wav2vec2-arabic-gpu-colab-similar-to-german-bigger-warm-up

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.6370
  • Wer: 0.4146

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: 2
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 6
  • total_train_batch_size: 12
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 5000
  • num_epochs: 40
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
9.4958 2.83 400 3.4822 1.0
3.2281 5.67 800 2.9404 1.0
2.942 8.51 1200 2.8690 1.0
2.6346 11.35 1600 1.5452 0.9994
1.3472 14.18 2000 0.8261 0.6853
0.8972 17.02 2400 0.6812 0.5737
0.6924 19.85 2800 0.6552 0.5291
0.5687 22.69 3200 0.6108 0.4909
0.4734 25.53 3600 0.5877 0.4674
0.4029 28.37 4000 0.6204 0.4662
0.3483 31.2 4400 0.5932 0.4451
0.307 34.04 4800 0.6445 0.4392
0.2722 36.88 5200 0.6126 0.4292
0.2247 39.71 5600 0.6370 0.4146

Framework versions

  • Transformers 4.11.3
  • Pytorch 1.10.0+cu113
  • Datasets 1.18.3
  • Tokenizers 0.10.3
Downloads last month
7
Inference Providers NEW
This model is not currently available via any of the supported third-party Inference Providers, and the model is not deployed on the HF Inference API.