Welcome to DRLX’s documentation!

DRLX (Diffuser Reinforcement Library X) is a library made to simplify the training of diffusion models with reinforcement learning, powered by diffusers and accelerate.

Setup:

• Installation and Setup

Getting Started:

• DRLX Example
  • Custom Prompt Pipeline
  • Custom Reward
  • Training Setup
  • Loading the Model and Performing Inference

Documentation:

• Pipeline
  • Pipeline
  • PromptPipeline
  • PickAPicPrompts
  • ImagenetAnimalPrompts
• Sampling
  • Sampler
  • DDPOSampler
• Denoisers
  • BaseConditionalDenoiser
  • LDMUNet
• Reward Modelling
  • Toy Rewards
  • Aesthetics
  • Pickscore (WIP)
• Trainers
  • BaseTrainer
  • DDPOTrainer

Indices and tables

• Index

• Module Index

• Search Page