Panga-Azazia/minianka-asr

wav2vec2-malian-languages-minianka-dataset

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the SUDOPING01/MALIAN-LANGUAGES-DATASET - MINIANKA dataset. It achieves the following results on the evaluation set:

• Loss: 0.1794
• Wer: 0.1271

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: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 15.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.4706	100	3.8351	1.0
No log	0.9412	200	2.8794	1.0
No log	1.4094	300	0.6950	0.5312
No log	1.88	400	0.3555	0.3603
2.6845	2.3482	500	0.2666	0.2772
2.6845	2.8188	600	0.2253	0.2401
2.6845	3.2871	700	0.2025	0.2246
2.6845	3.7576	800	0.1880	0.1979
2.6845	4.2259	900	0.1853	0.1959
0.2121	4.6965	1000	0.1887	0.1909
0.2121	5.1647	1100	0.1745	0.1709
0.2121	5.6353	1200	0.1602	0.1674
0.2121	6.1035	1300	0.1627	0.1588
0.2121	6.5741	1400	0.1593	0.1563
0.1148	7.0424	1500	0.1660	0.1593
0.1148	7.5129	1600	0.1655	0.1551
0.1148	7.9835	1700	0.1581	0.1520
0.1148	8.4518	1800	0.1771	0.1493
0.1148	8.9224	1900	0.1778	0.1482
0.0723	9.3906	2000	0.1683	0.1402
0.0723	9.8612	2100	0.1676	0.1378
0.0723	10.3294	2200	0.1672	0.1365
0.0723	10.8	2300	0.1646	0.1332
0.0723	11.2682	2400	0.1751	0.1337
0.0504	11.7388	2500	0.1742	0.1353
0.0504	12.2071	2600	0.1809	0.1329
0.0504	12.6776	2700	0.1769	0.1298
0.0504	13.1459	2800	0.1752	0.1289
0.0504	13.6165	2900	0.1782	0.1275
0.0382	14.0847	3000	0.1789	0.1289
0.0382	14.5553	3100	0.1783	0.1275

Framework versions

• Transformers 4.47.0
• Pytorch 2.5.1+cu121
• Datasets 3.3.1
• Tokenizers 0.21.0