danieleV9H's picture
update model card README.md
0d9392f
metadata
license: apache-2.0
tags:
  - generated_from_trainer
datasets:
  - librispeech_asr
model-index:
  - name: hubert-base-libri-clean-ft100h
    results: []

hubert-base-libri-clean-ft100h

This model is a fine-tuned version of facebook/hubert-base-ls960 on the librispeech_asr dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1324
  • Wer: 0.1597

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: 5e-05
  • train_batch_size: 8
  • eval_batch_size: 16
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 16
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 1000
  • num_epochs: 5
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
No log 0.14 250 4.1508 1.0000
4.4345 0.28 500 3.8766 1.0000
4.4345 0.42 750 3.4376 1.0000
2.8475 0.56 1000 2.7380 1.0
2.8475 0.7 1250 0.8803 0.6766
1.1877 0.84 1500 0.5671 0.5102
1.1877 0.98 1750 0.4537 0.4388
0.5802 1.12 2000 0.3566 0.3740
0.5802 1.26 2250 0.2925 0.3209
0.4301 1.4 2500 0.2613 0.2952
0.4301 1.54 2750 0.2363 0.2715
0.3591 1.68 3000 0.2155 0.2552
0.3591 1.82 3250 0.2062 0.2418
0.3015 1.96 3500 0.1951 0.2308
0.3015 2.1 3750 0.1842 0.2207
0.2698 2.24 4000 0.1900 0.2112
0.2698 2.38 4250 0.1745 0.2048
0.2561 2.52 4500 0.1718 0.2040
0.2561 2.66 4750 0.1625 0.1939
0.2348 2.8 5000 0.1568 0.1867
0.2348 2.94 5250 0.1517 0.1855
0.2278 3.08 5500 0.1501 0.1807
0.2278 3.22 5750 0.1445 0.1772
0.2166 3.36 6000 0.1422 0.1752
0.2166 3.5 6250 0.1418 0.1741
0.2017 3.64 6500 0.1404 0.1695
0.2017 3.78 6750 0.1356 0.1674
0.1922 3.92 7000 0.1350 0.1688
0.1922 4.06 7250 0.1346 0.1638
0.1979 4.2 7500 0.1359 0.1638
0.1979 4.34 7750 0.1336 0.1612
0.1836 4.48 8000 0.1324 0.1613
0.1836 4.62 8250 0.1320 0.1606
0.1891 4.76 8500 0.1325 0.1598
0.1891 4.9 8750 0.1324 0.1597

Framework versions

  • Transformers 4.17.0
  • Pytorch 1.11.0+cu113
  • Datasets 1.18.3
  • Tokenizers 0.12.1