xls-r-ur-large

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the common_voice_8_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.8056
• Wer: 0.4716

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: 1000
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
6.5282	3.25	1000	3.0650	0.9989
1.7351	6.49	2000	0.8798	0.6284
0.7662	9.74	3000	0.7720	0.5399
0.5675	12.99	4000	0.7661	0.5229
0.4591	16.23	5000	0.7849	0.5041
0.3881	19.48	6000	0.8065	0.4893
0.3522	22.73	7000	0.7915	0.4804
0.3127	25.97	8000	0.8119	0.4804
0.2932	29.22	9000	0.8056	0.4716

Framework versions

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