Simple HuggingFace

The idea behind this short course is to highlight some of the most commonly used functionality provided by Hugging Face in as simple terms as possible.

• Setting up Hugging Face
  • Hugging Face packages
    • Transformers
    • Datasets
    • Accelerate
    • Hugging Face Hub Client
    • Other libraries
  • Logging into Hugging Face hub
  • Environment variables and storage locations
• Model inference with Hugging Face
  • Hugging Face models
  • Transformers pipeline
  • Handling system prompts when using LLMs
    • Setting a system prompt
    • Reusing the system prompt
  • Batch inference
• Model training with Hugging Face
  • Model setup
  • Data loading and Trainer setup
  • Checking model results
  • Customizing Trainer with TrainingArguments
• Multi-GPU training with Accelerate
  • What is Accelerate?
  • Configuring Accelerate
  • Using Trainer with Accelerate
  • Using Accelerate with PyTorch code