Edit model card

wav2vec2-large-xlsr-53-german-cv8-dropout

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the MOZILLA-FOUNDATION/COMMON_VOICE_8_0 - DE dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1111
  • Wer: 0.1117

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: 16
  • eval_batch_size: 32
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 64
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 1000
  • num_epochs: 10.0
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
0.2081 1.0 6815 0.1784 0.1910
0.1686 2.0 13630 0.1621 0.1725
0.1515 3.0 20445 0.1569 0.1649
0.1426 4.0 27260 0.1466 0.1681
0.135 5.0 34075 0.1357 0.1410
0.1093 6.0 40890 0.1313 0.1436
0.1 7.0 47705 0.1242 0.1250
0.0999 8.0 54520 0.1191 0.1218
0.084 9.0 61335 0.1134 0.1164
0.0752 10.0 68150 0.1111 0.1117

Framework versions

  • Transformers 4.19.0.dev0
  • Pytorch 1.11.0+cu113
  • Datasets 2.0.0
  • Tokenizers 0.11.6
Downloads last month
26
Hosted inference API
or or
This model can be loaded on the Inference API on-demand.

Dataset used to train oliverguhr/wav2vec2-large-xlsr-53-german-cv8