Edit model card

xls-r-1b-bemba-10hrs

This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2350
  • Wer: 0.3524

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: 3e-05
  • train_batch_size: 4
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 400
  • num_epochs: 10
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
2.2547 0.54 400 0.4199 0.5888
0.5422 1.07 800 0.2689 0.4360
0.4154 1.61 1200 0.2342 0.4008
0.4075 2.15 1600 0.2172 0.3579
0.3326 2.68 2000 0.2151 0.3603
0.2837 3.22 2400 0.2117 0.3505
0.2688 3.76 2800 0.2040 0.3559
0.2401 4.3 3200 0.2099 0.3445
0.2176 4.83 3600 0.1973 0.3299
0.1913 5.37 4000 0.2123 0.3432
0.1683 5.91 4400 0.2032 0.3358
0.1445 6.44 4800 0.2350 0.3524

Framework versions

  • Transformers 4.19.0.dev0
  • Pytorch 1.10.0+cu111
  • Datasets 2.1.0
  • Tokenizers 0.12.1
Downloads last month
3