wav2vec2-xls-r-Wolof-10-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: 1.4744
• Wer: 0.5216
• 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: 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
5.797	2.3669	400	3.0600	1.0	1.0
2.6522	4.7337	800	1.5315	0.9461	0.3837
0.7997	7.1006	1200	0.9805	0.6680	0.2374
0.5559	9.4675	1600	0.9895	0.6360	0.2278
0.4535	11.8343	2000	0.7992	0.5788	0.2062
0.3639	14.2012	2400	0.9448	0.5772	0.2114
0.3102	16.5680	2800	0.8459	0.5738	0.2019
0.2516	18.9349	3200	0.9974	0.5601	0.1987
0.219	21.3018	3600	1.0617	0.5644	0.2016
0.189	23.6686	4000	1.0441	0.5556	0.1972
0.1645	26.0355	4400	1.1185	0.5484	0.2045
0.1427	28.4024	4800	1.0686	0.5666	0.2048
0.1342	30.7692	5200	1.2091	0.5409	0.1927
0.1192	33.1361	5600	1.2784	0.5367	0.1926
0.1064	35.5030	6000	1.3353	0.5421	0.1926
0.1021	37.8698	6400	1.4003	0.5377	0.1898
0.0889	40.2367	6800	1.4877	0.5409	0.1902
0.0876	42.6036	7200	1.4466	0.5306	0.1855
0.0788	44.9704	7600	1.4520	0.5242	0.1841
0.0744	47.3373	8000	1.4631	0.5177	0.1867
0.0732	49.7041	8400	1.4744	0.5216	0.1821

Framework versions

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