wav2vec2-darija

khaoulaoub/wav2vec2-darija

wav2vec2-darija is a automatic speech recognition model from khaoulaoub. Use it when you need speech turned into text. It is set up for transformers. The card lists the license as apache-2.0.

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

Downloads · 30 days

3% of all-time downloads

All-time downloads

358

Public

Repo size

2.5 GB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.bin1.3 GB · 100%

At a glance

Task: Automatic Speech Recognition
Library: transformers
License: apache-2.0
Model type: wav2vec2
Access: Public
Created: Jul 25, 2023
Updated: Oct 18, 2023
SHA: 0a841770

Base models

facebook/wav2vec2-large-xlsr-53

Model card

From the Hugging Face model README

wav2vec2-darija

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.4265
Wer: 0.4101
Cer: 0.1327

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: 0.0001
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 1000
num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
10.5265	0.49	500	3.2082	1.0	0.9971
2.9629	0.98	1000	2.9619	1.0000	0.9357
2.6527	1.46	1500	1.4934	1.0974	0.4550
1.2946	1.95	2000	0.6904	0.7809	0.2566
0.9059	2.44	2500	0.5906	0.7002	0.2190
0.8133	2.93	3000	0.5199	0.6382	0.2049
0.7026	3.41	3500	0.4570	0.5986	0.1919
0.6572	3.9	4000	0.4238	0.5792	0.1848
0.5904	4.39	4500	0.4116	0.5637	0.1807
0.5593	4.88	5000	0.3850	0.5414	0.1734
0.5084	5.37	5500	0.3951	0.5409	0.1733
0.5033	5.85	6000	0.4449	0.5176	0.1649
0.4712	6.34	6500	0.5485	0.5172	0.1658
0.4459	6.83	7000	0.5259	0.5061	0.1623
0.4246	7.32	7500	0.3686	0.4991	0.1605
0.4261	7.8	8000	0.3663	0.4898	0.1589
0.4078	8.29	8500	0.3740	0.4858	0.1564
0.3783	8.78	9000	0.3907	0.4824	0.1566
0.3647	9.27	9500	0.3424	0.4750	0.1525
0.3527	9.76	10000	0.3444	0.4692	0.1513
0.3482	10.24	10500	0.3856	0.4692	0.1507
0.3338	10.73	11000	0.3650	0.4664	0.1512
0.3198	11.22	11500	0.3516	0.4628	0.1492
0.3218	11.71	12000	0.3660	0.4644	0.1491
0.3115	12.2	12500	0.3490	0.4545	0.1475
0.2977	12.68	13000	0.3555	0.4504	0.1451
0.2958	13.17	13500	0.3425	0.4571	0.1449
0.278	13.66	14000	0.4035	0.4520	0.1446
0.2716	14.15	14500	0.3552	0.4492	0.1437
0.2729	14.63	15000	0.3665	0.4470	0.1432
0.2691	15.12	15500	0.3700	0.4498	0.1444
0.2563	15.61	16000	0.3658	0.4423	0.1421
0.2511	16.1	16500	0.4152	0.4408	0.1425
0.2348	16.59	17000	0.4889	0.4375	0.1416
0.2437	17.07	17500	0.4209	0.4382	0.1413
0.2388	17.56	18000	0.6032	0.4359	0.1408
0.2235	18.05	18500	0.4831	0.4369	0.1402
0.2197	18.54	19000	0.4989	0.4345	0.1402
0.2285	19.02	19500	0.5929	0.4342	0.1393
0.2224	19.51	20000	0.4098	0.4317	0.1398
0.2183	20.0	20500	0.3547	0.4254	0.1384
0.2113	20.49	21000	0.3926	0.4324	0.1385
0.2125	20.98	21500	0.3982	0.4299	0.1386
0.201	21.46	22000	0.3929	0.4293	0.1389
0.2002	21.95	22500	0.4047	0.4218	0.1372
0.2029	22.44	23000	0.5153	0.4235	0.1375
0.195	22.93	23500	0.5601	0.4198	0.1364
0.182	23.41	24000	0.4596	0.4168	0.1355
0.1889	23.9	24500	0.4165	0.4209	0.1353
0.1795	24.39	25000	0.4096	0.4185	0.1352
0.1809	24.88	25500	0.4732	0.4126	0.1341
0.1762	25.37	26000	0.4324	0.4146	0.1347
0.1764	25.85	26500	0.4462	0.4160	0.1347
0.1805	26.34	27000	0.3955	0.4107	0.1333
0.1733	26.83	27500	0.4182	0.4135	0.1336
0.1651	27.32	28000	0.4111	0.4104	0.1330
0.1713	27.8	28500	0.4426	0.4126	0.1332
0.1766	28.29	29000	0.4426	0.4085	0.1328
0.1631	28.78	29500	0.4248	0.4083	0.1328
0.1608	29.27	30000	0.4334	0.4096	0.1327
0.1688	29.76	30500	0.4265	0.4101	0.1327

Framework versions

Transformers 4.34.0
Pytorch 2.1.0+cu121
Datasets 2.14.5
Tokenizers 0.14.1

Task: Automatic Speech Recognition
Library: transformers
Type: wav2vec2
License: apache-2.0
Created: Jul 25, 2023
Updated: Oct 18, 2023