đ XLS-R Wav2Vec2 Polish by Jonatas Grosman
This is a fine - tuned XLS - R 1B model for Polish speech recognition, which can effectively convert Polish speech into text.
đ Quick Start
This model is a fine - tuned version of facebook/wav2vec2-xls-r-1b on Polish. It uses the train and validation splits of Common Voice 8.0, Multilingual LibriSpeech, and Voxpopuli. When using this model, ensure that your speech input is sampled at 16kHz.
This model has been fine - tuned by the HuggingSound tool, and thanks to the GPU credits generously given by the OVHcloud.
⨠Features
- Automatic Speech Recognition: Capable of accurately transcribing Polish speech.
- Fine - tuned on Multiple Datasets: Utilizes data from Common Voice 8.0, Multilingual LibriSpeech, and Voxpopuli for better performance.
đģ Usage Examples
Basic Usage
Using the HuggingSound library:
from huggingsound import SpeechRecognitionModel
model = SpeechRecognitionModel("jonatasgrosman/wav2vec2-xls-r-1b-polish")
audio_paths = ["/path/to/file.mp3", "/path/to/another_file.wav"]
transcriptions = model.transcribe(audio_paths)
Advanced Usage
Writing your own inference script:
import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "pl"
MODEL_ID = "jonatasgrosman/wav2vec2-xls-r-1b-polish"
SAMPLES = 10
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
batch["speech"] = speech_array
batch["sentence"] = batch["sentence"].upper()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], 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)
predicted_sentences = processor.batch_decode(predicted_ids)
đ§ Technical Details
Evaluation Results
Task |
Dataset |
Metric |
Value |
Automatic Speech Recognition |
Common Voice 8 |
Test WER |
11.01 |
Automatic Speech Recognition |
Common Voice 8 |
Test CER |
2.55 |
Automatic Speech Recognition |
Common Voice 8 |
Test WER (+LM) |
7.32 |
Automatic Speech Recognition |
Common Voice 8 |
Test CER (+LM) |
1.95 |
Automatic Speech Recognition |
Robust Speech Event - Dev Data |
Dev WER |
26.31 |
Automatic Speech Recognition |
Robust Speech Event - Dev Data |
Dev CER |
13.85 |
Automatic Speech Recognition |
Robust Speech Event - Dev Data |
Dev WER (+LM) |
20.33 |
Automatic Speech Recognition |
Robust Speech Event - Dev Data |
Dev CER (+LM) |
13.0 |
Automatic Speech Recognition |
Robust Speech Event - Test Data |
Test WER |
22.77 |
đĻ Installation
No specific installation steps are provided in the original document.
đ Documentation
Evaluation Commands
- To evaluate on
mozilla-foundation/common_voice_8_0
with split test
python eval.py --model_id jonatasgrosman/wav2vec2-xls-r-1b-polish --dataset mozilla-foundation/common_voice_8_0 --config pl --split test
- To evaluate on
speech-recognition-community-v2/dev_data
python eval.py --model_id jonatasgrosman/wav2vec2-xls-r-1b-polish --dataset speech-recognition-community-v2/dev_data --config pl --split validation --chunk_length_s 5.0 --stride_length_s 1.0
đ License
This model is licensed under the Apache - 2.0 license.
đ Citation
If you want to cite this model you can use this:
@misc{grosman2021xlsr-1b-polish,
title={Fine-tuned {XLS-R} 1{B} model for speech recognition in {P}olish},
author={Grosman, Jonatas},
howpublished={\url{https://huggingface.co/jonatasgrosman/wav2vec2-xls-r-1b-polish}},
year={2022}
}