**wav2vec2-model2-torgo Open-source Speech Recognition Model

Wav2vec2 Model2 Torgo

Developed by modhp

A speech recognition model based on the wav2vec2 architecture, trained on the TORGO dataset, but with limited details on training specifics and performance.

Speech Recognition

Transformers

#Speech Recognition #High Word Error Rate #Small Batch Training

Downloads 16

Release Time : 4/8/2022

Model Overview

This model is a speech recognition model based on the wav2vec2 architecture, likely used for converting speech to text. Due to incomplete model card information, its specific functionalities and applications require further verification.

Model Features

wav2vec2 Architecture

Utilizes the advanced wav2vec2 architecture, suitable for speech recognition tasks.

End-to-End Training

The model is trained from scratch, potentially offering better domain adaptability.

Model Capabilities

Speech Recognition

Audio-to-Text

Use Cases

Speech Transcription

Speech-to-Text

Convert spoken content into editable text format

🚀 wav2vec2-model2-torgo

This is a model trained from scratch on the None dataset. It achieves a Loss of 4.9975 and a Wer of 1.0 on the evaluation set.

🚀 Quick Start

This section provides a brief overview of the model and its performance.

📚 Documentation

Model description

This model, wav2vec2-model2-torgo, was trained from scratch on the None dataset.

Intended uses & limitations

More information about the intended uses and limitations of this model is yet to be provided.

Training and evaluation data

Details about the training and evaluation data are currently unavailable.

Training procedure

Training hyperparameters

The following hyperparameters were used during the training process:

learning_rate: 0.1
train_batch_size: 1
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 4
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 1000
num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
12.5453	0.76	500	14.6490	1.0
4.8036	1.53	1000	8.4523	1.0
5.0421	2.29	1500	5.4114	1.0
5.2055	3.05	2000	11.0507	1.0
4.6389	3.82	2500	4.6792	1.0
4.5523	4.58	3000	4.7855	1.0
4.7843	5.34	3500	11.2783	1.0
4.6066	6.11	4000	8.7807	1.0
4.7382	6.87	4500	2942.0220	1.0
130.5733	7.63	5000	5.8412	1.0
4.4972	8.4	5500	17.7038	1.0
4.5196	9.16	6000	11.4548	1.0
4.3198	9.92	6500	6.0885	1.0
4.4273	10.69	7000	6.7374	1.0
4.2783	11.45	7500	4.7276	1.0
4.2985	12.21	8000	6.1412	1.0
4.3262	12.98	8500	5.2621	1.0
4.1705	13.74	9000	5.2214	1.0
4.3176	14.5	9500	5.5359	1.0
3.9808	15.27	10000	4.1537	1.0
4.0228	16.03	10500	4.2962	1.0
4.0595	16.79	11000	7.6361	1.0
4.0088	17.56	11500	6.8715	1.0
3.8727	18.32	12000	8.8657	1.0
4.0073	19.08	12500	5.8170	1.0
3.8511	19.85	13000	13.9836	1.0
4.0899	20.61	13500	5.3287	1.0
3.8782	21.37	14000	8.0635	1.0
3.9235	22.14	14500	5.5129	1.0
3.7276	22.9	15000	5.0819	1.0
3.7908	23.66	15500	6.1458	1.0
3.9176	24.43	16000	4.6094	1.0
3.8477	25.19	16500	5.1406	1.0
3.6917	25.95	17000	4.5684	1.0
3.8568	26.72	17500	4.0306	1.0
3.7231	27.48	18000	5.6331	1.0
3.8145	28.24	18500	8.2997	1.0
3.7809	29.01	19000	5.7468	1.0
3.5995	29.77	19500	4.9975	1.0

Framework versions

Transformers 4.17.0
Pytorch 1.11.0
Datasets 1.18.3
Tokenizers 0.11.6

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