Name: wav2vec2 large xlsr 53 polish
Author: jonatasgrosman

Overview

The wav2vec2-large-xlsr-53-polish model is a specialized Automatic Speech Recognition (ASR) tool fine-tuned for the Polish language. Built upon Meta's cross-lingual wav2vec 2.0 architecture, it leverages large-scale self-supervised pre-training on 53 languages to achieve robust phonetic representations, which are then optimized for Polish transcription. For developers, this means a high-performance solution for converting Polish audio to text without needing to build a transformer pipeline from scratch. It integrates seamlessly with the Hugging Face Transformers library, making it straightforward to deploy in Python-based environments. Compared to general-purpose multilingual models, this version offers superior accuracy for Polish-specific nuances, making it ideal for localized voice assistants, automated transcription services, and accessibility tools.

Highlights

High-accuracy Polish speech-to-text transcription
Based on Meta's cross-lingual wav2vec 2.0 architecture
Easy integration via Hugging Face Transformers library
Optimized for Polish phonetic patterns and nuances
Permissive Apache-2.0 license for commercial deployment

Usage

Install

# Install Hugging Face transformers
pip install transformers torch

SDK Usage

# Load model with transformers
from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained("jonatasgrosman/wav2vec2-large-xlsr-53-polish")
tokenizer = AutoTokenizer.from_pretrained("jonatasgrosman/wav2vec2-large-xlsr-53-polish")

Hugging Face Download

We recommend downloading the model via the Hugging Face CLI or Hub SDK.

Guidance：Before downloading, install huggingface_hub with:

Guidance

pip install -U huggingface_hub

CLI Download

Download the full repository

huggingface-cli download jonatasgrosman/wav2vec2-large-xlsr-53-polish

Download a single file to a local folder (e.g. config.json into ./dir)

huggingface-cli download jonatasgrosman/wav2vec2-large-xlsr-53-polish config.json --local-dir ./dir

See the official docs for more CLI options

SDK Download

# 模型下载
from huggingface_hub import snapshot_download
model_dir = snapshot_download('jonatasgrosman/wav2vec2-large-xlsr-53-polish')

Git Download

Make sure git-lfs is installed first

Git Download

git lfs install
git clone https://huggingface.co/jonatasgrosman/wav2vec2-large-xlsr-53-polish

To skip LFS large-file downloads, use:

Skip LFS

GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/jonatasgrosman/wav2vec2-large-xlsr-53-polish

Model files are hosted on the Hugging Face Hub — download directly via HF CLI / SDK / Git, not through this site.

PyTorch / Transformers Usage

Install Transformers

pip install -U transformers torch

Load the model and run inference

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained('jonatasgrosman/wav2vec2-large-xlsr-53-polish')
tokenizer = AutoTokenizer.from_pretrained('jonatasgrosman/wav2vec2-large-xlsr-53-polish')

Model Download

We recommend downloading the model via the ModelScope CLI or SDK.

Guidance：Before downloading, install ModelScope with:

Guidance

pip install modelscope

CLI Download

Download the full repository

modelscope download --model jonatasgrosman/wav2vec2-large-xlsr-53-polish

Download a single file to a local folder (e.g. README.md into ./dir)

modelscope download --model jonatasgrosman/wav2vec2-large-xlsr-53-polish README.md --local_dir ./dir

See the docs for more CLI options

SDK Download

# 模型下载
from modelscope import snapshot_download
model_dir = snapshot_download('jonatasgrosman/wav2vec2-large-xlsr-53-polish')

Git Download

Make sure git-lfs is installed first

Git Download

git lfs install
git clone https://www.modelscope.cn/jonatasgrosman/wav2vec2-large-xlsr-53-polish.git

To skip LFS large-file downloads, use:

Skip LFS

GIT_LFS_SKIP_SMUDGE=1 git clone https://www.modelscope.cn/jonatasgrosman/wav2vec2-large-xlsr-53-polish.git

ModelScope 模型页直接下载模型文件；无需将模型文件放在本站服务器。

Notebook Quickstart

Install the ModelScope library

pip install "modelscope[audio,cv,nlp,multi-modal,science]" -f https://modelscope.oss-cn-beijing.aliyuncs.com/releases/repo.html

Load the model and run inference

from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks

p = pipeline('text-generation', 'jonatasgrosman/wav2vec2-large-xlsr-53-polish')

Full Documentation

来源: HuggingFace

---
language: pl
license: apache-2.0
datasets:

common_voice

mozilla-foundation/common_voice_6_0

metrics:

wer

cer

tags:

audio

automatic-speech-recognition

hf-asr-leaderboard

mozilla-foundation/common_voice_6_0

pl

robust-speech-event

speech

xlsr-fine-tuning-week

model-index:

name: XLSR Wav2Vec2 Polish by Jonatas Grosman

results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice pl
type: common_voice
args: pl
metrics:
- name: Test WER
type: wer
value: 14.21
- name: Test CER
type: cer
value: 3.49
- name: Test WER (+LM)
type: wer
value: 10.98
- name: Test CER (+LM)
type: cer
value: 2.93
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: Robust Speech Event - Dev Data
type: speech-recognition-community-v2/dev_data
args: pl
metrics:
- name: Dev WER
type: wer
value: 33.18
- name: Dev CER
type: cer
value: 15.92
- name: Dev WER (+LM)
type: wer
value: 29.31
- name: Dev CER (+LM)
type: cer
value: 15.17
---

Fine-tuned XLSR-53 large model for speech recognition in Polish

Fine-tuned facebook/wav2vec2-large-xlsr-53 on Polish using the train and validation splits of Common Voice 6.1.
When using this model, make sure that your speech input is sampled at 16kHz.

This model has been fine-tuned thanks to the GPU credits generously given by the OVHcloud :)

The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint

Usage

The model can be used directly (without a language model) as follows...

Using the HuggingSound library:

python

from huggingsound import SpeechRecognitionModel
model = SpeechRecognitionModel("jonatasgrosman/wav2vec2-large-xlsr-53-polish")
audio_paths = ["/path/to/file.mp3", "/path/to/another_file.wav"]
transcriptions = model.transcribe(audio_paths)

Writing your own inference script:

python

import torch
import librosa
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
LANG_ID = "pl"
MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-polish"
SAMPLES = 5
test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
Preprocessing the datasets.
We need to read the audio files as arrays
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)
for i, predicted_sentence in enumerate(predicted_sentences):
    print("-" * 100)
    print("Reference:", test_dataset[i]["sentence"])
    print("Prediction:", predicted_sentence)

| Reference | Prediction |
| ------------- | ------------- |
| """CZY DRZWI BYŁY ZAMKNIĘTE?""" | PRZY DRZWI BYŁY ZAMKNIĘTE |
| GDZIEŻ TU POWÓD DO WYRZUTÓW? | WGDZIEŻ TO POM DO WYRYDÓ |
| """O TEM JEDNAK NIE BYŁO MOWY.""" | O TEM JEDNAK NIE BYŁO MOWY |
| LUBIĘ GO. | LUBIĄ GO |
| — TO MI NIE POMAGA. | TO MNIE NIE POMAGA |
| WCIĄŻ LUDZIE WYSIADAJĄ PRZED ZAMKIEM, Z MIASTA, Z PRAGI. | WCIĄŻ LUDZIE WYSIADAJĄ PRZED ZAMKIEM Z MIASTA Z PRAGI |
| ALE ON WCALE INACZEJ NIE MYŚLAŁ. | ONY MONITCENIE PONACZUŁA NA MASU |
| A WY, CO TAK STOICIE? | A WY CO TAK STOICIE |
| A TEN PRZYRZĄD DO CZEGO SŁUŻY? | A TEN PRZYRZĄD DO CZEGO SŁUŻY |
| NA JUTRZEJSZYM KOLOKWIUM BĘDZIE PIĘĆ PYTAŃ OTWARTYCH I TEST WIELOKROTNEGO WYBORU. | NAJUTRZEJSZYM KOLOKWIUM BĘDZIE PIĘĆ PYTAŃ OTWARTYCH I TEST WIELOKROTNEGO WYBORU |

Evaluation

1. To evaluate on mozilla-foundation/common_voice_6_0 with split test

bash

python eval.py --model_id jonatasgrosman/wav2vec2-large-xlsr-53-polish --dataset mozilla-foundation/common_voice_6_0 --config pl --split test

2. To evaluate on speech-recognition-community-v2/dev_data

bash

python eval.py --model_id jonatasgrosman/wav2vec2-large-xlsr-53-polish --dataset speech-recognition-community-v2/dev_data --config pl --split validation --chunk_length_s 5.0 --stride_length_s 1.0

Citation

If you want to cite this model you can use this:

bibtex

@misc{grosman2021xlsr53-large-polish,
  title={Fine-tuned {XLSR}-53 large model for speech recognition in {P}olish},
  author={Grosman, Jonatas},
  howpublished={\url{https://huggingface.co/jonatasgrosman/wav2vec2-large-xlsr-53-polish}},
  year={2021}
}