fine-w2v2base-bs16-ep100-lr2e-05-linguistic-rmsnorm-focal_ctc_a0.5_g1.0-0.05_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: 2.0619
• Wer: 0.0997

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
1068.4723	0.94	50	533.5247	15.8344
704.929	1.89	100	149.4564	0.9983
104.7041	2.83	150	45.2165	1.0
57.2527	3.77	200	42.7504	1.0
55.1203	4.72	250	41.4518	1.0
52.9285	5.66	300	39.8048	1.0
51.1361	6.6	350	38.6911	1.0
49.1867	7.55	400	37.8983	1.0
48.293	8.49	450	37.5179	1.0
48.8025	9.43	500	37.2562	1.0
48.0603	10.38	550	37.0740	1.0
48.0837	11.32	600	37.0175	0.9999
45.7671	12.26	650	33.4394	0.9620
38.2468	13.21	700	20.5908	0.5614
22.0048	14.15	750	9.7715	0.2622
13.6453	15.09	800	6.5392	0.1925
10.3565	16.04	850	5.1822	0.1627
8.4776	16.98	900	4.4310	0.1547
7.2782	17.92	950	3.9109	0.1441
6.6759	18.87	1000	3.5788	0.1371
6.0682	19.81	1050	3.3775	0.1336
5.5782	20.75	1100	3.1172	0.1222
5.4805	21.7	1150	3.0142	0.1225
5.0893	22.64	1200	2.9002	0.1234
4.9178	23.58	1250	2.9029	0.1257
4.5324	24.53	1300	2.7464	0.1149
4.4924	25.47	1350	2.5754	0.1104
4.1324	26.42	1400	2.6028	0.1099
4.2581	27.36	1450	2.5399	0.1049
3.8897	28.3	1500	2.4484	0.1062
3.8507	29.25	1550	2.4717	0.1081
3.7424	30.19	1600	2.4559	0.1114
3.4716	31.13	1650	2.3895	0.1043
3.5385	32.08	1700	2.4023	0.1079
3.4308	33.02	1750	2.3014	0.1022
3.3027	33.96	1800	2.3091	0.1054
3.078	34.91	1850	2.2783	0.1000
3.1628	35.85	1900	2.2364	0.1029
3.1191	36.79	1950	2.1291	0.0963
2.9528	37.74	2000	2.1785	0.0975
2.9116	38.68	2050	2.1666	0.1006
2.7249	39.62	2100	2.1878	0.1053
2.7466	40.57	2150	2.1900	0.0997
2.6349	41.51	2200	2.1549	0.0963
2.6933	42.45	2250	2.1418	0.1030
2.5316	43.4	2300	2.1705	0.0982
2.5175	44.34	2350	2.1444	0.0991
2.5374	45.28	2400	2.1134	0.0970
2.4234	46.23	2450	2.1473	0.1052
2.318	47.17	2500	2.1129	0.1016
2.2632	48.11	2550	2.1011	0.0908
2.3666	49.06	2600	2.1168	0.0976
2.2127	50.0	2650	2.1183	0.0968
2.132	50.94	2700	2.0882	0.0943
2.2458	51.89	2750	2.0710	0.0934
1.9839	52.83	2800	2.0990	0.1026
2.147	53.77	2850	2.0917	0.1017
2.1353	54.72	2900	2.1009	0.1002
1.9557	55.66	2950	2.1425	0.1057
1.8819	56.6	3000	2.1140	0.0979
2.0495	57.55	3050	2.1637	0.1020
2.027	58.49	3100	2.1385	0.1025
1.9783	59.43	3150	2.1003	0.1002
1.9553	60.38	3200	2.1139	0.1043
1.7827	61.32	3250	2.1029	0.0967
1.9633	62.26	3300	2.0796	0.0941
1.7306	63.21	3350	2.0947	0.1009
1.8145	64.15	3400	2.1027	0.1029
1.7772	65.09	3450	2.1160	0.1014
1.784	66.04	3500	2.1080	0.1038
1.8016	66.98	3550	2.1155	0.0991
1.7837	67.92	3600	2.1112	0.1004
1.7027	68.87	3650	2.0888	0.0955
1.6968	69.81	3700	2.0739	0.0977
1.6873	70.75	3750	2.0948	0.0972
1.7168	71.7	3800	2.1186	0.0989
1.6195	72.64	3850	2.0967	0.0969
1.6414	73.58	3900	2.0811	0.1018
1.5118	74.53	3950	2.0674	0.0987
1.6768	75.47	4000	2.0616	0.0959
1.5945	76.42	4050	2.0632	0.1009
1.6417	77.36	4100	2.1003	0.1040
1.6208	78.3	4150	2.0939	0.1023
1.5037	79.25	4200	2.0788	0.0998
1.6181	80.19	4250	2.0641	0.0955
1.5608	81.13	4300	2.0864	0.1023
1.5658	82.08	4350	2.0802	0.1000
1.5369	83.02	4400	2.0750	0.0984
1.5474	83.96	4450	2.0582	0.0976
1.6031	84.91	4500	2.0666	0.0998
1.5224	85.85	4550	2.0695	0.0984
1.5687	86.79	4600	2.0645	0.0972
1.5393	87.74	4650	2.0702	0.0995
1.6074	88.68	4700	2.0673	0.0975
1.5601	89.62	4750	2.0622	0.0991
1.4178	90.57	4800	2.0666	0.0998
1.6219	91.51	4850	2.0620	0.1004
1.4044	92.45	4900	2.0572	0.0990
1.628	93.4	4950	2.0611	0.0993
1.5058	94.34	5000	2.0633	0.0995
1.4636	95.28	5050	2.0628	0.0998
1.5394	96.23	5100	2.0618	0.0999
1.4808	97.17	5150	2.0625	0.1004
1.5651	98.11	5200	2.0627	0.0998
1.499	99.06	5250	2.0618	0.0997
1.5463	100.0	5300	2.0619	0.0997

Framework versions

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