metadata
language:
- sv
license: apache-2.0
tags:
- generated_from_trainer
metrics:
- wer
datasets:
- jimregan/sbtal_riksdag_asr
model-index:
- name: Whisper Small Sv - Riksdag 100h
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: SBTal Riksdag ASR
type: jimregan/sbtal_riksdag_asr
metrics:
- name: Test WER
type: wer
value: 720.3756
Whisper Small Sv - Riksdag 100h
This model is a fine-tuned version of openai/whisper-small on the SBTal Riksdag ASR dataset. It achieves the following results on the evaluation set:
- Loss: 0.4019
- Wer: 720.3756
That's not an error, the results really are that bad, and this should not be used by anyone, ever, except to get a good laugh. I'll try to run fine-tuning again, but don't hold your breath.
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: 1e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 4000
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
0.1349 | 0.11 | 1000 | 0.4679 | 732.5701 |
0.1071 | 0.22 | 2000 | 0.4305 | 1417.2884 |
0.0959 | 0.33 | 3000 | 0.4077 | 787.1881 |
0.0691 | 0.43 | 4000 | 0.4019 | 720.3756 |
Framework versions
- Transformers 4.27.0.dev0
- Pytorch 1.13.1
- Datasets 2.10.1
- Tokenizers 0.13.2