wav2vec2-large-xls-r-300m-pt-colab Open-source Model - Free and Precise Portuguese Speech-to-Text Conversion

Wav2vec2 Large Xls R 300m Pt Colab

Developed by robertodtg

This model is a speech recognition model fine-tuned on the common_voice_9_0 dataset based on facebook/wav2vec2-xls-r-300m, supporting Portuguese speech-to-text tasks.

Speech Recognition

Transformers

Open Source License:Apache-2.0 #Portuguese Speech Recognition #Large Model Fine-tuning #Low Word Error Rate

Downloads 107

Release Time : 6/24/2022

Model Overview

This is a Portuguese-optimized speech recognition model, fine-tuned based on the wav2vec2-xls-r-300m architecture and trained on the Common Voice 9.0 dataset, suitable for Portuguese speech-to-text tasks.

Model Features

Portuguese Optimization

Specially fine-tuned and optimized for Portuguese speech recognition tasks

Based on Large-scale Pre-trained Model

Based on the facebook/wav2vec2-xls-r-300m pre-trained model with powerful speech feature extraction capabilities

Relatively Lightweight

300M parameter scale, relatively lightweight while maintaining performance

Model Capabilities

Portuguese Speech Recognition

Speech-to-Text

Automatic Speech Recognition

Use Cases

Speech Transcription

Portuguese Speech Transcription

Convert Portuguese speech content into text

Achieved a 17.36% word error rate on the evaluation set

Voice Assistants

Portuguese Voice Command Recognition

Used to build Portuguese voice assistants or voice control applications

🚀 wav2vec2-large-xls-r-300m-pt-colab

This model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_9_0 dataset. It can achieve certain performance on the evaluation set, which is useful for speech - related tasks.

🚀 Quick Start

This section is not provided in the original document, so it is skipped.

✨ Features

This model is a fine - tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_9_0 dataset. It achieves the following results on the evaluation set:

Loss: 0.2975
Wer: 0.1736

📚 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: 16
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e - 08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
num_epochs: 30
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
5.179	0.49	400	1.4554	0.9349
0.7545	0.98	800	0.5594	0.5174
0.4485	1.47	1200	0.3964	0.3749
0.4118	1.96	1600	0.3547	0.3172
0.3282	2.45	2000	0.3372	0.3061
0.3199	2.94	2400	0.3466	0.2910
0.2847	3.44	2800	0.3651	0.3310
0.2713	3.93	3200	0.3509	0.3016
0.2414	4.42	3600	0.3451	0.2908
0.2473	4.91	4000	0.3253	0.2747
0.2168	5.4	4400	0.3243	0.2680
0.219	5.89	4800	0.3067	0.2540
0.196	6.38	5200	0.3268	0.2824
0.1934	6.87	5600	0.3252	0.2736
0.1808	7.36	6000	0.3422	0.2737
0.177	7.85	6400	0.3292	0.2707
0.1626	8.34	6800	0.3089	0.2524
0.1605	8.83	7200	0.3062	0.2471
0.1505	9.32	7600	0.3229	0.2474
0.1491	9.82	8000	0.3098	0.2491
0.1433	10.31	8400	0.3449	0.2681
0.1431	10.8	8800	0.3439	0.2532
0.1349	11.29	9200	0.3112	0.2413
0.1236	11.78	9600	0.3248	0.2378
0.1253	12.27	10000	0.3393	0.2394
0.1195	12.76	10400	0.3050	0.2336
0.1194	13.25	10800	0.3494	0.2550
0.1125	13.74	11200	0.3332	0.2395
0.1063	14.23	11600	0.3134	0.2365
0.1044	14.72	12000	0.3101	0.2303
0.0999	15.21	12400	0.3162	0.2248
0.0986	15.71	12800	0.3183	0.2260
0.0958	16.2	13200	0.3300	0.2279
0.0907	16.69	13600	0.3136	0.2260
0.0875	17.18	14000	0.3492	0.2203
0.0823	17.67	14400	0.3214	0.2259
0.0839	18.16	14800	0.3194	0.2145
0.0783	18.65	15200	0.3122	0.2180
0.0789	19.14	15600	0.3158	0.2127
0.0732	19.63	16000	0.3076	0.2109
0.0715	20.12	16400	0.3216	0.2150
0.0649	20.61	16800	0.2958	0.2051
0.0647	21.1	17200	0.3022	0.2014
0.0649	21.59	17600	0.3045	0.2033
0.0621	22.09	18000	0.3194	0.2035
0.0561	22.58	18400	0.3197	0.2022
0.0582	23.07	18800	0.3109	0.1978
0.0533	23.56	19200	0.3121	0.1932
0.0515	24.05	19600	0.3125	0.1939
0.0484	24.54	20000	0.3081	0.1908
0.0485	25.03	20400	0.3042	0.1896
0.0444	25.52	20800	0.3038	0.1886
0.0426	26.01	21200	0.2985	0.1868
0.0415	26.5	21600	0.3066	0.1858
0.0398	26.99	22000	0.3117	0.1828
0.0397	27.48	22400	0.2980	0.1795
0.0394	27.97	22800	0.2950	0.1791
0.0364	28.47	23200	0.3025	0.1773
0.0365	28.96	23600	0.3022	0.1747
0.0376	29.45	24000	0.2978	0.1738
0.0344	29.94	24400	0.2975	0.1736

Framework versions

Transformers 4.20.1
Pytorch 1.10.0+cu111
Datasets 2.3.2
Tokenizers 0.12.1

📄 License

This model 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