wav2vec2-large-xlsr-coraa-exp-1

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: 0.5564
• Wer: 0.3555
• 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
39.0927	1.0	14	30.0708	1.0	0.9613
39.0927	2.0	28	9.0585	1.0	0.9619
39.0927	3.0	42	4.9408	1.0	0.9619
39.0927	4.0	56	4.1247	1.0	0.9619
39.0927	5.0	70	3.8053	1.0	0.9619
39.0927	6.0	84	3.6257	1.0	0.9619
39.0927	7.0	98	3.4882	1.0	0.9619
10.1374	8.0	112	3.3658	1.0	0.9619
10.1374	9.0	126	3.2659	1.0	0.9619
10.1374	10.0	140	3.1858	1.0	0.9619
10.1374	11.0	154	3.1509	1.0	0.9619
10.1374	12.0	168	3.1087	1.0	0.9619
10.1374	13.0	182	3.0901	1.0	0.9619
10.1374	14.0	196	3.0563	1.0	0.9619
3.0936	15.0	210	3.0725	1.0	0.9619
3.0936	16.0	224	3.0693	1.0	0.9619
3.0936	17.0	238	3.0401	1.0	0.9619
3.0936	18.0	252	3.0363	1.0	0.9619
3.0936	19.0	266	3.0399	1.0	0.9619
3.0936	20.0	280	3.0229	1.0	0.9619
3.0936	21.0	294	3.0135	1.0	0.9619
2.9599	22.0	308	3.0165	1.0	0.9619
2.9599	23.0	322	3.0048	1.0	0.9619
2.9599	24.0	336	3.0040	1.0	0.9619
2.9599	25.0	350	2.9981	1.0	0.9619
2.9599	26.0	364	3.0019	1.0	0.9619
2.9599	27.0	378	2.9937	1.0	0.9619
2.9599	28.0	392	2.9919	1.0	0.9619
2.9236	29.0	406	2.9904	1.0	0.9619
2.9236	30.0	420	2.9829	1.0	0.9619
2.9236	31.0	434	2.9772	1.0	0.9619
2.9236	32.0	448	2.9296	1.0	0.9619
2.9236	33.0	462	2.9011	1.0	0.9619
2.9236	34.0	476	2.8410	1.0	0.9619
2.9236	35.0	490	2.7730	1.0	0.9619
2.8623	36.0	504	2.7076	1.0	0.9618
2.8623	37.0	518	2.6268	1.0	0.9590
2.8623	38.0	532	2.4435	1.0	0.8387
2.8623	39.0	546	2.1666	1.0	0.6899
2.8623	40.0	560	1.8675	1.0	0.5183
2.8623	41.0	574	1.5864	1.0	0.4389
2.8623	42.0	588	1.4232	0.9996	0.3989
2.1779	43.0	602	1.2338	0.9976	0.3828
2.1779	44.0	616	1.1252	0.9431	0.3392
2.1779	45.0	630	1.0542	0.7505	0.2762
2.1779	46.0	644	0.9471	0.5510	0.2287
2.1779	47.0	658	0.8948	0.5154	0.2230
2.1779	48.0	672	0.8252	0.5055	0.2195
2.1779	49.0	686	0.7892	0.4691	0.2086
1.0861	50.0	700	0.7734	0.4464	0.2053
1.0861	51.0	714	0.7450	0.4466	0.2057
1.0861	52.0	728	0.7445	0.4421	0.2054
1.0861	53.0	742	0.7073	0.4291	0.2007
1.0861	54.0	756	0.7187	0.4279	0.2016
1.0861	55.0	770	0.7030	0.4185	0.1996
1.0861	56.0	784	0.6911	0.4130	0.1973
1.0861	57.0	798	0.6678	0.4055	0.1953
0.715	58.0	812	0.6554	0.4072	0.1947
0.715	59.0	826	0.6637	0.4110	0.1960
0.715	60.0	840	0.6606	0.4037	0.1962
0.715	61.0	854	0.6598	0.4069	0.1969
0.715	62.0	868	0.6365	0.4023	0.1946
0.715	63.0	882	0.6275	0.3937	0.1928
0.715	64.0	896	0.6460	0.3925	0.1941
0.5672	65.0	910	0.6349	0.3939	0.1945
0.5672	66.0	924	0.6282	0.3933	0.1938
0.5672	67.0	938	0.6014	0.3872	0.1901
0.5672	68.0	952	0.6073	0.3854	0.1899
0.5672	69.0	966	0.6144	0.3862	0.1914
0.5672	70.0	980	0.6038	0.3860	0.1912
0.5672	71.0	994	0.6110	0.3836	0.1916
0.4622	72.0	1008	0.6022	0.3781	0.1891
0.4622	73.0	1022	0.5961	0.3775	0.1890
0.4622	74.0	1036	0.5991	0.3753	0.1885
0.4622	75.0	1050	0.5966	0.3732	0.1887
0.4622	76.0	1064	0.5963	0.3785	0.1897
0.4622	77.0	1078	0.5902	0.3816	0.1896
0.4622	78.0	1092	0.5695	0.3738	0.1864
0.4311	79.0	1106	0.5828	0.3765	0.1869
0.4311	80.0	1120	0.5799	0.3748	0.1871
0.4311	81.0	1134	0.5753	0.3746	0.1874
0.4311	82.0	1148	0.5795	0.3738	0.1876
0.4311	83.0	1162	0.5899	0.3726	0.1884
0.4311	84.0	1176	0.5791	0.3671	0.1864
0.4311	85.0	1190	0.5711	0.3649	0.1850
0.3905	86.0	1204	0.5771	0.3692	0.1857
0.3905	87.0	1218	0.5769	0.3657	0.1850
0.3905	88.0	1232	0.5681	0.3663	0.1846
0.3905	89.0	1246	0.5772	0.3653	0.1846
0.3905	90.0	1260	0.5658	0.3623	0.1835
0.3905	91.0	1274	0.5706	0.3653	0.1853
0.3905	92.0	1288	0.5735	0.3600	0.1838
0.3626	93.0	1302	0.5607	0.3598	0.1833
0.3626	94.0	1316	0.5736	0.3610	0.1839
0.3626	95.0	1330	0.5701	0.3604	0.1847
0.3626	96.0	1344	0.5775	0.3637	0.1856
0.3626	97.0	1358	0.5564	0.3555	0.1821
0.3626	98.0	1372	0.5770	0.3580	0.1839
0.3626	99.0	1386	0.5692	0.3584	0.1831
0.3218	100.0	1400	0.5748	0.3582	0.1831
0.3218	101.0	1414	0.5647	0.3553	0.1822
0.3218	102.0	1428	0.5756	0.3584	0.1831
0.3218	103.0	1442	0.5739	0.3590	0.1833
0.3218	104.0	1456	0.5663	0.3586	0.1828
0.3218	105.0	1470	0.5631	0.3602	0.1829
0.3218	106.0	1484	0.5747	0.3616	0.1838
0.3218	107.0	1498	0.5691	0.3590	0.1838
0.3032	108.0	1512	0.5573	0.3582	0.1829
0.3032	109.0	1526	0.5605	0.3570	0.1834
0.3032	110.0	1540	0.5719	0.3568	0.1838
0.3032	111.0	1554	0.5595	0.3568	0.1826
0.3032	112.0	1568	0.5614	0.3570	0.1825
0.3032	113.0	1582	0.5676	0.3566	0.1832
0.3032	114.0	1596	0.5715	0.3572	0.1834
0.2957	115.0	1610	0.5735	0.3584	0.1831
0.2957	116.0	1624	0.5706	0.3588	0.1833
0.2957	117.0	1638	0.5708	0.3551	0.1828

Framework versions

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