w2v-bert-2.0-north-sami

jimregan/w2v-bert-2.0-north-sami

w2v-bert-2.0-north-sami is a automatic speech recognition model from jimregan. Use it when you need speech turned into text. It is set up for transformers. The card lists the license as mit.

should probably proofread and complete it, then remove this comment. --

Downloads · 30 days

17% of all-time downloads

All-time downloads

115

Public

Parameters

606M

29.1 GB on disk

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors2.4 GB · 100%

At a glance

Task: Automatic Speech Recognition
Library: transformers
License: mit
Model type: wav2vec2-bert
Access: Public
Created: Mar 6, 2024
Updated: Mar 13, 2024
SHA: b5425033

Base models

facebook/w2v-bert-2.0

Model card

From the Hugging Face model README

w2v-bert-2.0-north-sami

This model is a fine-tuned version of facebook/w2v-bert-2.0 on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.9778
Cer: 0.2753
Wer: 0.8105

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: 5e-05
train_batch_size: 4
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 8
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 1000
num_epochs: 6
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
3.9302	0.28	200	2.6591	0.6390	1.0156
1.9387	0.55	400	0.9924	0.2427	0.8257
1.5598	0.83	600	0.8492	0.2243	0.7684
1.4375	1.11	800	0.7714	0.2179	0.7595
1.389	1.38	1000	0.7074	0.1813	0.6970
1.3134	1.66	1200	0.6397	0.1668	0.6354
1.2836	1.94	1400	0.6282	0.1648	0.6241
1.1205	2.21	1600	0.6423	0.1708	0.6629
1.067	2.49	1800	0.5731	0.1636	0.6376
1.0965	2.76	2000	0.5496	0.1498	0.5932
0.975	3.04	2200	0.5305	0.1381	0.5460
0.8759	3.32	2400	0.5325	0.1404	0.5612
0.9537	3.59	2600	0.5397	0.1478	0.5570
0.887	3.87	2800	0.5264	0.1351	0.5354
0.8123	4.15	3000	0.5117	0.1413	0.5481
0.7681	4.42	3200	0.5105	0.1324	0.5283
0.811	4.7	3400	0.5330	0.1488	0.5612
0.8249	4.98	3600	0.6151	0.1592	0.5873
0.8376	5.25	3800	0.5518	0.1492	0.5755
0.9814	5.53	4000	0.8945	0.2293	0.7979
1.2369	5.81	4200	0.9778	0.2753	0.8105

Framework versions

Transformers 4.38.2
Pytorch 2.2.1+cu121
Datasets 2.18.0
Tokenizers 0.15.2

Task: Automatic Speech Recognition
Library: transformers
Type: wav2vec2-bert
License: mit
Created: Mar 6, 2024
Updated: Mar 13, 2024