---
license: apache-2.0
language:
- es
- ca
base_model:
- openai/whisper-large-v3
pipeline_tag: automatic-speech-recognition
library_name: transformers
tags:
- bsc
- projecte-aina
- barcelona-supercomputing-center
- automatic-speech-recognition
- whisper-large-v3
- code-switching
- spanish-catalan
- spanish
- catalan
---

# whisper-large-v3-tiny-caesar

## Table of Contents
<details>
<summary>Click to expand</summary>

- [Model Description](#model-description)
- [Intended Uses and Limitations](#intended-uses-and-limitations)
- [How to Get Started with the Model](#how-to-get-started-with-the-model)
- [Training Details](#training-details)
- [Citation](#citation)
- [Additional Information](#additional-information)

</details>

## Summary

The "whisper-large-v3-tiny-caesar" is an acoustic model based on ["openai/whisper-large-v3"](https://huggingface.co/openai/whisper-large-v3) suitable for Automatic Speech Recognition in code-switching conditions between Spanish and Catalan.

## Model Description

The "whisper-large-v3-tiny-caesar" is an acoustic model suitable for Automatic Speech Recognition in code-switching conditions between Spanish and Catalan. It is the result of fine-tuning the ["openai/whisper-large-v3"](https://huggingface.co/openai/whisper-large-v3) with [CAESAR-TINY](https://huggingface.co/datasets/BSC-LT/CAESAR-TINY), a 2-hour code-switching dataset in Spanish/Catalan.

## Intended Uses and Limitations

This model can be used for Automatic Speech Recognition (ASR) in code-switching conditions between Spanish and Catalan. The model is intended to transcribe audio files to plain text.

## How to Get Started with the Model

To see an updated and functional version of this code, please check our [Notebook](https://colab.research.google.com/drive/1MHiPrffNTwiyWeUyMQvSdSbfkef_8aJC?usp=sharing)

### Installation

To use this model, you may install [datasets](https://huggingface.co/docs/datasets/installation) and [transformers](https://huggingface.co/docs/transformers/installation):

Create a virtual environment:
```bash
python -m venv /path/to/venv
```
Activate the environment:
```bash
source /path/to/venv/bin/activate
```
Install the modules:
```bash
pip install datasets transformers 
```

### For Inference
To transcribe audio in Catalan using this model, you can follow this example:

```bash
#Install Prerequisites
pip install torch
pip install datasets
pip install 'transformers[torch]'
pip install evaluate
pip install jiwer
```

```python
#This code works with GPU

#Notice that: load_metric is no longer part of datasets.
# You have to remove it and use evaluate's load instead.
#(Note from November 2024)

import torch
from transformers import WhisperForConditionalGeneration, WhisperProcessor

#Load the processor and model.
MODEL_NAME="projecte-aina/whisper-large-v3-tiny-caesar"
processor = WhisperProcessor.from_pretrained(MODEL_NAME)
model = WhisperForConditionalGeneration.from_pretrained(MODEL_NAME).to("cuda")

#Load the dataset
from datasets import load_dataset, load_metric, Audio
ds=load_dataset("projecte-aina/3catparla_asr",split='test')

#Downsample to 16kHz
ds = ds.cast_column("audio", Audio(sampling_rate=16_000))

#Process the dataset
def map_to_pred(batch):
	audio = batch["audio"]
	input_features = processor(audio["array"], sampling_rate=audio["sampling_rate"], return_tensors="pt").input_features
	batch["reference"] = processor.tokenizer._normalize(batch['normalized_text'])

	with torch.no_grad():
		predicted_ids = model.generate(input_features.to("cuda"))[0]
	
	transcription = processor.decode(predicted_ids)
	batch["prediction"] = processor.tokenizer._normalize(transcription)
	
	return batch
	
#Do the evaluation
result = ds.map(map_to_pred)

#Compute the overall WER now.
from evaluate import load

wer = load("wer")
WER=100 * wer.compute(references=result["reference"], predictions=result["prediction"])
print(WER)
```

## Training Details

### Training data

The specific dataset used to create the model is a corpus called CAESAR-tiny, which has not been released at the moment.

### Training procedure

This model is the result of finetuning the model ["openai/whisper-large-v3"](https://huggingface.co/openai/whisper-large-v3) by following this [tutorial](https://huggingface.co/blog/fine-tune-whisper) provided by Hugging Face.

### Training Hyperparameters

* language: Spanish
* hours of training audio: 2
* learning rate: 1e-5
* sample rate: 16000
* train batch size: 32 (x4 GPUs)
  * gradient accumulation steps: 1
* eval batch size: 32
* save total limit: 3
* max steps: 80
* warmup steps: 8
* eval steps: 8
* save steps: 8
* shuffle buffer size: 480

## Citation
If this model contributes to your research, please cite the work:
```bibtex
@misc{mena2024whisperlarge3catparla,
      title={Acoustic Model in Catalan: whisper-large-v3-tiny-caesar.}, 
      author={Hernandez Mena, Carlos Daniel; Giraldo, Jose ;Armentano-Oller, Carme; Solito, Sarah; Messaoudi, Abir; Costa, Federico; Zeballos, Rodolfo},
      organization={Barcelona Supercomputing Center},
      url={https://huggingface.co/projecte-aina/whisper-large-v3-tiny-caesar},
      year={2024}
}
```

## Additional Information

### Author

The fine-tuning process was performed during November (2024) in the [Language Technologies Unit](https://huggingface.co/BSC-LT) of the [Barcelona Supercomputing Center](https://www.bsc.es/) by [Carlos Daniel Hernández Mena](https://huggingface.co/carlosdanielhernandezmena).

### Contact
For further information, please send an email to <langtech@bsc.es>.

### Copyright
Copyright(c) 2024 by Language Technologies Unit, Barcelona Supercomputing Center.

### License

[Apache-2.0](https://www.apache.org/licenses/LICENSE-2.0)

### Funding
This work has been promoted and financed by the Generalitat de Catalunya through the [Aina project](https://projecteaina.cat/).

The training of the model was possible thanks to the computing time provided by [Barcelona Supercomputing Center](https://www.bsc.es/) through MareNostrum 5.