metadata

language:
  - 'no'
license: apache-2.0
base_model: NbAiLab/nb-whisper-medium-RC1
tags:
  - audio
  - asr
  - automatic-speech-recognition
  - hf-asr-leaderboard
model-index:
  - name: nb-whisper-medium-v0.8-vad3
    results: []

nb-whisper-medium-v0.8-vad3

This model is a fine-tuned version of NbAiLab/nb-whisper-medium-RC1 on the NbAiLab/ncc_speech_styling_v2_vad3 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: 2.5e-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	0.4223	0.8343	2.3463	0.7206	2.9397	0.8105	0.6313	8.8868	5.7697	11.8752	6.2280
5000	0.4364	0.5289	2.6077	0.8063	3.2555	0.9057	0.6300	9.1071	5.8300	12.0840	6.3028
10000	0.4353	0.4901	2.4824	0.7765	3.0867	0.8709	0.6463	9.2563	5.9382	12.2144	6.4042
15000	0.4338	0.4760	2.4062	0.7290	3.0541	0.8324	0.6788	9.0052	5.7931	12.0816	6.2720
20000	0.4343	0.4553	2.5695	0.7868	3.1956	0.8819	0.7058	9.1710	5.9434	12.2168	6.4280

Framework versions

Transformers 4.34.1
Datasets 2.16.1
Tokenizers 0.14.1