Hugging Face LLM Course: Practical Transformers · Lesson 1
Pipeline API: Run Models in Three Lines
Hugging Face pipeline() lets you run NLP tasks without knowing the model's architecture.
Practical exercise
What to do after this lesson
Run three different pipelines (sentiment, ner, summarization) and swap the default model for a more recent Hub model. Benchmark default vs. newer on the same inputs.
Ready-to-use prompt
Template for this lesson
Copy and adapt to your context. Text in angle brackets should be replaced.
Help me pick a pipeline. Task: <…> Language: <…> GPU available: <yes/no> Suggest a pipeline type and 2-3 Hub models.
Common mistakes
What people get wrong
- Forgetting device=0 with a GPU — 10x slower on CPU.
- Using the default English model on non-English text without checking language support.
Pro tips
What works but no one documents
- batch_size=8 speeds up bulk processing.
- return_all_scores=True shows probabilities for all classes.