wav2vec2-large-xlsr-coraa-exp-13

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: 0.5561
• Wer: 0.3456
• Cer: 0.1803

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
• num_epochs: 150
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
38.268	1.0	14	32.0584	1.0	0.9430
38.268	2.0	28	10.3763	1.0	0.9619
38.268	3.0	42	4.8976	1.0	0.9619
38.268	4.0	56	4.0406	1.0	0.9619
38.268	5.0	70	3.7470	1.0	0.9619
38.268	6.0	84	3.5903	1.0	0.9619
38.268	7.0	98	3.4750	1.0	0.9619
10.1654	8.0	112	3.3406	1.0	0.9619
10.1654	9.0	126	3.2267	1.0	0.9619
10.1654	10.0	140	3.1887	1.0	0.9619
10.1654	11.0	154	3.1301	1.0	0.9619
10.1654	12.0	168	3.1046	1.0	0.9619
10.1654	13.0	182	3.0909	1.0	0.9619
10.1654	14.0	196	3.0603	1.0	0.9619
3.0823	15.0	210	3.0584	1.0	0.9619
3.0823	16.0	224	3.0485	1.0	0.9619
3.0823	17.0	238	3.0464	1.0	0.9619
3.0823	18.0	252	3.0242	1.0	0.9619
3.0823	19.0	266	3.0237	1.0	0.9619
3.0823	20.0	280	3.0304	1.0	0.9619
3.0823	21.0	294	3.0119	1.0	0.9619
2.9562	22.0	308	3.0148	1.0	0.9619
2.9562	23.0	322	3.0061	1.0	0.9619
2.9562	24.0	336	3.0042	1.0	0.9619
2.9562	25.0	350	3.0033	1.0	0.9619
2.9562	26.0	364	3.0029	1.0	0.9619
2.9562	27.0	378	3.0082	1.0	0.9619
2.9562	28.0	392	2.9956	1.0	0.9619
2.9262	29.0	406	2.9948	1.0	0.9619
2.9262	30.0	420	2.9982	1.0	0.9619
2.9262	31.0	434	2.9962	1.0	0.9619
2.9262	32.0	448	2.9931	1.0	0.9619
2.9262	33.0	462	2.9809	1.0	0.9619
2.9262	34.0	476	2.9804	1.0	0.9619
2.9262	35.0	490	2.9742	1.0	0.9619
2.9125	36.0	504	2.9522	1.0	0.9619
2.9125	37.0	518	2.9015	1.0	0.9619
2.9125	38.0	532	2.8522	1.0	0.9619
2.9125	39.0	546	2.8285	1.0	0.9619
2.9125	40.0	560	2.7294	1.0	0.9615
2.9125	41.0	574	2.6491	1.0	0.9605
2.9125	42.0	588	2.4883	1.0	0.8950
2.7205	43.0	602	2.3631	1.0	0.8365
2.7205	44.0	616	2.0546	1.0	0.6074
2.7205	45.0	630	1.7867	1.0	0.5148
2.7205	46.0	644	1.5453	1.0	0.4532
2.7205	47.0	658	1.3554	0.9990	0.4064
2.7205	48.0	672	1.2016	0.9829	0.3670
2.7205	49.0	686	1.0777	0.8805	0.3167
1.6469	50.0	700	0.9790	0.7030	0.2594
1.6469	51.0	714	0.8962	0.5270	0.2224
1.6469	52.0	728	0.8429	0.4974	0.2176
1.6469	53.0	742	0.8159	0.4659	0.2089
1.6469	54.0	756	0.7980	0.4512	0.2066
1.6469	55.0	770	0.7541	0.4441	0.2044
1.6469	56.0	784	0.7299	0.4273	0.2015
1.6469	57.0	798	0.7078	0.4092	0.1964
0.7997	58.0	812	0.7079	0.4110	0.1973
0.7997	59.0	826	0.6861	0.4137	0.1983
0.7997	60.0	840	0.7035	0.4011	0.1975
0.7997	61.0	854	0.6676	0.4000	0.1942
0.7997	62.0	868	0.6562	0.3980	0.1937
0.7997	63.0	882	0.6580	0.3850	0.1911
0.7997	64.0	896	0.6643	0.3911	0.1925
0.5379	65.0	910	0.6532	0.3929	0.1928
0.5379	66.0	924	0.6483	0.3866	0.1906
0.5379	67.0	938	0.6267	0.3757	0.1870
0.5379	68.0	952	0.6296	0.3793	0.1880
0.5379	69.0	966	0.6415	0.3785	0.1902
0.5379	70.0	980	0.6227	0.3746	0.1885
0.5379	71.0	994	0.6213	0.3738	0.1878
0.4372	72.0	1008	0.6110	0.3726	0.1872
0.4372	73.0	1022	0.6019	0.3696	0.1862
0.4372	74.0	1036	0.6037	0.3722	0.1867
0.4372	75.0	1050	0.5994	0.3657	0.1881
0.4372	76.0	1064	0.6083	0.3704	0.1881
0.4372	77.0	1078	0.5838	0.3696	0.1865
0.4372	78.0	1092	0.5795	0.3718	0.1855
0.3912	79.0	1106	0.6201	0.3714	0.1877
0.3912	80.0	1120	0.5915	0.3661	0.1854
0.3912	81.0	1134	0.5894	0.3651	0.1843
0.3912	82.0	1148	0.5994	0.3681	0.1859
0.3912	83.0	1162	0.6001	0.3655	0.1864
0.3912	84.0	1176	0.6008	0.3653	0.1865
0.3912	85.0	1190	0.5770	0.3602	0.1832
0.3485	86.0	1204	0.5905	0.3566	0.1836
0.3485	87.0	1218	0.5810	0.3580	0.1828
0.3485	88.0	1232	0.5765	0.3584	0.1830
0.3485	89.0	1246	0.5902	0.3641	0.1845
0.3485	90.0	1260	0.5812	0.3614	0.1831
0.3485	91.0	1274	0.5966	0.3586	0.1844
0.3485	92.0	1288	0.5686	0.3557	0.1822
0.3234	93.0	1302	0.5839	0.3553	0.1828
0.3234	94.0	1316	0.5765	0.3553	0.1820
0.3234	95.0	1330	0.5780	0.3566	0.1820
0.3234	96.0	1344	0.5862	0.3596	0.1834
0.3234	97.0	1358	0.5702	0.3555	0.1821
0.3234	98.0	1372	0.5787	0.3547	0.1821
0.3234	99.0	1386	0.5767	0.3531	0.1824
0.2803	100.0	1400	0.5778	0.3570	0.1818
0.2803	101.0	1414	0.5759	0.3543	0.1817
0.2803	102.0	1428	0.5838	0.3572	0.1824
0.2803	103.0	1442	0.5696	0.3541	0.1815
0.2803	104.0	1456	0.5724	0.3541	0.1820
0.2803	105.0	1470	0.5698	0.3543	0.1820
0.2803	106.0	1484	0.5727	0.3523	0.1816
0.2803	107.0	1498	0.5609	0.3511	0.1809
0.2718	108.0	1512	0.5655	0.3497	0.1807
0.2718	109.0	1526	0.5761	0.3535	0.1816
0.2718	110.0	1540	0.5753	0.3523	0.1815
0.2718	111.0	1554	0.5703	0.3503	0.1805
0.2718	112.0	1568	0.5623	0.3470	0.1802
0.2718	113.0	1582	0.5723	0.3511	0.1813
0.2718	114.0	1596	0.5608	0.3486	0.1803
0.2614	115.0	1610	0.5613	0.3511	0.1809
0.2614	116.0	1624	0.5742	0.3533	0.1817
0.2614	117.0	1638	0.5715	0.3523	0.1817
0.2614	118.0	1652	0.5695	0.3533	0.1817
0.2614	119.0	1666	0.5713	0.3531	0.1825
0.2614	120.0	1680	0.5664	0.3533	0.1821
0.2614	121.0	1694	0.5716	0.3531	0.1822
0.2463	122.0	1708	0.5680	0.3476	0.1810
0.2463	123.0	1722	0.5760	0.3527	0.1817
0.2463	124.0	1736	0.5561	0.3456	0.1803
0.2463	125.0	1750	0.5698	0.3478	0.1812
0.2463	126.0	1764	0.5667	0.3482	0.1811
0.2463	127.0	1778	0.5677	0.3478	0.1813
0.2463	128.0	1792	0.5681	0.3446	0.1805
0.2477	129.0	1806	0.5666	0.3470	0.1809
0.2477	130.0	1820	0.5696	0.3458	0.1804
0.2477	131.0	1834	0.5704	0.3478	0.1810
0.2477	132.0	1848	0.5656	0.3470	0.1808
0.2477	133.0	1862	0.5697	0.3472	0.1807
0.2477	134.0	1876	0.5716	0.3472	0.1810
0.2477	135.0	1890	0.5742	0.3484	0.1810
0.221	136.0	1904	0.5671	0.3472	0.1807
0.221	137.0	1918	0.5670	0.3462	0.1810
0.221	138.0	1932	0.5675	0.3460	0.1810
0.221	139.0	1946	0.5704	0.3462	0.1810
0.221	140.0	1960	0.5675	0.3458	0.1808
0.221	141.0	1974	0.5618	0.3444	0.1800
0.221	142.0	1988	0.5633	0.3454	0.1800
0.2217	143.0	2002	0.5664	0.3456	0.1807
0.2217	144.0	2016	0.5682	0.3462	0.1810

Framework versions

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