pere's picture
Saving weights and logs of step 0 - epoch 0
fb48bda
|
raw
history blame
No virus
1.94 kB
metadata
language:
  - 'no'
license: apache-2.0
base_model: NbAiLab/nb-whisper-tiny-v0.8-vad3
tags:
  - audio
  - asr
  - automatic-speech-recognition
  - hf-asr-leaderboard
model-index:
  - name: nb-whisper-tiny-v0.8-vad3-verbatim
    results: []

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

This model is a fine-tuned version of NbAiLab/nb-whisper-tiny-v0.8-vad3 on the NbAiLab/NPSC dataset.

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: 0.00015
  • 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: 250
  • starting_optimization_step: None
  • finishing_optimization_step: 250
  • num_train_dataset_workers: 32
  • num_hosts: 8
  • total_num_training_examples: 256,000
  • steps_per_epoch: To be computed after first epoch
  • 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 1.2428 1.2274 23.3208 12.5036 37.5903 15.7108

Framework versions

  • Transformers 4.34.1
  • Datasets 2.16.1
  • Tokenizers 0.14.1