metadata
language:
- en
license: apache-2.0
tags:
- generated_from_trainer
base_model: openai/whisper-small
datasets:
- Jzuluaga/atcosim_corpus
metrics:
- wer
model-index:
- name: Whisper Base ATCOSIM
results:
- task:
type: automatic-speech-recognition
name: Automatic Speech Recognition
dataset:
name: atcosim_corpus
type: Jzuluaga/atcosim_corpus
args: 'config: en, split: test'
metrics:
- type: wer
value: 4.914301017675415
name: Wer
Whisper Base ATCOSIM
This model is a fine-tuned version of openai/whisper-small on the atcosim_corpus dataset. It achieves the following results on the evaluation set:
- Loss: 0.0770
- Wer: 4.9143
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: 16
- 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: 1000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1.882 | 0.2092 | 100 | 1.5290 | 67.4143 |
| 0.4334 | 0.4184 | 200 | 0.4251 | 26.2587 |
| 0.1883 | 0.6276 | 300 | 0.2218 | 15.5932 |
| 0.1259 | 0.8368 | 400 | 0.1487 | 10.9869 |
| 0.0713 | 1.0460 | 500 | 0.1168 | 9.4135 |
| 0.0446 | 1.2552 | 600 | 0.1036 | 7.5656 |
| 0.0652 | 1.4644 | 700 | 0.0919 | 6.3203 |
| 0.0508 | 1.6736 | 800 | 0.0829 | 5.1152 |
| 0.038 | 1.8828 | 900 | 0.0776 | 4.9009 |
| 0.0186 | 2.0921 | 1000 | 0.0770 | 4.9143 |
Framework versions
- Transformers 4.42.0.dev0
- Pytorch 2.3.0+cu121
- Datasets 2.19.2
- Tokenizers 0.19.1