wav2vec2-large-xlsr-coraa-exp-15

This model is a fine-tuned version of Edresson/wav2vec2-large-xlsr-coraa-portuguese on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5572
• Wer: 0.3509
• Cer: 0.1800
• Per: 0.3419

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.01
• num_epochs: 150
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Per
38.389	1.0	14	41.3800	1.2235	1.2272	1.2229
38.389	2.0	28	9.4651	1.0	0.9619	1.0
38.389	3.0	42	4.3683	1.0	0.9619	1.0
38.389	4.0	56	3.7985	1.0	0.9619	1.0
38.389	5.0	70	3.5499	1.0	0.9619	1.0
38.389	6.0	84	3.3330	1.0	0.9619	1.0
38.389	7.0	98	3.2145	1.0	0.9619	1.0
11.0275	8.0	112	3.1199	1.0	0.9619	1.0
11.0275	9.0	126	3.0737	1.0	0.9619	1.0
11.0275	10.0	140	3.0931	1.0	0.9619	1.0
11.0275	11.0	154	3.0363	1.0	0.9619	1.0
11.0275	12.0	168	3.0763	1.0	0.9619	1.0
11.0275	13.0	182	3.0298	1.0	0.9619	1.0
11.0275	14.0	196	3.0064	1.0	0.9619	1.0
2.9809	15.0	210	3.0108	1.0	0.9619	1.0
2.9809	16.0	224	3.0049	1.0	0.9619	1.0
2.9809	17.0	238	3.0058	1.0	0.9619	1.0
2.9809	18.0	252	3.0002	1.0	0.9619	1.0
2.9809	19.0	266	2.9971	1.0	0.9619	1.0
2.9809	20.0	280	3.0035	1.0	0.9619	1.0
2.9809	21.0	294	2.9971	1.0	0.9619	1.0
2.9263	22.0	308	2.9916	1.0	0.9619	1.0
2.9263	23.0	322	2.9800	1.0	0.9619	1.0
2.9263	24.0	336	2.9640	1.0	0.9619	1.0
2.9263	25.0	350	2.8945	1.0	0.9619	1.0
2.9263	26.0	364	2.7927	1.0	0.9619	1.0
2.9263	27.0	378	2.6844	1.0	0.9589	1.0
2.9263	28.0	392	2.4282	1.0	0.8397	1.0
2.7713	29.0	406	1.9364	1.0	0.5755	1.0
2.7713	30.0	420	1.4716	1.0	0.4155	1.0
2.7713	31.0	434	1.1548	0.9970	0.3730	0.9970
2.7713	32.0	448	0.9997	0.9726	0.3450	0.9707
2.7713	33.0	462	0.9343	0.5624	0.2351	0.5414
2.7713	34.0	476	0.7747	0.4933	0.2163	0.4695
2.7713	35.0	490	0.7341	0.4307	0.2048	0.4094
1.1643	36.0	504	0.6947	0.4102	0.1987	0.3913
1.1643	37.0	518	0.7270	0.4273	0.2054	0.4082
1.1643	38.0	532	0.6677	0.4092	0.1979	0.3941
1.1643	39.0	546	0.6832	0.4130	0.1993	0.3986
1.1643	40.0	560	0.6691	0.3994	0.1978	0.3846
1.1643	41.0	574	0.6585	0.3941	0.1962	0.3818
1.1643	42.0	588	0.6199	0.3992	0.1942	0.3856
0.493	43.0	602	0.6268	0.3970	0.1929	0.3854
0.493	44.0	616	0.5925	0.3842	0.1902	0.3679
0.493	45.0	630	0.6071	0.3797	0.1908	0.3645
0.493	46.0	644	0.6037	0.3840	0.1902	0.3690
0.493	47.0	658	0.5819	0.3728	0.1879	0.3576
0.493	48.0	672	0.5930	0.3671	0.1862	0.3509
0.493	49.0	686	0.6292	0.3748	0.1914	0.3598
0.3417	50.0	700	0.6298	0.3777	0.1908	0.3623
0.3417	51.0	714	0.6183	0.3677	0.1881	0.3539
0.3417	52.0	728	0.6200	0.3728	0.1891	0.3590
0.3417	53.0	742	0.6202	0.3681	0.1883	0.3539
0.3417	54.0	756	0.5683	0.3663	0.1842	0.3537
0.3417	55.0	770	0.5832	0.3625	0.1849	0.3501
0.3417	56.0	784	0.5890	0.3655	0.1846	0.3529
0.3417	57.0	798	0.5770	0.3663	0.1843	0.3547
0.2811	58.0	812	0.5655	0.3606	0.1844	0.3488
0.2811	59.0	826	0.5698	0.3553	0.1818	0.3458
0.2811	60.0	840	0.5964	0.3633	0.1849	0.3545
0.2811	61.0	854	0.5919	0.3608	0.1859	0.3533
0.2811	62.0	868	0.5771	0.3549	0.1826	0.3472
0.2811	63.0	882	0.6055	0.3555	0.1858	0.3476
0.2811	64.0	896	0.5833	0.3588	0.1832	0.3488
0.2169	65.0	910	0.5864	0.3582	0.1841	0.3490
0.2169	66.0	924	0.5833	0.3578	0.1830	0.3490
0.2169	67.0	938	0.5663	0.3531	0.1811	0.3442
0.2169	68.0	952	0.5572	0.3509	0.1800	0.3419
0.2169	69.0	966	0.5641	0.3533	0.1800	0.3448
0.2169	70.0	980	0.5700	0.3525	0.1790	0.3446
0.2169	71.0	994	0.5835	0.3549	0.1813	0.3468
0.1834	72.0	1008	0.5718	0.3499	0.1811	0.3417
0.1834	73.0	1022	0.5938	0.3539	0.1837	0.3454
0.1834	74.0	1036	0.5955	0.3560	0.1833	0.3466
0.1834	75.0	1050	0.5658	0.3519	0.1819	0.3438
0.1834	76.0	1064	0.5671	0.3497	0.1806	0.3415
0.1834	77.0	1078	0.5772	0.3541	0.1827	0.3454
0.1834	78.0	1092	0.5744	0.3507	0.1806	0.3442
0.1877	79.0	1106	0.5727	0.3476	0.1796	0.3403
0.1877	80.0	1120	0.5696	0.3468	0.1797	0.3393
0.1877	81.0	1134	0.5846	0.3482	0.1805	0.3395
0.1877	82.0	1148	0.5943	0.3511	0.1799	0.3432
0.1877	83.0	1162	0.5738	0.3456	0.1791	0.3389
0.1877	84.0	1176	0.6163	0.3553	0.1834	0.3478
0.1877	85.0	1190	0.5756	0.3488	0.1801	0.3417
0.1684	86.0	1204	0.6208	0.3547	0.1831	0.3480
0.1684	87.0	1218	0.5924	0.3484	0.1811	0.3413
0.1684	88.0	1232	0.5953	0.3509	0.1813	0.3436

Framework versions

• Transformers 4.28.0
• Pytorch 2.4.1+cu121
• Datasets 3.2.0
• Tokenizers 0.13.3