Run Your First AI Model with Hugging Face's Inference API

1. 1
   Find a Model on the Hub

   Navigate to the Hugging Face Hub. Use the filters on the left to find a model for your task. For this guide, we'll use a text generation model. A good starting point is `distilgpt2`, a smaller version of GPT-2.
2. 2
   Get Your API Token

   Sign up or log in to your Hugging Face account. Go to your profile, then click 'Settings' -> 'Access Tokens'. Create a new token with the 'read' role. Copy this token, as you'll need it to authenticate API requests.
3. 3
   Query the Inference API

   Use a simple cURL command to send a request to the model's API endpoint. Replace `YOUR_TOKEN` with the key you just copied. This runs the model on Hugging Face's infrastructure and returns the result.
4. 4
   Run a Model Locally with `transformers`

   For more control, use the `transformers` library to download and run the model directly in your own environment. This is ideal for development, fine-tuning, and offline use. First, install the library: `pip install transformers torch`.

# This script demonstrates two ways to use Hugging Face models.
# Install required libraries first: pip install requests transformers torch

import os
import requests
import json
from transformers import pipeline

# --- Method 1: Hugging Face Inference API (Cloud) ---

# Get your token from https://huggingface.co/settings/tokens
# It's best to set this as an environment variable
API_TOKEN = os.getenv("HF_API_TOKEN", "YOUR_TOKEN_HERE")
API_URL = "https://api-inference.huggingface.co/models/distilgpt2"

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

def query_api(payload):
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json()

print("--- Querying Inference API ---")
if API_TOKEN == "YOUR_TOKEN_HERE":
    print("Please set your Hugging Face API token to run the API example.")
else:
    api_payload = {"inputs": "My favorite movie is"}
    api_output = query_api(api_payload)
    print(api_output)

print("\n" + "="*30 + "\n")

# --- Method 2: `transformers` library (Local) ---

print("--- Running model locally with transformers ---")
print("Downloading model (this might take a moment)...")

# Initialize the pipeline. This downloads the model on the first run.
generator = pipeline('text-generation', model='distilgpt2')

local_output = generator(
    "My favorite movie is", 
    max_length=30, 
    num_return_sequences=1
)

print("\nLocal generation result:")
print(local_output[0]['generated_text'])