 0.596,
+      "learning_rate": 5.4406250000000004e-05,
+      "loss": 0.6221,
       "step": 6840
     },
     {
       "epoch": 99.28,
-      "learning_rate": 9.152542372881356e-07,
-      "loss": 0.6763,
+      "learning_rate": 5.432812500000001e-05,
+      "loss": 0.7,
       "step": 6850
     },
     {
       "epoch": 99.42,
-      "learning_rate": 7.457627118644068e-07,
-      "loss": 0.6368,
+      "learning_rate": 5.4250000000000004e-05,
+      "loss": 0.6638,
       "step": 6860
     },
     {
       "epoch": 99.57,
-      "learning_rate": 5.76271186440678e-07,
-      "loss": 0.6463,
+      "learning_rate": 5.417187500000001e-05,
+      "loss": 0.6702,
       "step": 6870
     },
     {
       "epoch": 99.71,
-      "learning_rate": 4.0677966101694916e-07,
-      "loss": 0.62,
+      "learning_rate": 5.409375e-05,
+      "loss": 0.637,
       "step": 6880
     },
     {
       "epoch": 99.86,
-      "learning_rate": 2.3728813559322033e-07,
-      "loss": 0.6346,
+      "learning_rate": 5.4015625e-05,
+      "loss": 0.6518,
       "step": 6890
     },
     {
       "epoch": 100.0,
-      "learning_rate": 6.779661016949153e-08,
-      "loss": 0.5672,
+      "learning_rate": 5.393750000000001e-05,
+      "loss": 0.5912,
       "step": 6900
     },
     {
-      "epoch": 100.0,
-      "step": 6900,
-      "total_flos": 1.0025325448199992e+20,
-      "train_loss": 1.2894020353538402,
-      "train_runtime": 52505.7639,
-      "train_samples_per_second": 4.179,
-      "train_steps_per_second": 0.131
+      "epoch": 100.14,
+      "learning_rate": 5.3859375e-05,
+      "loss": 0.6999,
+      "step": 6910
+    },
+    {
+      "epoch": 100.29,
+      "learning_rate": 5.378125e-05,
+      "loss": 0.5682,
+      "step": 6920
+    },
+    {
+      "epoch": 100.43,
+      "learning_rate": 5.3703125e-05,
+      "loss": 0.6023,
+      "step": 6930
+    },
+    {
+      "epoch": 100.58,
+      "learning_rate": 5.3625e-05,
+      "loss": 0.6437,
+      "step": 6940
+    },
+    {
+      "epoch": 100.72,
+      "learning_rate": 5.3546875000000006e-05,
+      "loss": 0.7168,
+      "step": 6950
+    },
+    {
+      "epoch": 100.87,
+      "learning_rate": 5.346875e-05,
+      "loss": 0.5913,
+      "step": 6960
+    },
+    {
+      "epoch": 101.01,
+      "learning_rate": 5.3390625000000005e-05,
+      "loss": 0.6047,
+      "step": 6970
+    },
+    {
+      "epoch": 101.16,
+      "learning_rate": 5.33125e-05,
+      "loss": 0.6355,
+      "step": 6980
+    },
+    {
+      "epoch": 101.3,
+      "learning_rate": 5.3234375000000005e-05,
+      "loss": 0.7046,
+      "step": 6990
+    },
+    {
+      "epoch": 101.45,
+      "learning_rate": 5.315625000000001e-05,
+      "loss": 0.6099,
+      "step": 7000
+    },
+    {
+      "epoch": 101.45,
+      "eval_loss": 0.22652995586395264,
+      "eval_runtime": 583.6707,
+      "eval_samples_per_second": 5.815,
+      "eval_steps_per_second": 0.728,
+      "eval_wer": 0.1621839278817313,
+      "step": 7000
+    },
+    {
+      "epoch": 101.59,
+      "learning_rate": 5.3078125e-05,
+      "loss": 0.5706,
+      "step": 7010
+    },
+    {
+      "epoch": 101.74,
+      "learning_rate": 5.300000000000001e-05,
+      "loss": 0.5656,
+      "step": 7020
+    },
+    {
+      "epoch": 101.88,
+      "learning_rate": 5.2921875e-05,
+      "loss": 0.6755,
+      "step": 7030
+    },
+    {
+      "epoch": 102.03,
+      "learning_rate": 5.284375e-05,
+      "loss": 0.6343,
+      "step": 7040
+    },
+    {
+      "epoch": 102.17,
+      "learning_rate": 5.276562500000001e-05,
+      "loss": 0.6129,
+      "step": 7050
+    },
+    {
+      "epoch": 102.32,
+      "learning_rate": 5.26875e-05,
+      "loss": 0.6321,
+      "step": 7060
+    },
+    {
+      "epoch": 102.46,
+      "learning_rate": 5.2609375e-05,
+      "loss": 0.6189,
+      "step": 7070
+    },
+    {
+      "epoch": 102.61,
+      "learning_rate": 5.253125e-05,
+      "loss": 0.6267,
+      "step": 7080
+    },
+    {
+      "epoch": 102.75,
+      "learning_rate": 5.2453125e-05,
+      "loss": 0.601,
+      "step": 7090
+    },
+    {
+      "epoch": 102.9,
+      "learning_rate": 5.2375000000000006e-05,
+      "loss": 0.6051,
+      "step": 7100
+    },
+    {
+      "epoch": 103.04,
+      "learning_rate": 5.2296875e-05,
+      "loss": 0.538,
+      "step": 7110
+    },
+    {
+      "epoch": 103.19,
+      "learning_rate": 5.2218750000000006e-05,
+      "loss": 0.6136,
+      "step": 7120
+    },
+    {
+      "epoch": 103.33,
+      "learning_rate": 5.2140624999999995e-05,
+      "loss": 0.625,
+      "step": 7130
+    },
+    {
+      "epoch": 103.48,
+      "learning_rate": 5.2062500000000005e-05,
+      "loss": 0.5999,
+      "step": 7140
+    },
+    {
+      "epoch": 103.62,
+      "learning_rate": 5.198437500000001e-05,
+      "loss": 0.694,
+      "step": 7150
+    },
+    {
+      "epoch": 103.77,
+      "learning_rate": 5.190625e-05,
+      "loss": 0.6446,
+      "step": 7160
+    },
+    {
+      "epoch": 103.91,
+      "learning_rate": 5.182812500000001e-05,
+      "loss": 0.5879,
+      "step": 7170
+    },
+    {
+      "epoch": 104.06,
+      "learning_rate": 5.175e-05,
+      "loss": 0.6115,
+      "step": 7180
+    },
+    {
+      "epoch": 104.2,
+      "learning_rate": 5.1671875e-05,
+      "loss": 0.5663,
+      "step": 7190
+    },
+    {
+      "epoch": 104.35,
+      "learning_rate": 5.159375000000001e-05,
+      "loss": 0.5825,
+      "step": 7200
+    },
+    {
+      "epoch": 104.49,
+      "learning_rate": 5.1515625e-05,
+      "loss": 0.5692,
+      "step": 7210
+    },
+    {
+      "epoch": 104.64,
+      "learning_rate": 5.1437500000000003e-05,
+      "loss": 0.614,
+      "step": 7220
+    },
+    {
+      "epoch": 104.78,
+      "learning_rate": 5.1359375e-05,
+      "loss": 0.6005,
+      "step": 7230
+    },
+    {
+      "epoch": 104.93,
+      "learning_rate": 5.128125e-05,
+      "loss": 0.6169,
+      "step": 7240
+    },
+    {
+      "epoch": 105.07,
+      "learning_rate": 5.1203125000000006e-05,
+      "loss": 0.6383,
+      "step": 7250
+    },
+    {
+      "epoch": 105.22,
+      "learning_rate": 5.1125e-05,
+      "loss": 0.5705,
+      "step": 7260
+    },
+    {
+      "epoch": 105.36,
+      "learning_rate": 5.1046875000000006e-05,
+      "loss": 0.6212,
+      "step": 7270
+    },
+    {
+      "epoch": 105.51,
+      "learning_rate": 5.0968749999999995e-05,
+      "loss": 0.5747,
+      "step": 7280
+    },
+    {
+      "epoch": 105.65,
+      "learning_rate": 5.0890625000000005e-05,
+      "loss": 0.6407,
+      "step": 7290
+    },
+    {
+      "epoch": 105.8,
+      "learning_rate": 5.081250000000001e-05,
+      "loss": 0.5993,
+      "step": 7300
+    },
+    {
+      "epoch": 105.94,
+      "learning_rate": 5.0734375e-05,
+      "loss": 0.6226,
+      "step": 7310
+    },
+    {
+      "epoch": 106.09,
+      "learning_rate": 5.065625000000001e-05,
+      "loss": 0.6316,
+      "step": 7320
+    },
+    {
+      "epoch": 106.23,
+      "learning_rate": 5.0578125e-05,
+      "loss": 0.6922,
+      "step": 7330
+    },
+    {
+      "epoch": 106.38,
+      "learning_rate": 5.05e-05,
+      "loss": 0.6252,
+      "step": 7340
+    },
+    {
+      "epoch": 106.52,
+      "learning_rate": 5.0421875000000004e-05,
+      "loss": 0.5841,
+      "step": 7350
+    },
+    {
+      "epoch": 106.67,
+      "learning_rate": 5.034375e-05,
+      "loss": 0.6174,
+      "step": 7360
+    },
+    {
+      "epoch": 106.81,
+      "learning_rate": 5.0265625000000004e-05,
+      "loss": 0.6482,
+      "step": 7370
+    },
+    {
+      "epoch": 106.96,
+      "learning_rate": 5.01875e-05,
+      "loss": 0.5965,
+      "step": 7380
+    },
+    {
+      "epoch": 107.1,
+      "learning_rate": 5.0109375e-05,
+      "loss": 0.6417,
+      "step": 7390
+    },
+    {
+      "epoch": 107.25,
+      "learning_rate": 5.0031250000000007e-05,
+      "loss": 0.6373,
+      "step": 7400
+    },
+    {
+      "epoch": 107.39,
+      "learning_rate": 4.9953125e-05,
+      "loss": 0.6205,
+      "step": 7410
+    },
+    {
+      "epoch": 107.54,
+      "learning_rate": 4.9875000000000006e-05,
+      "loss": 0.5891,
+      "step": 7420
+    },
+    {
+      "epoch": 107.68,
+      "learning_rate": 4.9796875e-05,
+      "loss": 0.6333,
+      "step": 7430
+    },
+    {
+      "epoch": 107.83,
+      "learning_rate": 4.9718750000000006e-05,
+      "loss": 0.6479,
+      "step": 7440
+    },
+    {
+      "epoch": 107.97,
+      "learning_rate": 4.9640625e-05,
+      "loss": 0.5854,
+      "step": 7450
+    },
+    {
+      "epoch": 108.12,
+      "learning_rate": 4.95625e-05,
+      "loss": 0.602,
+      "step": 7460
+    },
+    {
+      "epoch": 108.26,
+      "learning_rate": 4.9484375e-05,
+      "loss": 0.6362,
+      "step": 7470
+    },
+    {
+      "epoch": 108.41,
+      "learning_rate": 4.9406250000000005e-05,
+      "loss": 0.7472,
+      "step": 7480
+    },
+    {
+      "epoch": 108.55,
+      "learning_rate": 4.9328125e-05,
+      "loss": 0.6158,
+      "step": 7490
+    },
+    {
+      "epoch": 108.7,
+      "learning_rate": 4.9250000000000004e-05,
+      "loss": 0.6069,
+      "step": 7500
+    },
+    {
+      "epoch": 108.7,
+      "eval_loss": 0.22459882497787476,
+      "eval_runtime": 574.1645,
+      "eval_samples_per_second": 5.911,
+      "eval_steps_per_second": 0.74,
+      "eval_wer": 0.1592502198226627,
+      "step": 7500
+    },
+    {
+      "epoch": 108.84,
+      "learning_rate": 4.9171875e-05,
+      "loss": 0.7501,
+      "step": 7510
+    },
+    {
+      "epoch": 108.99,
+      "learning_rate": 4.9093750000000004e-05,
+      "loss": 0.585,
+      "step": 7520
+    },
+    {
+      "epoch": 109.13,
+      "learning_rate": 4.901562500000001e-05,
+      "loss": 0.6811,
+      "step": 7530
+    },
+    {
+      "epoch": 109.28,
+      "learning_rate": 4.8937500000000004e-05,
+      "loss": 0.7343,
+      "step": 7540
+    },
+    {
+      "epoch": 109.42,
+      "learning_rate": 4.8859375e-05,
+      "loss": 0.6275,
+      "step": 7550
+    },
+    {
+      "epoch": 109.57,
+      "learning_rate": 4.878125e-05,
+      "loss": 0.6326,
+      "step": 7560
+    },
+    {
+      "epoch": 109.71,
+      "learning_rate": 4.8703125000000006e-05,
+      "loss": 0.6374,
+      "step": 7570
+    },
+    {
+      "epoch": 109.86,
+      "learning_rate": 4.8625e-05,
+      "loss": 0.6079,
+      "step": 7580
+    },
+    {
+      "epoch": 110.0,
+      "learning_rate": 4.8546875000000006e-05,
+      "loss": 0.5572,
+      "step": 7590
+    },
+    {
+      "epoch": 110.14,
+      "learning_rate": 4.846875e-05,
+      "loss": 0.6443,
+      "step": 7600
+    },
+    {
+      "epoch": 110.29,
+      "learning_rate": 4.8390625e-05,
+      "loss": 0.5898,
+      "step": 7610
+    },
+    {
+      "epoch": 110.43,
+      "learning_rate": 4.83125e-05,
+      "loss": 0.6258,
+      "step": 7620
+    },
+    {
+      "epoch": 110.58,
+      "learning_rate": 4.8234375000000005e-05,
+      "loss": 0.6048,
+      "step": 7630
+    },
+    {
+      "epoch": 110.72,
+      "learning_rate": 4.815625e-05,
+      "loss": 0.6092,
+      "step": 7640
+    },
+    {
+      "epoch": 110.87,
+      "learning_rate": 4.8078125000000005e-05,
+      "loss": 0.5939,
+      "step": 7650
+    },
+    {
+      "epoch": 111.01,
+      "learning_rate": 4.8e-05,
+      "loss": 0.6584,
+      "step": 7660
+    },
+    {
+      "epoch": 111.16,
+      "learning_rate": 4.7921875000000004e-05,
+      "loss": 0.6788,
+      "step": 7670
+    },
+    {
+      "epoch": 111.3,
+      "learning_rate": 4.784375e-05,
+      "loss": 0.6859,
+      "step": 7680
+    },
+    {
+      "epoch": 111.45,
+      "learning_rate": 4.7765625000000004e-05,
+      "loss": 0.6178,
+      "step": 7690
+    },
+    {
+      "epoch": 111.59,
+      "learning_rate": 4.76875e-05,
+      "loss": 0.5804,
+      "step": 7700
+    },
+    {
+      "epoch": 111.74,
+      "learning_rate": 4.7609375000000004e-05,
+      "loss": 0.693,
+      "step": 7710
+    },
+    {
+      "epoch": 111.88,
+      "learning_rate": 4.753125000000001e-05,
+      "loss": 0.6894,
+      "step": 7720
+    },
+    {
+      "epoch": 112.03,
+      "learning_rate": 4.7453125e-05,
+      "loss": 0.5552,
+      "step": 7730
+    },
+    {
+      "epoch": 112.17,
+      "learning_rate": 4.7375e-05,
+      "loss": 0.5774,
+      "step": 7740
+    },
+    {
+      "epoch": 112.32,
+      "learning_rate": 4.7296875e-05,
+      "loss": 0.5482,
+      "step": 7750
+    },
+    {
+      "epoch": 112.46,
+      "learning_rate": 4.721875e-05,
+      "loss": 0.6571,
+      "step": 7760
+    },
+    {
+      "epoch": 112.61,
+      "learning_rate": 4.7140625e-05,
+      "loss": 0.6455,
+      "step": 7770
+    },
+    {
+      "epoch": 112.75,
+      "learning_rate": 4.7062500000000006e-05,
+      "loss": 0.6115,
+      "step": 7780
+    },
+    {
+      "epoch": 112.9,
+      "learning_rate": 4.6984375e-05,
+      "loss": 0.6006,
+      "step": 7790
+    },
+    {
+      "epoch": 113.04,
+      "learning_rate": 4.690625e-05,
+      "loss": 0.5906,
+      "step": 7800
+    },
+    {
+      "epoch": 113.19,
+      "learning_rate": 4.6828125e-05,
+      "loss": 0.5477,
+      "step": 7810
+    },
+    {
+      "epoch": 113.33,
+      "learning_rate": 4.6750000000000005e-05,
+      "loss": 0.6492,
+      "step": 7820
+    },
+    {
+      "epoch": 113.48,
+      "learning_rate": 4.6671875e-05,
+      "loss": 0.5919,
+      "step": 7830
+    },
+    {
+      "epoch": 113.62,
+      "learning_rate": 4.6593750000000004e-05,
+      "loss": 0.5931,
+      "step": 7840
+    },
+    {
+      "epoch": 113.77,
+      "learning_rate": 4.6515625e-05,
+      "loss": 0.6467,
+      "step": 7850
+    },
+    {
+      "epoch": 113.91,
+      "learning_rate": 4.64375e-05,
+      "loss": 0.5646,
+      "step": 7860
+    },
+    {
+      "epoch": 114.06,
+      "learning_rate": 4.635937500000001e-05,
+      "loss": 0.6137,
+      "step": 7870
+    },
+    {
+      "epoch": 114.2,
+      "learning_rate": 4.6281250000000003e-05,
+      "loss": 0.5523,
+      "step": 7880
+    },
+    {
+      "epoch": 114.35,
+      "learning_rate": 4.6203125e-05,
+      "loss": 0.6965,
+      "step": 7890
+    },
+    {
+      "epoch": 114.49,
+      "learning_rate": 4.6125e-05,
+      "loss": 0.542,
+      "step": 7900
+    },
+    {
+      "epoch": 114.64,
+      "learning_rate": 4.6046875e-05,
+      "loss": 0.5662,
+      "step": 7910
+    },
+    {
+      "epoch": 114.78,
+      "learning_rate": 4.596875e-05,
+      "loss": 0.5677,
+      "step": 7920
+    },
+    {
+      "epoch": 114.93,
+      "learning_rate": 4.5890625000000006e-05,
+      "loss": 0.5547,
+      "step": 7930
+    },
+    {
+      "epoch": 115.07,
+      "learning_rate": 4.58125e-05,
+      "loss": 0.6085,
+      "step": 7940
+    },
+    {
+      "epoch": 115.22,
+      "learning_rate": 4.5734375e-05,
+      "loss": 0.5735,
+      "step": 7950
+    },
+    {
+      "epoch": 115.36,
+      "learning_rate": 4.565625e-05,
+      "loss": 0.5977,
+      "step": 7960
+    },
+    {
+      "epoch": 115.51,
+      "learning_rate": 4.5578125000000005e-05,
+      "loss": 0.6654,
+      "step": 7970
+    },
+    {
+      "epoch": 115.65,
+      "learning_rate": 4.55e-05,
+      "loss": 0.6036,
+      "step": 7980
+    },
+    {
+      "epoch": 115.8,
+      "learning_rate": 4.5421875000000005e-05,
+      "loss": 0.6112,
+      "step": 7990
+    },
+    {
+      "epoch": 115.94,
+      "learning_rate": 4.534375e-05,
+      "loss": 0.5929,
+      "step": 8000
+    },
+    {
+      "epoch": 115.94,
+      "eval_loss": 0.2322680950164795,
+      "eval_runtime": 572.4419,
+      "eval_samples_per_second": 5.929,
+      "eval_steps_per_second": 0.742,
+      "eval_wer": 0.16172373838226956,
+      "step": 8000
+    },
+    {
+      "epoch": 116.09,
+      "learning_rate": 4.5265625e-05,
+      "loss": 0.5724,
+      "step": 8010
+    },
+    {
+      "epoch": 116.23,
+      "learning_rate": 4.518750000000001e-05,
+      "loss": 0.589,
+      "step": 8020
+    },
+    {
+      "epoch": 116.38,
+      "learning_rate": 4.5109375000000004e-05,
+      "loss": 0.607,
+      "step": 8030
+    },
+    {
+      "epoch": 116.52,
+      "learning_rate": 4.503125e-05,
+      "loss": 0.6171,
+      "step": 8040
+    },
+    {
+      "epoch": 116.67,
+      "learning_rate": 4.4953125000000003e-05,
+      "loss": 0.6011,
+      "step": 8050
+    },
+    {
+      "epoch": 116.81,
+      "learning_rate": 4.4875e-05,
+      "loss": 0.6247,
+      "step": 8060
+    },
+    {
+      "epoch": 116.96,
+      "learning_rate": 4.4796875e-05,
+      "loss": 0.566,
+      "step": 8070
+    },
+    {
+      "epoch": 117.1,
+      "learning_rate": 4.4718750000000006e-05,
+      "loss": 0.6967,
+      "step": 8080
+    },
+    {
+      "epoch": 117.25,
+      "learning_rate": 4.4640625e-05,
+      "loss": 0.6296,
+      "step": 8090
+    },
+    {
+      "epoch": 117.39,
+      "learning_rate": 4.45625e-05,
+      "loss": 0.6186,
+      "step": 8100
+    },
+    {
+      "epoch": 117.54,
+      "learning_rate": 4.4484375e-05,
+      "loss": 0.5811,
+      "step": 8110
+    },
+    {
+      "epoch": 117.68,
+      "learning_rate": 4.4406250000000005e-05,
+      "loss": 0.5486,
+      "step": 8120
+    },
+    {
+      "epoch": 117.83,
+      "learning_rate": 4.4328125e-05,
+      "loss": 0.5805,
+      "step": 8130
+    },
+    {
+      "epoch": 117.97,
+      "learning_rate": 4.4250000000000005e-05,
+      "loss": 0.572,
+      "step": 8140
+    },
+    {
+      "epoch": 118.12,
+      "learning_rate": 4.4171875e-05,
+      "loss": 0.6108,
+      "step": 8150
+    },
+    {
+      "epoch": 118.26,
+      "learning_rate": 4.409375e-05,
+      "loss": 0.666,
+      "step": 8160
+    },
+    {
+      "epoch": 118.41,
+      "learning_rate": 4.401562500000001e-05,
+      "loss": 0.5758,
+      "step": 8170
+    },
+    {
+      "epoch": 118.55,
+      "learning_rate": 4.3937500000000004e-05,
+      "loss": 0.6212,
+      "step": 8180
+    },
+    {
+      "epoch": 118.7,
+      "learning_rate": 4.3859375e-05,
+      "loss": 0.6679,
+      "step": 8190
+    },
+    {
+      "epoch": 118.84,
+      "learning_rate": 4.3781250000000004e-05,
+      "loss": 0.6261,
+      "step": 8200
+    },
+    {
+      "epoch": 118.99,
+      "learning_rate": 4.3703125e-05,
+      "loss": 0.6025,
+      "step": 8210
+    },
+    {
+      "epoch": 119.13,
+      "learning_rate": 4.3625e-05,
+      "loss": 0.6722,
+      "step": 8220
+    },
+    {
+      "epoch": 119.28,
+      "learning_rate": 4.3546875000000006e-05,
+      "loss": 0.5637,
+      "step": 8230
+    },
+    {
+      "epoch": 119.42,
+      "learning_rate": 4.346875e-05,
+      "loss": 0.6138,
+      "step": 8240
+    },
+    {
+      "epoch": 119.57,
+      "learning_rate": 4.3390625e-05,
+      "loss": 0.6397,
+      "step": 8250
+    },
+    {
+      "epoch": 119.71,
+      "learning_rate": 4.33125e-05,
+      "loss": 0.5852,
+      "step": 8260
+    },
+    {
+      "epoch": 119.86,
+      "learning_rate": 4.3234375000000006e-05,
+      "loss": 0.616,
+      "step": 8270
+    },
+    {
+      "epoch": 120.0,
+      "learning_rate": 4.315625e-05,
+      "loss": 0.5824,
+      "step": 8280
+    },
+    {
+      "epoch": 120.14,
+      "learning_rate": 4.3078125000000005e-05,
+      "loss": 0.5909,
+      "step": 8290
+    },
+    {
+      "epoch": 120.29,
+      "learning_rate": 4.3e-05,
+      "loss": 0.6523,
+      "step": 8300
+    },
+    {
+      "epoch": 120.43,
+      "learning_rate": 4.2921875e-05,
+      "loss": 0.6134,
+      "step": 8310
+    },
+    {
+      "epoch": 120.58,
+      "learning_rate": 4.284375000000001e-05,
+      "loss": 0.5903,
+      "step": 8320
+    },
+    {
+      "epoch": 120.72,
+      "learning_rate": 4.2765625000000004e-05,
+      "loss": 0.5471,
+      "step": 8330
+    },
+    {
+      "epoch": 120.87,
+      "learning_rate": 4.26875e-05,
+      "loss": 0.62,
+      "step": 8340
+    },
+    {
+      "epoch": 121.01,
+      "learning_rate": 4.2609375000000004e-05,
+      "loss": 0.6299,
+      "step": 8350
+    },
+    {
+      "epoch": 121.16,
+      "learning_rate": 4.253125e-05,
+      "loss": 0.5991,
+      "step": 8360
+    },
+    {
+      "epoch": 121.3,
+      "learning_rate": 4.2453125000000004e-05,
+      "loss": 0.597,
+      "step": 8370
+    },
+    {
+      "epoch": 121.45,
+      "learning_rate": 4.237500000000001e-05,
+      "loss": 0.5925,
+      "step": 8380
+    },
+    {
+      "epoch": 121.59,
+      "learning_rate": 4.2296875e-05,
+      "loss": 0.6336,
+      "step": 8390
+    },
+    {
+      "epoch": 121.74,
+      "learning_rate": 4.221875e-05,
+      "loss": 0.5826,
+      "step": 8400
+    },
+    {
+      "epoch": 121.88,
+      "learning_rate": 4.2140625e-05,
+      "loss": 0.5964,
+      "step": 8410
+    },
+    {
+      "epoch": 122.03,
+      "learning_rate": 4.2062500000000006e-05,
+      "loss": 0.5978,
+      "step": 8420
+    },
+    {
+      "epoch": 122.17,
+      "learning_rate": 4.1984375e-05,
+      "loss": 0.6456,
+      "step": 8430
+    },
+    {
+      "epoch": 122.32,
+      "learning_rate": 4.1906250000000006e-05,
+      "loss": 0.6088,
+      "step": 8440
+    },
+    {
+      "epoch": 122.46,
+      "learning_rate": 4.1828125e-05,
+      "loss": 0.6204,
+      "step": 8450
+    },
+    {
+      "epoch": 122.61,
+      "learning_rate": 4.175e-05,
+      "loss": 0.5545,
+      "step": 8460
+    },
+    {
+      "epoch": 122.75,
+      "learning_rate": 4.1671875e-05,
+      "loss": 0.5823,
+      "step": 8470
+    },
+    {
+      "epoch": 122.9,
+      "learning_rate": 4.1593750000000005e-05,
+      "loss": 0.5386,
+      "step": 8480
+    },
+    {
+      "epoch": 123.04,
+      "learning_rate": 4.1515625e-05,
+      "loss": 0.5619,
+      "step": 8490
+    },
+    {
+      "epoch": 123.19,
+      "learning_rate": 4.1437500000000004e-05,
+      "loss": 0.6218,
+      "step": 8500
+    },
+    {
+      "epoch": 123.19,
+      "eval_loss": 0.22871814668178558,
+      "eval_runtime": 573.0196,
+      "eval_samples_per_second": 5.923,
+      "eval_steps_per_second": 0.742,
+      "eval_wer": 0.1565712595222247,
+      "step": 8500
+    },
+    {
+      "epoch": 123.33,
+      "learning_rate": 4.1359375e-05,
+      "loss": 0.6351,
+      "step": 8510
+    },
+    {
+      "epoch": 123.48,
+      "learning_rate": 4.1281250000000004e-05,
+      "loss": 0.5678,
+      "step": 8520
+    },
+    {
+      "epoch": 123.62,
+      "learning_rate": 4.1203125e-05,
+      "loss": 0.7051,
+      "step": 8530
+    },
+    {
+      "epoch": 123.77,
+      "learning_rate": 4.1125000000000004e-05,
+      "loss": 0.5785,
+      "step": 8540
+    },
+    {
+      "epoch": 123.91,
+      "learning_rate": 4.1046875e-05,
+      "loss": 0.6527,
+      "step": 8550
+    },
+    {
+      "epoch": 124.06,
+      "learning_rate": 4.096875e-05,
+      "loss": 0.5509,
+      "step": 8560
+    },
+    {
+      "epoch": 124.2,
+      "learning_rate": 4.0890625000000006e-05,
+      "loss": 0.6065,
+      "step": 8570
+    },
+    {
+      "epoch": 124.35,
+      "learning_rate": 4.08125e-05,
+      "loss": 0.6249,
+      "step": 8580
+    },
+    {
+      "epoch": 124.49,
+      "learning_rate": 4.0734375e-05,
+      "loss": 0.5477,
+      "step": 8590
+    },
+    {
+      "epoch": 124.64,
+      "learning_rate": 4.065625e-05,
+      "loss": 0.6066,
+      "step": 8600
+    },
+    {
+      "epoch": 124.78,
+      "learning_rate": 4.0585937500000007e-05,
+      "loss": 0.5868,
+      "step": 8610
+    },
+    {
+      "epoch": 124.93,
+      "learning_rate": 4.05078125e-05,
+      "loss": 0.5902,
+      "step": 8620
+    },
+    {
+      "epoch": 125.07,
+      "learning_rate": 4.04296875e-05,
+      "loss": 0.6138,
+      "step": 8630
+    },
+    {
+      "epoch": 125.22,
+      "learning_rate": 4.03515625e-05,
+      "loss": 0.6382,
+      "step": 8640
+    },
+    {
+      "epoch": 125.36,
+      "learning_rate": 4.02734375e-05,
+      "loss": 0.6279,
+      "step": 8650
+    },
+    {
+      "epoch": 125.51,
+      "learning_rate": 4.01953125e-05,
+      "loss": 0.6699,
+      "step": 8660
+    },
+    {
+      "epoch": 125.65,
+      "learning_rate": 4.0117187500000005e-05,
+      "loss": 0.6396,
+      "step": 8670
+    },
+    {
+      "epoch": 125.8,
+      "learning_rate": 4.00390625e-05,
+      "loss": 0.564,
+      "step": 8680
+    },
+    {
+      "epoch": 125.94,
+      "learning_rate": 3.99609375e-05,
+      "loss": 0.556,
+      "step": 8690
+    },
+    {
+      "epoch": 126.09,
+      "learning_rate": 3.98828125e-05,
+      "loss": 0.6425,
+      "step": 8700
+    },
+    {
+      "epoch": 126.23,
+      "learning_rate": 3.9804687500000004e-05,
+      "loss": 0.5753,
+      "step": 8710
+    },
+    {
+      "epoch": 126.38,
+      "learning_rate": 3.97265625e-05,
+      "loss": 0.5892,
+      "step": 8720
+    },
+    {
+      "epoch": 126.52,
+      "learning_rate": 3.9648437500000004e-05,
+      "loss": 0.5827,
+      "step": 8730
+    },
+    {
+      "epoch": 126.67,
+      "learning_rate": 3.95703125e-05,
+      "loss": 0.6517,
+      "step": 8740
+    },
+    {
+      "epoch": 126.81,
+      "learning_rate": 3.94921875e-05,
+      "loss": 0.5346,
+      "step": 8750
+    },
+    {
+      "epoch": 126.96,
+      "learning_rate": 3.941406250000001e-05,
+      "loss": 0.5868,
+      "step": 8760
+    },
+    {
+      "epoch": 127.1,
+      "learning_rate": 3.93359375e-05,
+      "loss": 0.5673,
+      "step": 8770
+    },
+    {
+      "epoch": 127.25,
+      "learning_rate": 3.92578125e-05,
+      "loss": 0.59,
+      "step": 8780
+    },
+    {
+      "epoch": 127.39,
+      "learning_rate": 3.91796875e-05,
+      "loss": 0.5705,
+      "step": 8790
+    },
+    {
+      "epoch": 127.54,
+      "learning_rate": 3.91015625e-05,
+      "loss": 0.5958,
+      "step": 8800
+    },
+    {
+      "epoch": 127.68,
+      "learning_rate": 3.90234375e-05,
+      "loss": 0.5522,
+      "step": 8810
+    },
+    {
+      "epoch": 127.83,
+      "learning_rate": 3.8945312500000006e-05,
+      "loss": 0.581,
+      "step": 8820
+    },
+    {
+      "epoch": 127.97,
+      "learning_rate": 3.88671875e-05,
+      "loss": 0.5469,
+      "step": 8830
+    },
+    {
+      "epoch": 128.12,
+      "learning_rate": 3.87890625e-05,
+      "loss": 0.6312,
+      "step": 8840
+    },
+    {
+      "epoch": 128.26,
+      "learning_rate": 3.87109375e-05,
+      "loss": 0.6486,
+      "step": 8850
+    },
+    {
+      "epoch": 128.41,
+      "learning_rate": 3.8632812500000005e-05,
+      "loss": 0.6362,
+      "step": 8860
+    },
+    {
+      "epoch": 128.55,
+      "learning_rate": 3.85546875e-05,
+      "loss": 0.5602,
+      "step": 8870
+    },
+    {
+      "epoch": 128.7,
+      "learning_rate": 3.8476562500000004e-05,
+      "loss": 0.5554,
+      "step": 8880
+    },
+    {
+      "epoch": 128.84,
+      "learning_rate": 3.83984375e-05,
+      "loss": 0.5491,
+      "step": 8890
+    },
+    {
+      "epoch": 128.99,
+      "learning_rate": 3.83203125e-05,
+      "loss": 0.5681,
+      "step": 8900
+    },
+    {
+      "epoch": 129.13,
+      "learning_rate": 3.824218750000001e-05,
+      "loss": 0.6239,
+      "step": 8910
+    },
+    {
+      "epoch": 129.28,
+      "learning_rate": 3.8164062500000004e-05,
+      "loss": 0.5759,
+      "step": 8920
+    },
+    {
+      "epoch": 129.42,
+      "learning_rate": 3.80859375e-05,
+      "loss": 0.5868,
+      "step": 8930
+    },
+    {
+      "epoch": 129.57,
+      "learning_rate": 3.8015625e-05,
+      "loss": 0.5796,
+      "step": 8940
+    },
+    {
+      "epoch": 129.71,
+      "learning_rate": 3.79375e-05,
+      "loss": 0.5501,
+      "step": 8950
+    },
+    {
+      "epoch": 129.86,
+      "learning_rate": 3.7859375000000004e-05,
+      "loss": 0.5976,
+      "step": 8960
+    },
+    {
+      "epoch": 130.0,
+      "learning_rate": 3.778125e-05,
+      "loss": 0.5488,
+      "step": 8970
+    },
+    {
+      "epoch": 130.14,
+      "learning_rate": 3.7703125e-05,
+      "loss": 0.5688,
+      "step": 8980
+    },
+    {
+      "epoch": 130.29,
+      "learning_rate": 3.7625e-05,
+      "loss": 0.5448,
+      "step": 8990
+    },
+    {
+      "epoch": 130.43,
+      "learning_rate": 3.7546875e-05,
+      "loss": 0.5751,
+      "step": 9000
+    },
+    {
+      "epoch": 130.43,
+      "eval_loss": 0.22747375071048737,
+      "eval_runtime": 574.7485,
+      "eval_samples_per_second": 5.905,
+      "eval_steps_per_second": 0.739,
+      "eval_wer": 0.1562672057457946,
+      "step": 9000
+    },
+    {
+      "epoch": 130.58,
+      "learning_rate": 3.746875e-05,
+      "loss": 0.6,
+      "step": 9010
+    },
+    {
+      "epoch": 130.72,
+      "learning_rate": 3.7390625e-05,
+      "loss": 0.5781,
+      "step": 9020
+    },
+    {
+      "epoch": 130.87,
+      "learning_rate": 3.73125e-05,
+      "loss": 0.5545,
+      "step": 9030
+    },
+    {
+      "epoch": 131.01,
+      "learning_rate": 3.7234375e-05,
+      "loss": 0.6116,
+      "step": 9040
+    },
+    {
+      "epoch": 131.16,
+      "learning_rate": 3.7156250000000005e-05,
+      "loss": 0.5634,
+      "step": 9050
+    },
+    {
+      "epoch": 131.3,
+      "learning_rate": 3.7078125e-05,
+      "loss": 0.6436,
+      "step": 9060
+    },
+    {
+      "epoch": 131.45,
+      "learning_rate": 3.7e-05,
+      "loss": 0.5882,
+      "step": 9070
+    },
+    {
+      "epoch": 131.59,
+      "learning_rate": 3.6921875e-05,
+      "loss": 0.5391,
+      "step": 9080
+    },
+    {
+      "epoch": 131.74,
+      "learning_rate": 3.684375e-05,
+      "loss": 0.5211,
+      "step": 9090
+    },
+    {
+      "epoch": 131.88,
+      "learning_rate": 3.6765625e-05,
+      "loss": 0.5971,
+      "step": 9100
+    },
+    {
+      "epoch": 132.03,
+      "learning_rate": 3.6687500000000004e-05,
+      "loss": 0.5314,
+      "step": 9110
+    },
+    {
+      "epoch": 132.17,
+      "learning_rate": 3.6609375e-05,
+      "loss": 0.573,
+      "step": 9120
+    },
+    {
+      "epoch": 132.32,
+      "learning_rate": 3.653125e-05,
+      "loss": 0.5678,
+      "step": 9130
+    },
+    {
+      "epoch": 132.46,
+      "learning_rate": 3.6453125e-05,
+      "loss": 0.5369,
+      "step": 9140
+    },
+    {
+      "epoch": 132.61,
+      "learning_rate": 3.6375e-05,
+      "loss": 0.5976,
+      "step": 9150
+    },
+    {
+      "epoch": 132.75,
+      "learning_rate": 3.6296875e-05,
+      "loss": 0.609,
+      "step": 9160
+    },
+    {
+      "epoch": 132.9,
+      "learning_rate": 3.621875e-05,
+      "loss": 0.5859,
+      "step": 9170
+    },
+    {
+      "epoch": 133.04,
+      "learning_rate": 3.6140625e-05,
+      "loss": 0.6005,
+      "step": 9180
+    },
+    {
+      "epoch": 133.19,
+      "learning_rate": 3.60625e-05,
+      "loss": 0.5796,
+      "step": 9190
+    },
+    {
+      "epoch": 133.33,
+      "learning_rate": 3.5984375000000006e-05,
+      "loss": 0.5125,
+      "step": 9200
+    },
+    {
+      "epoch": 133.48,
+      "learning_rate": 3.590625e-05,
+      "loss": 0.5465,
+      "step": 9210
+    },
+    {
+      "epoch": 133.62,
+      "learning_rate": 3.5828125e-05,
+      "loss": 0.5985,
+      "step": 9220
+    },
+    {
+      "epoch": 133.77,
+      "learning_rate": 3.575e-05,
+      "loss": 0.5687,
+      "step": 9230
+    },
+    {
+      "epoch": 133.91,
+      "learning_rate": 3.5671875e-05,
+      "loss": 0.6586,
+      "step": 9240
+    },
+    {
+      "epoch": 134.06,
+      "learning_rate": 3.559375e-05,
+      "loss": 0.5656,
+      "step": 9250
+    },
+    {
+      "epoch": 134.2,
+      "learning_rate": 3.5515625000000004e-05,
+      "loss": 0.6136,
+      "step": 9260
+    },
+    {
+      "epoch": 134.35,
+      "learning_rate": 3.54375e-05,
+      "loss": 0.6063,
+      "step": 9270
+    },
+    {
+      "epoch": 134.49,
+      "learning_rate": 3.5359375e-05,
+      "loss": 0.5242,
+      "step": 9280
+    },
+    {
+      "epoch": 134.64,
+      "learning_rate": 3.528125e-05,
+      "loss": 0.5605,
+      "step": 9290
+    },
+    {
+      "epoch": 134.78,
+      "learning_rate": 3.5203125000000004e-05,
+      "loss": 0.5774,
+      "step": 9300
+    },
+    {
+      "epoch": 134.93,
+      "learning_rate": 3.5125e-05,
+      "loss": 0.5937,
+      "step": 9310
+    },
+    {
+      "epoch": 135.07,
+      "learning_rate": 3.5046875e-05,
+      "loss": 0.5875,
+      "step": 9320
+    },
+    {
+      "epoch": 135.22,
+      "learning_rate": 3.496875e-05,
+      "loss": 0.5483,
+      "step": 9330
+    },
+    {
+      "epoch": 135.36,
+      "learning_rate": 3.4890624999999996e-05,
+      "loss": 0.5705,
+      "step": 9340
+    },
+    {
+      "epoch": 135.51,
+      "learning_rate": 3.4812500000000006e-05,
+      "loss": 0.6026,
+      "step": 9350
+    },
+    {
+      "epoch": 135.65,
+      "learning_rate": 3.4734375e-05,
+      "loss": 0.5888,
+      "step": 9360
+    },
+    {
+      "epoch": 135.8,
+      "learning_rate": 3.465625e-05,
+      "loss": 0.5913,
+      "step": 9370
+    },
+    {
+      "epoch": 135.94,
+      "learning_rate": 3.4578125e-05,
+      "loss": 0.5792,
+      "step": 9380
+    },
+    {
+      "epoch": 136.09,
+      "learning_rate": 3.45e-05,
+      "loss": 0.615,
+      "step": 9390
+    },
+    {
+      "epoch": 136.23,
+      "learning_rate": 3.4421875e-05,
+      "loss": 0.5827,
+      "step": 9400
+    },
+    {
+      "epoch": 136.38,
+      "learning_rate": 3.4343750000000005e-05,
+      "loss": 0.5956,
+      "step": 9410
+    },
+    {
+      "epoch": 136.52,
+      "learning_rate": 3.4265625e-05,
+      "loss": 0.4951,
+      "step": 9420
+    },
+    {
+      "epoch": 136.67,
+      "learning_rate": 3.41875e-05,
+      "loss": 0.5786,
+      "step": 9430
+    },
+    {
+      "epoch": 136.81,
+      "learning_rate": 3.4109375e-05,
+      "loss": 0.5307,
+      "step": 9440
+    },
+    {
+      "epoch": 136.96,
+      "learning_rate": 3.4031250000000004e-05,
+      "loss": 0.6652,
+      "step": 9450
+    },
+    {
+      "epoch": 137.1,
+      "learning_rate": 3.3953125e-05,
+      "loss": 0.5879,
+      "step": 9460
+    },
+    {
+      "epoch": 137.25,
+      "learning_rate": 3.3875000000000003e-05,
+      "loss": 0.6949,
+      "step": 9470
+    },
+    {
+      "epoch": 137.39,
+      "learning_rate": 3.3796875e-05,
+      "loss": 0.5537,
+      "step": 9480
+    },
+    {
+      "epoch": 137.54,
+      "learning_rate": 3.3718749999999996e-05,
+      "loss": 0.5923,
+      "step": 9490
+    },
+    {
+      "epoch": 137.68,
+      "learning_rate": 3.3640625000000006e-05,
+      "loss": 0.5181,
+      "step": 9500
+    },
+    {
+      "epoch": 137.68,
+      "eval_loss": 0.23162005841732025,
+      "eval_runtime": 574.0095,
+      "eval_samples_per_second": 5.913,
+      "eval_steps_per_second": 0.74,
+      "eval_wer": 0.15791895734207692,
+      "step": 9500
+    },
+    {
+      "epoch": 137.83,
+      "learning_rate": 3.35625e-05,
+      "loss": 0.5543,
+      "step": 9510
+    },
+    {
+      "epoch": 137.97,
+      "learning_rate": 3.3484375e-05,
+      "loss": 0.5875,
+      "step": 9520
+    },
+    {
+      "epoch": 138.12,
+      "learning_rate": 3.340625e-05,
+      "loss": 0.5757,
+      "step": 9530
+    },
+    {
+      "epoch": 138.26,
+      "learning_rate": 3.3328125e-05,
+      "loss": 0.5366,
+      "step": 9540
+    },
+    {
+      "epoch": 138.41,
+      "learning_rate": 3.325e-05,
+      "loss": 0.5328,
+      "step": 9550
+    },
+    {
+      "epoch": 138.55,
+      "learning_rate": 3.3171875000000005e-05,
+      "loss": 0.5193,
+      "step": 9560
+    },
+    {
+      "epoch": 138.7,
+      "learning_rate": 3.309375e-05,
+      "loss": 0.5377,
+      "step": 9570
+    },
+    {
+      "epoch": 138.84,
+      "learning_rate": 3.3015625e-05,
+      "loss": 0.5687,
+      "step": 9580
+    },
+    {
+      "epoch": 138.99,
+      "learning_rate": 3.29375e-05,
+      "loss": 0.5416,
+      "step": 9590
+    },
+    {
+      "epoch": 139.13,
+      "learning_rate": 3.2859375000000004e-05,
+      "loss": 0.5427,
+      "step": 9600
+    },
+    {
+      "epoch": 139.28,
+      "learning_rate": 3.278125e-05,
+      "loss": 0.577,
+      "step": 9610
+    },
+    {
+      "epoch": 139.42,
+      "learning_rate": 3.2703125000000004e-05,
+      "loss": 0.5943,
+      "step": 9620
+    },
+    {
+      "epoch": 139.57,
+      "learning_rate": 3.2625e-05,
+      "loss": 0.5481,
+      "step": 9630
+    },
+    {
+      "epoch": 139.71,
+      "learning_rate": 3.2546874999999997e-05,
+      "loss": 0.5904,
+      "step": 9640
+    },
+    {
+      "epoch": 139.86,
+      "learning_rate": 3.2468750000000007e-05,
+      "loss": 0.6591,
+      "step": 9650
+    },
+    {
+      "epoch": 140.0,
+      "learning_rate": 3.2390625e-05,
+      "loss": 0.6146,
+      "step": 9660
+    },
+    {
+      "epoch": 140.14,
+      "learning_rate": 3.23125e-05,
+      "loss": 0.5974,
+      "step": 9670
+    },
+    {
+      "epoch": 140.29,
+      "learning_rate": 3.2234375e-05,
+      "loss": 0.546,
+      "step": 9680
+    },
+    {
+      "epoch": 140.43,
+      "learning_rate": 3.215625e-05,
+      "loss": 0.5433,
+      "step": 9690
+    },
+    {
+      "epoch": 140.58,
+      "learning_rate": 3.2078125e-05,
+      "loss": 0.5357,
+      "step": 9700
+    },
+    {
+      "epoch": 140.72,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.6074,
+      "step": 9710
+    },
+    {
+      "epoch": 140.87,
+      "learning_rate": 3.1921875e-05,
+      "loss": 0.5543,
+      "step": 9720
+    },
+    {
+      "epoch": 141.01,
+      "learning_rate": 3.184375e-05,
+      "loss": 0.5415,
+      "step": 9730
+    },
+    {
+      "epoch": 141.16,
+      "learning_rate": 3.1765625e-05,
+      "loss": 0.5853,
+      "step": 9740
+    },
+    {
+      "epoch": 141.3,
+      "learning_rate": 3.1687500000000005e-05,
+      "loss": 0.5143,
+      "step": 9750
+    },
+    {
+      "epoch": 141.45,
+      "learning_rate": 3.1609375e-05,
+      "loss": 0.503,
+      "step": 9760
+    },
+    {
+      "epoch": 141.59,
+      "learning_rate": 3.1531250000000004e-05,
+      "loss": 0.5154,
+      "step": 9770
+    },
+    {
+      "epoch": 141.74,
+      "learning_rate": 3.1453125e-05,
+      "loss": 0.562,
+      "step": 9780
+    },
+    {
+      "epoch": 141.88,
+      "learning_rate": 3.1375e-05,
+      "loss": 0.5454,
+      "step": 9790
+    },
+    {
+      "epoch": 142.03,
+      "learning_rate": 3.1296875e-05,
+      "loss": 0.5722,
+      "step": 9800
+    },
+    {
+      "epoch": 142.17,
+      "learning_rate": 3.121875e-05,
+      "loss": 0.6911,
+      "step": 9810
+    },
+    {
+      "epoch": 142.32,
+      "learning_rate": 3.1140625e-05,
+      "loss": 0.5725,
+      "step": 9820
+    },
+    {
+      "epoch": 142.46,
+      "learning_rate": 3.10625e-05,
+      "loss": 0.5842,
+      "step": 9830
+    },
+    {
+      "epoch": 142.61,
+      "learning_rate": 3.0984375e-05,
+      "loss": 0.5597,
+      "step": 9840
+    },
+    {
+      "epoch": 142.75,
+      "learning_rate": 3.090625e-05,
+      "loss": 0.592,
+      "step": 9850
+    },
+    {
+      "epoch": 142.9,
+      "learning_rate": 3.0828125e-05,
+      "loss": 0.5506,
+      "step": 9860
+    },
+    {
+      "epoch": 143.04,
+      "learning_rate": 3.075e-05,
+      "loss": 0.6292,
+      "step": 9870
+    },
+    {
+      "epoch": 143.19,
+      "learning_rate": 3.0671875e-05,
+      "loss": 0.5672,
+      "step": 9880
+    },
+    {
+      "epoch": 143.33,
+      "learning_rate": 3.059375e-05,
+      "loss": 0.5315,
+      "step": 9890
+    },
+    {
+      "epoch": 143.48,
+      "learning_rate": 3.0515625000000005e-05,
+      "loss": 0.5563,
+      "step": 9900
+    },
+    {
+      "epoch": 143.62,
+      "learning_rate": 3.04375e-05,
+      "loss": 0.5877,
+      "step": 9910
+    },
+    {
+      "epoch": 143.77,
+      "learning_rate": 3.0359375e-05,
+      "loss": 0.546,
+      "step": 9920
+    },
+    {
+      "epoch": 143.91,
+      "learning_rate": 3.028125e-05,
+      "loss": 0.5425,
+      "step": 9930
+    },
+    {
+      "epoch": 144.06,
+      "learning_rate": 3.0203124999999997e-05,
+      "loss": 0.5368,
+      "step": 9940
+    },
+    {
+      "epoch": 144.2,
+      "learning_rate": 3.0125000000000004e-05,
+      "loss": 0.538,
+      "step": 9950
+    },
+    {
+      "epoch": 144.35,
+      "learning_rate": 3.0046875000000004e-05,
+      "loss": 0.5367,
+      "step": 9960
+    },
+    {
+      "epoch": 144.49,
+      "learning_rate": 2.996875e-05,
+      "loss": 0.5591,
+      "step": 9970
+    },
+    {
+      "epoch": 144.64,
+      "learning_rate": 2.9890625e-05,
+      "loss": 0.5554,
+      "step": 9980
+    },
+    {
+      "epoch": 144.78,
+      "learning_rate": 2.98125e-05,
+      "loss": 0.4985,
+      "step": 9990
+    },
+    {
+      "epoch": 144.93,
+      "learning_rate": 2.9734375000000003e-05,
+      "loss": 0.6306,
+      "step": 10000
+    },
+    {
+      "epoch": 144.93,
+      "eval_loss": 0.23715920746326447,
+      "eval_runtime": 577.4706,
+      "eval_samples_per_second": 5.877,
+      "eval_steps_per_second": 0.736,
+      "eval_wer": 0.15560157450550172,
+      "step": 10000
+    },
+    {
+      "epoch": 145.07,
+      "learning_rate": 2.9656250000000003e-05,
+      "loss": 0.5794,
+      "step": 10010
+    },
+    {
+      "epoch": 145.22,
+      "learning_rate": 2.9578125000000002e-05,
+      "loss": 0.5792,
+      "step": 10020
+    },
+    {
+      "epoch": 145.36,
+      "learning_rate": 2.95e-05,
+      "loss": 0.5664,
+      "step": 10030
+    },
+    {
+      "epoch": 145.51,
+      "learning_rate": 2.9421875e-05,
+      "loss": 0.5737,
+      "step": 10040
+    },
+    {
+      "epoch": 145.65,
+      "learning_rate": 2.9343750000000002e-05,
+      "loss": 0.5735,
+      "step": 10050
+    },
+    {
+      "epoch": 145.8,
+      "learning_rate": 2.9265625e-05,
+      "loss": 0.5972,
+      "step": 10060
+    },
+    {
+      "epoch": 145.94,
+      "learning_rate": 2.91875e-05,
+      "loss": 0.5998,
+      "step": 10070
+    },
+    {
+      "epoch": 146.09,
+      "learning_rate": 2.9109375e-05,
+      "loss": 0.5357,
+      "step": 10080
+    },
+    {
+      "epoch": 146.23,
+      "learning_rate": 2.9031249999999998e-05,
+      "loss": 0.522,
+      "step": 10090
+    },
+    {
+      "epoch": 146.38,
+      "learning_rate": 2.8953125000000004e-05,
+      "loss": 0.5231,
+      "step": 10100
+    },
+    {
+      "epoch": 146.52,
+      "learning_rate": 2.8875e-05,
+      "loss": 0.5343,
+      "step": 10110
+    },
+    {
+      "epoch": 146.67,
+      "learning_rate": 2.8796875e-05,
+      "loss": 0.5996,
+      "step": 10120
+    },
+    {
+      "epoch": 146.81,
+      "learning_rate": 2.871875e-05,
+      "loss": 0.5595,
+      "step": 10130
+    },
+    {
+      "epoch": 146.96,
+      "learning_rate": 2.8640625e-05,
+      "loss": 0.5386,
+      "step": 10140
+    },
+    {
+      "epoch": 147.1,
+      "learning_rate": 2.8562500000000003e-05,
+      "loss": 0.5338,
+      "step": 10150
+    },
+    {
+      "epoch": 147.25,
+      "learning_rate": 2.8484375000000003e-05,
+      "loss": 0.5273,
+      "step": 10160
+    },
+    {
+      "epoch": 147.39,
+      "learning_rate": 2.840625e-05,
+      "loss": 0.628,
+      "step": 10170
+    },
+    {
+      "epoch": 147.54,
+      "learning_rate": 2.8328125e-05,
+      "loss": 0.5602,
+      "step": 10180
+    },
+    {
+      "epoch": 147.68,
+      "learning_rate": 2.825e-05,
+      "loss": 0.621,
+      "step": 10190
+    },
+    {
+      "epoch": 147.83,
+      "learning_rate": 2.8171875000000002e-05,
+      "loss": 0.6112,
+      "step": 10200
+    },
+    {
+      "epoch": 147.97,
+      "learning_rate": 2.8093750000000002e-05,
+      "loss": 0.4686,
+      "step": 10210
+    },
+    {
+      "epoch": 148.12,
+      "learning_rate": 2.8015625e-05,
+      "loss": 0.5582,
+      "step": 10220
+    },
+    {
+      "epoch": 148.26,
+      "learning_rate": 2.79375e-05,
+      "loss": 0.5353,
+      "step": 10230
+    },
+    {
+      "epoch": 148.41,
+      "learning_rate": 2.7859374999999998e-05,
+      "loss": 0.5545,
+      "step": 10240
+    },
+    {
+      "epoch": 148.55,
+      "learning_rate": 2.7781250000000004e-05,
+      "loss": 0.497,
+      "step": 10250
+    },
+    {
+      "epoch": 148.7,
+      "learning_rate": 2.7703125e-05,
+      "loss": 0.6022,
+      "step": 10260
+    },
+    {
+      "epoch": 148.84,
+      "learning_rate": 2.7625e-05,
+      "loss": 0.5438,
+      "step": 10270
+    },
+    {
+      "epoch": 148.99,
+      "learning_rate": 2.7546875e-05,
+      "loss": 0.6067,
+      "step": 10280
+    },
+    {
+      "epoch": 149.13,
+      "learning_rate": 2.746875e-05,
+      "loss": 0.5966,
+      "step": 10290
+    },
+    {
+      "epoch": 149.28,
+      "learning_rate": 2.7390625000000003e-05,
+      "loss": 0.6345,
+      "step": 10300
+    },
+    {
+      "epoch": 149.42,
+      "learning_rate": 2.7312500000000003e-05,
+      "loss": 0.5105,
+      "step": 10310
+    },
+    {
+      "epoch": 149.57,
+      "learning_rate": 2.7234375e-05,
+      "loss": 0.573,
+      "step": 10320
+    },
+    {
+      "epoch": 149.71,
+      "learning_rate": 2.715625e-05,
+      "loss": 0.5542,
+      "step": 10330
+    },
+    {
+      "epoch": 149.86,
+      "learning_rate": 2.7078125e-05,
+      "loss": 0.6154,
+      "step": 10340
+    },
+    {
+      "epoch": 150.0,
+      "learning_rate": 2.7000000000000002e-05,
+      "loss": 0.5475,
+      "step": 10350
+    },
+    {
+      "epoch": 150.14,
+      "learning_rate": 2.6921875000000002e-05,
+      "loss": 0.5476,
+      "step": 10360
+    },
+    {
+      "epoch": 150.29,
+      "learning_rate": 2.6843750000000002e-05,
+      "loss": 0.5667,
+      "step": 10370
+    },
+    {
+      "epoch": 150.43,
+      "learning_rate": 2.6765625e-05,
+      "loss": 0.5752,
+      "step": 10380
+    },
+    {
+      "epoch": 150.58,
+      "learning_rate": 2.6687499999999998e-05,
+      "loss": 0.5797,
+      "step": 10390
+    },
+    {
+      "epoch": 150.72,
+      "learning_rate": 2.6609375000000005e-05,
+      "loss": 0.5585,
+      "step": 10400
+    },
+    {
+      "epoch": 150.87,
+      "learning_rate": 2.653125e-05,
+      "loss": 0.6719,
+      "step": 10410
+    },
+    {
+      "epoch": 151.01,
+      "learning_rate": 2.6453125e-05,
+      "loss": 0.535,
+      "step": 10420
+    },
+    {
+      "epoch": 151.16,
+      "learning_rate": 2.6375e-05,
+      "loss": 0.5853,
+      "step": 10430
+    },
+    {
+      "epoch": 151.3,
+      "learning_rate": 2.6296874999999997e-05,
+      "loss": 0.5341,
+      "step": 10440
+    },
+    {
+      "epoch": 151.45,
+      "learning_rate": 2.6218750000000004e-05,
+      "loss": 0.4932,
+      "step": 10450
+    },
+    {
+      "epoch": 151.59,
+      "learning_rate": 2.6140625000000004e-05,
+      "loss": 0.5924,
+      "step": 10460
+    },
+    {
+      "epoch": 151.74,
+      "learning_rate": 2.60625e-05,
+      "loss": 0.5665,
+      "step": 10470
+    },
+    {
+      "epoch": 151.88,
+      "learning_rate": 2.5984375e-05,
+      "loss": 0.5202,
+      "step": 10480
+    },
+    {
+      "epoch": 152.03,
+      "learning_rate": 2.590625e-05,
+      "loss": 0.5209,
+      "step": 10490
+    },
+    {
+      "epoch": 152.17,
+      "learning_rate": 2.5828125000000003e-05,
+      "loss": 0.5874,
+      "step": 10500
+    },
+    {
+      "epoch": 152.17,
+      "eval_loss": 0.23621943593025208,
+      "eval_runtime": 575.3543,
+      "eval_samples_per_second": 5.899,
+      "eval_steps_per_second": 0.739,
+      "eval_wer": 0.15330884467782627,
+      "step": 10500
+    },
+    {
+      "epoch": 152.32,
+      "learning_rate": 2.5750000000000002e-05,
+      "loss": 0.57,
+      "step": 10510
+    },
+    {
+      "epoch": 152.46,
+      "learning_rate": 2.5671875000000002e-05,
+      "loss": 0.58,
+      "step": 10520
+    },
+    {
+      "epoch": 152.61,
+      "learning_rate": 2.559375e-05,
+      "loss": 0.5269,
+      "step": 10530
+    },
+    {
+      "epoch": 152.75,
+      "learning_rate": 2.5515625e-05,
+      "loss": 0.5519,
+      "step": 10540
+    },
+    {
+      "epoch": 152.9,
+      "learning_rate": 2.54375e-05,
+      "loss": 0.497,
+      "step": 10550
+    },
+    {
+      "epoch": 153.04,
+      "learning_rate": 2.5359375e-05,
+      "loss": 0.5686,
+      "step": 10560
+    },
+    {
+      "epoch": 153.19,
+      "learning_rate": 2.528125e-05,
+      "loss": 0.5219,
+      "step": 10570
+    },
+    {
+      "epoch": 153.33,
+      "learning_rate": 2.5203125e-05,
+      "loss": 0.5695,
+      "step": 10580
+    },
+    {
+      "epoch": 153.48,
+      "learning_rate": 2.5124999999999997e-05,
+      "loss": 0.5747,
+      "step": 10590
+    },
+    {
+      "epoch": 153.62,
+      "learning_rate": 2.5046875000000004e-05,
+      "loss": 0.5297,
+      "step": 10600
+    },
+    {
+      "epoch": 153.77,
+      "learning_rate": 2.496875e-05,
+      "loss": 0.5956,
+      "step": 10610
+    },
+    {
+      "epoch": 153.91,
+      "learning_rate": 2.4890625e-05,
+      "loss": 0.5752,
+      "step": 10620
+    },
+    {
+      "epoch": 154.06,
+      "learning_rate": 2.4812500000000003e-05,
+      "loss": 0.5765,
+      "step": 10630
+    },
+    {
+      "epoch": 154.2,
+      "learning_rate": 2.4734375e-05,
+      "loss": 0.5627,
+      "step": 10640
+    },
+    {
+      "epoch": 154.35,
+      "learning_rate": 2.465625e-05,
+      "loss": 0.5591,
+      "step": 10650
+    },
+    {
+      "epoch": 154.49,
+      "learning_rate": 2.4578125000000003e-05,
+      "loss": 0.5595,
+      "step": 10660
+    },
+    {
+      "epoch": 154.64,
+      "learning_rate": 2.45e-05,
+      "loss": 0.5978,
+      "step": 10670
+    },
+    {
+      "epoch": 154.78,
+      "learning_rate": 2.4421875000000002e-05,
+      "loss": 0.5249,
+      "step": 10680
+    },
+    {
+      "epoch": 154.93,
+      "learning_rate": 2.4343750000000002e-05,
+      "loss": 0.5522,
+      "step": 10690
+    },
+    {
+      "epoch": 155.07,
+      "learning_rate": 2.4265625e-05,
+      "loss": 0.5152,
+      "step": 10700
+    },
+    {
+      "epoch": 155.22,
+      "learning_rate": 2.4187500000000002e-05,
+      "loss": 0.6,
+      "step": 10710
+    },
+    {
+      "epoch": 155.36,
+      "learning_rate": 2.4109375e-05,
+      "loss": 0.588,
+      "step": 10720
+    },
+    {
+      "epoch": 155.51,
+      "learning_rate": 2.403125e-05,
+      "loss": 0.52,
+      "step": 10730
+    },
+    {
+      "epoch": 155.65,
+      "learning_rate": 2.3953125e-05,
+      "loss": 0.5191,
+      "step": 10740
+    },
+    {
+      "epoch": 155.8,
+      "learning_rate": 2.3875e-05,
+      "loss": 0.5456,
+      "step": 10750
+    },
+    {
+      "epoch": 155.94,
+      "learning_rate": 2.3796875e-05,
+      "loss": 0.6365,
+      "step": 10760
+    },
+    {
+      "epoch": 156.09,
+      "learning_rate": 2.371875e-05,
+      "loss": 0.5344,
+      "step": 10770
+    },
+    {
+      "epoch": 156.23,
+      "learning_rate": 2.3640625000000004e-05,
+      "loss": 0.5185,
+      "step": 10780
+    },
+    {
+      "epoch": 156.38,
+      "learning_rate": 2.35625e-05,
+      "loss": 0.5619,
+      "step": 10790
+    },
+    {
+      "epoch": 156.52,
+      "learning_rate": 2.3484375e-05,
+      "loss": 0.5248,
+      "step": 10800
+    },
+    {
+      "epoch": 156.67,
+      "learning_rate": 2.3406250000000003e-05,
+      "loss": 0.499,
+      "step": 10810
+    },
+    {
+      "epoch": 156.81,
+      "learning_rate": 2.3328125e-05,
+      "loss": 0.5657,
+      "step": 10820
+    },
+    {
+      "epoch": 156.96,
+      "learning_rate": 2.3250000000000003e-05,
+      "loss": 0.52,
+      "step": 10830
+    },
+    {
+      "epoch": 157.1,
+      "learning_rate": 2.3171875000000003e-05,
+      "loss": 0.5119,
+      "step": 10840
+    },
+    {
+      "epoch": 157.25,
+      "learning_rate": 2.309375e-05,
+      "loss": 0.5749,
+      "step": 10850
+    },
+    {
+      "epoch": 157.39,
+      "learning_rate": 2.3015625000000002e-05,
+      "loss": 0.57,
+      "step": 10860
+    },
+    {
+      "epoch": 157.54,
+      "learning_rate": 2.2937500000000002e-05,
+      "loss": 0.5096,
+      "step": 10870
+    },
+    {
+      "epoch": 157.68,
+      "learning_rate": 2.2859375e-05,
+      "loss": 0.5025,
+      "step": 10880
+    },
+    {
+      "epoch": 157.83,
+      "learning_rate": 2.278125e-05,
+      "loss": 0.5884,
+      "step": 10890
+    },
+    {
+      "epoch": 157.97,
+      "learning_rate": 2.2703125e-05,
+      "loss": 0.4872,
+      "step": 10900
+    },
+    {
+      "epoch": 158.12,
+      "learning_rate": 2.2625e-05,
+      "loss": 0.5518,
+      "step": 10910
+    },
+    {
+      "epoch": 158.26,
+      "learning_rate": 2.2546875e-05,
+      "loss": 0.5348,
+      "step": 10920
+    },
+    {
+      "epoch": 158.41,
+      "learning_rate": 2.246875e-05,
+      "loss": 0.5997,
+      "step": 10930
+    },
+    {
+      "epoch": 158.55,
+      "learning_rate": 2.2390625e-05,
+      "loss": 0.5602,
+      "step": 10940
+    },
+    {
+      "epoch": 158.7,
+      "learning_rate": 2.23125e-05,
+      "loss": 0.5996,
+      "step": 10950
+    },
+    {
+      "epoch": 158.84,
+      "learning_rate": 2.2234375e-05,
+      "loss": 0.6152,
+      "step": 10960
+    },
+    {
+      "epoch": 158.99,
+      "learning_rate": 2.215625e-05,
+      "loss": 0.5291,
+      "step": 10970
+    },
+    {
+      "epoch": 159.13,
+      "learning_rate": 2.2078125000000003e-05,
+      "loss": 0.55,
+      "step": 10980
+    },
+    {
+      "epoch": 159.28,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 0.6242,
+      "step": 10990
+    },
+    {
+      "epoch": 159.42,
+      "learning_rate": 2.1921875e-05,
+      "loss": 0.5546,
+      "step": 11000
+    },
+    {
+      "epoch": 159.42,
+      "eval_loss": 0.23421239852905273,
+      "eval_runtime": 576.3026,
+      "eval_samples_per_second": 5.889,
+      "eval_steps_per_second": 0.737,
+      "eval_wer": 0.15427852969454922,
+      "step": 11000
+    },
+    {
+      "epoch": 159.57,
+      "learning_rate": 2.1843750000000002e-05,
+      "loss": 0.5893,
+      "step": 11010
+    },
+    {
+      "epoch": 159.71,
+      "learning_rate": 2.1765625000000002e-05,
+      "loss": 0.5654,
+      "step": 11020
+    },
+    {
+      "epoch": 159.86,
+      "learning_rate": 2.1687500000000002e-05,
+      "loss": 0.5748,
+      "step": 11030
+    },
+    {
+      "epoch": 160.0,
+      "learning_rate": 2.1609375000000002e-05,
+      "loss": 0.5317,
+      "step": 11040
+    },
+    {
+      "epoch": 160.14,
+      "learning_rate": 2.153125e-05,
+      "loss": 0.4892,
+      "step": 11050
+    },
+    {
+      "epoch": 160.29,
+      "learning_rate": 2.1453125e-05,
+      "loss": 0.544,
+      "step": 11060
+    },
+    {
+      "epoch": 160.43,
+      "learning_rate": 2.1375e-05,
+      "loss": 0.5197,
+      "step": 11070
+    },
+    {
+      "epoch": 160.58,
+      "learning_rate": 2.1296875e-05,
+      "loss": 0.5127,
+      "step": 11080
+    },
+    {
+      "epoch": 160.72,
+      "learning_rate": 2.121875e-05,
+      "loss": 0.5438,
+      "step": 11090
+    },
+    {
+      "epoch": 160.87,
+      "learning_rate": 2.1140625e-05,
+      "loss": 0.5184,
+      "step": 11100
+    },
+    {
+      "epoch": 161.01,
+      "learning_rate": 2.10625e-05,
+      "loss": 0.6122,
+      "step": 11110
+    },
+    {
+      "epoch": 161.16,
+      "learning_rate": 2.0984375e-05,
+      "loss": 0.5086,
+      "step": 11120
+    },
+    {
+      "epoch": 161.3,
+      "learning_rate": 2.0906250000000003e-05,
+      "loss": 0.6104,
+      "step": 11130
+    },
+    {
+      "epoch": 161.45,
+      "learning_rate": 2.0828125e-05,
+      "loss": 0.5168,
+      "step": 11140
+    },
+    {
+      "epoch": 161.59,
+      "learning_rate": 2.075e-05,
+      "loss": 0.5145,
+      "step": 11150
+    },
+    {
+      "epoch": 161.74,
+      "learning_rate": 2.0671875000000003e-05,
+      "loss": 0.5389,
+      "step": 11160
+    },
+    {
+      "epoch": 161.88,
+      "learning_rate": 2.059375e-05,
+      "loss": 0.671,
+      "step": 11170
+    },
+    {
+      "epoch": 162.03,
+      "learning_rate": 2.05234375e-05,
+      "loss": 0.5659,
+      "step": 11180
+    },
+    {
+      "epoch": 162.17,
+      "learning_rate": 2.0453125e-05,
+      "loss": 0.5035,
+      "step": 11190
+    },
+    {
+      "epoch": 162.32,
+      "learning_rate": 2.0375e-05,
+      "loss": 0.5038,
+      "step": 11200
+    },
+    {
+      "epoch": 162.46,
+      "learning_rate": 2.0296875e-05,
+      "loss": 0.6035,
+      "step": 11210
+    },
+    {
+      "epoch": 162.61,
+      "learning_rate": 2.021875e-05,
+      "loss": 0.5472,
+      "step": 11220
+    },
+    {
+      "epoch": 162.75,
+      "learning_rate": 2.0140625000000003e-05,
+      "loss": 0.5149,
+      "step": 11230
+    },
+    {
+      "epoch": 162.9,
+      "learning_rate": 2.00625e-05,
+      "loss": 0.6043,
+      "step": 11240
+    },
+    {
+      "epoch": 163.04,
+      "learning_rate": 1.9984375e-05,
+      "loss": 0.5377,
+      "step": 11250
+    },
+    {
+      "epoch": 163.19,
+      "learning_rate": 1.9906250000000003e-05,
+      "loss": 0.5328,
+      "step": 11260
+    },
+    {
+      "epoch": 163.33,
+      "learning_rate": 1.9828125e-05,
+      "loss": 0.4993,
+      "step": 11270
+    },
+    {
+      "epoch": 163.48,
+      "learning_rate": 1.9750000000000002e-05,
+      "loss": 0.5234,
+      "step": 11280
+    },
+    {
+      "epoch": 163.62,
+      "learning_rate": 1.9671875000000002e-05,
+      "loss": 0.6392,
+      "step": 11290
+    },
+    {
+      "epoch": 163.77,
+      "learning_rate": 1.959375e-05,
+      "loss": 0.5308,
+      "step": 11300
+    },
+    {
+      "epoch": 163.91,
+      "learning_rate": 1.9515625000000002e-05,
+      "loss": 0.5907,
+      "step": 11310
+    },
+    {
+      "epoch": 164.06,
+      "learning_rate": 1.94375e-05,
+      "loss": 0.5679,
+      "step": 11320
+    },
+    {
+      "epoch": 164.2,
+      "learning_rate": 1.9359375e-05,
+      "loss": 0.5094,
+      "step": 11330
+    },
+    {
+      "epoch": 164.35,
+      "learning_rate": 1.928125e-05,
+      "loss": 0.5877,
+      "step": 11340
+    },
+    {
+      "epoch": 164.49,
+      "learning_rate": 1.9203125e-05,
+      "loss": 0.5177,
+      "step": 11350
+    },
+    {
+      "epoch": 164.64,
+      "learning_rate": 1.9125e-05,
+      "loss": 0.5452,
+      "step": 11360
+    },
+    {
+      "epoch": 164.78,
+      "learning_rate": 1.9046875e-05,
+      "loss": 0.568,
+      "step": 11370
+    },
+    {
+      "epoch": 164.93,
+      "learning_rate": 1.896875e-05,
+      "loss": 0.5717,
+      "step": 11380
+    },
+    {
+      "epoch": 165.07,
+      "learning_rate": 1.8890625e-05,
+      "loss": 0.5861,
+      "step": 11390
+    },
+    {
+      "epoch": 165.22,
+      "learning_rate": 1.88125e-05,
+      "loss": 0.5153,
+      "step": 11400
+    },
+    {
+      "epoch": 165.36,
+      "learning_rate": 1.8734375e-05,
+      "loss": 0.5797,
+      "step": 11410
+    },
+    {
+      "epoch": 165.51,
+      "learning_rate": 1.865625e-05,
+      "loss": 0.5237,
+      "step": 11420
+    },
+    {
+      "epoch": 165.65,
+      "learning_rate": 1.8578125000000003e-05,
+      "loss": 0.5606,
+      "step": 11430
+    },
+    {
+      "epoch": 165.8,
+      "learning_rate": 1.85e-05,
+      "loss": 0.5389,
+      "step": 11440
+    },
+    {
+      "epoch": 165.94,
+      "learning_rate": 1.8421875e-05,
+      "loss": 0.572,
+      "step": 11450
+    },
+    {
+      "epoch": 166.09,
+      "learning_rate": 1.8343750000000002e-05,
+      "loss": 0.5169,
+      "step": 11460
+    },
+    {
+      "epoch": 166.23,
+      "learning_rate": 1.8265625e-05,
+      "loss": 0.506,
+      "step": 11470
+    },
+    {
+      "epoch": 166.38,
+      "learning_rate": 1.81875e-05,
+      "loss": 0.5406,
+      "step": 11480
+    },
+    {
+      "epoch": 166.52,
+      "learning_rate": 1.8109375e-05,
+      "loss": 0.5913,
+      "step": 11490
+    },
+    {
+      "epoch": 166.67,
+      "learning_rate": 1.803125e-05,
+      "loss": 0.6294,
+      "step": 11500
+    },
+    {
+      "epoch": 166.67,
+      "eval_loss": 0.23806829750537872,
+      "eval_runtime": 576.0097,
+      "eval_samples_per_second": 5.892,
+      "eval_steps_per_second": 0.738,
+      "eval_wer": 0.15362933379352284,
+      "step": 11500
+    },
+    {
+      "epoch": 166.81,
+      "learning_rate": 1.7953125e-05,
+      "loss": 0.51,
+      "step": 11510
+    },
+    {
+      "epoch": 166.96,
+      "learning_rate": 1.7875e-05,
+      "loss": 0.5439,
+      "step": 11520
+    },
+    {
+      "epoch": 167.1,
+      "learning_rate": 1.7796875e-05,
+      "loss": 0.4877,
+      "step": 11530
+    },
+    {
+      "epoch": 167.25,
+      "learning_rate": 1.771875e-05,
+      "loss": 0.5299,
+      "step": 11540
+    },
+    {
+      "epoch": 167.39,
+      "learning_rate": 1.7640625e-05,
+      "loss": 0.5475,
+      "step": 11550
+    },
+    {
+      "epoch": 167.54,
+      "learning_rate": 1.75625e-05,
+      "loss": 0.5661,
+      "step": 11560
+    },
+    {
+      "epoch": 167.68,
+      "learning_rate": 1.7484375e-05,
+      "loss": 0.5416,
+      "step": 11570
+    },
+    {
+      "epoch": 167.83,
+      "learning_rate": 1.7406250000000003e-05,
+      "loss": 0.4929,
+      "step": 11580
+    },
+    {
+      "epoch": 167.97,
+      "learning_rate": 1.7328125e-05,
+      "loss": 0.5195,
+      "step": 11590
+    },
+    {
+      "epoch": 168.12,
+      "learning_rate": 1.725e-05,
+      "loss": 0.5883,
+      "step": 11600
+    },
+    {
+      "epoch": 168.26,
+      "learning_rate": 1.7171875000000002e-05,
+      "loss": 0.5723,
+      "step": 11610
+    },
+    {
+      "epoch": 168.41,
+      "learning_rate": 1.709375e-05,
+      "loss": 0.5621,
+      "step": 11620
+    },
+    {
+      "epoch": 168.55,
+      "learning_rate": 1.7015625000000002e-05,
+      "loss": 0.5567,
+      "step": 11630
+    },
+    {
+      "epoch": 168.7,
+      "learning_rate": 1.6937500000000002e-05,
+      "loss": 0.4995,
+      "step": 11640
+    },
+    {
+      "epoch": 168.84,
+      "learning_rate": 1.6859374999999998e-05,
+      "loss": 0.5601,
+      "step": 11650
+    },
+    {
+      "epoch": 168.99,
+      "learning_rate": 1.678125e-05,
+      "loss": 0.5122,
+      "step": 11660
+    },
+    {
+      "epoch": 169.13,
+      "learning_rate": 1.6703125e-05,
+      "loss": 0.5406,
+      "step": 11670
+    },
+    {
+      "epoch": 169.28,
+      "learning_rate": 1.6625e-05,
+      "loss": 0.4914,
+      "step": 11680
+    },
+    {
+      "epoch": 169.42,
+      "learning_rate": 1.6546875e-05,
+      "loss": 0.6525,
+      "step": 11690
+    },
+    {
+      "epoch": 169.57,
+      "learning_rate": 1.646875e-05,
+      "loss": 0.5181,
+      "step": 11700
+    },
+    {
+      "epoch": 169.71,
+      "learning_rate": 1.6390625e-05,
+      "loss": 0.5328,
+      "step": 11710
+    },
+    {
+      "epoch": 169.86,
+      "learning_rate": 1.63125e-05,
+      "loss": 0.5583,
+      "step": 11720
+    },
+    {
+      "epoch": 170.0,
+      "learning_rate": 1.6234375000000003e-05,
+      "loss": 0.5376,
+      "step": 11730
+    },
+    {
+      "epoch": 170.14,
+      "learning_rate": 1.615625e-05,
+      "loss": 0.5182,
+      "step": 11740
+    },
+    {
+      "epoch": 170.29,
+      "learning_rate": 1.6078125e-05,
+      "loss": 0.4756,
+      "step": 11750
+    },
+    {
+      "epoch": 170.43,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.5295,
+      "step": 11760
+    },
+    {
+      "epoch": 170.58,
+      "learning_rate": 1.5921875e-05,
+      "loss": 0.5482,
+      "step": 11770
+    },
+    {
+      "epoch": 170.72,
+      "learning_rate": 1.5843750000000002e-05,
+      "loss": 0.5513,
+      "step": 11780
+    },
+    {
+      "epoch": 170.87,
+      "learning_rate": 1.5765625000000002e-05,
+      "loss": 0.5341,
+      "step": 11790
+    },
+    {
+      "epoch": 171.01,
+      "learning_rate": 1.56875e-05,
+      "loss": 0.4755,
+      "step": 11800
+    },
+    {
+      "epoch": 171.16,
+      "learning_rate": 1.5609375e-05,
+      "loss": 0.5352,
+      "step": 11810
+    },
+    {
+      "epoch": 171.3,
+      "learning_rate": 1.553125e-05,
+      "loss": 0.5318,
+      "step": 11820
+    },
+    {
+      "epoch": 171.45,
+      "learning_rate": 1.5453125e-05,
+      "loss": 0.4488,
+      "step": 11830
+    },
+    {
+      "epoch": 171.59,
+      "learning_rate": 1.5375e-05,
+      "loss": 0.6219,
+      "step": 11840
+    },
+    {
+      "epoch": 171.74,
+      "learning_rate": 1.5296875e-05,
+      "loss": 0.5449,
+      "step": 11850
+    },
+    {
+      "epoch": 171.88,
+      "learning_rate": 1.521875e-05,
+      "loss": 0.5278,
+      "step": 11860
+    },
+    {
+      "epoch": 172.03,
+      "learning_rate": 1.5140625e-05,
+      "loss": 0.539,
+      "step": 11870
+    },
+    {
+      "epoch": 172.17,
+      "learning_rate": 1.5062500000000002e-05,
+      "loss": 0.5436,
+      "step": 11880
+    },
+    {
+      "epoch": 172.32,
+      "learning_rate": 1.4984375e-05,
+      "loss": 0.5874,
+      "step": 11890
+    },
+    {
+      "epoch": 172.46,
+      "learning_rate": 1.490625e-05,
+      "loss": 0.5529,
+      "step": 11900
+    },
+    {
+      "epoch": 172.61,
+      "learning_rate": 1.4828125000000001e-05,
+      "loss": 0.5403,
+      "step": 11910
+    },
+    {
+      "epoch": 172.75,
+      "learning_rate": 1.475e-05,
+      "loss": 0.4999,
+      "step": 11920
+    },
+    {
+      "epoch": 172.9,
+      "learning_rate": 1.4671875000000001e-05,
+      "loss": 0.6045,
+      "step": 11930
+    },
+    {
+      "epoch": 173.04,
+      "learning_rate": 1.459375e-05,
+      "loss": 0.5237,
+      "step": 11940
+    },
+    {
+      "epoch": 173.19,
+      "learning_rate": 1.4515624999999999e-05,
+      "loss": 0.5556,
+      "step": 11950
+    },
+    {
+      "epoch": 173.33,
+      "learning_rate": 1.44375e-05,
+      "loss": 0.5837,
+      "step": 11960
+    },
+    {
+      "epoch": 173.48,
+      "learning_rate": 1.4359375e-05,
+      "loss": 0.5021,
+      "step": 11970
+    },
+    {
+      "epoch": 173.62,
+      "learning_rate": 1.4281250000000002e-05,
+      "loss": 0.5294,
+      "step": 11980
+    },
+    {
+      "epoch": 173.77,
+      "learning_rate": 1.4203125e-05,
+      "loss": 0.5379,
+      "step": 11990
+    },
+    {
+      "epoch": 173.91,
+      "learning_rate": 1.4125e-05,
+      "loss": 0.5989,
+      "step": 12000
+    },
+    {
+      "epoch": 173.91,
+      "eval_loss": 0.23597407341003418,
+      "eval_runtime": 575.4715,
+      "eval_samples_per_second": 5.898,
+      "eval_steps_per_second": 0.739,
+      "eval_wer": 0.1526843017856996,
+      "step": 12000
+    },
+    {
+      "epoch": 174.06,
+      "learning_rate": 1.4046875000000001e-05,
+      "loss": 0.5167,
+      "step": 12010
+    },
+    {
+      "epoch": 174.2,
+      "learning_rate": 1.396875e-05,
+      "loss": 0.5127,
+      "step": 12020
+    },
+    {
+      "epoch": 174.35,
+      "learning_rate": 1.3890625000000002e-05,
+      "loss": 0.613,
+      "step": 12030
+    },
+    {
+      "epoch": 174.49,
+      "learning_rate": 1.38125e-05,
+      "loss": 0.5287,
+      "step": 12040
+    },
+    {
+      "epoch": 174.64,
+      "learning_rate": 1.3734375e-05,
+      "loss": 0.5609,
+      "step": 12050
+    },
+    {
+      "epoch": 174.78,
+      "learning_rate": 1.3656250000000002e-05,
+      "loss": 0.5278,
+      "step": 12060
+    },
+    {
+      "epoch": 174.93,
+      "learning_rate": 1.3578125e-05,
+      "loss": 0.5928,
+      "step": 12070
+    },
+    {
+      "epoch": 175.07,
+      "learning_rate": 1.3500000000000001e-05,
+      "loss": 0.56,
+      "step": 12080
+    },
+    {
+      "epoch": 175.22,
+      "learning_rate": 1.3421875000000001e-05,
+      "loss": 0.5716,
+      "step": 12090
+    },
+    {
+      "epoch": 175.36,
+      "learning_rate": 1.3343749999999999e-05,
+      "loss": 0.5499,
+      "step": 12100
+    },
+    {
+      "epoch": 175.51,
+      "learning_rate": 1.3265625e-05,
+      "loss": 0.476,
+      "step": 12110
+    },
+    {
+      "epoch": 175.65,
+      "learning_rate": 1.31875e-05,
+      "loss": 0.5134,
+      "step": 12120
+    },
+    {
+      "epoch": 175.8,
+      "learning_rate": 1.3109375000000002e-05,
+      "loss": 0.5093,
+      "step": 12130
+    },
+    {
+      "epoch": 175.94,
+      "learning_rate": 1.303125e-05,
+      "loss": 0.579,
+      "step": 12140
+    },
+    {
+      "epoch": 176.09,
+      "learning_rate": 1.2953125e-05,
+      "loss": 0.4885,
+      "step": 12150
+    },
+    {
+      "epoch": 176.23,
+      "learning_rate": 1.2875000000000001e-05,
+      "loss": 0.5433,
+      "step": 12160
+    },
+    {
+      "epoch": 176.38,
+      "learning_rate": 1.2796875e-05,
+      "loss": 0.5137,
+      "step": 12170
+    },
+    {
+      "epoch": 176.52,
+      "learning_rate": 1.271875e-05,
+      "loss": 0.5202,
+      "step": 12180
+    },
+    {
+      "epoch": 176.67,
+      "learning_rate": 1.2640625e-05,
+      "loss": 0.5336,
+      "step": 12190
+    },
+    {
+      "epoch": 176.81,
+      "learning_rate": 1.2562499999999999e-05,
+      "loss": 0.5573,
+      "step": 12200
+    },
+    {
+      "epoch": 176.96,
+      "learning_rate": 1.2484375e-05,
+      "loss": 0.4605,
+      "step": 12210
+    },
+    {
+      "epoch": 177.1,
+      "learning_rate": 1.2406250000000002e-05,
+      "loss": 0.4848,
+      "step": 12220
+    },
+    {
+      "epoch": 177.25,
+      "learning_rate": 1.2328125e-05,
+      "loss": 0.5071,
+      "step": 12230
+    },
+    {
+      "epoch": 177.39,
+      "learning_rate": 1.225e-05,
+      "loss": 0.4797,
+      "step": 12240
+    },
+    {
+      "epoch": 177.54,
+      "learning_rate": 1.2171875000000001e-05,
+      "loss": 0.5308,
+      "step": 12250
+    },
+    {
+      "epoch": 177.68,
+      "learning_rate": 1.2093750000000001e-05,
+      "loss": 0.6085,
+      "step": 12260
+    },
+    {
+      "epoch": 177.83,
+      "learning_rate": 1.2015625e-05,
+      "loss": 0.5489,
+      "step": 12270
+    },
+    {
+      "epoch": 177.97,
+      "learning_rate": 1.19375e-05,
+      "loss": 0.4995,
+      "step": 12280
+    },
+    {
+      "epoch": 178.12,
+      "learning_rate": 1.1859375e-05,
+      "loss": 0.5686,
+      "step": 12290
+    },
+    {
+      "epoch": 178.26,
+      "learning_rate": 1.178125e-05,
+      "loss": 0.5471,
+      "step": 12300
+    },
+    {
+      "epoch": 178.41,
+      "learning_rate": 1.1703125000000002e-05,
+      "loss": 0.5439,
+      "step": 12310
+    },
+    {
+      "epoch": 178.55,
+      "learning_rate": 1.1625000000000001e-05,
+      "loss": 0.53,
+      "step": 12320
+    },
+    {
+      "epoch": 178.7,
+      "learning_rate": 1.1546875e-05,
+      "loss": 0.5401,
+      "step": 12330
+    },
+    {
+      "epoch": 178.84,
+      "learning_rate": 1.1468750000000001e-05,
+      "loss": 0.5493,
+      "step": 12340
+    },
+    {
+      "epoch": 178.99,
+      "learning_rate": 1.1390625e-05,
+      "loss": 0.5021,
+      "step": 12350
+    },
+    {
+      "epoch": 179.13,
+      "learning_rate": 1.13125e-05,
+      "loss": 0.5518,
+      "step": 12360
+    },
+    {
+      "epoch": 179.28,
+      "learning_rate": 1.1234375e-05,
+      "loss": 0.512,
+      "step": 12370
+    },
+    {
+      "epoch": 179.42,
+      "learning_rate": 1.115625e-05,
+      "loss": 0.503,
+      "step": 12380
+    },
+    {
+      "epoch": 179.57,
+      "learning_rate": 1.1078125e-05,
+      "loss": 0.5771,
+      "step": 12390
+    },
+    {
+      "epoch": 179.71,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 0.5227,
+      "step": 12400
+    },
+    {
+      "epoch": 179.86,
+      "learning_rate": 1.0921875000000001e-05,
+      "loss": 0.5993,
+      "step": 12410
+    },
+    {
+      "epoch": 180.0,
+      "learning_rate": 1.0843750000000001e-05,
+      "loss": 0.4581,
+      "step": 12420
+    },
+    {
+      "epoch": 180.14,
+      "learning_rate": 1.0765625e-05,
+      "loss": 0.523,
+      "step": 12430
+    },
+    {
+      "epoch": 180.29,
+      "learning_rate": 1.06875e-05,
+      "loss": 0.5472,
+      "step": 12440
+    },
+    {
+      "epoch": 180.43,
+      "learning_rate": 1.0609375e-05,
+      "loss": 0.595,
+      "step": 12450
+    },
+    {
+      "epoch": 180.58,
+      "learning_rate": 1.053125e-05,
+      "loss": 0.5503,
+      "step": 12460
+    },
+    {
+      "epoch": 180.72,
+      "learning_rate": 1.0453125000000002e-05,
+      "loss": 0.5245,
+      "step": 12470
+    },
+    {
+      "epoch": 180.87,
+      "learning_rate": 1.0375e-05,
+      "loss": 0.5567,
+      "step": 12480
+    },
+    {
+      "epoch": 181.01,
+      "learning_rate": 1.0296875e-05,
+      "loss": 0.4907,
+      "step": 12490
+    },
+    {
+      "epoch": 181.16,
+      "learning_rate": 1.0218750000000001e-05,
+      "loss": 0.5697,
+      "step": 12500
+    },
+    {
+      "epoch": 181.16,
+      "eval_loss": 0.23992973566055298,
+      "eval_runtime": 578.3824,
+      "eval_samples_per_second": 5.868,
+      "eval_steps_per_second": 0.735,
+      "eval_wer": 0.15259390741973392,
+      "step": 12500
+    },
+    {
+      "epoch": 181.3,
+      "learning_rate": 1.0140625000000001e-05,
+      "loss": 0.5403,
+      "step": 12510
+    },
+    {
+      "epoch": 181.45,
+      "learning_rate": 1.00625e-05,
+      "loss": 0.5109,
+      "step": 12520
+    },
+    {
+      "epoch": 181.59,
+      "learning_rate": 9.984375e-06,
+      "loss": 0.4829,
+      "step": 12530
+    },
+    {
+      "epoch": 181.74,
+      "learning_rate": 9.90625e-06,
+      "loss": 0.5025,
+      "step": 12540
+    },
+    {
+      "epoch": 181.88,
+      "learning_rate": 9.828125e-06,
+      "loss": 0.5442,
+      "step": 12550
+    },
+    {
+      "epoch": 182.03,
+      "learning_rate": 9.750000000000002e-06,
+      "loss": 0.5034,
+      "step": 12560
+    },
+    {
+      "epoch": 182.17,
+      "learning_rate": 9.671875000000001e-06,
+      "loss": 0.4793,
+      "step": 12570
+    },
+    {
+      "epoch": 182.32,
+      "learning_rate": 9.59375e-06,
+      "loss": 0.5074,
+      "step": 12580
+    },
+    {
+      "epoch": 182.46,
+      "learning_rate": 9.515625000000001e-06,
+      "loss": 0.5132,
+      "step": 12590
+    },
+    {
+      "epoch": 182.61,
+      "learning_rate": 9.4375e-06,
+      "loss": 0.5744,
+      "step": 12600
+    },
+    {
+      "epoch": 182.75,
+      "learning_rate": 9.359375e-06,
+      "loss": 0.4714,
+      "step": 12610
+    },
+    {
+      "epoch": 182.9,
+      "learning_rate": 9.28125e-06,
+      "loss": 0.566,
+      "step": 12620
+    },
+    {
+      "epoch": 183.04,
+      "learning_rate": 9.203125e-06,
+      "loss": 0.5804,
+      "step": 12630
+    },
+    {
+      "epoch": 183.19,
+      "learning_rate": 9.125e-06,
+      "loss": 0.4781,
+      "step": 12640
+    },
+    {
+      "epoch": 183.33,
+      "learning_rate": 9.046875e-06,
+      "loss": 0.4948,
+      "step": 12650
+    },
+    {
+      "epoch": 183.48,
+      "learning_rate": 8.968750000000001e-06,
+      "loss": 0.4981,
+      "step": 12660
+    },
+    {
+      "epoch": 183.62,
+      "learning_rate": 8.890625000000001e-06,
+      "loss": 0.5249,
+      "step": 12670
+    },
+    {
+      "epoch": 183.77,
+      "learning_rate": 8.8125e-06,
+      "loss": 0.5197,
+      "step": 12680
+    },
+    {
+      "epoch": 183.91,
+      "learning_rate": 8.734375e-06,
+      "loss": 0.5002,
+      "step": 12690
+    },
+    {
+      "epoch": 184.06,
+      "learning_rate": 8.65625e-06,
+      "loss": 0.5451,
+      "step": 12700
+    },
+    {
+      "epoch": 184.2,
+      "learning_rate": 8.578125e-06,
+      "loss": 0.5132,
+      "step": 12710
+    },
+    {
+      "epoch": 184.35,
+      "learning_rate": 8.500000000000002e-06,
+      "loss": 0.4806,
+      "step": 12720
+    },
+    {
+      "epoch": 184.49,
+      "learning_rate": 8.421875e-06,
+      "loss": 0.5207,
+      "step": 12730
+    },
+    {
+      "epoch": 184.64,
+      "learning_rate": 8.34375e-06,
+      "loss": 0.5618,
+      "step": 12740
+    },
+    {
+      "epoch": 184.78,
+      "learning_rate": 8.265625000000001e-06,
+      "loss": 0.5574,
+      "step": 12750
+    },
+    {
+      "epoch": 184.93,
+      "learning_rate": 8.1875e-06,
+      "loss": 0.5181,
+      "step": 12760
+    },
+    {
+      "epoch": 185.07,
+      "learning_rate": 8.109375e-06,
+      "loss": 0.5471,
+      "step": 12770
+    },
+    {
+      "epoch": 185.22,
+      "learning_rate": 8.03125e-06,
+      "loss": 0.4863,
+      "step": 12780
+    },
+    {
+      "epoch": 185.36,
+      "learning_rate": 7.953125e-06,
+      "loss": 0.547,
+      "step": 12790
+    },
+    {
+      "epoch": 185.51,
+      "learning_rate": 7.875e-06,
+      "loss": 0.5103,
+      "step": 12800
+    },
+    {
+      "epoch": 185.65,
+      "learning_rate": 7.796875000000001e-06,
+      "loss": 0.5406,
+      "step": 12810
+    },
+    {
+      "epoch": 185.8,
+      "learning_rate": 7.718750000000001e-06,
+      "loss": 0.5136,
+      "step": 12820
+    },
+    {
+      "epoch": 185.94,
+      "learning_rate": 7.640625e-06,
+      "loss": 0.4991,
+      "step": 12830
+    },
+    {
+      "epoch": 186.09,
+      "learning_rate": 7.5625e-06,
+      "loss": 0.5515,
+      "step": 12840
+    },
+    {
+      "epoch": 186.23,
+      "learning_rate": 7.484375000000001e-06,
+      "loss": 0.4893,
+      "step": 12850
+    },
+    {
+      "epoch": 186.38,
+      "learning_rate": 7.4062500000000005e-06,
+      "loss": 0.6101,
+      "step": 12860
+    },
+    {
+      "epoch": 186.52,
+      "learning_rate": 7.328125000000001e-06,
+      "loss": 0.5348,
+      "step": 12870
+    },
+    {
+      "epoch": 186.67,
+      "learning_rate": 7.25e-06,
+      "loss": 0.554,
+      "step": 12880
+    },
+    {
+      "epoch": 186.81,
+      "learning_rate": 7.171875e-06,
+      "loss": 0.5119,
+      "step": 12890
+    },
+    {
+      "epoch": 186.96,
+      "learning_rate": 7.0937500000000005e-06,
+      "loss": 0.5348,
+      "step": 12900
+    },
+    {
+      "epoch": 187.1,
+      "learning_rate": 7.015625e-06,
+      "loss": 0.5884,
+      "step": 12910
+    },
+    {
+      "epoch": 187.25,
+      "learning_rate": 6.937500000000001e-06,
+      "loss": 0.496,
+      "step": 12920
+    },
+    {
+      "epoch": 187.39,
+      "learning_rate": 6.859375e-06,
+      "loss": 0.5119,
+      "step": 12930
+    },
+    {
+      "epoch": 187.54,
+      "learning_rate": 6.7812500000000005e-06,
+      "loss": 0.5284,
+      "step": 12940
+    },
+    {
+      "epoch": 187.68,
+      "learning_rate": 6.703125e-06,
+      "loss": 0.5597,
+      "step": 12950
+    },
+    {
+      "epoch": 187.83,
+      "learning_rate": 6.625000000000001e-06,
+      "loss": 0.5816,
+      "step": 12960
+    },
+    {
+      "epoch": 187.97,
+      "learning_rate": 6.546875000000001e-06,
+      "loss": 0.5311,
+      "step": 12970
+    },
+    {
+      "epoch": 188.12,
+      "learning_rate": 6.46875e-06,
+      "loss": 0.5023,
+      "step": 12980
+    },
+    {
+      "epoch": 188.26,
+      "learning_rate": 6.390625e-06,
+      "loss": 0.5199,
+      "step": 12990
+    },
+    {
+      "epoch": 188.41,
+      "learning_rate": 6.3125e-06,
+      "loss": 0.5379,
+      "step": 13000
+    },
+    {
+      "epoch": 188.41,
+      "eval_loss": 0.2375340759754181,
+      "eval_runtime": 574.5803,
+      "eval_samples_per_second": 5.907,
+      "eval_steps_per_second": 0.74,
+      "eval_wer": 0.15229807131293707,
+      "step": 13000
+    },
+    {
+      "epoch": 188.55,
+      "learning_rate": 6.234375000000001e-06,
+      "loss": 0.5227,
+      "step": 13010
+    },
+    {
+      "epoch": 188.7,
+      "learning_rate": 6.1562500000000006e-06,
+      "loss": 0.4946,
+      "step": 13020
+    },
+    {
+      "epoch": 188.84,
+      "learning_rate": 6.078125e-06,
+      "loss": 0.5618,
+      "step": 13030
+    },
+    {
+      "epoch": 188.99,
+      "learning_rate": 6e-06,
+      "loss": 0.5023,
+      "step": 13040
+    },
+    {
+      "epoch": 189.13,
+      "learning_rate": 5.921875e-06,
+      "loss": 0.4884,
+      "step": 13050
+    },
+    {
+      "epoch": 189.28,
+      "learning_rate": 5.843750000000001e-06,
+      "loss": 0.4912,
+      "step": 13060
+    },
+    {
+      "epoch": 189.42,
+      "learning_rate": 5.765625e-06,
+      "loss": 0.5511,
+      "step": 13070
+    },
+    {
+      "epoch": 189.57,
+      "learning_rate": 5.6875e-06,
+      "loss": 0.542,
+      "step": 13080
+    },
+    {
+      "epoch": 189.71,
+      "learning_rate": 5.609375e-06,
+      "loss": 0.5275,
+      "step": 13090
+    },
+    {
+      "epoch": 189.86,
+      "learning_rate": 5.531250000000001e-06,
+      "loss": 0.5719,
+      "step": 13100
+    },
+    {
+      "epoch": 190.0,
+      "learning_rate": 5.453125e-06,
+      "loss": 0.5686,
+      "step": 13110
+    },
+    {
+      "epoch": 190.14,
+      "learning_rate": 5.375e-06,
+      "loss": 0.5425,
+      "step": 13120
+    },
+    {
+      "epoch": 190.29,
+      "learning_rate": 5.296875000000001e-06,
+      "loss": 0.5322,
+      "step": 13130
+    },
+    {
+      "epoch": 190.43,
+      "learning_rate": 5.21875e-06,
+      "loss": 0.5146,
+      "step": 13140
+    },
+    {
+      "epoch": 190.58,
+      "learning_rate": 5.1406250000000004e-06,
+      "loss": 0.5364,
+      "step": 13150
+    },
+    {
+      "epoch": 190.72,
+      "learning_rate": 5.0625e-06,
+      "loss": 0.496,
+      "step": 13160
+    },
+    {
+      "epoch": 190.87,
+      "learning_rate": 4.984375e-06,
+      "loss": 0.6167,
+      "step": 13170
+    },
+    {
+      "epoch": 191.01,
+      "learning_rate": 4.906250000000001e-06,
+      "loss": 0.4898,
+      "step": 13180
+    },
+    {
+      "epoch": 191.16,
+      "learning_rate": 4.828125e-06,
+      "loss": 0.5216,
+      "step": 13190
+    },
+    {
+      "epoch": 191.3,
+      "learning_rate": 4.75e-06,
+      "loss": 0.5122,
+      "step": 13200
+    },
+    {
+      "epoch": 191.45,
+      "learning_rate": 4.671875000000001e-06,
+      "loss": 0.5293,
+      "step": 13210
+    },
+    {
+      "epoch": 191.59,
+      "learning_rate": 4.59375e-06,
+      "loss": 0.5072,
+      "step": 13220
+    },
+    {
+      "epoch": 191.74,
+      "learning_rate": 4.5156250000000005e-06,
+      "loss": 0.4934,
+      "step": 13230
+    },
+    {
+      "epoch": 191.88,
+      "learning_rate": 4.4375e-06,
+      "loss": 0.5224,
+      "step": 13240
+    },
+    {
+      "epoch": 192.03,
+      "learning_rate": 4.359375e-06,
+      "loss": 0.4804,
+      "step": 13250
+    },
+    {
+      "epoch": 192.17,
+      "learning_rate": 4.281250000000001e-06,
+      "loss": 0.6039,
+      "step": 13260
+    },
+    {
+      "epoch": 192.32,
+      "learning_rate": 4.2031250000000005e-06,
+      "loss": 0.4833,
+      "step": 13270
+    },
+    {
+      "epoch": 192.46,
+      "learning_rate": 4.125e-06,
+      "loss": 0.5236,
+      "step": 13280
+    },
+    {
+      "epoch": 192.61,
+      "learning_rate": 4.046875e-06,
+      "loss": 0.5312,
+      "step": 13290
+    },
+    {
+      "epoch": 192.75,
+      "learning_rate": 3.96875e-06,
+      "loss": 0.5886,
+      "step": 13300
+    },
+    {
+      "epoch": 192.9,
+      "learning_rate": 3.8906250000000005e-06,
+      "loss": 0.5098,
+      "step": 13310
+    },
+    {
+      "epoch": 193.04,
+      "learning_rate": 3.8125e-06,
+      "loss": 0.5223,
+      "step": 13320
+    },
+    {
+      "epoch": 193.19,
+      "learning_rate": 3.7343750000000006e-06,
+      "loss": 0.581,
+      "step": 13330
+    },
+    {
+      "epoch": 193.33,
+      "learning_rate": 3.65625e-06,
+      "loss": 0.4837,
+      "step": 13340
+    },
+    {
+      "epoch": 193.48,
+      "learning_rate": 3.578125e-06,
+      "loss": 0.5184,
+      "step": 13350
+    },
+    {
+      "epoch": 193.62,
+      "learning_rate": 3.5000000000000004e-06,
+      "loss": 0.5145,
+      "step": 13360
+    },
+    {
+      "epoch": 193.77,
+      "learning_rate": 3.421875e-06,
+      "loss": 0.4956,
+      "step": 13370
+    },
+    {
+      "epoch": 193.91,
+      "learning_rate": 3.3437500000000004e-06,
+      "loss": 0.5211,
+      "step": 13380
+    },
+    {
+      "epoch": 194.06,
+      "learning_rate": 3.2656249999999998e-06,
+      "loss": 0.5218,
+      "step": 13390
+    },
+    {
+      "epoch": 194.2,
+      "learning_rate": 3.1875000000000004e-06,
+      "loss": 0.5457,
+      "step": 13400
+    },
+    {
+      "epoch": 194.35,
+      "learning_rate": 3.109375e-06,
+      "loss": 0.4737,
+      "step": 13410
+    },
+    {
+      "epoch": 194.49,
+      "learning_rate": 3.03125e-06,
+      "loss": 0.5828,
+      "step": 13420
+    },
+    {
+      "epoch": 194.64,
+      "learning_rate": 2.9531249999999998e-06,
+      "loss": 0.5597,
+      "step": 13430
+    },
+    {
+      "epoch": 194.78,
+      "learning_rate": 2.8750000000000004e-06,
+      "loss": 0.5,
+      "step": 13440
+    },
+    {
+      "epoch": 194.93,
+      "learning_rate": 2.7968750000000002e-06,
+      "loss": 0.5694,
+      "step": 13450
+    },
+    {
+      "epoch": 195.07,
+      "learning_rate": 2.71875e-06,
+      "loss": 0.5176,
+      "step": 13460
+    },
+    {
+      "epoch": 195.22,
+      "learning_rate": 2.6406250000000002e-06,
+      "loss": 0.5216,
+      "step": 13470
+    },
+    {
+      "epoch": 195.36,
+      "learning_rate": 2.5625e-06,
+      "loss": 0.489,
+      "step": 13480
+    },
+    {
+      "epoch": 195.51,
+      "learning_rate": 2.4843750000000002e-06,
+      "loss": 0.5186,
+      "step": 13490
+    },
+    {
+      "epoch": 195.65,
+      "learning_rate": 2.40625e-06,
+      "loss": 0.5022,
+      "step": 13500
+    },
+    {
+      "epoch": 195.65,
+      "eval_loss": 0.23950409889221191,
+      "eval_runtime": 592.332,
+      "eval_samples_per_second": 5.73,
+      "eval_steps_per_second": 0.718,
+      "eval_wer": 0.15194471151870753,
+      "step": 13500
+    },
+    {
+      "epoch": 195.8,
+      "learning_rate": 2.3281250000000003e-06,
+      "loss": 0.5106,
+      "step": 13510
+    },
+    {
+      "epoch": 195.94,
+      "learning_rate": 2.25e-06,
+      "loss": 0.4966,
+      "step": 13520
+    },
+    {
+      "epoch": 196.09,
+      "learning_rate": 2.171875e-06,
+      "loss": 0.5053,
+      "step": 13530
+    },
+    {
+      "epoch": 196.23,
+      "learning_rate": 2.09375e-06,
+      "loss": 0.5594,
+      "step": 13540
+    },
+    {
+      "epoch": 196.38,
+      "learning_rate": 2.0156250000000003e-06,
+      "loss": 0.5047,
+      "step": 13550
+    },
+    {
+      "epoch": 196.52,
+      "learning_rate": 1.9375e-06,
+      "loss": 0.5078,
+      "step": 13560
+    },
+    {
+      "epoch": 196.67,
+      "learning_rate": 1.859375e-06,
+      "loss": 0.5466,
+      "step": 13570
+    },
+    {
+      "epoch": 196.81,
+      "learning_rate": 1.7812499999999999e-06,
+      "loss": 0.5371,
+      "step": 13580
+    },
+    {
+      "epoch": 196.96,
+      "learning_rate": 1.703125e-06,
+      "loss": 0.5551,
+      "step": 13590
+    },
+    {
+      "epoch": 197.1,
+      "learning_rate": 1.6250000000000001e-06,
+      "loss": 0.5615,
+      "step": 13600
+    },
+    {
+      "epoch": 197.25,
+      "learning_rate": 1.5468750000000001e-06,
+      "loss": 0.5464,
+      "step": 13610
+    },
+    {
+      "epoch": 197.39,
+      "learning_rate": 1.46875e-06,
+      "loss": 0.5196,
+      "step": 13620
+    },
+    {
+      "epoch": 197.54,
+      "learning_rate": 1.3906250000000001e-06,
+      "loss": 0.4493,
+      "step": 13630
+    },
+    {
+      "epoch": 197.68,
+      "learning_rate": 1.3125e-06,
+      "loss": 0.5603,
+      "step": 13640
+    },
+    {
+      "epoch": 197.83,
+      "learning_rate": 1.2343750000000001e-06,
+      "loss": 0.5227,
+      "step": 13650
+    },
+    {
+      "epoch": 197.97,
+      "learning_rate": 1.15625e-06,
+      "loss": 0.4889,
+      "step": 13660
+    },
+    {
+      "epoch": 198.12,
+      "learning_rate": 1.078125e-06,
+      "loss": 0.5893,
+      "step": 13670
+    },
+    {
+      "epoch": 198.26,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.5509,
+      "step": 13680
+    },
+    {
+      "epoch": 198.41,
+      "learning_rate": 9.21875e-07,
+      "loss": 0.513,
+      "step": 13690
+    },
+    {
+      "epoch": 198.55,
+      "learning_rate": 8.437500000000001e-07,
+      "loss": 0.5744,
+      "step": 13700
+    },
+    {
+      "epoch": 198.7,
+      "learning_rate": 7.65625e-07,
+      "loss": 0.5457,
+      "step": 13710
+    },
+    {
+      "epoch": 198.84,
+      "learning_rate": 6.875000000000001e-07,
+      "loss": 0.5007,
+      "step": 13720
+    },
+    {
+      "epoch": 198.99,
+      "learning_rate": 6.093750000000001e-07,
+      "loss": 0.5829,
+      "step": 13730
+    },
+    {
+      "epoch": 199.13,
+      "learning_rate": 5.312500000000001e-07,
+      "loss": 0.5047,
+      "step": 13740
+    },
+    {
+      "epoch": 199.28,
+      "learning_rate": 4.53125e-07,
+      "loss": 0.5173,
+      "step": 13750
+    },
+    {
+      "epoch": 199.42,
+      "learning_rate": 3.75e-07,
+      "loss": 0.533,
+      "step": 13760
+    },
+    {
+      "epoch": 199.57,
+      "learning_rate": 2.96875e-07,
+      "loss": 0.5328,
+      "step": 13770
+    },
+    {
+      "epoch": 199.71,
+      "learning_rate": 2.1875000000000002e-07,
+      "loss": 0.5111,
+      "step": 13780
+    },
+    {
+      "epoch": 199.86,
+      "learning_rate": 1.40625e-07,
+      "loss": 0.5071,
+      "step": 13790
+    },
+    {
+      "epoch": 200.0,
+      "learning_rate": 6.250000000000001e-08,
+      "loss": 0.5151,
+      "step": 13800
+    },
+    {
+      "epoch": 200.0,
+      "step": 13800,
+      "total_flos": 2.0051176394243018e+20,
+      "train_loss": 0.32463424516760786,
+      "train_runtime": 60486.5281,
+      "train_samples_per_second": 7.255,
+      "train_steps_per_second": 0.228
     }
   ],
-  "max_steps": 6900,
-  "num_train_epochs": 100,
-  "total_flos": 1.0025325448199992e+20,
+  "max_steps": 13800,
+  "num_train_epochs": 200,
+  "total_flos": 2.0051176394243018e+20,
   "trial_name": null,
   "trial_params": null
 }