adrianSauer/wav2vec2-wer-extension2

Common Voice 16

This model is a fine-tuned version of glob-asr/wav2vec2-large-xls-r-300m-guarani-small on the Common Voice 16 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4335
• Wer: 49.7002

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 16
• 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: constant_with_warmup
• lr_scheduler_warmup_steps: 3000
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.258	0.4955	500	0.3710	53.1646
0.921	0.9911	1000	0.3282	49.2338
0.7458	1.4866	1500	0.2940	46.7022
0.6763	1.9822	2000	0.2628	44.9700
0.568	2.4777	2500	0.2616	43.3711
0.5414	2.9732	3000	0.2504	39.8401
0.484	3.4688	3500	0.2462	41.0393
0.5281	3.9643	4000	0.3584	43.5043
0.5756	4.4599	4500	0.4220	44.3038
0.721	4.9554	5000	0.4335	49.7002

Framework versions

• Transformers 4.44.1
• Pytorch 2.3.1+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1