wav2vec2-xls-r-Wolof-28-hours-alffa-plus-fleurs-dataset

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the fleurs dataset. It achieves the following results on the evaluation set:

• Loss: 1.2602
• Wer: 0.4408
• Cer: 0.1556

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: 0.0003
• 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: 0.1
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
7.4709	0.7286	400	3.3573	1.0	1.0
3.0751	1.4572	800	3.2424	1.0	1.0
2.3348	2.1858	1200	1.1856	0.8107	0.2961
0.7384	2.9144	1600	0.9239	0.6241	0.2184
0.5978	3.6430	2000	0.7904	0.5831	0.2078
0.5443	4.3716	2400	0.7605	0.5613	0.2024
0.517	5.1002	2800	0.7471	0.5778	0.2085
0.4672	5.8288	3200	0.7260	0.5458	0.1966
0.4348	6.5574	3600	0.7106	0.5102	0.1839
0.4117	7.2860	4000	0.6811	0.5134	0.1873
0.3845	8.0146	4400	0.6847	0.5192	0.1877
0.3617	8.7432	4800	0.6870	0.5747	0.2167
0.3395	9.4718	5200	0.6642	0.5366	0.1938
0.323	10.2004	5600	0.6752	0.5127	0.1880
0.309	10.9290	6000	0.6696	0.5075	0.1838
0.2853	11.6576	6400	0.7383	0.5071	0.1852
0.2652	12.3862	6800	0.6571	0.5063	0.1834
0.2557	13.1148	7200	0.6866	0.4921	0.1791
0.2443	13.8434	7600	0.6916	0.4945	0.1800
0.224	14.5719	8000	0.6833	0.5365	0.1904
0.2202	15.3005	8400	0.6920	0.5205	0.1886
0.2117	16.0291	8800	0.7279	0.5241	0.1881
0.1932	16.7577	9200	0.6772	0.5071	0.1837
0.1874	17.4863	9600	0.7134	0.4961	0.1769
0.1733	18.2149	10000	0.7350	0.5096	0.1849
0.1705	18.9435	10400	0.7188	0.5021	0.1808
0.1631	19.6721	10800	0.7608	0.5126	0.1861
0.1518	20.4007	11200	0.7117	0.4884	0.1760
0.147	21.1293	11600	0.7853	0.4754	0.1698
0.1421	21.8579	12000	0.8173	0.4717	0.1682
0.1365	22.5865	12400	0.7970	0.4797	0.1734
0.1328	23.3151	12800	0.8331	0.4799	0.1752
0.1273	24.0437	13200	0.7806	0.4729	0.1713
0.1202	24.7723	13600	0.8037	0.4704	0.1666
0.1185	25.5009	14000	0.8201	0.4650	0.1690
0.1119	26.2295	14400	0.9294	0.4774	0.1720
0.1087	26.9581	14800	0.8380	0.4718	0.1697
0.1043	27.6867	15200	0.9948	0.4677	0.1670
0.1034	28.4153	15600	0.9864	0.4707	0.1689
0.0979	29.1439	16000	1.0066	0.4694	0.1704
0.0947	29.8725	16400	0.8745	0.4802	0.1709
0.0876	30.6011	16800	0.9511	0.4910	0.1750
0.0857	31.3297	17200	0.9594	0.4525	0.1625
0.0842	32.0583	17600	1.0274	0.4681	0.1662
0.08	32.7869	18000	0.9747	0.4623	0.1647
0.08	33.5155	18400	0.9912	0.4676	0.1640
0.077	34.2441	18800	1.1352	0.4629	0.1634
0.0753	34.9727	19200	1.0100	0.4542	0.1614
0.0712	35.7013	19600	1.0493	0.4554	0.1605
0.0679	36.4299	20000	1.1336	0.4528	0.1620
0.0664	37.1585	20400	1.1095	0.4496	0.1599
0.0679	37.8871	20800	1.0197	0.4576	0.1621
0.0615	38.6157	21200	1.1053	0.4567	0.1606
0.0588	39.3443	21600	1.1809	0.4469	0.1594
0.0594	40.0729	22000	1.1607	0.4538	0.1619
0.0566	40.8015	22400	1.1570	0.4498	0.1586
0.0567	41.5301	22800	1.1453	0.4505	0.1590
0.0554	42.2587	23200	1.1740	0.4563	0.1586
0.0524	42.9872	23600	1.1408	0.4557	0.1598
0.0504	43.7158	24000	1.1360	0.4511	0.1587
0.0492	44.4444	24400	1.2167	0.4487	0.1576
0.0505	45.1730	24800	1.1709	0.4442	0.1571
0.0479	45.9016	25200	1.2109	0.4443	0.1569
0.047	46.6302	25600	1.2031	0.4430	0.1555
0.0454	47.3588	26000	1.2316	0.4401	0.1555
0.0442	48.0874	26400	1.2515	0.4390	0.1554
0.0445	48.8160	26800	1.2538	0.4409	0.1554
0.041	49.5446	27200	1.2602	0.4408	0.1556

Framework versions

• Transformers 4.44.1
• Pytorch 2.1.0+cu118
• Datasets 2.17.0
• Tokenizers 0.19.1