wav2vec2-large-xls-r-300m-breton-colab

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:

• Loss: 1.2211
• Wer: 0.4937

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
5.3288	1.34	400	1.7076	0.9809
1.2014	2.69	800	1.0803	0.7733
0.7687	4.03	1200	0.9806	0.6642
0.5539	5.38	1600	0.9914	0.6301
0.4456	6.72	2000	0.9797	0.6265
0.3586	8.07	2400	1.0354	0.5803
0.2922	9.41	2800	0.9996	0.5821
0.2628	10.76	3200	1.0250	0.5708
0.2284	12.1	3600	1.0865	0.5722
0.1908	13.45	4000	1.0674	0.5450
0.1732	14.79	4400	1.1775	0.5614
0.153	16.13	4800	1.1542	0.5435
0.14	17.48	5200	1.1807	0.5449
0.1302	18.82	5600	1.1679	0.5376
0.1142	20.17	6000	1.1441	0.5276
0.104	21.51	6400	1.2243	0.5355
0.0882	22.86	6800	1.1837	0.5316
0.0807	24.2	7200	1.1986	0.5132
0.0744	25.55	7600	1.2182	0.5108
0.0646	26.89	8000	1.2116	0.5047
0.0551	28.24	8400	1.2009	0.4948
0.0503	29.58	8800	1.2211	0.4937

Framework versions

• Transformers 4.32.1
• Pytorch 2.0.1+cu118
• Datasets 2.14.4
• Tokenizers 0.13.3