metadata
library_name: transformers
language:
- zh
license: apache-2.0
base_model: openai/whisper-small
tags:
- generated_from_trainer
model-index:
- name: Whisper-Small-squeezeformer-architecture
results: []
Whisper-Small-squeezeformer-architecture
This model is a fine-tuned version of openai/whisper-small on the Voice_Data_Collection_second_edition dataset. It achieves the following results on the evaluation set:
- Loss: 0.4281
- Cer: 24.9352
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: 3750
- training_steps: 60000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Cer | Validation Loss |
|---|---|---|---|---|
| 3.0206 | 1.0 | 3750 | 101.4703 | 2.9448 |
| 2.6624 | 2.0 | 7500 | 99.0366 | 2.7587 |
| 1.4476 | 3.0 | 11250 | 80.5695 | 1.4832 |
| 0.7219 | 4.0 | 15000 | 49.1456 | 0.8063 |
| 0.5387 | 5.0 | 18750 | 41.2985 | 0.6620 |
| 0.3824 | 6.0 | 22500 | 39.1312 | 0.6240 |
| 0.3445 | 7.0 | 26250 | 37.5137 | 0.5984 |
| 0.2819 | 8.0 | 30000 | 36.6006 | 0.5823 |
| 0.538 | 9.0 | 33750 | 0.5156 | 30.9133 |
| 0.3792 | 10.0 | 37500 | 0.4562 | 27.3205 |
| 0.2974 | 11.0 | 41250 | 0.4408 | 26.4521 |
| 0.2281 | 12.0 | 45000 | 0.4225 | 25.7254 |
| 0.1644 | 13.0 | 48750 | 0.4207 | 25.1123 |
| 0.1208 | 14.0 | 52500 | 0.4241 | 25.0098 |
| 0.0877 | 15.0 | 56250 | 0.4232 | 24.9781 |
| 0.0652 | 16.0 | 60000 | 0.4281 | 24.9352 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.4.0
- Datasets 2.21.0
- Tokenizers 0.19.1