Complete Hugging Face Transformers Tutorial: Pretrained Models for NLP and Vision

Hugging Face Transformers is the most popular library for state-of-the-art pretrained models. With thousands of models available, you can easily use BERT, GPT, T5, and more for text classification, generation, translation, and computer vision tasks.

Why Hugging Face Transformers?

• Thousands of models: Access to 200,000+ pretrained models
• Multiple frameworks: PyTorch, TensorFlow, JAX support
• Easy to use: Simple pipelines API
• Production ready: Optimized inference
• Active community: Regular updates and new models

• Text classification and NER
• Question answering
• Text generation (GPT, LLaMA)
• Translation
• Image classification and object detection

Installation

pip install transformers

With PyTorch
pip install transformers torch

With TensorFlow
pip install transformers tensorflow

With all extras
pip install transformers[torch,sentencepiece,tokenizers]

Verify
python -c "import transformers; print(transformers.version)"

Quick Start with Pipelines

1. Text Classification

from transformers import pipeline

Sentiment analysis
classifier = pipeline("sentiment-analysis")
result = classifier("I love this product! It's amazing!")
print(result)  # [{'label': 'POSITIVE', 'score': 0.9998}]

Multiple texts
results = classifier([
    "This is great!",
    "This is terrible!",
    "It's okay, nothing special."
])

Custom model
classifier = pipeline(
    "sentiment-analysis",
    model="nlptown/bert-base-multilingual-uncased-sentiment"
)

2. Named Entity Recognition

from transformers import pipeline

ner = pipeline("ner", groupedentities=True)

result = ner("My name is John and I work at Google in New York.")
print(result)
[{'entitygroup': 'PER', 'word': 'John', ...},
{'entitygroup': 'ORG', 'word': 'Google', ...},

{'entitygroup': 'LOC', 'word': 'New York', ...}]

3. Question Answering

from transformers import pipeline

qa = pipeline("question-answering")
context = """
Hugging Face is a company that develops tools for machine learning.
They are best known for the Transformers library and the Hugging Face Hub.
The company was founded in 2016.
"""

result = qa(
    question="When was Hugging Face founded?",
    context=context
)
print(result)  # {'answer': '2016', 'score': 0.98, ...}

4. Text Generation

from transformers import pipeline

GPT-2
generator = pipeline("text-generation", model="gpt2")
result = generator(
    "The future of AI is",
    maxlength=50,

    numreturnsequences=3

)

With better model
generator = pipeline("text-generation", model="meta-llama/Llama-2-7b-hf")
result = generator(
    "Explain machine learning:",
    maxnewtokens=100,

    temperature=0.7
)

from transformers import pipeline

English to French
translator = pipeline("translationentofr")
result = translator("Hello, how are you?")
print(result)  # [{'translationtext': 'Bonjour, comment allez-vous?'}]


Multi-language
translator = pipeline("translation", model="Helsinki-NLP/opus-mt-en-id")
result = translator("Machine learning is fascinating.")

from transformers import pipeline

summarizer = pipeline("summarization")
article = """
Long article text here...
"""

summary = summarizer(
    article,
    maxlength=100,
    minlength=30,

    dosample=False
)
print(summary[0]['summarytext'])