w2v-bert-grain-lg-v5

sulaimank/w2v-bert-grain-lg-v5

w2v-bert-grain-lg-v5 is a automatic speech recognition model from sulaimank. 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

32% of all-time downloads

All-time downloads

Public

Parameters

606M

104 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: Oct 21, 2024
Updated: Oct 21, 2024
SHA: 0dfb634b

Base models

facebook/w2v-bert-2.0

Datasets

Grain

Model card

From the Hugging Face model README

w

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

Loss: 0.0084
Wer: 0.0055
Cer: 0.0011

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: 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: 80
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.2656	1.0	1296	0.0886	0.0909	0.0165
0.0752	2.0	2592	0.0589	0.0620	0.0117
0.0529	3.0	3888	0.0448	0.0408	0.0081
0.0391	4.0	5184	0.0409	0.0374	0.0073
0.032	5.0	6480	0.0323	0.0299	0.0058
0.0268	6.0	7776	0.0326	0.0348	0.0065
0.0234	7.0	9072	0.0236	0.0243	0.0050
0.0207	8.0	10368	0.0228	0.0289	0.0057
0.0179	9.0	11664	0.0235	0.0240	0.0048
0.0163	10.0	12960	0.0268	0.0280	0.0054
0.0157	11.0	14256	0.0258	0.0352	0.0067
0.0125	12.0	15552	0.0205	0.0221	0.0046
0.0116	13.0	16848	0.0187	0.0161	0.0035
0.0113	14.0	18144	0.0193	0.0215	0.0041
0.0111	15.0	19440	0.0185	0.0209	0.0041
0.01	16.0	20736	0.0188	0.0191	0.0038
0.0098	17.0	22032	0.0132	0.0143	0.0027
0.0082	18.0	23328	0.0155	0.0161	0.0032
0.0077	19.0	24624	0.0180	0.0214	0.0041
0.0073	20.0	25920	0.0170	0.0145	0.0029
0.0075	21.0	27216	0.0134	0.0170	0.0030
0.0067	22.0	28512	0.0120	0.0130	0.0026
0.0061	23.0	29808	0.0125	0.0155	0.0031
0.0054	24.0	31104	0.0141	0.0130	0.0024
0.0051	25.0	32400	0.0134	0.0109	0.0022
0.0052	26.0	33696	0.0103	0.0108	0.0022
0.0046	27.0	34992	0.0092	0.0095	0.0018
0.004	28.0	36288	0.0140	0.0123	0.0023
0.004	29.0	37584	0.0110	0.0133	0.0024
0.0035	30.0	38880	0.0110	0.0103	0.0021
0.0035	31.0	40176	0.0101	0.0064	0.0016
0.0035	32.0	41472	0.0148	0.0124	0.0024
0.003	33.0	42768	0.0090	0.0053	0.0012
0.0031	34.0	44064	0.0096	0.0073	0.0015
0.0032	35.0	45360	0.0071	0.0057	0.0011
0.0025	36.0	46656	0.0097	0.0078	0.0017
0.0023	37.0	47952	0.0116	0.0066	0.0014
0.0024	38.0	49248	0.0087	0.0076	0.0015
0.003	39.0	50544	0.0098	0.0074	0.0015
0.002	40.0	51840	0.0122	0.0108	0.0019
0.0017	41.0	53136	0.0089	0.0054	0.0012
0.0018	42.0	54432	0.0094	0.0064	0.0015
0.0019	43.0	55728	0.0084	0.0055	0.0011

Framework versions

Transformers 4.45.2
Pytorch 2.1.0+cu118
Datasets 3.0.1
Tokenizers 0.20.1

Task: Automatic Speech Recognition
Library: transformers
Type: wav2vec2-bert
License: mit
Languages: lg
Created: Oct 21, 2024
Updated: Oct 21, 2024