|
---
|
|
language:
|
|
- he
|
|
license: apache-2.0
|
|
base_model: openai/whisper-small
|
|
tags:
|
|
- generated_from_trainer
|
|
datasets:
|
|
- OverloadedOperator/tests-101
|
|
model-index:
|
|
- name: Whisper Small He
|
|
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 Small He
|
|
|
|
This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the TestDS dataset.
|
|
|
|
## 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: 10
|
|
- mixed_precision_training: Native AMP
|
|
|
|
### Training results
|
|
|
|
|
|
|
|
### Framework versions
|
|
|
|
- Transformers 4.38.2
|
|
- Pytorch 2.2.1+cu121
|
|
- Datasets 2.18.0
|
|
- Tokenizers 0.15.0
|
|
|
