wav2vec2-large-xlsr-coraa-exp-9

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: 0.5553
• Wer: 0.3466
• Cer: 0.1788

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
37.5508	1.0	14	23.1376	1.0	0.9619
37.5508	2.0	28	6.5036	1.0	0.9619
37.5508	3.0	42	4.3919	1.0	0.9619
37.5508	4.0	56	3.9441	1.0	0.9619
37.5508	5.0	70	3.7306	1.0	0.9619
37.5508	6.0	84	3.5762	1.0	0.9619
37.5508	7.0	98	3.4129	1.0	0.9619
8.6902	8.0	112	3.2859	1.0	0.9619
8.6902	9.0	126	3.2192	1.0	0.9619
8.6902	10.0	140	3.1479	1.0	0.9619
8.6902	11.0	154	3.1063	1.0	0.9619
8.6902	12.0	168	3.0897	1.0	0.9619
8.6902	13.0	182	3.0849	1.0	0.9619
8.6902	14.0	196	3.0485	1.0	0.9619
3.059	15.0	210	3.0496	1.0	0.9619
3.059	16.0	224	3.0510	1.0	0.9619
3.059	17.0	238	3.0428	1.0	0.9619
3.059	18.0	252	3.0331	1.0	0.9619
3.059	19.0	266	3.0353	1.0	0.9619
3.059	20.0	280	3.0217	1.0	0.9619
3.059	21.0	294	3.0107	1.0	0.9619
2.9492	22.0	308	3.0068	1.0	0.9619
2.9492	23.0	322	2.9950	1.0	0.9619
2.9492	24.0	336	2.9896	1.0	0.9619
2.9492	25.0	350	2.9687	1.0	0.9619
2.9492	26.0	364	2.9474	1.0	0.9619
2.9492	27.0	378	2.9414	1.0	0.9619
2.9492	28.0	392	2.8425	1.0	0.9619
2.8892	29.0	406	2.7813	1.0	0.9619
2.8892	30.0	420	2.7270	1.0	0.9619
2.8892	31.0	434	2.6645	1.0	0.9606
2.8892	32.0	448	2.5593	1.0	0.9139
2.8892	33.0	462	2.3230	1.0	0.7003
2.8892	34.0	476	1.9706	1.0	0.5358
2.8892	35.0	490	1.7085	0.9998	0.4548
2.3937	36.0	504	1.4494	1.0	0.4064
2.3937	37.0	518	1.2865	1.0	0.3847
2.3937	38.0	532	1.1509	0.9947	0.3659
2.3937	39.0	546	1.0467	0.9031	0.3183
2.3937	40.0	560	0.9832	0.5961	0.2404
2.3937	41.0	574	0.8921	0.5049	0.2223
2.3937	42.0	588	0.8306	0.4687	0.2123
1.0877	43.0	602	0.8017	0.4563	0.2088
1.0877	44.0	616	0.7716	0.4405	0.2046
1.0877	45.0	630	0.7694	0.4407	0.2054
1.0877	46.0	644	0.7451	0.4315	0.2037
1.0877	47.0	658	0.7112	0.4250	0.1996
1.0877	48.0	672	0.7008	0.4116	0.1958
1.0877	49.0	686	0.7140	0.4057	0.1980
0.6292	50.0	700	0.7208	0.4114	0.1988
0.6292	51.0	714	0.6675	0.4033	0.1937
0.6292	52.0	728	0.6650	0.4015	0.1938
0.6292	53.0	742	0.6550	0.4013	0.1938
0.6292	54.0	756	0.6477	0.3990	0.1932
0.6292	55.0	770	0.6362	0.3960	0.1932
0.6292	56.0	784	0.6323	0.3919	0.1930
0.6292	57.0	798	0.6264	0.3870	0.1921
0.4739	58.0	812	0.6290	0.3872	0.1921
0.4739	59.0	826	0.6207	0.3864	0.1925
0.4739	60.0	840	0.6178	0.3858	0.1918
0.4739	61.0	854	0.6217	0.3860	0.1918
0.4739	62.0	868	0.6078	0.3799	0.1900
0.4739	63.0	882	0.6072	0.3781	0.1889
0.4739	64.0	896	0.6068	0.3761	0.1883
0.3855	65.0	910	0.5945	0.3748	0.1870
0.3855	66.0	924	0.6194	0.3799	0.1900
0.3855	67.0	938	0.6044	0.3793	0.1885
0.3855	68.0	952	0.5946	0.3751	0.1880
0.3855	69.0	966	0.6116	0.3714	0.1880
0.3855	70.0	980	0.5877	0.3679	0.1861
0.3855	71.0	994	0.5861	0.3679	0.1863
0.3302	72.0	1008	0.5805	0.3685	0.1856
0.3302	73.0	1022	0.5862	0.3714	0.1862
0.3302	74.0	1036	0.5921	0.3720	0.1866
0.3302	75.0	1050	0.5692	0.3683	0.1854
0.3302	76.0	1064	0.5922	0.3702	0.1878
0.3302	77.0	1078	0.6105	0.3710	0.1883
0.3302	78.0	1092	0.5873	0.3683	0.1856
0.3046	79.0	1106	0.5826	0.3681	0.1859
0.3046	80.0	1120	0.5792	0.3633	0.1845
0.3046	81.0	1134	0.5738	0.3610	0.1835
0.3046	82.0	1148	0.5794	0.3625	0.1843
0.3046	83.0	1162	0.5766	0.3564	0.1829
0.3046	84.0	1176	0.5745	0.3578	0.1830
0.3046	85.0	1190	0.5615	0.3555	0.1814
0.2927	86.0	1204	0.5854	0.3614	0.1828
0.2927	87.0	1218	0.5818	0.3625	0.1835
0.2927	88.0	1232	0.5613	0.3578	0.1815
0.2927	89.0	1246	0.5661	0.3549	0.1813
0.2927	90.0	1260	0.5795	0.3604	0.1820
0.2927	91.0	1274	0.5604	0.3521	0.1802
0.2927	92.0	1288	0.5738	0.3590	0.1822
0.2576	93.0	1302	0.5658	0.3574	0.1814
0.2576	94.0	1316	0.5620	0.3511	0.1808
0.2576	95.0	1330	0.5709	0.3541	0.1810
0.2576	96.0	1344	0.5675	0.3503	0.1799
0.2576	97.0	1358	0.5788	0.3549	0.1815
0.2576	98.0	1372	0.5730	0.3525	0.1810
0.2576	99.0	1386	0.5694	0.3511	0.1803
0.2273	100.0	1400	0.5748	0.3527	0.1807
0.2273	101.0	1414	0.5688	0.3513	0.1797
0.2273	102.0	1428	0.5767	0.3553	0.1805
0.2273	103.0	1442	0.5758	0.3529	0.1812
0.2273	104.0	1456	0.5641	0.3507	0.1793
0.2273	105.0	1470	0.5628	0.3495	0.1789
0.2273	106.0	1484	0.5729	0.3466	0.1789
0.2273	107.0	1498	0.5722	0.3497	0.1798
0.2181	108.0	1512	0.5553	0.3466	0.1788
0.2181	109.0	1526	0.5582	0.3484	0.1792
0.2181	110.0	1540	0.5702	0.3521	0.1802
0.2181	111.0	1554	0.5691	0.3505	0.1798
0.2181	112.0	1568	0.5604	0.3470	0.1786
0.2181	113.0	1582	0.5661	0.3482	0.1795
0.2181	114.0	1596	0.5683	0.3511	0.1796
0.2171	115.0	1610	0.5738	0.3509	0.1798
0.2171	116.0	1624	0.5730	0.3458	0.1793
0.2171	117.0	1638	0.5705	0.3456	0.1789
0.2171	118.0	1652	0.5814	0.3466	0.1796
0.2171	119.0	1666	0.5715	0.3442	0.1791
0.2171	120.0	1680	0.5720	0.3470	0.1798
0.2171	121.0	1694	0.5769	0.3470	0.1797
0.1986	122.0	1708	0.5711	0.3464	0.1792
0.1986	123.0	1722	0.5728	0.3442	0.1790
0.1986	124.0	1736	0.5668	0.3450	0.1783
0.1986	125.0	1750	0.5855	0.3484	0.1797
0.1986	126.0	1764	0.5667	0.3427	0.1783
0.1986	127.0	1778	0.5711	0.3460	0.1789
0.1986	128.0	1792	0.5682	0.3444	0.1781

Framework versions

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