fine-w2v2base-bs16-ep100-lr2e-05-linguistic-rmsnorm-focal_ctc_a0.99_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: 3.9977
• Wer: 0.0897

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
2069.4439	0.94	50	995.1141	12.4711
1363.7322	1.89	100	368.4300	0.9986
280.6324	2.83	150	95.5134	1.0
113.4733	3.77	200	82.3901	1.0
107.4746	4.72	250	80.1978	1.0
103.9621	5.66	300	77.6156	1.0
101.1517	6.6	350	75.8253	1.0
97.3946	7.55	400	74.4834	1.0
95.6127	8.49	450	73.5984	1.0
96.617	9.43	500	73.0784	1.0
95.0988	10.38	550	72.0475	1.0
91.6734	11.32	600	56.5232	0.7296
61.804	12.26	650	31.1295	0.4563
36.8773	13.21	700	17.1019	0.2573
24.3159	14.15	750	12.2619	0.1894
18.8062	15.09	800	9.9440	0.1692
16.0671	16.04	850	8.5476	0.1489
14.1267	16.98	900	7.8480	0.1537
12.5625	17.92	950	7.0957	0.1342
11.8023	18.87	1000	6.7007	0.1328
10.925	19.81	1050	6.3680	0.1286
10.1188	20.75	1100	6.0900	0.1204
9.9066	21.7	1150	5.8583	0.1236
9.299	22.64	1200	5.6292	0.1191
8.8681	23.58	1250	5.5766	0.1184
8.382	24.53	1300	5.5799	0.1159
8.4245	25.47	1350	5.2661	0.1140
7.7502	26.42	1400	5.1468	0.1096
7.8289	27.36	1450	5.0905	0.1088
7.729	28.3	1500	5.0458	0.1138
7.5606	29.25	1550	4.9833	0.1082
7.095	30.19	1600	4.8245	0.1023
6.6926	31.13	1650	4.7635	0.1013
7.1713	32.08	1700	4.8290	0.1019
6.9475	33.02	1750	4.7296	0.1051
6.4841	33.96	1800	4.6856	0.0998
6.1089	34.91	1850	4.6471	0.1053
6.1785	35.85	1900	4.6827	0.1005
6.2511	36.79	1950	4.6419	0.1024
6.0869	37.74	2000	4.6392	0.1082
5.9902	38.68	2050	4.4272	0.1044
5.3991	39.62	2100	4.4101	0.1037
5.3874	40.57	2150	4.2949	0.1004
5.1402	41.51	2200	4.3465	0.1024
5.1982	42.45	2250	4.3083	0.1066
4.9461	43.4	2300	4.3431	0.0967
4.9867	44.34	2350	4.2870	0.0981
4.929	45.28	2400	4.1885	0.0988
4.6383	46.23	2450	4.2757	0.1025
4.634	47.17	2500	4.1286	0.0998
4.4744	48.11	2550	4.1603	0.0963
4.5322	49.06	2600	4.2124	0.1020
4.4657	50.0	2650	4.2822	0.1023
4.457	50.94	2700	4.2232	0.0947
4.5771	51.89	2750	4.1734	0.0939
4.0174	52.83	2800	4.2091	0.0976
4.4103	53.77	2850	4.1129	0.0954
4.2772	54.72	2900	4.2312	0.0966
3.9499	55.66	2950	4.2480	0.1010
3.9354	56.6	3000	4.1807	0.1013
4.1267	57.55	3050	4.1947	0.0988
4.0414	58.49	3100	4.1440	0.1031
3.8629	59.43	3150	4.1198	0.0988
3.9139	60.38	3200	4.1225	0.0992
3.5162	61.32	3250	4.1204	0.0974
3.8449	62.26	3300	4.1679	0.0948
3.5105	63.21	3350	4.1508	0.0967
3.6462	64.15	3400	4.1263	0.0989
3.4621	65.09	3450	4.0731	0.0933
3.5294	66.04	3500	4.0631	0.0922
3.4527	66.98	3550	4.0846	0.0938
3.4661	67.92	3600	4.0970	0.0945
3.3707	68.87	3650	4.1004	0.0925
3.4103	69.81	3700	4.0616	0.0913
3.3892	70.75	3750	4.0689	0.0945
3.4639	71.7	3800	4.1037	0.0976
3.2201	72.64	3850	4.0952	0.0948
3.2162	73.58	3900	4.0467	0.0938
2.9934	74.53	3950	4.0333	0.0908
3.3083	75.47	4000	4.0238	0.0896
3.1558	76.42	4050	4.0116	0.0908
3.2886	77.36	4100	4.0221	0.0921
3.1983	78.3	4150	4.0320	0.0908
2.9773	79.25	4200	3.9992	0.0913
3.1754	80.19	4250	3.9902	0.0910
3.1124	81.13	4300	3.9905	0.0914
3.0626	82.08	4350	4.0065	0.0911
3.0676	83.02	4400	3.9876	0.0908
3.0152	83.96	4450	3.9895	0.0910
3.2035	84.91	4500	3.9913	0.0907
3.0765	85.85	4550	4.0102	0.0929
3.1136	86.79	4600	3.9986	0.0922
3.0455	87.74	4650	4.0120	0.0934
3.1063	88.68	4700	3.9956	0.0916
3.0908	89.62	4750	3.9880	0.0904
2.7746	90.57	4800	4.0073	0.0893
3.3105	91.51	4850	4.0035	0.0893
2.7955	92.45	4900	3.9969	0.0897
3.2318	93.4	4950	3.9973	0.0900
3.0095	94.34	5000	3.9992	0.0901
2.9096	95.28	5050	3.9995	0.0899
3.0568	96.23	5100	3.9999	0.0899
2.9639	97.17	5150	3.9983	0.0898
3.0901	98.11	5200	3.9998	0.0901
3.0161	99.06	5250	3.9981	0.0898
3.0623	100.0	5300	3.9977	0.0897

Framework versions

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