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: cv-uk-ft
+  results: []
+---
+
+
+# Ukrainian STT model (with Language Model)
+
+This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the MOZILLA-FOUNDATION/COMMON_VOICE_7_0 - UK dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.3015
+- Wer: 0.3377
+- Cer: 0.0708
+
+The above results present evaluation without the language model.
+
+
+## Model description
+
+On 100 test example the model shows the following results:
+
+Without LM:
+
+- WER: 0.2647058823529412
+- CER: 0.046974185357596274
+
+With LM:
+
+- WER: 0.1568627450980392
+- CER: 0.028988573846804908
+
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 0.0003
+- 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    |
+|:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
+| 3.0255        | 7.93  | 500  | 2.5514          | 0.9921 | 0.9047 |
+| 1.3809        | 15.86 | 1000 | 0.4065          | 0.5361 | 0.1201 |
+| 1.2355        | 23.8  | 1500 | 0.3474          | 0.4618 | 0.1033 |
+| 1.1956        | 31.74 | 2000 | 0.3617          | 0.4580 | 0.1005 |
+| 1.1416        | 39.67 | 2500 | 0.3182          | 0.4074 | 0.0891 |
+| 1.0996        | 47.61 | 3000 | 0.3166          | 0.3985 | 0.0875 |
+| 1.0427        | 55.55 | 3500 | 0.3116          | 0.3835 | 0.0828 |
+| 0.9961        | 63.49 | 4000 | 0.3137          | 0.3757 | 0.0807 |
+| 0.9575        | 71.42 | 4500 | 0.2992          | 0.3632 | 0.0771 |
+| 0.9154        | 79.36 | 5000 | 0.3015          | 0.3502 | 0.0740 |
+| 0.8994        | 87.3  | 5500 | 0.3004          | 0.3425 | 0.0723 |
+| 0.871         | 95.24 | 6000 | 0.3016          | 0.3394 | 0.0713 |
+
+
+### 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>": 59, "</s>": 60}

all_results.json

@@ -0,0 +1,15 @@
+{
+    "epoch": 99.99,
+    "eval_cer": 0.07081761068660654,
+    "eval_loss": 0.3015391230583191,
+    "eval_runtime": 185.2936,
+    "eval_samples": 4332,
+    "eval_samples_per_second": 23.379,
+    "eval_steps_per_second": 2.925,
+    "eval_wer": 0.337661929317067,
+    "train_loss": 1.3816477500067816,
+    "train_runtime": 56302.7905,
+    "train_samples": 10193,
+    "train_samples_per_second": 18.104,
+    "train_steps_per_second": 0.112
+}

alphabet.json

@@ -0,0 +1 @@
+{"labels": [" ", "`", "a", "c", "e", "i", "j", "k", "l", "m", "n", "o", "p", "u", "x", "y", "\u00ab", "\u00bb", "\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", "\u044b", "\u044c", "\u044e", "\u044f", "\u0454", "\u0456", "\u0457", "\u0491", "\u055a", "\u2013", "\u2019", "\u2026", "\u2047", "", "<s>", "</s>"], "is_bpe": false}

config.json

@@ -0,0 +1,107 @@
+{
+  "_name_or_path": "facebook/wav2vec2-xls-r-300m",
+  "activation_dropout": 0.1,
+  "adapter_kernel_size": 3,
+  "adapter_stride": 2,
+  "add_adapter": false,
+  "apply_spec_augment": true,
+  "architectures": [
+    "Wav2Vec2ForPreTraining"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "classifier_proj_size": 256,
+  "codevector_dim": 768,
+  "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": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "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": 24,
+  "num_negatives": 100,
+  "output_hidden_size": 1024,
+  "pad_token_id": 58,
+  "proj_codevector_dim": 768,
+  "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": 61,
+  "xvector_output_dim": 512
+}

eval_results.json

@@ -0,0 +1,10 @@
+{
+    "epoch": 99.99,
+    "eval_cer": 0.07081761068660654,
+    "eval_loss": 0.3015391230583191,
+    "eval_runtime": 185.2936,
+    "eval_samples": 4332,
+    "eval_samples_per_second": 23.379,
+    "eval_steps_per_second": 2.925,
+    "eval_wer": 0.337661929317067
+}

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:f117bc1aac94b5e497f6b0e89aff83d344d6f7c53c7917cd3d521eef64eb28de
+size 1262173745

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}, {"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": "ukrainian/cv-uk-ft/last_best", "tokenizer_class": "Wav2Vec2CTCTokenizer", "processor_class": "Wav2Vec2ProcessorWithLM"}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 99.99,
+    "train_loss": 1.3816477500067816,
+    "train_runtime": 56302.7905,
+    "train_samples": 10193,
+    "train_samples_per_second": 18.104,
+    "train_steps_per_second": 0.112
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59b1aec03e4565877ee8f27aac574277990c22479c57beee3a3de80f12e522cf
+size 3055

vocab.json

@@ -0,0 +1 @@
+{"`": 1, "a": 2, "c": 3, "e": 4, "i": 5, "j": 6, "k": 7, "l": 8, "m": 9, "n": 10, "o": 11, "p": 12, "u": 13, "x": 14, "y": 15, "«": 16, "»": 17, "́": 18, "а": 19, "б": 20, "в": 21, "г": 22, "д": 23, "е": 24, "ж": 25, "з": 26, "и": 27, "й": 28, "к": 29, "л": 30, "м": 31, "н": 32, "о": 33, "п": 34, "р": 35, "с": 36, "т": 37, "у": 38, "ф": 39, "х": 40, "ц": 41, "ч": 42, "ш": 43, "щ": 44, "ы": 45, "ь": 46, "ю": 47, "я": 48, "є": 49, "і": 50, "ї": 51, "ґ": 52, "՚": 53, "–": 54, "’": 55, "…": 56, "|": 0, "[UNK]": 57, "[PAD]": 58}