metadata
language:
- multilingual
license: apache-2.0
base_model: openai/whisper-small
tags:
- hf-asr-leaderboard
- generated_from_trainer
datasets:
- mozilla-foundation/common_voice_11_0
metrics:
- wer
model-index:
- name: whisper small trained on 5000 en de en
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: xbilek25/xbilek25/train_set_5000_en_de_en
type: mozilla-foundation/common_voice_11_0
args: 'config: ende, split: train'
metrics:
- name: Wer
type: wer
value: 6.089633937735203
basic_train_basic_test 1000 similar params: per_device_train_batch_size=32, # bylo 16 a pod tim 1 gradient_accumulation_steps=2, warmup_steps=300, max_steps=3000
This model is a fine-tuned version of openai/whisper-small on the xbilek25/xbilek25/train_set_5000_en_de_en dataset. It achieves the following results on the evaluation set:
- Loss: 0.1263
- Wer: 6.0896
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: 4000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.1048 | 1.05 | 500 | 0.1559 | 9.2371 |
| 0.0147 | 3.02 | 1000 | 0.1256 | 7.3212 |
| 0.004 | 4.06 | 1500 | 0.1147 | 6.7054 |
| 0.0013 | 6.03 | 2000 | 0.1188 | 6.1238 |
| 0.0011 | 7.08 | 2500 | 0.1216 | 6.1923 |
| 0.0012 | 9.05 | 3000 | 0.1234 | 5.9528 |
| 0.0007 | 11.01 | 3500 | 0.1258 | 6.0896 |
| 0.0007 | 12.06 | 4000 | 0.1263 | 6.0896 |
Framework versions
- Transformers 4.37.2
- Pytorch 2.2.1+cu121
- Datasets 2.19.0
- Tokenizers 0.15.2