setfit ag news endpoint LLM Model

Introduction

The SETFIT AG News model is a text classification model fine-tuned using the SetFit approach on the AG News dataset. It achieves high accuracy with minimal training data, outperforming larger models like GPT-3. The model is available for use as an inference endpoint.

Architecture

SetFit (arxiv:2209.11055) is an efficient approach for text classification, allowing for high accuracy with limited labeled data. The model utilizes the sentence-transformers/all-mpnet-base-v2 as its base architecture and has been trained with only 32 samples.

Training

Dataset: AG News
Training Samples: 32 (8 per class)
Test Accuracy: 0.87

The model leverages the SetFit methodology, which is shown to outperform GPT-3 in several tasks despite being significantly smaller.

Guide: Running Locally

To run the model locally, you can interact with the inference endpoint using Python or cURL.

Python Example

Install necessary packages:
```
pip install requests
```

Use the following script to send a request:

import requests as r

ENDPOINT_URL = ""  # URL of your endpoint
HF_TOKEN = ""  # Your Hugging Face token

headers = {
    "Authorization": f"Bearer {HF_TOKEN}",
    "Content-Type": "application/json"
}

payload = { "inputs": "Your input text here" }
response = r.post(ENDPOINT_URL, headers=headers, json=payload)
print(response.json())

cURL Example

curl https://YOURDOMAIN.us-east-1.aws.endpoints.huggingface.cloud \
-X POST \
-d '{"inputs": "Your input text here"}' \
-H "Authorization: Bearer YOUR_HF_TOKEN" \
-H "Content-Type: application/json"

Cloud GPUs

For enhanced performance, consider using cloud GPU services like AWS or Azure to deploy the model.

License

The model is licensed under the MIT License, allowing for open use and distribution with minimal restrictions.

More Related APIs in Text Classification