wav2vec2-base-finetune-vi-v4

This model is a fine-tuned version of nguyenvulebinh/wav2vec2-base-vietnamese-250h on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2829
• Wer: 0.1587

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
5.0547	0.49	500	3.4652	1.0
2.9133	0.99	1000	1.1003	0.5035
1.0644	1.48	1500	0.5960	0.3199
0.8294	1.98	2000	0.5098	0.2809
0.6965	2.47	2500	0.5010	0.2596
0.6646	2.96	3000	0.4209	0.2398
0.5753	3.46	3500	0.4089	0.2361
0.5265	3.95	4000	0.3868	0.2195
0.4701	4.45	4500	0.3626	0.2171
0.4617	4.94	5000	0.3693	0.2160
0.4343	5.43	5500	0.3661	0.2058
0.4246	5.93	6000	0.3618	0.2067
0.3881	6.42	6500	0.3654	0.2044
0.3948	6.92	7000	0.3586	0.2009
0.367	7.41	7500	0.3431	0.1961
0.3449	7.91	8000	0.3196	0.1944
0.3168	8.4	8500	0.3310	0.1912
0.3393	8.89	9000	0.3418	0.1879
0.3197	9.39	9500	0.3434	0.1888
0.2954	9.88	10000	0.3371	0.1863
0.2968	10.38	10500	0.2941	0.1899
0.2802	10.87	11000	0.3095	0.1836
0.2783	11.36	11500	0.3275	0.1822
0.3027	11.86	12000	0.3103	0.1806
0.2645	12.35	12500	0.3247	0.1842
0.2958	12.85	13000	0.3242	0.1801
0.2648	13.34	13500	0.3169	0.1775
0.2461	13.83	14000	0.2926	0.1764
0.247	14.33	14500	0.3033	0.1741
0.2212	14.82	15000	0.2901	0.1749
0.2239	15.32	15500	0.3237	0.1758
0.2093	15.81	16000	0.2972	0.1759
0.2284	16.3	16500	0.3025	0.1749
0.228	16.8	17000	0.2862	0.1708
0.2033	17.29	17500	0.3039	0.1745
0.189	17.79	18000	0.3084	0.1708
0.1992	18.28	18500	0.2931	0.1735
0.1989	18.77	19000	0.2964	0.1693
0.1953	19.27	19500	0.3082	0.1715
0.1813	19.76	20000	0.2859	0.1702
0.1703	20.26	20500	0.2936	0.1680
0.1939	20.75	21000	0.2871	0.1684
0.1769	21.25	21500	0.2994	0.1646
0.1795	21.74	22000	0.2990	0.1669
0.17	22.23	22500	0.2839	0.1663
0.1507	22.73	23000	0.3125	0.1666
0.1676	23.22	23500	0.2867	0.1611
0.1675	23.72	24000	0.3099	0.1607
0.171	24.21	24500	0.3000	0.1627
0.1483	24.7	25000	0.3010	0.1629
0.1452	25.2	25500	0.2910	0.1641
0.1394	25.69	26000	0.2878	0.1605
0.1478	26.19	26500	0.2881	0.1617
0.1426	26.68	27000	0.2714	0.1607
0.1342	27.17	27500	0.2941	0.1615
0.1385	27.67	28000	0.2758	0.1594
0.1541	28.16	28500	0.2830	0.1592
0.153	28.66	29000	0.2789	0.1575
0.1359	29.15	29500	0.2819	0.1588
0.1276	29.64	30000	0.2829	0.1587

Framework versions

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