Text Generation

Hugging Face 🤗 Pipeline

Jan Kirenz

Table of contents

• Python setup
• Intuition
  • Use Cases
  • Task Variants
• Pipeline example
  • Create pipeline with GPT-2
  • Provide text
  • Make inference
  • Output

Python setup

• You need Anaconda or Miniconda

• Install this virtual environment env-transformers.yml

from transformers import pipeline

Intuition

Use Cases

• Stories Generation

• Code Generation: can help programmers in their repetitive coding tasks.

Task Variants

• Completion Generation Models
  • Given an incomplete sentence, complete it.
  • Continue a story given the first sentences.
  • Provided a code description, generate the code.
• Text-to-Text Generation Models
  • Translation
  • Summarization
  • Text classification
• Inference
  • takes an incomplete text and returns multiple outputs with which the text can be completed.

Pipeline example

Create pipeline with GPT-2

generator = pipeline('text-generation', model='gpt2')

Provide text

my_text = "Hello, I study online media management"

Make inference

generator(my_text, max_length=30, num_return_sequences=3)

Output

[{'generated_text': 'Hello, I study online media management, and while most of my courses focus on the business, I specialize in online development.\n\nIf you know'},
 {'generated_text': 'Hello, I study online media management at a computer-focused high school. One of my favorite exercises is watching and reading my students write things.\n'},
 {'generated_text': 'Hello, I study online media management, so I know that it\'s not the best idea to be a part of the digital environment yourself."\n\n'}]