fine-w2v2base-bs16-ep100-lr2e-05-linguistic-rmsnorm-focal_ctc_a0.75_g2.0-0.05_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: 3.0758
• Wer: 0.0993

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
1625.3502	0.94	50	806.0434	15.8991
1410.0119	1.89	100	629.8961	16.0119
709.9929	2.83	150	85.9791	0.9988
90.0141	3.77	200	65.2759	1.0
83.76	4.72	250	63.1484	1.0
80.6192	5.66	300	61.0757	1.0
77.8424	6.6	350	59.5632	1.0
74.3949	7.55	400	58.2095	1.0
72.7863	8.49	450	57.4982	1.0
73.4921	9.43	500	57.0917	1.0
72.5605	10.38	550	57.0573	1.0
73.0506	11.32	600	57.1200	1.0
70.646	12.26	650	57.6753	0.9994
68.9098	13.21	700	52.4541	0.9763
55.5991	14.15	750	26.6825	0.4343
30.2222	15.09	800	14.2918	0.2538
20.0258	16.04	850	10.1003	0.1932
15.5053	16.98	900	8.1504	0.1790
12.741	17.92	950	6.8064	0.1565
11.3321	18.87	1000	6.2718	0.1492
10.2277	19.81	1050	5.6648	0.1443
9.2196	20.75	1100	5.0958	0.1292
8.7783	21.7	1150	5.0376	0.1368
8.1951	22.64	1200	4.7069	0.1261
7.5671	23.58	1250	4.5874	0.1290
7.0327	24.53	1300	4.3482	0.1169
7.0547	25.47	1350	4.0875	0.1154
6.4848	26.42	1400	4.0527	0.1185
6.4467	27.36	1450	3.9355	0.1162
6.1179	28.3	1500	3.8567	0.1187
5.8745	29.25	1550	3.8254	0.1162
5.6889	30.19	1600	3.7496	0.1079
5.3704	31.13	1650	3.8083	0.1093
5.5541	32.08	1700	3.7809	0.1057
5.353	33.02	1750	3.6145	0.1032
5.1094	33.96	1800	3.5845	0.1086
4.8619	34.91	1850	3.6174	0.1077
4.9216	35.85	1900	3.4465	0.1014
4.8789	36.79	1950	3.3949	0.0954
4.4445	37.74	2000	3.3273	0.0945
4.5174	38.68	2050	3.3895	0.1003
4.3901	39.62	2100	3.3033	0.0997
4.3151	40.57	2150	3.2694	0.0963
4.1073	41.51	2200	3.3102	0.1007
4.0983	42.45	2250	3.3103	0.1111
4.0334	43.4	2300	3.3534	0.1086
4.0756	44.34	2350	3.3035	0.1055
3.9379	45.28	2400	3.3220	0.1031
3.6192	46.23	2450	3.2740	0.1019
3.6579	47.17	2500	3.2513	0.1012
3.4756	48.11	2550	3.2198	0.0973
3.548	49.06	2600	3.1895	0.1028
3.4199	50.0	2650	3.2025	0.1011
3.3485	50.94	2700	3.1908	0.1033
3.4955	51.89	2750	3.2124	0.1018
3.0636	52.83	2800	3.2538	0.1074
3.3026	53.77	2850	3.1777	0.1017
3.2866	54.72	2900	3.2008	0.1018
3.0405	55.66	2950	3.2269	0.1057
2.9104	56.6	3000	3.2111	0.1017
3.1029	57.55	3050	3.2158	0.1001
3.0138	58.49	3100	3.2130	0.1039
2.9072	59.43	3150	3.1995	0.0991
3.0077	60.38	3200	3.2179	0.1016
2.7565	61.32	3250	3.1346	0.0953
2.9245	62.26	3300	3.1797	0.1009
2.6777	63.21	3350	3.1870	0.1069
2.7601	64.15	3400	3.2077	0.1004
2.6767	65.09	3450	3.1921	0.0982
2.7252	66.04	3500	3.1475	0.0972
2.7375	66.98	3550	3.1569	0.0992
2.7624	67.92	3600	3.1613	0.0980
2.5993	68.87	3650	3.1353	0.0948
2.6334	69.81	3700	3.0991	0.0961
2.5925	70.75	3750	3.0972	0.0988
2.6538	71.7	3800	3.1004	0.0954
2.5137	72.64	3850	3.0980	0.0956
2.4971	73.58	3900	3.0919	0.0968
2.3636	74.53	3950	3.0861	0.0938
2.6246	75.47	4000	3.1040	0.0959
2.5092	76.42	4050	3.0918	0.0949
2.5617	77.36	4100	3.0869	0.0960
2.4673	78.3	4150	3.0836	0.0988
2.4177	79.25	4200	3.0862	0.0962
2.533	80.19	4250	3.0585	0.0979
2.4484	81.13	4300	3.0875	0.1022
2.4034	82.08	4350	3.0803	0.0993
2.3971	83.02	4400	3.0964	0.0991
2.3776	83.96	4450	3.0840	0.0972
2.529	84.91	4500	3.0982	0.1002
2.3854	85.85	4550	3.0903	0.0992
2.4461	86.79	4600	3.0840	0.0981
2.4031	87.74	4650	3.0963	0.1006
2.4072	88.68	4700	3.0897	0.0976
2.398	89.62	4750	3.0871	0.0998
2.1919	90.57	4800	3.0778	0.0995
2.5859	91.51	4850	3.0781	0.1003
2.1701	92.45	4900	3.0706	0.0983
2.4991	93.4	4950	3.0815	0.0992
2.3048	94.34	5000	3.0771	0.0988
2.2576	95.28	5050	3.0785	0.0993
2.3997	96.23	5100	3.0770	0.0987
2.3028	97.17	5150	3.0766	0.0986
2.397	98.11	5200	3.0765	0.0991
2.3609	99.06	5250	3.0757	0.0990
2.4194	100.0	5300	3.0758	0.0993

Framework versions

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