Open-Source Gen AI with HuggingFace

The HuggingFace Platform

HuggingFace hosts 500K+ models, 100K+ datasets, and Spaces for demo deployment. Use Colab with free GPUs for inference.

pip install transformers torch diffusers accelerate

HuggingFace Pipelines

Quick inference for common tasks without boilerplate.

from transformers import pipeline

# Sentiment
classifier = pipeline("sentiment-analysis")
result = classifier("I love AI agents!")
print(result)

# NER
ner = pipeline("ner")
print(ner("John works at Microsoft in Redmond."))

# Question Answering
qa = pipeline("question-answering")
result = qa(question="What is AI?", context="AI is the simulation of human intelligence.")
print(result)

# Image generation
from diffusers import StableDiffusionPipeline
pipe = StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5")
image = pipe("A robot writing code").images[0]

Tokenizers

Tokenizers convert text to token IDs and back. Different models use different tokenizers.

from transformers import AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
tokens = tokenizer("Hello, how are you?", return_tensors="pt")
print(tokenizer.decode(tokens["input_ids"][0]))

Transformers & Quantization

Use 8-bit or 4-bit quantization to run large models on limited hardware.

from transformers import BitsAndBytesConfig

quant_config = BitsAndBytesConfig(load_in_4bit=True)
model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-2-7b-chat-hf",
    quantization_config=quant_config
)

Applications

• Token-by-token visualization — See how models generate text
• Meeting Minutes with Whisper — Transcribe audio to text
• Synthetic Data Generation — Use LLMs to create training datasets