wav2vec2-base-vi-vivos

This model is a fine-tuned version of nguyenvulebinh/wav2vec2-base-vi on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3990
• Wer: 0.2339

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: 1e-05
• train_batch_size: 32
• 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
2.2335	1.37	500	1.9337	0.9589
2.0861	2.74	1000	1.6892	0.9078
1.8044	4.11	1500	1.3953	0.7989
1.5782	5.48	2000	1.1773	0.7221
1.3843	6.85	2500	1.0011	0.6243
1.2181	8.22	3000	0.8656	0.5361
1.1115	9.59	3500	0.7775	0.4933
0.9948	10.96	4000	0.6933	0.4286
0.9307	12.33	4500	0.6314	0.3959
0.8529	13.7	5000	0.5832	0.3560
0.8094	15.07	5500	0.5446	0.3292
0.7517	16.44	6000	0.5156	0.3064
0.701	17.81	6500	0.4899	0.2907
0.6753	19.18	7000	0.4668	0.2742
0.6621	20.55	7500	0.4528	0.2621
0.6455	21.92	8000	0.4345	0.2564
0.6159	23.29	8500	0.4258	0.2475
0.596	24.66	9000	0.4143	0.2435
0.5833	26.03	9500	0.4063	0.2387
0.5899	27.4	10000	0.4029	0.2357
0.5729	28.77	10500	0.3990	0.2339

Framework versions

• Transformers 4.32.1
• Pytorch 2.0.1+cu118
• Datasets 2.14.4
• Tokenizers 0.13.3