Training Config

Definitions

1. The set of configuration values used for a training run, such as learning rate, batch size, optimizer choice, schedule, and data settings. Training config changes can materially alter model behavior and efficiency.

   In plain English: The settings that define how a training run is configured.

   Example: "The regression was traced to a training config change that increased sequence length without adjusting the learning schedule."

Related Terms

• Knowledge Cutoff
• LLM Config
• Max Tokens
• Model Config
• Safety Training
• Supervised
• Training Budget
• Training Efficiency
• Training Loss
• Training Objective
• Training Schedule
• Training Strategy
• Warm Start