add new model

README.md

@@ -10,13 +10,13 @@ datasets:
 metrics:
 - wer
 model-index:
-- name: Whisper Small Th Additional Data - biodatlab
+- name: Whisper Small Thai Combined Concat
   results:
   - task:
       name: Automatic Speech Recognition
       type: automatic-speech-recognition
     dataset:
-      name: Common Voice 11.0
+      name: mozilla-foundation/common_voice_11_0 th
       type: mozilla-foundation/common_voice_11_0
       config: th
       split: test
@@ -24,18 +24,18 @@ model-index:
     metrics:
     - name: Wer
       type: wer
-      value: 62.71786878276888
+      value: 27.279438445464898
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 should probably proofread and complete it, then remove this comment. -->
 
-# Whisper Small Th Additional Data - biodatlab
+# Whisper Small Thai Combined Concat
 
-This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the Common Voice 11.0 dataset.
+This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the mozilla-foundation/common_voice_11_0 th dataset and additional scraped data.
 It achieves the following results on the evaluation set:
-- Loss: 0.1986
-- Wer: 62.7179
+- Loss: 0.5034
+- Wer: 27.2794 (without tokenizer)
 
 ## Model description
 
@@ -55,31 +55,25 @@ More information needed
 
 The following hyperparameters were used during training:
 - learning_rate: 1e-05
-- train_batch_size: 32
-- eval_batch_size: 8
+- train_batch_size: 64
+- eval_batch_size: 32
 - seed: 42
-- gradient_accumulation_steps: 2
-- total_train_batch_size: 64
+- distributed_type: multi-GPU
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
 - lr_scheduler_warmup_steps: 500
 - training_steps: 5000
-- mixed_precision_training: Native AMP
 
 ### Training results
 
 | Training Loss | Epoch | Step | Validation Loss | Wer     |
 |:-------------:|:-----:|:----:|:---------------:|:-------:|
-| 0.3104        | 1.46  | 1000 | 0.1969          | 67.1036 |
-| 0.2026        | 2.93  | 2000 | 0.1686          | 63.4264 |
-| 0.0975        | 4.39  | 3000 | 0.1780          | 61.8818 |
-| 0.0591        | 5.86  | 4000 | 0.1877          | 62.6045 |
-| 0.0328        | 7.32  | 5000 | 0.1986          | 62.7179 |
+| 0.0002        | 83.33 | 5000 | 0.5034          | 27.2794 |
 
 
 ### Framework versions
 
-- Transformers 4.26.0.dev0
-- Pytorch 1.13.0
-- Datasets 2.7.1
+- Transformers 4.27.0.dev0
+- Pytorch 1.13.1
+- Datasets 2.9.1.dev0
 - Tokenizers 0.13.2

added_tokens.json

@@ -17,7 +17,6 @@
   "<|da|>": 50285,
   "<|de|>": 50261,
   "<|el|>": 50281,
-  "<|endoftext|>": 50257,
   "<|en|>": 50259,
   "<|es|>": 50262,
   "<|et|>": 50307,
@@ -30,6 +29,7 @@
   "<|gu|>": 50333,
   "<|haw|>": 50352,
   "<|ha|>": 50354,
+  "<|he|>": 50279,
   "<|hi|>": 50276,
   "<|hr|>": 50291,
   "<|ht|>": 50339,
@@ -38,7 +38,6 @@
   "<|id|>": 50275,
   "<|is|>": 50311,
   "<|it|>": 50274,
-  "<|iw|>": 50279,
   "<|ja|>": 50266,
   "<|jw|>": 50356,
   "<|ka|>": 50329,

all_results.json

@@ -0,0 +1,12 @@
+{
+    "epoch": 83.33,
+    "eval_loss": 0.50341796875,
+    "eval_runtime": 154.187,
+    "eval_samples_per_second": 13.166,
+    "eval_steps_per_second": 0.415,
+    "eval_wer": 27.279438445464898,
+    "train_loss": 0.031859876942634584,
+    "train_runtime": 20088.5298,
+    "train_samples_per_second": 15.929,
+    "train_steps_per_second": 0.249
+}

config.json

@@ -34,9 +34,8 @@
   "num_mel_bins": 80,
   "pad_token_id": 50257,
   "scale_embedding": false,
-  "suppress_tokens": [],
-  "torch_dtype": "float32",
-  "transformers_version": "4.26.0.dev0",
+  "torch_dtype": "float16",
+  "transformers_version": "4.27.0.dev0",
   "use_cache": false,
   "vocab_size": 51865
 }

dsconfig.json

@@ -0,0 +1,50 @@
+{
+    "fp16": {
+        "enabled": "auto",
+        "loss_scale": 0,
+        "loss_scale_window": 1000,
+        "initial_scale_power": 16,
+        "hysteresis": 2,
+        "min_loss_scale": 1
+    },
+
+    "optimizer": {
+        "type": "AdamW",
+        "params": {
+            "lr": "auto",
+            "betas": "auto",
+            "eps": "auto",
+            "weight_decay": "auto"
+        }
+    },
+
+   "scheduler": {
+         "type": "WarmupDecayLR",
+         "params": {
+             "last_batch_iteration": -1,
+             "total_num_steps": "auto",
+             "warmup_min_lr": "auto",
+             "warmup_max_lr": "auto",
+             "warmup_num_steps": "auto"
+         }
+     },
+
+    "zero_optimization": {
+        "stage": 2,
+        "offload_optimizer": {
+            "device": "cpu",
+            "pin_memory": true
+        },
+        "allgather_partitions": true,
+        "allgather_bucket_size": 2e8,
+        "overlap_comm": true,
+        "reduce_scatter": true,
+        "reduce_bucket_size": 2e8,
+        "contiguous_gradients": true
+    },
+
+    "gradient_accumulation_steps": "auto",
+    "gradient_clipping": "auto",
+    "train_batch_size": "auto",
+    "train_micro_batch_size_per_gpu": "auto"
+}

eval_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 83.33,
+    "eval_loss": 0.50341796875,
+    "eval_runtime": 154.187,
+    "eval_samples_per_second": 13.166,
+    "eval_steps_per_second": 0.415,
+    "eval_wer": 27.279438445464898
+}

generation_config.json

@@ -0,0 +1,223 @@
+{
+  "begin_suppress_tokens": [
+    220,
+    50257
+  ],
+  "bos_token_id": 50257,
+  "decoder_start_token_id": 50258,
+  "eos_token_id": 50257,
+  "forced_decoder_ids": [
+    [
+      1,
+      null
+    ],
+    [
+      2,
+      50359
+    ],
+    [
+      3,
+      50363
+    ]
+  ],
+  "is_multilingual": true,
+  "lang_to_id": {
+    "<|af|>": 50327,
+    "<|am|>": 50334,
+    "<|ar|>": 50272,
+    "<|as|>": 50350,
+    "<|az|>": 50304,
+    "<|ba|>": 50355,
+    "<|be|>": 50330,
+    "<|bg|>": 50292,
+    "<|bn|>": 50302,
+    "<|bo|>": 50347,
+    "<|br|>": 50309,
+    "<|bs|>": 50315,
+    "<|ca|>": 50270,
+    "<|cs|>": 50283,
+    "<|cy|>": 50297,
+    "<|da|>": 50285,
+    "<|de|>": 50261,
+    "<|el|>": 50281,
+    "<|en|>": 50259,
+    "<|es|>": 50262,
+    "<|et|>": 50307,
+    "<|eu|>": 50310,
+    "<|fa|>": 50300,
+    "<|fi|>": 50277,
+    "<|fo|>": 50338,
+    "<|fr|>": 50265,
+    "<|gl|>": 50319,
+    "<|gu|>": 50333,
+    "<|haw|>": 50352,
+    "<|ha|>": 50354,
+    "<|he|>": 50279,
+    "<|hi|>": 50276,
+    "<|hr|>": 50291,
+    "<|ht|>": 50339,
+    "<|hu|>": 50286,
+    "<|hy|>": 50312,
+    "<|id|>": 50275,
+    "<|is|>": 50311,
+    "<|it|>": 50274,
+    "<|ja|>": 50266,
+    "<|jw|>": 50356,
+    "<|ka|>": 50329,
+    "<|kk|>": 50316,
+    "<|km|>": 50323,
+    "<|kn|>": 50306,
+    "<|ko|>": 50264,
+    "<|la|>": 50294,
+    "<|lb|>": 50345,
+    "<|ln|>": 50353,
+    "<|lo|>": 50336,
+    "<|lt|>": 50293,
+    "<|lv|>": 50301,
+    "<|mg|>": 50349,
+    "<|mi|>": 50295,
+    "<|mk|>": 50308,
+    "<|ml|>": 50296,
+    "<|mn|>": 50314,
+    "<|mr|>": 50320,
+    "<|ms|>": 50282,
+    "<|mt|>": 50343,
+    "<|my|>": 50346,
+    "<|ne|>": 50313,
+    "<|nl|>": 50271,
+    "<|nn|>": 50342,
+    "<|no|>": 50288,
+    "<|oc|>": 50328,
+    "<|pa|>": 50321,
+    "<|pl|>": 50269,
+    "<|ps|>": 50340,
+    "<|pt|>": 50267,
+    "<|ro|>": 50284,
+    "<|ru|>": 50263,
+    "<|sa|>": 50344,
+    "<|sd|>": 50332,
+    "<|si|>": 50322,
+    "<|sk|>": 50298,
+    "<|sl|>": 50305,
+    "<|sn|>": 50324,
+    "<|so|>": 50326,
+    "<|sq|>": 50317,
+    "<|sr|>": 50303,
+    "<|su|>": 50357,
+    "<|sv|>": 50273,
+    "<|sw|>": 50318,
+    "<|ta|>": 50287,
+    "<|te|>": 50299,
+    "<|tg|>": 50331,
+    "<|th|>": 50289,
+    "<|tk|>": 50341,
+    "<|tl|>": 50348,
+    "<|tr|>": 50268,
+    "<|tt|>": 50351,
+    "<|uk|>": 50280,
+    "<|ur|>": 50290,
+    "<|uz|>": 50337,
+    "<|vi|>": 50278,
+    "<|yi|>": 50335,
+    "<|yo|>": 50325,
+    "<|zh|>": 50260
+  },
+  "max_initial_timestamp_index": 1,
+  "max_length": 448,
+  "no_timestamps_token_id": 50363,
+  "pad_token_id": 50256,
+  "return_timestamps": false,
+  "suppress_tokens": [
+    1,
+    2,
+    7,
+    8,
+    9,
+    10,
+    14,
+    25,
+    26,
+    27,
+    28,
+    29,
+    31,
+    58,
+    59,
+    60,
+    61,
+    62,
+    63,
+    90,
+    91,
+    92,
+    93,
+    359,
+    503,
+    522,
+    542,
+    873,
+    893,
+    902,
+    918,
+    922,
+    931,
+    1350,
+    1853,
+    1982,
+    2460,
+    2627,
+    3246,
+    3253,
+    3268,
+    3536,
+    3846,
+    3961,
+    4183,
+    4667,
+    6585,
+    6647,
+    7273,
+    9061,
+    9383,
+    10428,
+    10929,
+    11938,
+    12033,
+    12331,
+    12562,
+    13793,
+    14157,
+    14635,
+    15265,
+    15618,
+    16553,
+    16604,
+    18362,
+    18956,
+    20075,
+    21675,
+    22520,
+    26130,
+    26161,
+    26435,
+    28279,
+    29464,
+    31650,
+    32302,
+    32470,
+    36865,
+    42863,
+    47425,
+    49870,
+    50254,
+    50258,
+    50360,
+    50361,
+    50362
+  ],
+  "task_to_id": {
+    "transcribe": 50359,
+    "translate": 50358
+  },
+  "transformers_version": "4.27.0.dev0"
+}

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a2da5dd823173d2bb0639e696dd9b08106865184949bad079b2e1c22341e78d6
-size 967102601
+oid sha256:dc8d8ce98f7b8ed8ae3fe0154f1911f137be21ce23cfaa8fece1f7a80bd8aaf1
+size 483536061

runs/Dec06_22-57-28_rtx-i7/events.out.tfevents.1670342356.rtx-i7.82376.0

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:dd11c1a5aa32bb3584d848844a2d494d689dd8b837eea9b4be6930caa20e0f46
-size 24414

runs/{Dec06_14-37-40_rtx-i7/1670312302.611461/events.out.tfevents.1670312302.rtx-i7.150024.1 → Feb05_19-39-15_rtx-i7/1675600822.880342/events.out.tfevents.1675600822.rtx-i7}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c8b58f96d68854ca5a11fe8ae8e24cf1f936873e4213256826415da0a6f9ec13
-size 5771
+oid sha256:293238aa87be01a6b1962e6b3dc488ce4a5a65aefcb3a0d48a86c9927293f5b6
+size 5667

runs/{Dec06_14-37-40_rtx-i7/events.out.tfevents.1670312302.rtx-i7.150024.0 → Feb05_19-39-15_rtx-i7/events.out.tfevents.1675600822.rtx-i7}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4bfc01ffe1d6e8dd5c422f9ec93e6fe5f1e593c58265b2b6039942aa5be26aa3
-size 18075
+oid sha256:1cf4cac77a972e11e49522adfea296d163095148616dfbb0a0edf48b79729bd2
+size 36381

runs/{Dec06_22-57-28_rtx-i7/1670342356.6395447/events.out.tfevents.1670342356.rtx-i7.82376.1 → Feb05_19-39-15_rtx-i7/events.out.tfevents.1675621065.rtx-i7}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1d6df8ed5293a00fd6d407231712b19c03da6763addc4a2683d6eb2037fcf98c
-size 5771
+oid sha256:717c7fedcce3f5f5a022b22c64e6ed0dd34cba77e2363e0c130ca0f9eae0fd4a
+size 358

special_tokens_map.json

@@ -22,7 +22,7 @@
     "<|hi|>",
     "<|fi|>",
     "<|vi|>",
-    "<|iw|>",
+    "<|he|>",
     "<|uk|>",
     "<|el|>",
     "<|ms|>",
@@ -124,7 +124,7 @@
   },
   "pad_token": "<|endoftext|>",
   "unk_token": {
-    "content": "",
+    "content": "<|endoftext|>",
     "lstrip": false,
     "normalized": true,
     "rstrip": false,

tokenizer_config.json

@@ -27,7 +27,7 @@
   "tokenizer_class": "WhisperTokenizer",
   "unk_token": {
     "__type": "AddedToken",
-    "content": "",
+    "content": "<|endoftext|>",
     "lstrip": false,
     "normalized": true,
     "rstrip": false,

train_results.json

@@ -0,0 +1,7 @@
+{
+    "epoch": 83.33,
+    "train_loss": 0.031859876942634584,
+    "train_runtime": 20088.5298,
+    "train_samples_per_second": 15.929,
+    "train_steps_per_second": 0.249
+}

trainer_state.json

@@ -0,0 +1,1234 @@
+{
+  "best_metric": 27.279438445464898,
+  "best_model_checkpoint": "./checkpoint-5000",
+  "epoch": 83.33333333333333,
+  "global_step": 5000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.973833272194737e-06,
+      "loss": 0.7626,
+      "step": 25
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.195318418690893e-06,
+      "loss": 0.6204,
+      "step": 50
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 6.881634451095711e-06,
+      "loss": 0.5306,
+      "step": 75
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 7.361221988663844e-06,
+      "loss": 0.4583,
+      "step": 100
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.730207550743121e-06,
+      "loss": 0.4212,
+      "step": 125
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 8.03016458599496e-06,
+      "loss": 0.3621,
+      "step": 150
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 8.282894746203441e-06,
+      "loss": 0.3433,
+      "step": 175
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 8.501266121799902e-06,
+      "loss": 0.2989,
+      "step": 200
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 8.693512601774437e-06,
+      "loss": 0.2812,
+      "step": 225
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 8.865222471593567e-06,
+      "loss": 0.2559,
+      "step": 250
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 9.020362953730323e-06,
+      "loss": 0.2217,
+      "step": 275
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 9.161852281961698e-06,
+      "loss": 0.2143,
+      "step": 300
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 9.29189975311636e-06,
+      "loss": 0.1723,
+      "step": 325
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 9.412218256259678e-06,
+      "loss": 0.1688,
+      "step": 350
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 9.524162683365145e-06,
+      "loss": 0.1442,
+      "step": 375
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.62882322733502e-06,
+      "loss": 0.1264,
+      "step": 400
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 9.727090137141168e-06,
+      "loss": 0.1243,
+      "step": 425
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 9.819699807237934e-06,
+      "loss": 0.094,
+      "step": 450
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 9.907268307310855e-06,
+      "loss": 0.0933,
+      "step": 475
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 9.990316248055788e-06,
+      "loss": 0.0702,
+      "step": 500
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 9.953333333333333e-06,
+      "loss": 0.0678,
+      "step": 525
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 9.89777777777778e-06,
+      "loss": 0.0585,
+      "step": 550
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 9.842222222222223e-06,
+      "loss": 0.046,
+      "step": 575
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 9.786666666666667e-06,
+      "loss": 0.0464,
+      "step": 600
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 9.731111111111113e-06,
+      "loss": 0.0319,
+      "step": 625
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 9.675555555555555e-06,
+      "loss": 0.0315,
+      "step": 650
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 9.620000000000001e-06,
+      "loss": 0.026,
+      "step": 675
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 9.564444444444445e-06,
+      "loss": 0.0215,
+      "step": 700
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 9.508888888888889e-06,
+      "loss": 0.0195,
+      "step": 725
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 9.453333333333335e-06,
+      "loss": 0.0147,
+      "step": 750
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 9.397777777777779e-06,
+      "loss": 0.0145,
+      "step": 775
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 9.342222222222223e-06,
+      "loss": 0.0116,
+      "step": 800
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 9.286666666666667e-06,
+      "loss": 0.0109,
+      "step": 825
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 9.231111111111111e-06,
+      "loss": 0.0093,
+      "step": 850
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 9.175555555555557e-06,
+      "loss": 0.0079,
+      "step": 875
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 9.12e-06,
+      "loss": 0.0077,
+      "step": 900
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 9.064444444444447e-06,
+      "loss": 0.0064,
+      "step": 925
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 9.008888888888889e-06,
+      "loss": 0.006,
+      "step": 950
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 8.953333333333335e-06,
+      "loss": 0.0051,
+      "step": 975
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 8.897777777777779e-06,
+      "loss": 0.0046,
+      "step": 1000
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 8.842222222222223e-06,
+      "loss": 0.005,
+      "step": 1025
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 8.786666666666668e-06,
+      "loss": 0.0043,
+      "step": 1050
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 8.73111111111111e-06,
+      "loss": 0.0039,
+      "step": 1075
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 8.675555555555556e-06,
+      "loss": 0.003,
+      "step": 1100
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 8.62e-06,
+      "loss": 0.0029,
+      "step": 1125
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 8.564444444444445e-06,
+      "loss": 0.0029,
+      "step": 1150
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 8.50888888888889e-06,
+      "loss": 0.0025,
+      "step": 1175
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 8.453333333333334e-06,
+      "loss": 0.0023,
+      "step": 1200
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 8.397777777777778e-06,
+      "loss": 0.002,
+      "step": 1225
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 8.342222222222222e-06,
+      "loss": 0.0019,
+      "step": 1250
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 8.286666666666668e-06,
+      "loss": 0.0018,
+      "step": 1275
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 8.231111111111112e-06,
+      "loss": 0.0017,
+      "step": 1300
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 8.175555555555556e-06,
+      "loss": 0.0016,
+      "step": 1325
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 8.120000000000002e-06,
+      "loss": 0.0015,
+      "step": 1350
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 8.064444444444444e-06,
+      "loss": 0.0015,
+      "step": 1375
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 8.00888888888889e-06,
+      "loss": 0.0014,
+      "step": 1400
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 7.953333333333334e-06,
+      "loss": 0.0013,
+      "step": 1425
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 7.897777777777778e-06,
+      "loss": 0.0013,
+      "step": 1450
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 7.842222222222224e-06,
+      "loss": 0.0012,
+      "step": 1475
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 7.786666666666666e-06,
+      "loss": 0.0012,
+      "step": 1500
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 7.731111111111112e-06,
+      "loss": 0.0011,
+      "step": 1525
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 7.675555555555556e-06,
+      "loss": 0.0011,
+      "step": 1550
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 7.620000000000001e-06,
+      "loss": 0.001,
+      "step": 1575
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 7.564444444444446e-06,
+      "loss": 0.001,
+      "step": 1600
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 7.50888888888889e-06,
+      "loss": 0.0014,
+      "step": 1625
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 7.453333333333334e-06,
+      "loss": 0.0024,
+      "step": 1650
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 7.3977777777777786e-06,
+      "loss": 0.0033,
+      "step": 1675
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 7.342222222222223e-06,
+      "loss": 0.0051,
+      "step": 1700
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 7.2866666666666675e-06,
+      "loss": 0.0061,
+      "step": 1725
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 7.231111111111112e-06,
+      "loss": 0.0063,
+      "step": 1750
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 7.1755555555555556e-06,
+      "loss": 0.0058,
+      "step": 1775
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 7.1200000000000004e-06,
+      "loss": 0.0055,
+      "step": 1800
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 7.0644444444444445e-06,
+      "loss": 0.0043,
+      "step": 1825
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 7.008888888888889e-06,
+      "loss": 0.0047,
+      "step": 1850
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 6.953333333333334e-06,
+      "loss": 0.0047,
+      "step": 1875
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 6.897777777777779e-06,
+      "loss": 0.0044,
+      "step": 1900
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 6.842222222222222e-06,
+      "loss": 0.0034,
+      "step": 1925
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 6.786666666666667e-06,
+      "loss": 0.0034,
+      "step": 1950
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 6.731111111111111e-06,
+      "loss": 0.0029,
+      "step": 1975
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 6.675555555555556e-06,
+      "loss": 0.0026,
+      "step": 2000
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 6.6244444444444445e-06,
+      "loss": 0.0019,
+      "step": 2025
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 6.568888888888889e-06,
+      "loss": 0.0015,
+      "step": 2050
+    },
+    {
+      "epoch": 34.58,
+      "learning_rate": 6.513333333333333e-06,
+      "loss": 0.0012,
+      "step": 2075
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 6.457777777777778e-06,
+      "loss": 0.0011,
+      "step": 2100
+    },
+    {
+      "epoch": 35.42,
+      "learning_rate": 6.402222222222223e-06,
+      "loss": 0.0009,
+      "step": 2125
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 6.346666666666668e-06,
+      "loss": 0.001,
+      "step": 2150
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 6.291111111111111e-06,
+      "loss": 0.0008,
+      "step": 2175
+    },
+    {
+      "epoch": 36.67,
+      "learning_rate": 6.235555555555556e-06,
+      "loss": 0.0007,
+      "step": 2200
+    },
+    {
+      "epoch": 37.08,
+      "learning_rate": 6.18e-06,
+      "loss": 0.0007,
+      "step": 2225
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 6.124444444444445e-06,
+      "loss": 0.0007,
+      "step": 2250
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 6.06888888888889e-06,
+      "loss": 0.0008,
+      "step": 2275
+    },
+    {
+      "epoch": 38.33,
+      "learning_rate": 6.013333333333335e-06,
+      "loss": 0.0008,
+      "step": 2300
+    },
+    {
+      "epoch": 38.75,
+      "learning_rate": 5.957777777777778e-06,
+      "loss": 0.0008,
+      "step": 2325
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 5.902222222222223e-06,
+      "loss": 0.0007,
+      "step": 2350
+    },
+    {
+      "epoch": 39.58,
+      "learning_rate": 5.846666666666667e-06,
+      "loss": 0.0008,
+      "step": 2375
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 5.791111111111112e-06,
+      "loss": 0.0007,
+      "step": 2400
+    },
+    {
+      "epoch": 40.42,
+      "learning_rate": 5.735555555555557e-06,
+      "loss": 0.0006,
+      "step": 2425
+    },
+    {
+      "epoch": 40.83,
+      "learning_rate": 5.68e-06,
+      "loss": 0.0006,
+      "step": 2450
+    },
+    {
+      "epoch": 41.25,
+      "learning_rate": 5.624444444444445e-06,
+      "loss": 0.0005,
+      "step": 2475
+    },
+    {
+      "epoch": 41.67,
+      "learning_rate": 5.56888888888889e-06,
+      "loss": 0.0005,
+      "step": 2500
+    },
+    {
+      "epoch": 42.08,
+      "learning_rate": 5.513333333333334e-06,
+      "loss": 0.0005,
+      "step": 2525
+    },
+    {
+      "epoch": 42.5,
+      "learning_rate": 5.4577777777777785e-06,
+      "loss": 0.0005,
+      "step": 2550
+    },
+    {
+      "epoch": 42.92,
+      "learning_rate": 5.402222222222223e-06,
+      "loss": 0.0005,
+      "step": 2575
+    },
+    {
+      "epoch": 43.33,
+      "learning_rate": 5.346666666666667e-06,
+      "loss": 0.0005,
+      "step": 2600
+    },
+    {
+      "epoch": 43.75,
+      "learning_rate": 5.2911111111111115e-06,
+      "loss": 0.0005,
+      "step": 2625
+    },
+    {
+      "epoch": 44.17,
+      "learning_rate": 5.235555555555556e-06,
+      "loss": 0.0004,
+      "step": 2650
+    },
+    {
+      "epoch": 44.58,
+      "learning_rate": 5.18e-06,
+      "loss": 0.0005,
+      "step": 2675
+    },
+    {
+      "epoch": 45.0,
+      "learning_rate": 5.124444444444445e-06,
+      "loss": 0.0004,
+      "step": 2700
+    },
+    {
+      "epoch": 45.42,
+      "learning_rate": 5.06888888888889e-06,
+      "loss": 0.0004,
+      "step": 2725
+    },
+    {
+      "epoch": 45.83,
+      "learning_rate": 5.013333333333333e-06,
+      "loss": 0.0004,
+      "step": 2750
+    },
+    {
+      "epoch": 46.25,
+      "learning_rate": 4.957777777777778e-06,
+      "loss": 0.0004,
+      "step": 2775
+    },
+    {
+      "epoch": 46.67,
+      "learning_rate": 4.902222222222222e-06,
+      "loss": 0.0004,
+      "step": 2800
+    },
+    {
+      "epoch": 47.08,
+      "learning_rate": 4.846666666666667e-06,
+      "loss": 0.0004,
+      "step": 2825
+    },
+    {
+      "epoch": 47.5,
+      "learning_rate": 4.791111111111111e-06,
+      "loss": 0.0004,
+      "step": 2850
+    },
+    {
+      "epoch": 47.92,
+      "learning_rate": 4.735555555555556e-06,
+      "loss": 0.0004,
+      "step": 2875
+    },
+    {
+      "epoch": 48.33,
+      "learning_rate": 4.680000000000001e-06,
+      "loss": 0.0004,
+      "step": 2900
+    },
+    {
+      "epoch": 48.75,
+      "learning_rate": 4.624444444444445e-06,
+      "loss": 0.0004,
+      "step": 2925
+    },
+    {
+      "epoch": 49.17,
+      "learning_rate": 4.568888888888889e-06,
+      "loss": 0.0003,
+      "step": 2950
+    },
+    {
+      "epoch": 49.58,
+      "learning_rate": 4.513333333333333e-06,
+      "loss": 0.0003,
+      "step": 2975
+    },
+    {
+      "epoch": 50.0,
+      "learning_rate": 4.457777777777778e-06,
+      "loss": 0.0003,
+      "step": 3000
+    },
+    {
+      "epoch": 50.42,
+      "learning_rate": 4.406666666666667e-06,
+      "loss": 0.0003,
+      "step": 3025
+    },
+    {
+      "epoch": 50.83,
+      "learning_rate": 4.351111111111111e-06,
+      "loss": 0.0003,
+      "step": 3050
+    },
+    {
+      "epoch": 51.25,
+      "learning_rate": 4.295555555555556e-06,
+      "loss": 0.0003,
+      "step": 3075
+    },
+    {
+      "epoch": 51.67,
+      "learning_rate": 4.24e-06,
+      "loss": 0.0003,
+      "step": 3100
+    },
+    {
+      "epoch": 52.08,
+      "learning_rate": 4.184444444444445e-06,
+      "loss": 0.0003,
+      "step": 3125
+    },
+    {
+      "epoch": 52.5,
+      "learning_rate": 4.12888888888889e-06,
+      "loss": 0.0003,
+      "step": 3150
+    },
+    {
+      "epoch": 52.92,
+      "learning_rate": 4.073333333333334e-06,
+      "loss": 0.0003,
+      "step": 3175
+    },
+    {
+      "epoch": 53.33,
+      "learning_rate": 4.017777777777778e-06,
+      "loss": 0.0003,
+      "step": 3200
+    },
+    {
+      "epoch": 53.75,
+      "learning_rate": 3.962222222222222e-06,
+      "loss": 0.0003,
+      "step": 3225
+    },
+    {
+      "epoch": 54.17,
+      "learning_rate": 3.906666666666667e-06,
+      "loss": 0.0003,
+      "step": 3250
+    },
+    {
+      "epoch": 54.58,
+      "learning_rate": 3.851111111111112e-06,
+      "loss": 0.0003,
+      "step": 3275
+    },
+    {
+      "epoch": 55.0,
+      "learning_rate": 3.7955555555555557e-06,
+      "loss": 0.0003,
+      "step": 3300
+    },
+    {
+      "epoch": 55.42,
+      "learning_rate": 3.74e-06,
+      "loss": 0.0003,
+      "step": 3325
+    },
+    {
+      "epoch": 55.83,
+      "learning_rate": 3.684444444444445e-06,
+      "loss": 0.0003,
+      "step": 3350
+    },
+    {
+      "epoch": 56.25,
+      "learning_rate": 3.628888888888889e-06,
+      "loss": 0.0003,
+      "step": 3375
+    },
+    {
+      "epoch": 56.67,
+      "learning_rate": 3.5733333333333336e-06,
+      "loss": 0.0003,
+      "step": 3400
+    },
+    {
+      "epoch": 57.08,
+      "learning_rate": 3.5177777777777784e-06,
+      "loss": 0.0003,
+      "step": 3425
+    },
+    {
+      "epoch": 57.5,
+      "learning_rate": 3.4622222222222225e-06,
+      "loss": 0.0003,
+      "step": 3450
+    },
+    {
+      "epoch": 57.92,
+      "learning_rate": 3.406666666666667e-06,
+      "loss": 0.0003,
+      "step": 3475
+    },
+    {
+      "epoch": 58.33,
+      "learning_rate": 3.351111111111112e-06,
+      "loss": 0.0003,
+      "step": 3500
+    },
+    {
+      "epoch": 58.75,
+      "learning_rate": 3.295555555555556e-06,
+      "loss": 0.0003,
+      "step": 3525
+    },
+    {
+      "epoch": 59.17,
+      "learning_rate": 3.2400000000000003e-06,
+      "loss": 0.0003,
+      "step": 3550
+    },
+    {
+      "epoch": 59.58,
+      "learning_rate": 3.1844444444444444e-06,
+      "loss": 0.0003,
+      "step": 3575
+    },
+    {
+      "epoch": 60.0,
+      "learning_rate": 3.1288888888888892e-06,
+      "loss": 0.0003,
+      "step": 3600
+    },
+    {
+      "epoch": 60.42,
+      "learning_rate": 3.0733333333333337e-06,
+      "loss": 0.0002,
+      "step": 3625
+    },
+    {
+      "epoch": 60.83,
+      "learning_rate": 3.0177777777777777e-06,
+      "loss": 0.0003,
+      "step": 3650
+    },
+    {
+      "epoch": 61.25,
+      "learning_rate": 2.9622222222222226e-06,
+      "loss": 0.0002,
+      "step": 3675
+    },
+    {
+      "epoch": 61.67,
+      "learning_rate": 2.906666666666667e-06,
+      "loss": 0.0002,
+      "step": 3700
+    },
+    {
+      "epoch": 62.08,
+      "learning_rate": 2.851111111111111e-06,
+      "loss": 0.0002,
+      "step": 3725
+    },
+    {
+      "epoch": 62.5,
+      "learning_rate": 2.795555555555556e-06,
+      "loss": 0.0002,
+      "step": 3750
+    },
+    {
+      "epoch": 62.92,
+      "learning_rate": 2.7400000000000004e-06,
+      "loss": 0.0002,
+      "step": 3775
+    },
+    {
+      "epoch": 63.33,
+      "learning_rate": 2.6844444444444445e-06,
+      "loss": 0.0002,
+      "step": 3800
+    },
+    {
+      "epoch": 63.75,
+      "learning_rate": 2.6288888888888894e-06,
+      "loss": 0.0002,
+      "step": 3825
+    },
+    {
+      "epoch": 64.17,
+      "learning_rate": 2.573333333333334e-06,
+      "loss": 0.0002,
+      "step": 3850
+    },
+    {
+      "epoch": 64.58,
+      "learning_rate": 2.517777777777778e-06,
+      "loss": 0.0002,
+      "step": 3875
+    },
+    {
+      "epoch": 65.0,
+      "learning_rate": 2.4622222222222223e-06,
+      "loss": 0.0002,
+      "step": 3900
+    },
+    {
+      "epoch": 65.42,
+      "learning_rate": 2.4066666666666668e-06,
+      "loss": 0.0002,
+      "step": 3925
+    },
+    {
+      "epoch": 65.83,
+      "learning_rate": 2.3511111111111112e-06,
+      "loss": 0.0002,
+      "step": 3950
+    },
+    {
+      "epoch": 66.25,
+      "learning_rate": 2.2955555555555557e-06,
+      "loss": 0.0002,
+      "step": 3975
+    },
+    {
+      "epoch": 66.67,
+      "learning_rate": 2.24e-06,
+      "loss": 0.0002,
+      "step": 4000
+    },
+    {
+      "epoch": 67.08,
+      "learning_rate": 2.188888888888889e-06,
+      "loss": 0.0002,
+      "step": 4025
+    },
+    {
+      "epoch": 67.5,
+      "learning_rate": 2.133333333333334e-06,
+      "loss": 0.0002,
+      "step": 4050
+    },
+    {
+      "epoch": 67.92,
+      "learning_rate": 2.077777777777778e-06,
+      "loss": 0.0002,
+      "step": 4075
+    },
+    {
+      "epoch": 68.33,
+      "learning_rate": 2.0222222222222223e-06,
+      "loss": 0.0002,
+      "step": 4100
+    },
+    {
+      "epoch": 68.75,
+      "learning_rate": 1.9666666666666668e-06,
+      "loss": 0.0002,
+      "step": 4125
+    },
+    {
+      "epoch": 69.17,
+      "learning_rate": 1.9111111111111112e-06,
+      "loss": 0.0002,
+      "step": 4150
+    },
+    {
+      "epoch": 69.58,
+      "learning_rate": 1.8555555555555557e-06,
+      "loss": 0.0002,
+      "step": 4175
+    },
+    {
+      "epoch": 70.0,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 0.0002,
+      "step": 4200
+    },
+    {
+      "epoch": 70.42,
+      "learning_rate": 1.7444444444444448e-06,
+      "loss": 0.0002,
+      "step": 4225
+    },
+    {
+      "epoch": 70.83,
+      "learning_rate": 1.688888888888889e-06,
+      "loss": 0.0002,
+      "step": 4250
+    },
+    {
+      "epoch": 71.25,
+      "learning_rate": 1.6333333333333335e-06,
+      "loss": 0.0002,
+      "step": 4275
+    },
+    {
+      "epoch": 71.67,
+      "learning_rate": 1.5777777777777778e-06,
+      "loss": 0.0002,
+      "step": 4300
+    },
+    {
+      "epoch": 72.08,
+      "learning_rate": 1.5222222222222224e-06,
+      "loss": 0.0002,
+      "step": 4325
+    },
+    {
+      "epoch": 72.5,
+      "learning_rate": 1.4666666666666669e-06,
+      "loss": 0.0002,
+      "step": 4350
+    },
+    {
+      "epoch": 72.92,
+      "learning_rate": 1.4111111111111111e-06,
+      "loss": 0.0002,
+      "step": 4375
+    },
+    {
+      "epoch": 73.33,
+      "learning_rate": 1.3555555555555558e-06,
+      "loss": 0.0002,
+      "step": 4400
+    },
+    {
+      "epoch": 73.75,
+      "learning_rate": 1.3e-06,
+      "loss": 0.0002,
+      "step": 4425
+    },
+    {
+      "epoch": 74.17,
+      "learning_rate": 1.2444444444444445e-06,
+      "loss": 0.0002,
+      "step": 4450
+    },
+    {
+      "epoch": 74.58,
+      "learning_rate": 1.188888888888889e-06,
+      "loss": 0.0002,
+      "step": 4475
+    },
+    {
+      "epoch": 75.0,
+      "learning_rate": 1.1333333333333334e-06,
+      "loss": 0.0002,
+      "step": 4500
+    },
+    {
+      "epoch": 75.42,
+      "learning_rate": 1.0777777777777779e-06,
+      "loss": 0.0002,
+      "step": 4525
+    },
+    {
+      "epoch": 75.83,
+      "learning_rate": 1.0222222222222223e-06,
+      "loss": 0.0002,
+      "step": 4550
+    },
+    {
+      "epoch": 76.25,
+      "learning_rate": 9.666666666666668e-07,
+      "loss": 0.0002,
+      "step": 4575
+    },
+    {
+      "epoch": 76.67,
+      "learning_rate": 9.111111111111113e-07,
+      "loss": 0.0002,
+      "step": 4600
+    },
+    {
+      "epoch": 77.08,
+      "learning_rate": 8.555555555555556e-07,
+      "loss": 0.0002,
+      "step": 4625
+    },
+    {
+      "epoch": 77.5,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 0.0002,
+      "step": 4650
+    },
+    {
+      "epoch": 77.92,
+      "learning_rate": 7.444444444444444e-07,
+      "loss": 0.0002,
+      "step": 4675
+    },
+    {
+      "epoch": 78.33,
+      "learning_rate": 6.88888888888889e-07,
+      "loss": 0.0002,
+      "step": 4700
+    },
+    {
+      "epoch": 78.75,
+      "learning_rate": 6.333333333333334e-07,
+      "loss": 0.0002,
+      "step": 4725
+    },
+    {
+      "epoch": 79.17,
+      "learning_rate": 5.777777777777778e-07,
+      "loss": 0.0002,
+      "step": 4750
+    },
+    {
+      "epoch": 79.58,
+      "learning_rate": 5.222222222222223e-07,
+      "loss": 0.0002,
+      "step": 4775
+    },
+    {
+      "epoch": 80.0,
+      "learning_rate": 4.666666666666667e-07,
+      "loss": 0.0002,
+      "step": 4800
+    },
+    {
+      "epoch": 80.42,
+      "learning_rate": 4.111111111111112e-07,
+      "loss": 0.0002,
+      "step": 4825
+    },
+    {
+      "epoch": 80.83,
+      "learning_rate": 3.555555555555556e-07,
+      "loss": 0.0002,
+      "step": 4850
+    },
+    {
+      "epoch": 81.25,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 0.0002,
+      "step": 4875
+    },
+    {
+      "epoch": 81.67,
+      "learning_rate": 2.444444444444445e-07,
+      "loss": 0.0002,
+      "step": 4900
+    },
+    {
+      "epoch": 82.08,
+      "learning_rate": 1.888888888888889e-07,
+      "loss": 0.0002,
+      "step": 4925
+    },
+    {
+      "epoch": 82.5,
+      "learning_rate": 1.3333333333333336e-07,
+      "loss": 0.0002,
+      "step": 4950
+    },
+    {
+      "epoch": 82.92,
+      "learning_rate": 7.777777777777778e-08,
+      "loss": 0.0002,
+      "step": 4975
+    },
+    {
+      "epoch": 83.33,
+      "learning_rate": 2.2222222222222224e-08,
+      "loss": 0.0002,
+      "step": 5000
+    },
+    {
+      "epoch": 83.33,
+      "eval_loss": 0.50341796875,
+      "eval_runtime": 154.2054,
+      "eval_samples_per_second": 13.164,
+      "eval_steps_per_second": 0.415,
+      "eval_wer": 27.279438445464898,
+      "step": 5000
+    },
+    {
+      "epoch": 83.33,
+      "step": 5000,
+      "total_flos": 9.17485126904384e+19,
+      "train_loss": 0.031859876942634584,
+      "train_runtime": 20088.5298,
+      "train_samples_per_second": 15.929,
+      "train_steps_per_second": 0.249
+    }
+  ],
+  "max_steps": 5000,
+  "num_train_epochs": 84,
+  "total_flos": 9.17485126904384e+19,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fe0f907f13fd2cb0efddce2dce0e573993526bec55f9f4af7959a52fd9f3b66d
-size 3579
+oid sha256:c2fb67b34d1aaaf327e038dd1e7ff6fe1ac4201d6dc60f7bcbc9a0a6b743801e
+size 4731

vocab.json

@@ -314,6 +314,7 @@
   ";;": 35746,
   "<": 27,
   "</": 3433,
+  "<|endoftext|>": 50257,
   "=": 28,
   "=\"": 13114,
   "=\"#": 34106,