nb-whisper-large / stats.md
pere's picture
Bulk update
78cb1b6
metadata
language:
  - 'no'
license: apache-2.0
base_model: NbAiLab/nb-whisper-large-v3-RC4
tags:
  - audio
  - asr
  - automatic-speech-recognition
  - hf-asr-leaderboard
model-index:
  - name: nb-whisper-large-v0.8-vad3
    results: []

nb-whisper-large-v0.8-vad3

This model is a fine-tuned version of NbAiLab/nb-whisper-large-v3-RC4 on the NbAiLab/ncc_speech_styling_v2_vad3 dataset. It achieves the following results on the evaluation set:

  • step: 49999
  • validation_nst_loss: 0.4292
  • train_loss: 0.4893
  • validation_nst_wer: 2.2211
  • validation_nst_cer: 0.6628
  • validation_nst_exact_wer: 2.8145
  • validation_nst_exact_cer: 0.7555
  • validation_clean_stortinget_no_loss: 0.7534
  • validation_clean_stortinget_no_wer: 8.9128
  • validation_clean_stortinget_no_cer: 5.6979
  • validation_clean_stortinget_no_exact_wer: 11.8159
  • validation_clean_stortinget_no_exact_cer: 6.1484

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
  • lr_scheduler_type: linear
  • per_device_train_batch_size: 8
  • total_train_batch_size_per_node: 32
  • total_train_batch_size: 1024
  • total_optimization_steps: 15,000
  • starting_optimization_step: 35,000
  • finishing_optimization_step: 50,000
  • num_train_dataset_workers: 32
  • num_hosts: 32
  • total_num_training_examples: 51,200,000
  • steps_per_epoch: 24982
  • 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 0.4259 0.9588 2.1721 0.6246 2.7111 0.7079 0.6807 8.5931 5.4608 11.4221 5.8946
5000 0.4376 0.5822 2.5859 0.7793 3.0867 0.8563 0.6738 9.1686 5.8478 12.0792 6.3020
10000 0.4368 0.5675 2.5913 0.7271 3.2337 0.8269 0.6875 9.2705 5.9200 12.1741 6.3750
15000 0.4335 0.5403 2.3409 0.6936 2.9180 0.7821 0.7187 9.0834 5.7344 11.9962 6.1944
20000 0.4324 0.5187 2.3518 0.6945 2.9561 0.7857 0.7357 8.9839 5.7154 11.8610 6.1664
25000 0.4307 0.5158 2.3028 0.6712 2.9343 0.7711 0.7228 9.1284 5.8704 11.9915 6.3161
30000 0.4312 0.5108 2.2810 0.6656 2.8690 0.7564 0.7428 8.9010 5.6726 11.8349 6.1305
35000 0.4299 0.4908 2.2320 0.6768 2.8417 0.7729 0.7513 8.8015 5.6123 11.6854 6.0642
40000 0.4313 0.4865 2.2973 0.6917 2.8907 0.7839 0.7545 8.9057 5.6912 11.8491 6.1465
45000 0.4303 0.4849 2.2429 0.6665 2.8254 0.7564 0.7484 8.9578 5.7320 11.8752 6.1851
49999 0.4292 0.4893 2.2211 0.6628 2.8145 0.7555
49999 0.7534 0.4893 8.9128 5.6979 11.8159 6.1484

Framework versions

  • Transformers 4.36.2
  • Datasets 2.16.1
  • Tokenizers 0.15.0