This Whisper model has been fine-tuned specifically for the Tamil language using the Common Voice 11.0 dataset. It is designed to handle tasks such as speech-to-text transcription and language identification, making it suitable for applications where Tamil is a primary language of interest. The fine-tuning process focused on enhancing performance for Tamil, aiming to reduce the error rate in transcriptions and improve general accuracy.

Intended uses & limitations

Intended Uses: Speech-to-text transcription in Tamil

Limitations: May not perform as well on languages or dialects that are not well-represented in the Common Voice dataset. Higher Word Error Rate (WER) in noisy environments or with speakers who have heavy accents not covered in the training data. The model is optimized for Tamil; performance in other languages may be suboptimal.

Training and evaluation data

The training data for this model consists of voice recordings in Tamil from the Mozilla-foundation/Common Voice 11.0 dataset. The dataset is a crowd-sourced collection of transcribed speech, ensuring diversity in terms of speaker accents, age groups, and speech styles.

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: 4000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1753	0.2992	1000	0.2705	51.0174
0.1404	0.5984	2000	0.2368	46.9969
0.1344	0.8977	3000	0.2196	44.5325
0.0947	1.1969	4000	0.2150	43.3196

Framework versions

Transformers 4.45.2
Pytorch 2.4.1+cu121
Datasets 3.0.1
Tokenizers 0.20.1

Example Usage

Here is an example of how to use the model for Tamil speech recognition with Gradio:

python

import gradio as gr
from transformers import pipeline

# Initialize the pipeline with the specified model
pipe = pipeline(model="Lingalingeswaran/whisper-small-ta")

def transcribe(audio):
    # Transcribe the audio file to text
    text = pipe(audio)["text"]
    return text

# Create the Gradio interface

iface = gr.Interface(
    fn=transcribe,
    inputs=gr.Audio(sources=["microphone", "upload"], type="filepath"),
    outputs="text",
    title="Whisper Small Tamil",
    description="Realtime demo for Tamil speech recognition using a fine-tuned Whisper small model.",
)

# Launch the interface
if __name__ == "__main__":
    iface.launch()