diff --git a/model-bin/finetune/base/checkpoint-3111/config.json b/model-bin/finetune/base/checkpoint-3111/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..db8911bce1701d03e9c1d6ddaff48d2b5b4e4bb3
--- /dev/null
+++ b/model-bin/finetune/base/checkpoint-3111/config.json
@@ -0,0 +1,78 @@
+{
+  "_name_or_path": "./model-bin/pretrained/base",
+  "activation_dropout": 0.1,
+  "apply_spec_augment": true,
+  "architectures": [
+    "Wav2Vec2ForCTC"
+  ],
+  "attention_dropout": 0.1,
+  "bos_token_id": 1,
+  "codevector_dim": 256,
+  "contrastive_logits_temperature": 0.1,
+  "conv_bias": false,
+  "conv_dim": [
+    512,
+    512,
+    512,
+    512,
+    512,
+    512,
+    512
+  ],
+  "conv_kernel": [
+    10,
+    3,
+    3,
+    3,
+    3,
+    2,
+    2
+  ],
+  "conv_stride": [
+    5,
+    2,
+    2,
+    2,
+    2,
+    2,
+    2
+  ],
+  "ctc_loss_reduction": "mean",
+  "ctc_zero_infinity": false,
+  "diversity_loss_weight": 0.1,
+  "do_stable_layer_norm": false,
+  "eos_token_id": 2,
+  "feat_extract_activation": "gelu",
+  "feat_extract_dropout": 0.0,
+  "feat_extract_norm": "group",
+  "feat_proj_dropout": 0.1,
+  "feat_quantizer_dropout": 0.0,
+  "final_dropout": 0.1,
+  "gradient_checkpointing": true,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.1,
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.1,
+  "mask_feature_length": 10,
+  "mask_feature_prob": 0.0,
+  "mask_time_length": 10,
+  "mask_time_prob": 0.05,
+  "model_type": "wav2vec2",
+  "num_attention_heads": 12,
+  "num_codevector_groups": 2,
+  "num_codevectors_per_group": 320,
+  "num_conv_pos_embedding_groups": 16,
+  "num_conv_pos_embeddings": 128,
+  "num_feat_extract_layers": 7,
+  "num_hidden_layers": 12,
+  "num_negatives": 100,
+  "pad_token_id": 109,
+  "proj_codevector_dim": 256,
+  "torch_dtype": "float32",
+  "transformers_version": "4.9.2",
+  "vocab_size": 110
+}
diff --git a/model-bin/finetune/base/checkpoint-3111/optimizer.pt b/model-bin/finetune/base/checkpoint-3111/optimizer.pt
new file mode 100644
index 0000000000000000000000000000000000000000..3873fe274a41638c16bc0b8e23454291770f0dd2
--- /dev/null
+++ b/model-bin/finetune/base/checkpoint-3111/optimizer.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bc8ccdd293c90b9f04c144f3894211ff701487fcee9b3bbe40d3e691cfb01dc
+size 722165009
diff --git a/model-bin/finetune/base/checkpoint-3111/preprocessor_config.json b/model-bin/finetune/base/checkpoint-3111/preprocessor_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0b7227fc1d916e469b14f6c154ad6dfea1e6891
--- /dev/null
+++ b/model-bin/finetune/base/checkpoint-3111/preprocessor_config.json
@@ -0,0 +1,9 @@
+{
+  "do_normalize": true,
+  "feature_extractor_type": "Wav2Vec2FeatureExtractor",
+  "feature_size": 1,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "return_attention_mask": false,
+  "sampling_rate": 16000
+}
diff --git a/model-bin/finetune/base/checkpoint-3111/pytorch_model.bin b/model-bin/finetune/base/checkpoint-3111/pytorch_model.bin
new file mode 100644
index 0000000000000000000000000000000000000000..f0ea0afc460269e40131bb4b6d7d6c55b9cb3e14
--- /dev/null
+++ b/model-bin/finetune/base/checkpoint-3111/pytorch_model.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff0e57112fe21250625da4085b0f85faed64e109a4e1f1c4a6e97ffa3b86315c
+size 377909911
diff --git a/model-bin/finetune/base/checkpoint-3111/rng_state.pth b/model-bin/finetune/base/checkpoint-3111/rng_state.pth
new file mode 100644
index 0000000000000000000000000000000000000000..4954b1f723c295deed87c003c980d9e23f55a0bd
--- /dev/null
+++ b/model-bin/finetune/base/checkpoint-3111/rng_state.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2971a50aa8ef06effb33c5ff0b927d39811fdf9b458c3ee980ee447619d9e4f2
+size 14567
diff --git a/model-bin/finetune/base/checkpoint-3111/scaler.pt b/model-bin/finetune/base/checkpoint-3111/scaler.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4841fb475b9baac69b477bd0c62689bdef4ef98b
--- /dev/null
+++ b/model-bin/finetune/base/checkpoint-3111/scaler.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60e4ff8f28c2f89e9490f4fb5046568afc1cf91eb19eb2dc17694ebc7aa075ff
+size 559
diff --git a/model-bin/finetune/base/checkpoint-3111/scheduler.pt b/model-bin/finetune/base/checkpoint-3111/scheduler.pt
new file mode 100644
index 0000000000000000000000000000000000000000..2da4a4cbc885f5b258438f395ea293ab8715b22c
--- /dev/null
+++ b/model-bin/finetune/base/checkpoint-3111/scheduler.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c7506a3a798294cb47ae8fd0f8adcda828f320a4074703b394a74aace940463
+size 623
diff --git a/model-bin/finetune/base/checkpoint-3111/trainer_state.json b/model-bin/finetune/base/checkpoint-3111/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b1fb34efbb68322dcc3e4fe76ce022966959f278
--- /dev/null
+++ b/model-bin/finetune/base/checkpoint-3111/trainer_state.json
@@ -0,0 +1,18907 @@
+{
+  "best_metric": 1.0,
+  "best_model_checkpoint": "./model-bin/finetune/base/checkpoint-124",
+  "epoch": 24.99598393574297,
+  "global_step": 3111,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0000000000000001e-07,
+      "loss": 16.0797,
+      "step": 1
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 18.3622,
+      "step": 2
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 15.1391,
+      "step": 3
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 17.0028,
+      "step": 4
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 16.5122,
+      "step": 5
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 18.0639,
+      "step": 6
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.000000000000001e-07,
+      "loss": 14.4356,
+      "step": 7
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 14.5066,
+      "step": 8
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9e-07,
+      "loss": 13.9063,
+      "step": 9
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 13.5473,
+      "step": 10
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.1e-06,
+      "loss": 13.3916,
+      "step": 11
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 12.7812,
+      "step": 12
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3e-06,
+      "loss": 12.9837,
+      "step": 13
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 12.7311,
+      "step": 14
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5e-06,
+      "loss": 12.6939,
+      "step": 15
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 11.7367,
+      "step": 16
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7000000000000002e-06,
+      "loss": 12.1929,
+      "step": 17
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8e-06,
+      "loss": 11.4885,
+      "step": 18
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9e-06,
+      "loss": 12.3475,
+      "step": 19
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 11.4102,
+      "step": 20
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.1000000000000002e-06,
+      "loss": 11.7961,
+      "step": 21
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.2e-06,
+      "loss": 16.5972,
+      "step": 22
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.3e-06,
+      "loss": 16.2498,
+      "step": 23
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.3e-06,
+      "loss": 22.289,
+      "step": 24
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.3e-06,
+      "loss": 35.8739,
+      "step": 25
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 15.6146,
+      "step": 26
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.5e-06,
+      "loss": 18.2366,
+      "step": 27
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.6e-06,
+      "loss": 16.2926,
+      "step": 28
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.7e-06,
+      "loss": 16.2131,
+      "step": 29
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 15.348,
+      "step": 30
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9e-06,
+      "loss": 15.1917,
+      "step": 31
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3e-06,
+      "loss": 14.7583,
+      "step": 32
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.1e-06,
+      "loss": 14.6317,
+      "step": 33
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 14.1954,
+      "step": 34
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.3e-06,
+      "loss": 13.7497,
+      "step": 35
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 13.431,
+      "step": 36
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.5000000000000004e-06,
+      "loss": 13.402,
+      "step": 37
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.6e-06,
+      "loss": 12.7428,
+      "step": 38
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.7e-06,
+      "loss": 12.3585,
+      "step": 39
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.8e-06,
+      "loss": 12.127,
+      "step": 40
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9e-06,
+      "loss": 11.7666,
+      "step": 41
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 11.7605,
+      "step": 42
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.1000000000000006e-06,
+      "loss": 11.2904,
+      "step": 43
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 11.4718,
+      "step": 44
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.2999999999999995e-06,
+      "loss": 13.2095,
+      "step": 45
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.4e-06,
+      "loss": 12.6655,
+      "step": 46
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.5e-06,
+      "loss": 13.656,
+      "step": 47
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.6e-06,
+      "loss": 17.9423,
+      "step": 48
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.7e-06,
+      "loss": 17.1848,
+      "step": 49
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 32.635,
+      "step": 50
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.9000000000000005e-06,
+      "loss": 17.4454,
+      "step": 51
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5e-06,
+      "loss": 16.7005,
+      "step": 52
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.1e-06,
+      "loss": 16.6912,
+      "step": 53
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.2e-06,
+      "loss": 15.4944,
+      "step": 54
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.3e-06,
+      "loss": 15.7256,
+      "step": 55
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.4e-06,
+      "loss": 15.2067,
+      "step": 56
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 14.4251,
+      "step": 57
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.600000000000001e-06,
+      "loss": 13.6545,
+      "step": 58
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7000000000000005e-06,
+      "loss": 14.4077,
+      "step": 59
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.8e-06,
+      "loss": 14.1156,
+      "step": 60
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.9e-06,
+      "loss": 13.2192,
+      "step": 61
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 6e-06,
+      "loss": 13.0185,
+      "step": 62
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 6.1e-06,
+      "loss": 12.5215,
+      "step": 63
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 6.2e-06,
+      "loss": 11.9673,
+      "step": 64
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 6.300000000000001e-06,
+      "loss": 12.308,
+      "step": 65
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 6.4000000000000006e-06,
+      "loss": 11.7839,
+      "step": 66
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 11.7278,
+      "step": 67
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.6e-06,
+      "loss": 12.0906,
+      "step": 68
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.700000000000001e-06,
+      "loss": 11.1648,
+      "step": 69
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.800000000000001e-06,
+      "loss": 13.5982,
+      "step": 70
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 6.900000000000001e-06,
+      "loss": 15.0072,
+      "step": 71
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.000000000000001e-06,
+      "loss": 14.676,
+      "step": 72
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1e-06,
+      "loss": 17.0978,
+      "step": 73
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.2e-06,
+      "loss": 31.465,
+      "step": 74
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2999999999999996e-06,
+      "loss": 38.9536,
+      "step": 75
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.4e-06,
+      "loss": 16.3893,
+      "step": 76
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.5e-06,
+      "loss": 16.4709,
+      "step": 77
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 7.6e-06,
+      "loss": 16.0627,
+      "step": 78
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 7.7e-06,
+      "loss": 15.3402,
+      "step": 79
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 7.8e-06,
+      "loss": 14.7176,
+      "step": 80
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.9e-06,
+      "loss": 14.8858,
+      "step": 81
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 14.3208,
+      "step": 82
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.1e-06,
+      "loss": 13.6955,
+      "step": 83
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.200000000000001e-06,
+      "loss": 13.553,
+      "step": 84
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.3e-06,
+      "loss": 13.2848,
+      "step": 85
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 12.8201,
+      "step": 86
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.500000000000002e-06,
+      "loss": 12.2694,
+      "step": 87
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.599999999999999e-06,
+      "loss": 12.3244,
+      "step": 88
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.7e-06,
+      "loss": 12.1961,
+      "step": 89
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.8e-06,
+      "loss": 12.5786,
+      "step": 90
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.9e-06,
+      "loss": 11.8884,
+      "step": 91
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9e-06,
+      "loss": 11.0813,
+      "step": 92
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.100000000000001e-06,
+      "loss": 12.5929,
+      "step": 93
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.2e-06,
+      "loss": 11.1881,
+      "step": 94
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.3e-06,
+      "loss": 10.8623,
+      "step": 95
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.4e-06,
+      "loss": 12.0579,
+      "step": 96
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.5e-06,
+      "loss": 12.5371,
+      "step": 97
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 17.5842,
+      "step": 98
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.7e-06,
+      "loss": 17.8802,
+      "step": 99
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.800000000000001e-06,
+      "loss": 35.328,
+      "step": 100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.900000000000002e-06,
+      "loss": 14.9651,
+      "step": 101
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1e-05,
+      "loss": 15.501,
+      "step": 102
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0100000000000002e-05,
+      "loss": 14.2666,
+      "step": 103
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.02e-05,
+      "loss": 13.7764,
+      "step": 104
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.03e-05,
+      "loss": 13.2375,
+      "step": 105
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.04e-05,
+      "loss": 13.1426,
+      "step": 106
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.05e-05,
+      "loss": 13.4613,
+      "step": 107
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.06e-05,
+      "loss": 12.4776,
+      "step": 108
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0700000000000001e-05,
+      "loss": 12.1265,
+      "step": 109
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.08e-05,
+      "loss": 11.3284,
+      "step": 110
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.09e-05,
+      "loss": 10.9315,
+      "step": 111
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 11.1953,
+      "step": 112
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.11e-05,
+      "loss": 10.1608,
+      "step": 113
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1200000000000001e-05,
+      "loss": 10.278,
+      "step": 114
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.13e-05,
+      "loss": 10.6152,
+      "step": 115
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1400000000000001e-05,
+      "loss": 10.1285,
+      "step": 116
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1500000000000002e-05,
+      "loss": 9.3729,
+      "step": 117
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.16e-05,
+      "loss": 10.9062,
+      "step": 118
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1700000000000001e-05,
+      "loss": 9.2814,
+      "step": 119
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.18e-05,
+      "loss": 11.0057,
+      "step": 120
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.19e-05,
+      "loss": 11.4806,
+      "step": 121
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2e-05,
+      "loss": 13.8106,
+      "step": 122
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2100000000000001e-05,
+      "loss": 17.0643,
+      "step": 123
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2100000000000001e-05,
+      "loss": 28.3301,
+      "step": 124
+    },
+    {
+      "epoch": 1.0,
+      "eval_loss": 25.103845596313477,
+      "eval_runtime": 38.5441,
+      "eval_samples_per_second": 21.793,
+      "eval_steps_per_second": 0.7,
+      "eval_wer": 1.0,
+      "step": 124
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.22e-05,
+      "loss": 12.2581,
+      "step": 125
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.23e-05,
+      "loss": 12.3587,
+      "step": 126
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.24e-05,
+      "loss": 13.2348,
+      "step": 127
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.25e-05,
+      "loss": 12.0815,
+      "step": 128
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.2600000000000001e-05,
+      "loss": 12.9259,
+      "step": 129
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.27e-05,
+      "loss": 10.3132,
+      "step": 130
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2800000000000001e-05,
+      "loss": 9.8928,
+      "step": 131
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.29e-05,
+      "loss": 13.5352,
+      "step": 132
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 9.6943,
+      "step": 133
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.3100000000000002e-05,
+      "loss": 11.5528,
+      "step": 134
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.32e-05,
+      "loss": 10.3106,
+      "step": 135
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3300000000000001e-05,
+      "loss": 8.9044,
+      "step": 136
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3400000000000002e-05,
+      "loss": 9.0137,
+      "step": 137
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3500000000000001e-05,
+      "loss": 9.2745,
+      "step": 138
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3600000000000002e-05,
+      "loss": 9.7839,
+      "step": 139
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3700000000000001e-05,
+      "loss": 9.4428,
+      "step": 140
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3800000000000002e-05,
+      "loss": 8.5327,
+      "step": 141
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3900000000000002e-05,
+      "loss": 9.0275,
+      "step": 142
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.4000000000000001e-05,
+      "loss": 7.8265,
+      "step": 143
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.4099999999999999e-05,
+      "loss": 9.2273,
+      "step": 144
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.42e-05,
+      "loss": 9.0895,
+      "step": 145
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.43e-05,
+      "loss": 9.8122,
+      "step": 146
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.44e-05,
+      "loss": 10.8855,
+      "step": 147
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.45e-05,
+      "loss": 14.9699,
+      "step": 148
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.4599999999999999e-05,
+      "loss": 17.7406,
+      "step": 149
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.47e-05,
+      "loss": 10.4401,
+      "step": 150
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.48e-05,
+      "loss": 14.2482,
+      "step": 151
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.49e-05,
+      "loss": 10.177,
+      "step": 152
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.5e-05,
+      "loss": 10.841,
+      "step": 153
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.51e-05,
+      "loss": 9.4118,
+      "step": 154
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.52e-05,
+      "loss": 11.3901,
+      "step": 155
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.53e-05,
+      "loss": 9.1406,
+      "step": 156
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.54e-05,
+      "loss": 9.4707,
+      "step": 157
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.55e-05,
+      "loss": 9.1934,
+      "step": 158
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.56e-05,
+      "loss": 10.0078,
+      "step": 159
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.5700000000000002e-05,
+      "loss": 8.3876,
+      "step": 160
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.58e-05,
+      "loss": 8.3182,
+      "step": 161
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.59e-05,
+      "loss": 8.4264,
+      "step": 162
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 8.2627,
+      "step": 163
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.6100000000000002e-05,
+      "loss": 8.7069,
+      "step": 164
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.62e-05,
+      "loss": 8.4502,
+      "step": 165
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.63e-05,
+      "loss": 7.9165,
+      "step": 166
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.6400000000000002e-05,
+      "loss": 7.7637,
+      "step": 167
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.65e-05,
+      "loss": 8.1296,
+      "step": 168
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.66e-05,
+      "loss": 8.1479,
+      "step": 169
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.6700000000000003e-05,
+      "loss": 8.6646,
+      "step": 170
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.6800000000000002e-05,
+      "loss": 12.0686,
+      "step": 171
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.69e-05,
+      "loss": 10.0614,
+      "step": 172
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 18.4941,
+      "step": 173
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.7100000000000002e-05,
+      "loss": 17.3924,
+      "step": 174
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.7199999999999998e-05,
+      "loss": 9.7856,
+      "step": 175
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.73e-05,
+      "loss": 9.631,
+      "step": 176
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.74e-05,
+      "loss": 9.5508,
+      "step": 177
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.75e-05,
+      "loss": 8.5556,
+      "step": 178
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.76e-05,
+      "loss": 8.7867,
+      "step": 179
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.77e-05,
+      "loss": 8.6881,
+      "step": 180
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.78e-05,
+      "loss": 8.5451,
+      "step": 181
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.79e-05,
+      "loss": 8.4048,
+      "step": 182
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.8e-05,
+      "loss": 8.3765,
+      "step": 183
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.81e-05,
+      "loss": 7.705,
+      "step": 184
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.8200000000000002e-05,
+      "loss": 7.7412,
+      "step": 185
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.83e-05,
+      "loss": 7.5259,
+      "step": 186
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.84e-05,
+      "loss": 7.4572,
+      "step": 187
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.85e-05,
+      "loss": 8.5334,
+      "step": 188
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 1.86e-05,
+      "loss": 7.203,
+      "step": 189
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 1.87e-05,
+      "loss": 8.414,
+      "step": 190
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 1.88e-05,
+      "loss": 7.3297,
+      "step": 191
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.8900000000000002e-05,
+      "loss": 6.9508,
+      "step": 192
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.9e-05,
+      "loss": 6.8011,
+      "step": 193
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.91e-05,
+      "loss": 9.4001,
+      "step": 194
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 1.9200000000000003e-05,
+      "loss": 7.4166,
+      "step": 195
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.93e-05,
+      "loss": 12.1686,
+      "step": 196
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.94e-05,
+      "loss": 8.7785,
+      "step": 197
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.9500000000000003e-05,
+      "loss": 12.4968,
+      "step": 198
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9500000000000003e-05,
+      "loss": 29.9715,
+      "step": 199
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9600000000000002e-05,
+      "loss": 9.6982,
+      "step": 200
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.97e-05,
+      "loss": 8.9671,
+      "step": 201
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.9800000000000004e-05,
+      "loss": 9.0367,
+      "step": 202
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.9900000000000003e-05,
+      "loss": 8.6784,
+      "step": 203
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2e-05,
+      "loss": 9.6911,
+      "step": 204
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.01e-05,
+      "loss": 8.0665,
+      "step": 205
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.0200000000000003e-05,
+      "loss": 8.1636,
+      "step": 206
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.0300000000000002e-05,
+      "loss": 7.9764,
+      "step": 207
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.04e-05,
+      "loss": 8.1983,
+      "step": 208
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.05e-05,
+      "loss": 8.772,
+      "step": 209
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.06e-05,
+      "loss": 7.3145,
+      "step": 210
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.07e-05,
+      "loss": 7.6946,
+      "step": 211
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.08e-05,
+      "loss": 7.1942,
+      "step": 212
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.09e-05,
+      "loss": 7.0128,
+      "step": 213
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.1e-05,
+      "loss": 7.319,
+      "step": 214
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.11e-05,
+      "loss": 6.8166,
+      "step": 215
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.12e-05,
+      "loss": 6.5882,
+      "step": 216
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.13e-05,
+      "loss": 6.5567,
+      "step": 217
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.1400000000000002e-05,
+      "loss": 6.5641,
+      "step": 218
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.15e-05,
+      "loss": 7.2986,
+      "step": 219
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.16e-05,
+      "loss": 7.1401,
+      "step": 220
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.1700000000000002e-05,
+      "loss": 11.3617,
+      "step": 221
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.18e-05,
+      "loss": 7.9772,
+      "step": 222
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.19e-05,
+      "loss": 9.9042,
+      "step": 223
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 13.5206,
+      "step": 224
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.2100000000000002e-05,
+      "loss": 8.5632,
+      "step": 225
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.22e-05,
+      "loss": 8.8155,
+      "step": 226
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.23e-05,
+      "loss": 8.1444,
+      "step": 227
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.2400000000000002e-05,
+      "loss": 8.1459,
+      "step": 228
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.25e-05,
+      "loss": 7.7481,
+      "step": 229
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.26e-05,
+      "loss": 10.1725,
+      "step": 230
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.2700000000000003e-05,
+      "loss": 7.3734,
+      "step": 231
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.2800000000000002e-05,
+      "loss": 7.2943,
+      "step": 232
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.29e-05,
+      "loss": 8.1402,
+      "step": 233
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.3000000000000003e-05,
+      "loss": 7.026,
+      "step": 234
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.3100000000000002e-05,
+      "loss": 6.8895,
+      "step": 235
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.32e-05,
+      "loss": 6.9352,
+      "step": 236
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.3300000000000004e-05,
+      "loss": 7.5332,
+      "step": 237
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.3400000000000003e-05,
+      "loss": 6.6096,
+      "step": 238
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.35e-05,
+      "loss": 6.3894,
+      "step": 239
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.36e-05,
+      "loss": 6.3214,
+      "step": 240
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.37e-05,
+      "loss": 7.0104,
+      "step": 241
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.38e-05,
+      "loss": 6.1961,
+      "step": 242
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.39e-05,
+      "loss": 8.2539,
+      "step": 243
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4e-05,
+      "loss": 6.5452,
+      "step": 244
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.41e-05,
+      "loss": 7.3944,
+      "step": 245
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4200000000000002e-05,
+      "loss": 6.8491,
+      "step": 246
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.43e-05,
+      "loss": 8.3832,
+      "step": 247
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.44e-05,
+      "loss": 15.2116,
+      "step": 248
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": 14.279631614685059,
+      "eval_runtime": 39.8591,
+      "eval_samples_per_second": 21.074,
+      "eval_steps_per_second": 0.677,
+      "eval_wer": 1.0,
+      "step": 248
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.45e-05,
+      "loss": 8.1804,
+      "step": 249
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.46e-05,
+      "loss": 8.2228,
+      "step": 250
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.47e-05,
+      "loss": 7.7418,
+      "step": 251
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.48e-05,
+      "loss": 7.5864,
+      "step": 252
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.4900000000000002e-05,
+      "loss": 7.4361,
+      "step": 253
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.5e-05,
+      "loss": 8.8981,
+      "step": 254
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.51e-05,
+      "loss": 8.002,
+      "step": 255
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.5200000000000003e-05,
+      "loss": 7.0949,
+      "step": 256
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.5300000000000002e-05,
+      "loss": 6.7862,
+      "step": 257
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.54e-05,
+      "loss": 7.8525,
+      "step": 258
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.5500000000000003e-05,
+      "loss": 6.5721,
+      "step": 259
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.5600000000000002e-05,
+      "loss": 6.5643,
+      "step": 260
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.57e-05,
+      "loss": 6.3822,
+      "step": 261
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.58e-05,
+      "loss": 6.3616,
+      "step": 262
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.5900000000000003e-05,
+      "loss": 6.5943,
+      "step": 263
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 6.8219,
+      "step": 264
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.61e-05,
+      "loss": 5.8965,
+      "step": 265
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6200000000000003e-05,
+      "loss": 5.9806,
+      "step": 266
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6300000000000002e-05,
+      "loss": 6.1043,
+      "step": 267
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.64e-05,
+      "loss": 6.9121,
+      "step": 268
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6500000000000004e-05,
+      "loss": 6.7837,
+      "step": 269
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6600000000000003e-05,
+      "loss": 6.3355,
+      "step": 270
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6700000000000002e-05,
+      "loss": 9.7755,
+      "step": 271
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6800000000000004e-05,
+      "loss": 8.7245,
+      "step": 272
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6900000000000003e-05,
+      "loss": 13.1759,
+      "step": 273
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.7000000000000002e-05,
+      "loss": 7.7578,
+      "step": 274
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.7100000000000005e-05,
+      "loss": 7.8589,
+      "step": 275
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.7200000000000004e-05,
+      "loss": 7.558,
+      "step": 276
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.7300000000000003e-05,
+      "loss": 7.0113,
+      "step": 277
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.7400000000000002e-05,
+      "loss": 6.9333,
+      "step": 278
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.7500000000000004e-05,
+      "loss": 6.9015,
+      "step": 279
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.7600000000000003e-05,
+      "loss": 8.1645,
+      "step": 280
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.7700000000000002e-05,
+      "loss": 6.4343,
+      "step": 281
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.7800000000000005e-05,
+      "loss": 6.6459,
+      "step": 282
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.7900000000000004e-05,
+      "loss": 6.2265,
+      "step": 283
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 6.1436,
+      "step": 284
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.8100000000000005e-05,
+      "loss": 6.21,
+      "step": 285
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.8199999999999998e-05,
+      "loss": 6.0527,
+      "step": 286
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.83e-05,
+      "loss": 5.9419,
+      "step": 287
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.84e-05,
+      "loss": 5.9473,
+      "step": 288
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.8499999999999998e-05,
+      "loss": 5.7468,
+      "step": 289
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.86e-05,
+      "loss": 5.7283,
+      "step": 290
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.87e-05,
+      "loss": 6.0418,
+      "step": 291
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.88e-05,
+      "loss": 7.4134,
+      "step": 292
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.8899999999999998e-05,
+      "loss": 5.6938,
+      "step": 293
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.9e-05,
+      "loss": 8.3107,
+      "step": 294
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.91e-05,
+      "loss": 6.6061,
+      "step": 295
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.9199999999999998e-05,
+      "loss": 7.6899,
+      "step": 296
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.93e-05,
+      "loss": 8.0845,
+      "step": 297
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.94e-05,
+      "loss": 15.5426,
+      "step": 298
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.95e-05,
+      "loss": 6.8855,
+      "step": 299
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.96e-05,
+      "loss": 7.9526,
+      "step": 300
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.97e-05,
+      "loss": 6.6894,
+      "step": 301
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.98e-05,
+      "loss": 6.7512,
+      "step": 302
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.9900000000000002e-05,
+      "loss": 6.567,
+      "step": 303
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3e-05,
+      "loss": 7.2903,
+      "step": 304
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.01e-05,
+      "loss": 6.259,
+      "step": 305
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.02e-05,
+      "loss": 6.2604,
+      "step": 306
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.03e-05,
+      "loss": 5.9724,
+      "step": 307
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.04e-05,
+      "loss": 5.8859,
+      "step": 308
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.05e-05,
+      "loss": 5.8829,
+      "step": 309
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.06e-05,
+      "loss": 6.4364,
+      "step": 310
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.07e-05,
+      "loss": 6.045,
+      "step": 311
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.08e-05,
+      "loss": 5.6413,
+      "step": 312
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 3.09e-05,
+      "loss": 6.1185,
+      "step": 313
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 3.1e-05,
+      "loss": 5.4294,
+      "step": 314
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 3.1100000000000004e-05,
+      "loss": 5.2197,
+      "step": 315
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 3.12e-05,
+      "loss": 5.3347,
+      "step": 316
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 3.13e-05,
+      "loss": 5.8226,
+      "step": 317
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 3.1400000000000004e-05,
+      "loss": 5.4393,
+      "step": 318
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 3.15e-05,
+      "loss": 5.4959,
+      "step": 319
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 3.16e-05,
+      "loss": 6.1362,
+      "step": 320
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 3.1700000000000005e-05,
+      "loss": 6.3625,
+      "step": 321
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 3.18e-05,
+      "loss": 9.4302,
+      "step": 322
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 3.19e-05,
+      "loss": 11.9141,
+      "step": 323
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 6.549,
+      "step": 324
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.21e-05,
+      "loss": 6.9,
+      "step": 325
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.2200000000000003e-05,
+      "loss": 6.3619,
+      "step": 326
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.2300000000000006e-05,
+      "loss": 6.4263,
+      "step": 327
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.24e-05,
+      "loss": 6.2014,
+      "step": 328
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.2500000000000004e-05,
+      "loss": 5.8565,
+      "step": 329
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.26e-05,
+      "loss": 5.931,
+      "step": 330
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.27e-05,
+      "loss": 5.7223,
+      "step": 331
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.2800000000000004e-05,
+      "loss": 5.5335,
+      "step": 332
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.29e-05,
+      "loss": 5.6481,
+      "step": 333
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.3e-05,
+      "loss": 5.7051,
+      "step": 334
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.3100000000000005e-05,
+      "loss": 5.3221,
+      "step": 335
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.32e-05,
+      "loss": 5.2614,
+      "step": 336
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.33e-05,
+      "loss": 5.4341,
+      "step": 337
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 3.3400000000000005e-05,
+      "loss": 5.117,
+      "step": 338
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 3.35e-05,
+      "loss": 5.0734,
+      "step": 339
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.3600000000000004e-05,
+      "loss": 5.2825,
+      "step": 340
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.3700000000000006e-05,
+      "loss": 5.0451,
+      "step": 341
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.38e-05,
+      "loss": 5.0239,
+      "step": 342
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.3900000000000004e-05,
+      "loss": 5.1983,
+      "step": 343
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 5.7491,
+      "step": 344
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.41e-05,
+      "loss": 5.7687,
+      "step": 345
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.4200000000000005e-05,
+      "loss": 6.5379,
+      "step": 346
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.430000000000001e-05,
+      "loss": 6.825,
+      "step": 347
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.4399999999999996e-05,
+      "loss": 13.0222,
+      "step": 348
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 3.45e-05,
+      "loss": 6.3984,
+      "step": 349
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 3.46e-05,
+      "loss": 5.9707,
+      "step": 350
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 3.4699999999999996e-05,
+      "loss": 5.8836,
+      "step": 351
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.48e-05,
+      "loss": 5.8106,
+      "step": 352
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.49e-05,
+      "loss": 5.5632,
+      "step": 353
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.5e-05,
+      "loss": 5.4655,
+      "step": 354
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 3.51e-05,
+      "loss": 5.4477,
+      "step": 355
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 3.52e-05,
+      "loss": 5.5167,
+      "step": 356
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 3.53e-05,
+      "loss": 5.2506,
+      "step": 357
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 3.54e-05,
+      "loss": 5.3548,
+      "step": 358
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 3.55e-05,
+      "loss": 5.0665,
+      "step": 359
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 3.56e-05,
+      "loss": 5.1102,
+      "step": 360
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 3.57e-05,
+      "loss": 4.9738,
+      "step": 361
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 3.58e-05,
+      "loss": 4.9472,
+      "step": 362
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.59e-05,
+      "loss": 4.8614,
+      "step": 363
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.6e-05,
+      "loss": 4.8071,
+      "step": 364
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.61e-05,
+      "loss": 4.7996,
+      "step": 365
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.62e-05,
+      "loss": 4.7922,
+      "step": 366
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.63e-05,
+      "loss": 5.0222,
+      "step": 367
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 3.6400000000000004e-05,
+      "loss": 5.3838,
+      "step": 368
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 3.65e-05,
+      "loss": 6.0561,
+      "step": 369
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.66e-05,
+      "loss": 4.9871,
+      "step": 370
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.6700000000000004e-05,
+      "loss": 5.2072,
+      "step": 371
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.68e-05,
+      "loss": 5.9062,
+      "step": 372
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.69e-05,
+      "loss": 10.7258,
+      "step": 373
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": 10.808684349060059,
+      "eval_runtime": 39.3917,
+      "eval_samples_per_second": 21.324,
+      "eval_steps_per_second": 0.685,
+      "eval_wer": 1.0,
+      "step": 373
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 3.7e-05,
+      "loss": 5.5652,
+      "step": 374
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 3.71e-05,
+      "loss": 5.3923,
+      "step": 375
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 3.72e-05,
+      "loss": 5.4629,
+      "step": 376
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 3.73e-05,
+      "loss": 5.2987,
+      "step": 377
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 3.74e-05,
+      "loss": 5.1875,
+      "step": 378
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 3.7500000000000003e-05,
+      "loss": 5.1286,
+      "step": 379
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 3.76e-05,
+      "loss": 5.0511,
+      "step": 380
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 3.77e-05,
+      "loss": 4.9293,
+      "step": 381
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 3.7800000000000004e-05,
+      "loss": 4.8372,
+      "step": 382
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 3.79e-05,
+      "loss": 4.6864,
+      "step": 383
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 3.8e-05,
+      "loss": 4.7565,
+      "step": 384
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 3.8100000000000005e-05,
+      "loss": 4.6983,
+      "step": 385
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 3.82e-05,
+      "loss": 4.5762,
+      "step": 386
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 3.83e-05,
+      "loss": 4.584,
+      "step": 387
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.8400000000000005e-05,
+      "loss": 4.4756,
+      "step": 388
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.85e-05,
+      "loss": 4.48,
+      "step": 389
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.86e-05,
+      "loss": 4.4843,
+      "step": 390
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8700000000000006e-05,
+      "loss": 4.4373,
+      "step": 391
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.88e-05,
+      "loss": 4.4628,
+      "step": 392
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.8900000000000004e-05,
+      "loss": 4.4452,
+      "step": 393
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.9000000000000006e-05,
+      "loss": 4.5501,
+      "step": 394
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.91e-05,
+      "loss": 4.6419,
+      "step": 395
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.9200000000000004e-05,
+      "loss": 5.2376,
+      "step": 396
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.9300000000000007e-05,
+      "loss": 5.5678,
+      "step": 397
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.94e-05,
+      "loss": 9.2255,
+      "step": 398
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.9500000000000005e-05,
+      "loss": 4.9797,
+      "step": 399
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.960000000000001e-05,
+      "loss": 4.9101,
+      "step": 400
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.97e-05,
+      "loss": 4.9319,
+      "step": 401
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.9800000000000005e-05,
+      "loss": 5.089,
+      "step": 402
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.99e-05,
+      "loss": 4.7739,
+      "step": 403
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4e-05,
+      "loss": 4.5654,
+      "step": 404
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.0100000000000006e-05,
+      "loss": 4.6165,
+      "step": 405
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.02e-05,
+      "loss": 4.5005,
+      "step": 406
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.0300000000000004e-05,
+      "loss": 4.4904,
+      "step": 407
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.0400000000000006e-05,
+      "loss": 4.5285,
+      "step": 408
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.05e-05,
+      "loss": 4.3915,
+      "step": 409
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.0600000000000004e-05,
+      "loss": 4.3923,
+      "step": 410
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.07e-05,
+      "loss": 4.2785,
+      "step": 411
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.08e-05,
+      "loss": 4.342,
+      "step": 412
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.09e-05,
+      "loss": 4.2636,
+      "step": 413
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.1e-05,
+      "loss": 4.2375,
+      "step": 414
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.11e-05,
+      "loss": 4.169,
+      "step": 415
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.12e-05,
+      "loss": 4.2448,
+      "step": 416
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.13e-05,
+      "loss": 4.1744,
+      "step": 417
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.14e-05,
+      "loss": 4.3531,
+      "step": 418
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.15e-05,
+      "loss": 4.4133,
+      "step": 419
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.16e-05,
+      "loss": 4.7164,
+      "step": 420
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.17e-05,
+      "loss": 4.6919,
+      "step": 421
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.18e-05,
+      "loss": 5.7254,
+      "step": 422
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.19e-05,
+      "loss": 7.8026,
+      "step": 423
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.2e-05,
+      "loss": 4.7145,
+      "step": 424
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.21e-05,
+      "loss": 4.6025,
+      "step": 425
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.22e-05,
+      "loss": 4.4541,
+      "step": 426
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.23e-05,
+      "loss": 4.5301,
+      "step": 427
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 4.24e-05,
+      "loss": 4.5181,
+      "step": 428
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.25e-05,
+      "loss": 4.3071,
+      "step": 429
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.26e-05,
+      "loss": 4.2593,
+      "step": 430
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.27e-05,
+      "loss": 4.2182,
+      "step": 431
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.2800000000000004e-05,
+      "loss": 4.1614,
+      "step": 432
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 4.29e-05,
+      "loss": 4.166,
+      "step": 433
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.3e-05,
+      "loss": 4.0794,
+      "step": 434
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.3100000000000004e-05,
+      "loss": 4.1143,
+      "step": 435
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.32e-05,
+      "loss": 4.0252,
+      "step": 436
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.33e-05,
+      "loss": 4.0411,
+      "step": 437
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.3400000000000005e-05,
+      "loss": 4.0253,
+      "step": 438
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.35e-05,
+      "loss": 3.9661,
+      "step": 439
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.36e-05,
+      "loss": 3.9899,
+      "step": 440
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.3700000000000005e-05,
+      "loss": 3.9575,
+      "step": 441
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.38e-05,
+      "loss": 3.9065,
+      "step": 442
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.39e-05,
+      "loss": 3.9266,
+      "step": 443
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 3.9945,
+      "step": 444
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.41e-05,
+      "loss": 4.2228,
+      "step": 445
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.4200000000000004e-05,
+      "loss": 4.3721,
+      "step": 446
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.43e-05,
+      "loss": 4.7031,
+      "step": 447
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.44e-05,
+      "loss": 13.3247,
+      "step": 448
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.4500000000000004e-05,
+      "loss": 5.3875,
+      "step": 449
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.46e-05,
+      "loss": 4.2368,
+      "step": 450
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.47e-05,
+      "loss": 4.1829,
+      "step": 451
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.4800000000000005e-05,
+      "loss": 4.1418,
+      "step": 452
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.49e-05,
+      "loss": 5.5886,
+      "step": 453
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.5e-05,
+      "loss": 4.0515,
+      "step": 454
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.5100000000000005e-05,
+      "loss": 3.9857,
+      "step": 455
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.52e-05,
+      "loss": 5.032,
+      "step": 456
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.53e-05,
+      "loss": 3.8938,
+      "step": 457
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.5400000000000006e-05,
+      "loss": 3.8972,
+      "step": 458
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.55e-05,
+      "loss": 3.8636,
+      "step": 459
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.5600000000000004e-05,
+      "loss": 3.8983,
+      "step": 460
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.5700000000000006e-05,
+      "loss": 3.8283,
+      "step": 461
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.58e-05,
+      "loss": 3.8267,
+      "step": 462
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.5900000000000004e-05,
+      "loss": 3.785,
+      "step": 463
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 3.7596,
+      "step": 464
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.61e-05,
+      "loss": 3.7847,
+      "step": 465
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.6200000000000005e-05,
+      "loss": 3.8716,
+      "step": 466
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.630000000000001e-05,
+      "loss": 3.7767,
+      "step": 467
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.64e-05,
+      "loss": 3.8877,
+      "step": 468
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.6500000000000005e-05,
+      "loss": 3.8786,
+      "step": 469
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.660000000000001e-05,
+      "loss": 3.9419,
+      "step": 470
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.6700000000000003e-05,
+      "loss": 3.9412,
+      "step": 471
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.6800000000000006e-05,
+      "loss": 4.7002,
+      "step": 472
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.69e-05,
+      "loss": 6.7147,
+      "step": 473
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.7e-05,
+      "loss": 3.9492,
+      "step": 474
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.71e-05,
+      "loss": 3.9398,
+      "step": 475
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.72e-05,
+      "loss": 4.1639,
+      "step": 476
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.73e-05,
+      "loss": 3.8474,
+      "step": 477
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 4.74e-05,
+      "loss": 3.7996,
+      "step": 478
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.75e-05,
+      "loss": 3.8546,
+      "step": 479
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.76e-05,
+      "loss": 3.7909,
+      "step": 480
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.77e-05,
+      "loss": 4.6526,
+      "step": 481
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.78e-05,
+      "loss": 3.6987,
+      "step": 482
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.79e-05,
+      "loss": 3.6992,
+      "step": 483
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.8e-05,
+      "loss": 3.7142,
+      "step": 484
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.8100000000000004e-05,
+      "loss": 3.6886,
+      "step": 485
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.82e-05,
+      "loss": 3.6595,
+      "step": 486
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.83e-05,
+      "loss": 3.9014,
+      "step": 487
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 4.8400000000000004e-05,
+      "loss": 3.6571,
+      "step": 488
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.85e-05,
+      "loss": 3.6503,
+      "step": 489
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.86e-05,
+      "loss": 3.6311,
+      "step": 490
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.87e-05,
+      "loss": 3.6099,
+      "step": 491
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 4.88e-05,
+      "loss": 3.6616,
+      "step": 492
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.89e-05,
+      "loss": 3.6864,
+      "step": 493
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.9e-05,
+      "loss": 3.7382,
+      "step": 494
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.91e-05,
+      "loss": 3.8174,
+      "step": 495
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.92e-05,
+      "loss": 3.804,
+      "step": 496
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.93e-05,
+      "loss": 3.9716,
+      "step": 497
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.94e-05,
+      "loss": 4.9653,
+      "step": 498
+    },
+    {
+      "epoch": 3.0,
+      "eval_loss": 6.089671611785889,
+      "eval_runtime": 39.0564,
+      "eval_samples_per_second": 21.482,
+      "eval_steps_per_second": 0.691,
+      "eval_wer": 1.0,
+      "step": 498
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.9500000000000004e-05,
+      "loss": 3.738,
+      "step": 499
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.96e-05,
+      "loss": 3.7489,
+      "step": 500
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.97e-05,
+      "loss": 3.6982,
+      "step": 501
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.9800000000000004e-05,
+      "loss": 3.677,
+      "step": 502
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 4.99e-05,
+      "loss": 3.6467,
+      "step": 503
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5e-05,
+      "loss": 3.6049,
+      "step": 504
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.0100000000000005e-05,
+      "loss": 3.598,
+      "step": 505
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.02e-05,
+      "loss": 3.5649,
+      "step": 506
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.03e-05,
+      "loss": 3.6033,
+      "step": 507
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.0400000000000005e-05,
+      "loss": 3.5588,
+      "step": 508
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.05e-05,
+      "loss": 3.5565,
+      "step": 509
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.0600000000000003e-05,
+      "loss": 3.5482,
+      "step": 510
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.0700000000000006e-05,
+      "loss": 3.5433,
+      "step": 511
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.08e-05,
+      "loss": 3.5699,
+      "step": 512
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.0900000000000004e-05,
+      "loss": 3.6706,
+      "step": 513
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.1000000000000006e-05,
+      "loss": 3.527,
+      "step": 514
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.11e-05,
+      "loss": 3.5105,
+      "step": 515
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.1200000000000004e-05,
+      "loss": 3.5149,
+      "step": 516
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.130000000000001e-05,
+      "loss": 3.5255,
+      "step": 517
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.14e-05,
+      "loss": 3.5306,
+      "step": 518
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.1500000000000005e-05,
+      "loss": 3.6149,
+      "step": 519
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.16e-05,
+      "loss": 3.6858,
+      "step": 520
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.17e-05,
+      "loss": 3.9652,
+      "step": 521
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.1800000000000005e-05,
+      "loss": 4.3416,
+      "step": 522
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.19e-05,
+      "loss": 5.4813,
+      "step": 523
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 3.6919,
+      "step": 524
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.2100000000000006e-05,
+      "loss": 3.6024,
+      "step": 525
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.22e-05,
+      "loss": 3.5839,
+      "step": 526
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.2300000000000004e-05,
+      "loss": 3.5264,
+      "step": 527
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.2400000000000007e-05,
+      "loss": 3.5435,
+      "step": 528
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.25e-05,
+      "loss": 3.5134,
+      "step": 529
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.2600000000000005e-05,
+      "loss": 3.4728,
+      "step": 530
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.270000000000001e-05,
+      "loss": 3.4911,
+      "step": 531
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.28e-05,
+      "loss": 3.4774,
+      "step": 532
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.2900000000000005e-05,
+      "loss": 3.4689,
+      "step": 533
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.300000000000001e-05,
+      "loss": 3.4727,
+      "step": 534
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.31e-05,
+      "loss": 3.496,
+      "step": 535
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.3200000000000006e-05,
+      "loss": 3.4633,
+      "step": 536
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.330000000000001e-05,
+      "loss": 3.5826,
+      "step": 537
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.3400000000000004e-05,
+      "loss": 3.4758,
+      "step": 538
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.3500000000000006e-05,
+      "loss": 3.4754,
+      "step": 539
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.360000000000001e-05,
+      "loss": 3.556,
+      "step": 540
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.3700000000000004e-05,
+      "loss": 3.4639,
+      "step": 541
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.380000000000001e-05,
+      "loss": 3.4725,
+      "step": 542
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.390000000000001e-05,
+      "loss": 3.5395,
+      "step": 543
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.4000000000000005e-05,
+      "loss": 3.5235,
+      "step": 544
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.410000000000001e-05,
+      "loss": 3.5668,
+      "step": 545
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.420000000000001e-05,
+      "loss": 3.7506,
+      "step": 546
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.4300000000000005e-05,
+      "loss": 4.2632,
+      "step": 547
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.440000000000001e-05,
+      "loss": 5.2235,
+      "step": 548
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.45e-05,
+      "loss": 3.6157,
+      "step": 549
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.4600000000000006e-05,
+      "loss": 3.5356,
+      "step": 550
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.470000000000001e-05,
+      "loss": 3.5316,
+      "step": 551
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.4800000000000004e-05,
+      "loss": 3.4951,
+      "step": 552
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.4900000000000006e-05,
+      "loss": 3.4724,
+      "step": 553
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.500000000000001e-05,
+      "loss": 3.4851,
+      "step": 554
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.5100000000000004e-05,
+      "loss": 3.4287,
+      "step": 555
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.520000000000001e-05,
+      "loss": 3.4458,
+      "step": 556
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.530000000000001e-05,
+      "loss": 3.4678,
+      "step": 557
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.5400000000000005e-05,
+      "loss": 3.4363,
+      "step": 558
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.550000000000001e-05,
+      "loss": 3.4351,
+      "step": 559
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.560000000000001e-05,
+      "loss": 3.4291,
+      "step": 560
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.5700000000000005e-05,
+      "loss": 3.4484,
+      "step": 561
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.580000000000001e-05,
+      "loss": 3.4312,
+      "step": 562
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.590000000000001e-05,
+      "loss": 3.4503,
+      "step": 563
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 3.4308,
+      "step": 564
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.610000000000001e-05,
+      "loss": 3.4465,
+      "step": 565
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.620000000000001e-05,
+      "loss": 3.4391,
+      "step": 566
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.63e-05,
+      "loss": 3.4702,
+      "step": 567
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.6399999999999995e-05,
+      "loss": 3.5524,
+      "step": 568
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.65e-05,
+      "loss": 3.506,
+      "step": 569
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.66e-05,
+      "loss": 3.6138,
+      "step": 570
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.6699999999999996e-05,
+      "loss": 3.7177,
+      "step": 571
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.68e-05,
+      "loss": 3.6899,
+      "step": 572
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.69e-05,
+      "loss": 5.2196,
+      "step": 573
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.6999999999999996e-05,
+      "loss": 3.5806,
+      "step": 574
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.71e-05,
+      "loss": 3.5998,
+      "step": 575
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.72e-05,
+      "loss": 3.4949,
+      "step": 576
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.73e-05,
+      "loss": 3.4563,
+      "step": 577
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.74e-05,
+      "loss": 3.539,
+      "step": 578
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.7499999999999995e-05,
+      "loss": 3.4364,
+      "step": 579
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.76e-05,
+      "loss": 3.4038,
+      "step": 580
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.77e-05,
+      "loss": 3.448,
+      "step": 581
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.7799999999999995e-05,
+      "loss": 3.4455,
+      "step": 582
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.79e-05,
+      "loss": 3.4237,
+      "step": 583
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.8e-05,
+      "loss": 3.4235,
+      "step": 584
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.8099999999999996e-05,
+      "loss": 3.4206,
+      "step": 585
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.82e-05,
+      "loss": 3.4364,
+      "step": 586
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.83e-05,
+      "loss": 3.4334,
+      "step": 587
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.8399999999999997e-05,
+      "loss": 3.4457,
+      "step": 588
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.85e-05,
+      "loss": 3.4184,
+      "step": 589
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.86e-05,
+      "loss": 3.4109,
+      "step": 590
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.87e-05,
+      "loss": 3.4305,
+      "step": 591
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.88e-05,
+      "loss": 3.4668,
+      "step": 592
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.89e-05,
+      "loss": 3.47,
+      "step": 593
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.9e-05,
+      "loss": 3.4843,
+      "step": 594
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.91e-05,
+      "loss": 3.5973,
+      "step": 595
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.92e-05,
+      "loss": 3.8848,
+      "step": 596
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.93e-05,
+      "loss": 4.1709,
+      "step": 597
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.94e-05,
+      "loss": 5.4604,
+      "step": 598
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.95e-05,
+      "loss": 3.5526,
+      "step": 599
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.96e-05,
+      "loss": 3.4681,
+      "step": 600
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.97e-05,
+      "loss": 3.472,
+      "step": 601
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.9800000000000003e-05,
+      "loss": 3.4856,
+      "step": 602
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.99e-05,
+      "loss": 3.4429,
+      "step": 603
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 6e-05,
+      "loss": 3.4357,
+      "step": 604
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 6.0100000000000004e-05,
+      "loss": 3.4201,
+      "step": 605
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 6.02e-05,
+      "loss": 3.4123,
+      "step": 606
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 6.03e-05,
+      "loss": 3.4089,
+      "step": 607
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 6.04e-05,
+      "loss": 3.476,
+      "step": 608
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 6.05e-05,
+      "loss": 3.4117,
+      "step": 609
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 6.06e-05,
+      "loss": 3.4184,
+      "step": 610
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 6.07e-05,
+      "loss": 3.419,
+      "step": 611
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 6.08e-05,
+      "loss": 3.4269,
+      "step": 612
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 6.09e-05,
+      "loss": 3.4376,
+      "step": 613
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 6.1e-05,
+      "loss": 3.5009,
+      "step": 614
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 6.110000000000001e-05,
+      "loss": 3.4316,
+      "step": 615
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 6.12e-05,
+      "loss": 3.4929,
+      "step": 616
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 6.13e-05,
+      "loss": 3.4376,
+      "step": 617
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 6.14e-05,
+      "loss": 3.5242,
+      "step": 618
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 6.15e-05,
+      "loss": 3.8807,
+      "step": 619
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 6.16e-05,
+      "loss": 4.0315,
+      "step": 620
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 6.170000000000001e-05,
+      "loss": 3.8248,
+      "step": 621
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 6.18e-05,
+      "loss": 5.0871,
+      "step": 622
+    },
+    {
+      "epoch": 5.0,
+      "eval_loss": 4.694390296936035,
+      "eval_runtime": 39.6874,
+      "eval_samples_per_second": 21.14,
+      "eval_steps_per_second": 0.68,
+      "eval_wer": 1.0,
+      "step": 622
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.19e-05,
+      "loss": 3.4912,
+      "step": 623
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 6.2e-05,
+      "loss": 3.5015,
+      "step": 624
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 6.21e-05,
+      "loss": 3.4894,
+      "step": 625
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 6.220000000000001e-05,
+      "loss": 3.4278,
+      "step": 626
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 6.23e-05,
+      "loss": 3.42,
+      "step": 627
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 6.24e-05,
+      "loss": 3.4302,
+      "step": 628
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 6.25e-05,
+      "loss": 3.5188,
+      "step": 629
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 6.26e-05,
+      "loss": 3.4634,
+      "step": 630
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 6.27e-05,
+      "loss": 3.4053,
+      "step": 631
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 6.280000000000001e-05,
+      "loss": 3.4071,
+      "step": 632
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 6.29e-05,
+      "loss": 3.4126,
+      "step": 633
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 6.3e-05,
+      "loss": 3.4357,
+      "step": 634
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 6.31e-05,
+      "loss": 3.4169,
+      "step": 635
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 6.32e-05,
+      "loss": 3.41,
+      "step": 636
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 6.330000000000001e-05,
+      "loss": 3.4351,
+      "step": 637
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 6.340000000000001e-05,
+      "loss": 3.4346,
+      "step": 638
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 6.35e-05,
+      "loss": 3.4348,
+      "step": 639
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 6.36e-05,
+      "loss": 3.473,
+      "step": 640
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 6.37e-05,
+      "loss": 3.4477,
+      "step": 641
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 6.38e-05,
+      "loss": 3.5072,
+      "step": 642
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 6.390000000000001e-05,
+      "loss": 3.5109,
+      "step": 643
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 3.5351,
+      "step": 644
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 6.41e-05,
+      "loss": 3.6902,
+      "step": 645
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 6.42e-05,
+      "loss": 4.354,
+      "step": 646
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 6.43e-05,
+      "loss": 5.6441,
+      "step": 647
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 6.440000000000001e-05,
+      "loss": 3.4895,
+      "step": 648
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 6.450000000000001e-05,
+      "loss": 3.5194,
+      "step": 649
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 6.460000000000001e-05,
+      "loss": 3.4674,
+      "step": 650
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 6.47e-05,
+      "loss": 3.4137,
+      "step": 651
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 6.48e-05,
+      "loss": 3.431,
+      "step": 652
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 6.49e-05,
+      "loss": 3.514,
+      "step": 653
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 6.500000000000001e-05,
+      "loss": 3.4105,
+      "step": 654
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 6.510000000000001e-05,
+      "loss": 3.396,
+      "step": 655
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 6.52e-05,
+      "loss": 3.3931,
+      "step": 656
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 6.53e-05,
+      "loss": 3.4178,
+      "step": 657
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 6.54e-05,
+      "loss": 3.3882,
+      "step": 658
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 6.55e-05,
+      "loss": 3.4607,
+      "step": 659
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 6.560000000000001e-05,
+      "loss": 3.4425,
+      "step": 660
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 6.570000000000001e-05,
+      "loss": 3.402,
+      "step": 661
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 6.58e-05,
+      "loss": 3.4192,
+      "step": 662
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 6.59e-05,
+      "loss": 3.431,
+      "step": 663
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 6.6e-05,
+      "loss": 3.4302,
+      "step": 664
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 6.610000000000001e-05,
+      "loss": 3.4079,
+      "step": 665
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 6.620000000000001e-05,
+      "loss": 3.4481,
+      "step": 666
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 6.630000000000001e-05,
+      "loss": 3.4441,
+      "step": 667
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 6.64e-05,
+      "loss": 3.6276,
+      "step": 668
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 6.65e-05,
+      "loss": 3.6745,
+      "step": 669
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 6.66e-05,
+      "loss": 4.0938,
+      "step": 670
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 6.670000000000001e-05,
+      "loss": 4.0446,
+      "step": 671
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 6.680000000000001e-05,
+      "loss": 7.0792,
+      "step": 672
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 6.690000000000001e-05,
+      "loss": 3.4913,
+      "step": 673
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 6.7e-05,
+      "loss": 3.4613,
+      "step": 674
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 6.71e-05,
+      "loss": 3.4476,
+      "step": 675
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 6.720000000000001e-05,
+      "loss": 3.4333,
+      "step": 676
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 6.730000000000001e-05,
+      "loss": 3.4461,
+      "step": 677
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 6.740000000000001e-05,
+      "loss": 3.396,
+      "step": 678
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 6.750000000000001e-05,
+      "loss": 3.39,
+      "step": 679
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 6.76e-05,
+      "loss": 3.3948,
+      "step": 680
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 6.77e-05,
+      "loss": 3.4059,
+      "step": 681
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 6.780000000000001e-05,
+      "loss": 3.398,
+      "step": 682
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 6.790000000000001e-05,
+      "loss": 3.4053,
+      "step": 683
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 3.4109,
+      "step": 684
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 6.81e-05,
+      "loss": 3.4208,
+      "step": 685
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 6.82e-05,
+      "loss": 3.4157,
+      "step": 686
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 6.83e-05,
+      "loss": 3.423,
+      "step": 687
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 6.840000000000001e-05,
+      "loss": 3.4023,
+      "step": 688
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 6.850000000000001e-05,
+      "loss": 3.385,
+      "step": 689
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 6.860000000000001e-05,
+      "loss": 3.3905,
+      "step": 690
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 6.87e-05,
+      "loss": 3.4698,
+      "step": 691
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 6.879999999999999e-05,
+      "loss": 3.5215,
+      "step": 692
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 6.89e-05,
+      "loss": 3.4541,
+      "step": 693
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 6.9e-05,
+      "loss": 3.6146,
+      "step": 694
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 6.91e-05,
+      "loss": 3.7586,
+      "step": 695
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 6.92e-05,
+      "loss": 4.3523,
+      "step": 696
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 6.93e-05,
+      "loss": 5.1112,
+      "step": 697
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 6.939999999999999e-05,
+      "loss": 3.5266,
+      "step": 698
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 6.95e-05,
+      "loss": 3.4923,
+      "step": 699
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 6.96e-05,
+      "loss": 3.4556,
+      "step": 700
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 6.97e-05,
+      "loss": 3.42,
+      "step": 701
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 6.98e-05,
+      "loss": 3.3979,
+      "step": 702
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 6.99e-05,
+      "loss": 3.4013,
+      "step": 703
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 7e-05,
+      "loss": 3.3969,
+      "step": 704
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 7.01e-05,
+      "loss": 3.4211,
+      "step": 705
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 7.02e-05,
+      "loss": 3.4007,
+      "step": 706
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 7.03e-05,
+      "loss": 3.4008,
+      "step": 707
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 7.04e-05,
+      "loss": 3.4345,
+      "step": 708
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 7.05e-05,
+      "loss": 3.3983,
+      "step": 709
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 7.06e-05,
+      "loss": 3.4058,
+      "step": 710
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 7.07e-05,
+      "loss": 3.4266,
+      "step": 711
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 7.08e-05,
+      "loss": 3.4206,
+      "step": 712
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 7.09e-05,
+      "loss": 3.3828,
+      "step": 713
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 7.1e-05,
+      "loss": 3.4341,
+      "step": 714
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 7.11e-05,
+      "loss": 3.4127,
+      "step": 715
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 7.12e-05,
+      "loss": 3.4122,
+      "step": 716
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 7.13e-05,
+      "loss": 3.401,
+      "step": 717
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 7.14e-05,
+      "loss": 3.4634,
+      "step": 718
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 7.15e-05,
+      "loss": 3.5251,
+      "step": 719
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 7.16e-05,
+      "loss": 3.9569,
+      "step": 720
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 7.17e-05,
+      "loss": 3.9862,
+      "step": 721
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 7.18e-05,
+      "loss": 5.2295,
+      "step": 722
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 7.19e-05,
+      "loss": 3.6121,
+      "step": 723
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 7.2e-05,
+      "loss": 3.5585,
+      "step": 724
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 7.21e-05,
+      "loss": 3.4609,
+      "step": 725
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 7.22e-05,
+      "loss": 3.4284,
+      "step": 726
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 7.23e-05,
+      "loss": 3.4155,
+      "step": 727
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 7.24e-05,
+      "loss": 3.4147,
+      "step": 728
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 7.25e-05,
+      "loss": 3.3923,
+      "step": 729
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 7.26e-05,
+      "loss": 3.3738,
+      "step": 730
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 7.27e-05,
+      "loss": 3.3914,
+      "step": 731
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 7.280000000000001e-05,
+      "loss": 3.3995,
+      "step": 732
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 7.29e-05,
+      "loss": 3.4029,
+      "step": 733
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 7.3e-05,
+      "loss": 3.3871,
+      "step": 734
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 7.31e-05,
+      "loss": 3.3976,
+      "step": 735
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 7.32e-05,
+      "loss": 3.3947,
+      "step": 736
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 7.33e-05,
+      "loss": 3.4503,
+      "step": 737
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 7.340000000000001e-05,
+      "loss": 3.4057,
+      "step": 738
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 7.35e-05,
+      "loss": 3.3943,
+      "step": 739
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 7.36e-05,
+      "loss": 3.4517,
+      "step": 740
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 7.37e-05,
+      "loss": 3.4956,
+      "step": 741
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 7.38e-05,
+      "loss": 3.4466,
+      "step": 742
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 7.390000000000001e-05,
+      "loss": 3.4633,
+      "step": 743
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 7.4e-05,
+      "loss": 3.6725,
+      "step": 744
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 7.41e-05,
+      "loss": 4.0028,
+      "step": 745
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 7.42e-05,
+      "loss": 4.3562,
+      "step": 746
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 7.43e-05,
+      "loss": 6.8324,
+      "step": 747
+    },
+    {
+      "epoch": 5.0,
+      "eval_loss": 4.721221923828125,
+      "eval_runtime": 39.6107,
+      "eval_samples_per_second": 21.181,
+      "eval_steps_per_second": 0.682,
+      "eval_wer": 1.0,
+      "step": 747
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 7.44e-05,
+      "loss": 3.4562,
+      "step": 748
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 7.450000000000001e-05,
+      "loss": 3.4754,
+      "step": 749
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 7.46e-05,
+      "loss": 3.4425,
+      "step": 750
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 7.47e-05,
+      "loss": 3.3959,
+      "step": 751
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 7.48e-05,
+      "loss": 3.3904,
+      "step": 752
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 7.49e-05,
+      "loss": 3.4053,
+      "step": 753
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 7.500000000000001e-05,
+      "loss": 3.3995,
+      "step": 754
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 7.510000000000001e-05,
+      "loss": 3.3951,
+      "step": 755
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 7.52e-05,
+      "loss": 3.3967,
+      "step": 756
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 7.53e-05,
+      "loss": 3.4139,
+      "step": 757
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 7.54e-05,
+      "loss": 3.4363,
+      "step": 758
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 7.55e-05,
+      "loss": 3.4064,
+      "step": 759
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 7.560000000000001e-05,
+      "loss": 3.3958,
+      "step": 760
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 7.570000000000001e-05,
+      "loss": 3.3962,
+      "step": 761
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 7.58e-05,
+      "loss": 3.4321,
+      "step": 762
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 7.59e-05,
+      "loss": 3.4447,
+      "step": 763
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 7.6e-05,
+      "loss": 3.4108,
+      "step": 764
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 7.61e-05,
+      "loss": 3.4402,
+      "step": 765
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 7.620000000000001e-05,
+      "loss": 3.4114,
+      "step": 766
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 7.630000000000001e-05,
+      "loss": 3.4616,
+      "step": 767
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 7.64e-05,
+      "loss": 3.6044,
+      "step": 768
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 7.65e-05,
+      "loss": 3.628,
+      "step": 769
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 7.66e-05,
+      "loss": 3.6594,
+      "step": 770
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 7.670000000000001e-05,
+      "loss": 4.2139,
+      "step": 771
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 7.680000000000001e-05,
+      "loss": 6.1836,
+      "step": 772
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 7.69e-05,
+      "loss": 3.4886,
+      "step": 773
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 7.7e-05,
+      "loss": 3.5179,
+      "step": 774
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 7.71e-05,
+      "loss": 3.4309,
+      "step": 775
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 7.72e-05,
+      "loss": 3.4183,
+      "step": 776
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 7.730000000000001e-05,
+      "loss": 3.3985,
+      "step": 777
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 7.740000000000001e-05,
+      "loss": 3.4481,
+      "step": 778
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 7.75e-05,
+      "loss": 3.3889,
+      "step": 779
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 7.76e-05,
+      "loss": 3.3918,
+      "step": 780
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 7.77e-05,
+      "loss": 3.404,
+      "step": 781
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 7.780000000000001e-05,
+      "loss": 3.3787,
+      "step": 782
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 7.790000000000001e-05,
+      "loss": 3.4095,
+      "step": 783
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 7.800000000000001e-05,
+      "loss": 3.3804,
+      "step": 784
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 7.81e-05,
+      "loss": 3.4595,
+      "step": 785
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 7.82e-05,
+      "loss": 3.403,
+      "step": 786
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 7.83e-05,
+      "loss": 3.379,
+      "step": 787
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 7.840000000000001e-05,
+      "loss": 3.4168,
+      "step": 788
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 7.850000000000001e-05,
+      "loss": 3.3958,
+      "step": 789
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 7.860000000000001e-05,
+      "loss": 3.4155,
+      "step": 790
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 7.87e-05,
+      "loss": 3.4197,
+      "step": 791
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 7.88e-05,
+      "loss": 3.6726,
+      "step": 792
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 7.890000000000001e-05,
+      "loss": 3.4682,
+      "step": 793
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 7.900000000000001e-05,
+      "loss": 3.5124,
+      "step": 794
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 7.910000000000001e-05,
+      "loss": 3.541,
+      "step": 795
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 7.920000000000001e-05,
+      "loss": 4.5525,
+      "step": 796
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 7.93e-05,
+      "loss": 5.6761,
+      "step": 797
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 7.94e-05,
+      "loss": 3.4846,
+      "step": 798
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 7.950000000000001e-05,
+      "loss": 3.4553,
+      "step": 799
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 7.960000000000001e-05,
+      "loss": 3.4217,
+      "step": 800
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 7.970000000000001e-05,
+      "loss": 3.3876,
+      "step": 801
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 7.98e-05,
+      "loss": 3.3914,
+      "step": 802
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 7.99e-05,
+      "loss": 3.4051,
+      "step": 803
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 8e-05,
+      "loss": 3.3868,
+      "step": 804
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 8.010000000000001e-05,
+      "loss": 3.3821,
+      "step": 805
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 8.020000000000001e-05,
+      "loss": 3.3859,
+      "step": 806
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 8.030000000000001e-05,
+      "loss": 3.3884,
+      "step": 807
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 8.04e-05,
+      "loss": 3.4002,
+      "step": 808
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 8.05e-05,
+      "loss": 3.4144,
+      "step": 809
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 8.060000000000001e-05,
+      "loss": 3.3961,
+      "step": 810
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 8.070000000000001e-05,
+      "loss": 3.3971,
+      "step": 811
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 8.080000000000001e-05,
+      "loss": 3.3995,
+      "step": 812
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 8.090000000000001e-05,
+      "loss": 3.3895,
+      "step": 813
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 8.1e-05,
+      "loss": 3.3982,
+      "step": 814
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 8.11e-05,
+      "loss": 3.4148,
+      "step": 815
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 8.120000000000001e-05,
+      "loss": 3.4126,
+      "step": 816
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 8.13e-05,
+      "loss": 3.4553,
+      "step": 817
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 8.14e-05,
+      "loss": 3.443,
+      "step": 818
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 8.15e-05,
+      "loss": 3.7768,
+      "step": 819
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 8.16e-05,
+      "loss": 3.7885,
+      "step": 820
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 8.17e-05,
+      "loss": 3.736,
+      "step": 821
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 8.18e-05,
+      "loss": 6.0843,
+      "step": 822
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 8.19e-05,
+      "loss": 3.4285,
+      "step": 823
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 8.2e-05,
+      "loss": 3.4857,
+      "step": 824
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 8.21e-05,
+      "loss": 3.4263,
+      "step": 825
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 8.22e-05,
+      "loss": 3.4211,
+      "step": 826
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 8.23e-05,
+      "loss": 3.3926,
+      "step": 827
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 8.24e-05,
+      "loss": 3.3909,
+      "step": 828
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 8.25e-05,
+      "loss": 3.46,
+      "step": 829
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 8.26e-05,
+      "loss": 3.3973,
+      "step": 830
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 8.27e-05,
+      "loss": 3.3739,
+      "step": 831
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 8.28e-05,
+      "loss": 3.3859,
+      "step": 832
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 8.29e-05,
+      "loss": 3.3815,
+      "step": 833
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 8.3e-05,
+      "loss": 3.3895,
+      "step": 834
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 8.31e-05,
+      "loss": 3.397,
+      "step": 835
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 8.32e-05,
+      "loss": 3.3991,
+      "step": 836
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 8.33e-05,
+      "loss": 3.4068,
+      "step": 837
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 8.34e-05,
+      "loss": 3.3908,
+      "step": 838
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 8.35e-05,
+      "loss": 3.407,
+      "step": 839
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 8.36e-05,
+      "loss": 3.4071,
+      "step": 840
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 8.37e-05,
+      "loss": 3.4199,
+      "step": 841
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 8.38e-05,
+      "loss": 3.3986,
+      "step": 842
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 8.39e-05,
+      "loss": 3.4503,
+      "step": 843
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 8.4e-05,
+      "loss": 3.5765,
+      "step": 844
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 8.41e-05,
+      "loss": 3.5476,
+      "step": 845
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 8.42e-05,
+      "loss": 3.9662,
+      "step": 846
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 8.43e-05,
+      "loss": 5.2809,
+      "step": 847
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 8.44e-05,
+      "loss": 3.4424,
+      "step": 848
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 8.450000000000001e-05,
+      "loss": 3.5335,
+      "step": 849
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 8.46e-05,
+      "loss": 3.463,
+      "step": 850
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 8.47e-05,
+      "loss": 3.4037,
+      "step": 851
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 8.48e-05,
+      "loss": 3.3916,
+      "step": 852
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 8.49e-05,
+      "loss": 3.3907,
+      "step": 853
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 8.5e-05,
+      "loss": 3.393,
+      "step": 854
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 8.510000000000001e-05,
+      "loss": 3.3936,
+      "step": 855
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 8.52e-05,
+      "loss": 3.3901,
+      "step": 856
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 8.53e-05,
+      "loss": 3.3925,
+      "step": 857
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 8.54e-05,
+      "loss": 3.3838,
+      "step": 858
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 8.55e-05,
+      "loss": 3.3828,
+      "step": 859
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 8.560000000000001e-05,
+      "loss": 3.4088,
+      "step": 860
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 8.57e-05,
+      "loss": 3.4213,
+      "step": 861
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 8.58e-05,
+      "loss": 3.429,
+      "step": 862
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 8.59e-05,
+      "loss": 3.426,
+      "step": 863
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 8.6e-05,
+      "loss": 3.4201,
+      "step": 864
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 8.61e-05,
+      "loss": 3.4385,
+      "step": 865
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 8.620000000000001e-05,
+      "loss": 3.4607,
+      "step": 866
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 8.63e-05,
+      "loss": 3.4441,
+      "step": 867
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 8.64e-05,
+      "loss": 3.5602,
+      "step": 868
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 8.65e-05,
+      "loss": 3.7873,
+      "step": 869
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 8.66e-05,
+      "loss": 4.4211,
+      "step": 870
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.67e-05,
+      "loss": 5.8494,
+      "step": 871
+    },
+    {
+      "epoch": 7.0,
+      "eval_loss": 5.310427188873291,
+      "eval_runtime": 36.9127,
+      "eval_samples_per_second": 22.783,
+      "eval_steps_per_second": 0.731,
+      "eval_wer": 1.0,
+      "step": 871
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 8.680000000000001e-05,
+      "loss": 3.461,
+      "step": 872
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 8.69e-05,
+      "loss": 3.8809,
+      "step": 873
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 8.7e-05,
+      "loss": 3.4618,
+      "step": 874
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 8.71e-05,
+      "loss": 3.4032,
+      "step": 875
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 8.72e-05,
+      "loss": 3.3962,
+      "step": 876
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 8.730000000000001e-05,
+      "loss": 3.4039,
+      "step": 877
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 8.740000000000001e-05,
+      "loss": 3.3814,
+      "step": 878
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 8.75e-05,
+      "loss": 3.405,
+      "step": 879
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 8.76e-05,
+      "loss": 3.3953,
+      "step": 880
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 8.77e-05,
+      "loss": 3.4088,
+      "step": 881
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 8.78e-05,
+      "loss": 3.3837,
+      "step": 882
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 8.790000000000001e-05,
+      "loss": 3.3818,
+      "step": 883
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 3.381,
+      "step": 884
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 8.81e-05,
+      "loss": 3.4145,
+      "step": 885
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 8.82e-05,
+      "loss": 3.3978,
+      "step": 886
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 8.83e-05,
+      "loss": 3.3993,
+      "step": 887
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 8.840000000000001e-05,
+      "loss": 3.4631,
+      "step": 888
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 8.850000000000001e-05,
+      "loss": 3.4078,
+      "step": 889
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 8.86e-05,
+      "loss": 3.486,
+      "step": 890
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 8.87e-05,
+      "loss": 3.4268,
+      "step": 891
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 8.88e-05,
+      "loss": 3.4362,
+      "step": 892
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 8.89e-05,
+      "loss": 3.5654,
+      "step": 893
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 8.900000000000001e-05,
+      "loss": 3.6173,
+      "step": 894
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 8.910000000000001e-05,
+      "loss": 4.0168,
+      "step": 895
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 8.92e-05,
+      "loss": 4.973,
+      "step": 896
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 8.93e-05,
+      "loss": 3.5004,
+      "step": 897
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 8.94e-05,
+      "loss": 3.5823,
+      "step": 898
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 8.950000000000001e-05,
+      "loss": 3.4741,
+      "step": 899
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 8.960000000000001e-05,
+      "loss": 3.4438,
+      "step": 900
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 8.970000000000001e-05,
+      "loss": 3.4202,
+      "step": 901
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 8.98e-05,
+      "loss": 3.4107,
+      "step": 902
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 8.99e-05,
+      "loss": 3.3951,
+      "step": 903
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 9e-05,
+      "loss": 3.3939,
+      "step": 904
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 9.010000000000001e-05,
+      "loss": 3.3988,
+      "step": 905
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 9.020000000000001e-05,
+      "loss": 3.3947,
+      "step": 906
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 9.030000000000001e-05,
+      "loss": 3.3589,
+      "step": 907
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 9.04e-05,
+      "loss": 3.3827,
+      "step": 908
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 9.05e-05,
+      "loss": 3.399,
+      "step": 909
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 9.06e-05,
+      "loss": 3.4123,
+      "step": 910
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 9.070000000000001e-05,
+      "loss": 3.4028,
+      "step": 911
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 9.080000000000001e-05,
+      "loss": 3.4054,
+      "step": 912
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 9.090000000000001e-05,
+      "loss": 3.3936,
+      "step": 913
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 9.1e-05,
+      "loss": 3.437,
+      "step": 914
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 9.11e-05,
+      "loss": 3.4241,
+      "step": 915
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 9.120000000000001e-05,
+      "loss": 3.4708,
+      "step": 916
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 9.130000000000001e-05,
+      "loss": 3.4484,
+      "step": 917
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 9.140000000000001e-05,
+      "loss": 3.5468,
+      "step": 918
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 9.15e-05,
+      "loss": 3.8123,
+      "step": 919
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 9.16e-05,
+      "loss": 4.0886,
+      "step": 920
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 9.17e-05,
+      "loss": 4.9131,
+      "step": 921
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 9.180000000000001e-05,
+      "loss": 3.5434,
+      "step": 922
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 9.190000000000001e-05,
+      "loss": 3.4536,
+      "step": 923
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 3.4465,
+      "step": 924
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 9.21e-05,
+      "loss": 3.4371,
+      "step": 925
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 9.22e-05,
+      "loss": 3.4232,
+      "step": 926
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 9.230000000000001e-05,
+      "loss": 3.3906,
+      "step": 927
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 9.240000000000001e-05,
+      "loss": 3.3992,
+      "step": 928
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 9.250000000000001e-05,
+      "loss": 3.3921,
+      "step": 929
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 9.260000000000001e-05,
+      "loss": 3.3851,
+      "step": 930
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 9.27e-05,
+      "loss": 3.3857,
+      "step": 931
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 9.28e-05,
+      "loss": 3.3886,
+      "step": 932
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 9.290000000000001e-05,
+      "loss": 3.3741,
+      "step": 933
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 9.300000000000001e-05,
+      "loss": 3.3997,
+      "step": 934
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 9.310000000000001e-05,
+      "loss": 3.4278,
+      "step": 935
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 9.320000000000002e-05,
+      "loss": 3.3854,
+      "step": 936
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 9.33e-05,
+      "loss": 3.3974,
+      "step": 937
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 9.340000000000001e-05,
+      "loss": 3.4039,
+      "step": 938
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 9.350000000000001e-05,
+      "loss": 3.4221,
+      "step": 939
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 9.360000000000001e-05,
+      "loss": 3.4255,
+      "step": 940
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 9.370000000000001e-05,
+      "loss": 3.471,
+      "step": 941
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 9.38e-05,
+      "loss": 3.4719,
+      "step": 942
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 9.39e-05,
+      "loss": 3.5246,
+      "step": 943
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 9.4e-05,
+      "loss": 3.5401,
+      "step": 944
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 9.41e-05,
+      "loss": 3.7673,
+      "step": 945
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 9.42e-05,
+      "loss": 5.3016,
+      "step": 946
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 9.43e-05,
+      "loss": 3.5359,
+      "step": 947
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 9.44e-05,
+      "loss": 3.4756,
+      "step": 948
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 9.449999999999999e-05,
+      "loss": 3.4442,
+      "step": 949
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 9.46e-05,
+      "loss": 3.4462,
+      "step": 950
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 9.47e-05,
+      "loss": 3.4066,
+      "step": 951
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 9.48e-05,
+      "loss": 3.4047,
+      "step": 952
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 9.49e-05,
+      "loss": 3.3845,
+      "step": 953
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 9.5e-05,
+      "loss": 3.3732,
+      "step": 954
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.51e-05,
+      "loss": 3.4269,
+      "step": 955
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.52e-05,
+      "loss": 3.3724,
+      "step": 956
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.53e-05,
+      "loss": 3.3864,
+      "step": 957
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.54e-05,
+      "loss": 3.3777,
+      "step": 958
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.55e-05,
+      "loss": 3.3995,
+      "step": 959
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.56e-05,
+      "loss": 3.4015,
+      "step": 960
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.57e-05,
+      "loss": 3.3791,
+      "step": 961
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.58e-05,
+      "loss": 3.3999,
+      "step": 962
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.59e-05,
+      "loss": 3.4725,
+      "step": 963
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.6e-05,
+      "loss": 3.4273,
+      "step": 964
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.61e-05,
+      "loss": 3.4752,
+      "step": 965
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.620000000000001e-05,
+      "loss": 3.4574,
+      "step": 966
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.63e-05,
+      "loss": 3.5005,
+      "step": 967
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.64e-05,
+      "loss": 3.642,
+      "step": 968
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.65e-05,
+      "loss": 3.6436,
+      "step": 969
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.66e-05,
+      "loss": 4.0689,
+      "step": 970
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.67e-05,
+      "loss": 5.4368,
+      "step": 971
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.680000000000001e-05,
+      "loss": 3.5517,
+      "step": 972
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.69e-05,
+      "loss": 3.4409,
+      "step": 973
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.7e-05,
+      "loss": 3.4217,
+      "step": 974
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.71e-05,
+      "loss": 3.4154,
+      "step": 975
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.72e-05,
+      "loss": 3.3877,
+      "step": 976
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.730000000000001e-05,
+      "loss": 3.3719,
+      "step": 977
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.74e-05,
+      "loss": 3.3926,
+      "step": 978
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.75e-05,
+      "loss": 3.3969,
+      "step": 979
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.76e-05,
+      "loss": 3.3755,
+      "step": 980
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.77e-05,
+      "loss": 3.3894,
+      "step": 981
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.78e-05,
+      "loss": 3.3776,
+      "step": 982
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.790000000000001e-05,
+      "loss": 3.3971,
+      "step": 983
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.8e-05,
+      "loss": 3.374,
+      "step": 984
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.81e-05,
+      "loss": 3.4223,
+      "step": 985
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.82e-05,
+      "loss": 3.4032,
+      "step": 986
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.83e-05,
+      "loss": 3.398,
+      "step": 987
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.84e-05,
+      "loss": 3.3999,
+      "step": 988
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.850000000000001e-05,
+      "loss": 3.4509,
+      "step": 989
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.86e-05,
+      "loss": 3.4446,
+      "step": 990
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.87e-05,
+      "loss": 3.5017,
+      "step": 991
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.88e-05,
+      "loss": 3.6567,
+      "step": 992
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.89e-05,
+      "loss": 3.641,
+      "step": 993
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.900000000000001e-05,
+      "loss": 3.6632,
+      "step": 994
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.910000000000001e-05,
+      "loss": 4.4714,
+      "step": 995
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.92e-05,
+      "loss": 5.3396,
+      "step": 996
+    },
+    {
+      "epoch": 7.0,
+      "eval_loss": 5.220389366149902,
+      "eval_runtime": 39.5757,
+      "eval_samples_per_second": 21.25,
+      "eval_steps_per_second": 0.682,
+      "eval_wer": 1.0,
+      "step": 996
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 9.93e-05,
+      "loss": 3.4391,
+      "step": 997
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 9.94e-05,
+      "loss": 3.4615,
+      "step": 998
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 9.95e-05,
+      "loss": 3.4253,
+      "step": 999
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 9.960000000000001e-05,
+      "loss": 3.3917,
+      "step": 1000
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 9.970000000000001e-05,
+      "loss": 3.404,
+      "step": 1001
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 9.98e-05,
+      "loss": 3.3803,
+      "step": 1002
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 9.99e-05,
+      "loss": 3.384,
+      "step": 1003
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.0001,
+      "loss": 3.3955,
+      "step": 1004
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 9.999983844911147e-05,
+      "loss": 3.368,
+      "step": 1005
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 9.999967689822295e-05,
+      "loss": 3.369,
+      "step": 1006
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 9.999951534733442e-05,
+      "loss": 3.4097,
+      "step": 1007
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 9.999935379644588e-05,
+      "loss": 3.4022,
+      "step": 1008
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 9.999919224555737e-05,
+      "loss": 3.3879,
+      "step": 1009
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 9.999903069466883e-05,
+      "loss": 3.4259,
+      "step": 1010
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 9.999886914378029e-05,
+      "loss": 3.3831,
+      "step": 1011
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 9.999870759289176e-05,
+      "loss": 3.4025,
+      "step": 1012
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 9.999854604200324e-05,
+      "loss": 3.4024,
+      "step": 1013
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 9.99983844911147e-05,
+      "loss": 3.4255,
+      "step": 1014
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 9.999822294022618e-05,
+      "loss": 3.4349,
+      "step": 1015
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 9.999806138933765e-05,
+      "loss": 3.4384,
+      "step": 1016
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 9.999789983844911e-05,
+      "loss": 3.4449,
+      "step": 1017
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 9.999773828756059e-05,
+      "loss": 3.7325,
+      "step": 1018
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 9.999757673667206e-05,
+      "loss": 3.5972,
+      "step": 1019
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 9.999741518578352e-05,
+      "loss": 4.3002,
+      "step": 1020
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 9.9997253634895e-05,
+      "loss": 5.7384,
+      "step": 1021
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 9.999709208400647e-05,
+      "loss": 3.5591,
+      "step": 1022
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 9.999693053311794e-05,
+      "loss": 3.4809,
+      "step": 1023
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 9.999676898222941e-05,
+      "loss": 3.4305,
+      "step": 1024
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 9.999660743134087e-05,
+      "loss": 3.502,
+      "step": 1025
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 9.999644588045235e-05,
+      "loss": 3.4397,
+      "step": 1026
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 9.999628432956382e-05,
+      "loss": 3.3616,
+      "step": 1027
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 9.999612277867528e-05,
+      "loss": 3.3901,
+      "step": 1028
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 9.999596122778676e-05,
+      "loss": 3.4012,
+      "step": 1029
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 9.999579967689822e-05,
+      "loss": 3.3935,
+      "step": 1030
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 9.99956381260097e-05,
+      "loss": 3.3888,
+      "step": 1031
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 9.999547657512117e-05,
+      "loss": 3.3742,
+      "step": 1032
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 9.999531502423263e-05,
+      "loss": 3.3868,
+      "step": 1033
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 9.999515347334411e-05,
+      "loss": 3.4084,
+      "step": 1034
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 9.999499192245558e-05,
+      "loss": 3.4094,
+      "step": 1035
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 9.999483037156704e-05,
+      "loss": 3.3967,
+      "step": 1036
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 9.999466882067852e-05,
+      "loss": 3.3964,
+      "step": 1037
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 9.999450726978998e-05,
+      "loss": 3.4033,
+      "step": 1038
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 9.999434571890146e-05,
+      "loss": 3.4991,
+      "step": 1039
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 9.999418416801293e-05,
+      "loss": 3.4055,
+      "step": 1040
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 9.999402261712439e-05,
+      "loss": 3.4862,
+      "step": 1041
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 9.999386106623587e-05,
+      "loss": 3.4238,
+      "step": 1042
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 9.999369951534734e-05,
+      "loss": 3.5208,
+      "step": 1043
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 9.99935379644588e-05,
+      "loss": 3.6646,
+      "step": 1044
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 9.999337641357028e-05,
+      "loss": 3.7567,
+      "step": 1045
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 9.999321486268175e-05,
+      "loss": 6.3611,
+      "step": 1046
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 9.999305331179322e-05,
+      "loss": 3.5013,
+      "step": 1047
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 9.999289176090469e-05,
+      "loss": 3.4481,
+      "step": 1048
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 9.999273021001617e-05,
+      "loss": 3.4308,
+      "step": 1049
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 9.999256865912763e-05,
+      "loss": 3.4308,
+      "step": 1050
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 9.99924071082391e-05,
+      "loss": 3.4079,
+      "step": 1051
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 9.999224555735058e-05,
+      "loss": 3.3887,
+      "step": 1052
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 9.999208400646204e-05,
+      "loss": 3.3715,
+      "step": 1053
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 9.99919224555735e-05,
+      "loss": 3.3801,
+      "step": 1054
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 9.999176090468499e-05,
+      "loss": 3.3912,
+      "step": 1055
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 9.999159935379645e-05,
+      "loss": 3.3731,
+      "step": 1056
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 9.999143780290791e-05,
+      "loss": 3.3875,
+      "step": 1057
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 9.99912762520194e-05,
+      "loss": 3.3544,
+      "step": 1058
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 9.999111470113086e-05,
+      "loss": 3.3853,
+      "step": 1059
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 9.999095315024232e-05,
+      "loss": 3.3797,
+      "step": 1060
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 9.999079159935381e-05,
+      "loss": 3.427,
+      "step": 1061
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 9.999063004846527e-05,
+      "loss": 3.4031,
+      "step": 1062
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 9.999046849757674e-05,
+      "loss": 3.386,
+      "step": 1063
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 9.999030694668822e-05,
+      "loss": 3.4627,
+      "step": 1064
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 9.999014539579969e-05,
+      "loss": 3.4078,
+      "step": 1065
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 9.998998384491115e-05,
+      "loss": 3.4639,
+      "step": 1066
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 9.998982229402262e-05,
+      "loss": 3.5078,
+      "step": 1067
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 9.99896607431341e-05,
+      "loss": 3.978,
+      "step": 1068
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 9.998949919224556e-05,
+      "loss": 3.753,
+      "step": 1069
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 9.998933764135703e-05,
+      "loss": 3.6902,
+      "step": 1070
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 9.998917609046851e-05,
+      "loss": 4.9658,
+      "step": 1071
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 9.998901453957997e-05,
+      "loss": 3.5267,
+      "step": 1072
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 9.998885298869144e-05,
+      "loss": 3.4399,
+      "step": 1073
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 9.998869143780292e-05,
+      "loss": 3.415,
+      "step": 1074
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 9.998852988691438e-05,
+      "loss": 3.4607,
+      "step": 1075
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 9.998836833602586e-05,
+      "loss": 3.4128,
+      "step": 1076
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 9.998820678513733e-05,
+      "loss": 3.4415,
+      "step": 1077
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 9.998804523424879e-05,
+      "loss": 3.4104,
+      "step": 1078
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 9.998788368336027e-05,
+      "loss": 3.3783,
+      "step": 1079
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 9.998772213247173e-05,
+      "loss": 3.4035,
+      "step": 1080
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 9.99875605815832e-05,
+      "loss": 3.3918,
+      "step": 1081
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 9.998739903069467e-05,
+      "loss": 3.3942,
+      "step": 1082
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 9.998723747980614e-05,
+      "loss": 3.3841,
+      "step": 1083
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 9.998707592891762e-05,
+      "loss": 3.3582,
+      "step": 1084
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 9.998691437802908e-05,
+      "loss": 3.3813,
+      "step": 1085
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 9.998675282714055e-05,
+      "loss": 3.4151,
+      "step": 1086
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 9.998659127625203e-05,
+      "loss": 3.4284,
+      "step": 1087
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 9.998642972536349e-05,
+      "loss": 3.4158,
+      "step": 1088
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 9.998626817447496e-05,
+      "loss": 3.4446,
+      "step": 1089
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 9.998610662358644e-05,
+      "loss": 3.4477,
+      "step": 1090
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 9.99859450726979e-05,
+      "loss": 3.4937,
+      "step": 1091
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 9.998578352180938e-05,
+      "loss": 3.5366,
+      "step": 1092
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 9.998562197092084e-05,
+      "loss": 3.6949,
+      "step": 1093
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 9.998546042003231e-05,
+      "loss": 4.0449,
+      "step": 1094
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 9.998529886914379e-05,
+      "loss": 4.28,
+      "step": 1095
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 9.998513731825525e-05,
+      "loss": 5.2403,
+      "step": 1096
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 9.998497576736672e-05,
+      "loss": 3.4899,
+      "step": 1097
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 9.99848142164782e-05,
+      "loss": 3.4207,
+      "step": 1098
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 9.998465266558966e-05,
+      "loss": 3.4033,
+      "step": 1099
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 9.998449111470114e-05,
+      "loss": 3.4102,
+      "step": 1100
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 9.998432956381261e-05,
+      "loss": 3.3837,
+      "step": 1101
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 9.998416801292407e-05,
+      "loss": 3.4049,
+      "step": 1102
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 9.998400646203555e-05,
+      "loss": 3.3739,
+      "step": 1103
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 9.998384491114702e-05,
+      "loss": 3.3761,
+      "step": 1104
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 9.998368336025848e-05,
+      "loss": 3.3799,
+      "step": 1105
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 9.998352180936995e-05,
+      "loss": 3.3733,
+      "step": 1106
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 9.998336025848143e-05,
+      "loss": 3.3693,
+      "step": 1107
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 9.99831987075929e-05,
+      "loss": 3.3638,
+      "step": 1108
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 9.998303715670436e-05,
+      "loss": 3.3785,
+      "step": 1109
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 9.998287560581585e-05,
+      "loss": 3.3859,
+      "step": 1110
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 9.998271405492731e-05,
+      "loss": 3.455,
+      "step": 1111
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 9.998255250403877e-05,
+      "loss": 3.4195,
+      "step": 1112
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 9.998239095315026e-05,
+      "loss": 3.4314,
+      "step": 1113
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 9.998222940226172e-05,
+      "loss": 3.4077,
+      "step": 1114
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 9.998206785137318e-05,
+      "loss": 3.5853,
+      "step": 1115
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 9.998190630048467e-05,
+      "loss": 3.4744,
+      "step": 1116
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 9.998174474959613e-05,
+      "loss": 3.7246,
+      "step": 1117
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 9.998158319870759e-05,
+      "loss": 3.6937,
+      "step": 1118
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 9.998142164781907e-05,
+      "loss": 3.7955,
+      "step": 1119
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 9.998126009693054e-05,
+      "loss": 4.4085,
+      "step": 1120
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 9.9981098546042e-05,
+      "loss": 4.9125,
+      "step": 1121
+    },
+    {
+      "epoch": 8.0,
+      "eval_loss": 4.6698760986328125,
+      "eval_runtime": 39.0358,
+      "eval_samples_per_second": 21.544,
+      "eval_steps_per_second": 0.692,
+      "eval_wer": 1.0,
+      "step": 1121
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 9.998093699515348e-05,
+      "loss": 3.4491,
+      "step": 1122
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 9.998077544426495e-05,
+      "loss": 3.4483,
+      "step": 1123
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 9.998061389337642e-05,
+      "loss": 3.4148,
+      "step": 1124
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 9.998045234248789e-05,
+      "loss": 3.3895,
+      "step": 1125
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 9.998029079159937e-05,
+      "loss": 3.4017,
+      "step": 1126
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 9.998012924071083e-05,
+      "loss": 3.4107,
+      "step": 1127
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 9.99799676898223e-05,
+      "loss": 3.3896,
+      "step": 1128
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 9.997980613893378e-05,
+      "loss": 3.3726,
+      "step": 1129
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 9.997964458804524e-05,
+      "loss": 3.3539,
+      "step": 1130
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 9.997948303715671e-05,
+      "loss": 3.4048,
+      "step": 1131
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 9.997932148626818e-05,
+      "loss": 3.3927,
+      "step": 1132
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 9.997915993537965e-05,
+      "loss": 3.3872,
+      "step": 1133
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 9.997899838449111e-05,
+      "loss": 3.4085,
+      "step": 1134
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 9.997883683360259e-05,
+      "loss": 3.406,
+      "step": 1135
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 9.997867528271406e-05,
+      "loss": 3.3907,
+      "step": 1136
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 9.997851373182552e-05,
+      "loss": 3.4142,
+      "step": 1137
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 9.9978352180937e-05,
+      "loss": 3.4349,
+      "step": 1138
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 9.997819063004847e-05,
+      "loss": 3.4087,
+      "step": 1139
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 9.997802907915993e-05,
+      "loss": 3.4383,
+      "step": 1140
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 9.997786752827141e-05,
+      "loss": 3.4465,
+      "step": 1141
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 9.997770597738289e-05,
+      "loss": 3.4623,
+      "step": 1142
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 9.997754442649435e-05,
+      "loss": 3.4872,
+      "step": 1143
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 9.997738287560582e-05,
+      "loss": 3.6398,
+      "step": 1144
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 9.99772213247173e-05,
+      "loss": 4.033,
+      "step": 1145
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 9.997705977382876e-05,
+      "loss": 5.2918,
+      "step": 1146
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 9.997689822294023e-05,
+      "loss": 3.5296,
+      "step": 1147
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 9.99767366720517e-05,
+      "loss": 3.4976,
+      "step": 1148
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 9.997657512116317e-05,
+      "loss": 3.4582,
+      "step": 1149
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 9.997641357027464e-05,
+      "loss": 3.3995,
+      "step": 1150
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 9.99762520193861e-05,
+      "loss": 3.4191,
+      "step": 1151
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 9.997609046849758e-05,
+      "loss": 3.4158,
+      "step": 1152
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 9.997592891760906e-05,
+      "loss": 3.3898,
+      "step": 1153
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 9.997576736672052e-05,
+      "loss": 3.3729,
+      "step": 1154
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 9.997560581583199e-05,
+      "loss": 3.3822,
+      "step": 1155
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 9.997544426494347e-05,
+      "loss": 3.367,
+      "step": 1156
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 9.997528271405493e-05,
+      "loss": 3.3844,
+      "step": 1157
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 9.99751211631664e-05,
+      "loss": 3.3832,
+      "step": 1158
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 9.997495961227788e-05,
+      "loss": 3.4329,
+      "step": 1159
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 9.997479806138934e-05,
+      "loss": 3.387,
+      "step": 1160
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 9.99746365105008e-05,
+      "loss": 3.3906,
+      "step": 1161
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 9.997447495961229e-05,
+      "loss": 3.4006,
+      "step": 1162
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 9.997431340872375e-05,
+      "loss": 3.3988,
+      "step": 1163
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 9.997415185783521e-05,
+      "loss": 3.4054,
+      "step": 1164
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 9.99739903069467e-05,
+      "loss": 3.4151,
+      "step": 1165
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 9.997382875605816e-05,
+      "loss": 3.4567,
+      "step": 1166
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 9.997366720516963e-05,
+      "loss": 3.5884,
+      "step": 1167
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 9.99735056542811e-05,
+      "loss": 3.5689,
+      "step": 1168
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 9.997334410339258e-05,
+      "loss": 3.738,
+      "step": 1169
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 9.997318255250404e-05,
+      "loss": 4.3097,
+      "step": 1170
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 9.997302100161551e-05,
+      "loss": 6.1488,
+      "step": 1171
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 9.997285945072699e-05,
+      "loss": 3.5111,
+      "step": 1172
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 9.997269789983845e-05,
+      "loss": 3.4751,
+      "step": 1173
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 9.997253634894992e-05,
+      "loss": 3.4113,
+      "step": 1174
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 9.99723747980614e-05,
+      "loss": 3.4133,
+      "step": 1175
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 9.997221324717286e-05,
+      "loss": 3.4085,
+      "step": 1176
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 9.997205169628434e-05,
+      "loss": 3.3827,
+      "step": 1177
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 9.997189014539581e-05,
+      "loss": 3.3832,
+      "step": 1178
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 9.997172859450727e-05,
+      "loss": 3.3611,
+      "step": 1179
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 9.997156704361875e-05,
+      "loss": 3.3576,
+      "step": 1180
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 9.997140549273022e-05,
+      "loss": 3.3898,
+      "step": 1181
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 9.997124394184168e-05,
+      "loss": 3.4046,
+      "step": 1182
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 9.997108239095316e-05,
+      "loss": 3.39,
+      "step": 1183
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 9.997092084006463e-05,
+      "loss": 3.4155,
+      "step": 1184
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 9.99707592891761e-05,
+      "loss": 3.4019,
+      "step": 1185
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 9.997059773828756e-05,
+      "loss": 3.3949,
+      "step": 1186
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 9.997043618739903e-05,
+      "loss": 3.425,
+      "step": 1187
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 9.997027463651051e-05,
+      "loss": 3.4208,
+      "step": 1188
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 9.997011308562197e-05,
+      "loss": 3.4065,
+      "step": 1189
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 9.996995153473344e-05,
+      "loss": 3.4367,
+      "step": 1190
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 9.996978998384492e-05,
+      "loss": 3.8457,
+      "step": 1191
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 9.996962843295638e-05,
+      "loss": 3.4878,
+      "step": 1192
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 9.996946688206786e-05,
+      "loss": 3.4507,
+      "step": 1193
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 9.996930533117933e-05,
+      "loss": 3.643,
+      "step": 1194
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 9.996914378029079e-05,
+      "loss": 4.0981,
+      "step": 1195
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 9.996898222940227e-05,
+      "loss": 5.1538,
+      "step": 1196
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 9.996882067851374e-05,
+      "loss": 3.459,
+      "step": 1197
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 9.99686591276252e-05,
+      "loss": 3.4915,
+      "step": 1198
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 9.996849757673668e-05,
+      "loss": 3.4582,
+      "step": 1199
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 9.996833602584814e-05,
+      "loss": 3.4252,
+      "step": 1200
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 9.996817447495962e-05,
+      "loss": 3.4025,
+      "step": 1201
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 9.996801292407109e-05,
+      "loss": 3.3916,
+      "step": 1202
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 9.996785137318255e-05,
+      "loss": 3.3966,
+      "step": 1203
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 9.996768982229403e-05,
+      "loss": 3.3714,
+      "step": 1204
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 9.99675282714055e-05,
+      "loss": 3.4027,
+      "step": 1205
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 9.996736672051696e-05,
+      "loss": 3.396,
+      "step": 1206
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 9.996720516962844e-05,
+      "loss": 3.3744,
+      "step": 1207
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 9.996704361873991e-05,
+      "loss": 3.3786,
+      "step": 1208
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 9.996688206785138e-05,
+      "loss": 3.3905,
+      "step": 1209
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 9.996672051696285e-05,
+      "loss": 3.4102,
+      "step": 1210
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 9.996655896607433e-05,
+      "loss": 3.413,
+      "step": 1211
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 9.996639741518579e-05,
+      "loss": 3.4086,
+      "step": 1212
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 9.996623586429725e-05,
+      "loss": 3.4051,
+      "step": 1213
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 9.996607431340874e-05,
+      "loss": 3.4011,
+      "step": 1214
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 9.99659127625202e-05,
+      "loss": 3.4307,
+      "step": 1215
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 9.996575121163166e-05,
+      "loss": 3.4706,
+      "step": 1216
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 9.996558966074315e-05,
+      "loss": 3.5299,
+      "step": 1217
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 9.996542810985461e-05,
+      "loss": 3.4886,
+      "step": 1218
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 9.996526655896607e-05,
+      "loss": 3.6451,
+      "step": 1219
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 9.996510500807755e-05,
+      "loss": 4.2715,
+      "step": 1220
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 9.996494345718902e-05,
+      "loss": 5.2142,
+      "step": 1221
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 9.996478190630048e-05,
+      "loss": 3.4578,
+      "step": 1222
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 9.996462035541196e-05,
+      "loss": 3.4484,
+      "step": 1223
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 9.996445880452343e-05,
+      "loss": 3.4359,
+      "step": 1224
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 9.99642972536349e-05,
+      "loss": 3.4089,
+      "step": 1225
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 9.996413570274637e-05,
+      "loss": 3.4101,
+      "step": 1226
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 9.996397415185784e-05,
+      "loss": 3.4087,
+      "step": 1227
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 9.99638126009693e-05,
+      "loss": 3.3663,
+      "step": 1228
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 9.996365105008078e-05,
+      "loss": 3.374,
+      "step": 1229
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 9.996348949919226e-05,
+      "loss": 3.3867,
+      "step": 1230
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 9.996332794830372e-05,
+      "loss": 3.3652,
+      "step": 1231
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 9.996316639741519e-05,
+      "loss": 3.3753,
+      "step": 1232
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 9.996300484652667e-05,
+      "loss": 3.399,
+      "step": 1233
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 9.996284329563813e-05,
+      "loss": 3.407,
+      "step": 1234
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 9.996268174474959e-05,
+      "loss": 3.3957,
+      "step": 1235
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 9.996252019386108e-05,
+      "loss": 3.4113,
+      "step": 1236
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 9.996235864297254e-05,
+      "loss": 3.3976,
+      "step": 1237
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 9.9962197092084e-05,
+      "loss": 3.4092,
+      "step": 1238
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 9.996203554119549e-05,
+      "loss": 3.4195,
+      "step": 1239
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 9.996187399030695e-05,
+      "loss": 3.4528,
+      "step": 1240
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 9.996171243941841e-05,
+      "loss": 3.4191,
+      "step": 1241
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 9.996155088852989e-05,
+      "loss": 3.5653,
+      "step": 1242
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 9.996138933764136e-05,
+      "loss": 3.5675,
+      "step": 1243
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 9.996122778675283e-05,
+      "loss": 3.7032,
+      "step": 1244
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 9.99610662358643e-05,
+      "loss": 3.7538,
+      "step": 1245
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 9.996090468497578e-05,
+      "loss": 5.5621,
+      "step": 1246
+    },
+    {
+      "epoch": 9.0,
+      "eval_loss": 5.817511558532715,
+      "eval_runtime": 38.2346,
+      "eval_samples_per_second": 21.996,
+      "eval_steps_per_second": 0.706,
+      "eval_wer": 1.0,
+      "step": 1246
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 9.996074313408724e-05,
+      "loss": 3.475,
+      "step": 1247
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 9.996058158319871e-05,
+      "loss": 3.4527,
+      "step": 1248
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 9.996042003231019e-05,
+      "loss": 3.3908,
+      "step": 1249
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 9.996025848142165e-05,
+      "loss": 3.4666,
+      "step": 1250
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 9.996009693053312e-05,
+      "loss": 3.3853,
+      "step": 1251
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 9.99599353796446e-05,
+      "loss": 3.3962,
+      "step": 1252
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 9.995977382875606e-05,
+      "loss": 3.4309,
+      "step": 1253
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 9.995961227786754e-05,
+      "loss": 3.3723,
+      "step": 1254
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 9.9959450726979e-05,
+      "loss": 3.3733,
+      "step": 1255
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 9.995928917609047e-05,
+      "loss": 3.3608,
+      "step": 1256
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 9.995912762520195e-05,
+      "loss": 3.3593,
+      "step": 1257
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 9.995896607431341e-05,
+      "loss": 3.361,
+      "step": 1258
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 9.995880452342488e-05,
+      "loss": 3.3998,
+      "step": 1259
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 9.995864297253636e-05,
+      "loss": 3.3948,
+      "step": 1260
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 9.995848142164782e-05,
+      "loss": 3.432,
+      "step": 1261
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 9.99583198707593e-05,
+      "loss": 3.4246,
+      "step": 1262
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 9.995815831987077e-05,
+      "loss": 3.4307,
+      "step": 1263
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 9.995799676898223e-05,
+      "loss": 3.4283,
+      "step": 1264
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 9.995783521809371e-05,
+      "loss": 3.4099,
+      "step": 1265
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 9.995767366720518e-05,
+      "loss": 3.4347,
+      "step": 1266
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 9.995751211631664e-05,
+      "loss": 3.5066,
+      "step": 1267
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 9.99573505654281e-05,
+      "loss": 3.5359,
+      "step": 1268
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 9.99571890145396e-05,
+      "loss": 6.298,
+      "step": 1269
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 9.995702746365106e-05,
+      "loss": 3.9315,
+      "step": 1270
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 9.995686591276252e-05,
+      "loss": 6.4339,
+      "step": 1271
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 9.995670436187399e-05,
+      "loss": 3.504,
+      "step": 1272
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 9.995654281098547e-05,
+      "loss": 3.4667,
+      "step": 1273
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 9.995638126009693e-05,
+      "loss": 3.4291,
+      "step": 1274
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 9.99562197092084e-05,
+      "loss": 3.4167,
+      "step": 1275
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 9.995605815831988e-05,
+      "loss": 3.357,
+      "step": 1276
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 9.995589660743134e-05,
+      "loss": 3.4011,
+      "step": 1277
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 9.995573505654282e-05,
+      "loss": 3.3918,
+      "step": 1278
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 9.995557350565429e-05,
+      "loss": 3.3848,
+      "step": 1279
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 9.995541195476575e-05,
+      "loss": 3.3736,
+      "step": 1280
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 9.995525040387723e-05,
+      "loss": 3.3983,
+      "step": 1281
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 9.99550888529887e-05,
+      "loss": 3.4085,
+      "step": 1282
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 9.995492730210016e-05,
+      "loss": 3.3818,
+      "step": 1283
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 9.995476575121164e-05,
+      "loss": 3.3738,
+      "step": 1284
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 9.995460420032311e-05,
+      "loss": 3.4054,
+      "step": 1285
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 9.995444264943458e-05,
+      "loss": 3.3794,
+      "step": 1286
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 9.995428109854604e-05,
+      "loss": 3.421,
+      "step": 1287
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 9.995411954765753e-05,
+      "loss": 3.3891,
+      "step": 1288
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 9.995395799676899e-05,
+      "loss": 3.4039,
+      "step": 1289
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 9.995379644588045e-05,
+      "loss": 3.4149,
+      "step": 1290
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 9.995363489499194e-05,
+      "loss": 3.4364,
+      "step": 1291
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 9.99534733441034e-05,
+      "loss": 3.4565,
+      "step": 1292
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 9.995331179321486e-05,
+      "loss": 3.5349,
+      "step": 1293
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 9.995315024232633e-05,
+      "loss": 3.6394,
+      "step": 1294
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 9.995298869143781e-05,
+      "loss": 4.5086,
+      "step": 1295
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 9.995282714054927e-05,
+      "loss": 5.3028,
+      "step": 1296
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 9.995266558966075e-05,
+      "loss": 3.479,
+      "step": 1297
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 9.995250403877222e-05,
+      "loss": 3.4439,
+      "step": 1298
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 9.995234248788368e-05,
+      "loss": 3.432,
+      "step": 1299
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 9.995218093699516e-05,
+      "loss": 3.4205,
+      "step": 1300
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 9.995201938610663e-05,
+      "loss": 3.4167,
+      "step": 1301
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 9.99518578352181e-05,
+      "loss": 3.4003,
+      "step": 1302
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 9.995169628432957e-05,
+      "loss": 3.381,
+      "step": 1303
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 9.995153473344104e-05,
+      "loss": 3.3834,
+      "step": 1304
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 9.99513731825525e-05,
+      "loss": 3.3563,
+      "step": 1305
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 9.995121163166398e-05,
+      "loss": 3.4086,
+      "step": 1306
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 9.995105008077544e-05,
+      "loss": 3.398,
+      "step": 1307
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 9.995088852988692e-05,
+      "loss": 3.3857,
+      "step": 1308
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 9.995072697899839e-05,
+      "loss": 3.3816,
+      "step": 1309
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 9.995056542810985e-05,
+      "loss": 3.4095,
+      "step": 1310
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 9.995040387722133e-05,
+      "loss": 3.4172,
+      "step": 1311
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 9.99502423263328e-05,
+      "loss": 3.4108,
+      "step": 1312
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 9.995008077544427e-05,
+      "loss": 3.3943,
+      "step": 1313
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 9.994991922455574e-05,
+      "loss": 3.4235,
+      "step": 1314
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 9.994975767366722e-05,
+      "loss": 3.4258,
+      "step": 1315
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 9.994959612277868e-05,
+      "loss": 3.4619,
+      "step": 1316
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 9.994943457189015e-05,
+      "loss": 3.5207,
+      "step": 1317
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 9.994927302100163e-05,
+      "loss": 3.4845,
+      "step": 1318
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 9.994911147011309e-05,
+      "loss": 3.6282,
+      "step": 1319
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 9.994894991922456e-05,
+      "loss": 3.7106,
+      "step": 1320
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 9.994878836833604e-05,
+      "loss": 4.9415,
+      "step": 1321
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 9.99486268174475e-05,
+      "loss": 3.4682,
+      "step": 1322
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 9.994846526655896e-05,
+      "loss": 3.4756,
+      "step": 1323
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 9.994830371567044e-05,
+      "loss": 3.4589,
+      "step": 1324
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 9.994814216478191e-05,
+      "loss": 3.3998,
+      "step": 1325
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 9.994798061389337e-05,
+      "loss": 3.3959,
+      "step": 1326
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 9.994781906300485e-05,
+      "loss": 3.3914,
+      "step": 1327
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 9.994765751211632e-05,
+      "loss": 3.4033,
+      "step": 1328
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 9.994749596122779e-05,
+      "loss": 3.3697,
+      "step": 1329
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 9.994733441033926e-05,
+      "loss": 3.3652,
+      "step": 1330
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 9.994717285945074e-05,
+      "loss": 3.3865,
+      "step": 1331
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 9.99470113085622e-05,
+      "loss": 3.3778,
+      "step": 1332
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 9.994684975767367e-05,
+      "loss": 3.3662,
+      "step": 1333
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 9.994668820678515e-05,
+      "loss": 3.3863,
+      "step": 1334
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 9.994652665589661e-05,
+      "loss": 3.3942,
+      "step": 1335
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 9.994636510500808e-05,
+      "loss": 3.4118,
+      "step": 1336
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 9.994620355411956e-05,
+      "loss": 3.4134,
+      "step": 1337
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 9.994604200323102e-05,
+      "loss": 3.3981,
+      "step": 1338
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 9.994588045234248e-05,
+      "loss": 3.4061,
+      "step": 1339
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 9.994571890145397e-05,
+      "loss": 3.425,
+      "step": 1340
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 9.994555735056543e-05,
+      "loss": 3.4379,
+      "step": 1341
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 9.99453957996769e-05,
+      "loss": 3.4427,
+      "step": 1342
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 9.994523424878838e-05,
+      "loss": 3.5137,
+      "step": 1343
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 9.994507269789984e-05,
+      "loss": 3.5183,
+      "step": 1344
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 9.99449111470113e-05,
+      "loss": 3.7058,
+      "step": 1345
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 9.99447495961228e-05,
+      "loss": 5.7257,
+      "step": 1346
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 9.994458804523426e-05,
+      "loss": 3.4616,
+      "step": 1347
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 9.994442649434572e-05,
+      "loss": 3.448,
+      "step": 1348
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 9.994426494345719e-05,
+      "loss": 3.5266,
+      "step": 1349
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 9.994410339256867e-05,
+      "loss": 3.4218,
+      "step": 1350
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 9.994394184168013e-05,
+      "loss": 3.3676,
+      "step": 1351
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 9.99437802907916e-05,
+      "loss": 3.3902,
+      "step": 1352
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 9.994361873990308e-05,
+      "loss": 3.3808,
+      "step": 1353
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 9.994345718901454e-05,
+      "loss": 3.3777,
+      "step": 1354
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 9.994329563812602e-05,
+      "loss": 3.3983,
+      "step": 1355
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 9.994313408723749e-05,
+      "loss": 3.3726,
+      "step": 1356
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 9.994297253634895e-05,
+      "loss": 3.3692,
+      "step": 1357
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 9.994281098546043e-05,
+      "loss": 3.4006,
+      "step": 1358
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 9.99426494345719e-05,
+      "loss": 3.3811,
+      "step": 1359
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 9.994248788368336e-05,
+      "loss": 3.4083,
+      "step": 1360
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 9.994232633279484e-05,
+      "loss": 3.3727,
+      "step": 1361
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 9.99421647819063e-05,
+      "loss": 3.4117,
+      "step": 1362
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 9.994200323101777e-05,
+      "loss": 3.4982,
+      "step": 1363
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 9.994184168012925e-05,
+      "loss": 3.4018,
+      "step": 1364
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 9.994168012924071e-05,
+      "loss": 3.4772,
+      "step": 1365
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 9.994151857835219e-05,
+      "loss": 3.4558,
+      "step": 1366
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 9.994135702746366e-05,
+      "loss": 3.4814,
+      "step": 1367
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 9.994119547657512e-05,
+      "loss": 3.5135,
+      "step": 1368
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 9.99410339256866e-05,
+      "loss": 3.9354,
+      "step": 1369
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 9.994087237479807e-05,
+      "loss": 4.8459,
+      "step": 1370
+    },
+    {
+      "epoch": 11.0,
+      "eval_loss": 5.171015739440918,
+      "eval_runtime": 38.3538,
+      "eval_samples_per_second": 21.927,
+      "eval_steps_per_second": 0.704,
+      "eval_wer": 1.0,
+      "step": 1370
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 9.994071082390953e-05,
+      "loss": 3.5137,
+      "step": 1371
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 9.994054927302101e-05,
+      "loss": 3.5314,
+      "step": 1372
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 9.994038772213247e-05,
+      "loss": 3.431,
+      "step": 1373
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 9.994022617124395e-05,
+      "loss": 3.4569,
+      "step": 1374
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 9.994006462035541e-05,
+      "loss": 3.4215,
+      "step": 1375
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 9.993990306946688e-05,
+      "loss": 3.4102,
+      "step": 1376
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 9.993974151857836e-05,
+      "loss": 3.3799,
+      "step": 1377
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 9.993957996768982e-05,
+      "loss": 3.3668,
+      "step": 1378
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 9.99394184168013e-05,
+      "loss": 3.3965,
+      "step": 1379
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 9.993925686591277e-05,
+      "loss": 3.3836,
+      "step": 1380
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 9.993909531502423e-05,
+      "loss": 3.3488,
+      "step": 1381
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 9.99389337641357e-05,
+      "loss": 3.3811,
+      "step": 1382
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 9.993877221324718e-05,
+      "loss": 3.3805,
+      "step": 1383
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 9.993861066235864e-05,
+      "loss": 3.4057,
+      "step": 1384
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 9.993844911147012e-05,
+      "loss": 3.4224,
+      "step": 1385
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 9.993828756058159e-05,
+      "loss": 3.3963,
+      "step": 1386
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 9.993812600969305e-05,
+      "loss": 3.3978,
+      "step": 1387
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 9.993796445880453e-05,
+      "loss": 3.4247,
+      "step": 1388
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 9.9937802907916e-05,
+      "loss": 3.4483,
+      "step": 1389
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 9.993764135702747e-05,
+      "loss": 3.5112,
+      "step": 1390
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 9.993747980613893e-05,
+      "loss": 3.4816,
+      "step": 1391
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 9.993731825525042e-05,
+      "loss": 3.5297,
+      "step": 1392
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 9.993715670436188e-05,
+      "loss": 4.1088,
+      "step": 1393
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 9.993699515347334e-05,
+      "loss": 3.9624,
+      "step": 1394
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 9.993683360258483e-05,
+      "loss": 5.6379,
+      "step": 1395
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 9.993667205169629e-05,
+      "loss": 3.4408,
+      "step": 1396
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 9.993651050080775e-05,
+      "loss": 3.4602,
+      "step": 1397
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 9.993634894991924e-05,
+      "loss": 3.4087,
+      "step": 1398
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 9.99361873990307e-05,
+      "loss": 3.4044,
+      "step": 1399
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 9.993602584814216e-05,
+      "loss": 3.3967,
+      "step": 1400
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 9.993586429725365e-05,
+      "loss": 3.3841,
+      "step": 1401
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 9.993570274636511e-05,
+      "loss": 3.3796,
+      "step": 1402
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 9.993554119547657e-05,
+      "loss": 3.3893,
+      "step": 1403
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 9.993537964458805e-05,
+      "loss": 3.3957,
+      "step": 1404
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 9.993521809369952e-05,
+      "loss": 3.385,
+      "step": 1405
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 9.993505654281099e-05,
+      "loss": 3.3527,
+      "step": 1406
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 9.993489499192246e-05,
+      "loss": 3.3839,
+      "step": 1407
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 9.993473344103394e-05,
+      "loss": 3.3729,
+      "step": 1408
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 9.99345718901454e-05,
+      "loss": 3.397,
+      "step": 1409
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 9.993441033925687e-05,
+      "loss": 3.4039,
+      "step": 1410
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 9.993424878836835e-05,
+      "loss": 3.3957,
+      "step": 1411
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 9.993408723747981e-05,
+      "loss": 3.415,
+      "step": 1412
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 9.993392568659128e-05,
+      "loss": 3.4726,
+      "step": 1413
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 9.993376413570276e-05,
+      "loss": 3.4426,
+      "step": 1414
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 9.993360258481422e-05,
+      "loss": 3.4845,
+      "step": 1415
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 9.99334410339257e-05,
+      "loss": 3.5886,
+      "step": 1416
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 9.993327948303716e-05,
+      "loss": 3.6212,
+      "step": 1417
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 9.993311793214863e-05,
+      "loss": 3.6058,
+      "step": 1418
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 9.993295638126011e-05,
+      "loss": 4.2419,
+      "step": 1419
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 9.993279483037157e-05,
+      "loss": 5.4044,
+      "step": 1420
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 9.993263327948304e-05,
+      "loss": 3.4599,
+      "step": 1421
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 9.993247172859452e-05,
+      "loss": 3.4317,
+      "step": 1422
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 9.993231017770598e-05,
+      "loss": 3.4455,
+      "step": 1423
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 9.993214862681746e-05,
+      "loss": 3.4077,
+      "step": 1424
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 9.993198707592892e-05,
+      "loss": 3.422,
+      "step": 1425
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 9.993182552504039e-05,
+      "loss": 3.3735,
+      "step": 1426
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 9.993166397415187e-05,
+      "loss": 3.3674,
+      "step": 1427
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 9.993150242326333e-05,
+      "loss": 3.3866,
+      "step": 1428
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 9.99313408723748e-05,
+      "loss": 3.4026,
+      "step": 1429
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 9.993117932148626e-05,
+      "loss": 3.3939,
+      "step": 1430
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 9.993101777059774e-05,
+      "loss": 3.3895,
+      "step": 1431
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 9.993085621970922e-05,
+      "loss": 3.3964,
+      "step": 1432
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 9.993069466882068e-05,
+      "loss": 3.4241,
+      "step": 1433
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 9.993053311793215e-05,
+      "loss": 3.3954,
+      "step": 1434
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 9.993037156704363e-05,
+      "loss": 3.3885,
+      "step": 1435
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 9.993021001615509e-05,
+      "loss": 3.3952,
+      "step": 1436
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 9.993004846526656e-05,
+      "loss": 3.4125,
+      "step": 1437
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 9.992988691437804e-05,
+      "loss": 3.3884,
+      "step": 1438
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 9.99297253634895e-05,
+      "loss": 3.4362,
+      "step": 1439
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 9.992956381260097e-05,
+      "loss": 3.4431,
+      "step": 1440
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 9.992940226171245e-05,
+      "loss": 3.4637,
+      "step": 1441
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 9.992924071082391e-05,
+      "loss": 3.5293,
+      "step": 1442
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 9.992907915993537e-05,
+      "loss": 3.6574,
+      "step": 1443
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 9.992891760904686e-05,
+      "loss": 3.9874,
+      "step": 1444
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 9.992875605815832e-05,
+      "loss": 5.7415,
+      "step": 1445
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 9.992859450726978e-05,
+      "loss": 3.4532,
+      "step": 1446
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 9.992843295638127e-05,
+      "loss": 3.4616,
+      "step": 1447
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 9.992827140549273e-05,
+      "loss": 3.4298,
+      "step": 1448
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 9.99281098546042e-05,
+      "loss": 3.4565,
+      "step": 1449
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 9.992794830371568e-05,
+      "loss": 3.3999,
+      "step": 1450
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 9.992778675282715e-05,
+      "loss": 3.3847,
+      "step": 1451
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 9.992762520193861e-05,
+      "loss": 3.3974,
+      "step": 1452
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 9.99274636510501e-05,
+      "loss": 3.3501,
+      "step": 1453
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 9.992730210016156e-05,
+      "loss": 3.3685,
+      "step": 1454
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 9.992714054927302e-05,
+      "loss": 3.3836,
+      "step": 1455
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 9.99269789983845e-05,
+      "loss": 3.3671,
+      "step": 1456
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 9.992681744749597e-05,
+      "loss": 3.4176,
+      "step": 1457
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 9.992665589660743e-05,
+      "loss": 3.3777,
+      "step": 1458
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 9.99264943457189e-05,
+      "loss": 3.422,
+      "step": 1459
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 9.992633279483038e-05,
+      "loss": 3.4093,
+      "step": 1460
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 9.992617124394184e-05,
+      "loss": 3.3971,
+      "step": 1461
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 9.992600969305332e-05,
+      "loss": 3.4107,
+      "step": 1462
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 9.992584814216479e-05,
+      "loss": 3.5351,
+      "step": 1463
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 9.992568659127625e-05,
+      "loss": 3.4201,
+      "step": 1464
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 9.992552504038773e-05,
+      "loss": 3.5549,
+      "step": 1465
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 9.99253634894992e-05,
+      "loss": 3.4975,
+      "step": 1466
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 9.992520193861067e-05,
+      "loss": 3.5676,
+      "step": 1467
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 9.992504038772214e-05,
+      "loss": 3.6648,
+      "step": 1468
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 9.99248788368336e-05,
+      "loss": 4.6306,
+      "step": 1469
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 9.992471728594508e-05,
+      "loss": 5.5506,
+      "step": 1470
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 9.992455573505655e-05,
+      "loss": 3.4904,
+      "step": 1471
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 9.992439418416801e-05,
+      "loss": 3.484,
+      "step": 1472
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 9.992423263327949e-05,
+      "loss": 3.4449,
+      "step": 1473
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 9.992407108239096e-05,
+      "loss": 3.4263,
+      "step": 1474
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 9.992390953150243e-05,
+      "loss": 3.3899,
+      "step": 1475
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 9.99237479806139e-05,
+      "loss": 3.3852,
+      "step": 1476
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 9.992358642972536e-05,
+      "loss": 3.3755,
+      "step": 1477
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 9.992342487883684e-05,
+      "loss": 3.3835,
+      "step": 1478
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 9.992326332794831e-05,
+      "loss": 3.4005,
+      "step": 1479
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 9.992310177705977e-05,
+      "loss": 3.3833,
+      "step": 1480
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 9.992294022617125e-05,
+      "loss": 3.3648,
+      "step": 1481
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 9.992277867528271e-05,
+      "loss": 3.3895,
+      "step": 1482
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 9.992261712439419e-05,
+      "loss": 3.3854,
+      "step": 1483
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 9.992245557350566e-05,
+      "loss": 3.3995,
+      "step": 1484
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 9.992229402261712e-05,
+      "loss": 3.4631,
+      "step": 1485
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 9.99221324717286e-05,
+      "loss": 3.4043,
+      "step": 1486
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 9.992197092084007e-05,
+      "loss": 3.4356,
+      "step": 1487
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 9.992180936995153e-05,
+      "loss": 3.4208,
+      "step": 1488
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 9.992164781906301e-05,
+      "loss": 3.4519,
+      "step": 1489
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 9.992148626817448e-05,
+      "loss": 3.5097,
+      "step": 1490
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 9.992132471728595e-05,
+      "loss": 3.5482,
+      "step": 1491
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 9.992116316639742e-05,
+      "loss": 3.6338,
+      "step": 1492
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 9.99210016155089e-05,
+      "loss": 4.5229,
+      "step": 1493
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 9.992084006462036e-05,
+      "loss": 5.396,
+      "step": 1494
+    },
+    {
+      "epoch": 12.0,
+      "eval_loss": 5.10784912109375,
+      "eval_runtime": 37.5728,
+      "eval_samples_per_second": 22.383,
+      "eval_steps_per_second": 0.719,
+      "eval_wer": 1.0,
+      "step": 1494
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 9.992067851373183e-05,
+      "loss": 3.4834,
+      "step": 1495
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 9.992051696284331e-05,
+      "loss": 3.4458,
+      "step": 1496
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 9.992035541195477e-05,
+      "loss": 3.4417,
+      "step": 1497
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 9.992019386106623e-05,
+      "loss": 3.4098,
+      "step": 1498
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 9.992003231017772e-05,
+      "loss": 3.4088,
+      "step": 1499
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 9.991987075928918e-05,
+      "loss": 3.4095,
+      "step": 1500
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 9.991970920840064e-05,
+      "loss": 3.385,
+      "step": 1501
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 9.991954765751213e-05,
+      "loss": 3.3835,
+      "step": 1502
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 9.991938610662359e-05,
+      "loss": 3.3762,
+      "step": 1503
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 9.991922455573505e-05,
+      "loss": 3.4102,
+      "step": 1504
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 9.991906300484654e-05,
+      "loss": 3.3938,
+      "step": 1505
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 9.9918901453958e-05,
+      "loss": 3.3696,
+      "step": 1506
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 9.991873990306946e-05,
+      "loss": 3.3632,
+      "step": 1507
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 9.991857835218095e-05,
+      "loss": 3.3968,
+      "step": 1508
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 9.991841680129242e-05,
+      "loss": 3.4201,
+      "step": 1509
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 9.991825525040388e-05,
+      "loss": 3.3984,
+      "step": 1510
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 9.991809369951535e-05,
+      "loss": 3.412,
+      "step": 1511
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 9.991793214862683e-05,
+      "loss": 3.3836,
+      "step": 1512
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 9.991777059773829e-05,
+      "loss": 3.4056,
+      "step": 1513
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 9.991760904684976e-05,
+      "loss": 3.5592,
+      "step": 1514
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 9.991744749596124e-05,
+      "loss": 3.4849,
+      "step": 1515
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 9.99172859450727e-05,
+      "loss": 3.7149,
+      "step": 1516
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 9.991712439418417e-05,
+      "loss": 3.7231,
+      "step": 1517
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 9.991696284329565e-05,
+      "loss": 4.0842,
+      "step": 1518
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 9.991680129240711e-05,
+      "loss": 5.8199,
+      "step": 1519
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 9.991663974151859e-05,
+      "loss": 3.4434,
+      "step": 1520
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 9.991647819063006e-05,
+      "loss": 3.4583,
+      "step": 1521
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 9.991631663974152e-05,
+      "loss": 3.4464,
+      "step": 1522
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 9.9916155088853e-05,
+      "loss": 3.4643,
+      "step": 1523
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 9.991599353796446e-05,
+      "loss": 3.4334,
+      "step": 1524
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 9.991583198707593e-05,
+      "loss": 3.3907,
+      "step": 1525
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 9.991567043618741e-05,
+      "loss": 3.3634,
+      "step": 1526
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 9.991550888529887e-05,
+      "loss": 3.3636,
+      "step": 1527
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 9.991534733441035e-05,
+      "loss": 3.3834,
+      "step": 1528
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 9.991518578352181e-05,
+      "loss": 3.3826,
+      "step": 1529
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 9.991502423263328e-05,
+      "loss": 3.3666,
+      "step": 1530
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 9.991486268174476e-05,
+      "loss": 3.3856,
+      "step": 1531
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 9.991470113085622e-05,
+      "loss": 3.3887,
+      "step": 1532
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 9.99145395799677e-05,
+      "loss": 3.3702,
+      "step": 1533
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 9.991437802907917e-05,
+      "loss": 3.383,
+      "step": 1534
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 9.991421647819063e-05,
+      "loss": 3.4071,
+      "step": 1535
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 9.99140549273021e-05,
+      "loss": 3.4003,
+      "step": 1536
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 9.991389337641357e-05,
+      "loss": 3.4011,
+      "step": 1537
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 9.991373182552504e-05,
+      "loss": 3.4826,
+      "step": 1538
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 9.991357027463652e-05,
+      "loss": 3.4672,
+      "step": 1539
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 9.991340872374798e-05,
+      "loss": 3.492,
+      "step": 1540
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 9.991324717285945e-05,
+      "loss": 3.5799,
+      "step": 1541
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 9.991308562197093e-05,
+      "loss": 3.6138,
+      "step": 1542
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 9.991292407108239e-05,
+      "loss": 4.4718,
+      "step": 1543
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 9.991276252019387e-05,
+      "loss": 6.0988,
+      "step": 1544
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 9.991260096930534e-05,
+      "loss": 3.4806,
+      "step": 1545
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 9.99124394184168e-05,
+      "loss": 3.4431,
+      "step": 1546
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 9.991227786752828e-05,
+      "loss": 3.4469,
+      "step": 1547
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 9.991211631663975e-05,
+      "loss": 3.4412,
+      "step": 1548
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 9.991195476575121e-05,
+      "loss": 3.3916,
+      "step": 1549
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 9.991179321486268e-05,
+      "loss": 3.4102,
+      "step": 1550
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 9.991163166397416e-05,
+      "loss": 3.3908,
+      "step": 1551
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 9.991147011308563e-05,
+      "loss": 3.3792,
+      "step": 1552
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 9.991130856219709e-05,
+      "loss": 3.3897,
+      "step": 1553
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 9.991114701130858e-05,
+      "loss": 3.3737,
+      "step": 1554
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 9.991098546042004e-05,
+      "loss": 3.3835,
+      "step": 1555
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 9.99108239095315e-05,
+      "loss": 3.3768,
+      "step": 1556
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 9.991066235864299e-05,
+      "loss": 3.384,
+      "step": 1557
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 9.991050080775445e-05,
+      "loss": 3.3771,
+      "step": 1558
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 9.991033925686591e-05,
+      "loss": 3.4017,
+      "step": 1559
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 9.99101777059774e-05,
+      "loss": 3.4108,
+      "step": 1560
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 9.991001615508886e-05,
+      "loss": 3.3812,
+      "step": 1561
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 9.990985460420032e-05,
+      "loss": 3.4319,
+      "step": 1562
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 9.99096930533118e-05,
+      "loss": 3.4447,
+      "step": 1563
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 9.990953150242327e-05,
+      "loss": 3.4511,
+      "step": 1564
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 9.990936995153473e-05,
+      "loss": 3.424,
+      "step": 1565
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 9.990920840064621e-05,
+      "loss": 3.5205,
+      "step": 1566
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 9.990904684975768e-05,
+      "loss": 3.5745,
+      "step": 1567
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 9.990888529886915e-05,
+      "loss": 3.833,
+      "step": 1568
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 9.990872374798062e-05,
+      "loss": 5.6812,
+      "step": 1569
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 9.99085621970921e-05,
+      "loss": 3.5017,
+      "step": 1570
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 9.990840064620356e-05,
+      "loss": 3.4314,
+      "step": 1571
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 9.990823909531503e-05,
+      "loss": 3.4185,
+      "step": 1572
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 9.990807754442651e-05,
+      "loss": 3.4068,
+      "step": 1573
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 9.990791599353797e-05,
+      "loss": 3.3819,
+      "step": 1574
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 9.990775444264944e-05,
+      "loss": 3.3964,
+      "step": 1575
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 9.990759289176092e-05,
+      "loss": 3.378,
+      "step": 1576
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 9.990743134087238e-05,
+      "loss": 3.3813,
+      "step": 1577
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 9.990726978998384e-05,
+      "loss": 3.4017,
+      "step": 1578
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 9.990710823909532e-05,
+      "loss": 3.3706,
+      "step": 1579
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 9.990694668820679e-05,
+      "loss": 3.3875,
+      "step": 1580
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 9.990678513731825e-05,
+      "loss": 3.3951,
+      "step": 1581
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 9.990662358642973e-05,
+      "loss": 3.398,
+      "step": 1582
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 9.99064620355412e-05,
+      "loss": 3.3811,
+      "step": 1583
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 9.990630048465266e-05,
+      "loss": 3.3951,
+      "step": 1584
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 9.990613893376414e-05,
+      "loss": 3.4046,
+      "step": 1585
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 9.990597738287561e-05,
+      "loss": 3.4109,
+      "step": 1586
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 9.990581583198708e-05,
+      "loss": 3.4118,
+      "step": 1587
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 9.990565428109855e-05,
+      "loss": 3.4194,
+      "step": 1588
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 9.990549273021003e-05,
+      "loss": 3.491,
+      "step": 1589
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 9.990533117932149e-05,
+      "loss": 3.4697,
+      "step": 1590
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 9.990516962843296e-05,
+      "loss": 3.4914,
+      "step": 1591
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 9.990500807754442e-05,
+      "loss": 3.7132,
+      "step": 1592
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 9.99048465266559e-05,
+      "loss": 4.2858,
+      "step": 1593
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 9.990468497576737e-05,
+      "loss": 5.9496,
+      "step": 1594
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 9.990452342487884e-05,
+      "loss": 3.4364,
+      "step": 1595
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 9.990436187399031e-05,
+      "loss": 3.474,
+      "step": 1596
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 9.990420032310179e-05,
+      "loss": 3.4272,
+      "step": 1597
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 9.990403877221325e-05,
+      "loss": 3.4,
+      "step": 1598
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 9.990387722132472e-05,
+      "loss": 3.3933,
+      "step": 1599
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 9.99037156704362e-05,
+      "loss": 3.3931,
+      "step": 1600
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 9.990355411954766e-05,
+      "loss": 3.3552,
+      "step": 1601
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 9.990339256865913e-05,
+      "loss": 3.3769,
+      "step": 1602
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 9.990323101777061e-05,
+      "loss": 3.3784,
+      "step": 1603
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 9.990306946688207e-05,
+      "loss": 3.3671,
+      "step": 1604
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 9.990290791599353e-05,
+      "loss": 3.3704,
+      "step": 1605
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 9.990274636510502e-05,
+      "loss": 3.3845,
+      "step": 1606
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 9.990258481421648e-05,
+      "loss": 3.4382,
+      "step": 1607
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 9.990242326332794e-05,
+      "loss": 3.4034,
+      "step": 1608
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 9.990226171243943e-05,
+      "loss": 3.3932,
+      "step": 1609
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 9.99021001615509e-05,
+      "loss": 3.3965,
+      "step": 1610
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 9.990193861066236e-05,
+      "loss": 3.3803,
+      "step": 1611
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 9.990177705977384e-05,
+      "loss": 3.4853,
+      "step": 1612
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 9.99016155088853e-05,
+      "loss": 3.4269,
+      "step": 1613
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 9.990145395799677e-05,
+      "loss": 3.4679,
+      "step": 1614
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 9.990129240710824e-05,
+      "loss": 3.5142,
+      "step": 1615
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 9.990113085621972e-05,
+      "loss": 3.7542,
+      "step": 1616
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 9.990096930533118e-05,
+      "loss": 3.7533,
+      "step": 1617
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 9.990080775444265e-05,
+      "loss": 4.4184,
+      "step": 1618
+    },
+    {
+      "epoch": 13.0,
+      "eval_loss": 5.155089378356934,
+      "eval_runtime": 38.0295,
+      "eval_samples_per_second": 22.114,
+      "eval_steps_per_second": 0.71,
+      "eval_wer": 1.0,
+      "step": 1618
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 9.990064620355413e-05,
+      "loss": 3.4642,
+      "step": 1619
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 9.990048465266559e-05,
+      "loss": 3.5058,
+      "step": 1620
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 9.990032310177707e-05,
+      "loss": 3.4687,
+      "step": 1621
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 9.990016155088854e-05,
+      "loss": 3.4321,
+      "step": 1622
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 9.99e-05,
+      "loss": 3.4402,
+      "step": 1623
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 9.989983844911148e-05,
+      "loss": 3.3815,
+      "step": 1624
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 9.989967689822295e-05,
+      "loss": 3.377,
+      "step": 1625
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 9.989951534733441e-05,
+      "loss": 3.386,
+      "step": 1626
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 9.989935379644589e-05,
+      "loss": 3.3931,
+      "step": 1627
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 9.989919224555736e-05,
+      "loss": 3.384,
+      "step": 1628
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 9.989903069466883e-05,
+      "loss": 3.3896,
+      "step": 1629
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 9.989886914378029e-05,
+      "loss": 3.3926,
+      "step": 1630
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 9.989870759289176e-05,
+      "loss": 3.395,
+      "step": 1631
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 9.989854604200324e-05,
+      "loss": 3.3855,
+      "step": 1632
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 9.98983844911147e-05,
+      "loss": 3.3853,
+      "step": 1633
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 9.989822294022617e-05,
+      "loss": 3.3876,
+      "step": 1634
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 9.989806138933765e-05,
+      "loss": 3.4022,
+      "step": 1635
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 9.989789983844911e-05,
+      "loss": 3.4191,
+      "step": 1636
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 9.989773828756059e-05,
+      "loss": 3.578,
+      "step": 1637
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 9.989757673667206e-05,
+      "loss": 3.4506,
+      "step": 1638
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 9.989741518578352e-05,
+      "loss": 3.5101,
+      "step": 1639
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 9.9897253634895e-05,
+      "loss": 3.5199,
+      "step": 1640
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 9.989709208400647e-05,
+      "loss": 3.7695,
+      "step": 1641
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 9.989693053311793e-05,
+      "loss": 4.6444,
+      "step": 1642
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 9.989676898222941e-05,
+      "loss": 6.0861,
+      "step": 1643
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 9.989660743134087e-05,
+      "loss": 3.4851,
+      "step": 1644
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 9.989644588045235e-05,
+      "loss": 3.4779,
+      "step": 1645
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 9.989628432956382e-05,
+      "loss": 3.4148,
+      "step": 1646
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 9.989612277867528e-05,
+      "loss": 3.3925,
+      "step": 1647
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 9.989596122778676e-05,
+      "loss": 3.4136,
+      "step": 1648
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 9.989579967689823e-05,
+      "loss": 3.4054,
+      "step": 1649
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 9.98956381260097e-05,
+      "loss": 3.3769,
+      "step": 1650
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 9.989547657512117e-05,
+      "loss": 3.3624,
+      "step": 1651
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 9.989531502423264e-05,
+      "loss": 3.3768,
+      "step": 1652
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 9.98951534733441e-05,
+      "loss": 3.3884,
+      "step": 1653
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 9.989499192245558e-05,
+      "loss": 3.3958,
+      "step": 1654
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 9.989483037156706e-05,
+      "loss": 3.3857,
+      "step": 1655
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 9.989466882067852e-05,
+      "loss": 3.3814,
+      "step": 1656
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 9.989450726978998e-05,
+      "loss": 3.3995,
+      "step": 1657
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 9.989434571890147e-05,
+      "loss": 3.3907,
+      "step": 1658
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 9.989418416801293e-05,
+      "loss": 3.4116,
+      "step": 1659
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 9.989402261712439e-05,
+      "loss": 3.4086,
+      "step": 1660
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 9.989386106623588e-05,
+      "loss": 3.4176,
+      "step": 1661
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 9.989369951534734e-05,
+      "loss": 3.4156,
+      "step": 1662
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 9.98935379644588e-05,
+      "loss": 3.4867,
+      "step": 1663
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 9.989337641357029e-05,
+      "loss": 3.4673,
+      "step": 1664
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 9.989321486268175e-05,
+      "loss": 3.5474,
+      "step": 1665
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 9.989305331179321e-05,
+      "loss": 3.6242,
+      "step": 1666
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 9.989289176090469e-05,
+      "loss": 4.1575,
+      "step": 1667
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 9.989273021001616e-05,
+      "loss": 5.3613,
+      "step": 1668
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 9.989256865912762e-05,
+      "loss": 3.4788,
+      "step": 1669
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 9.98924071082391e-05,
+      "loss": 3.4578,
+      "step": 1670
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 9.989224555735057e-05,
+      "loss": 3.4383,
+      "step": 1671
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 9.989208400646204e-05,
+      "loss": 3.4026,
+      "step": 1672
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 9.989192245557351e-05,
+      "loss": 3.402,
+      "step": 1673
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 9.989176090468499e-05,
+      "loss": 3.411,
+      "step": 1674
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 9.989159935379645e-05,
+      "loss": 3.396,
+      "step": 1675
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 9.989143780290792e-05,
+      "loss": 3.3881,
+      "step": 1676
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 9.98912762520194e-05,
+      "loss": 3.3822,
+      "step": 1677
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 9.989111470113086e-05,
+      "loss": 3.3947,
+      "step": 1678
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 9.989095315024233e-05,
+      "loss": 3.3775,
+      "step": 1679
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 9.989079159935381e-05,
+      "loss": 3.3957,
+      "step": 1680
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 9.989063004846527e-05,
+      "loss": 3.3993,
+      "step": 1681
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 9.989046849757673e-05,
+      "loss": 3.3953,
+      "step": 1682
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 9.989030694668822e-05,
+      "loss": 3.4004,
+      "step": 1683
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 9.989014539579968e-05,
+      "loss": 3.3972,
+      "step": 1684
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 9.988998384491114e-05,
+      "loss": 3.4162,
+      "step": 1685
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 9.988982229402262e-05,
+      "loss": 3.4219,
+      "step": 1686
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 9.98896607431341e-05,
+      "loss": 3.4551,
+      "step": 1687
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 9.988949919224556e-05,
+      "loss": 3.4359,
+      "step": 1688
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 9.988933764135703e-05,
+      "loss": 3.4462,
+      "step": 1689
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 9.98891760904685e-05,
+      "loss": 3.5601,
+      "step": 1690
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 9.988901453957997e-05,
+      "loss": 3.6097,
+      "step": 1691
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 9.988885298869144e-05,
+      "loss": 4.3425,
+      "step": 1692
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 9.988869143780292e-05,
+      "loss": 5.3673,
+      "step": 1693
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 9.988852988691438e-05,
+      "loss": 3.479,
+      "step": 1694
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 9.988836833602585e-05,
+      "loss": 3.4327,
+      "step": 1695
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 9.988820678513733e-05,
+      "loss": 3.4137,
+      "step": 1696
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 9.988804523424879e-05,
+      "loss": 3.4163,
+      "step": 1697
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 9.988788368336027e-05,
+      "loss": 3.3777,
+      "step": 1698
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 9.988772213247173e-05,
+      "loss": 3.3966,
+      "step": 1699
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 9.98875605815832e-05,
+      "loss": 3.4235,
+      "step": 1700
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 9.988739903069468e-05,
+      "loss": 3.3658,
+      "step": 1701
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 9.988723747980614e-05,
+      "loss": 3.3642,
+      "step": 1702
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 9.988707592891761e-05,
+      "loss": 3.3649,
+      "step": 1703
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 9.988691437802909e-05,
+      "loss": 3.3946,
+      "step": 1704
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 9.988675282714055e-05,
+      "loss": 3.38,
+      "step": 1705
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 9.988659127625203e-05,
+      "loss": 3.3848,
+      "step": 1706
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 9.98864297253635e-05,
+      "loss": 3.4155,
+      "step": 1707
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 9.988626817447496e-05,
+      "loss": 3.391,
+      "step": 1708
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 9.988610662358644e-05,
+      "loss": 3.4297,
+      "step": 1709
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 9.988594507269791e-05,
+      "loss": 3.4079,
+      "step": 1710
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 9.988578352180937e-05,
+      "loss": 3.4186,
+      "step": 1711
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 9.988562197092084e-05,
+      "loss": 3.4196,
+      "step": 1712
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 9.988546042003232e-05,
+      "loss": 3.4304,
+      "step": 1713
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 9.988529886914379e-05,
+      "loss": 3.4969,
+      "step": 1714
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 9.988513731825525e-05,
+      "loss": 3.6277,
+      "step": 1715
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 9.988497576736672e-05,
+      "loss": 3.6389,
+      "step": 1716
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 9.98848142164782e-05,
+      "loss": 4.4056,
+      "step": 1717
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 9.988465266558966e-05,
+      "loss": 5.1644,
+      "step": 1718
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 9.988449111470113e-05,
+      "loss": 3.4799,
+      "step": 1719
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 9.988432956381261e-05,
+      "loss": 3.4707,
+      "step": 1720
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 9.988416801292407e-05,
+      "loss": 3.4475,
+      "step": 1721
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 9.988400646203555e-05,
+      "loss": 3.394,
+      "step": 1722
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 9.988384491114702e-05,
+      "loss": 3.3855,
+      "step": 1723
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 9.988368336025848e-05,
+      "loss": 3.4094,
+      "step": 1724
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 9.988352180936996e-05,
+      "loss": 3.3969,
+      "step": 1725
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 9.988336025848143e-05,
+      "loss": 3.3866,
+      "step": 1726
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 9.98831987075929e-05,
+      "loss": 3.3871,
+      "step": 1727
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 9.988303715670437e-05,
+      "loss": 3.3671,
+      "step": 1728
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 9.988287560581584e-05,
+      "loss": 3.3756,
+      "step": 1729
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 9.98827140549273e-05,
+      "loss": 3.4044,
+      "step": 1730
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 9.988255250403878e-05,
+      "loss": 3.3589,
+      "step": 1731
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 9.988239095315026e-05,
+      "loss": 3.3827,
+      "step": 1732
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 9.988222940226172e-05,
+      "loss": 3.4257,
+      "step": 1733
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 9.988206785137318e-05,
+      "loss": 3.4071,
+      "step": 1734
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 9.988190630048467e-05,
+      "loss": 3.4039,
+      "step": 1735
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 9.988174474959613e-05,
+      "loss": 3.3947,
+      "step": 1736
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 9.988158319870759e-05,
+      "loss": 3.4234,
+      "step": 1737
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 9.988142164781906e-05,
+      "loss": 3.4375,
+      "step": 1738
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 9.988126009693054e-05,
+      "loss": 3.509,
+      "step": 1739
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 9.9881098546042e-05,
+      "loss": 3.4687,
+      "step": 1740
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 9.988093699515348e-05,
+      "loss": 3.4954,
+      "step": 1741
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 9.988077544426495e-05,
+      "loss": 4.0691,
+      "step": 1742
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 9.988061389337641e-05,
+      "loss": 4.8546,
+      "step": 1743
+    },
+    {
+      "epoch": 13.0,
+      "eval_loss": 4.945601463317871,
+      "eval_runtime": 37.8982,
+      "eval_samples_per_second": 22.191,
+      "eval_steps_per_second": 0.712,
+      "eval_wer": 1.0,
+      "step": 1743
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 9.988045234248789e-05,
+      "loss": 3.5314,
+      "step": 1744
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.988029079159936e-05,
+      "loss": 3.4513,
+      "step": 1745
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.988012924071082e-05,
+      "loss": 3.4456,
+      "step": 1746
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.98799676898223e-05,
+      "loss": 3.4378,
+      "step": 1747
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 9.987980613893377e-05,
+      "loss": 3.4041,
+      "step": 1748
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 9.987964458804524e-05,
+      "loss": 3.3798,
+      "step": 1749
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.987948303715671e-05,
+      "loss": 3.3872,
+      "step": 1750
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.987932148626819e-05,
+      "loss": 3.391,
+      "step": 1751
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 9.987915993537965e-05,
+      "loss": 3.3874,
+      "step": 1752
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 9.987899838449112e-05,
+      "loss": 3.3884,
+      "step": 1753
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 9.987883683360258e-05,
+      "loss": 3.3986,
+      "step": 1754
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.987867528271406e-05,
+      "loss": 3.4015,
+      "step": 1755
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.987851373182553e-05,
+      "loss": 3.3776,
+      "step": 1756
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 9.9878352180937e-05,
+      "loss": 3.3688,
+      "step": 1757
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 9.987819063004847e-05,
+      "loss": 3.4058,
+      "step": 1758
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.987802907915995e-05,
+      "loss": 3.4006,
+      "step": 1759
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.987786752827141e-05,
+      "loss": 3.4002,
+      "step": 1760
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.987770597738288e-05,
+      "loss": 3.4037,
+      "step": 1761
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 9.987754442649436e-05,
+      "loss": 3.4571,
+      "step": 1762
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 9.987738287560582e-05,
+      "loss": 3.4351,
+      "step": 1763
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 9.98772213247173e-05,
+      "loss": 3.443,
+      "step": 1764
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.987705977382877e-05,
+      "loss": 3.5821,
+      "step": 1765
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.987689822294023e-05,
+      "loss": 3.5351,
+      "step": 1766
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 9.987673667205169e-05,
+      "loss": 4.5021,
+      "step": 1767
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 9.987657512116317e-05,
+      "loss": 4.6276,
+      "step": 1768
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 9.987641357027464e-05,
+      "loss": 3.4463,
+      "step": 1769
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.98762520193861e-05,
+      "loss": 3.4625,
+      "step": 1770
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.987609046849758e-05,
+      "loss": 3.4441,
+      "step": 1771
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 9.987592891760905e-05,
+      "loss": 3.4192,
+      "step": 1772
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 9.987576736672052e-05,
+      "loss": 3.4043,
+      "step": 1773
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 9.987560581583199e-05,
+      "loss": 3.3956,
+      "step": 1774
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.987544426494347e-05,
+      "loss": 3.3834,
+      "step": 1775
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.987528271405493e-05,
+      "loss": 3.3904,
+      "step": 1776
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.98751211631664e-05,
+      "loss": 3.3554,
+      "step": 1777
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 9.987495961227788e-05,
+      "loss": 3.3979,
+      "step": 1778
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 9.987479806138934e-05,
+      "loss": 3.3884,
+      "step": 1779
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 9.987463651050081e-05,
+      "loss": 3.3834,
+      "step": 1780
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 9.987447495961229e-05,
+      "loss": 3.3888,
+      "step": 1781
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 9.987431340872375e-05,
+      "loss": 3.3863,
+      "step": 1782
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 9.987415185783521e-05,
+      "loss": 3.3743,
+      "step": 1783
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 9.98739903069467e-05,
+      "loss": 3.3611,
+      "step": 1784
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 9.987382875605816e-05,
+      "loss": 3.4306,
+      "step": 1785
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 9.987366720516962e-05,
+      "loss": 3.4041,
+      "step": 1786
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 9.987350565428111e-05,
+      "loss": 3.4268,
+      "step": 1787
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 9.987334410339257e-05,
+      "loss": 3.4377,
+      "step": 1788
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 9.987318255250404e-05,
+      "loss": 3.5122,
+      "step": 1789
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 9.987302100161552e-05,
+      "loss": 3.5678,
+      "step": 1790
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 9.987285945072699e-05,
+      "loss": 3.5956,
+      "step": 1791
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 9.987269789983845e-05,
+      "loss": 4.3408,
+      "step": 1792
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 9.987253634894992e-05,
+      "loss": 5.1214,
+      "step": 1793
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 9.98723747980614e-05,
+      "loss": 3.4711,
+      "step": 1794
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 9.987221324717286e-05,
+      "loss": 3.4781,
+      "step": 1795
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 9.987205169628433e-05,
+      "loss": 3.413,
+      "step": 1796
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 9.987189014539581e-05,
+      "loss": 3.4533,
+      "step": 1797
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 9.987172859450727e-05,
+      "loss": 3.3939,
+      "step": 1798
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 9.987156704361875e-05,
+      "loss": 3.3718,
+      "step": 1799
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 9.987140549273022e-05,
+      "loss": 3.3692,
+      "step": 1800
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 9.987124394184168e-05,
+      "loss": 3.3737,
+      "step": 1801
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 9.987108239095316e-05,
+      "loss": 3.3845,
+      "step": 1802
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 9.987092084006463e-05,
+      "loss": 3.3839,
+      "step": 1803
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 9.98707592891761e-05,
+      "loss": 3.3826,
+      "step": 1804
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 9.987059773828757e-05,
+      "loss": 3.3715,
+      "step": 1805
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 9.987043618739903e-05,
+      "loss": 3.4002,
+      "step": 1806
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 9.98702746365105e-05,
+      "loss": 3.3916,
+      "step": 1807
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 9.987011308562198e-05,
+      "loss": 3.3955,
+      "step": 1808
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 9.986995153473344e-05,
+      "loss": 3.3978,
+      "step": 1809
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 9.986978998384492e-05,
+      "loss": 3.406,
+      "step": 1810
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 9.986962843295639e-05,
+      "loss": 3.4348,
+      "step": 1811
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 9.986946688206785e-05,
+      "loss": 3.4252,
+      "step": 1812
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 9.986930533117933e-05,
+      "loss": 3.4158,
+      "step": 1813
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 9.98691437802908e-05,
+      "loss": 3.4417,
+      "step": 1814
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 9.986898222940226e-05,
+      "loss": 3.5018,
+      "step": 1815
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 9.986882067851374e-05,
+      "loss": 3.9239,
+      "step": 1816
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 9.986865912762521e-05,
+      "loss": 4.249,
+      "step": 1817
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 9.986849757673668e-05,
+      "loss": 5.8619,
+      "step": 1818
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 9.986833602584814e-05,
+      "loss": 3.4918,
+      "step": 1819
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 9.986817447495961e-05,
+      "loss": 3.4494,
+      "step": 1820
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 9.986801292407109e-05,
+      "loss": 3.5096,
+      "step": 1821
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 9.986785137318255e-05,
+      "loss": 3.4067,
+      "step": 1822
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 9.986768982229402e-05,
+      "loss": 3.4232,
+      "step": 1823
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 9.98675282714055e-05,
+      "loss": 3.3747,
+      "step": 1824
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 9.986736672051696e-05,
+      "loss": 3.3777,
+      "step": 1825
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 9.986720516962844e-05,
+      "loss": 3.3625,
+      "step": 1826
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 9.986704361873991e-05,
+      "loss": 3.3819,
+      "step": 1827
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 9.986688206785137e-05,
+      "loss": 3.368,
+      "step": 1828
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 9.986672051696285e-05,
+      "loss": 3.3753,
+      "step": 1829
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 9.986655896607432e-05,
+      "loss": 3.3701,
+      "step": 1830
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 9.986639741518578e-05,
+      "loss": 3.3777,
+      "step": 1831
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 9.986623586429726e-05,
+      "loss": 3.4002,
+      "step": 1832
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 9.986607431340873e-05,
+      "loss": 3.4013,
+      "step": 1833
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 9.98659127625202e-05,
+      "loss": 3.6573,
+      "step": 1834
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 9.986575121163166e-05,
+      "loss": 3.4195,
+      "step": 1835
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 9.986558966074315e-05,
+      "loss": 3.4679,
+      "step": 1836
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 9.986542810985461e-05,
+      "loss": 3.4261,
+      "step": 1837
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 9.986526655896607e-05,
+      "loss": 3.4559,
+      "step": 1838
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 9.986510500807756e-05,
+      "loss": 3.4514,
+      "step": 1839
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 9.986494345718902e-05,
+      "loss": 3.7859,
+      "step": 1840
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 9.986478190630048e-05,
+      "loss": 3.6416,
+      "step": 1841
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 9.986462035541197e-05,
+      "loss": 3.6602,
+      "step": 1842
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 9.986445880452343e-05,
+      "loss": 5.4816,
+      "step": 1843
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 9.986429725363489e-05,
+      "loss": 3.4479,
+      "step": 1844
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 9.986413570274638e-05,
+      "loss": 3.4496,
+      "step": 1845
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 9.986397415185784e-05,
+      "loss": 3.4492,
+      "step": 1846
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 9.98638126009693e-05,
+      "loss": 3.3963,
+      "step": 1847
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 9.986365105008078e-05,
+      "loss": 3.4388,
+      "step": 1848
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 9.986348949919225e-05,
+      "loss": 3.3817,
+      "step": 1849
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 9.986332794830372e-05,
+      "loss": 3.3609,
+      "step": 1850
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 9.986316639741519e-05,
+      "loss": 3.3993,
+      "step": 1851
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 9.986300484652667e-05,
+      "loss": 3.3932,
+      "step": 1852
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 9.986284329563813e-05,
+      "loss": 3.3697,
+      "step": 1853
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 9.98626817447496e-05,
+      "loss": 3.3669,
+      "step": 1854
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 9.986252019386108e-05,
+      "loss": 3.3733,
+      "step": 1855
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 9.986235864297254e-05,
+      "loss": 3.3821,
+      "step": 1856
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 9.986219709208401e-05,
+      "loss": 3.3727,
+      "step": 1857
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 9.986203554119549e-05,
+      "loss": 3.397,
+      "step": 1858
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 9.986187399030695e-05,
+      "loss": 3.4129,
+      "step": 1859
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 9.986171243941843e-05,
+      "loss": 3.4601,
+      "step": 1860
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 9.986155088852989e-05,
+      "loss": 3.4424,
+      "step": 1861
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 9.986138933764136e-05,
+      "loss": 3.4003,
+      "step": 1862
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 9.986122778675284e-05,
+      "loss": 3.4314,
+      "step": 1863
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 9.98610662358643e-05,
+      "loss": 3.521,
+      "step": 1864
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 9.986090468497577e-05,
+      "loss": 3.547,
+      "step": 1865
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 9.986074313408725e-05,
+      "loss": 3.8098,
+      "step": 1866
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 9.986058158319871e-05,
+      "loss": 4.744,
+      "step": 1867
+    },
+    {
+      "epoch": 15.0,
+      "eval_loss": 5.2505598068237305,
+      "eval_runtime": 39.1786,
+      "eval_samples_per_second": 21.466,
+      "eval_steps_per_second": 0.689,
+      "eval_wer": 1.0,
+      "step": 1867
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 9.986042003231019e-05,
+      "loss": 3.4636,
+      "step": 1868
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.986025848142166e-05,
+      "loss": 3.4864,
+      "step": 1869
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.986009693053312e-05,
+      "loss": 3.4756,
+      "step": 1870
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.98599353796446e-05,
+      "loss": 3.3943,
+      "step": 1871
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 9.985977382875606e-05,
+      "loss": 3.4303,
+      "step": 1872
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 9.985961227786753e-05,
+      "loss": 3.3743,
+      "step": 1873
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.9859450726979e-05,
+      "loss": 3.3826,
+      "step": 1874
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.985928917609047e-05,
+      "loss": 3.3532,
+      "step": 1875
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 9.985912762520195e-05,
+      "loss": 3.3893,
+      "step": 1876
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 9.98589660743134e-05,
+      "loss": 3.3753,
+      "step": 1877
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 9.985880452342488e-05,
+      "loss": 3.3924,
+      "step": 1878
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.985864297253636e-05,
+      "loss": 3.3669,
+      "step": 1879
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 9.985848142164782e-05,
+      "loss": 3.3872,
+      "step": 1880
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 9.985831987075929e-05,
+      "loss": 3.4195,
+      "step": 1881
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 9.985815831987077e-05,
+      "loss": 3.401,
+      "step": 1882
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.985799676898223e-05,
+      "loss": 3.4244,
+      "step": 1883
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.98578352180937e-05,
+      "loss": 3.3836,
+      "step": 1884
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 9.985767366720518e-05,
+      "loss": 3.4067,
+      "step": 1885
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 9.985751211631664e-05,
+      "loss": 3.4543,
+      "step": 1886
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 9.98573505654281e-05,
+      "loss": 3.4272,
+      "step": 1887
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 9.985718901453959e-05,
+      "loss": 3.5068,
+      "step": 1888
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.985702746365105e-05,
+      "loss": 3.5036,
+      "step": 1889
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 9.985686591276251e-05,
+      "loss": 3.6338,
+      "step": 1890
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 9.9856704361874e-05,
+      "loss": 3.9178,
+      "step": 1891
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 9.985654281098546e-05,
+      "loss": 4.907,
+      "step": 1892
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 9.985638126009693e-05,
+      "loss": 3.4446,
+      "step": 1893
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.985621970920841e-05,
+      "loss": 3.457,
+      "step": 1894
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 9.985605815831988e-05,
+      "loss": 3.4414,
+      "step": 1895
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 9.985589660743134e-05,
+      "loss": 3.4024,
+      "step": 1896
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 9.985573505654283e-05,
+      "loss": 3.3916,
+      "step": 1897
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 9.985557350565429e-05,
+      "loss": 3.3933,
+      "step": 1898
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.985541195476575e-05,
+      "loss": 3.3905,
+      "step": 1899
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 9.985525040387722e-05,
+      "loss": 3.3867,
+      "step": 1900
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.98550888529887e-05,
+      "loss": 3.3747,
+      "step": 1901
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 9.985492730210016e-05,
+      "loss": 3.3815,
+      "step": 1902
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 9.985476575121164e-05,
+      "loss": 3.3783,
+      "step": 1903
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 9.985460420032311e-05,
+      "loss": 3.3767,
+      "step": 1904
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 9.985444264943457e-05,
+      "loss": 3.4112,
+      "step": 1905
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 9.985428109854605e-05,
+      "loss": 3.3787,
+      "step": 1906
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 9.985411954765752e-05,
+      "loss": 3.3962,
+      "step": 1907
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 9.985395799676898e-05,
+      "loss": 3.3884,
+      "step": 1908
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 9.985379644588046e-05,
+      "loss": 3.4009,
+      "step": 1909
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 9.985363489499193e-05,
+      "loss": 3.3937,
+      "step": 1910
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 9.98534733441034e-05,
+      "loss": 3.4264,
+      "step": 1911
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 9.985331179321487e-05,
+      "loss": 3.4468,
+      "step": 1912
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 9.985315024232633e-05,
+      "loss": 3.6839,
+      "step": 1913
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 9.985298869143781e-05,
+      "loss": 3.5753,
+      "step": 1914
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 9.985282714054928e-05,
+      "loss": 4.2657,
+      "step": 1915
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 9.985266558966074e-05,
+      "loss": 4.154,
+      "step": 1916
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 9.985250403877222e-05,
+      "loss": 5.2054,
+      "step": 1917
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 9.98523424878837e-05,
+      "loss": 3.5003,
+      "step": 1918
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 9.985218093699516e-05,
+      "loss": 3.4627,
+      "step": 1919
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 9.985201938610663e-05,
+      "loss": 3.4064,
+      "step": 1920
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 9.985185783521809e-05,
+      "loss": 3.4062,
+      "step": 1921
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 9.985169628432957e-05,
+      "loss": 3.4017,
+      "step": 1922
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 9.985153473344104e-05,
+      "loss": 3.3824,
+      "step": 1923
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 9.98513731825525e-05,
+      "loss": 3.3939,
+      "step": 1924
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 9.985121163166398e-05,
+      "loss": 3.3745,
+      "step": 1925
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 9.985105008077545e-05,
+      "loss": 3.3643,
+      "step": 1926
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 9.985088852988692e-05,
+      "loss": 3.4041,
+      "step": 1927
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 9.985072697899839e-05,
+      "loss": 3.3807,
+      "step": 1928
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 9.985056542810985e-05,
+      "loss": 3.393,
+      "step": 1929
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 9.985040387722133e-05,
+      "loss": 3.4,
+      "step": 1930
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 9.98502423263328e-05,
+      "loss": 3.3761,
+      "step": 1931
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 9.985008077544426e-05,
+      "loss": 3.3903,
+      "step": 1932
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 9.984991922455574e-05,
+      "loss": 3.4099,
+      "step": 1933
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 9.984975767366721e-05,
+      "loss": 3.427,
+      "step": 1934
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 9.984959612277868e-05,
+      "loss": 3.4301,
+      "step": 1935
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 9.984943457189015e-05,
+      "loss": 3.4176,
+      "step": 1936
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 9.984927302100163e-05,
+      "loss": 3.445,
+      "step": 1937
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 9.984911147011309e-05,
+      "loss": 3.548,
+      "step": 1938
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 9.984894991922456e-05,
+      "loss": 3.4985,
+      "step": 1939
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 9.984878836833604e-05,
+      "loss": 3.6898,
+      "step": 1940
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 9.98486268174475e-05,
+      "loss": 3.6492,
+      "step": 1941
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 9.984846526655896e-05,
+      "loss": 4.6221,
+      "step": 1942
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 9.984830371567045e-05,
+      "loss": 3.4604,
+      "step": 1943
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 9.984814216478191e-05,
+      "loss": 3.4616,
+      "step": 1944
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 9.984798061389337e-05,
+      "loss": 3.4389,
+      "step": 1945
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 9.984781906300486e-05,
+      "loss": 3.4178,
+      "step": 1946
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 9.984765751211632e-05,
+      "loss": 3.4,
+      "step": 1947
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 9.984749596122778e-05,
+      "loss": 3.3561,
+      "step": 1948
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 9.984733441033927e-05,
+      "loss": 3.3834,
+      "step": 1949
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 9.984717285945073e-05,
+      "loss": 3.3876,
+      "step": 1950
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 9.98470113085622e-05,
+      "loss": 3.3746,
+      "step": 1951
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 9.984684975767368e-05,
+      "loss": 3.4804,
+      "step": 1952
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 9.984668820678514e-05,
+      "loss": 3.3701,
+      "step": 1953
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 9.98465266558966e-05,
+      "loss": 3.3863,
+      "step": 1954
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 9.984636510500808e-05,
+      "loss": 3.3817,
+      "step": 1955
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 9.984620355411956e-05,
+      "loss": 3.3958,
+      "step": 1956
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 9.984604200323102e-05,
+      "loss": 3.3921,
+      "step": 1957
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 9.984588045234249e-05,
+      "loss": 3.399,
+      "step": 1958
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 9.984571890145397e-05,
+      "loss": 3.4248,
+      "step": 1959
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 9.984555735056543e-05,
+      "loss": 3.408,
+      "step": 1960
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 9.98453957996769e-05,
+      "loss": 3.4182,
+      "step": 1961
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 9.984523424878838e-05,
+      "loss": 3.4355,
+      "step": 1962
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 9.984507269789984e-05,
+      "loss": 3.4948,
+      "step": 1963
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 9.984491114701132e-05,
+      "loss": 3.572,
+      "step": 1964
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 9.984474959612279e-05,
+      "loss": 3.5348,
+      "step": 1965
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 9.984458804523425e-05,
+      "loss": 3.9518,
+      "step": 1966
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 9.984442649434573e-05,
+      "loss": 4.8724,
+      "step": 1967
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 9.984426494345719e-05,
+      "loss": 3.4344,
+      "step": 1968
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 9.984410339256866e-05,
+      "loss": 3.4423,
+      "step": 1969
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 9.984394184168014e-05,
+      "loss": 3.4241,
+      "step": 1970
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 9.98437802907916e-05,
+      "loss": 3.4129,
+      "step": 1971
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 9.984361873990308e-05,
+      "loss": 3.402,
+      "step": 1972
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 9.984345718901454e-05,
+      "loss": 3.3878,
+      "step": 1973
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 9.984329563812601e-05,
+      "loss": 3.3824,
+      "step": 1974
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 9.984313408723749e-05,
+      "loss": 3.3919,
+      "step": 1975
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 9.984297253634895e-05,
+      "loss": 3.3806,
+      "step": 1976
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 9.984281098546042e-05,
+      "loss": 3.3777,
+      "step": 1977
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 9.98426494345719e-05,
+      "loss": 3.418,
+      "step": 1978
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 9.984248788368336e-05,
+      "loss": 3.3896,
+      "step": 1979
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 9.984232633279484e-05,
+      "loss": 3.3631,
+      "step": 1980
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 9.98421647819063e-05,
+      "loss": 3.384,
+      "step": 1981
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 9.984200323101777e-05,
+      "loss": 3.3722,
+      "step": 1982
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 9.984184168012925e-05,
+      "loss": 3.3914,
+      "step": 1983
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 9.984168012924071e-05,
+      "loss": 3.4091,
+      "step": 1984
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 9.984151857835218e-05,
+      "loss": 3.4843,
+      "step": 1985
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 9.984135702746366e-05,
+      "loss": 3.4427,
+      "step": 1986
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 9.984119547657512e-05,
+      "loss": 3.4347,
+      "step": 1987
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 9.98410339256866e-05,
+      "loss": 3.4934,
+      "step": 1988
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 9.984087237479807e-05,
+      "loss": 3.639,
+      "step": 1989
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 9.984071082390953e-05,
+      "loss": 3.5962,
+      "step": 1990
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 9.984054927302101e-05,
+      "loss": 4.3664,
+      "step": 1991
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 9.984038772213248e-05,
+      "loss": 5.2138,
+      "step": 1992
+    },
+    {
+      "epoch": 15.0,
+      "eval_loss": 5.1763596534729,
+      "eval_runtime": 39.7016,
+      "eval_samples_per_second": 21.183,
+      "eval_steps_per_second": 0.68,
+      "eval_wer": 1.0,
+      "step": 1992
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 9.984022617124394e-05,
+      "loss": 3.4346,
+      "step": 1993
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 9.98400646203554e-05,
+      "loss": 3.4439,
+      "step": 1994
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 9.98399030694669e-05,
+      "loss": 3.4001,
+      "step": 1995
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 9.983974151857836e-05,
+      "loss": 3.4203,
+      "step": 1996
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 9.983957996768982e-05,
+      "loss": 3.3986,
+      "step": 1997
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 9.98394184168013e-05,
+      "loss": 3.3735,
+      "step": 1998
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 9.983925686591277e-05,
+      "loss": 3.4019,
+      "step": 1999
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 9.983909531502423e-05,
+      "loss": 3.3893,
+      "step": 2000
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 9.983893376413572e-05,
+      "loss": 3.3738,
+      "step": 2001
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 9.983877221324718e-05,
+      "loss": 3.3906,
+      "step": 2002
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 9.983861066235864e-05,
+      "loss": 3.3834,
+      "step": 2003
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 9.983844911147013e-05,
+      "loss": 3.4069,
+      "step": 2004
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 9.983828756058159e-05,
+      "loss": 3.4005,
+      "step": 2005
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 9.983812600969305e-05,
+      "loss": 3.3971,
+      "step": 2006
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 9.983796445880453e-05,
+      "loss": 3.4005,
+      "step": 2007
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 9.9837802907916e-05,
+      "loss": 3.3984,
+      "step": 2008
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 9.983764135702746e-05,
+      "loss": 3.4196,
+      "step": 2009
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 9.983747980613894e-05,
+      "loss": 3.4134,
+      "step": 2010
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 9.983731825525041e-05,
+      "loss": 3.4137,
+      "step": 2011
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 9.983715670436188e-05,
+      "loss": 3.4869,
+      "step": 2012
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 9.983699515347335e-05,
+      "loss": 3.5241,
+      "step": 2013
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 9.983683360258483e-05,
+      "loss": 3.6443,
+      "step": 2014
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 9.983667205169629e-05,
+      "loss": 3.5845,
+      "step": 2015
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 9.983651050080776e-05,
+      "loss": 3.7464,
+      "step": 2016
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 9.983634894991924e-05,
+      "loss": 4.7983,
+      "step": 2017
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 9.98361873990307e-05,
+      "loss": 3.4519,
+      "step": 2018
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 9.983602584814217e-05,
+      "loss": 3.4439,
+      "step": 2019
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 9.983586429725365e-05,
+      "loss": 3.4557,
+      "step": 2020
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 9.983570274636511e-05,
+      "loss": 3.4125,
+      "step": 2021
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 9.983554119547659e-05,
+      "loss": 3.4107,
+      "step": 2022
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 9.983537964458805e-05,
+      "loss": 3.3936,
+      "step": 2023
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 9.983521809369952e-05,
+      "loss": 3.3825,
+      "step": 2024
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 9.983505654281098e-05,
+      "loss": 3.3919,
+      "step": 2025
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 9.983489499192246e-05,
+      "loss": 3.383,
+      "step": 2026
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 9.983473344103393e-05,
+      "loss": 3.4045,
+      "step": 2027
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 9.98345718901454e-05,
+      "loss": 3.3671,
+      "step": 2028
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 9.983441033925687e-05,
+      "loss": 3.3864,
+      "step": 2029
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 9.983424878836834e-05,
+      "loss": 3.3602,
+      "step": 2030
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 9.98340872374798e-05,
+      "loss": 3.3867,
+      "step": 2031
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 9.983392568659128e-05,
+      "loss": 3.4051,
+      "step": 2032
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 9.983376413570276e-05,
+      "loss": 3.3865,
+      "step": 2033
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 9.983360258481422e-05,
+      "loss": 3.4365,
+      "step": 2034
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 9.983344103392569e-05,
+      "loss": 3.3984,
+      "step": 2035
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 9.983327948303715e-05,
+      "loss": 3.4567,
+      "step": 2036
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 9.983311793214863e-05,
+      "loss": 3.4488,
+      "step": 2037
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 9.98329563812601e-05,
+      "loss": 3.5055,
+      "step": 2038
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 9.983279483037157e-05,
+      "loss": 3.4472,
+      "step": 2039
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 9.983263327948304e-05,
+      "loss": 3.6984,
+      "step": 2040
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 9.983247172859452e-05,
+      "loss": 4.125,
+      "step": 2041
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 9.983231017770598e-05,
+      "loss": 5.34,
+      "step": 2042
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 9.983214862681745e-05,
+      "loss": 3.4269,
+      "step": 2043
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 9.983198707592893e-05,
+      "loss": 3.6023,
+      "step": 2044
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 9.983182552504039e-05,
+      "loss": 3.4286,
+      "step": 2045
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 9.983166397415186e-05,
+      "loss": 3.4368,
+      "step": 2046
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 9.983150242326334e-05,
+      "loss": 3.4209,
+      "step": 2047
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 9.98313408723748e-05,
+      "loss": 3.3745,
+      "step": 2048
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 9.983117932148626e-05,
+      "loss": 3.3699,
+      "step": 2049
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 9.983101777059775e-05,
+      "loss": 3.3814,
+      "step": 2050
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 9.983085621970921e-05,
+      "loss": 3.3741,
+      "step": 2051
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 9.983069466882067e-05,
+      "loss": 3.3765,
+      "step": 2052
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 9.983053311793216e-05,
+      "loss": 3.3839,
+      "step": 2053
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 9.983037156704362e-05,
+      "loss": 3.3696,
+      "step": 2054
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 9.983021001615509e-05,
+      "loss": 3.414,
+      "step": 2055
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 9.983004846526657e-05,
+      "loss": 3.3824,
+      "step": 2056
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 9.982988691437804e-05,
+      "loss": 3.3895,
+      "step": 2057
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 9.98297253634895e-05,
+      "loss": 3.4463,
+      "step": 2058
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 9.982956381260097e-05,
+      "loss": 3.4266,
+      "step": 2059
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 9.982940226171245e-05,
+      "loss": 3.4468,
+      "step": 2060
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 9.982924071082391e-05,
+      "loss": 3.4166,
+      "step": 2061
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 9.982907915993538e-05,
+      "loss": 3.4391,
+      "step": 2062
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 9.982891760904686e-05,
+      "loss": 3.4411,
+      "step": 2063
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 9.982875605815832e-05,
+      "loss": 3.4493,
+      "step": 2064
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 9.98285945072698e-05,
+      "loss": 3.6914,
+      "step": 2065
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 9.982843295638127e-05,
+      "loss": 3.9023,
+      "step": 2066
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 9.982827140549273e-05,
+      "loss": 5.7024,
+      "step": 2067
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 9.982810985460421e-05,
+      "loss": 3.5259,
+      "step": 2068
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 9.982794830371568e-05,
+      "loss": 3.455,
+      "step": 2069
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 9.982778675282714e-05,
+      "loss": 3.4347,
+      "step": 2070
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 9.982762520193862e-05,
+      "loss": 3.407,
+      "step": 2071
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 9.98274636510501e-05,
+      "loss": 3.395,
+      "step": 2072
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 9.982730210016156e-05,
+      "loss": 3.3943,
+      "step": 2073
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 9.982714054927303e-05,
+      "loss": 3.3873,
+      "step": 2074
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 9.982697899838449e-05,
+      "loss": 3.3637,
+      "step": 2075
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 9.982681744749597e-05,
+      "loss": 3.3987,
+      "step": 2076
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 9.982665589660743e-05,
+      "loss": 3.3838,
+      "step": 2077
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 9.98264943457189e-05,
+      "loss": 3.3689,
+      "step": 2078
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 9.982633279483038e-05,
+      "loss": 3.3653,
+      "step": 2079
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 9.982617124394184e-05,
+      "loss": 3.3827,
+      "step": 2080
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 9.982600969305332e-05,
+      "loss": 3.423,
+      "step": 2081
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 9.982584814216479e-05,
+      "loss": 3.3828,
+      "step": 2082
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 9.982568659127625e-05,
+      "loss": 3.3915,
+      "step": 2083
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 9.982552504038773e-05,
+      "loss": 3.4034,
+      "step": 2084
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 9.98253634894992e-05,
+      "loss": 3.3961,
+      "step": 2085
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 9.982520193861066e-05,
+      "loss": 3.4151,
+      "step": 2086
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 9.982504038772214e-05,
+      "loss": 3.4602,
+      "step": 2087
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 9.98248788368336e-05,
+      "loss": 3.4405,
+      "step": 2088
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 9.982471728594508e-05,
+      "loss": 3.4921,
+      "step": 2089
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 9.982455573505655e-05,
+      "loss": 3.4674,
+      "step": 2090
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 9.982439418416801e-05,
+      "loss": 4.1452,
+      "step": 2091
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 9.982423263327949e-05,
+      "loss": 5.6197,
+      "step": 2092
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 9.982407108239096e-05,
+      "loss": 3.4692,
+      "step": 2093
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 9.982390953150242e-05,
+      "loss": 3.4237,
+      "step": 2094
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 9.98237479806139e-05,
+      "loss": 3.4159,
+      "step": 2095
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 9.982358642972537e-05,
+      "loss": 3.4136,
+      "step": 2096
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 9.982342487883683e-05,
+      "loss": 3.3817,
+      "step": 2097
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 9.982326332794831e-05,
+      "loss": 3.3718,
+      "step": 2098
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 9.982310177705979e-05,
+      "loss": 3.3845,
+      "step": 2099
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 9.982294022617125e-05,
+      "loss": 3.3736,
+      "step": 2100
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 9.982277867528272e-05,
+      "loss": 3.3724,
+      "step": 2101
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 9.98226171243942e-05,
+      "loss": 3.3764,
+      "step": 2102
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 9.982245557350566e-05,
+      "loss": 3.3823,
+      "step": 2103
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 9.982229402261712e-05,
+      "loss": 3.4981,
+      "step": 2104
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 9.982213247172861e-05,
+      "loss": 3.3711,
+      "step": 2105
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 9.982197092084007e-05,
+      "loss": 3.3958,
+      "step": 2106
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 9.982180936995153e-05,
+      "loss": 3.39,
+      "step": 2107
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 9.982164781906302e-05,
+      "loss": 3.4537,
+      "step": 2108
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 9.982148626817448e-05,
+      "loss": 3.4317,
+      "step": 2109
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 9.982132471728594e-05,
+      "loss": 3.4186,
+      "step": 2110
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 9.982116316639742e-05,
+      "loss": 3.4916,
+      "step": 2111
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 9.982100161550889e-05,
+      "loss": 3.6911,
+      "step": 2112
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 9.982084006462035e-05,
+      "loss": 3.4662,
+      "step": 2113
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 9.982067851373183e-05,
+      "loss": 3.5233,
+      "step": 2114
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 9.98205169628433e-05,
+      "loss": 3.6269,
+      "step": 2115
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 9.982035541195477e-05,
+      "loss": 4.7412,
+      "step": 2116
+    },
+    {
+      "epoch": 17.0,
+      "eval_loss": 4.64270544052124,
+      "eval_runtime": 38.8096,
+      "eval_samples_per_second": 21.67,
+      "eval_steps_per_second": 0.696,
+      "eval_wer": 1.0,
+      "step": 2116
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 9.982019386106624e-05,
+      "loss": 3.475,
+      "step": 2117
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 9.982003231017772e-05,
+      "loss": 3.4318,
+      "step": 2118
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 9.981987075928918e-05,
+      "loss": 3.4111,
+      "step": 2119
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 9.981970920840065e-05,
+      "loss": 3.4218,
+      "step": 2120
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 9.981954765751213e-05,
+      "loss": 3.3785,
+      "step": 2121
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 9.981938610662359e-05,
+      "loss": 3.3729,
+      "step": 2122
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 9.981922455573506e-05,
+      "loss": 3.3694,
+      "step": 2123
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 9.981906300484654e-05,
+      "loss": 3.3732,
+      "step": 2124
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 9.9818901453958e-05,
+      "loss": 3.3739,
+      "step": 2125
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 9.981873990306946e-05,
+      "loss": 3.4061,
+      "step": 2126
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 9.981857835218095e-05,
+      "loss": 3.3878,
+      "step": 2127
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 9.981841680129241e-05,
+      "loss": 3.3927,
+      "step": 2128
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 9.981825525040387e-05,
+      "loss": 3.3978,
+      "step": 2129
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 9.981809369951535e-05,
+      "loss": 3.3986,
+      "step": 2130
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 9.981793214862682e-05,
+      "loss": 3.3952,
+      "step": 2131
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 9.981777059773829e-05,
+      "loss": 3.4128,
+      "step": 2132
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 9.981760904684976e-05,
+      "loss": 3.4022,
+      "step": 2133
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 9.981744749596124e-05,
+      "loss": 3.4327,
+      "step": 2134
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 9.98172859450727e-05,
+      "loss": 3.4581,
+      "step": 2135
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 9.981712439418417e-05,
+      "loss": 3.5364,
+      "step": 2136
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 9.981696284329565e-05,
+      "loss": 3.7411,
+      "step": 2137
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 9.981680129240711e-05,
+      "loss": 3.6514,
+      "step": 2138
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 9.981663974151858e-05,
+      "loss": 3.5979,
+      "step": 2139
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 9.981647819063006e-05,
+      "loss": 3.7142,
+      "step": 2140
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 9.981631663974152e-05,
+      "loss": 5.117,
+      "step": 2141
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 9.9816155088853e-05,
+      "loss": 3.4463,
+      "step": 2142
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 9.981599353796446e-05,
+      "loss": 3.447,
+      "step": 2143
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 9.981583198707593e-05,
+      "loss": 3.4161,
+      "step": 2144
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 9.981567043618741e-05,
+      "loss": 3.4293,
+      "step": 2145
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 9.981550888529887e-05,
+      "loss": 3.4267,
+      "step": 2146
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 9.981534733441034e-05,
+      "loss": 3.4337,
+      "step": 2147
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 9.981518578352182e-05,
+      "loss": 3.3793,
+      "step": 2148
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 9.981502423263328e-05,
+      "loss": 3.3865,
+      "step": 2149
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 9.981486268174476e-05,
+      "loss": 3.3796,
+      "step": 2150
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 9.981470113085623e-05,
+      "loss": 3.387,
+      "step": 2151
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 9.981453957996769e-05,
+      "loss": 3.3662,
+      "step": 2152
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 9.981437802907917e-05,
+      "loss": 3.4009,
+      "step": 2153
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 9.981421647819064e-05,
+      "loss": 3.3814,
+      "step": 2154
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 9.98140549273021e-05,
+      "loss": 3.3678,
+      "step": 2155
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 9.981389337641357e-05,
+      "loss": 3.4002,
+      "step": 2156
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 9.981373182552505e-05,
+      "loss": 3.4278,
+      "step": 2157
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 9.981357027463652e-05,
+      "loss": 3.4044,
+      "step": 2158
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 9.981340872374798e-05,
+      "loss": 3.4197,
+      "step": 2159
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 9.981324717285947e-05,
+      "loss": 3.4327,
+      "step": 2160
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 9.981308562197093e-05,
+      "loss": 3.4362,
+      "step": 2161
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 9.981292407108239e-05,
+      "loss": 3.5289,
+      "step": 2162
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 9.981276252019386e-05,
+      "loss": 3.5798,
+      "step": 2163
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 9.981260096930534e-05,
+      "loss": 3.5013,
+      "step": 2164
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 9.98124394184168e-05,
+      "loss": 4.3013,
+      "step": 2165
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 9.981227786752828e-05,
+      "loss": 5.2187,
+      "step": 2166
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 9.981211631663975e-05,
+      "loss": 3.4744,
+      "step": 2167
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 9.981195476575121e-05,
+      "loss": 3.4555,
+      "step": 2168
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 9.981179321486269e-05,
+      "loss": 3.4575,
+      "step": 2169
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 9.981163166397416e-05,
+      "loss": 3.4332,
+      "step": 2170
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 9.981147011308562e-05,
+      "loss": 3.4172,
+      "step": 2171
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 9.98113085621971e-05,
+      "loss": 3.4152,
+      "step": 2172
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 9.981114701130857e-05,
+      "loss": 3.374,
+      "step": 2173
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 9.981098546042003e-05,
+      "loss": 3.3815,
+      "step": 2174
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 9.981082390953151e-05,
+      "loss": 3.3756,
+      "step": 2175
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 9.981066235864298e-05,
+      "loss": 3.3781,
+      "step": 2176
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 9.981050080775445e-05,
+      "loss": 3.3872,
+      "step": 2177
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 9.981033925686591e-05,
+      "loss": 3.3879,
+      "step": 2178
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 9.98101777059774e-05,
+      "loss": 3.3967,
+      "step": 2179
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 9.981001615508886e-05,
+      "loss": 3.3747,
+      "step": 2180
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 9.980985460420032e-05,
+      "loss": 3.3994,
+      "step": 2181
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 9.980969305331181e-05,
+      "loss": 3.3946,
+      "step": 2182
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 9.980953150242327e-05,
+      "loss": 3.3892,
+      "step": 2183
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 9.980936995153473e-05,
+      "loss": 3.4247,
+      "step": 2184
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 9.98092084006462e-05,
+      "loss": 3.5314,
+      "step": 2185
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 9.980904684975768e-05,
+      "loss": 3.4556,
+      "step": 2186
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 9.980888529886914e-05,
+      "loss": 3.4682,
+      "step": 2187
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 9.980872374798062e-05,
+      "loss": 3.538,
+      "step": 2188
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 9.980856219709209e-05,
+      "loss": 3.6367,
+      "step": 2189
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 9.980840064620355e-05,
+      "loss": 3.8164,
+      "step": 2190
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 9.980823909531503e-05,
+      "loss": 5.3578,
+      "step": 2191
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 9.98080775444265e-05,
+      "loss": 3.4821,
+      "step": 2192
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 9.980791599353797e-05,
+      "loss": 3.4856,
+      "step": 2193
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 9.980775444264944e-05,
+      "loss": 3.4223,
+      "step": 2194
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 9.980759289176092e-05,
+      "loss": 3.4078,
+      "step": 2195
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 9.980743134087238e-05,
+      "loss": 3.4067,
+      "step": 2196
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 9.980726978998385e-05,
+      "loss": 3.3952,
+      "step": 2197
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 9.980710823909531e-05,
+      "loss": 3.3871,
+      "step": 2198
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 9.980694668820679e-05,
+      "loss": 3.3811,
+      "step": 2199
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 9.980678513731826e-05,
+      "loss": 3.371,
+      "step": 2200
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 9.980662358642973e-05,
+      "loss": 3.3717,
+      "step": 2201
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 9.98064620355412e-05,
+      "loss": 3.377,
+      "step": 2202
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 9.980630048465268e-05,
+      "loss": 3.3976,
+      "step": 2203
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 9.980613893376414e-05,
+      "loss": 3.3919,
+      "step": 2204
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 9.980597738287561e-05,
+      "loss": 3.4054,
+      "step": 2205
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 9.980581583198709e-05,
+      "loss": 3.3821,
+      "step": 2206
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 9.980565428109855e-05,
+      "loss": 3.3926,
+      "step": 2207
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 9.980549273021002e-05,
+      "loss": 3.4109,
+      "step": 2208
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 9.98053311793215e-05,
+      "loss": 3.4147,
+      "step": 2209
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 9.980516962843296e-05,
+      "loss": 3.4873,
+      "step": 2210
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 9.980500807754442e-05,
+      "loss": 3.4573,
+      "step": 2211
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 9.980484652665591e-05,
+      "loss": 3.5282,
+      "step": 2212
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 9.980468497576737e-05,
+      "loss": 3.5363,
+      "step": 2213
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 9.980452342487883e-05,
+      "loss": 3.4939,
+      "step": 2214
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 9.980436187399031e-05,
+      "loss": 4.4445,
+      "step": 2215
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 9.980420032310178e-05,
+      "loss": 5.6252,
+      "step": 2216
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 9.980403877221325e-05,
+      "loss": 3.4699,
+      "step": 2217
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 9.980387722132472e-05,
+      "loss": 3.4391,
+      "step": 2218
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 9.98037156704362e-05,
+      "loss": 3.4223,
+      "step": 2219
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 9.980355411954766e-05,
+      "loss": 3.4265,
+      "step": 2220
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 9.980339256865913e-05,
+      "loss": 3.3975,
+      "step": 2221
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 9.980323101777061e-05,
+      "loss": 3.3642,
+      "step": 2222
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 9.980306946688207e-05,
+      "loss": 3.3805,
+      "step": 2223
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 9.980290791599354e-05,
+      "loss": 3.3992,
+      "step": 2224
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 9.980274636510502e-05,
+      "loss": 3.3801,
+      "step": 2225
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 9.980258481421648e-05,
+      "loss": 3.3609,
+      "step": 2226
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 9.980242326332796e-05,
+      "loss": 3.384,
+      "step": 2227
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 9.980226171243943e-05,
+      "loss": 3.3794,
+      "step": 2228
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 9.980210016155089e-05,
+      "loss": 3.3888,
+      "step": 2229
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 9.980193861066235e-05,
+      "loss": 3.3791,
+      "step": 2230
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 9.980177705977384e-05,
+      "loss": 3.3992,
+      "step": 2231
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 9.98016155088853e-05,
+      "loss": 3.3833,
+      "step": 2232
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 9.980145395799677e-05,
+      "loss": 3.4642,
+      "step": 2233
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 9.980129240710825e-05,
+      "loss": 3.4389,
+      "step": 2234
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 9.980113085621972e-05,
+      "loss": 3.4382,
+      "step": 2235
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 9.980096930533118e-05,
+      "loss": 3.4324,
+      "step": 2236
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 9.980080775444265e-05,
+      "loss": 3.6596,
+      "step": 2237
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 9.980064620355413e-05,
+      "loss": 3.4792,
+      "step": 2238
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 9.980048465266559e-05,
+      "loss": 3.6209,
+      "step": 2239
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 9.980032310177706e-05,
+      "loss": 5.0611,
+      "step": 2240
+    },
+    {
+      "epoch": 18.0,
+      "eval_loss": 4.816751480102539,
+      "eval_runtime": 37.1095,
+      "eval_samples_per_second": 22.528,
+      "eval_steps_per_second": 0.728,
+      "eval_wer": 1.0,
+      "step": 2240
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 9.980016155088854e-05,
+      "loss": 3.4903,
+      "step": 2241
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 9.98e-05,
+      "loss": 3.4504,
+      "step": 2242
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 9.979983844911147e-05,
+      "loss": 3.4449,
+      "step": 2243
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 9.979967689822295e-05,
+      "loss": 3.4294,
+      "step": 2244
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 9.979951534733441e-05,
+      "loss": 3.391,
+      "step": 2245
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 9.979935379644589e-05,
+      "loss": 3.3744,
+      "step": 2246
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 9.979919224555736e-05,
+      "loss": 3.3982,
+      "step": 2247
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 9.979903069466882e-05,
+      "loss": 3.3653,
+      "step": 2248
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 9.97988691437803e-05,
+      "loss": 3.3716,
+      "step": 2249
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 9.979870759289176e-05,
+      "loss": 3.3703,
+      "step": 2250
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 9.979854604200323e-05,
+      "loss": 3.4018,
+      "step": 2251
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 9.979838449111471e-05,
+      "loss": 3.3859,
+      "step": 2252
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 9.979822294022617e-05,
+      "loss": 3.3735,
+      "step": 2253
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 9.979806138933765e-05,
+      "loss": 3.3819,
+      "step": 2254
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 9.979789983844912e-05,
+      "loss": 3.3949,
+      "step": 2255
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 9.979773828756058e-05,
+      "loss": 3.4041,
+      "step": 2256
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 9.979757673667206e-05,
+      "loss": 3.4147,
+      "step": 2257
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 9.979741518578353e-05,
+      "loss": 3.4296,
+      "step": 2258
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 9.9797253634895e-05,
+      "loss": 3.4317,
+      "step": 2259
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 9.979709208400647e-05,
+      "loss": 3.4559,
+      "step": 2260
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 9.979693053311794e-05,
+      "loss": 3.4686,
+      "step": 2261
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 9.97967689822294e-05,
+      "loss": 3.6497,
+      "step": 2262
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 9.979660743134087e-05,
+      "loss": 3.5004,
+      "step": 2263
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 9.979644588045234e-05,
+      "loss": 3.85,
+      "step": 2264
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 9.979628432956382e-05,
+      "loss": 4.8379,
+      "step": 2265
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 9.979612277867528e-05,
+      "loss": 3.5011,
+      "step": 2266
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 9.979596122778675e-05,
+      "loss": 3.448,
+      "step": 2267
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 9.979579967689823e-05,
+      "loss": 3.4296,
+      "step": 2268
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 9.979563812600969e-05,
+      "loss": 3.4346,
+      "step": 2269
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 9.979547657512117e-05,
+      "loss": 3.4059,
+      "step": 2270
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 9.979531502423264e-05,
+      "loss": 3.3911,
+      "step": 2271
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 9.97951534733441e-05,
+      "loss": 3.39,
+      "step": 2272
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 9.979499192245558e-05,
+      "loss": 3.3609,
+      "step": 2273
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 9.979483037156705e-05,
+      "loss": 3.3678,
+      "step": 2274
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 9.979466882067851e-05,
+      "loss": 3.4446,
+      "step": 2275
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 9.979450726978999e-05,
+      "loss": 3.3857,
+      "step": 2276
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 9.979434571890146e-05,
+      "loss": 3.3623,
+      "step": 2277
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 9.979418416801293e-05,
+      "loss": 3.3968,
+      "step": 2278
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 9.97940226171244e-05,
+      "loss": 3.4056,
+      "step": 2279
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 9.979386106623588e-05,
+      "loss": 3.4207,
+      "step": 2280
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 9.979369951534734e-05,
+      "loss": 3.4021,
+      "step": 2281
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 9.97935379644588e-05,
+      "loss": 3.4143,
+      "step": 2282
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 9.979337641357029e-05,
+      "loss": 3.5715,
+      "step": 2283
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 9.979321486268175e-05,
+      "loss": 3.4729,
+      "step": 2284
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 9.979305331179321e-05,
+      "loss": 3.4849,
+      "step": 2285
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 9.97928917609047e-05,
+      "loss": 3.5226,
+      "step": 2286
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 9.979273021001616e-05,
+      "loss": 3.5109,
+      "step": 2287
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 9.979256865912762e-05,
+      "loss": 3.987,
+      "step": 2288
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 9.979240710823911e-05,
+      "loss": 4.2548,
+      "step": 2289
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 9.979224555735057e-05,
+      "loss": 5.8598,
+      "step": 2290
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 9.979208400646203e-05,
+      "loss": 3.4647,
+      "step": 2291
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 9.979192245557351e-05,
+      "loss": 3.4205,
+      "step": 2292
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 9.979176090468498e-05,
+      "loss": 3.4323,
+      "step": 2293
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 9.979159935379645e-05,
+      "loss": 3.4299,
+      "step": 2294
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 9.979143780290792e-05,
+      "loss": 3.3842,
+      "step": 2295
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 9.97912762520194e-05,
+      "loss": 3.3856,
+      "step": 2296
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 9.979111470113086e-05,
+      "loss": 3.4071,
+      "step": 2297
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 9.979095315024233e-05,
+      "loss": 3.3733,
+      "step": 2298
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 9.979079159935381e-05,
+      "loss": 3.3758,
+      "step": 2299
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 9.979063004846527e-05,
+      "loss": 3.3602,
+      "step": 2300
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 9.979046849757674e-05,
+      "loss": 3.3851,
+      "step": 2301
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 9.979030694668822e-05,
+      "loss": 3.3903,
+      "step": 2302
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 9.979014539579968e-05,
+      "loss": 3.4033,
+      "step": 2303
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 9.978998384491116e-05,
+      "loss": 3.4022,
+      "step": 2304
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 9.978982229402262e-05,
+      "loss": 3.4041,
+      "step": 2305
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 9.978966074313409e-05,
+      "loss": 3.3768,
+      "step": 2306
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 9.978949919224557e-05,
+      "loss": 3.3949,
+      "step": 2307
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 9.978933764135703e-05,
+      "loss": 3.4039,
+      "step": 2308
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 9.97891760904685e-05,
+      "loss": 3.4075,
+      "step": 2309
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 9.978901453957998e-05,
+      "loss": 3.435,
+      "step": 2310
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 9.978885298869144e-05,
+      "loss": 3.4469,
+      "step": 2311
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 9.978869143780292e-05,
+      "loss": 3.6279,
+      "step": 2312
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 9.978852988691439e-05,
+      "loss": 3.559,
+      "step": 2313
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 9.978836833602585e-05,
+      "loss": 3.7567,
+      "step": 2314
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 9.978820678513733e-05,
+      "loss": 5.2817,
+      "step": 2315
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 9.978804523424879e-05,
+      "loss": 3.4572,
+      "step": 2316
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 9.978788368336026e-05,
+      "loss": 3.4653,
+      "step": 2317
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 9.978772213247172e-05,
+      "loss": 3.441,
+      "step": 2318
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 9.97875605815832e-05,
+      "loss": 3.4082,
+      "step": 2319
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 9.978739903069467e-05,
+      "loss": 3.3845,
+      "step": 2320
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 9.978723747980614e-05,
+      "loss": 3.3827,
+      "step": 2321
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 9.978707592891761e-05,
+      "loss": 3.3949,
+      "step": 2322
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 9.978691437802909e-05,
+      "loss": 3.3764,
+      "step": 2323
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 9.978675282714055e-05,
+      "loss": 3.3821,
+      "step": 2324
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 9.978659127625202e-05,
+      "loss": 3.3671,
+      "step": 2325
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 9.97864297253635e-05,
+      "loss": 3.3834,
+      "step": 2326
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 9.978626817447496e-05,
+      "loss": 3.3813,
+      "step": 2327
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 9.978610662358643e-05,
+      "loss": 3.3875,
+      "step": 2328
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 9.978594507269791e-05,
+      "loss": 3.3888,
+      "step": 2329
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 9.978578352180937e-05,
+      "loss": 3.4054,
+      "step": 2330
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 9.978562197092083e-05,
+      "loss": 3.4129,
+      "step": 2331
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 9.978546042003232e-05,
+      "loss": 3.4231,
+      "step": 2332
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 9.978529886914378e-05,
+      "loss": 3.4561,
+      "step": 2333
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 9.978513731825524e-05,
+      "loss": 3.4123,
+      "step": 2334
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 9.978497576736673e-05,
+      "loss": 3.4649,
+      "step": 2335
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 9.97848142164782e-05,
+      "loss": 3.4226,
+      "step": 2336
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 9.978465266558966e-05,
+      "loss": 3.4794,
+      "step": 2337
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 9.978449111470114e-05,
+      "loss": 3.8768,
+      "step": 2338
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 9.97843295638126e-05,
+      "loss": 3.6947,
+      "step": 2339
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 9.978416801292407e-05,
+      "loss": 5.0777,
+      "step": 2340
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 9.978400646203556e-05,
+      "loss": 3.5085,
+      "step": 2341
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 9.978384491114702e-05,
+      "loss": 3.4583,
+      "step": 2342
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 9.978368336025848e-05,
+      "loss": 3.4343,
+      "step": 2343
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 9.978352180936995e-05,
+      "loss": 3.4315,
+      "step": 2344
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 9.978336025848143e-05,
+      "loss": 3.3885,
+      "step": 2345
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 9.978319870759289e-05,
+      "loss": 3.3839,
+      "step": 2346
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 9.978303715670437e-05,
+      "loss": 3.3826,
+      "step": 2347
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 9.978287560581584e-05,
+      "loss": 3.3875,
+      "step": 2348
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 9.97827140549273e-05,
+      "loss": 3.3743,
+      "step": 2349
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 9.978255250403878e-05,
+      "loss": 3.3949,
+      "step": 2350
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 9.978239095315025e-05,
+      "loss": 3.3819,
+      "step": 2351
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 9.978222940226171e-05,
+      "loss": 3.3753,
+      "step": 2352
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 9.978206785137319e-05,
+      "loss": 3.3839,
+      "step": 2353
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 9.978190630048466e-05,
+      "loss": 3.4425,
+      "step": 2354
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 9.978174474959613e-05,
+      "loss": 3.4059,
+      "step": 2355
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 9.97815831987076e-05,
+      "loss": 3.4113,
+      "step": 2356
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 9.978142164781908e-05,
+      "loss": 3.4104,
+      "step": 2357
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 9.978126009693054e-05,
+      "loss": 3.4673,
+      "step": 2358
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 9.978109854604201e-05,
+      "loss": 3.4496,
+      "step": 2359
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 9.978093699515347e-05,
+      "loss": 3.4464,
+      "step": 2360
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 9.978077544426495e-05,
+      "loss": 3.4392,
+      "step": 2361
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 9.978061389337642e-05,
+      "loss": 3.5256,
+      "step": 2362
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 9.978045234248789e-05,
+      "loss": 3.8012,
+      "step": 2363
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 9.978029079159936e-05,
+      "loss": 4.912,
+      "step": 2364
+    },
+    {
+      "epoch": 19.0,
+      "eval_loss": 4.449174880981445,
+      "eval_runtime": 38.7485,
+      "eval_samples_per_second": 21.575,
+      "eval_steps_per_second": 0.697,
+      "eval_wer": 1.0,
+      "step": 2364
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 9.978012924071084e-05,
+      "loss": 3.5797,
+      "step": 2365
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 9.97799676898223e-05,
+      "loss": 3.4596,
+      "step": 2366
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 9.977980613893377e-05,
+      "loss": 3.4696,
+      "step": 2367
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 9.977964458804523e-05,
+      "loss": 3.4248,
+      "step": 2368
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 9.977948303715671e-05,
+      "loss": 3.4958,
+      "step": 2369
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 9.977932148626818e-05,
+      "loss": 3.3933,
+      "step": 2370
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 9.977915993537965e-05,
+      "loss": 3.3717,
+      "step": 2371
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 9.977899838449112e-05,
+      "loss": 3.388,
+      "step": 2372
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 9.977883683360258e-05,
+      "loss": 3.3815,
+      "step": 2373
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 9.977867528271406e-05,
+      "loss": 3.3734,
+      "step": 2374
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 9.977851373182553e-05,
+      "loss": 3.3731,
+      "step": 2375
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 9.9778352180937e-05,
+      "loss": 3.3545,
+      "step": 2376
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 9.977819063004847e-05,
+      "loss": 3.3661,
+      "step": 2377
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 9.977802907915994e-05,
+      "loss": 3.3928,
+      "step": 2378
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 9.97778675282714e-05,
+      "loss": 3.3929,
+      "step": 2379
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 9.977770597738288e-05,
+      "loss": 3.384,
+      "step": 2380
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 9.977754442649436e-05,
+      "loss": 3.3908,
+      "step": 2381
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 9.977738287560582e-05,
+      "loss": 3.4058,
+      "step": 2382
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 9.977722132471729e-05,
+      "loss": 3.4327,
+      "step": 2383
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 9.977705977382877e-05,
+      "loss": 3.4359,
+      "step": 2384
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 9.977689822294023e-05,
+      "loss": 3.6267,
+      "step": 2385
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 9.977673667205169e-05,
+      "loss": 3.6344,
+      "step": 2386
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 9.977657512116318e-05,
+      "loss": 3.7014,
+      "step": 2387
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 9.977641357027464e-05,
+      "loss": 3.8403,
+      "step": 2388
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 9.97762520193861e-05,
+      "loss": 5.2715,
+      "step": 2389
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 9.977609046849759e-05,
+      "loss": 3.4577,
+      "step": 2390
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 9.977592891760905e-05,
+      "loss": 3.4206,
+      "step": 2391
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 9.977576736672051e-05,
+      "loss": 3.5281,
+      "step": 2392
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 9.9775605815832e-05,
+      "loss": 3.3872,
+      "step": 2393
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 9.977544426494346e-05,
+      "loss": 3.3947,
+      "step": 2394
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 9.977528271405492e-05,
+      "loss": 3.3882,
+      "step": 2395
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 9.977512116316641e-05,
+      "loss": 3.3745,
+      "step": 2396
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 9.977495961227787e-05,
+      "loss": 3.3704,
+      "step": 2397
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 9.977479806138934e-05,
+      "loss": 3.3845,
+      "step": 2398
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 9.977463651050081e-05,
+      "loss": 3.3693,
+      "step": 2399
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 9.977447495961229e-05,
+      "loss": 3.3904,
+      "step": 2400
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 9.977431340872375e-05,
+      "loss": 3.3743,
+      "step": 2401
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 9.977415185783522e-05,
+      "loss": 3.3848,
+      "step": 2402
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 9.97739903069467e-05,
+      "loss": 3.3936,
+      "step": 2403
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 9.977382875605816e-05,
+      "loss": 3.4194,
+      "step": 2404
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 9.977366720516963e-05,
+      "loss": 3.3894,
+      "step": 2405
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 9.977350565428111e-05,
+      "loss": 3.3854,
+      "step": 2406
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 9.977334410339257e-05,
+      "loss": 3.4209,
+      "step": 2407
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 9.977318255250405e-05,
+      "loss": 3.5246,
+      "step": 2408
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 9.977302100161552e-05,
+      "loss": 3.4056,
+      "step": 2409
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 9.977285945072698e-05,
+      "loss": 3.4411,
+      "step": 2410
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 9.977269789983846e-05,
+      "loss": 3.5543,
+      "step": 2411
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 9.977253634894992e-05,
+      "loss": 3.762,
+      "step": 2412
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 9.97723747980614e-05,
+      "loss": 3.6924,
+      "step": 2413
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 9.977221324717287e-05,
+      "loss": 5.1231,
+      "step": 2414
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 9.977205169628433e-05,
+      "loss": 3.4251,
+      "step": 2415
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 9.97718901453958e-05,
+      "loss": 3.4652,
+      "step": 2416
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 9.977172859450728e-05,
+      "loss": 3.444,
+      "step": 2417
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 9.977156704361874e-05,
+      "loss": 3.4032,
+      "step": 2418
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 9.977140549273022e-05,
+      "loss": 3.4117,
+      "step": 2419
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 9.977124394184168e-05,
+      "loss": 3.3999,
+      "step": 2420
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 9.977108239095315e-05,
+      "loss": 3.396,
+      "step": 2421
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 9.977092084006463e-05,
+      "loss": 3.3682,
+      "step": 2422
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 9.977075928917609e-05,
+      "loss": 3.3694,
+      "step": 2423
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 9.977059773828757e-05,
+      "loss": 3.4,
+      "step": 2424
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 9.977043618739903e-05,
+      "loss": 3.4227,
+      "step": 2425
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 9.97702746365105e-05,
+      "loss": 3.3839,
+      "step": 2426
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 9.977011308562198e-05,
+      "loss": 3.4268,
+      "step": 2427
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 9.976995153473344e-05,
+      "loss": 3.394,
+      "step": 2428
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 9.976978998384491e-05,
+      "loss": 3.3983,
+      "step": 2429
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 9.976962843295639e-05,
+      "loss": 3.3913,
+      "step": 2430
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 9.976946688206785e-05,
+      "loss": 3.4026,
+      "step": 2431
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 9.976930533117933e-05,
+      "loss": 3.4077,
+      "step": 2432
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 9.97691437802908e-05,
+      "loss": 3.4354,
+      "step": 2433
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 9.976898222940226e-05,
+      "loss": 3.4339,
+      "step": 2434
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 9.976882067851374e-05,
+      "loss": 3.5052,
+      "step": 2435
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 9.976865912762521e-05,
+      "loss": 3.9427,
+      "step": 2436
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 9.976849757673667e-05,
+      "loss": 3.6308,
+      "step": 2437
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 9.976833602584814e-05,
+      "loss": 4.4985,
+      "step": 2438
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 9.976817447495962e-05,
+      "loss": 5.2888,
+      "step": 2439
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 9.976801292407109e-05,
+      "loss": 3.4772,
+      "step": 2440
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 9.976785137318255e-05,
+      "loss": 3.4269,
+      "step": 2441
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 9.976768982229404e-05,
+      "loss": 3.4189,
+      "step": 2442
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 9.97675282714055e-05,
+      "loss": 3.4379,
+      "step": 2443
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 9.976736672051696e-05,
+      "loss": 3.394,
+      "step": 2444
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 9.976720516962845e-05,
+      "loss": 3.3755,
+      "step": 2445
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 9.976704361873991e-05,
+      "loss": 3.4178,
+      "step": 2446
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 9.976688206785137e-05,
+      "loss": 3.3834,
+      "step": 2447
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 9.976672051696286e-05,
+      "loss": 3.3951,
+      "step": 2448
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 9.976655896607432e-05,
+      "loss": 3.3534,
+      "step": 2449
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 9.976639741518578e-05,
+      "loss": 3.3831,
+      "step": 2450
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 9.976623586429727e-05,
+      "loss": 3.3801,
+      "step": 2451
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 9.976607431340873e-05,
+      "loss": 3.3844,
+      "step": 2452
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 9.97659127625202e-05,
+      "loss": 3.3988,
+      "step": 2453
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 9.976575121163167e-05,
+      "loss": 3.3935,
+      "step": 2454
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 9.976558966074314e-05,
+      "loss": 3.4046,
+      "step": 2455
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 9.97654281098546e-05,
+      "loss": 3.4007,
+      "step": 2456
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 9.976526655896608e-05,
+      "loss": 3.404,
+      "step": 2457
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 9.976510500807756e-05,
+      "loss": 3.4078,
+      "step": 2458
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 9.976494345718902e-05,
+      "loss": 3.4418,
+      "step": 2459
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 9.976478190630049e-05,
+      "loss": 3.426,
+      "step": 2460
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 9.976462035541197e-05,
+      "loss": 3.9064,
+      "step": 2461
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 9.976445880452343e-05,
+      "loss": 3.7009,
+      "step": 2462
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 9.97642972536349e-05,
+      "loss": 3.8488,
+      "step": 2463
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 9.976413570274638e-05,
+      "loss": 5.911,
+      "step": 2464
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 9.976397415185784e-05,
+      "loss": 3.5043,
+      "step": 2465
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 9.976381260096932e-05,
+      "loss": 3.4579,
+      "step": 2466
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 9.976365105008078e-05,
+      "loss": 3.4174,
+      "step": 2467
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 9.976348949919225e-05,
+      "loss": 3.4184,
+      "step": 2468
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 9.976332794830371e-05,
+      "loss": 3.4015,
+      "step": 2469
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 9.976316639741519e-05,
+      "loss": 3.3738,
+      "step": 2470
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 9.976300484652666e-05,
+      "loss": 3.3791,
+      "step": 2471
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 9.976284329563812e-05,
+      "loss": 3.3726,
+      "step": 2472
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 9.97626817447496e-05,
+      "loss": 3.3632,
+      "step": 2473
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 9.976252019386107e-05,
+      "loss": 3.3756,
+      "step": 2474
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 9.976235864297254e-05,
+      "loss": 3.3719,
+      "step": 2475
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 9.976219709208401e-05,
+      "loss": 3.3708,
+      "step": 2476
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 9.976203554119549e-05,
+      "loss": 3.3846,
+      "step": 2477
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 9.976187399030695e-05,
+      "loss": 3.3912,
+      "step": 2478
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 9.976171243941842e-05,
+      "loss": 3.3919,
+      "step": 2479
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 9.976155088852988e-05,
+      "loss": 3.3911,
+      "step": 2480
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 9.976138933764136e-05,
+      "loss": 3.4317,
+      "step": 2481
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 9.976122778675283e-05,
+      "loss": 3.3929,
+      "step": 2482
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 9.97610662358643e-05,
+      "loss": 3.4376,
+      "step": 2483
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 9.976090468497577e-05,
+      "loss": 3.4719,
+      "step": 2484
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 9.976074313408725e-05,
+      "loss": 3.6938,
+      "step": 2485
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 9.976058158319871e-05,
+      "loss": 3.6373,
+      "step": 2486
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 9.976042003231018e-05,
+      "loss": 3.597,
+      "step": 2487
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 9.976025848142166e-05,
+      "loss": 4.4762,
+      "step": 2488
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 9.976009693053312e-05,
+      "loss": 6.6753,
+      "step": 2489
+    },
+    {
+      "epoch": 19.0,
+      "eval_loss": 4.967015266418457,
+      "eval_runtime": 37.9609,
+      "eval_samples_per_second": 22.023,
+      "eval_steps_per_second": 0.711,
+      "eval_wer": 1.0,
+      "step": 2489
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 9.97599353796446e-05,
+      "loss": 3.4625,
+      "step": 2490
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 9.975977382875607e-05,
+      "loss": 3.4501,
+      "step": 2491
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 9.975961227786753e-05,
+      "loss": 3.4751,
+      "step": 2492
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 9.975945072697899e-05,
+      "loss": 3.3902,
+      "step": 2493
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 9.975928917609048e-05,
+      "loss": 3.3771,
+      "step": 2494
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 9.975912762520194e-05,
+      "loss": 3.3677,
+      "step": 2495
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 9.97589660743134e-05,
+      "loss": 3.3828,
+      "step": 2496
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 9.975880452342489e-05,
+      "loss": 3.7759,
+      "step": 2497
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 9.975864297253635e-05,
+      "loss": 3.3855,
+      "step": 2498
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 9.975848142164782e-05,
+      "loss": 3.38,
+      "step": 2499
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 9.97583198707593e-05,
+      "loss": 3.3825,
+      "step": 2500
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 9.975815831987077e-05,
+      "loss": 3.3733,
+      "step": 2501
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 9.975799676898223e-05,
+      "loss": 3.3752,
+      "step": 2502
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 9.975783521809372e-05,
+      "loss": 3.3849,
+      "step": 2503
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 9.975767366720518e-05,
+      "loss": 3.3878,
+      "step": 2504
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 9.975751211631664e-05,
+      "loss": 3.4139,
+      "step": 2505
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 9.975735056542811e-05,
+      "loss": 3.4064,
+      "step": 2506
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 9.975718901453959e-05,
+      "loss": 3.4747,
+      "step": 2507
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 9.975702746365105e-05,
+      "loss": 3.4238,
+      "step": 2508
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 9.975686591276253e-05,
+      "loss": 3.5044,
+      "step": 2509
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 9.9756704361874e-05,
+      "loss": 3.5413,
+      "step": 2510
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 9.975654281098546e-05,
+      "loss": 3.4993,
+      "step": 2511
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 9.975638126009694e-05,
+      "loss": 3.6455,
+      "step": 2512
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 9.975621970920841e-05,
+      "loss": 3.761,
+      "step": 2513
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 9.975605815831987e-05,
+      "loss": 5.178,
+      "step": 2514
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 9.975589660743135e-05,
+      "loss": 3.4367,
+      "step": 2515
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 9.975573505654282e-05,
+      "loss": 3.4529,
+      "step": 2516
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 9.975557350565429e-05,
+      "loss": 3.455,
+      "step": 2517
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 9.975541195476576e-05,
+      "loss": 3.4163,
+      "step": 2518
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 9.975525040387722e-05,
+      "loss": 3.3843,
+      "step": 2519
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 9.97550888529887e-05,
+      "loss": 3.3635,
+      "step": 2520
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 9.975492730210016e-05,
+      "loss": 3.3773,
+      "step": 2521
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 9.975476575121163e-05,
+      "loss": 3.3772,
+      "step": 2522
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 9.975460420032311e-05,
+      "loss": 3.3577,
+      "step": 2523
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 9.975444264943457e-05,
+      "loss": 3.3865,
+      "step": 2524
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 9.975428109854605e-05,
+      "loss": 3.3666,
+      "step": 2525
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 9.975411954765752e-05,
+      "loss": 3.3901,
+      "step": 2526
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 9.975395799676898e-05,
+      "loss": 3.4064,
+      "step": 2527
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 9.975379644588046e-05,
+      "loss": 3.3754,
+      "step": 2528
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 9.975363489499193e-05,
+      "loss": 3.4597,
+      "step": 2529
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 9.97534733441034e-05,
+      "loss": 3.4076,
+      "step": 2530
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 9.975331179321487e-05,
+      "loss": 3.4302,
+      "step": 2531
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 9.975315024232634e-05,
+      "loss": 3.4385,
+      "step": 2532
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 9.97529886914378e-05,
+      "loss": 3.4551,
+      "step": 2533
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 9.975282714054928e-05,
+      "loss": 3.4475,
+      "step": 2534
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 9.975266558966074e-05,
+      "loss": 3.4292,
+      "step": 2535
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 9.975250403877222e-05,
+      "loss": 3.458,
+      "step": 2536
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 9.975234248788369e-05,
+      "loss": 3.5835,
+      "step": 2537
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 9.975218093699515e-05,
+      "loss": 4.3801,
+      "step": 2538
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 9.975201938610663e-05,
+      "loss": 5.2584,
+      "step": 2539
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 9.97518578352181e-05,
+      "loss": 3.5006,
+      "step": 2540
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 9.975169628432956e-05,
+      "loss": 3.4702,
+      "step": 2541
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 9.975153473344104e-05,
+      "loss": 3.4337,
+      "step": 2542
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 9.975137318255251e-05,
+      "loss": 3.428,
+      "step": 2543
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 9.975121163166398e-05,
+      "loss": 3.3885,
+      "step": 2544
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 9.975105008077545e-05,
+      "loss": 3.3825,
+      "step": 2545
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 9.975088852988693e-05,
+      "loss": 3.393,
+      "step": 2546
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 9.975072697899839e-05,
+      "loss": 3.3827,
+      "step": 2547
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 9.975056542810985e-05,
+      "loss": 3.3749,
+      "step": 2548
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 9.975040387722134e-05,
+      "loss": 3.3612,
+      "step": 2549
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 9.97502423263328e-05,
+      "loss": 3.3756,
+      "step": 2550
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 9.975008077544426e-05,
+      "loss": 4.2908,
+      "step": 2551
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 9.974991922455575e-05,
+      "loss": 3.3676,
+      "step": 2552
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 9.974975767366721e-05,
+      "loss": 3.369,
+      "step": 2553
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 9.974959612277867e-05,
+      "loss": 3.4014,
+      "step": 2554
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 9.974943457189015e-05,
+      "loss": 3.375,
+      "step": 2555
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 9.974927302100162e-05,
+      "loss": 3.3959,
+      "step": 2556
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 9.974911147011308e-05,
+      "loss": 3.4226,
+      "step": 2557
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 9.974894991922456e-05,
+      "loss": 3.4247,
+      "step": 2558
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 9.974878836833603e-05,
+      "loss": 3.4926,
+      "step": 2559
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 9.97486268174475e-05,
+      "loss": 3.443,
+      "step": 2560
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 9.974846526655897e-05,
+      "loss": 3.4672,
+      "step": 2561
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 9.974830371567045e-05,
+      "loss": 4.5105,
+      "step": 2562
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 9.974814216478191e-05,
+      "loss": 4.1278,
+      "step": 2563
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 9.974798061389338e-05,
+      "loss": 4.8835,
+      "step": 2564
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 9.974781906300486e-05,
+      "loss": 3.5905,
+      "step": 2565
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 9.974765751211632e-05,
+      "loss": 3.4835,
+      "step": 2566
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 9.97474959612278e-05,
+      "loss": 3.4353,
+      "step": 2567
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 9.974733441033927e-05,
+      "loss": 3.4341,
+      "step": 2568
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 9.974717285945073e-05,
+      "loss": 3.4114,
+      "step": 2569
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 9.97470113085622e-05,
+      "loss": 3.3655,
+      "step": 2570
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 9.974684975767368e-05,
+      "loss": 3.3844,
+      "step": 2571
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 9.974668820678514e-05,
+      "loss": 3.354,
+      "step": 2572
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 9.97465266558966e-05,
+      "loss": 3.3741,
+      "step": 2573
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 9.974636510500808e-05,
+      "loss": 3.3687,
+      "step": 2574
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 9.974620355411955e-05,
+      "loss": 3.3734,
+      "step": 2575
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 9.974604200323102e-05,
+      "loss": 3.3866,
+      "step": 2576
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 9.974588045234249e-05,
+      "loss": 3.4057,
+      "step": 2577
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 9.974571890145397e-05,
+      "loss": 3.3888,
+      "step": 2578
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 9.974555735056543e-05,
+      "loss": 3.4033,
+      "step": 2579
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 9.97453957996769e-05,
+      "loss": 3.3936,
+      "step": 2580
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 9.974523424878838e-05,
+      "loss": 3.4037,
+      "step": 2581
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 9.974507269789984e-05,
+      "loss": 3.426,
+      "step": 2582
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 9.974491114701131e-05,
+      "loss": 3.4597,
+      "step": 2583
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 9.974474959612279e-05,
+      "loss": 3.4577,
+      "step": 2584
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 9.974458804523425e-05,
+      "loss": 3.6236,
+      "step": 2585
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 9.974442649434573e-05,
+      "loss": 3.4944,
+      "step": 2586
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 9.974426494345719e-05,
+      "loss": 3.6443,
+      "step": 2587
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 9.974410339256866e-05,
+      "loss": 4.0037,
+      "step": 2588
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 9.974394184168014e-05,
+      "loss": 5.5272,
+      "step": 2589
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 9.97437802907916e-05,
+      "loss": 3.4626,
+      "step": 2590
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 9.974361873990307e-05,
+      "loss": 3.4488,
+      "step": 2591
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 9.974345718901455e-05,
+      "loss": 3.4804,
+      "step": 2592
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 9.974329563812601e-05,
+      "loss": 3.4073,
+      "step": 2593
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 9.974313408723749e-05,
+      "loss": 3.3686,
+      "step": 2594
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 9.974297253634896e-05,
+      "loss": 3.3827,
+      "step": 2595
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 9.974281098546042e-05,
+      "loss": 3.407,
+      "step": 2596
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 9.97426494345719e-05,
+      "loss": 3.3804,
+      "step": 2597
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 9.974248788368337e-05,
+      "loss": 3.3754,
+      "step": 2598
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 9.974232633279483e-05,
+      "loss": 3.3859,
+      "step": 2599
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 9.97421647819063e-05,
+      "loss": 3.3815,
+      "step": 2600
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 9.974200323101778e-05,
+      "loss": 3.3838,
+      "step": 2601
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 9.974184168012925e-05,
+      "loss": 3.3533,
+      "step": 2602
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 9.97416801292407e-05,
+      "loss": 3.3858,
+      "step": 2603
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 9.97415185783522e-05,
+      "loss": 3.3711,
+      "step": 2604
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 9.974135702746366e-05,
+      "loss": 3.3866,
+      "step": 2605
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 9.974119547657512e-05,
+      "loss": 3.3836,
+      "step": 2606
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 9.97410339256866e-05,
+      "loss": 3.414,
+      "step": 2607
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 9.974087237479807e-05,
+      "loss": 3.4791,
+      "step": 2608
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 9.974071082390953e-05,
+      "loss": 3.4612,
+      "step": 2609
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 9.9740549273021e-05,
+      "loss": 3.4849,
+      "step": 2610
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 9.974038772213248e-05,
+      "loss": 3.819,
+      "step": 2611
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 9.974022617124394e-05,
+      "loss": 4.0566,
+      "step": 2612
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 9.974006462035542e-05,
+      "loss": 5.1,
+      "step": 2613
+    },
+    {
+      "epoch": 21.0,
+      "eval_loss": 4.653445243835449,
+      "eval_runtime": 38.8714,
+      "eval_samples_per_second": 21.507,
+      "eval_steps_per_second": 0.695,
+      "eval_wer": 1.0,
+      "step": 2613
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 9.973990306946689e-05,
+      "loss": 3.4468,
+      "step": 2614
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 9.973974151857835e-05,
+      "loss": 3.4202,
+      "step": 2615
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 9.973957996768983e-05,
+      "loss": 3.3857,
+      "step": 2616
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 9.97394184168013e-05,
+      "loss": 3.3873,
+      "step": 2617
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 9.973925686591276e-05,
+      "loss": 3.3983,
+      "step": 2618
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 9.973909531502424e-05,
+      "loss": 3.3773,
+      "step": 2619
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 9.973893376413571e-05,
+      "loss": 3.3826,
+      "step": 2620
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 9.973877221324718e-05,
+      "loss": 3.3893,
+      "step": 2621
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 9.973861066235865e-05,
+      "loss": 3.3842,
+      "step": 2622
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 9.973844911147013e-05,
+      "loss": 3.3979,
+      "step": 2623
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 9.973828756058159e-05,
+      "loss": 3.401,
+      "step": 2624
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 9.973812600969305e-05,
+      "loss": 3.3977,
+      "step": 2625
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 9.973796445880454e-05,
+      "loss": 3.3874,
+      "step": 2626
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 9.9737802907916e-05,
+      "loss": 3.3781,
+      "step": 2627
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 9.973764135702746e-05,
+      "loss": 3.385,
+      "step": 2628
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 9.973747980613894e-05,
+      "loss": 3.4236,
+      "step": 2629
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 9.973731825525041e-05,
+      "loss": 3.3881,
+      "step": 2630
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 9.973715670436187e-05,
+      "loss": 3.3661,
+      "step": 2631
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 9.973699515347335e-05,
+      "loss": 3.3944,
+      "step": 2632
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 9.973683360258482e-05,
+      "loss": 3.4197,
+      "step": 2633
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 9.973667205169628e-05,
+      "loss": 3.5153,
+      "step": 2634
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 9.973651050080776e-05,
+      "loss": 3.7084,
+      "step": 2635
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 9.973634894991923e-05,
+      "loss": 3.8485,
+      "step": 2636
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 9.97361873990307e-05,
+      "loss": 3.7631,
+      "step": 2637
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 9.973602584814217e-05,
+      "loss": 4.821,
+      "step": 2638
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 9.973586429725365e-05,
+      "loss": 3.4297,
+      "step": 2639
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 9.973570274636511e-05,
+      "loss": 3.4193,
+      "step": 2640
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 9.973554119547658e-05,
+      "loss": 3.4189,
+      "step": 2641
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 9.973537964458804e-05,
+      "loss": 3.3946,
+      "step": 2642
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 9.973521809369952e-05,
+      "loss": 3.3684,
+      "step": 2643
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 9.9735056542811e-05,
+      "loss": 3.3775,
+      "step": 2644
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 9.973489499192246e-05,
+      "loss": 3.3593,
+      "step": 2645
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 9.973473344103393e-05,
+      "loss": 3.3805,
+      "step": 2646
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 9.97345718901454e-05,
+      "loss": 3.3626,
+      "step": 2647
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 9.973441033925687e-05,
+      "loss": 3.377,
+      "step": 2648
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 9.973424878836834e-05,
+      "loss": 3.3959,
+      "step": 2649
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 9.973408723747982e-05,
+      "loss": 3.3876,
+      "step": 2650
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 9.973392568659128e-05,
+      "loss": 3.3985,
+      "step": 2651
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 9.973376413570275e-05,
+      "loss": 3.3968,
+      "step": 2652
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 9.973360258481423e-05,
+      "loss": 3.3909,
+      "step": 2653
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 9.973344103392569e-05,
+      "loss": 3.3792,
+      "step": 2654
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 9.973327948303715e-05,
+      "loss": 3.3881,
+      "step": 2655
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 9.973311793214864e-05,
+      "loss": 3.3909,
+      "step": 2656
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 9.97329563812601e-05,
+      "loss": 3.4028,
+      "step": 2657
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 9.973279483037156e-05,
+      "loss": 3.4637,
+      "step": 2658
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 9.973263327948304e-05,
+      "loss": 3.5684,
+      "step": 2659
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 9.973247172859451e-05,
+      "loss": 3.5673,
+      "step": 2660
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 9.973231017770598e-05,
+      "loss": 3.5892,
+      "step": 2661
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 9.973214862681745e-05,
+      "loss": 4.3663,
+      "step": 2662
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 9.973198707592893e-05,
+      "loss": 6.2201,
+      "step": 2663
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 9.973182552504039e-05,
+      "loss": 3.4924,
+      "step": 2664
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 9.973166397415186e-05,
+      "loss": 3.4442,
+      "step": 2665
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 9.973150242326334e-05,
+      "loss": 3.4098,
+      "step": 2666
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 9.97313408723748e-05,
+      "loss": 3.3916,
+      "step": 2667
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 9.973117932148627e-05,
+      "loss": 3.384,
+      "step": 2668
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 9.973101777059775e-05,
+      "loss": 3.384,
+      "step": 2669
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 9.973085621970921e-05,
+      "loss": 3.3746,
+      "step": 2670
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 9.973069466882069e-05,
+      "loss": 3.3835,
+      "step": 2671
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 9.973053311793216e-05,
+      "loss": 3.3665,
+      "step": 2672
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 9.973037156704362e-05,
+      "loss": 3.3755,
+      "step": 2673
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 9.973021001615508e-05,
+      "loss": 3.4112,
+      "step": 2674
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 9.973004846526657e-05,
+      "loss": 3.3803,
+      "step": 2675
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 9.972988691437803e-05,
+      "loss": 3.3797,
+      "step": 2676
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 9.97297253634895e-05,
+      "loss": 3.4046,
+      "step": 2677
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 9.972956381260098e-05,
+      "loss": 3.4318,
+      "step": 2678
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 9.972940226171245e-05,
+      "loss": 3.3904,
+      "step": 2679
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 9.97292407108239e-05,
+      "loss": 3.3822,
+      "step": 2680
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 9.972907915993538e-05,
+      "loss": 3.4215,
+      "step": 2681
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 9.972891760904686e-05,
+      "loss": 3.3861,
+      "step": 2682
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 9.972875605815832e-05,
+      "loss": 3.4016,
+      "step": 2683
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 9.97285945072698e-05,
+      "loss": 3.5072,
+      "step": 2684
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 9.972843295638127e-05,
+      "loss": 3.6512,
+      "step": 2685
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 9.972827140549273e-05,
+      "loss": 4.0526,
+      "step": 2686
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 9.97281098546042e-05,
+      "loss": 4.296,
+      "step": 2687
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 9.972794830371568e-05,
+      "loss": 5.4486,
+      "step": 2688
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 9.972778675282714e-05,
+      "loss": 3.5229,
+      "step": 2689
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 9.972762520193862e-05,
+      "loss": 3.4806,
+      "step": 2690
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 9.972746365105009e-05,
+      "loss": 3.4172,
+      "step": 2691
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 9.972730210016155e-05,
+      "loss": 3.3832,
+      "step": 2692
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 9.972714054927303e-05,
+      "loss": 3.3795,
+      "step": 2693
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 9.972697899838449e-05,
+      "loss": 3.3758,
+      "step": 2694
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 9.972681744749596e-05,
+      "loss": 3.3869,
+      "step": 2695
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 9.972665589660744e-05,
+      "loss": 3.3915,
+      "step": 2696
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 9.97264943457189e-05,
+      "loss": 3.3878,
+      "step": 2697
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 9.972633279483038e-05,
+      "loss": 3.3692,
+      "step": 2698
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 9.972617124394185e-05,
+      "loss": 3.3958,
+      "step": 2699
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 9.972600969305331e-05,
+      "loss": 3.3674,
+      "step": 2700
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 9.972584814216479e-05,
+      "loss": 3.3865,
+      "step": 2701
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 9.972568659127626e-05,
+      "loss": 3.3897,
+      "step": 2702
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 9.972552504038772e-05,
+      "loss": 3.3673,
+      "step": 2703
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 9.97253634894992e-05,
+      "loss": 3.3729,
+      "step": 2704
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 9.972520193861067e-05,
+      "loss": 3.3821,
+      "step": 2705
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 9.972504038772214e-05,
+      "loss": 3.3947,
+      "step": 2706
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 9.972487883683361e-05,
+      "loss": 3.3782,
+      "step": 2707
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 9.972471728594509e-05,
+      "loss": 3.4019,
+      "step": 2708
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 9.972455573505655e-05,
+      "loss": 3.4954,
+      "step": 2709
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 9.972439418416801e-05,
+      "loss": 3.8591,
+      "step": 2710
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 9.972423263327948e-05,
+      "loss": 3.9441,
+      "step": 2711
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 9.972407108239096e-05,
+      "loss": 3.764,
+      "step": 2712
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 9.972390953150242e-05,
+      "loss": 5.9534,
+      "step": 2713
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 9.97237479806139e-05,
+      "loss": 3.4387,
+      "step": 2714
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 9.972358642972537e-05,
+      "loss": 3.436,
+      "step": 2715
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 9.972342487883683e-05,
+      "loss": 3.409,
+      "step": 2716
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 9.972326332794831e-05,
+      "loss": 3.4173,
+      "step": 2717
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 9.972310177705978e-05,
+      "loss": 3.3772,
+      "step": 2718
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 9.972294022617124e-05,
+      "loss": 3.3598,
+      "step": 2719
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 9.972277867528272e-05,
+      "loss": 3.3856,
+      "step": 2720
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 9.97226171243942e-05,
+      "loss": 3.3771,
+      "step": 2721
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 9.972245557350566e-05,
+      "loss": 3.3784,
+      "step": 2722
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 9.972229402261713e-05,
+      "loss": 3.3823,
+      "step": 2723
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 9.97221324717286e-05,
+      "loss": 3.3885,
+      "step": 2724
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 9.972197092084007e-05,
+      "loss": 3.3948,
+      "step": 2725
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 9.972180936995153e-05,
+      "loss": 3.3889,
+      "step": 2726
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 9.972164781906302e-05,
+      "loss": 3.394,
+      "step": 2727
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 9.972148626817448e-05,
+      "loss": 3.4012,
+      "step": 2728
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 9.972132471728594e-05,
+      "loss": 3.3926,
+      "step": 2729
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 9.972116316639743e-05,
+      "loss": 3.3902,
+      "step": 2730
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 9.972100161550889e-05,
+      "loss": 3.432,
+      "step": 2731
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 9.972084006462035e-05,
+      "loss": 3.4049,
+      "step": 2732
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 9.972067851373184e-05,
+      "loss": 3.5215,
+      "step": 2733
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 9.97205169628433e-05,
+      "loss": 3.5147,
+      "step": 2734
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 9.972035541195476e-05,
+      "loss": 3.6495,
+      "step": 2735
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 9.972019386106624e-05,
+      "loss": 4.1999,
+      "step": 2736
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 9.972003231017771e-05,
+      "loss": 4.1069,
+      "step": 2737
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 9.971987075928918e-05,
+      "loss": 5.0655,
+      "step": 2738
+    },
+    {
+      "epoch": 21.0,
+      "eval_loss": 4.395701885223389,
+      "eval_runtime": 36.102,
+      "eval_samples_per_second": 23.157,
+      "eval_steps_per_second": 0.748,
+      "eval_wer": 1.0,
+      "step": 2738
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 9.971970920840065e-05,
+      "loss": 3.5447,
+      "step": 2739
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 9.971954765751213e-05,
+      "loss": 3.4406,
+      "step": 2740
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 9.971938610662359e-05,
+      "loss": 3.4056,
+      "step": 2741
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 9.971922455573506e-05,
+      "loss": 3.3886,
+      "step": 2742
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 9.971906300484654e-05,
+      "loss": 3.3982,
+      "step": 2743
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 9.9718901453958e-05,
+      "loss": 3.3784,
+      "step": 2744
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 9.971873990306947e-05,
+      "loss": 3.3852,
+      "step": 2745
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 9.971857835218095e-05,
+      "loss": 3.4095,
+      "step": 2746
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 9.971841680129241e-05,
+      "loss": 3.3863,
+      "step": 2747
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 9.971825525040389e-05,
+      "loss": 3.3807,
+      "step": 2748
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 9.971809369951535e-05,
+      "loss": 3.3928,
+      "step": 2749
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 9.971793214862682e-05,
+      "loss": 3.3901,
+      "step": 2750
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 9.97177705977383e-05,
+      "loss": 3.3797,
+      "step": 2751
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 9.971760904684976e-05,
+      "loss": 3.405,
+      "step": 2752
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 9.971744749596123e-05,
+      "loss": 3.4716,
+      "step": 2753
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 9.971728594507271e-05,
+      "loss": 3.394,
+      "step": 2754
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 9.971712439418417e-05,
+      "loss": 3.4194,
+      "step": 2755
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 9.971696284329565e-05,
+      "loss": 3.3884,
+      "step": 2756
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 9.971680129240712e-05,
+      "loss": 3.5228,
+      "step": 2757
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 9.971663974151858e-05,
+      "loss": 3.5231,
+      "step": 2758
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 9.971647819063006e-05,
+      "loss": 3.5085,
+      "step": 2759
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 9.971631663974152e-05,
+      "loss": 3.617,
+      "step": 2760
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 9.971615508885299e-05,
+      "loss": 4.0493,
+      "step": 2761
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 9.971599353796445e-05,
+      "loss": 3.8282,
+      "step": 2762
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 9.971583198707593e-05,
+      "loss": 6.1947,
+      "step": 2763
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 9.97156704361874e-05,
+      "loss": 3.4812,
+      "step": 2764
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 9.971550888529887e-05,
+      "loss": 3.4663,
+      "step": 2765
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 9.971534733441034e-05,
+      "loss": 3.4026,
+      "step": 2766
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 9.971518578352182e-05,
+      "loss": 3.3853,
+      "step": 2767
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 9.971502423263328e-05,
+      "loss": 3.3853,
+      "step": 2768
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 9.971486268174475e-05,
+      "loss": 3.3832,
+      "step": 2769
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 9.971470113085623e-05,
+      "loss": 3.3781,
+      "step": 2770
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 9.971453957996769e-05,
+      "loss": 3.3886,
+      "step": 2771
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 9.971437802907916e-05,
+      "loss": 3.3704,
+      "step": 2772
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 9.971421647819064e-05,
+      "loss": 3.3947,
+      "step": 2773
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 9.97140549273021e-05,
+      "loss": 3.3831,
+      "step": 2774
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 9.971389337641358e-05,
+      "loss": 3.4166,
+      "step": 2775
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 9.971373182552505e-05,
+      "loss": 3.4281,
+      "step": 2776
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 9.971357027463651e-05,
+      "loss": 3.4012,
+      "step": 2777
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 9.971340872374797e-05,
+      "loss": 3.3981,
+      "step": 2778
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 9.971324717285946e-05,
+      "loss": 3.4362,
+      "step": 2779
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 9.971308562197092e-05,
+      "loss": 3.4196,
+      "step": 2780
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 9.971292407108239e-05,
+      "loss": 3.4304,
+      "step": 2781
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 9.971276252019387e-05,
+      "loss": 3.4158,
+      "step": 2782
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 9.971260096930534e-05,
+      "loss": 3.4947,
+      "step": 2783
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 9.97124394184168e-05,
+      "loss": 3.4854,
+      "step": 2784
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 9.971227786752829e-05,
+      "loss": 3.8496,
+      "step": 2785
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 9.971211631663975e-05,
+      "loss": 3.7608,
+      "step": 2786
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 9.971195476575121e-05,
+      "loss": 4.1597,
+      "step": 2787
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 9.971179321486268e-05,
+      "loss": 5.6508,
+      "step": 2788
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 9.971163166397416e-05,
+      "loss": 3.4493,
+      "step": 2789
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 9.971147011308562e-05,
+      "loss": 3.4631,
+      "step": 2790
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 9.97113085621971e-05,
+      "loss": 3.4285,
+      "step": 2791
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 9.971114701130857e-05,
+      "loss": 3.4079,
+      "step": 2792
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 9.971098546042003e-05,
+      "loss": 3.4032,
+      "step": 2793
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 9.971082390953151e-05,
+      "loss": 3.3756,
+      "step": 2794
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 9.971066235864298e-05,
+      "loss": 3.3677,
+      "step": 2795
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 9.971050080775444e-05,
+      "loss": 3.3942,
+      "step": 2796
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 9.971033925686592e-05,
+      "loss": 3.3911,
+      "step": 2797
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 9.97101777059774e-05,
+      "loss": 3.4213,
+      "step": 2798
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 9.971001615508886e-05,
+      "loss": 3.3863,
+      "step": 2799
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 9.970985460420033e-05,
+      "loss": 3.3892,
+      "step": 2800
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 9.97096930533118e-05,
+      "loss": 3.3848,
+      "step": 2801
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 9.970953150242327e-05,
+      "loss": 3.405,
+      "step": 2802
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 9.970936995153474e-05,
+      "loss": 3.3937,
+      "step": 2803
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 9.97092084006462e-05,
+      "loss": 3.4001,
+      "step": 2804
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 9.970904684975768e-05,
+      "loss": 3.4161,
+      "step": 2805
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 9.970888529886915e-05,
+      "loss": 3.4149,
+      "step": 2806
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 9.970872374798062e-05,
+      "loss": 3.4247,
+      "step": 2807
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 9.970856219709209e-05,
+      "loss": 3.4808,
+      "step": 2808
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 9.970840064620357e-05,
+      "loss": 3.4322,
+      "step": 2809
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 9.970823909531503e-05,
+      "loss": 3.4845,
+      "step": 2810
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 9.97080775444265e-05,
+      "loss": 3.9003,
+      "step": 2811
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 9.970791599353796e-05,
+      "loss": 4.2638,
+      "step": 2812
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 9.970775444264944e-05,
+      "loss": 4.548,
+      "step": 2813
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 9.970759289176091e-05,
+      "loss": 3.4714,
+      "step": 2814
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 9.970743134087238e-05,
+      "loss": 3.4712,
+      "step": 2815
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 9.970726978998385e-05,
+      "loss": 3.4295,
+      "step": 2816
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 9.970710823909531e-05,
+      "loss": 3.4043,
+      "step": 2817
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 9.970694668820679e-05,
+      "loss": 3.4081,
+      "step": 2818
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 9.970678513731826e-05,
+      "loss": 3.4053,
+      "step": 2819
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 9.970662358642972e-05,
+      "loss": 3.3695,
+      "step": 2820
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 9.97064620355412e-05,
+      "loss": 3.3877,
+      "step": 2821
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 9.970630048465267e-05,
+      "loss": 3.3884,
+      "step": 2822
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 9.970613893376414e-05,
+      "loss": 3.3644,
+      "step": 2823
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 9.970597738287561e-05,
+      "loss": 3.3739,
+      "step": 2824
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 9.970581583198709e-05,
+      "loss": 3.5161,
+      "step": 2825
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 9.970565428109855e-05,
+      "loss": 3.378,
+      "step": 2826
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 9.970549273021002e-05,
+      "loss": 3.3871,
+      "step": 2827
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 9.97053311793215e-05,
+      "loss": 3.3923,
+      "step": 2828
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 9.970516962843296e-05,
+      "loss": 3.3913,
+      "step": 2829
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 9.970500807754442e-05,
+      "loss": 3.4236,
+      "step": 2830
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 9.970484652665591e-05,
+      "loss": 3.4159,
+      "step": 2831
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 9.970468497576737e-05,
+      "loss": 3.4279,
+      "step": 2832
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 9.970452342487883e-05,
+      "loss": 3.4415,
+      "step": 2833
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 9.970436187399032e-05,
+      "loss": 3.5152,
+      "step": 2834
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 9.970420032310178e-05,
+      "loss": 3.5232,
+      "step": 2835
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 9.970403877221324e-05,
+      "loss": 4.108,
+      "step": 2836
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 9.970387722132473e-05,
+      "loss": 4.1242,
+      "step": 2837
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 9.970371567043619e-05,
+      "loss": 5.9955,
+      "step": 2838
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 9.970355411954765e-05,
+      "loss": 3.5008,
+      "step": 2839
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 9.970339256865914e-05,
+      "loss": 3.4408,
+      "step": 2840
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 9.97032310177706e-05,
+      "loss": 3.4262,
+      "step": 2841
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 9.970306946688207e-05,
+      "loss": 3.4044,
+      "step": 2842
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 9.970290791599354e-05,
+      "loss": 3.3944,
+      "step": 2843
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 9.970274636510502e-05,
+      "loss": 3.3965,
+      "step": 2844
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 9.970258481421648e-05,
+      "loss": 3.3936,
+      "step": 2845
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 9.970242326332795e-05,
+      "loss": 3.3722,
+      "step": 2846
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 9.970226171243943e-05,
+      "loss": 3.3739,
+      "step": 2847
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 9.970210016155089e-05,
+      "loss": 3.3747,
+      "step": 2848
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 9.970193861066236e-05,
+      "loss": 3.3813,
+      "step": 2849
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 9.970177705977384e-05,
+      "loss": 3.3827,
+      "step": 2850
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 9.97016155088853e-05,
+      "loss": 3.3927,
+      "step": 2851
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 9.970145395799678e-05,
+      "loss": 3.3881,
+      "step": 2852
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 9.970129240710825e-05,
+      "loss": 3.4036,
+      "step": 2853
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 9.970113085621971e-05,
+      "loss": 3.3998,
+      "step": 2854
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 9.970096930533119e-05,
+      "loss": 3.4221,
+      "step": 2855
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 9.970080775444265e-05,
+      "loss": 3.401,
+      "step": 2856
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 9.970064620355412e-05,
+      "loss": 3.4053,
+      "step": 2857
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 9.97004846526656e-05,
+      "loss": 3.4697,
+      "step": 2858
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 9.970032310177706e-05,
+      "loss": 3.5169,
+      "step": 2859
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 9.970016155088854e-05,
+      "loss": 3.578,
+      "step": 2860
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 9.970000000000001e-05,
+      "loss": 3.6169,
+      "step": 2861
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 9.969983844911147e-05,
+      "loss": 4.0273,
+      "step": 2862
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 9.969967689822295e-05,
+      "loss": 4.7717,
+      "step": 2863
+    },
+    {
+      "epoch": 22.0,
+      "eval_loss": 4.710836410522461,
+      "eval_runtime": 39.2428,
+      "eval_samples_per_second": 21.278,
+      "eval_steps_per_second": 0.688,
+      "eval_wer": 1.0,
+      "step": 2863
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 9.969951534733441e-05,
+      "loss": 3.4672,
+      "step": 2864
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 9.969935379644588e-05,
+      "loss": 3.4347,
+      "step": 2865
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 9.969919224555736e-05,
+      "loss": 3.4201,
+      "step": 2866
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 9.969903069466882e-05,
+      "loss": 3.3869,
+      "step": 2867
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 9.96988691437803e-05,
+      "loss": 3.3862,
+      "step": 2868
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 9.969870759289176e-05,
+      "loss": 3.3816,
+      "step": 2869
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 9.969854604200323e-05,
+      "loss": 3.4089,
+      "step": 2870
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 9.969838449111471e-05,
+      "loss": 3.3699,
+      "step": 2871
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 9.969822294022617e-05,
+      "loss": 3.3818,
+      "step": 2872
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 9.969806138933764e-05,
+      "loss": 3.3657,
+      "step": 2873
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 9.969789983844912e-05,
+      "loss": 3.3893,
+      "step": 2874
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 9.969773828756058e-05,
+      "loss": 3.3942,
+      "step": 2875
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 9.969757673667206e-05,
+      "loss": 3.4177,
+      "step": 2876
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 9.969741518578353e-05,
+      "loss": 3.3947,
+      "step": 2877
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 9.969725363489499e-05,
+      "loss": 3.3839,
+      "step": 2878
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 9.969709208400647e-05,
+      "loss": 3.3643,
+      "step": 2879
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 9.969693053311794e-05,
+      "loss": 3.4058,
+      "step": 2880
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 9.96967689822294e-05,
+      "loss": 3.434,
+      "step": 2881
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 9.969660743134088e-05,
+      "loss": 3.4404,
+      "step": 2882
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 9.969644588045235e-05,
+      "loss": 3.5301,
+      "step": 2883
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 9.969628432956382e-05,
+      "loss": 3.4364,
+      "step": 2884
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 9.969612277867528e-05,
+      "loss": 3.6236,
+      "step": 2885
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 9.969596122778677e-05,
+      "loss": 3.6576,
+      "step": 2886
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 9.969579967689823e-05,
+      "loss": 4.0947,
+      "step": 2887
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 9.969563812600969e-05,
+      "loss": 5.8602,
+      "step": 2888
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 9.969547657512118e-05,
+      "loss": 3.5211,
+      "step": 2889
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 9.969531502423264e-05,
+      "loss": 3.4539,
+      "step": 2890
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 9.96951534733441e-05,
+      "loss": 3.4338,
+      "step": 2891
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 9.969499192245559e-05,
+      "loss": 3.4263,
+      "step": 2892
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 9.969483037156705e-05,
+      "loss": 3.4004,
+      "step": 2893
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 9.969466882067851e-05,
+      "loss": 3.4049,
+      "step": 2894
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 9.969450726979e-05,
+      "loss": 3.3972,
+      "step": 2895
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 9.969434571890146e-05,
+      "loss": 3.3776,
+      "step": 2896
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 9.969418416801292e-05,
+      "loss": 3.3849,
+      "step": 2897
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 9.96940226171244e-05,
+      "loss": 3.382,
+      "step": 2898
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 9.969386106623587e-05,
+      "loss": 3.3482,
+      "step": 2899
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 9.969369951534733e-05,
+      "loss": 3.398,
+      "step": 2900
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 9.969353796445881e-05,
+      "loss": 3.3839,
+      "step": 2901
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 9.969337641357029e-05,
+      "loss": 3.3803,
+      "step": 2902
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 9.969321486268175e-05,
+      "loss": 3.3836,
+      "step": 2903
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 9.969305331179322e-05,
+      "loss": 3.3929,
+      "step": 2904
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 9.96928917609047e-05,
+      "loss": 3.4115,
+      "step": 2905
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 9.969273021001616e-05,
+      "loss": 3.4186,
+      "step": 2906
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 9.969256865912763e-05,
+      "loss": 3.4561,
+      "step": 2907
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 9.969240710823911e-05,
+      "loss": 3.4414,
+      "step": 2908
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 9.969224555735057e-05,
+      "loss": 3.5185,
+      "step": 2909
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 9.969208400646204e-05,
+      "loss": 3.637,
+      "step": 2910
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 9.96919224555735e-05,
+      "loss": 3.6047,
+      "step": 2911
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 9.969176090468498e-05,
+      "loss": 3.9394,
+      "step": 2912
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 9.969159935379646e-05,
+      "loss": 6.0172,
+      "step": 2913
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 9.969143780290792e-05,
+      "loss": 3.4634,
+      "step": 2914
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 9.969127625201939e-05,
+      "loss": 3.4609,
+      "step": 2915
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 9.969111470113085e-05,
+      "loss": 3.4202,
+      "step": 2916
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 9.969095315024233e-05,
+      "loss": 3.4157,
+      "step": 2917
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 9.96907915993538e-05,
+      "loss": 3.397,
+      "step": 2918
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 9.969063004846527e-05,
+      "loss": 3.3747,
+      "step": 2919
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 9.969046849757674e-05,
+      "loss": 3.4073,
+      "step": 2920
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 9.969030694668822e-05,
+      "loss": 3.3822,
+      "step": 2921
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 9.969014539579968e-05,
+      "loss": 3.3935,
+      "step": 2922
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 9.968998384491115e-05,
+      "loss": 3.3873,
+      "step": 2923
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 9.968982229402261e-05,
+      "loss": 3.3824,
+      "step": 2924
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 9.968966074313409e-05,
+      "loss": 3.3985,
+      "step": 2925
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 9.968949919224556e-05,
+      "loss": 3.3881,
+      "step": 2926
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 9.968933764135703e-05,
+      "loss": 3.3977,
+      "step": 2927
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 9.96891760904685e-05,
+      "loss": 3.3913,
+      "step": 2928
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 9.968901453957998e-05,
+      "loss": 3.3928,
+      "step": 2929
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 9.968885298869144e-05,
+      "loss": 3.4115,
+      "step": 2930
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 9.968869143780291e-05,
+      "loss": 3.4258,
+      "step": 2931
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 9.968852988691439e-05,
+      "loss": 3.466,
+      "step": 2932
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 9.968836833602585e-05,
+      "loss": 3.4269,
+      "step": 2933
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 9.968820678513732e-05,
+      "loss": 3.4687,
+      "step": 2934
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 9.96880452342488e-05,
+      "loss": 3.6543,
+      "step": 2935
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 9.968788368336026e-05,
+      "loss": 3.686,
+      "step": 2936
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 9.968772213247172e-05,
+      "loss": 4.2211,
+      "step": 2937
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 9.968756058158321e-05,
+      "loss": 5.3919,
+      "step": 2938
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 9.968739903069467e-05,
+      "loss": 3.4792,
+      "step": 2939
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 9.968723747980613e-05,
+      "loss": 3.4365,
+      "step": 2940
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 9.968707592891762e-05,
+      "loss": 3.4109,
+      "step": 2941
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 9.968691437802908e-05,
+      "loss": 3.4298,
+      "step": 2942
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 9.968675282714055e-05,
+      "loss": 3.4358,
+      "step": 2943
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 9.968659127625203e-05,
+      "loss": 3.4355,
+      "step": 2944
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 9.96864297253635e-05,
+      "loss": 3.3943,
+      "step": 2945
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 9.968626817447496e-05,
+      "loss": 3.3913,
+      "step": 2946
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 9.968610662358645e-05,
+      "loss": 3.3828,
+      "step": 2947
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 9.968594507269791e-05,
+      "loss": 3.3892,
+      "step": 2948
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 9.968578352180937e-05,
+      "loss": 3.39,
+      "step": 2949
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 9.968562197092084e-05,
+      "loss": 3.3733,
+      "step": 2950
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 9.968546042003232e-05,
+      "loss": 3.3856,
+      "step": 2951
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 9.968529886914378e-05,
+      "loss": 3.3878,
+      "step": 2952
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 9.968513731825526e-05,
+      "loss": 3.3853,
+      "step": 2953
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 9.968497576736673e-05,
+      "loss": 3.3867,
+      "step": 2954
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 9.968481421647819e-05,
+      "loss": 3.4203,
+      "step": 2955
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 9.968465266558967e-05,
+      "loss": 3.4136,
+      "step": 2956
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 9.968449111470114e-05,
+      "loss": 3.4027,
+      "step": 2957
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 9.96843295638126e-05,
+      "loss": 3.4336,
+      "step": 2958
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 9.968416801292408e-05,
+      "loss": 3.455,
+      "step": 2959
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 9.968400646203555e-05,
+      "loss": 3.5638,
+      "step": 2960
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 9.968384491114702e-05,
+      "loss": 3.6982,
+      "step": 2961
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 9.968368336025849e-05,
+      "loss": 4.1518,
+      "step": 2962
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 9.968352180936995e-05,
+      "loss": 5.1366,
+      "step": 2963
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 9.968336025848143e-05,
+      "loss": 3.488,
+      "step": 2964
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 9.96831987075929e-05,
+      "loss": 3.4609,
+      "step": 2965
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 9.968303715670436e-05,
+      "loss": 3.409,
+      "step": 2966
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 9.968287560581584e-05,
+      "loss": 3.4252,
+      "step": 2967
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 9.96827140549273e-05,
+      "loss": 3.3938,
+      "step": 2968
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 9.968255250403878e-05,
+      "loss": 3.3695,
+      "step": 2969
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 9.968239095315025e-05,
+      "loss": 3.3974,
+      "step": 2970
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 9.968222940226171e-05,
+      "loss": 3.394,
+      "step": 2971
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 9.968206785137319e-05,
+      "loss": 3.4182,
+      "step": 2972
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 9.968190630048466e-05,
+      "loss": 3.3794,
+      "step": 2973
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 9.968174474959612e-05,
+      "loss": 3.374,
+      "step": 2974
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 9.96815831987076e-05,
+      "loss": 3.3745,
+      "step": 2975
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 9.968142164781907e-05,
+      "loss": 3.414,
+      "step": 2976
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 9.968126009693053e-05,
+      "loss": 3.3847,
+      "step": 2977
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 9.968109854604201e-05,
+      "loss": 3.3722,
+      "step": 2978
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 9.968093699515347e-05,
+      "loss": 3.3923,
+      "step": 2979
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 9.968077544426495e-05,
+      "loss": 3.4104,
+      "step": 2980
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 9.968061389337642e-05,
+      "loss": 3.4447,
+      "step": 2981
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 9.968045234248788e-05,
+      "loss": 3.43,
+      "step": 2982
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 9.968029079159936e-05,
+      "loss": 3.6307,
+      "step": 2983
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 9.968012924071083e-05,
+      "loss": 3.5688,
+      "step": 2984
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 9.96799676898223e-05,
+      "loss": 3.764,
+      "step": 2985
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 9.967980613893377e-05,
+      "loss": 3.8299,
+      "step": 2986
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 9.967964458804524e-05,
+      "loss": 5.838,
+      "step": 2987
+    },
+    {
+      "epoch": 24.0,
+      "eval_loss": 4.48088264465332,
+      "eval_runtime": 36.7477,
+      "eval_samples_per_second": 22.723,
+      "eval_steps_per_second": 0.735,
+      "eval_wer": 1.0,
+      "step": 2987
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 9.96794830371567e-05,
+      "loss": 3.5044,
+      "step": 2988
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 9.967932148626818e-05,
+      "loss": 3.478,
+      "step": 2989
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 9.967915993537966e-05,
+      "loss": 3.4067,
+      "step": 2990
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 9.967899838449112e-05,
+      "loss": 3.5408,
+      "step": 2991
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 9.967883683360258e-05,
+      "loss": 3.3875,
+      "step": 2992
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 9.967867528271407e-05,
+      "loss": 3.3913,
+      "step": 2993
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 9.967851373182553e-05,
+      "loss": 3.3695,
+      "step": 2994
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 9.967835218093699e-05,
+      "loss": 3.3612,
+      "step": 2995
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 9.967819063004848e-05,
+      "loss": 3.3702,
+      "step": 2996
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 9.967802907915994e-05,
+      "loss": 3.3902,
+      "step": 2997
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 9.96778675282714e-05,
+      "loss": 3.3696,
+      "step": 2998
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 9.967770597738289e-05,
+      "loss": 3.373,
+      "step": 2999
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 9.967754442649435e-05,
+      "loss": 3.3945,
+      "step": 3000
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 9.967738287560581e-05,
+      "loss": 3.3987,
+      "step": 3001
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 9.967722132471729e-05,
+      "loss": 3.4012,
+      "step": 3002
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 9.967705977382876e-05,
+      "loss": 3.403,
+      "step": 3003
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 9.967689822294023e-05,
+      "loss": 3.3783,
+      "step": 3004
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 9.96767366720517e-05,
+      "loss": 3.4035,
+      "step": 3005
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 9.967657512116318e-05,
+      "loss": 3.4478,
+      "step": 3006
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 9.967641357027464e-05,
+      "loss": 3.4171,
+      "step": 3007
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 9.967625201938611e-05,
+      "loss": 3.4625,
+      "step": 3008
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 9.967609046849759e-05,
+      "loss": 3.5077,
+      "step": 3009
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 9.967592891760905e-05,
+      "loss": 3.577,
+      "step": 3010
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 9.967576736672052e-05,
+      "loss": 4.3753,
+      "step": 3011
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 9.9675605815832e-05,
+      "loss": 5.2648,
+      "step": 3012
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 9.967544426494346e-05,
+      "loss": 3.4105,
+      "step": 3013
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 9.967528271405494e-05,
+      "loss": 3.4486,
+      "step": 3014
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 9.967512116316641e-05,
+      "loss": 3.4099,
+      "step": 3015
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 9.967495961227787e-05,
+      "loss": 3.4368,
+      "step": 3016
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 9.967479806138933e-05,
+      "loss": 3.3944,
+      "step": 3017
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 9.967463651050081e-05,
+      "loss": 3.3913,
+      "step": 3018
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 9.967447495961228e-05,
+      "loss": 3.4029,
+      "step": 3019
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 9.967431340872375e-05,
+      "loss": 3.3947,
+      "step": 3020
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 9.967415185783522e-05,
+      "loss": 3.3768,
+      "step": 3021
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 9.96739903069467e-05,
+      "loss": 3.3722,
+      "step": 3022
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 9.967382875605816e-05,
+      "loss": 3.3855,
+      "step": 3023
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 9.967366720516963e-05,
+      "loss": 3.3783,
+      "step": 3024
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 9.967350565428111e-05,
+      "loss": 3.3979,
+      "step": 3025
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 9.967334410339257e-05,
+      "loss": 3.4023,
+      "step": 3026
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 9.967318255250404e-05,
+      "loss": 3.3839,
+      "step": 3027
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 9.967302100161552e-05,
+      "loss": 3.4244,
+      "step": 3028
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 9.967285945072698e-05,
+      "loss": 3.4168,
+      "step": 3029
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 9.967269789983846e-05,
+      "loss": 3.4073,
+      "step": 3030
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 9.967253634894992e-05,
+      "loss": 3.4112,
+      "step": 3031
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 9.967237479806139e-05,
+      "loss": 3.464,
+      "step": 3032
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 9.967221324717287e-05,
+      "loss": 3.4549,
+      "step": 3033
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 9.967205169628433e-05,
+      "loss": 3.5703,
+      "step": 3034
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 9.96718901453958e-05,
+      "loss": 3.5683,
+      "step": 3035
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 9.967172859450728e-05,
+      "loss": 3.7426,
+      "step": 3036
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 9.967156704361874e-05,
+      "loss": 4.5105,
+      "step": 3037
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 9.967140549273022e-05,
+      "loss": 3.4899,
+      "step": 3038
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 9.967124394184169e-05,
+      "loss": 3.4567,
+      "step": 3039
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 9.967108239095315e-05,
+      "loss": 3.4152,
+      "step": 3040
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 9.967092084006463e-05,
+      "loss": 3.3946,
+      "step": 3041
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 9.96707592891761e-05,
+      "loss": 3.4025,
+      "step": 3042
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 9.967059773828756e-05,
+      "loss": 3.3912,
+      "step": 3043
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 9.967043618739902e-05,
+      "loss": 3.3921,
+      "step": 3044
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 9.967027463651051e-05,
+      "loss": 3.3647,
+      "step": 3045
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 9.967011308562198e-05,
+      "loss": 3.3675,
+      "step": 3046
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 9.966995153473344e-05,
+      "loss": 3.3763,
+      "step": 3047
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 9.966978998384493e-05,
+      "loss": 3.3919,
+      "step": 3048
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 9.966962843295639e-05,
+      "loss": 3.4341,
+      "step": 3049
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 9.966946688206785e-05,
+      "loss": 3.4069,
+      "step": 3050
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 9.966930533117934e-05,
+      "loss": 3.3844,
+      "step": 3051
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 9.96691437802908e-05,
+      "loss": 3.4068,
+      "step": 3052
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 9.966898222940226e-05,
+      "loss": 3.4125,
+      "step": 3053
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 9.966882067851373e-05,
+      "loss": 3.4203,
+      "step": 3054
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 9.966865912762521e-05,
+      "loss": 3.4191,
+      "step": 3055
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 9.966849757673667e-05,
+      "loss": 3.4408,
+      "step": 3056
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 9.966833602584815e-05,
+      "loss": 3.5334,
+      "step": 3057
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 9.966817447495962e-05,
+      "loss": 3.4807,
+      "step": 3058
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 9.966801292407108e-05,
+      "loss": 3.5753,
+      "step": 3059
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 9.966785137318256e-05,
+      "loss": 3.8766,
+      "step": 3060
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 9.966768982229403e-05,
+      "loss": 3.8627,
+      "step": 3061
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 9.96675282714055e-05,
+      "loss": 6.4975,
+      "step": 3062
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 9.966736672051697e-05,
+      "loss": 3.4605,
+      "step": 3063
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 9.966720516962844e-05,
+      "loss": 3.4622,
+      "step": 3064
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 9.96670436187399e-05,
+      "loss": 3.4167,
+      "step": 3065
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 9.966688206785138e-05,
+      "loss": 3.3938,
+      "step": 3066
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 9.966672051696286e-05,
+      "loss": 3.3978,
+      "step": 3067
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 9.966655896607432e-05,
+      "loss": 3.3727,
+      "step": 3068
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 9.966639741518578e-05,
+      "loss": 3.395,
+      "step": 3069
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 9.966623586429727e-05,
+      "loss": 3.4072,
+      "step": 3070
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 9.966607431340873e-05,
+      "loss": 3.3797,
+      "step": 3071
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 9.966591276252019e-05,
+      "loss": 3.3546,
+      "step": 3072
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 9.966575121163167e-05,
+      "loss": 3.3716,
+      "step": 3073
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 9.966558966074314e-05,
+      "loss": 3.3749,
+      "step": 3074
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 9.96654281098546e-05,
+      "loss": 3.3973,
+      "step": 3075
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 9.966526655896608e-05,
+      "loss": 3.4179,
+      "step": 3076
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 9.966510500807755e-05,
+      "loss": 3.3927,
+      "step": 3077
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 9.966494345718901e-05,
+      "loss": 3.4081,
+      "step": 3078
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 9.966478190630049e-05,
+      "loss": 3.3895,
+      "step": 3079
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 9.966462035541196e-05,
+      "loss": 3.4006,
+      "step": 3080
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 9.966445880452343e-05,
+      "loss": 3.4478,
+      "step": 3081
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 9.96642972536349e-05,
+      "loss": 3.4376,
+      "step": 3082
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 9.966413570274638e-05,
+      "loss": 3.4906,
+      "step": 3083
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 9.966397415185784e-05,
+      "loss": 3.5361,
+      "step": 3084
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 9.966381260096931e-05,
+      "loss": 3.6314,
+      "step": 3085
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 9.966365105008077e-05,
+      "loss": 4.1448,
+      "step": 3086
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 9.966348949919225e-05,
+      "loss": 5.4056,
+      "step": 3087
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 9.966332794830372e-05,
+      "loss": 3.4807,
+      "step": 3088
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 9.966316639741519e-05,
+      "loss": 3.4455,
+      "step": 3089
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 9.966300484652666e-05,
+      "loss": 3.4607,
+      "step": 3090
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 9.966284329563814e-05,
+      "loss": 3.4168,
+      "step": 3091
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 9.96626817447496e-05,
+      "loss": 3.4005,
+      "step": 3092
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 9.966252019386107e-05,
+      "loss": 3.3944,
+      "step": 3093
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 9.966235864297255e-05,
+      "loss": 3.3708,
+      "step": 3094
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 9.966219709208401e-05,
+      "loss": 3.3833,
+      "step": 3095
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 9.966203554119548e-05,
+      "loss": 3.3707,
+      "step": 3096
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 9.966187399030696e-05,
+      "loss": 3.3774,
+      "step": 3097
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 9.966171243941842e-05,
+      "loss": 3.4236,
+      "step": 3098
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 9.966155088852988e-05,
+      "loss": 3.3872,
+      "step": 3099
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 9.966138933764137e-05,
+      "loss": 3.4044,
+      "step": 3100
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 9.966122778675283e-05,
+      "loss": 3.3935,
+      "step": 3101
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 9.96610662358643e-05,
+      "loss": 3.4333,
+      "step": 3102
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 9.966090468497577e-05,
+      "loss": 3.4338,
+      "step": 3103
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 9.966074313408724e-05,
+      "loss": 3.4192,
+      "step": 3104
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 9.96605815831987e-05,
+      "loss": 3.43,
+      "step": 3105
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 9.966042003231018e-05,
+      "loss": 3.5429,
+      "step": 3106
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 9.966025848142166e-05,
+      "loss": 3.452,
+      "step": 3107
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 9.966009693053312e-05,
+      "loss": 3.6104,
+      "step": 3108
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 9.965993537964459e-05,
+      "loss": 3.9334,
+      "step": 3109
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 9.965977382875607e-05,
+      "loss": 3.876,
+      "step": 3110
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 9.965961227786753e-05,
+      "loss": 4.7946,
+      "step": 3111
+    },
+    {
+      "epoch": 25.0,
+      "eval_loss": 4.620550632476807,
+      "eval_runtime": 39.7464,
+      "eval_samples_per_second": 21.033,
+      "eval_steps_per_second": 0.679,
+      "eval_wer": 1.0,
+      "step": 3111
+    }
+  ],
+  "max_steps": 620000,
+  "num_train_epochs": 5000,
+  "total_flos": 8.739857551774381e+18,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/model-bin/finetune/base/checkpoint-3111/training_args.bin b/model-bin/finetune/base/checkpoint-3111/training_args.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e7a8718b9eaed8ec9d5ea95aafd35e0369b532df
--- /dev/null
+++ b/model-bin/finetune/base/checkpoint-3111/training_args.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de83b8f37cb6c77cab3d4f164a060f8ddc9ad6805223463da65d9ee2ef8ede6e
+size 2671
diff --git a/model-bin/finetune/base/log/1629376492.385729/events.out.tfevents.1629376492.feefc9cbc27a.91257.1 b/model-bin/finetune/base/log/1629376492.385729/events.out.tfevents.1629376492.feefc9cbc27a.91257.1
new file mode 100644
index 0000000000000000000000000000000000000000..133526a7495a2d09a4790364045d6c01d5428a9f
--- /dev/null
+++ b/model-bin/finetune/base/log/1629376492.385729/events.out.tfevents.1629376492.feefc9cbc27a.91257.1
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6383e39b12e15cc8d252a4ae2592f94a688aaacd6e17263ba1d69f87db4381d2
+size 4194
diff --git a/model-bin/finetune/base/log/1629377138.4521065/events.out.tfevents.1629377138.feefc9cbc27a.91257.3 b/model-bin/finetune/base/log/1629377138.4521065/events.out.tfevents.1629377138.feefc9cbc27a.91257.3
new file mode 100644
index 0000000000000000000000000000000000000000..01bd12f4b59c3578d03d9bce6f9e1e7dd834137f
--- /dev/null
+++ b/model-bin/finetune/base/log/1629377138.4521065/events.out.tfevents.1629377138.feefc9cbc27a.91257.3
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86739f3fe8ab8d6112d194f238ffbb19b570f0812a4c84ff7486901d51974da4
+size 4194
diff --git a/model-bin/finetune/base/log/1629377787.9352672/events.out.tfevents.1629377787.feefc9cbc27a.91257.5 b/model-bin/finetune/base/log/1629377787.9352672/events.out.tfevents.1629377787.feefc9cbc27a.91257.5
new file mode 100644
index 0000000000000000000000000000000000000000..414c1520803b5fb2fbe38410947c8079af43bd99
--- /dev/null
+++ b/model-bin/finetune/base/log/1629377787.9352672/events.out.tfevents.1629377787.feefc9cbc27a.91257.5
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f76de54d1c1f8c2ab3fe1e1b20b3d636b82b3a7d312d92b5d8d1f6299052fe5
+size 4194
diff --git a/model-bin/finetune/base/log/1629378426.0399842/events.out.tfevents.1629378426.feefc9cbc27a.91257.7 b/model-bin/finetune/base/log/1629378426.0399842/events.out.tfevents.1629378426.feefc9cbc27a.91257.7
new file mode 100644
index 0000000000000000000000000000000000000000..ea5dbf35b322777b1f142c872eca34158dd08423
--- /dev/null
+++ b/model-bin/finetune/base/log/1629378426.0399842/events.out.tfevents.1629378426.feefc9cbc27a.91257.7
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c108de9ef1024cf7c797ddd3bbba08d6376fb5a5f5e202d5193b211ef79be415
+size 4194
diff --git a/model-bin/finetune/base/log/1629379071.5436237/events.out.tfevents.1629379071.feefc9cbc27a.91257.9 b/model-bin/finetune/base/log/1629379071.5436237/events.out.tfevents.1629379071.feefc9cbc27a.91257.9
new file mode 100644
index 0000000000000000000000000000000000000000..fd2ea4d37508382a44eb7cea0f8a4d5d7ae63515
--- /dev/null
+++ b/model-bin/finetune/base/log/1629379071.5436237/events.out.tfevents.1629379071.feefc9cbc27a.91257.9
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:51173bd32802291a8027155ff1e74047f3b6657946ba560f7d4693c8a4d8f0eb
+size 4194
diff --git a/model-bin/finetune/base/log/1629379702.0305626/events.out.tfevents.1629379702.feefc9cbc27a.91257.11 b/model-bin/finetune/base/log/1629379702.0305626/events.out.tfevents.1629379702.feefc9cbc27a.91257.11
new file mode 100644
index 0000000000000000000000000000000000000000..ba59d335b3d230824a1fbc9914a4a76630cd7339
--- /dev/null
+++ b/model-bin/finetune/base/log/1629379702.0305626/events.out.tfevents.1629379702.feefc9cbc27a.91257.11
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da0662a6af9d5294ff5390449a3a6a51b8447c208c984eb6ca3dcbd4244d57a1
+size 4194
diff --git a/model-bin/finetune/base/log/1629380423.6802466/events.out.tfevents.1629380423.feefc9cbc27a.91257.13 b/model-bin/finetune/base/log/1629380423.6802466/events.out.tfevents.1629380423.feefc9cbc27a.91257.13
new file mode 100644
index 0000000000000000000000000000000000000000..25cd6fe3cc768ae69c9eead65c199a31f67262b4
--- /dev/null
+++ b/model-bin/finetune/base/log/1629380423.6802466/events.out.tfevents.1629380423.feefc9cbc27a.91257.13
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:327682483ef721132df1ba0488666ad205761a9514a6f39ff6303fef7429ec84
+size 4194
diff --git a/model-bin/finetune/base/log/1629381079.3164322/events.out.tfevents.1629381079.feefc9cbc27a.91257.15 b/model-bin/finetune/base/log/1629381079.3164322/events.out.tfevents.1629381079.feefc9cbc27a.91257.15
new file mode 100644
index 0000000000000000000000000000000000000000..e27f9107be6acaddbec119a7f1fe33804bb2bf5d
--- /dev/null
+++ b/model-bin/finetune/base/log/1629381079.3164322/events.out.tfevents.1629381079.feefc9cbc27a.91257.15
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90356458a0a4c034e5c53cb1d105739aecf38c96f44fd06892f3480b0261a456
+size 4194
diff --git a/model-bin/finetune/base/log/1629381718.8383589/events.out.tfevents.1629381718.feefc9cbc27a.91257.17 b/model-bin/finetune/base/log/1629381718.8383589/events.out.tfevents.1629381718.feefc9cbc27a.91257.17
new file mode 100644
index 0000000000000000000000000000000000000000..0c4f2c686cfc6300523c87e78c1dd37d9763e676
--- /dev/null
+++ b/model-bin/finetune/base/log/1629381718.8383589/events.out.tfevents.1629381718.feefc9cbc27a.91257.17
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11c641288fd5193daf61f45d0e7318b257a7da374fc81743168169dc1c7bdcf4
+size 4194
diff --git a/model-bin/finetune/base/log/1629382361.3283687/events.out.tfevents.1629382361.feefc9cbc27a.91257.19 b/model-bin/finetune/base/log/1629382361.3283687/events.out.tfevents.1629382361.feefc9cbc27a.91257.19
new file mode 100644
index 0000000000000000000000000000000000000000..dccb215967b7659d9eeb4fd189874417f1ebae22
--- /dev/null
+++ b/model-bin/finetune/base/log/1629382361.3283687/events.out.tfevents.1629382361.feefc9cbc27a.91257.19
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:add8c6eb0817ea2a046f328a0c896bc22f5f4423bc0361d603d53e61de8ac116
+size 4194
diff --git a/model-bin/finetune/base/log/1629382988.8115258/events.out.tfevents.1629382988.feefc9cbc27a.91257.21 b/model-bin/finetune/base/log/1629382988.8115258/events.out.tfevents.1629382988.feefc9cbc27a.91257.21
new file mode 100644
index 0000000000000000000000000000000000000000..00284dad603a9eac0cf79fe33ba3d8879b0ccaec
--- /dev/null
+++ b/model-bin/finetune/base/log/1629382988.8115258/events.out.tfevents.1629382988.feefc9cbc27a.91257.21
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d027315a41463d93518239261bc5d0feb69eed666c015db9c7126acc9b3c3205
+size 4194
diff --git a/model-bin/finetune/base/log/1629383614.8360982/events.out.tfevents.1629383614.feefc9cbc27a.91257.23 b/model-bin/finetune/base/log/1629383614.8360982/events.out.tfevents.1629383614.feefc9cbc27a.91257.23
new file mode 100644
index 0000000000000000000000000000000000000000..e9eecf7704fee6090e3e40cd9f2e46db6ce4d080
--- /dev/null
+++ b/model-bin/finetune/base/log/1629383614.8360982/events.out.tfevents.1629383614.feefc9cbc27a.91257.23
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d88759fa4ce640cc382805f79ae8a7a0fb9adebaaf5da4e8ee3200831c91971
+size 4194
diff --git a/model-bin/finetune/base/log/1629384242.0507712/events.out.tfevents.1629384242.feefc9cbc27a.91257.25 b/model-bin/finetune/base/log/1629384242.0507712/events.out.tfevents.1629384242.feefc9cbc27a.91257.25
new file mode 100644
index 0000000000000000000000000000000000000000..dfe8407a6ec5ce896f02aac86f09bdc59133ee25
--- /dev/null
+++ b/model-bin/finetune/base/log/1629384242.0507712/events.out.tfevents.1629384242.feefc9cbc27a.91257.25
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:531d040c7663b1772e9c19493a3a1c1f99216e4aec39926269f3b3a92699f834
+size 4194
diff --git a/model-bin/finetune/base/log/1629384879.444011/events.out.tfevents.1629384879.feefc9cbc27a.91257.27 b/model-bin/finetune/base/log/1629384879.444011/events.out.tfevents.1629384879.feefc9cbc27a.91257.27
new file mode 100644
index 0000000000000000000000000000000000000000..bdd856e9a71893becb9bddf6a5d5afdd91c6924e
--- /dev/null
+++ b/model-bin/finetune/base/log/1629384879.444011/events.out.tfevents.1629384879.feefc9cbc27a.91257.27
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a5c192f97d4c2587d6f1d7545e37bd46bf683dd5c6dc7bcf94ac8df18304235
+size 4194
diff --git a/model-bin/finetune/base/log/1629385514.3230636/events.out.tfevents.1629385514.feefc9cbc27a.91257.29 b/model-bin/finetune/base/log/1629385514.3230636/events.out.tfevents.1629385514.feefc9cbc27a.91257.29
new file mode 100644
index 0000000000000000000000000000000000000000..40e3b4c06a190d236905043b38f7445ea21df1c5
--- /dev/null
+++ b/model-bin/finetune/base/log/1629385514.3230636/events.out.tfevents.1629385514.feefc9cbc27a.91257.29
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1da241e86f767b547d3c12732a6a9fbb4b8d5020d58fc8a161ed8b7b06b19558
+size 4194
diff --git a/model-bin/finetune/base/log/1629386161.4312515/events.out.tfevents.1629386161.feefc9cbc27a.91257.31 b/model-bin/finetune/base/log/1629386161.4312515/events.out.tfevents.1629386161.feefc9cbc27a.91257.31
new file mode 100644
index 0000000000000000000000000000000000000000..b8d1f4eb10b7e3e116b628da4297976de9a3963d
--- /dev/null
+++ b/model-bin/finetune/base/log/1629386161.4312515/events.out.tfevents.1629386161.feefc9cbc27a.91257.31
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8561f35a05c17d18ecedda4b674d789769ba4df402629b47fa3c4114d4b9a79d
+size 4194
diff --git a/model-bin/finetune/base/log/1629386793.9539597/events.out.tfevents.1629386793.feefc9cbc27a.91257.33 b/model-bin/finetune/base/log/1629386793.9539597/events.out.tfevents.1629386793.feefc9cbc27a.91257.33
new file mode 100644
index 0000000000000000000000000000000000000000..d51c5c60a022a6dd45d7b916926af37d4bcb8efe
--- /dev/null
+++ b/model-bin/finetune/base/log/1629386793.9539597/events.out.tfevents.1629386793.feefc9cbc27a.91257.33
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8dce362da443f6b2a4f05bc148657cb0af31458aa5b6afbaf925db45d6686d3
+size 4194
diff --git a/model-bin/finetune/base/log/1629387516.694571/events.out.tfevents.1629387516.feefc9cbc27a.91257.35 b/model-bin/finetune/base/log/1629387516.694571/events.out.tfevents.1629387516.feefc9cbc27a.91257.35
new file mode 100644
index 0000000000000000000000000000000000000000..bad0bc49c380d0d654b5f97c54e35d1f6a79eb19
--- /dev/null
+++ b/model-bin/finetune/base/log/1629387516.694571/events.out.tfevents.1629387516.feefc9cbc27a.91257.35
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c53b2829af1ce63e9c2c05c3f43befbbdbeff65ccd874fa4f56fac2eee54a6d2
+size 4194
diff --git a/model-bin/finetune/base/log/1629388160.4914525/events.out.tfevents.1629388160.feefc9cbc27a.91257.37 b/model-bin/finetune/base/log/1629388160.4914525/events.out.tfevents.1629388160.feefc9cbc27a.91257.37
new file mode 100644
index 0000000000000000000000000000000000000000..255300b8ac9010f8b340f94abe1a6b1e7650c92b
--- /dev/null
+++ b/model-bin/finetune/base/log/1629388160.4914525/events.out.tfevents.1629388160.feefc9cbc27a.91257.37
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:453f8c9117c9fc3c3dbefa374e9a62bed0577b7a2edac0173187c55dc55c9409
+size 4194
diff --git a/model-bin/finetune/base/log/1629388781.7681358/events.out.tfevents.1629388781.feefc9cbc27a.91257.39 b/model-bin/finetune/base/log/1629388781.7681358/events.out.tfevents.1629388781.feefc9cbc27a.91257.39
new file mode 100644
index 0000000000000000000000000000000000000000..e73db9e020747ffad80d973f5d92824daa59140a
--- /dev/null
+++ b/model-bin/finetune/base/log/1629388781.7681358/events.out.tfevents.1629388781.feefc9cbc27a.91257.39
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f86ac640f6b82b7b51b34fa2653c509d11282c5c59dcb8aa011cb8ada7cfab51
+size 4194
diff --git a/model-bin/finetune/base/log/1629389410.0638642/events.out.tfevents.1629389410.feefc9cbc27a.91257.41 b/model-bin/finetune/base/log/1629389410.0638642/events.out.tfevents.1629389410.feefc9cbc27a.91257.41
new file mode 100644
index 0000000000000000000000000000000000000000..aa87e0d6166c2157003c25e5756c741bf4f4be69
--- /dev/null
+++ b/model-bin/finetune/base/log/1629389410.0638642/events.out.tfevents.1629389410.feefc9cbc27a.91257.41
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ded5460b4287550772197110df30d3972c20f8ea42c6aaa883593cbb5beb7ab
+size 4194
diff --git a/model-bin/finetune/base/log/1629390060.0094845/events.out.tfevents.1629390060.feefc9cbc27a.91257.43 b/model-bin/finetune/base/log/1629390060.0094845/events.out.tfevents.1629390060.feefc9cbc27a.91257.43
new file mode 100644
index 0000000000000000000000000000000000000000..a6bed0818a7ac57162799ac22403d3ac46f438de
--- /dev/null
+++ b/model-bin/finetune/base/log/1629390060.0094845/events.out.tfevents.1629390060.feefc9cbc27a.91257.43
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c293638752616b9fd6d6e72b508200455325e3d3f488d47b4476eec2f8e278f
+size 4194
diff --git a/model-bin/finetune/base/log/1629390689.30961/events.out.tfevents.1629390689.feefc9cbc27a.91257.45 b/model-bin/finetune/base/log/1629390689.30961/events.out.tfevents.1629390689.feefc9cbc27a.91257.45
new file mode 100644
index 0000000000000000000000000000000000000000..7d4d5853a8c01981e2517be450528f9c28e11d23
--- /dev/null
+++ b/model-bin/finetune/base/log/1629390689.30961/events.out.tfevents.1629390689.feefc9cbc27a.91257.45
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:855863bc5e04752d9073c00856de702ad2e65943809195fb9f698c9f8b39dc15
+size 4194
diff --git a/model-bin/finetune/base/log/1629391334.0100863/events.out.tfevents.1629391334.feefc9cbc27a.91257.47 b/model-bin/finetune/base/log/1629391334.0100863/events.out.tfevents.1629391334.feefc9cbc27a.91257.47
new file mode 100644
index 0000000000000000000000000000000000000000..26575ddedc577d109768c79618fc575281bdd682
--- /dev/null
+++ b/model-bin/finetune/base/log/1629391334.0100863/events.out.tfevents.1629391334.feefc9cbc27a.91257.47
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df8487d7a005f32bafdddeb8f62ce2879f4da279d45572d85b308924f3a037ea
+size 4194
diff --git a/model-bin/finetune/base/log/1629391969.854197/events.out.tfevents.1629391969.feefc9cbc27a.91257.49 b/model-bin/finetune/base/log/1629391969.854197/events.out.tfevents.1629391969.feefc9cbc27a.91257.49
new file mode 100644
index 0000000000000000000000000000000000000000..2d6937491cb5b5a5224b6dff87f6dadf92622c0e
--- /dev/null
+++ b/model-bin/finetune/base/log/1629391969.854197/events.out.tfevents.1629391969.feefc9cbc27a.91257.49
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96002d7c4f05b6ce45aff10edc7b78ee5940d7db5e5e53f7d4b44303e51b1bde
+size 4194
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629376492.feefc9cbc27a.91257.0 b/model-bin/finetune/base/log/events.out.tfevents.1629376492.feefc9cbc27a.91257.0
new file mode 100644
index 0000000000000000000000000000000000000000..d4ca9b11a5fd0b59baf977dcfd980f66d56abd8d
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629376492.feefc9cbc27a.91257.0
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:163bb1e8564bc726aaf2b355c436a663fdfe7a6d9b2a7b1c1e1a0cb37c629b3e
+size 23703
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629377138.feefc9cbc27a.91257.2 b/model-bin/finetune/base/log/events.out.tfevents.1629377138.feefc9cbc27a.91257.2
new file mode 100644
index 0000000000000000000000000000000000000000..e275c08f345fa6f1a54c7e1a45cc99efded202ec
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629377138.feefc9cbc27a.91257.2
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f0fe3792b2cabad6662d0f1093c4542cf733681ff353fd245601412064c787e
+size 24070
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629377787.feefc9cbc27a.91257.4 b/model-bin/finetune/base/log/events.out.tfevents.1629377787.feefc9cbc27a.91257.4
new file mode 100644
index 0000000000000000000000000000000000000000..1fd02a92865f82feda906c3bd9aa1fef70c297f5
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629377787.feefc9cbc27a.91257.4
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0498cfe3bc38164fba47a0adfab9ff7aa02f1a385a9d172f95a4208420568a60
+size 24236
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629378426.feefc9cbc27a.91257.6 b/model-bin/finetune/base/log/events.out.tfevents.1629378426.feefc9cbc27a.91257.6
new file mode 100644
index 0000000000000000000000000000000000000000..c316c8caf13bfa6ec6c87d83e2c6718226d77de7
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629378426.feefc9cbc27a.91257.6
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a5b078c66105d613c50b5f202b396b0bb48439c02216034ebda0ca5b4b0d6a5
+size 24236
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629379071.feefc9cbc27a.91257.8 b/model-bin/finetune/base/log/events.out.tfevents.1629379071.feefc9cbc27a.91257.8
new file mode 100644
index 0000000000000000000000000000000000000000..8bbd625cc5bee2f8dfbd1e308d83a0ec482112fc
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629379071.feefc9cbc27a.91257.8
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d7211e4b6d406387e53f745d0750f466756d8c9f42801e155e9ec0b0efe015f
+size 24079
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629379702.feefc9cbc27a.91257.10 b/model-bin/finetune/base/log/events.out.tfevents.1629379702.feefc9cbc27a.91257.10
new file mode 100644
index 0000000000000000000000000000000000000000..93c590ffb8d53649a40f57dca18e344938695cb1
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629379702.feefc9cbc27a.91257.10
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d57e2fdd271edaf03df88d00402b54a69084ee36677729c98a5ce69098372f4
+size 24236
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629380423.feefc9cbc27a.91257.12 b/model-bin/finetune/base/log/events.out.tfevents.1629380423.feefc9cbc27a.91257.12
new file mode 100644
index 0000000000000000000000000000000000000000..c8e10de14d586108aedc7e1bfc87eba33b12d072
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629380423.feefc9cbc27a.91257.12
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82fdcff7ca49bd77913d5aac71b84e9ef4474fcea272caa7d924ad842ecf1b6d
+size 24079
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629381079.feefc9cbc27a.91257.14 b/model-bin/finetune/base/log/events.out.tfevents.1629381079.feefc9cbc27a.91257.14
new file mode 100644
index 0000000000000000000000000000000000000000..aab3a78b56fb7b260ad158eee5e99b288899c41c
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629381079.feefc9cbc27a.91257.14
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ab6ef83a526a9db877aeefac567ebe983523d4816fe25c330181fc8c5c4c12d
+size 24236
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629381718.feefc9cbc27a.91257.16 b/model-bin/finetune/base/log/events.out.tfevents.1629381718.feefc9cbc27a.91257.16
new file mode 100644
index 0000000000000000000000000000000000000000..588f0db9d9676ddfb7b26dc0f7cfa16929098c77
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629381718.feefc9cbc27a.91257.16
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba658ffef6036965740d3afe1b26e76aa625a3e4321636404c75530c9c60fc7d
+size 24236
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629382361.feefc9cbc27a.91257.18 b/model-bin/finetune/base/log/events.out.tfevents.1629382361.feefc9cbc27a.91257.18
new file mode 100644
index 0000000000000000000000000000000000000000..f198ac9876462308c114abaada6d3b80bc6366b5
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629382361.feefc9cbc27a.91257.18
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8dd166f861407a26cd94be68396400ddb33551a93aca83dfb40c2a96d06b6b56
+size 24236
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629382988.feefc9cbc27a.91257.20 b/model-bin/finetune/base/log/events.out.tfevents.1629382988.feefc9cbc27a.91257.20
new file mode 100644
index 0000000000000000000000000000000000000000..86751443cfea261309b867ff52c472b9435bb353
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629382988.feefc9cbc27a.91257.20
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af1db21b7917a4efad55c7c25dc56c58bc31cabab30d366fffb6f0a78f096f3f
+size 24079
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629383614.feefc9cbc27a.91257.22 b/model-bin/finetune/base/log/events.out.tfevents.1629383614.feefc9cbc27a.91257.22
new file mode 100644
index 0000000000000000000000000000000000000000..5116bd168d53c9de1b0b062682e04bf76448ffd4
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629383614.feefc9cbc27a.91257.22
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:112ebb4713f0703c67fabd19cfe41516c1af66244df334e4b352f791f3566661
+size 24079
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629384242.feefc9cbc27a.91257.24 b/model-bin/finetune/base/log/events.out.tfevents.1629384242.feefc9cbc27a.91257.24
new file mode 100644
index 0000000000000000000000000000000000000000..ecd1faef883264c826b3ad177eff719b1e60b75e
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629384242.feefc9cbc27a.91257.24
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88bb19db4db3ca0e20b588c1714efadd9ca3fb6f2f14405101eb2be6db92b60f
+size 24079
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629384879.feefc9cbc27a.91257.26 b/model-bin/finetune/base/log/events.out.tfevents.1629384879.feefc9cbc27a.91257.26
new file mode 100644
index 0000000000000000000000000000000000000000..2b80f87b59da10345723d17cb52285d4fd142270
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629384879.feefc9cbc27a.91257.26
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:686cca7bf20df5008c990d38a54d29156fa09936f70fc02059565fbb5b00c2b7
+size 24236
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629385514.feefc9cbc27a.91257.28 b/model-bin/finetune/base/log/events.out.tfevents.1629385514.feefc9cbc27a.91257.28
new file mode 100644
index 0000000000000000000000000000000000000000..d897fec37d2103e700a4074309e9ef2e1d0a5eea
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629385514.feefc9cbc27a.91257.28
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:17de1c5252ee375a535299d5da76b54c5ad386414e2d5dd242ba55c096f8293b
+size 24079
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629386161.feefc9cbc27a.91257.30 b/model-bin/finetune/base/log/events.out.tfevents.1629386161.feefc9cbc27a.91257.30
new file mode 100644
index 0000000000000000000000000000000000000000..e4ab8baf52639749983099468a564c50adcc2e58
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629386161.feefc9cbc27a.91257.30
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73f1f4887601a07d0b5126f1c531eb1b982d107d6261d9ab40c4ea6505643f9b
+size 24236
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629386793.feefc9cbc27a.91257.32 b/model-bin/finetune/base/log/events.out.tfevents.1629386793.feefc9cbc27a.91257.32
new file mode 100644
index 0000000000000000000000000000000000000000..159404662c526be11548e666be066950d2b4a376
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629386793.feefc9cbc27a.91257.32
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f47652815300602b2de9742501dd2a9d7b0b7fd575ddf6ac1bbe3636465d65aa
+size 24079
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629387516.feefc9cbc27a.91257.34 b/model-bin/finetune/base/log/events.out.tfevents.1629387516.feefc9cbc27a.91257.34
new file mode 100644
index 0000000000000000000000000000000000000000..3834b2120b6b957caa023b5796e11d624ced3dc0
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629387516.feefc9cbc27a.91257.34
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2bb21b25d4baafb17b24d666ad51e6bda7460b4ce95be9af2d823ac1677c769c
+size 24079
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629388160.feefc9cbc27a.91257.36 b/model-bin/finetune/base/log/events.out.tfevents.1629388160.feefc9cbc27a.91257.36
new file mode 100644
index 0000000000000000000000000000000000000000..ff6b19e6c308cba003390b162ad7423f850f944b
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629388160.feefc9cbc27a.91257.36
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19ff8372bc8022832f5922e2f65d090e9cc6572f8fb1c2e1f776a4f4c1c1b5ef
+size 24079
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629388781.feefc9cbc27a.91257.38 b/model-bin/finetune/base/log/events.out.tfevents.1629388781.feefc9cbc27a.91257.38
new file mode 100644
index 0000000000000000000000000000000000000000..83f0b2c329887d414c066d4c697c23b5ea5d7bc8
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629388781.feefc9cbc27a.91257.38
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:749a74faed1bd806770c56683b0242fa14829fb62ef161dcec6401c46748acdd
+size 24236
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629389410.feefc9cbc27a.91257.40 b/model-bin/finetune/base/log/events.out.tfevents.1629389410.feefc9cbc27a.91257.40
new file mode 100644
index 0000000000000000000000000000000000000000..cb545192f8b6f86587840b52faa215e2b6b96de4
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629389410.feefc9cbc27a.91257.40
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f4a30b9b822d19b6aac5a6f5cdbd71b4c858ce401347f60e81f7416ab118281
+size 24079
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629390060.feefc9cbc27a.91257.42 b/model-bin/finetune/base/log/events.out.tfevents.1629390060.feefc9cbc27a.91257.42
new file mode 100644
index 0000000000000000000000000000000000000000..02a4c15832dcb6b3afb23fcd355a7d23f69eb6a9
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629390060.feefc9cbc27a.91257.42
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80c0ab159a14ded71e2ae0ba6d39940b7b9f6dbd117e3c841717b8123f919829
+size 24236
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629390689.feefc9cbc27a.91257.44 b/model-bin/finetune/base/log/events.out.tfevents.1629390689.feefc9cbc27a.91257.44
new file mode 100644
index 0000000000000000000000000000000000000000..3345593ae9fb95bb3c2044da17247250cb053e37
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629390689.feefc9cbc27a.91257.44
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6fc888b5fab1772fbf7757e1bf780e00e0454a985352ad9df871d7007c9fbe68
+size 24236
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629391334.feefc9cbc27a.91257.46 b/model-bin/finetune/base/log/events.out.tfevents.1629391334.feefc9cbc27a.91257.46
new file mode 100644
index 0000000000000000000000000000000000000000..42ed0d5dd220fa340a08eaa2293bcd226c28b5bd
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629391334.feefc9cbc27a.91257.46
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42d244f8f52cab52a157f9025360f095ce898229655754cbe17af573e2920f61
+size 24079
diff --git a/model-bin/finetune/base/log/events.out.tfevents.1629391969.feefc9cbc27a.91257.48 b/model-bin/finetune/base/log/events.out.tfevents.1629391969.feefc9cbc27a.91257.48
new file mode 100644
index 0000000000000000000000000000000000000000..b85a6095244c884129da47775516eb58d4eb2630
--- /dev/null
+++ b/model-bin/finetune/base/log/events.out.tfevents.1629391969.feefc9cbc27a.91257.48
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:97c5d84d09d8a78ac4e39796ac70f077dc3ee8a2627c1e97a216a26846ddff4b
+size 24079