fine-w2v2base-bs16-ep100-lr2e-05-linguistic-rmsnorm-focal_ctc_a0.25_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: 1.0890
• Wer: 0.0978

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
522.579	0.94	50	251.0458	14.9639
317.1515	1.89	100	84.8688	0.9971
57.5912	2.83	150	24.7802	1.0
28.4209	3.77	200	21.7164	1.0
27.1215	4.72	250	21.1519	1.0
26.1663	5.66	300	20.5749	1.0
25.4374	6.6	350	20.1536	1.0
24.5548	7.55	400	19.6697	1.0
24.1548	8.49	450	19.5588	1.0
24.4262	9.43	500	19.4870	1.0
24.0949	10.38	550	19.5979	1.0
24.1762	11.32	600	20.2140	1.0
23.2554	12.26	650	20.0865	1.0
22.7304	13.21	700	19.4624	0.9999
22.0028	14.15	750	17.8907	0.9991
20.0064	15.09	800	11.9386	0.7020
12.6884	16.04	850	5.5360	0.3161
7.2843	16.98	900	3.4633	0.2105
5.2335	17.92	950	2.6680	0.1726
4.2601	18.87	1000	2.1859	0.1505
3.6512	19.81	1050	1.9664	0.1471
3.2164	20.75	1100	1.7849	0.1351
3.0286	21.7	1150	1.6425	0.1313
2.776	22.64	1200	1.5509	0.1323
2.5805	23.58	1250	1.5048	0.1281
2.372	24.53	1300	1.4450	0.1169
2.3566	25.47	1350	1.3800	0.1136
2.137	26.42	1400	1.3534	0.1164
2.1112	27.36	1450	1.3263	0.1132
1.9889	28.3	1500	1.3027	0.1091
1.9183	29.25	1550	1.2998	0.1117
1.8744	30.19	1600	1.2637	0.1093
1.75	31.13	1650	1.2712	0.1059
1.7865	32.08	1700	1.2368	0.1092
1.6976	33.02	1750	1.2081	0.1039
1.6891	33.96	1800	1.2146	0.1064
1.5919	34.91	1850	1.2082	0.1080
1.5751	35.85	1900	1.2008	0.1033
1.5628	36.79	1950	1.1641	0.1024
1.4812	37.74	2000	1.2022	0.1054
1.4784	38.68	2050	1.1667	0.1025
1.4142	39.62	2100	1.1611	0.1054
1.3841	40.57	2150	1.1252	0.0979
1.3636	41.51	2200	1.1582	0.1022
1.3526	42.45	2250	1.1616	0.1080
1.2923	43.4	2300	1.1714	0.1045
1.2576	44.34	2350	1.1561	0.1035
1.2791	45.28	2400	1.1193	0.1006
1.2104	46.23	2450	1.1346	0.1026
1.1839	47.17	2500	1.1126	0.1009
1.1314	48.11	2550	1.1136	0.0996
1.1772	49.06	2600	1.1369	0.1029
1.1137	50.0	2650	1.1157	0.1012
1.1125	50.94	2700	1.1241	0.1015
1.1536	51.89	2750	1.1277	0.1012
1.0589	52.83	2800	1.1413	0.1142
1.1234	53.77	2850	1.1188	0.1034
1.1047	54.72	2900	1.1186	0.1068
0.9979	55.66	2950	1.1079	0.1007
0.9788	56.6	3000	1.0918	0.0939
1.009	57.55	3050	1.1172	0.1024
0.9942	58.49	3100	1.1139	0.0990
0.9602	59.43	3150	1.1063	0.1017
0.9813	60.38	3200	1.1151	0.1047
0.9112	61.32	3250	1.0930	0.0970
0.9705	62.26	3300	1.0990	0.0993
0.8753	63.21	3350	1.1053	0.1039
0.9259	64.15	3400	1.0978	0.0984
0.8877	65.09	3450	1.1047	0.0987
0.9111	66.04	3500	1.0937	0.1009
0.9103	66.98	3550	1.0963	0.0998
0.9031	67.92	3600	1.0969	0.1024
0.876	68.87	3650	1.0920	0.0964
0.8722	69.81	3700	1.0868	0.0958
0.8751	70.75	3750	1.0880	0.0966
0.8816	71.7	3800	1.0879	0.0974
0.8488	72.64	3850	1.0898	0.0974
0.8327	73.58	3900	1.0848	0.0978
0.7818	74.53	3950	1.0878	0.0957
0.8569	75.47	4000	1.0838	0.0997
0.8078	76.42	4050	1.0725	0.0983
0.8557	77.36	4100	1.0776	0.1000
0.8361	78.3	4150	1.0857	0.0978
0.7911	79.25	4200	1.0816	0.0953
0.8146	80.19	4250	1.0816	0.0969
0.8237	81.13	4300	1.0928	0.1005
0.7944	82.08	4350	1.0918	0.0965
0.8108	83.02	4400	1.0946	0.0968
0.7892	83.96	4450	1.0921	0.0968
0.8261	84.91	4500	1.0867	0.0975
0.7909	85.85	4550	1.0858	0.0964
0.804	86.79	4600	1.0832	0.0966
0.7981	87.74	4650	1.0888	0.0984
0.7975	88.68	4700	1.0890	0.0986
0.7966	89.62	4750	1.0862	0.0962
0.7295	90.57	4800	1.0895	0.0968
0.8447	91.51	4850	1.0907	0.0981
0.7192	92.45	4900	1.0872	0.0967
0.8368	93.4	4950	1.0875	0.0971
0.7808	94.34	5000	1.0887	0.0977
0.76	95.28	5050	1.0896	0.0978
0.7858	96.23	5100	1.0896	0.0974
0.766	97.17	5150	1.0894	0.0978
0.7899	98.11	5200	1.0898	0.0978
0.784	99.06	5250	1.0889	0.0978
0.801	100.0	5300	1.0890	0.0978

Framework versions

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