fine-w2v2base-bs16-ep100-lr2e-05-linguistic-rmsnorm-focal_ctc_a0.5_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: 2.0217
• Wer: 0.1027

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
1087.6127	0.94	50	544.0147	15.9400
883.6459	1.89	100	263.3972	1.2869
154.8555	2.83	150	46.7449	1.0
56.6812	3.77	200	42.9581	1.0
54.5616	4.72	250	41.7305	1.0
52.6183	5.66	300	40.4653	1.0
51.0696	6.6	350	39.4718	1.0
49.2717	7.55	400	38.6904	1.0
48.4209	8.49	450	38.1454	1.0
48.9467	9.43	500	37.8410	1.0
48.2942	10.38	550	37.5333	1.0
48.4695	11.32	600	37.3824	1.0
43.8977	12.26	650	27.3106	0.7221
28.474	13.21	700	13.0357	0.3541
16.5311	14.15	750	7.8864	0.2232
11.7013	15.09	800	5.9173	0.1789
9.495	16.04	850	4.9400	0.1665
8.0993	16.98	900	4.2435	0.1529
7.0221	17.92	950	3.7539	0.1342
6.4537	18.87	1000	3.5368	0.1375
6.2301	19.81	1050	3.5333	0.1416
5.7669	20.75	1100	3.1940	0.1263
5.4415	21.7	1150	2.9780	0.1260
5.1487	22.64	1200	2.8402	0.1238
4.7512	23.58	1250	2.7171	0.1174
4.4345	24.53	1300	2.6491	0.1089
4.5031	25.47	1350	2.5340	0.1143
4.1201	26.42	1400	2.5977	0.1204
4.1617	27.36	1450	2.4961	0.1209
3.8342	28.3	1500	2.3904	0.1123
3.8148	29.25	1550	2.4308	0.1169
3.6714	30.19	1600	2.3569	0.1108
3.4776	31.13	1650	2.3612	0.1158
3.5231	32.08	1700	2.2809	0.1065
3.396	33.02	1750	2.2409	0.1043
3.2426	33.96	1800	2.2213	0.1009
3.0984	34.91	1850	2.2880	0.1131
3.1632	35.85	1900	2.3351	0.1122
3.1377	36.79	1950	2.2338	0.1046
2.9159	37.74	2000	2.2138	0.1015
2.9623	38.68	2050	2.1678	0.1003
2.8043	39.62	2100	2.1545	0.1056
2.7835	40.57	2150	2.1060	0.1025
2.6223	41.51	2200	2.1277	0.1041
2.6276	42.45	2250	2.1640	0.1099
2.5713	43.4	2300	2.1470	0.1088
2.5732	44.34	2350	2.1019	0.1064
2.5823	45.28	2400	2.0942	0.1104
2.401	46.23	2450	2.1207	0.1038
2.3553	47.17	2500	2.0486	0.1027
2.2568	48.11	2550	2.0719	0.1019
2.3041	49.06	2600	2.1119	0.1054
2.1967	50.0	2650	2.0949	0.1047
2.1611	50.94	2700	2.0584	0.0992
2.2721	51.89	2750	2.0706	0.1034
1.9844	52.83	2800	2.0582	0.1079
2.1597	53.77	2850	2.0510	0.1054
2.0874	54.72	2900	2.0830	0.1075
1.968	55.66	2950	2.0899	0.1078
1.9349	56.6	3000	2.0793	0.1022
2.0729	57.55	3050	2.0744	0.1026
2.0062	58.49	3100	2.0859	0.1083
1.9635	59.43	3150	2.0448	0.1011
1.9711	60.38	3200	2.1105	0.1045
1.7753	61.32	3250	2.0405	0.1004
1.918	62.26	3300	2.0738	0.1047
1.737	63.21	3350	2.0535	0.1060
1.861	64.15	3400	2.0935	0.1029
1.7855	65.09	3450	2.0630	0.1041
1.7638	66.04	3500	2.0319	0.1058
1.7905	66.98	3550	2.0325	0.1049
1.8109	67.92	3600	2.0527	0.1073
1.7491	68.87	3650	2.0453	0.1074
1.778	69.81	3700	2.0238	0.1034
1.7323	70.75	3750	2.0391	0.1078
1.7734	71.7	3800	2.0206	0.1061
1.6741	72.64	3850	2.0337	0.1057
1.6221	73.58	3900	2.0290	0.1068
1.5371	74.53	3950	2.0296	0.1039
1.7238	75.47	4000	2.0393	0.1063
1.6034	76.42	4050	2.0171	0.1053
1.6784	77.36	4100	2.0421	0.1073
1.6036	78.3	4150	2.0327	0.1049
1.5265	79.25	4200	2.0292	0.1048
1.6041	80.19	4250	2.0262	0.1029
1.5758	81.13	4300	2.0568	0.1063
1.5859	82.08	4350	2.0255	0.1044
1.5839	83.02	4400	2.0236	0.1028
1.5533	83.96	4450	2.0242	0.1039
1.6382	84.91	4500	2.0066	0.1049
1.5394	85.85	4550	2.0174	0.1039
1.6017	86.79	4600	2.0138	0.1014
1.567	87.74	4650	2.0214	0.1049
1.5778	88.68	4700	2.0198	0.1038
1.5495	89.62	4750	2.0208	0.1042
1.4151	90.57	4800	2.0247	0.1039
1.6654	91.51	4850	2.0195	0.1036
1.4209	92.45	4900	2.0172	0.1037
1.6416	93.4	4950	2.0206	0.1031
1.5105	94.34	5000	2.0220	0.1034
1.4753	95.28	5050	2.0204	0.1030
1.5578	96.23	5100	2.0227	0.1034
1.5149	97.17	5150	2.0222	0.1031
1.557	98.11	5200	2.0224	0.1029
1.5528	99.06	5250	2.0216	0.1027
1.5551	100.0	5300	2.0217	0.1027

Framework versions

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