nb-whisper-small / stats.md
pere's picture
Bulk update
8d53c43
metadata
language:
  - 'no'
license: apache-2.0
base_model: NbAiLab/nb-whisper-small-RC1
tags:
  - audio
  - asr
  - automatic-speech-recognition
  - hf-asr-leaderboard
model-index:
  - name: nb-whisper-small-v0.8-vad3
    results: []

nb-whisper-small-v0.8-vad3

This model is a fine-tuned version of NbAiLab/nb-whisper-small-RC1 on the NbAiLab/ncc_speech_styling_v2_vad3 dataset. It achieves the following results on the evaluation set:

  • step: 49999
  • validation_nst_loss: 0.4444
  • train_loss: 0.4400
  • validation_nst_wer: 3.0595
  • validation_nst_cer: 0.9443
  • validation_nst_exact_wer: 3.7237
  • validation_nst_exact_cer: 1.0431
  • validation_clean_stortinget_no_loss: 0.7056
  • validation_clean_stortinget_no_wer: 10.0663
  • validation_clean_stortinget_no_cer: 6.2768
  • validation_clean_stortinget_no_exact_wer: 13.4172
  • validation_clean_stortinget_no_exact_cer: 6.8042

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
  • lr_scheduler_type: linear
  • per_device_train_batch_size: 32
  • total_train_batch_size_per_node: 128
  • total_train_batch_size: 1024
  • total_optimization_steps: 50,000
  • starting_optimization_step: None
  • finishing_optimization_step: 50,000
  • num_train_dataset_workers: 32
  • num_hosts: 8
  • total_num_training_examples: 51,200,000
  • steps_per_epoch: 7455
  • num_beams: None
  • weight_decay: 0.01
  • adam_beta1: 0.9
  • adam_beta2: 0.98
  • adam_epsilon: 1e-06
  • dropout: True
  • bpe_dropout_probability: 0.2
  • activation_dropout_probability: 0.1

Training results

step validation_nst_loss train_loss validation_nst_wer validation_nst_cer validation_nst_exact_wer validation_nst_exact_cer validation_clean_stortinget_no_loss validation_clean_stortinget_no_wer validation_clean_stortinget_no_cer validation_clean_stortinget_no_exact_wer validation_clean_stortinget_no_exact_cer
0 0.4313 1.0396 2.8254 0.8865 3.5168 0.9900 0.5547 9.6092 5.9949 12.6794 6.4755
5000 0.4484 0.5692 3.2010 1.0142 3.8870 1.1172 0.6138 10.1824 6.1896 13.4124 6.6954
10000 0.4477 0.5317 3.3589 1.0347 4.0176 1.1337 0.6275 10.3316 6.4310 13.6022 6.9442
15000 0.4493 0.5132 3.3099 1.0086 3.9904 1.1145 0.6599 10.2203 6.3042 13.4100 6.8175
20000 0.4491 0.4911 3.2283 1.0226 3.8924 1.1227 0.6755 10.1421 6.3188 13.4409 6.8428
25000 0.4441 0.4766 3.1575 0.9816 3.8924 1.0898 0.6763 10.2700 6.3383 13.5951 6.8658
30000 0.4498 0.4632 3.1357 0.9741 3.8543 1.0797 0.6599 10.2274 6.3787 13.5144 6.8974
35000 0.4480 0.4523 3.0432 0.9378 3.7727 1.0486 0.6948 10.2416 6.3617 13.5547 6.8912
40000 0.4471 0.4606 3.0486 0.9080 3.7291 1.0101 0.6754 10.2155 6.3375 13.5097 6.8506
45000 0.4442 0.4412 2.9778 0.9275 3.6366 1.0229 0.7021 10.1468 6.2994 13.5286 6.8358
49999 0.4444 0.4400 3.0595 0.9443 3.7237 1.0431
49999 0.7056 0.4400 10.0663 6.2768 13.4172 6.8042

Framework versions

  • Transformers 4.34.1
  • Datasets 2.16.1
  • Tokenizers 0.14.1