wav2vec2-base-vietnamese-clean-dataset-20-epochs

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: 0.5701
• Wer: 0.2489

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: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
15.8906	0.41	500	22.2498	1.0
10.675	0.81	1000	16.8372	1.0
7.7286	1.22	1500	10.6552	1.0
5.3176	1.63	2000	6.4350	1.0
4.004	2.04	2500	4.2915	1.0
3.5239	2.44	3000	3.8151	1.0
3.4366	2.85	3500	3.5758	1.0
3.3874	3.26	4000	3.4953	1.0
3.3758	3.66	4500	3.4716	1.0
3.3647	4.07	5000	3.6072	1.0
3.3574	4.48	5500	3.5273	1.0
3.303	4.89	6000	3.4187	1.0000
3.0766	5.29	6500	2.9887	0.9993
2.7324	5.7	7000	2.5486	1.0010
2.3984	6.11	7500	2.2322	0.9850
2.1125	6.51	8000	1.9550	0.8958
1.8964	6.92	8500	1.7719	0.8172
1.7212	7.33	9000	1.5676	0.7549
1.5851	7.74	9500	1.4595	0.7091
1.49	8.14	10000	1.2293	0.6449
1.3883	8.55	10500	1.1185	0.6026
1.2862	8.96	11000	1.0546	0.5747
1.2146	9.36	11500	0.9808	0.5227
1.153	9.77	12000	0.9699	0.4917
1.0782	10.18	12500	0.9498	0.4544
1.0517	10.59	13000	0.9242	0.4206
1.0001	10.99	13500	0.8411	0.3910
0.9578	11.4	14000	0.8315	0.3708
0.9302	11.81	14500	0.8107	0.3521
0.8978	12.21	15000	0.7713	0.3351
0.8738	12.62	15500	0.7798	0.3253
0.8932	13.03	16000	0.7182	0.3117
0.8267	13.44	16500	0.7165	0.3054
0.8007	13.84	17000	0.6838	0.2973
0.7854	14.25	17500	0.6783	0.2913
0.7878	14.66	18000	0.6394	0.2851
0.7738	15.07	18500	0.5956	0.2771
0.7626	15.47	19000	0.6121	0.2708
0.7342	15.88	19500	0.5865	0.2661
0.7297	16.29	20000	0.5963	0.2646
0.7113	16.69	20500	0.5828	0.2601
0.7302	17.1	21000	0.5981	0.2601
0.721	17.51	21500	0.5881	0.2555
0.7089	17.92	22000	0.5841	0.2545
0.7059	18.32	22500	0.5794	0.2525
0.6969	18.73	23000	0.5910	0.2507
0.7065	19.14	23500	0.5707	0.2498
0.6869	19.54	24000	0.5736	0.2496
0.7308	19.95	24500	0.5701	0.2489

Framework versions

• Transformers 4.32.1
• Pytorch 2.0.1+cu117
• Datasets 2.14.4
• Tokenizers 0.13.3