Fine-Tuning

Learning Rate

Crucial hyperparameter controlling the magnitude of model weight updates at each iteration, generally requiring a lower value during fine-tuning to preserve pre-learned knowledge.

← Wstecz