fine-w2v2base-bs16-ep100-lr2e-05-linguistic-rmsnorm-focal_ctc_a0.5_g0.5-0.05_10_0.004_40

This model is a fine-tuned version of nguyenvulebinh/wav2vec2-base-vietnamese-250h on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.1651
• Wer: 0.0985

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 64
• total_eval_batch_size: 32
• optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1068.6505	0.94	50	527.9987	15.8840
924.9194	1.89	100	365.2998	15.7290
268.1159	2.83	150	45.3076	1.0
56.3914	3.77	200	42.2653	1.0
54.5992	4.72	250	41.2272	1.0
52.7823	5.66	300	40.2126	1.0
51.1032	6.6	350	39.6254	1.0
49.2081	7.55	400	38.7989	1.0
48.3538	8.49	450	38.5792	1.0
48.8615	9.43	500	38.4622	1.0
48.1912	10.38	550	38.1422	1.0
48.3589	11.32	600	38.6144	1.0
46.5985	12.26	650	39.6394	1.0473
45.5769	13.21	700	37.7580	0.9992
44.1749	14.15	750	36.0692	0.9991
41.8932	15.09	800	27.9404	0.9316
29.8551	16.04	850	14.1211	0.3930
16.9135	16.98	900	7.9824	0.2228
11.5569	17.92	950	5.8073	0.1693
9.1965	18.87	1000	4.6891	0.1577
7.6846	19.81	1050	4.0938	0.1444
6.6186	20.75	1100	3.7074	0.1337
6.1733	21.7	1150	3.3769	0.1279
5.5833	22.64	1200	3.1933	0.1288
5.1097	23.58	1250	3.0785	0.1232
4.8098	24.53	1300	3.0687	0.1210
4.784	25.47	1350	2.7771	0.1152
4.3574	26.42	1400	2.7347	0.1200
4.2972	27.36	1450	2.6853	0.1147
4.1072	28.3	1500	2.5680	0.1185
3.9651	29.25	1550	2.5938	0.1200
4.0325	30.19	1600	2.5324	0.1180
3.6586	31.13	1650	2.5847	0.1113
3.7213	32.08	1700	2.5886	0.1116
3.4746	33.02	1750	2.4285	0.1005
3.3572	33.96	1800	2.4607	0.1073
3.2202	34.91	1850	2.4459	0.1103
3.2437	35.85	1900	2.3630	0.1027
3.1303	36.79	1950	2.3281	0.1025
3.0037	37.74	2000	2.3129	0.1019
3.0523	38.68	2050	2.2962	0.0988
2.8943	39.62	2100	2.3238	0.1021
2.8502	40.57	2150	2.3549	0.1044
2.7045	41.51	2200	2.3680	0.1018
2.7291	42.45	2250	2.4172	0.1129
2.6162	43.4	2300	2.3216	0.1018
2.5643	44.34	2350	2.2663	0.0979
2.5842	45.28	2400	2.2408	0.0986
2.4498	46.23	2450	2.2695	0.1017
2.4177	47.17	2500	2.2029	0.0980
2.3297	48.11	2550	2.2254	0.0938
2.3637	49.06	2600	2.2551	0.1011
2.2528	50.0	2650	2.2350	0.1012
2.2221	50.94	2700	2.2253	0.0968
2.3083	51.89	2750	2.2426	0.0958
2.0585	52.83	2800	2.2169	0.0972
2.2349	53.77	2850	2.2151	0.1004
2.1969	54.72	2900	2.2562	0.1024
2.0415	55.66	2950	2.2862	0.1027
2.0126	56.6	3000	2.2167	0.1015
2.1	57.55	3050	2.2360	0.1024
2.0739	58.49	3100	2.2198	0.1056
1.9875	59.43	3150	2.1716	0.0987
2.0259	60.38	3200	2.2143	0.0999
1.8519	61.32	3250	2.1837	0.0958
1.9733	62.26	3300	2.1865	0.1008
1.8496	63.21	3350	2.2045	0.1054
1.9354	64.15	3400	2.1783	0.1002
1.8247	65.09	3450	2.1670	0.0989
1.8418	66.04	3500	2.1823	0.0993
1.8259	66.98	3550	2.1875	0.0990
1.8458	67.92	3600	2.2048	0.1000
1.7796	68.87	3650	2.2019	0.0975
1.7931	69.81	3700	2.1673	0.0955
1.789	70.75	3750	2.1924	0.0985
1.8166	71.7	3800	2.1839	0.0964
1.692	72.64	3850	2.1771	0.0950
1.6898	73.58	3900	2.1621	0.0944
1.5916	74.53	3950	2.1718	0.0973
1.7778	75.47	4000	2.1617	0.0973
1.6884	76.42	4050	2.1566	0.0982
1.7182	77.36	4100	2.1699	0.0968
1.6774	78.3	4150	2.1849	0.0964
1.5921	79.25	4200	2.1785	0.0962
1.7108	80.19	4250	2.1642	0.0981
1.7039	81.13	4300	2.1836	0.0997
1.6068	82.08	4350	2.1924	0.1002
1.6267	83.02	4400	2.1808	0.0979
1.6209	83.96	4450	2.1808	0.0976
1.6989	84.91	4500	2.1661	0.0976
1.6126	85.85	4550	2.1738	0.0988
1.6623	86.79	4600	2.1695	0.0979
1.637	87.74	4650	2.1702	0.0989
1.63	88.68	4700	2.1631	0.0973
1.6153	89.62	4750	2.1659	0.0985
1.4989	90.57	4800	2.1691	0.0991
1.7316	91.51	4850	2.1688	0.0986
1.4623	92.45	4900	2.1635	0.0980
1.6932	93.4	4950	2.1671	0.0986
1.5762	94.34	5000	2.1678	0.0990
1.5346	95.28	5050	2.1654	0.0984
1.6015	96.23	5100	2.1667	0.0986
1.5609	97.17	5150	2.1653	0.0982
1.6414	98.11	5200	2.1648	0.0983
1.581	99.06	5250	2.1666	0.0987
1.6469	100.0	5300	2.1651	0.0985

Framework versions

• Transformers 4.34.0
• Pytorch 2.0.1
• Datasets 2.14.5
• Tokenizers 0.14.1