README.md

---
library_name: transformers
license: apache-2.0
base_model: openai/whisper-small
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: whisper-a-norm-ls-5
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# whisper-a-norm-ls-5

This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0448
- Wer: 184.1866

## 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: 0.0004
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 132
- num_epochs: 5
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch  | Step | Validation Loss | Wer      |
|:-------------:|:------:|:----:|:---------------:|:--------:|
| No log        | 1.0    | 70   | 0.2880          | 157.4516 |
| 0.7821        | 2.0    | 140  | 0.0673          | 14.2207  |
| 0.1757        | 3.0    | 210  | 0.0536          | 175.0853 |
| 0.1757        | 4.0    | 280  | 0.0507          | 199.3174 |
| 0.0113        | 4.9353 | 345  | 0.0448          | 184.1866 |


### Framework versions

- Transformers 4.47.0.dev0
- Pytorch 2.4.0
- Datasets 3.0.1
- Tokenizers 0.20.0