DrishtiSharma's picture
update model card README.md
ebf3100
|
raw
history blame
2.49 kB
metadata
license: apache-2.0
tags:
  - generated_from_trainer
datasets:
  - common_voice
model-index:
  - name: ''
    results: []

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: 0.7060
  • Wer: 0.5559

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: 7e-05
  • train_batch_size: 32
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 1200
  • num_epochs: 50.0
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
3.9554 2.61 300 3.8166 1.0
2.9853 5.22 600 2.9793 1.0
0.6939 7.83 900 0.9723 0.9149
0.352 10.43 1200 0.6556 0.7296
0.2369 13.04 1500 0.6315 0.6928
0.1757 15.65 1800 0.6387 0.6724
0.1517 18.26 2100 0.6607 0.6265
0.1229 20.87 2400 0.6520 0.6183
0.1201 23.48 2700 0.6750 0.6116
0.1076 26.09 3000 0.6698 0.6006
0.1006 28.7 3300 0.6804 0.5872
0.0952 31.3 3600 0.7010 0.6003
0.0894 33.91 3900 0.7069 0.5840
0.0873 36.52 4200 0.6765 0.5781
0.0798 39.13 4500 0.6822 0.5617
0.0767 41.74 4800 0.7003 0.5638
0.0717 44.35 5100 0.6970 0.5591
0.0687 46.96 5400 0.7073 0.5603
0.0669 49.57 5700 0.7060 0.5559

Framework versions

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