Deberta V3 Nli Onnx Quantized

🚀 DeBERTa-v3-base-onnx-quantized

This model is a quantized version, offering efficient zero - shot classification. It's optimized for CPUs with AVX2 support and easy to integrate into your projects.

🚀 Quick Start

This model has been quantized using the base model: sileod/deberta-v3-base-tasksource-nli. To use this model, you need to have onnxruntime installed on your machine. You can check out my Huggingface Spaces for usage. The source code for the Huggingface Application can be found on GitHub.

📦 Installation

To run this model on your machine, follow these steps:

Step 1: Install dependencies

pip install transformers optimum[onnxruntime]

💻 Usage Examples

Basic Usage

# load libraries
from transformers import AutoTokenizer
from optimum.onnxruntime import ORTModelForSequenceClassification
from optimum.pipelines import pipeline

# load model components
MODEL_ID = "pitangent-ds/deberta-v3-nli-onnx-quantized"
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
model = ORTModelForSequenceClassification.from_pretrained(MODEL_ID)

# load the pipeline
classifier = pipeline("zero-shot-classification", tokenizer=tokenizer, model=model)

# inference
text = "The jacket that I bought is awesome"
candidate_labels = ["positive", "negative"]

results = classifier(text, candidate_labels)

📄 License

This project is licensed under the Apache-2.0 license.