Training vs inference

Training is when the model learns from data (expensive, slow). Inference is when it generates responses for users (fast). Most users only see inference.

• Training teaches the model; inference uses it.
• Training is expensive; inference is fast.