metadata
language:
- en
license: apache-2.0
base_model: openai/whisper-small
tags:
- generated_from_trainer
datasets:
- Jzuluaga/atcosim_corpus
metrics:
- wer
model-index:
- name: Whisper Base ATCOSIM
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: atcosim_corpus_numbers_converted
type: Jzuluaga/atcosim_corpus
args: 'config: en, split: test'
metrics:
- name: Wer
type: wer
value: 404.9337012855893
Whisper Base ATCOSIM
This model is a fine-tuned version of openai/whisper-small on the atcosim_corpus_numbers_converted dataset. It achieves the following results on the evaluation set:
- Loss: 0.5964
- Wer: 404.9337
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 |
|---|---|---|---|---|
| 8.7316 | 0.2092 | 100 | 8.1521 | 110.7020 |
| 5.3768 | 0.4184 | 200 | 5.2758 | 110.9713 |
| 3.2842 | 0.6276 | 300 | 3.4243 | 119.1492 |
| 1.892 | 0.8368 | 400 | 2.1743 | 121.4377 |
| 0.9999 | 1.0460 | 500 | 1.4513 | 168.8160 |
| 0.6472 | 1.2552 | 600 | 1.0157 | 369.7449 |
| 0.529 | 1.4644 | 700 | 0.8372 | 356.8554 |
| 0.4086 | 1.6736 | 800 | 0.6973 | 300.4644 |
| 0.3559 | 1.8828 | 900 | 0.6164 | 361.0150 |
| 0.2302 | 2.0921 | 1000 | 0.5964 | 404.9337 |
Framework versions
- Transformers 4.42.0.dev0
- Pytorch 2.3.0+cu121
- Datasets 2.19.2
- Tokenizers 0.19.1