Wav2vec2 Large Xlsr 53 Finnish

🚀 Wav2Vec2-Large-XLSR-53-Finnish

This model is fine - tuned from facebook/wav2vec2-large-xlsr-53 on Finnish datasets, enabling accurate automatic speech recognition.

⚠️ Important Note

This is an old model and should not be used anymore!! There are a lot better newer models available at our organization hub: Finnish - NLP/wav2vec2-xlsr-1b-finnish-lm-v2 and Finnish - NLP/wav2vec2-xlsr-300m-finnish-lm

🚀 Quick Start

This model is fine - tuned facebook/wav2vec2-large-xlsr-53 on Finnish using the Common Voice, CSS10 Finnish and Finnish parliament session 2 datasets. When using this model, make sure that your speech input is sampled at 16kHz.

✨ Features

• Audio Processing: Capable of handling audio data for speech recognition tasks.
• Fine - Tuned: Specifically fine - tuned on Finnish datasets for better performance in Finnish speech recognition.

📦 Installation

No specific installation steps are provided in the original document.

💻 Usage Examples

Basic Usage

import librosa
import torch
import torchaudio
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor

test_dataset = load_dataset("common_voice", "fi", split="test[:2%]")

processor = Wav2Vec2Processor.from_pretrained("aapot/wav2vec2-large-xlsr-53-finnish")
model = Wav2Vec2ForCTC.from_pretrained("aapot/wav2vec2-large-xlsr-53-finnish")

resampler = lambda sr, y: librosa.resample(y.numpy().squeeze(), sr, 16_000)

# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
    speech_array, sampling_rate = torchaudio.load(batch["path"])
    batch["speech"] = resampler(sampling_rate, speech_array).squeeze()
    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])

Advanced Usage

import librosa
import torch
import torchaudio
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import re

test_dataset = load_dataset("common_voice", "fi", split="test")
wer = load_metric("wer")

processor = Wav2Vec2Processor.from_pretrained("aapot/wav2vec2-large-xlsr-53-finnish")
model = Wav2Vec2ForCTC.from_pretrained("aapot/wav2vec2-large-xlsr-53-finnish")
model.to("cuda")

chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"\“\%\‘\”\�\'\...\…\–\é]'
resampler = lambda sr, y: librosa.resample(y.numpy().squeeze(), sr, 16_000)

# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
  batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
  speech_array, sampling_rate = torchaudio.load(batch["path"])
  batch["speech"] = resampler(sampling_rate, speech_array).squeeze()
  return batch

test_dataset = test_dataset.map(speech_file_to_array_fn)

# Preprocessing the datasets.
# We need to read the audio 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: 32.378771 %

📚 Documentation

Training

The Common Voice train, validation and other datasets were used for training as well as CSS10 Finnish and Finnish parliament session 2 datasets.

The script used for training can be found from Google Colab

📄 License

This model is licensed under the apache - 2.0 license.

📋 Information Table