Init

.gitattributes

@@ -24,4 +24,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.xz filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zstandard filter=lfs diff=lfs merge=lfs -text
+*.arpa filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,3 +1,85 @@
----
-license: apache-2.0
----
+---
+language:
+- uk
+license: apache-2.0
+tags:
+- automatic-speech-recognition
+- mozilla-foundation/common_voice_7_0
+- generated_from_trainer
+- uk
+- robust-speech-event
+datasets:
+- common_voice
+model-index:
+- name: wav2vec2-xls-r-1b-uk-with-lm
+  results: []
+---
+
+# Ukrainian STT model (with Language Model)
+
+This model is a fine-tuned version of [facebook/wav2vec2-xls-r-1b](https://huggingface.co/facebook/wav2vec2-xls-r-1b) on the MOZILLA-FOUNDATION/COMMON_VOICE_7_0 - UK dataset.
+
+It achieves the following results on the evaluation set without the language model:
+
+- Loss: 0.1875
+- Wer: 0.2033
+- Cer: 0.0384
+
+Follow our comminity in Telegram: https://t.me/speech_recognition_uk 
+
+## Model description
+
+On 100 test example the model shows the following results:
+
+Without LM:
+
+- WER: 0.1862
+- CER: 0.0277
+
+With LM:
+
+- WER: 0.1218
+- CER: 0.0190
+
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 5e-05
+- train_batch_size: 8
+- eval_batch_size: 8
+- seed: 42
+- gradient_accumulation_steps: 20
+- total_train_batch_size: 160
+- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
+- lr_scheduler_type: linear
+- lr_scheduler_warmup_steps: 500
+- num_epochs: 100.0
+- mixed_precision_training: Native AMP
+
+### Training results
+
+| Training Loss | Epoch | Step | Validation Loss | Wer    | Cer    |
+|:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
+| 1.2815        | 7.93  | 500  | 0.3536          | 0.4753 | 0.1009 |
+| 1.0869        | 15.86 | 1000 | 0.2317          | 0.3111 | 0.0614 |
+| 0.9984        | 23.8  | 1500 | 0.2022          | 0.2676 | 0.0521 |
+| 0.975         | 31.74 | 2000 | 0.1948          | 0.2469 | 0.0487 |
+| 0.9306        | 39.67 | 2500 | 0.1916          | 0.2377 | 0.0464 |
+| 0.8868        | 47.61 | 3000 | 0.1903          | 0.2257 | 0.0439 |
+| 0.8424        | 55.55 | 3500 | 0.1786          | 0.2206 | 0.0423 |
+| 0.8126        | 63.49 | 4000 | 0.1849          | 0.2160 | 0.0416 |
+| 0.7901        | 71.42 | 4500 | 0.1869          | 0.2138 | 0.0413 |
+| 0.7671        | 79.36 | 5000 | 0.1855          | 0.2075 | 0.0394 |
+| 0.7467        | 87.3  | 5500 | 0.1884          | 0.2049 | 0.0389 |
+| 0.731         | 95.24 | 6000 | 0.1877          | 0.2060 | 0.0387 |
+
+
+### Framework versions
+
+- Transformers 4.16.0.dev0
+- Pytorch 1.10.1+cu102
+- Datasets 1.18.1.dev0
+- Tokenizers 0.11.0

added_tokens.json

@@ -0,0 +1 @@
+{"<s>": 51, "</s>": 52}

all_results.json

@@ -0,0 +1,15 @@
+{
+    "epoch": 99.99,
+    "eval_cer": 0.03843434091927693,
+    "eval_loss": 0.18747110664844513,
+    "eval_runtime": 268.1183,
+    "eval_samples": 4332,
+    "eval_samples_per_second": 16.157,
+    "eval_steps_per_second": 2.021,
+    "eval_wer": 0.20326104163368688,
+    "train_loss": 1.049089940994505,
+    "train_runtime": 95054.1856,
+    "train_samples": 10193,
+    "train_samples_per_second": 10.723,
+    "train_steps_per_second": 0.066
+}

alphabet.json

@@ -0,0 +1 @@
+{"labels": [" ", "a", "c", "e", "i", "j", "k", "l", "m", "n", "o", "p", "u", "x", "y", "\u0301", "\u0430", "\u0431", "\u0432", "\u0433", "\u0434", "\u0435", "\u0436", "\u0437", "\u0438", "\u0439", "\u043a", "\u043b", "\u043c", "\u043d", "\u043e", "\u043f", "\u0440", "\u0441", "\u0442", "\u0443", "\u0444", "\u0445", "\u0446", "\u0447", "\u0448", "\u0449", "\u044c", "\u044e", "\u044f", "\u0454", "\u0456", "\u0457", "\u0491", "\u2047", "", "<s>", "</s>"], "is_bpe": false}

config.json

@@ -0,0 +1,107 @@
+{
+  "_name_or_path": "facebook/wav2vec2-xls-r-1b",
+  "activation_dropout": 0.1,
+  "adapter_kernel_size": 3,
+  "adapter_stride": 2,
+  "add_adapter": false,
+  "apply_spec_augment": true,
+  "architectures": [
+    "Wav2Vec2ForCTC"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "classifier_proj_size": 256,
+  "codevector_dim": 1024,
+  "contrastive_logits_temperature": 0.1,
+  "conv_bias": true,
+  "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": true,
+  "eos_token_id": 2,
+  "feat_extract_activation": "gelu",
+  "feat_extract_dropout": 0.0,
+  "feat_extract_norm": "layer",
+  "feat_proj_dropout": 0.0,
+  "feat_quantizer_dropout": 0.0,
+  "final_dropout": 0.0,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.0,
+  "hidden_size": 1280,
+  "initializer_range": 0.02,
+  "intermediate_size": 5120,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.0,
+  "mask_feature_length": 64,
+  "mask_feature_min_masks": 0,
+  "mask_feature_prob": 0.25,
+  "mask_time_length": 10,
+  "mask_time_min_masks": 2,
+  "mask_time_prob": 0.75,
+  "model_type": "wav2vec2",
+  "num_adapter_layers": 3,
+  "num_attention_heads": 16,
+  "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": 48,
+  "num_negatives": 100,
+  "output_hidden_size": 1280,
+  "pad_token_id": 50,
+  "proj_codevector_dim": 1024,
+  "tdnn_dilation": [
+    1,
+    2,
+    3,
+    1,
+    1
+  ],
+  "tdnn_dim": [
+    512,
+    512,
+    512,
+    512,
+    1500
+  ],
+  "tdnn_kernel": [
+    5,
+    3,
+    3,
+    1,
+    1
+  ],
+  "torch_dtype": "float32",
+  "transformers_version": "4.16.0.dev0",
+  "use_weighted_layer_sum": false,
+  "vocab_size": 53,
+  "xvector_output_dim": 512
+}

eval_results.json

@@ -0,0 +1,10 @@
+{
+    "epoch": 99.99,
+    "eval_cer": 0.03843434091927693,
+    "eval_loss": 0.18747110664844513,
+    "eval_runtime": 268.1183,
+    "eval_samples": 4332,
+    "eval_samples_per_second": 16.157,
+    "eval_steps_per_second": 2.021,
+    "eval_wer": 0.20326104163368688
+}

language_model/5gram_correct.arpa

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8720d88d4f86f9bdd24a806713c0d4be2eb694227daed83bedade8ba69c16516
+size 28311207

language_model/attrs.json

@@ -0,0 +1 @@
+{"alpha": 0.5, "beta": 1.5, "unk_score_offset": -10.0, "score_boundary": true}

preprocessor_config.json

@@ -0,0 +1,10 @@
+{
+  "do_normalize": true,
+  "feature_extractor_type": "Wav2Vec2FeatureExtractor",
+  "feature_size": 1,
+  "padding_side": "right",
+  "padding_value": 0,
+  "processor_class": "Wav2Vec2ProcessorWithLM",
+  "return_attention_mask": true,
+  "sampling_rate": 16000
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6933edff2d809c0035d6ce0fffe2bf975c5d0d3f2932f2c4545f47a769d04ce4
+size 3850584305

runs/Jan26_11-22-42_job-df329b21-d243-4736-8f96-d11192aeb370/1643196546.6195295/events.out.tfevents.1643196546.job-df329b21-d243-4736-8f96-d11192aeb370.13810.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dccefa87dfb9f1a70748307d90324380c39d8b93d9af111a89d51ba81de70d24
+size 4778

runs/Jan26_11-22-42_job-df329b21-d243-4736-8f96-d11192aeb370/events.out.tfevents.1643196546.job-df329b21-d243-4736-8f96-d11192aeb370.13810.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1dd1de0119ae5b997385fc2a7feb994debd200ac3055775715f8355a07eb0d2
+size 29217

runs/Jan26_11-22-42_job-df329b21-d243-4736-8f96-d11192aeb370/events.out.tfevents.1643291875.job-df329b21-d243-4736-8f96-d11192aeb370.13810.2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cddb9d99a97b8c4323dfd96a1f6fef25c3c7ff9009b73aa856629224e0ce4325
+size 405

special_tokens_map.json

@@ -0,0 +1 @@
+{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "[UNK]", "pad_token": "[PAD]", "additional_special_tokens": [{"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}]}

tokenizer_config.json

@@ -0,0 +1 @@
+{"unk_token": "[UNK]", "bos_token": "<s>", "eos_token": "</s>", "pad_token": "[PAD]", "do_lower_case": false, "word_delimiter_token": "|", "special_tokens_map_file": null, "tokenizer_file": null, "name_or_path": "final", "tokenizer_class": "Wav2Vec2CTCTokenizer", "processor_class": "Wav2Vec2ProcessorWithLM"}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 99.99,
+    "train_loss": 1.049089940994505,
+    "train_runtime": 95054.1856,
+    "train_samples": 10193,
+    "train_samples_per_second": 10.723,
+    "train_steps_per_second": 0.066
+}

trainer_state.json

@@ -0,0 +1,901 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 99.98823529411764,
+  "global_step": 6300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.9000000000000005e-06,
+      "loss": 11.1133,
+      "step": 50
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.900000000000002e-06,
+      "loss": 3.3967,
+      "step": 100
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 1.49e-05,
+      "loss": 3.2205,
+      "step": 150
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 1.9900000000000003e-05,
+      "loss": 2.8143,
+      "step": 200
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 2.4900000000000002e-05,
+      "loss": 1.9249,
+      "step": 250
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 2.9900000000000002e-05,
+      "loss": 1.6708,
+      "step": 300
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.49e-05,
+      "loss": 1.5501,
+      "step": 350
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.99e-05,
+      "loss": 1.4258,
+      "step": 400
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 4.49e-05,
+      "loss": 1.332,
+      "step": 450
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 4.99e-05,
+      "loss": 1.2815,
+      "step": 500
+    },
+    {
+      "epoch": 7.93,
+      "eval_cer": 0.10093122852447588,
+      "eval_loss": 0.35359087586402893,
+      "eval_runtime": 257.0651,
+      "eval_samples_per_second": 16.852,
+      "eval_steps_per_second": 2.108,
+      "eval_wer": 0.47525724236188066,
+      "step": 500
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 4.957758620689655e-05,
+      "loss": 1.2632,
+      "step": 550
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 4.9146551724137934e-05,
+      "loss": 1.2239,
+      "step": 600
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 4.871551724137931e-05,
+      "loss": 1.2044,
+      "step": 650
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 4.828448275862069e-05,
+      "loss": 1.1918,
+      "step": 700
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 4.785344827586207e-05,
+      "loss": 1.1641,
+      "step": 750
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 4.742241379310345e-05,
+      "loss": 1.1718,
+      "step": 800
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 4.699137931034483e-05,
+      "loss": 1.1638,
+      "step": 850
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 4.656034482758621e-05,
+      "loss": 1.1317,
+      "step": 900
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 4.612931034482759e-05,
+      "loss": 1.1334,
+      "step": 950
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 4.569827586206897e-05,
+      "loss": 1.0869,
+      "step": 1000
+    },
+    {
+      "epoch": 15.86,
+      "eval_cer": 0.06135152631841044,
+      "eval_loss": 0.23165984451770782,
+      "eval_runtime": 262.4285,
+      "eval_samples_per_second": 16.507,
+      "eval_steps_per_second": 2.065,
+      "eval_wer": 0.3110653791356657,
+      "step": 1000
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 4.526724137931035e-05,
+      "loss": 1.104,
+      "step": 1050
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 4.4836206896551726e-05,
+      "loss": 1.109,
+      "step": 1100
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.440517241379311e-05,
+      "loss": 1.0902,
+      "step": 1150
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 4.397413793103449e-05,
+      "loss": 1.0676,
+      "step": 1200
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.3543103448275865e-05,
+      "loss": 1.0453,
+      "step": 1250
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 4.311206896551725e-05,
+      "loss": 1.0489,
+      "step": 1300
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 4.268103448275862e-05,
+      "loss": 1.0495,
+      "step": 1350
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 4.2250000000000004e-05,
+      "loss": 1.0325,
+      "step": 1400
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 4.181896551724138e-05,
+      "loss": 1.0298,
+      "step": 1450
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 4.138793103448276e-05,
+      "loss": 0.9984,
+      "step": 1500
+    },
+    {
+      "epoch": 23.8,
+      "eval_cer": 0.052054180568696776,
+      "eval_loss": 0.20215292274951935,
+      "eval_runtime": 259.0562,
+      "eval_samples_per_second": 16.722,
+      "eval_steps_per_second": 2.092,
+      "eval_wer": 0.26762703815102107,
+      "step": 1500
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 4.0956896551724136e-05,
+      "loss": 1.0118,
+      "step": 1550
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 4.053448275862069e-05,
+      "loss": 1.0165,
+      "step": 1600
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 4.0103448275862074e-05,
+      "loss": 1.0075,
+      "step": 1650
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 3.967241379310345e-05,
+      "loss": 1.003,
+      "step": 1700
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 3.924137931034483e-05,
+      "loss": 0.9905,
+      "step": 1750
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 3.8810344827586206e-05,
+      "loss": 1.0019,
+      "step": 1800
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 3.837931034482759e-05,
+      "loss": 1.0085,
+      "step": 1850
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 3.794827586206896e-05,
+      "loss": 0.9868,
+      "step": 1900
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 3.7517241379310345e-05,
+      "loss": 0.9816,
+      "step": 1950
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 3.708620689655173e-05,
+      "loss": 0.975,
+      "step": 2000
+    },
+    {
+      "epoch": 31.74,
+      "eval_cer": 0.04868781435187491,
+      "eval_loss": 0.19483695924282074,
+      "eval_runtime": 259.0795,
+      "eval_samples_per_second": 16.721,
+      "eval_steps_per_second": 2.092,
+      "eval_wer": 0.24688934620864333,
+      "step": 2000
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 3.66551724137931e-05,
+      "loss": 0.9552,
+      "step": 2050
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 3.6224137931034484e-05,
+      "loss": 0.9649,
+      "step": 2100
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 3.5793103448275866e-05,
+      "loss": 0.9632,
+      "step": 2150
+    },
+    {
+      "epoch": 34.91,
+      "learning_rate": 3.536206896551724e-05,
+      "loss": 0.9542,
+      "step": 2200
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 3.493103448275862e-05,
+      "loss": 0.9686,
+      "step": 2250
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 3.45e-05,
+      "loss": 0.9418,
+      "step": 2300
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 3.406896551724138e-05,
+      "loss": 0.9295,
+      "step": 2350
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 3.363793103448276e-05,
+      "loss": 0.9372,
+      "step": 2400
+    },
+    {
+      "epoch": 38.88,
+      "learning_rate": 3.320689655172414e-05,
+      "loss": 0.9205,
+      "step": 2450
+    },
+    {
+      "epoch": 39.67,
+      "learning_rate": 3.277586206896552e-05,
+      "loss": 0.9306,
+      "step": 2500
+    },
+    {
+      "epoch": 39.67,
+      "eval_cer": 0.046377172451571136,
+      "eval_loss": 0.19161736965179443,
+      "eval_runtime": 258.3157,
+      "eval_samples_per_second": 16.77,
+      "eval_steps_per_second": 2.098,
+      "eval_wer": 0.2377394332752889,
+      "step": 2500
+    },
+    {
+      "epoch": 40.47,
+      "learning_rate": 3.23448275862069e-05,
+      "loss": 0.9331,
+      "step": 2550
+    },
+    {
+      "epoch": 41.27,
+      "learning_rate": 3.1913793103448276e-05,
+      "loss": 0.8936,
+      "step": 2600
+    },
+    {
+      "epoch": 42.06,
+      "learning_rate": 3.148275862068966e-05,
+      "loss": 0.8987,
+      "step": 2650
+    },
+    {
+      "epoch": 42.85,
+      "learning_rate": 3.105172413793104e-05,
+      "loss": 0.8853,
+      "step": 2700
+    },
+    {
+      "epoch": 43.64,
+      "learning_rate": 3.0620689655172415e-05,
+      "loss": 0.9106,
+      "step": 2750
+    },
+    {
+      "epoch": 44.44,
+      "learning_rate": 3.0189655172413794e-05,
+      "loss": 0.8932,
+      "step": 2800
+    },
+    {
+      "epoch": 45.24,
+      "learning_rate": 2.9758620689655176e-05,
+      "loss": 0.9096,
+      "step": 2850
+    },
+    {
+      "epoch": 46.03,
+      "learning_rate": 2.932758620689655e-05,
+      "loss": 0.8919,
+      "step": 2900
+    },
+    {
+      "epoch": 46.82,
+      "learning_rate": 2.8896551724137933e-05,
+      "loss": 0.8744,
+      "step": 2950
+    },
+    {
+      "epoch": 47.61,
+      "learning_rate": 2.8465517241379315e-05,
+      "loss": 0.8868,
+      "step": 3000
+    },
+    {
+      "epoch": 47.61,
+      "eval_cer": 0.04391713560081669,
+      "eval_loss": 0.19031885266304016,
+      "eval_runtime": 265.4438,
+      "eval_samples_per_second": 16.32,
+      "eval_steps_per_second": 2.042,
+      "eval_wer": 0.2257400664872566,
+      "step": 3000
+    },
+    {
+      "epoch": 48.41,
+      "learning_rate": 2.803448275862069e-05,
+      "loss": 0.8793,
+      "step": 3050
+    },
+    {
+      "epoch": 49.2,
+      "learning_rate": 2.7603448275862072e-05,
+      "loss": 0.8739,
+      "step": 3100
+    },
+    {
+      "epoch": 49.99,
+      "learning_rate": 2.717241379310345e-05,
+      "loss": 0.8696,
+      "step": 3150
+    },
+    {
+      "epoch": 50.78,
+      "learning_rate": 2.674137931034483e-05,
+      "loss": 0.863,
+      "step": 3200
+    },
+    {
+      "epoch": 51.58,
+      "learning_rate": 2.6310344827586207e-05,
+      "loss": 0.8612,
+      "step": 3250
+    },
+    {
+      "epoch": 52.38,
+      "learning_rate": 2.587931034482759e-05,
+      "loss": 0.8639,
+      "step": 3300
+    },
+    {
+      "epoch": 53.17,
+      "learning_rate": 2.5448275862068964e-05,
+      "loss": 0.8523,
+      "step": 3350
+    },
+    {
+      "epoch": 53.96,
+      "learning_rate": 2.5017241379310346e-05,
+      "loss": 0.8577,
+      "step": 3400
+    },
+    {
+      "epoch": 54.75,
+      "learning_rate": 2.4586206896551725e-05,
+      "loss": 0.8465,
+      "step": 3450
+    },
+    {
+      "epoch": 55.55,
+      "learning_rate": 2.4155172413793103e-05,
+      "loss": 0.8424,
+      "step": 3500
+    },
+    {
+      "epoch": 55.55,
+      "eval_cer": 0.042293710472586024,
+      "eval_loss": 0.17861121892929077,
+      "eval_runtime": 259.0288,
+      "eval_samples_per_second": 16.724,
+      "eval_steps_per_second": 2.092,
+      "eval_wer": 0.22061104954883648,
+      "step": 3500
+    },
+    {
+      "epoch": 56.35,
+      "learning_rate": 2.3724137931034485e-05,
+      "loss": 0.8436,
+      "step": 3550
+    },
+    {
+      "epoch": 57.14,
+      "learning_rate": 2.3293103448275864e-05,
+      "loss": 0.8404,
+      "step": 3600
+    },
+    {
+      "epoch": 57.93,
+      "learning_rate": 2.2862068965517242e-05,
+      "loss": 0.8304,
+      "step": 3650
+    },
+    {
+      "epoch": 58.72,
+      "learning_rate": 2.2431034482758624e-05,
+      "loss": 0.8331,
+      "step": 3700
+    },
+    {
+      "epoch": 59.52,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 0.824,
+      "step": 3750
+    },
+    {
+      "epoch": 60.31,
+      "learning_rate": 2.1568965517241378e-05,
+      "loss": 0.8328,
+      "step": 3800
+    },
+    {
+      "epoch": 61.11,
+      "learning_rate": 2.113793103448276e-05,
+      "loss": 0.8234,
+      "step": 3850
+    },
+    {
+      "epoch": 61.89,
+      "learning_rate": 2.070689655172414e-05,
+      "loss": 0.8098,
+      "step": 3900
+    },
+    {
+      "epoch": 62.69,
+      "learning_rate": 2.0275862068965517e-05,
+      "loss": 0.8287,
+      "step": 3950
+    },
+    {
+      "epoch": 63.49,
+      "learning_rate": 1.98448275862069e-05,
+      "loss": 0.8126,
+      "step": 4000
+    },
+    {
+      "epoch": 63.49,
+      "eval_cer": 0.04164135252228475,
+      "eval_loss": 0.18486249446868896,
+      "eval_runtime": 261.7127,
+      "eval_samples_per_second": 16.553,
+      "eval_steps_per_second": 2.071,
+      "eval_wer": 0.2159886021845813,
+      "step": 4000
+    },
+    {
+      "epoch": 64.28,
+      "learning_rate": 1.9413793103448277e-05,
+      "loss": 0.8089,
+      "step": 4050
+    },
+    {
+      "epoch": 65.08,
+      "learning_rate": 1.8982758620689656e-05,
+      "loss": 0.8126,
+      "step": 4100
+    },
+    {
+      "epoch": 65.86,
+      "learning_rate": 1.8551724137931034e-05,
+      "loss": 0.7975,
+      "step": 4150
+    },
+    {
+      "epoch": 66.66,
+      "learning_rate": 1.8120689655172416e-05,
+      "loss": 0.8049,
+      "step": 4200
+    },
+    {
+      "epoch": 67.45,
+      "learning_rate": 1.7698275862068966e-05,
+      "loss": 0.8088,
+      "step": 4250
+    },
+    {
+      "epoch": 68.25,
+      "learning_rate": 1.7267241379310344e-05,
+      "loss": 0.8038,
+      "step": 4300
+    },
+    {
+      "epoch": 69.05,
+      "learning_rate": 1.6836206896551726e-05,
+      "loss": 0.7886,
+      "step": 4350
+    },
+    {
+      "epoch": 69.83,
+      "learning_rate": 1.6405172413793105e-05,
+      "loss": 0.7735,
+      "step": 4400
+    },
+    {
+      "epoch": 70.63,
+      "learning_rate": 1.5974137931034483e-05,
+      "loss": 0.7837,
+      "step": 4450
+    },
+    {
+      "epoch": 71.42,
+      "learning_rate": 1.5543103448275865e-05,
+      "loss": 0.7901,
+      "step": 4500
+    },
+    {
+      "epoch": 71.42,
+      "eval_cer": 0.04126786514615806,
+      "eval_loss": 0.18691900372505188,
+      "eval_runtime": 261.5464,
+      "eval_samples_per_second": 16.563,
+      "eval_steps_per_second": 2.072,
+      "eval_wer": 0.21383568149438023,
+      "step": 4500
+    },
+    {
+      "epoch": 72.22,
+      "learning_rate": 1.5112068965517242e-05,
+      "loss": 0.7949,
+      "step": 4550
+    },
+    {
+      "epoch": 73.02,
+      "learning_rate": 1.468103448275862e-05,
+      "loss": 0.7893,
+      "step": 4600
+    },
+    {
+      "epoch": 73.8,
+      "learning_rate": 1.4249999999999999e-05,
+      "loss": 0.7603,
+      "step": 4650
+    },
+    {
+      "epoch": 74.6,
+      "learning_rate": 1.3818965517241381e-05,
+      "loss": 0.776,
+      "step": 4700
+    },
+    {
+      "epoch": 75.39,
+      "learning_rate": 1.338793103448276e-05,
+      "loss": 0.7755,
+      "step": 4750
+    },
+    {
+      "epoch": 76.19,
+      "learning_rate": 1.2956896551724138e-05,
+      "loss": 0.7751,
+      "step": 4800
+    },
+    {
+      "epoch": 76.97,
+      "learning_rate": 1.2525862068965518e-05,
+      "loss": 0.7608,
+      "step": 4850
+    },
+    {
+      "epoch": 77.77,
+      "learning_rate": 1.2094827586206897e-05,
+      "loss": 0.7663,
+      "step": 4900
+    },
+    {
+      "epoch": 78.56,
+      "learning_rate": 1.1663793103448277e-05,
+      "loss": 0.7656,
+      "step": 4950
+    },
+    {
+      "epoch": 79.36,
+      "learning_rate": 1.1232758620689656e-05,
+      "loss": 0.7671,
+      "step": 5000
+    },
+    {
+      "epoch": 79.36,
+      "eval_cer": 0.03937054927543449,
+      "eval_loss": 0.18550464510917664,
+      "eval_runtime": 260.3539,
+      "eval_samples_per_second": 16.639,
+      "eval_steps_per_second": 2.082,
+      "eval_wer": 0.20747190121893302,
+      "step": 5000
+    },
+    {
+      "epoch": 80.16,
+      "learning_rate": 1.0801724137931036e-05,
+      "loss": 0.7694,
+      "step": 5050
+    },
+    {
+      "epoch": 80.94,
+      "learning_rate": 1.0370689655172414e-05,
+      "loss": 0.7672,
+      "step": 5100
+    },
+    {
+      "epoch": 81.74,
+      "learning_rate": 9.939655172413793e-06,
+      "loss": 0.7444,
+      "step": 5150
+    },
+    {
+      "epoch": 82.53,
+      "learning_rate": 9.508620689655173e-06,
+      "loss": 0.7534,
+      "step": 5200
+    },
+    {
+      "epoch": 83.33,
+      "learning_rate": 9.077586206896552e-06,
+      "loss": 0.7453,
+      "step": 5250
+    },
+    {
+      "epoch": 84.13,
+      "learning_rate": 8.646551724137932e-06,
+      "loss": 0.7494,
+      "step": 5300
+    },
+    {
+      "epoch": 84.91,
+      "learning_rate": 8.224137931034483e-06,
+      "loss": 0.7425,
+      "step": 5350
+    },
+    {
+      "epoch": 85.71,
+      "learning_rate": 7.793103448275863e-06,
+      "loss": 0.7499,
+      "step": 5400
+    },
+    {
+      "epoch": 86.5,
+      "learning_rate": 7.370689655172413e-06,
+      "loss": 0.735,
+      "step": 5450
+    },
+    {
+      "epoch": 87.3,
+      "learning_rate": 6.939655172413794e-06,
+      "loss": 0.7467,
+      "step": 5500
+    },
+    {
+      "epoch": 87.3,
+      "eval_cer": 0.03894228375080922,
+      "eval_loss": 0.18841499090194702,
+      "eval_runtime": 261.43,
+      "eval_samples_per_second": 16.57,
+      "eval_steps_per_second": 2.073,
+      "eval_wer": 0.20490739274972297,
+      "step": 5500
+    },
+    {
+      "epoch": 88.09,
+      "learning_rate": 6.508620689655173e-06,
+      "loss": 0.7348,
+      "step": 5550
+    },
+    {
+      "epoch": 88.88,
+      "learning_rate": 6.0775862068965515e-06,
+      "loss": 0.7244,
+      "step": 5600
+    },
+    {
+      "epoch": 89.67,
+      "learning_rate": 5.646551724137932e-06,
+      "loss": 0.7394,
+      "step": 5650
+    },
+    {
+      "epoch": 90.47,
+      "learning_rate": 5.21551724137931e-06,
+      "loss": 0.7423,
+      "step": 5700
+    },
+    {
+      "epoch": 91.27,
+      "learning_rate": 4.78448275862069e-06,
+      "loss": 0.7251,
+      "step": 5750
+    },
+    {
+      "epoch": 92.06,
+      "learning_rate": 4.353448275862069e-06,
+      "loss": 0.7304,
+      "step": 5800
+    },
+    {
+      "epoch": 92.85,
+      "learning_rate": 3.9224137931034484e-06,
+      "loss": 0.7153,
+      "step": 5850
+    },
+    {
+      "epoch": 93.64,
+      "learning_rate": 3.491379310344828e-06,
+      "loss": 0.7287,
+      "step": 5900
+    },
+    {
+      "epoch": 94.44,
+      "learning_rate": 3.0603448275862068e-06,
+      "loss": 0.7349,
+      "step": 5950
+    },
+    {
+      "epoch": 95.24,
+      "learning_rate": 2.6293103448275866e-06,
+      "loss": 0.731,
+      "step": 6000
+    },
+    {
+      "epoch": 95.24,
+      "eval_cer": 0.03871819132513321,
+      "eval_loss": 0.1877404898405075,
+      "eval_runtime": 259.3367,
+      "eval_samples_per_second": 16.704,
+      "eval_steps_per_second": 2.09,
+      "eval_wer": 0.2059838530948235,
+      "step": 6000
+    },
+    {
+      "epoch": 96.03,
+      "learning_rate": 2.1982758620689655e-06,
+      "loss": 0.7151,
+      "step": 6050
+    },
+    {
+      "epoch": 96.82,
+      "learning_rate": 1.7672413793103449e-06,
+      "loss": 0.713,
+      "step": 6100
+    },
+    {
+      "epoch": 97.61,
+      "learning_rate": 1.3362068965517243e-06,
+      "loss": 0.7257,
+      "step": 6150
+    },
+    {
+      "epoch": 98.41,
+      "learning_rate": 9.051724137931035e-07,
+      "loss": 0.7287,
+      "step": 6200
+    },
+    {
+      "epoch": 99.2,
+      "learning_rate": 4.7413793103448276e-07,
+      "loss": 0.7273,
+      "step": 6250
+    },
+    {
+      "epoch": 99.99,
+      "learning_rate": 4.310344827586207e-08,
+      "loss": 0.7082,
+      "step": 6300
+    },
+    {
+      "epoch": 99.99,
+      "step": 6300,
+      "total_flos": 4.0887685530877926e+20,
+      "train_loss": 1.049089940994505,
+      "train_runtime": 95054.1856,
+      "train_samples_per_second": 10.723,
+      "train_steps_per_second": 0.066
+    }
+  ],
+  "max_steps": 6300,
+  "num_train_epochs": 100,
+  "total_flos": 4.0887685530877926e+20,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d28ad45d3721baa2d9d1ae37754a308d7348290cd7784d5d9c00aa9c1ffe0061
+size 3055

vocab.json

@@ -0,0 +1 @@
+{"a": 1, "c": 2, "e": 3, "i": 4, "j": 5, "k": 6, "l": 7, "m": 8, "n": 9, "o": 10, "p": 11, "u": 12, "x": 13, "y": 14, "́": 15, "а": 16, "б": 17, "в": 18, "г": 19, "д": 20, "е": 21, "ж": 22, "з": 23, "и": 24, "й": 25, "к": 26, "л": 27, "м": 28, "н": 29, "о": 30, "п": 31, "р": 32, "с": 33, "т": 34, "у": 35, "ф": 36, "х": 37, "ц": 38, "ч": 39, "ш": 40, "щ": 41, "ь": 42, "ю": 43, "я": 44, "є": 45, "і": 46, "ї": 47, "ґ": 48, "|": 0, "[UNK]": 49, "[PAD]": 50}