xls-r_cv_ur

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4888
• Wer: 1.0

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
8.4361	1.14	100	2.3733	1.0
1.9609	2.27	200	1.8280	1.0
1.8191	3.41	300	1.8259	1.0
1.8359	4.55	400	1.8069	1.0
1.7667	5.68	500	1.8038	1.0
1.7751	6.82	600	1.7473	1.0
1.7428	7.95	700	1.6996	1.0
1.697	9.09	800	1.6364	1.0
1.6532	10.23	900	1.4985	1.0
1.5217	11.36	1000	1.2836	1.0
1.3385	12.5	1100	1.0293	1.0
1.1596	13.64	1200	0.8294	1.0
1.0655	14.77	1300	0.7150	1.0
0.9951	15.91	1400	0.6364	1.0
0.9013	17.05	1500	0.5548	1.0
0.8276	18.18	1600	0.5200	1.0
0.8129	19.32	1700	0.4888	1.0

Framework versions

• Transformers 4.21.0
• Pytorch 1.11.0+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1