w2v-bert-2.0-tamil-gpu-custom_clean_v2

Sajjo/w2v-bert-2.0-tamil-gpu-custom_clean_v2

w2v-bert-2.0-tamil-gpu-custom_clean_v2 is a automatic speech recognition model from Sajjo. 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

28% of all-time downloads

All-time downloads

Public

Parameters

606M

46 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: May 1, 2024
Updated: May 2, 2024
SHA: 6505b992

Base models

facebook/w2v-bert-2.0

Model card

From the Hugging Face model README

w2v-bert-2.0-tamil-gpu-custom_clean_v2

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: 0.1042
Wer: 0.1892

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: 2.5356e-05
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: 10
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.9949	0.25	300	0.5158	0.6736
0.4051	0.5	600	0.1858	0.3035
0.2789	0.76	900	0.1670	0.2730
0.2352	1.01	1200	0.1479	0.2594
0.1988	1.26	1500	0.1428	0.2464
0.1879	1.51	1800	0.1388	0.2391
0.1757	1.76	2100	0.1244	0.2412
0.1683	2.01	2400	0.1278	0.2231
0.1475	2.27	2700	0.1231	0.2240
0.1438	2.52	3000	0.1219	0.2192
0.1422	2.77	3300	0.1216	0.2128
0.1337	3.02	3600	0.1145	0.2087
0.1212	3.27	3900	0.1131	0.2061
0.1195	3.52	4200	0.1159	0.2147
0.1169	3.78	4500	0.1111	0.2083
0.1162	4.03	4800	0.1116	0.2058
0.1073	4.28	5100	0.1138	0.2114
0.1011	4.53	5400	0.1103	0.2057
0.1008	4.78	5700	0.1096	0.2018
0.1016	5.03	6000	0.1045	0.2008
0.092	5.29	6300	0.1104	0.2008
0.0889	5.54	6600	0.1079	0.2005
0.0936	5.79	6900	0.1036	0.2026
0.0888	6.04	7200	0.1106	0.2109
0.0836	6.29	7500	0.1115	0.2103
0.0807	6.54	7800	0.1104	0.2045
0.0807	6.8	8100	0.1051	0.2039
0.0784	7.05	8400	0.1067	0.1947
0.0719	7.3	8700	0.1051	0.1957
0.0735	7.55	9000	0.1084	0.1894
0.0715	7.8	9300	0.1029	0.1916
0.0732	8.05	9600	0.1059	0.1894
0.0673	8.31	9900	0.1053	0.1890
0.0642	8.56	10200	0.1042	0.1879
0.0669	8.81	10500	0.1039	0.1877
0.0665	9.06	10800	0.1043	0.1881
0.0606	9.31	11100	0.1027	0.1870
0.0615	9.56	11400	0.1046	0.1887
0.0602	9.82	11700	0.1042	0.1892

Framework versions

Transformers 4.39.3
Pytorch 2.1.2+cu121
Datasets 2.18.0
Tokenizers 0.15.2

Task: Automatic Speech Recognition
Library: transformers
Type: wav2vec2-bert
License: mit
Created: May 1, 2024
Updated: May 2, 2024