|
--- |
|
license: apache-2.0 |
|
tags: |
|
- whisper-event |
|
metrics: |
|
- wer |
|
model-index: |
|
- name: agnesluhtaru/whisper-medium-et-ERR2020 |
|
results: |
|
- task: |
|
type: automatic-speech-recognition |
|
name: Automatic Speech Recognition |
|
dataset: |
|
name: mozilla-foundation/common_voice_11_0 |
|
type: mozilla-foundation/common_voice_11_0 |
|
config: et |
|
split: test |
|
metrics: |
|
- type: wer |
|
value: 20.56 |
|
name: WER |
|
--- |
|
|
|
# whisper-medium-et with ERR2020 data |
|
|
|
This model is a fine-tuned version of [openai/whisper-medium](https://huggingface.co/openai/whisper-medium) on the following datasets: Common Voice 11, VoxPopuli, FLEURS and [ERR2020](http://bark.phon.ioc.ee/lw/korpused/ERR2020.html). |
|
The model is stopped a little early because the Whisper fine-tuning event was ending :) |
|
|
|
## Model description |
|
|
|
More information needed |
|
|
|
## Intended uses & limitations |
|
|
|
More information needed |
|
|
|
## Training and evaluation data |
|
|
|
Estonian data from Common Voice 11, VoxPopuli, FLEURS and ERR2020 corpora as both training and validation sets. Tested on Common Voice 11 test set. |
|
|
|
## Training procedure |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 1e-05 |
|
- train_batch_size: 32 |
|
- eval_batch_size: 16 |
|
- 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: 6000 |
|
- mixed_precision_training: Native AMP |
|
|
|
### Framework versions |
|
|
|
- Transformers 4.26.0.dev0 |
|
- Pytorch 1.12.1+rocm5.1.1 |
|
- Datasets 2.7.1.dev0 |
|
- Tokenizers 0.13.2 |