wav2vec2-large-xlsr-coraa-exp-6

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: 0.5587
• Wer: 0.3549
• Cer: 0.1821

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
38.8729	1.0	14	29.8196	1.0	0.9615
38.8729	2.0	28	9.8857	1.0	0.9619
38.8729	3.0	42	5.0224	1.0	0.9619
38.8729	4.0	56	4.1451	1.0	0.9619
38.8729	5.0	70	3.8275	1.0	0.9619
38.8729	6.0	84	3.6444	1.0	0.9619
38.8729	7.0	98	3.5199	1.0	0.9619
10.2228	8.0	112	3.4075	1.0	0.9619
10.2228	9.0	126	3.2890	1.0	0.9619
10.2228	10.0	140	3.2052	1.0	0.9619
10.2228	11.0	154	3.1952	1.0	0.9619
10.2228	12.0	168	3.1434	1.0	0.9619
10.2228	13.0	182	3.1563	1.0	0.9619
10.2228	14.0	196	3.1021	1.0	0.9619
3.1118	15.0	210	3.0893	1.0	0.9619
3.1118	16.0	224	3.0649	1.0	0.9619
3.1118	17.0	238	3.0773	1.0	0.9619
3.1118	18.0	252	3.0653	1.0	0.9619
3.1118	19.0	266	3.0556	1.0	0.9619
3.1118	20.0	280	3.0357	1.0	0.9619
3.1118	21.0	294	3.0212	1.0	0.9619
2.9669	22.0	308	3.0193	1.0	0.9619
2.9669	23.0	322	3.0091	1.0	0.9619
2.9669	24.0	336	3.0074	1.0	0.9619
2.9669	25.0	350	3.0022	1.0	0.9619
2.9669	26.0	364	3.0021	1.0	0.9619
2.9669	27.0	378	3.0017	1.0	0.9619
2.9669	28.0	392	2.9958	1.0	0.9619
2.9277	29.0	406	3.0000	1.0	0.9619
2.9277	30.0	420	2.9948	1.0	0.9619
2.9277	31.0	434	2.9936	1.0	0.9619
2.9277	32.0	448	2.9882	1.0	0.9619
2.9277	33.0	462	2.9866	1.0	0.9619
2.9277	34.0	476	2.9898	1.0	0.9619
2.9277	35.0	490	2.9887	1.0	0.9619
2.9186	36.0	504	2.9847	1.0	0.9619
2.9186	37.0	518	2.9852	1.0	0.9619
2.9186	38.0	532	2.9846	1.0	0.9619
2.9186	39.0	546	2.9766	1.0	0.9619
2.9186	40.0	560	2.9774	1.0	0.9619
2.9186	41.0	574	2.9617	1.0	0.9619
2.9186	42.0	588	2.9321	1.0	0.9619
2.8987	43.0	602	2.8981	1.0	0.9619
2.8987	44.0	616	2.8334	1.0	0.9619
2.8987	45.0	630	2.7951	1.0	0.9619
2.8987	46.0	644	2.7475	1.0	0.9619
2.8987	47.0	658	2.6645	1.0	0.9527
2.8987	48.0	672	2.6282	1.0	0.9541
2.8987	49.0	686	2.4763	1.0	0.8417
2.6692	50.0	700	2.2491	1.0	0.6703
2.6692	51.0	714	2.0035	1.0	0.5842
2.6692	52.0	728	1.7479	1.0	0.4960
2.6692	53.0	742	1.5238	1.0	0.4234
2.6692	54.0	756	1.3688	1.0	0.4067
2.6692	55.0	770	1.2722	0.9994	0.3967
2.6692	56.0	784	1.1672	0.9967	0.3818
2.6692	57.0	798	1.0985	0.8992	0.3311
1.6908	58.0	812	1.0185	0.8799	0.3196
1.6908	59.0	826	0.9546	0.8033	0.2906
1.6908	60.0	840	0.8950	0.6337	0.2470
1.6908	61.0	854	0.8433	0.5293	0.2234
1.6908	62.0	868	0.8039	0.4925	0.2156
1.6908	63.0	882	0.7747	0.4754	0.2117
1.6908	64.0	896	0.7777	0.4569	0.2091
0.9571	65.0	910	0.7666	0.4516	0.2074
0.9571	66.0	924	0.7772	0.4429	0.2072
0.9571	67.0	938	0.7258	0.4315	0.2026
0.9571	68.0	952	0.7159	0.4236	0.2013
0.9571	69.0	966	0.6914	0.4256	0.2006
0.9571	70.0	980	0.6768	0.4163	0.1992
0.9571	71.0	994	0.6966	0.4094	0.1981
0.6701	72.0	1008	0.6756	0.4108	0.1974
0.6701	73.0	1022	0.6746	0.4051	0.1964
0.6701	74.0	1036	0.6620	0.4007	0.1956
0.6701	75.0	1050	0.6627	0.4031	0.1957
0.6701	76.0	1064	0.6529	0.4007	0.1963
0.6701	77.0	1078	0.6478	0.3974	0.1947
0.6701	78.0	1092	0.6381	0.4017	0.1947
0.5683	79.0	1106	0.6425	0.3944	0.1935
0.5683	80.0	1120	0.6374	0.3917	0.1931
0.5683	81.0	1134	0.6219	0.3862	0.1911
0.5683	82.0	1148	0.6318	0.3854	0.1914
0.5683	83.0	1162	0.6325	0.3895	0.1933
0.5683	84.0	1176	0.6222	0.3852	0.1913
0.5683	85.0	1190	0.6149	0.3818	0.1897
0.4891	86.0	1204	0.6181	0.3805	0.1899
0.4891	87.0	1218	0.6089	0.3769	0.1889
0.4891	88.0	1232	0.6029	0.3748	0.1885
0.4891	89.0	1246	0.5954	0.3751	0.1872
0.4891	90.0	1260	0.5977	0.3755	0.1864
0.4891	91.0	1274	0.6000	0.3722	0.1873
0.4891	92.0	1288	0.5896	0.3740	0.1876
0.44	93.0	1302	0.5874	0.3781	0.1884
0.44	94.0	1316	0.5871	0.3716	0.1870
0.44	95.0	1330	0.5927	0.3740	0.1872
0.44	96.0	1344	0.6053	0.3755	0.1884
0.44	97.0	1358	0.5858	0.3718	0.1863
0.44	98.0	1372	0.5933	0.3736	0.1869
0.44	99.0	1386	0.5861	0.3722	0.1859
0.3835	100.0	1400	0.5969	0.3742	0.1872
0.3835	101.0	1414	0.5779	0.3681	0.1856
0.3835	102.0	1428	0.5938	0.3732	0.1872
0.3835	103.0	1442	0.5759	0.3663	0.1850
0.3835	104.0	1456	0.5893	0.3714	0.1876
0.3835	105.0	1470	0.5816	0.3665	0.1857
0.3835	106.0	1484	0.5775	0.3659	0.1860
0.3835	107.0	1498	0.5809	0.3704	0.1868
0.3613	108.0	1512	0.5722	0.3635	0.1854
0.3613	109.0	1526	0.5721	0.3623	0.1847
0.3613	110.0	1540	0.5774	0.3610	0.1847
0.3613	111.0	1554	0.5723	0.3631	0.1842
0.3613	112.0	1568	0.5748	0.3588	0.1837
0.3613	113.0	1582	0.5801	0.3623	0.1841
0.3613	114.0	1596	0.5773	0.3614	0.1838
0.3396	115.0	1610	0.5742	0.3623	0.1845
0.3396	116.0	1624	0.5832	0.3604	0.1848
0.3396	117.0	1638	0.5818	0.3592	0.1852
0.3396	118.0	1652	0.5700	0.3560	0.1836
0.3396	119.0	1666	0.5796	0.3608	0.1846
0.3396	120.0	1680	0.5706	0.3578	0.1837
0.3396	121.0	1694	0.5750	0.3584	0.1842
0.3327	122.0	1708	0.5764	0.3580	0.1842
0.3327	123.0	1722	0.5690	0.3551	0.1834
0.3327	124.0	1736	0.5587	0.3549	0.1821
0.3327	125.0	1750	0.5637	0.3543	0.1827
0.3327	126.0	1764	0.5634	0.3543	0.1823
0.3327	127.0	1778	0.5625	0.3531	0.1817
0.3327	128.0	1792	0.5737	0.3545	0.1826
0.3237	129.0	1806	0.5653	0.3539	0.1817
0.3237	130.0	1820	0.5671	0.3545	0.1824
0.3237	131.0	1834	0.5711	0.3549	0.1823
0.3237	132.0	1848	0.5682	0.3533	0.1819
0.3237	133.0	1862	0.5685	0.3545	0.1829
0.3237	134.0	1876	0.5662	0.3539	0.1826
0.3237	135.0	1890	0.5706	0.3535	0.1827
0.3038	136.0	1904	0.5678	0.3539	0.1821
0.3038	137.0	1918	0.5648	0.3543	0.1823
0.3038	138.0	1932	0.5638	0.3539	0.1819
0.3038	139.0	1946	0.5689	0.3541	0.1823
0.3038	140.0	1960	0.5710	0.3541	0.1825
0.3038	141.0	1974	0.5648	0.3533	0.1821
0.3038	142.0	1988	0.5656	0.3533	0.1820
0.2933	143.0	2002	0.5654	0.3539	0.1825
0.2933	144.0	2016	0.5667	0.3529	0.1828

Framework versions

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