Modelpublic

jharshraj/whisper-indian-names-onnx

sourceHugging Faceupdated 2y agoView on Hugging Face

0likes3downloads

Model Card

Quantized ONNX version of jharshraj/whisper-indian-names

This repository contains a quantized ONNX version of the Whisper model fine-tuned for Indian names.

Model Details

—Original model: jharshraj/whisper-indian-names
—Quantization: INT8 dynamic quantization
—Format: ONNX

Usage

python

import onnxruntime as ort
from transformers import WhisperProcessor

# Load processor
processor = WhisperProcessor.from_pretrained('MODEL_PATH')

# Load ONNX sessions
encoder_path = 'MODEL_PATH/encoder_quantized.onnx'
decoder_path = 'MODEL_PATH/decoder_quantized.onnx'
ort_session_encoder = ort.InferenceSession(encoder_path)
ort_session_decoder = ort.InferenceSession(decoder_path)