pere's picture
Update stats.md
c953c33
metadata
language:
  - 'no'
license: apache-2.0
base_model: NbAiLabBeta/nb-whisper-large
tags:
  - audio
  - asr
  - automatic-speech-recognition
  - hf-asr-leaderboard
model-index:
  - name: nb-whisper-large-v0.8-vad3-verbatim
    results: []

nb-whisper-large-v0.8-vad3-verbatim

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

  • step: 249
  • validation_loss: 0.5839
  • train_loss: 0.4632
  • validation_wer: 7.9358
  • validation_cer: 2.5127
  • validation_exact_wer: 8.0494
  • validation_exact_cer: 2.5279

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: 250
  • starting_optimization_step: None
  • finishing_optimization_step: 250
  • num_train_dataset_workers: 32
  • num_hosts: 32
  • total_num_training_examples: 256,000
  • steps_per_epoch: 97
  • 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_loss train_loss validation_wer validation_cer validation_exact_wer validation_exact_cer
0 1.2831 1.1864 18.9083 11.8409 33.9801 15.0322
40 0.5952 0.4958 8.9760 2.9212 9.1099 2.9390
80 0.5848 0.4761 8.3105 2.6432 8.4330 2.6621
120 0.5831 0.4492 8.1204 2.5679 8.2356 2.5821
160 0.5811 0.4678 7.9302 2.5051 8.0438 2.5193
200 0.5840 0.4692 7.9861 2.5346 8.0945 2.5498
240 0.5844 0.4543 7.9246 2.5051 8.0381 2.5193
249 0.5839 0.4632 7.9358 2.5127 8.0494 2.5279

Framework versions

  • Transformers 4.36.2
  • Datasets 2.16.1
  • Tokenizers 0.15.0