w2v-bert-2.0-bangala-gpu-CV16.0_v2

Sajjo/w2v-bert-2.0-bangala-gpu-CV16.0_v2

w2v-bert-2.0-bangala-gpu-CV16.0_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

2% of all-time downloads

All-time downloads

252

Public

Parameters

606M

33.9 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: Apr 5, 2024
Updated: Apr 8, 2024
SHA: be84686d

Base models

facebook/w2v-bert-2.0

Datasets

common_voice_16_0

Spaces

cdactvm/demoASR

Model card

From the Hugging Face model README

w2v-bert-2.0-bangala-gpu-CV16.0_v2

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the common_voice_16_0 dataset. It achieves the following results on the evaluation set:

Loss: 0.4490
Wer: 0.4811

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: 4.42184e-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
3.5221	0.31	300	0.5900	0.6271
1.2024	0.63	600	0.4088	0.4071
0.9149	0.94	900	0.3200	0.3270
0.8124	1.26	1200	0.2965	0.3080
0.7028	1.57	1500	0.2759	0.2884
0.6301	1.89	1800	0.2435	0.2671
0.6147	2.2	2100	0.2335	0.2477
0.6304	2.52	2400	0.2248	0.2458
0.5921	2.83	2700	0.2326	0.2441
0.495	3.15	3000	0.2180	0.2378
0.4987	3.46	3300	0.2139	0.2227
0.5669	3.78	3600	0.2097	0.2236
0.5904	4.09	3900	0.2038	0.2178
0.6016	4.41	4200	0.2091	0.2131
0.5325	4.72	4500	0.2064	0.2147
0.5271	5.04	4800	0.2002	0.2159
0.5229	5.35	5100	0.2069	0.2209
0.5843	5.67	5400	0.2090	0.2202
0.5477	5.98	5700	0.2085	0.2175
0.508	6.3	6000	0.2046	0.2158
0.5226	6.61	6300	0.2515	0.3250
0.7576	6.93	6600	0.2343	0.2364
1.0089	7.24	6900	0.2731	0.2713
0.9462	7.56	7200	0.2588	0.2648
0.8648	7.87	7500	0.2916	0.3393
1.1282	8.19	7800	0.3830	0.4583
1.3279	8.5	8100	0.3910	0.4117
1.2722	8.82	8400	0.4424	0.3442
1.2886	9.13	8700	0.4421	0.4011
1.3274	9.45	9000	0.4483	0.4769
1.3235	9.76	9300	0.4490	0.4811

Framework versions

Transformers 4.39.3
Pytorch 2.1.2+cu121
Datasets 2.18.0
Tokenizers 0.15.0

Task: Automatic Speech Recognition
Library: transformers
Type: wav2vec2-bert
License: mit
Created: Apr 5, 2024
Updated: Apr 8, 2024