wav2vec2-base-german

This model is a fine-tuned version of wav2vec2-base-german on the MOZILLA-FOUNDATION/COMMON_VOICE_10_0 - DE dataset. It achieves the following results on the evaluation set:

• Loss: 0.9302
• Wer: 0.7428

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: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 1024
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.8427	1.0	451	1.0878	0.8091
0.722	2.0	902	0.9732	0.7593
0.6589	3.0	1353	0.9302	0.7428

Framework versions

• Transformers 4.21.1
• Pytorch 1.12.0
• Datasets 2.4.0
• Tokenizers 0.12.1