wav2vec2-large-xlsr-coraa-exp-17

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: 0.5486
• Wer: 0.3511
• Cer: 0.1797
• Per: 0.3387

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: 4e-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: 0.01
• num_epochs: 150
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Per
38.4208	1.0	14	41.8095	1.0057	1.2146	1.0057
38.4208	2.0	28	12.2873	1.0	0.9619	1.0
38.4208	3.0	42	4.8093	1.0	0.9619	1.0
38.4208	4.0	56	3.9738	1.0	0.9619	1.0
38.4208	5.0	70	3.6684	1.0	0.9619	1.0
38.4208	6.0	84	3.5007	1.0	0.9619	1.0
38.4208	7.0	98	3.3854	1.0	0.9619	1.0
11.8009	8.0	112	3.4506	1.0	0.9619	1.0
11.8009	9.0	126	3.1789	1.0	0.9619	1.0
11.8009	10.0	140	3.1274	1.0	0.9619	1.0
11.8009	11.0	154	3.1624	1.0	0.9619	1.0
11.8009	12.0	168	3.1066	1.0	0.9619	1.0
11.8009	13.0	182	3.0580	1.0	0.9619	1.0
11.8009	14.0	196	3.0477	1.0	0.9619	1.0
3.0395	15.0	210	3.0519	1.0	0.9619	1.0
3.0395	16.0	224	3.0364	1.0	0.9619	1.0
3.0395	17.0	238	3.0152	1.0	0.9619	1.0
3.0395	18.0	252	3.0167	1.0	0.9619	1.0
3.0395	19.0	266	3.0130	1.0	0.9619	1.0
3.0395	20.0	280	3.0103	1.0	0.9619	1.0
3.0395	21.0	294	2.9994	1.0	0.9619	1.0
2.9424	22.0	308	2.9999	1.0	0.9619	1.0
2.9424	23.0	322	3.0009	1.0	0.9619	1.0
2.9424	24.0	336	3.0024	1.0	0.9619	1.0
2.9424	25.0	350	3.0001	1.0	0.9619	1.0
2.9424	26.0	364	2.9891	1.0	0.9619	1.0
2.9424	27.0	378	2.9881	1.0	0.9619	1.0
2.9424	28.0	392	2.9703	1.0	0.9619	1.0
2.9154	29.0	406	2.9531	1.0	0.9619	1.0
2.9154	30.0	420	2.9208	1.0	0.9619	1.0
2.9154	31.0	434	2.8981	1.0	0.9619	1.0
2.9154	32.0	448	2.8321	1.0	0.9619	1.0
2.9154	33.0	462	2.7583	1.0	0.9619	1.0
2.9154	34.0	476	2.6405	1.0	0.9616	1.0
2.9154	35.0	490	2.5072	1.0	0.8832	1.0
2.7552	36.0	504	2.1547	1.0	0.6144	1.0
2.7552	37.0	518	1.7565	1.0	0.4996	1.0
2.7552	38.0	532	1.4602	1.0	0.4065	1.0
2.7552	39.0	546	1.2269	0.9896	0.3658	0.9892
2.7552	40.0	560	1.0906	0.8881	0.3205	0.8834
2.7552	41.0	574	0.9941	0.6772	0.2631	0.6603
2.7552	42.0	588	0.9133	0.5423	0.2322	0.5154
1.4599	43.0	602	0.8487	0.5142	0.2241	0.4882
1.4599	44.0	616	0.8211	0.4898	0.2207	0.4626
1.4599	45.0	630	0.7672	0.4803	0.2140	0.4518
1.4599	46.0	644	0.7432	0.4707	0.2092	0.4445
1.4599	47.0	658	0.7390	0.4492	0.2059	0.4262
1.4599	48.0	672	0.6994	0.4348	0.2011	0.4106
1.4599	49.0	686	0.6999	0.4230	0.1991	0.3998
0.7585	50.0	700	0.6738	0.4122	0.1959	0.3883
0.7585	51.0	714	0.6697	0.4094	0.1963	0.3858
0.7585	52.0	728	0.6707	0.4163	0.1996	0.3954
0.7585	53.0	742	0.6397	0.4031	0.1942	0.3832
0.7585	54.0	756	0.6293	0.4039	0.1939	0.3836
0.7585	55.0	770	0.6479	0.4027	0.1946	0.3852
0.7585	56.0	784	0.6307	0.3982	0.1934	0.3822
0.7585	57.0	798	0.6166	0.3844	0.1908	0.3673
0.5473	58.0	812	0.6099	0.3860	0.1906	0.3708
0.5473	59.0	826	0.6007	0.3868	0.1904	0.3730
0.5473	60.0	840	0.6191	0.3885	0.1928	0.3744
0.5473	61.0	854	0.6015	0.3885	0.1892	0.3732
0.5473	62.0	868	0.5965	0.3838	0.1902	0.3688
0.5473	63.0	882	0.5926	0.3826	0.1904	0.3667
0.5473	64.0	896	0.6188	0.3921	0.1921	0.3765
0.443	65.0	910	0.5835	0.3830	0.1892	0.3690
0.443	66.0	924	0.5914	0.3870	0.1903	0.3722
0.443	67.0	938	0.5828	0.3779	0.1876	0.3627
0.443	68.0	952	0.5745	0.3722	0.1857	0.3576
0.443	69.0	966	0.5786	0.3795	0.1882	0.3633
0.443	70.0	980	0.5869	0.3751	0.1884	0.3604
0.443	71.0	994	0.5923	0.3753	0.1888	0.3596
0.3564	72.0	1008	0.5707	0.3714	0.1859	0.3578
0.3564	73.0	1022	0.5733	0.3700	0.1857	0.3551
0.3564	74.0	1036	0.5731	0.3706	0.1854	0.3566
0.3564	75.0	1050	0.5644	0.3669	0.1847	0.3531
0.3564	76.0	1064	0.5661	0.3702	0.1852	0.3555
0.3564	77.0	1078	0.5705	0.3675	0.1847	0.3513
0.3564	78.0	1092	0.5631	0.3671	0.1835	0.3527
0.3456	79.0	1106	0.5675	0.3651	0.1831	0.3503
0.3456	80.0	1120	0.5697	0.3645	0.1846	0.3507
0.3456	81.0	1134	0.5644	0.3631	0.1841	0.3492
0.3456	82.0	1148	0.5657	0.3627	0.1843	0.3480
0.3456	83.0	1162	0.5831	0.3679	0.1876	0.3523
0.3456	84.0	1176	0.5824	0.3659	0.1862	0.3523
0.3456	85.0	1190	0.5567	0.3653	0.1833	0.3509
0.3073	86.0	1204	0.5755	0.3649	0.1852	0.3507
0.3073	87.0	1218	0.5590	0.3586	0.1829	0.3450
0.3073	88.0	1232	0.5663	0.3610	0.1835	0.3480
0.3073	89.0	1246	0.5734	0.3618	0.1851	0.3468
0.3073	90.0	1260	0.5657	0.3602	0.1830	0.3458
0.3073	91.0	1274	0.5651	0.3578	0.1828	0.3442
0.3073	92.0	1288	0.5608	0.3557	0.1820	0.3415
0.2836	93.0	1302	0.5505	0.3525	0.1807	0.3389
0.2836	94.0	1316	0.5495	0.3501	0.1798	0.3375
0.2836	95.0	1330	0.5693	0.3557	0.1816	0.3432
0.2836	96.0	1344	0.5638	0.3564	0.1822	0.3417
0.2836	97.0	1358	0.5486	0.3511	0.1797	0.3387
0.2836	98.0	1372	0.5618	0.3545	0.1810	0.3415
0.2836	99.0	1386	0.5637	0.3515	0.1800	0.3399
0.2502	100.0	1400	0.5658	0.3555	0.1810	0.3438
0.2502	101.0	1414	0.5527	0.3525	0.1795	0.3411
0.2502	102.0	1428	0.5701	0.3562	0.1807	0.3440
0.2502	103.0	1442	0.5543	0.3497	0.1794	0.3389
0.2502	104.0	1456	0.5660	0.3509	0.1803	0.3399
0.2502	105.0	1470	0.5543	0.3501	0.1795	0.3399
0.2502	106.0	1484	0.5742	0.3547	0.1817	0.3432
0.2502	107.0	1498	0.5527	0.3454	0.1789	0.3350
0.2368	108.0	1512	0.5577	0.3497	0.1789	0.3379
0.2368	109.0	1526	0.5539	0.3452	0.1789	0.3356
0.2368	110.0	1540	0.5700	0.3517	0.1802	0.3417
0.2368	111.0	1554	0.5627	0.3501	0.1794	0.3397
0.2368	112.0	1568	0.5622	0.3497	0.1797	0.3405
0.2368	113.0	1582	0.5708	0.3495	0.1801	0.3403
0.2368	114.0	1596	0.5733	0.3511	0.1805	0.3401
0.2288	115.0	1610	0.5615	0.3486	0.1795	0.3387
0.2288	116.0	1624	0.5741	0.3497	0.1809	0.3397
0.2288	117.0	1638	0.5610	0.3460	0.1796	0.3373

Framework versions

• Transformers 4.28.0
• Pytorch 2.6.0+cu124
• Datasets 3.3.2
• Tokenizers 0.13.3