whisper-large-v3-ca-3catparla

Table of Contents

Click to expand

• Model Description
• Intended Uses and Limitations
• How to Get Started with the Model
• Training Details
• Citation
• Additional Information

Summary

The "whisper-large-v3-ca-3catparla" is an acoustic model based on "openai/whisper-large-v3" suitable for Automatic Speech Recognition in Catalan.

Model Description

The "whisper-large-v3-ca-3catparla" is an acoustic model suitable for Automatic Speech Recognition in Catalan. It is the result of finetuning the model "openai/whisper-large-v3" with 710 hours of Catalan data released by the Projecte AINA from Barcelona, Spain.

Intended Uses and Limitations

This model can used for Automatic Speech Recognition (ASR) in Catalan. The model is intended to transcribe audio files in Catalan to plain text without punctuation.

How to Get Started with the Model

Installation

In order to use this model, you may install datasets and transformers:

Create a virtual environment:

python -m venv /path/to/venv

Activate the environment:

source /path/to/venv/bin/activate

Install the modules:

pip install datasets transformers 

For Inference

In order to transcribe audio in Catalan using this model, you can follow this example:

import torch
from transformers import WhisperForConditionalGeneration, WhisperProcessor

#Load the processor and model.
MODEL_NAME="projecte-aina/whisper-large-v3-ca-3catparla"
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)

Test Result: 0.96

Training Details

Training data

The specific dataset used to create the model is called "3CatParla".

Training procedure

This model is the result of finetuning the model "openai/whisper-large-v3" by following this tutorial provided by Hugging Face.

Training Hyperparameters

• language: catalan
• hours of training audio: 710
• learning rate: 1.95e-07
• sample rate: 16000
• train batch size: 32 (x4 GPUs)
  • gradient accumulation steps: 1
• eval batch size: 32
• save total limit: 3
• max steps: 19842
• warmup steps: 1984
• eval steps: 3307
• save steps: 3307
• shuffle buffer size: 480

Citation

If this model contributes to your research, please cite the work:

@misc{mena2024whisperlarge3catparla,
      title={Acoustic Model in Catalan: whisper-large-v3-ca-3catparla.}, 
      author={Hernandez Mena, Carlos Daniel; Armentano-Oller, Carme; Solito, Sarah; Külebi, Baybars},
      organization={Barcelona Supercomputing Center},
      url={https://huggingface.co/projecte-aina/whisper-large-v3-ca-3catparla},
      year={2024}
}

Additional Information

Author

The fine-tuning process was perform during July (2024) in the Language Technologies Unit of the Barcelona Supercomputing Center by Carlos Daniel Hernández Mena.

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

Funding

This work has been promoted and financed by the Generalitat de Catalunya through the Aina project.

The training of the model was possible thanks to the compute time provided by Barcelona Supercomputing Center through MareNostrum 5.