wav2vec2-large-xlsr-coraa-exp-14

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: 0.5485
• Wer: 0.3417
• Cer: 0.1776
• Per: 0.3310

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: 3e-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: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 150
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Per
37.6216	1.0	14	46.2465	1.0071	4.0178	1.0071
37.6216	2.0	28	44.1710	1.0041	2.7949	1.0039
37.6216	3.0	42	39.3999	1.0	0.9284	1.0
37.6216	4.0	56	31.8074	1.0	0.9589	1.0
37.6216	5.0	70	20.4155	1.0	0.9619	1.0
37.6216	6.0	84	9.6948	1.0	0.9619	1.0
37.6216	7.0	98	5.8283	1.0	0.9619	1.0
25.7765	8.0	112	4.5548	1.0	0.9619	1.0
25.7765	9.0	126	4.0980	1.0	0.9619	1.0
25.7765	10.0	140	3.8329	1.0	0.9619	1.0
25.7765	11.0	154	3.6665	1.0	0.9619	1.0
25.7765	12.0	168	3.5348	1.0	0.9619	1.0
25.7765	13.0	182	3.3872	1.0	0.9619	1.0
25.7765	14.0	196	3.2821	1.0	0.9619	1.0
3.8439	15.0	210	3.2034	1.0	0.9619	1.0
3.8439	16.0	224	3.1549	1.0	0.9619	1.0
3.8439	17.0	238	3.1379	1.0	0.9619	1.0
3.8439	18.0	252	3.0903	1.0	0.9619	1.0
3.8439	19.0	266	3.0685	1.0	0.9619	1.0
3.8439	20.0	280	3.0650	1.0	0.9619	1.0
3.8439	21.0	294	3.0469	1.0	0.9619	1.0
3.035	22.0	308	3.0442	1.0	0.9619	1.0
3.035	23.0	322	3.0488	1.0	0.9619	1.0
3.035	24.0	336	3.0284	1.0	0.9619	1.0
3.035	25.0	350	3.0219	1.0	0.9619	1.0
3.035	26.0	364	3.0185	1.0	0.9619	1.0
3.035	27.0	378	3.0074	1.0	0.9619	1.0
3.035	28.0	392	3.0130	1.0	0.9619	1.0
2.9429	29.0	406	3.0014	1.0	0.9619	1.0
2.9429	30.0	420	2.9969	1.0	0.9619	1.0
2.9429	31.0	434	3.0056	1.0	0.9619	1.0
2.9429	32.0	448	3.0042	1.0	0.9619	1.0
2.9429	33.0	462	2.9842	1.0	0.9619	1.0
2.9429	34.0	476	2.9850	1.0	0.9619	1.0
2.9429	35.0	490	2.9796	1.0	0.9619	1.0
2.9201	36.0	504	2.9647	1.0	0.9619	1.0
2.9201	37.0	518	2.9327	1.0	0.9619	1.0
2.9201	38.0	532	2.8997	1.0	0.9619	1.0
2.9201	39.0	546	2.8702	1.0	0.9619	1.0
2.9201	40.0	560	2.8146	1.0	0.9619	1.0
2.9201	41.0	574	2.6820	1.0	0.9600	1.0
2.9201	42.0	588	2.5411	1.0	0.9084	1.0
2.7676	43.0	602	2.3459	1.0	0.7805	1.0
2.7676	44.0	616	2.0668	1.0	0.5817	1.0
2.7676	45.0	630	1.7420	1.0	0.4851	1.0
2.7676	46.0	644	1.4977	1.0	0.4356	1.0
2.7676	47.0	658	1.2985	0.9986	0.3912	0.9986
2.7676	48.0	672	1.1510	0.9441	0.3408	0.9403
2.7676	49.0	686	1.0227	0.7387	0.2760	0.7249
1.6224	50.0	700	0.9463	0.5246	0.2268	0.4965
1.6224	51.0	714	0.8637	0.4951	0.2179	0.4640
1.6224	52.0	728	0.8114	0.4732	0.2128	0.4449
1.6224	53.0	742	0.7687	0.4388	0.2038	0.4102
1.6224	54.0	756	0.7736	0.4340	0.2035	0.4063
1.6224	55.0	770	0.7564	0.4275	0.2018	0.4019
1.6224	56.0	784	0.7419	0.4210	0.2015	0.3988
1.6224	57.0	798	0.7089	0.4074	0.1982	0.3856
0.7727	58.0	812	0.6822	0.4061	0.1962	0.3866
0.7727	59.0	826	0.6505	0.4067	0.1952	0.3866
0.7727	60.0	840	0.6542	0.3933	0.1938	0.3767
0.7727	61.0	854	0.6346	0.3958	0.1926	0.3789
0.7727	62.0	868	0.6301	0.3885	0.1912	0.3722
0.7727	63.0	882	0.6702	0.3944	0.1944	0.3755
0.7727	64.0	896	0.6394	0.3846	0.1905	0.3675
0.5159	65.0	910	0.6235	0.3775	0.1892	0.3606
0.5159	66.0	924	0.6329	0.3795	0.1911	0.3633
0.5159	67.0	938	0.6074	0.3732	0.1891	0.3582
0.5159	68.0	952	0.5993	0.3742	0.1876	0.3592
0.5159	69.0	966	0.6088	0.3663	0.1871	0.3513
0.5159	70.0	980	0.6132	0.3771	0.1888	0.3606
0.5159	71.0	994	0.6205	0.3779	0.1892	0.3616
0.41	72.0	1008	0.6048	0.3732	0.1885	0.3562
0.41	73.0	1022	0.5851	0.3700	0.1873	0.3543
0.41	74.0	1036	0.5975	0.3706	0.1872	0.3543
0.41	75.0	1050	0.5996	0.3722	0.1890	0.3555
0.41	76.0	1064	0.5951	0.3690	0.1874	0.3533
0.41	77.0	1078	0.5791	0.3637	0.1857	0.3486
0.41	78.0	1092	0.5675	0.3616	0.1851	0.3466
0.371	79.0	1106	0.6022	0.3659	0.1880	0.3486
0.371	80.0	1120	0.5954	0.3669	0.1854	0.3507
0.371	81.0	1134	0.5832	0.3629	0.1841	0.3470
0.371	82.0	1148	0.5867	0.3620	0.1843	0.3456
0.371	83.0	1162	0.5971	0.3669	0.1870	0.3519
0.371	84.0	1176	0.5926	0.3633	0.1859	0.3478
0.371	85.0	1190	0.5774	0.3596	0.1837	0.3438
0.3317	86.0	1204	0.5779	0.3610	0.1846	0.3462
0.3317	87.0	1218	0.5797	0.3604	0.1845	0.3446
0.3317	88.0	1232	0.5750	0.3578	0.1843	0.3425
0.3317	89.0	1246	0.5651	0.3584	0.1824	0.3436
0.3317	90.0	1260	0.5749	0.3592	0.1830	0.3427
0.3317	91.0	1274	0.5791	0.3572	0.1834	0.3419
0.3317	92.0	1288	0.5589	0.3541	0.1814	0.3389
0.3039	93.0	1302	0.5670	0.3543	0.1816	0.3387
0.3039	94.0	1316	0.5619	0.3521	0.1805	0.3385
0.3039	95.0	1330	0.5628	0.3539	0.1801	0.3393
0.3039	96.0	1344	0.5800	0.3572	0.1820	0.3417
0.3039	97.0	1358	0.5605	0.3509	0.1805	0.3371
0.3039	98.0	1372	0.5619	0.3513	0.1805	0.3369
0.3039	99.0	1386	0.5704	0.3523	0.1825	0.3373
0.2596	100.0	1400	0.5618	0.3531	0.1810	0.3393
0.2596	101.0	1414	0.5591	0.3458	0.1799	0.3310
0.2596	102.0	1428	0.5675	0.3492	0.1817	0.3352
0.2596	103.0	1442	0.5614	0.3537	0.1808	0.3397
0.2596	104.0	1456	0.5652	0.3527	0.1810	0.3389
0.2596	105.0	1470	0.5576	0.3497	0.1798	0.3354
0.2596	106.0	1484	0.5653	0.3499	0.1794	0.3360
0.2596	107.0	1498	0.5543	0.3474	0.1796	0.3324
0.2488	108.0	1512	0.5540	0.3484	0.1794	0.3344
0.2488	109.0	1526	0.5626	0.3484	0.1804	0.3346
0.2488	110.0	1540	0.5648	0.3499	0.1809	0.3352
0.2488	111.0	1554	0.5588	0.3495	0.1803	0.3358
0.2488	112.0	1568	0.5574	0.3466	0.1790	0.3338
0.2488	113.0	1582	0.5624	0.3486	0.1798	0.3358
0.2488	114.0	1596	0.5538	0.3488	0.1791	0.3348
0.2409	115.0	1610	0.5577	0.3474	0.1790	0.3332
0.2409	116.0	1624	0.5612	0.3472	0.1793	0.3322
0.2409	117.0	1638	0.5640	0.3482	0.1797	0.3334
0.2409	118.0	1652	0.5598	0.3484	0.1791	0.3346
0.2409	119.0	1666	0.5660	0.3468	0.1792	0.3334
0.2409	120.0	1680	0.5578	0.3460	0.1784	0.3328
0.2409	121.0	1694	0.5592	0.3466	0.1790	0.3338
0.2278	122.0	1708	0.5624	0.3456	0.1788	0.3332
0.2278	123.0	1722	0.5639	0.3436	0.1789	0.3308
0.2278	124.0	1736	0.5486	0.3438	0.1781	0.3320
0.2278	125.0	1750	0.5649	0.3417	0.1790	0.3291
0.2278	126.0	1764	0.5518	0.3425	0.1783	0.3302
0.2278	127.0	1778	0.5538	0.3415	0.1780	0.3295
0.2278	128.0	1792	0.5591	0.3413	0.1779	0.3302
0.2261	129.0	1806	0.5575	0.3417	0.1780	0.3295
0.2261	130.0	1820	0.5548	0.3403	0.1778	0.3287
0.2261	131.0	1834	0.5608	0.3401	0.1781	0.3289
0.2261	132.0	1848	0.5485	0.3417	0.1776	0.3310
0.2261	133.0	1862	0.5508	0.3393	0.1776	0.3281
0.2261	134.0	1876	0.5572	0.3383	0.1777	0.3269
0.2261	135.0	1890	0.5605	0.3389	0.1783	0.3269
0.2015	136.0	1904	0.5549	0.3387	0.1776	0.3271
0.2015	137.0	1918	0.5534	0.3399	0.1775	0.3281
0.2015	138.0	1932	0.5525	0.3385	0.1775	0.3271
0.2015	139.0	1946	0.5551	0.3383	0.1774	0.3269
0.2015	140.0	1960	0.5531	0.3375	0.1770	0.3259
0.2015	141.0	1974	0.5504	0.3354	0.1765	0.3236
0.2015	142.0	1988	0.5518	0.3379	0.1770	0.3259
0.2036	143.0	2002	0.5525	0.3377	0.1770	0.3255
0.2036	144.0	2016	0.5528	0.3383	0.1772	0.3269
0.2036	145.0	2030	0.5534	0.3379	0.1773	0.3263
0.2036	146.0	2044	0.5548	0.3373	0.1774	0.3259
0.2036	147.0	2058	0.5542	0.3364	0.1771	0.3249
0.2036	148.0	2072	0.5550	0.3373	0.1772	0.3257
0.2036	149.0	2086	0.5555	0.3373	0.1772	0.3257
0.2013	150.0	2100	0.5558	0.3383	0.1774	0.3265

Framework versions

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