Iloko_model Open-source Automatic Speech Recognition Model - Free Deployment for Ilocano Speech Recognition

Iloko Model

Developed by denden

This model is a fine-tuned automatic speech recognition (ASR) model based on facebook/wav2vec2-large-xlsr-53 on an unknown dataset, supporting Iloko language speech recognition.

Speech Recognition

Transformers

Open Source License:Apache-2.0 #Low Word Error Rate Speech Recognition #Multilingual Fine-tuned Model #High Accuracy Speech-to-Text

Downloads 28

Release Time : 3/2/2022

Model Overview

This is a fine-tuned model for automatic speech recognition, based on Facebook's wav2vec2-large-xlsr-53 architecture, specifically optimized for the Iloko language.

Model Features

Low Word Error Rate

Achieved an excellent word error rate (WER) of 0.084 on the evaluation set.

Efficient Training

Implemented efficient training with a relatively small batch size (8) and gradient accumulation (2).

Learning Rate Optimization

Adopted linear learning rate scheduling and warm-up strategies to optimize the training process.

Model Capabilities

Iloko speech recognition

Speech-to-text

Use Cases

Speech Transcription

Iloko Speech Transcription

Convert Iloko speech into text

Word error rate 0.084

Voice-Assisted Applications

Iloko Voice Assistant

Build voice assistant applications supporting the Iloko language

🚀 iloko_model

This iloko_model is a fine - tuned automatic speech recognition model. It is based on facebook/wav2vec2-large-xlsr-53 and trained on an unknown dataset. It achieves a Loss of 0.0095 and a Wer of 0.0840 on the evaluation set.

🚀 Quick Start

This section provides a high - level overview of the model. For more detailed information, please refer to the following sections.

📚 Documentation

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: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e - 08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.2784	1.11	100	2.9875	1.0
2.6899	2.22	200	2.6741	1.0
2.6177	3.33	300	2.6516	1.0
2.5327	4.44	400	2.4530	1.0
0.8653	5.56	500	0.5227	0.6547
0.3414	6.67	600	0.1830	0.2487
0.2299	7.78	700	0.1212	0.1877
0.1739	8.89	800	0.0843	0.1441
0.1242	10.0	900	0.0766	0.1441
0.1116	11.11	1000	0.0530	0.1145
0.0861	12.22	1100	0.0442	0.1047
0.1007	13.33	1200	0.0379	0.1023
0.0613	14.44	1300	0.0291	0.1006
0.0629	15.56	1400	0.0264	0.0961
0.047	16.67	1500	0.0238	0.0935
0.0797	17.78	1600	0.0226	0.0913
0.034	18.89	1700	0.0197	0.0893
0.0485	20.0	1800	0.0173	0.0905
0.0402	21.11	1900	0.0148	0.0902
0.0231	22.22	2000	0.0135	0.0891
0.0512	23.33	2100	0.0134	0.0861
0.0181	24.44	2200	0.0118	0.0842
0.0371	25.56	2300	0.0116	0.0867
0.0342	26.67	2400	0.0104	0.0863
0.0344	27.78	2500	0.0100	0.0850
0.0182	28.89	2600	0.0096	0.0839
0.0171	30.0	2700	0.0095	0.0840

Framework versions

Transformers 4.11.3
Pytorch 1.10.0+cu102
Datasets 1.13.3
Tokenizers 0.10.3

📄 License

This model is released 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