Wav2Vec2-XLS-R-300M-LG Open-source Speech Recognition Model - Free Support for Automatic Luganda Language Recognition

Wav2vec2 Xls R 300m Lg

Developed by samitizerxu

This model is a fine-tuned speech recognition model based on facebook/wav2vec2-xls-r-300m on the COMMON_VOICE - LG dataset, supporting automatic speech recognition tasks for Luganda (lg).

Speech Recognition

Transformers

OtherOpen Source License:Apache-2.0 #Luganda speech recognition #Low-resource speech processing #XLS-R architecture

Downloads 22

Release Time : 3/2/2022

Model Overview

This is an optimized automatic speech recognition (ASR) model for Luganda, fine-tuned based on the wav2vec2-xls-r-300m architecture, suitable for converting Luganda speech to text.

Model Features

Robust Speech Recognition

Optimized speech recognition capability for Luganda, capable of handling different accents and pronunciation variants

Based on Large-scale Pretrained Model

Fine-tuned from the facebook/wav2vec2-xls-r-300m pretrained model, with powerful speech feature extraction capabilities

Trained on Common Voice Dataset

Trained using the Mozilla Common Voice dataset, covering diverse speech scenarios

Model Capabilities

Luganda speech recognition

Speech-to-text

Robust speech processing

Use Cases

Speech Transcription

Luganda Speech Transcription

Convert Luganda speech content into text

WER of 78.89% and CER of 15.16% on the Common Voice test set

Voice Assistants

Luganda Voice Command Recognition

Used for voice command recognition systems in Luganda

🚀 wav2vec2-xls-r-300m-lg

This model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the COMMON_VOICE - LG dataset. It is designed for automatic speech recognition tasks and can achieve certain performance on the evaluation set, providing valuable support for related speech processing applications.

🚀 Quick Start

This model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the COMMON_VOICE - LG dataset. It achieves the following results on the evaluation set:

Loss: 0.6989
Wer: 0.8529

✨ Features

Tags: automatic - speech - recognition, robust - speech - event, common_voice, lg, generated_from_trainer, hf - asr - leaderboard
Datasets: common_voice

📚 Documentation

Model Index

Property	Details
Name	wav2vec2-xls-r-300m-lg
Results	- Task: - Name: Automatic Speech Recognition - Type: automatic - speech - recognition - Dataset: - Name: Common Voice 7 - Type: mozilla - foundation/common_voice_7_0 - Args: sv - SE - Metrics: - Name: Test WER - Type: wer - Value: 78.89 - Name: Test CER - Type: cer - Value: 15.16

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0003
train_batch_size: 16
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e - 08
lr_scheduler_type: linear
num_epochs: 20.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.9089	6.33	500	2.8983	1.0002
2.5754	12.66	1000	1.8710	1.0
1.4093	18.99	1500	0.7195	0.8547

Framework versions

Transformers 4.17.0.dev0
Pytorch 1.10.2+cu102
Datasets 1.18.2.dev0
Tokenizers 0.11.0

Evaluation Commands

Basic Usage

# To evaluate on `mozilla-foundation/common_voice_7_0` with split `test`
python eval.py --model_id samitizerxu/wav2vec2-xls-r-300m-lg --dataset mozilla-foundation/common_voice_7_0 --config lg --split test

📄 License

This project is licensed under the Apache - 2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご