wav2vec2-base-finetune-vi-v5

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

• Loss: 0.2373
• Wer: 0.1681

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: 5e-05
• train_batch_size: 16
• 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: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
20.4793	0.99	500	8.0023	1.0
4.5661	1.98	1000	3.4221	1.0
3.356	2.96	1500	3.2739	1.0
2.2533	3.95	2000	0.8992	0.4884
0.8412	4.94	2500	0.5181	0.3061
0.5738	5.93	3000	0.3986	0.2586
0.4608	6.92	3500	0.3545	0.2230
0.399	7.91	4000	0.3220	0.2003
0.341	8.89	4500	0.2936	0.1928
0.3094	9.88	5000	0.2727	0.1873
0.2834	10.87	5500	0.2721	0.1813
0.2761	11.86	6000	0.2704	0.1817
0.2505	12.85	6500	0.2597	0.1766
0.2472	13.83	7000	0.2460	0.1744
0.2335	14.82	7500	0.2486	0.1728
0.2183	15.81	8000	0.2430	0.1714
0.2153	16.8	8500	0.2433	0.1697
0.2029	17.79	9000	0.2408	0.1688
0.2094	18.77	9500	0.2349	0.1694
0.2045	19.76	10000	0.2373	0.1681

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.0
• Datasets 2.8.0
• Tokenizers 0.13.3