hindi_wav2vec2_optimized

TheAIchemist13/hindi_wav2vec2_optimized

hindi_wav2vec2_optimized is a automatic speech recognition model from TheAIchemist13. 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

45% of all-time downloads

All-time downloads

Public

Repo size

21.5 GB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.bin378 MB · 100%

At a glance

Task: Automatic Speech Recognition
Library: transformers
License: mit
Model type: wav2vec2
Access: Public
Created: Oct 12, 2023
Updated: Oct 13, 2023
SHA: e42bbdeb

Base models

Harveenchadha/vakyansh-wav2vec2-hindi-him-4200

Model card

From the Hugging Face model README

hindi_wav2vec2_optimized

This model is a fine-tuned version of Harveenchadha/vakyansh-wav2vec2-hindi-him-4200 on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.4854
Wer: 0.3632

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: 100
num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.1601	0.11	25	1.5548	0.4892
2.8628	0.22	50	1.4072	0.6004
1.5585	0.33	75	1.3488	0.5654
2.8588	0.44	100	1.5733	0.5500
1.3384	0.54	125	1.2629	0.5381
1.7817	0.65	150	1.4159	0.5626
1.124	0.76	175	1.1186	0.5003
1.502	0.87	200	1.3352	0.5738
1.0237	0.98	225	1.2497	0.6165
0.9385	1.09	250	0.9858	0.5073
1.5313	1.2	275	1.1366	0.5619
0.9124	1.31	300	0.9704	0.4787
0.649	1.42	325	1.1915	0.5458
0.838	1.53	350	1.5229	0.5836
0.6835	1.63	375	1.0692	0.5052
0.823	1.74	400	0.9683	0.4640
0.968	1.85	425	0.9629	0.4836
0.8596	1.96	450	0.8242	0.4682
0.6729	2.07	475	0.7999	0.4346
0.6426	2.18	500	0.9678	0.4885
0.7213	2.29	525	1.1779	0.5353
0.4131	2.4	550	0.7007	0.4738
0.4188	2.51	575	0.6085	0.4199
0.3784	2.61	600	0.6526	0.4542
0.4181	2.72	625	0.6716	0.4157
0.3194	2.83	650	0.6058	0.4185
0.3553	2.94	675	0.6023	0.4276
0.3242	3.05	700	0.5864	0.4178
0.2655	3.16	725	0.5705	0.3989
0.3149	3.27	750	0.5275	0.3961
0.2454	3.38	775	0.5401	0.3968
0.2792	3.49	800	0.5303	0.3919
0.2488	3.59	825	0.5459	0.4010
0.3013	3.7	850	0.5180	0.3856
0.2306	3.81	875	0.5179	0.3814
0.2578	3.92	900	0.5168	0.3765
0.2344	4.03	925	0.5212	0.3737
0.2002	4.14	950	0.5026	0.3674
0.2443	4.25	975	0.5045	0.3716
0.1857	4.36	1000	0.5083	0.3751
0.2158	4.47	1025	0.4941	0.3681
0.1922	4.58	1050	0.4996	0.3723
0.217	4.68	1075	0.4943	0.3695
0.1888	4.79	1100	0.4849	0.3653
0.1971	4.9	1125	0.4854	0.3632

Framework versions

Transformers 4.34.0
Pytorch 2.0.1+cu118
Datasets 2.14.5
Tokenizers 0.14.1

Task: Automatic Speech Recognition
Library: transformers
Type: wav2vec2
License: mit
Created: Oct 12, 2023
Updated: Oct 13, 2023