wav2vec2-xls-r-300m-vivos

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5156
• Wer: 0.3337

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
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
8.3411	0.66	500	3.5728	1.0
3.418	1.31	1000	3.1432	1.0
1.6726	1.97	1500	0.7995	0.7146
0.8244	2.62	2000	0.5569	0.5370
0.6392	3.28	2500	0.4773	0.4744
0.5537	3.94	3000	0.4592	0.4631
0.4956	4.59	3500	0.4649	0.4536
0.4539	5.25	4000	0.4345	0.4175
0.4144	5.91	4500	0.4291	0.4204
0.3899	6.56	5000	0.4325	0.4105
0.3748	7.22	5500	0.4151	0.3954
0.3543	7.87	6000	0.4320	0.4070
0.3335	8.53	6500	0.4061	0.3776
0.3266	9.19	7000	0.4307	0.3899
0.3107	9.84	7500	0.4404	0.3866
0.2886	10.5	8000	0.4528	0.3825
0.2897	11.15	8500	0.4027	0.3731
0.2757	11.81	9000	0.4423	0.3837
0.2582	12.47	9500	0.4412	0.3717
0.2598	13.12	10000	0.4410	0.3609
0.2421	13.78	10500	0.4398	0.3651
0.2414	14.44	11000	0.4488	0.3585
0.2259	15.09	11500	0.4528	0.3572
0.2269	15.75	12000	0.4613	0.3590
0.2109	16.4	12500	0.4492	0.3610
0.2097	17.06	13000	0.4468	0.3522
0.1992	17.72	13500	0.4520	0.3531
0.1949	18.37	14000	0.4782	0.3525
0.1924	19.03	14500	0.4643	0.3459
0.1906	19.69	15000	0.4839	0.3519
0.1837	20.34	15500	0.4891	0.3427
0.1744	21.0	16000	0.4905	0.3481
0.1705	21.65	16500	0.4758	0.3445
0.1697	22.31	17000	0.4765	0.3441
0.1657	22.97	17500	0.5059	0.3447
0.1582	23.62	18000	0.4941	0.3446
0.159	24.28	18500	0.4977	0.3469
0.1562	24.93	19000	0.4966	0.3415
0.1516	25.59	19500	0.5130	0.3403
0.144	26.25	20000	0.5049	0.3390
0.1429	26.9	20500	0.5130	0.3355
0.1378	27.56	21000	0.5140	0.3371
0.1436	28.22	21500	0.5172	0.3351
0.1363	28.87	22000	0.5215	0.3361
0.1332	29.53	22500	0.5156	0.3337

Framework versions

• Transformers 4.39.3
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.2