---
tags:
- generated_from_trainer
model-index:
- name: ls-timit-100percent-supervised-aug
  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. -->

# ls-timit-100percent-supervised-aug

This model was trained from scratch on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0519
- Wer: 0.0292

## 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: 0.0001
- train_batch_size: 32
- 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: 1000
- num_epochs: 20
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch | Step | Validation Loss | Wer    |
|:-------------:|:-----:|:----:|:---------------:|:------:|
| 0.2985        | 7.04  | 1000 | 0.0556          | 0.0380 |
| 0.1718        | 14.08 | 2000 | 0.0519          | 0.0292 |


### Framework versions

- Transformers 4.16.2
- Pytorch 1.10.2
- Datasets 1.18.2
- Tokenizers 0.10.3