Vision Language Models

Index

Symbols

• 4-bit Normal Float (NF4), Training with Quantization, Quantization for VLMs
• π0.6, π0.6: VLAs that learn from experience

A

• A/B mixture testing, Measuring mixture impact
• ablation, Ablation methodology-Measuring mixture impact
• accelerate library, Accelerate
• action chunks, Closing the Perception-Action Loop
• action expert/action head, The action expert
• adapter approach, Connecting the spectrum to modern architectures
• adapter architecture, The Adapter Approach: Cross-Attention-The Unified Sequence Approach: Self-Attention
• affine quantization, Quantization for VLMs
• agentic vision language models
  • coding examples, First toy example
  • computer use agents, Computer Use Agents-Vision browser example
  • connecting models, Connecting your model
  • evaluating agents, Evaluating agents
  • introduction to agents, Agentic Vision Language Models
  • introduction to smolagents, Introduction to Smolagents
  • ReAct loop, The ReAct loop
  • using Hugging Face Spaces as tools, Leveraging Hugging Face Spaces as tools
• alignment, The When: A Model’s Training Stages, Multimodal alignment, Direct Preference Optimization and Mixed Preference Optimization
• annotation, Data Annotation and Quality Validation at Scale-Preparing the Dataset for Consumption
• any-to-any models
  • approaches to, Introducing the Three Approaches
  • characteristics of, Any-to-Any Models
  • hybrid multiobjective models, Hybrid Multiobjective Models-Diffusion pseudocode
  • late-conditioning models, Late-Conditioning Models-Hands On: Qwen-Image-Edit, a Modular Diffusion Editor
  • purpose of ...