Initial commit (commonvoice only)

README.md

@@ -0,0 +1,136 @@
+---
+language: ru
+datasets:
+- common_voice
+metrics:
+- wer
+tags:
+- audio
+- automatic-speech-recognition
+- speech
+- xlsr-fine-tuning-week
+license: apache-2.0
+model-index:
+- name: [Anton Lozhkov] Russian XLSR Wav2Vec2 Large 53
+  results:
+  - task: 
+      name: Speech Recognition
+      type: automatic-speech-recognition
+    dataset:
+      name: Common Voice ru
+      type: common_voice
+      args: ru
+    metrics:
+       - name: Test WER
+         type: wer
+         value: 22.36
+---
+
+# Wav2Vec2-Large-XLSR-53-Russian
+
+Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Russian using the [Common Voice](https://huggingface.co/datasets/common_voice) dataset.
+When using this model, make sure that your speech input is sampled at 16kHz.
+
+## Usage
+
+The model can be used directly (without a language model) as follows:
+
+```python
+import torch
+import torchaudio
+from datasets import load_dataset
+from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
+
+test_dataset = load_dataset("common_voice", "ru", split="test[:2%]")
+
+processor = Wav2Vec2Processor.from_pretrained("anton-l/wav2vec2-large-xlsr-53-russian")
+model = Wav2Vec2ForCTC.from_pretrained("anton-l/wav2vec2-large-xlsr-53-russian")
+
+resampler = torchaudio.transforms.Resample(48_000, 16_000)
+
+# Preprocessing the datasets.
+# We need to read the aduio files as arrays
+def speech_file_to_array_fn(batch):
+	speech_array, sampling_rate = torchaudio.load(batch["path"])
+	batch["speech"] = resampler(speech_array).squeeze().numpy()
+	return batch
+
+test_dataset = test_dataset.map(speech_file_to_array_fn)
+inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
+
+with torch.no_grad():
+	logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
+
+predicted_ids = torch.argmax(logits, dim=-1)
+
+print("Prediction:", processor.batch_decode(predicted_ids))
+print("Reference:", test_dataset["sentence"][:2])
+```
+
+
+## Evaluation
+
+The model can be evaluated as follows on the {language} test data of Common Voice.  # TODO: replace #TODO: replace language with your {language}, *e.g.* French
+
+
+```python
+import torch
+import torchaudio
+from datasets import load_dataset, load_metric
+from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
+import re
+
+test_dataset = load_dataset("common_voice", "ru", split="test")
+wer = load_metric("wer")
+
+processor = Wav2Vec2Processor.from_pretrained("anton-l/wav2vec2-large-xlsr-53-russian")
+model = Wav2Vec2ForCTC.from_pretrained("anton-l/wav2vec2-large-xlsr-53-russian")
+model.to("cuda")
+
+resampler = torchaudio.transforms.Resample(48_000, 16_000)
+
+def clean_sentence(sent):
+	sent = sent.lower()
+	# replace non-alphanumeric characters with space ("какой-то, вот" -> "какой то  вот")
+	sent = "".join(ch if ch.isalnum() else " " for ch in sent)
+	# remove repeated spaces
+	sent = " ".join(sent.split())
+	# these letters are considered equivalent in written Russian
+	sent = sent.replace('ё', 'е')
+	return sent
+
+# Preprocessing the datasets.
+# We need to read the aduio files as arrays
+def speech_file_to_array_fn(batch):
+	batch["sentence"] = clean_sentence(batch["sentence"])
+	speech_array, sampling_rate = torchaudio.load(batch["path"])
+	batch["speech"] = resampler(speech_array).squeeze().numpy()
+	return batch
+
+test_dataset = test_dataset.map(speech_file_to_array_fn)
+
+# Preprocessing the datasets.
+# We need to read the aduio files as arrays
+def evaluate(batch):
+	inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
+
+	with torch.no_grad():
+		logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
+
+	pred_ids = torch.argmax(logits, dim=-1)
+	batch["pred_strings"] = processor.batch_decode(pred_ids)
+	return batch
+
+result = test_dataset.map(evaluate, batched=True, batch_size=8)
+
+print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
+```
+
+**Test Result**: 22.36 %  
+
+
+## Training
+
+The Common Voice `train` and `validation` datasets were used for training.
+
+The script used for training can be found [here](github.com)

config.json

@@ -0,0 +1,76 @@
+{
+  "_name_or_path": "anton-l/wav2vec2-large-xlsr-53-russian",
+  "activation_dropout": 0.0,
+  "apply_spec_augment": true,
+  "architectures": [
+    "Wav2Vec2ForCTC"
+  ],
+  "attention_dropout": 0.1,
+  "bos_token_id": 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,
+  "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,
+  "final_dropout": 0.0,
+  "gradient_checkpointing": true,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.1,
+  "mask_channel_length": 10,
+  "mask_channel_min_space": 1,
+  "mask_channel_other": 0.0,
+  "mask_channel_prob": 0.0,
+  "mask_channel_selection": "static",
+  "mask_feature_length": 10,
+  "mask_feature_prob": 0.0,
+  "mask_time_length": 10,
+  "mask_time_min_space": 1,
+  "mask_time_other": 0.0,
+  "mask_time_prob": 0.05,
+  "mask_time_selection": "static",
+  "model_type": "wav2vec2",
+  "num_attention_heads": 16,
+  "num_conv_pos_embedding_groups": 16,
+  "num_conv_pos_embeddings": 128,
+  "num_feat_extract_layers": 7,
+  "num_hidden_layers": 24,
+  "pad_token_id": 53,
+  "transformers_version": "4.4.0",
+  "vocab_size": 54
+}

preprocessor_config.json

@@ -0,0 +1,8 @@
+{
+  "do_normalize": true,
+  "feature_size": 1,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "return_attention_mask": true,
+  "sampling_rate": 16000
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9e062de78eecef4bcd92288f9b31e91609d98274e473b2da9390dfc182fc686
+size 1262149336

special_tokens_map.json

@@ -0,0 +1 @@
+{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "[UNK]", "pad_token": "[PAD]"}

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": "|"}

vocab.json

@@ -0,0 +1 @@
+{"a": 1, "b": 2, "c": 3, "e": 4, "f": 5, "g": 6, "h": 7, "i": 8, "k": 9, "l": 10, "m": 11, "n": 12, "o": 13, "p": 14, "r": 15, "s": 16, "t": 17, "x": 18, "z": 19, "а": 20, "б": 21, "в": 22, "г": 23, "д": 24, "е": 25, "ж": 26, "з": 27, "и": 28, "й": 29, "к": 30, "л": 31, "м": 32, "н": 33, "о": 34, "п": 35, "р": 36, "с": 37, "т": 38, "у": 39, "ф": 40, "х": 41, "ц": 42, "ч": 43, "ш": 44, "щ": 45, "ъ": 46, "ы": 47, "ь": 48, "э": 49, "ю": 50, "я": 51, "|": 0, "[UNK]": 52, "[PAD]": 53}