nouamanetazi's picture
nouamanetazi HF staff
update model card README.md
16c3cc2
|
raw
history blame
2.48 kB
metadata
license: apache-2.0
tags:
  - generated_from_trainer
datasets:
  - common_voice
model-index:
  - name: wav2vec2-xls-r-300m-ar
    results: []

wav2vec2-xls-r-300m-ar

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. It achieves the following results on the evaluation set:

  • Loss: 6.9375
  • Wer: 1.0

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: 7.5e-05
  • train_batch_size: 64
  • eval_batch_size: 64
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 20
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
19.0933 1.0 2 32.5547 1.0
18.6757 2.0 4 31.0408 1.0
17.1649 3.0 6 29.5776 1.0
14.7415 4.0 8 22.9183 1.0
11.8071 5.0 10 17.5073 1.0
9.7675 6.0 12 14.1743 1.0
8.4193 7.0 14 12.1225 1.0
7.4746 8.0 16 10.7446 1.0
6.8442 9.0 18 9.7949 1.0
6.3765 10.0 20 9.1148 1.0
6.0321 11.0 22 8.5934 1.0
5.7783 12.0 24 8.1983 1.0
5.5827 13.0 26 7.8780 1.0
5.4249 14.0 28 7.6288 1.0
5.3088 15.0 30 7.4279 1.0
5.2078 16.0 32 7.2688 1.0
5.1289 17.0 34 7.1459 1.0
5.0697 18.0 36 7.0528 1.0
5.0227 19.0 38 6.9835 1.0
4.9853 20.0 40 6.9375 1.0

Framework versions

  • Transformers 4.17.0.dev0
  • Pytorch 1.10.2+cu102
  • Datasets 1.18.2.dev0
  • Tokenizers 0.11.0