|
|
--- |
|
|
language: |
|
|
- 'no' |
|
|
license: apache-2.0 |
|
|
base_model: NbAiLab/nb-whisper-tiny-v0.7 |
|
|
tags: |
|
|
- audio |
|
|
- asr |
|
|
- automatic-speech-recognition |
|
|
- hf-asr-leaderboard |
|
|
model-index: |
|
|
- name: nb-whisper-tiny-v0.7-semantic |
|
|
results: [] |
|
|
--- |
|
|
|
|
|
<!-- This model card has been generated automatically according to the information Keras had access to. You should |
|
|
probably proofread and complete it, then remove this comment. --> |
|
|
|
|
|
# nb-whisper-tiny-v0.7-semantic |
|
|
|
|
|
This model is a fine-tuned version of [NbAiLab/nb-whisper-tiny-v0.7](https://huggingface.co/NbAiLab/nb-whisper-tiny-v0.7) on the NbAiLab/ncc_speech_styling_v4 dataset. |
|
|
It achieves the following results on the evaluation set: |
|
|
- step: 249 |
|
|
- validation_nst_loss: 0.6579 |
|
|
- train_loss: 1.2508 |
|
|
- validation_nst_wer: 8.8029 |
|
|
- validation_nst_cer: 2.9662 |
|
|
- validation_nst_exact_wer: 9.6358 |
|
|
- validation_nst_exact_cer: 3.0880 |
|
|
- validation_clean_stortinget_no_loss: 0.7202 |
|
|
- validation_clean_stortinget_no_wer: 16.5324 |
|
|
- validation_clean_stortinget_no_cer: 9.1926 |
|
|
- validation_clean_stortinget_no_exact_wer: 20.6476 |
|
|
- validation_clean_stortinget_no_exact_cer: 9.9178 |
|
|
|
|
|
## 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: 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: 0.00015 |
|
|
- 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 | 0.5155 | 1.4782 | 7.9155 | 2.5430 | 8.7103 | 2.6640 | 0.7153 | 15.8148 | 8.6090 | 19.7201 | 9.3061 | |
|
|
| 40 | 0.5328 | 1.3581 | 8.6232 | 2.8926 | 9.6031 | 3.0477 | 0.7058 | 17.0843 | 9.6747 | 21.3213 | 10.4075 | |
|
|
| 80 | 0.6321 | 1.1988 | 8.8464 | 3.1266 | 9.7447 | 3.2629 | 0.7131 | 16.8712 | 9.3548 | 20.9062 | 10.0570 | |
|
|
| 120 | 0.6456 | 1.1956 | 8.7757 | 3.0306 | 9.6630 | 3.1585 | 0.7165 | 16.7243 | 9.2521 | 20.8113 | 9.9704 | |
|
|
| 160 | 0.6455 | 1.2386 | 8.7648 | 2.9784 | 9.6195 | 3.1081 | 0.7188 | 16.7030 | 9.2259 | 20.8967 | 9.9658 | |
|
|
| 200 | 0.6576 | 1.1753 | 8.6777 | 2.9336 | 9.4888 | 3.0559 | 0.7202 | 16.5917 | 9.1859 | 20.7188 | 9.9205 | |
|
|
| 240 | 0.6577 | 1.1923 | 8.9226 | 2.9923 | 9.7338 | 3.1100 | 0.7214 | 16.6272 | 9.2176 | 20.7211 | 9.9396 | |
|
|
| 249 | 0.6579 | 1.2508 | 8.8029 | 2.9662 | 9.6358 | 3.0880 | |
|
|
| 249 | 0.7202 | 1.2508 | 16.5324 | 9.1926 | 20.6476 | 9.9178 | |
|
|
|
|
|
|
|
|
### Framework versions |
|
|
|
|
|
- Transformers 4.34.1 |
|
|
- Datasets 2.15.0 |
|
|
- Tokenizers 0.14.1 |
|
|
|
