Updates may affect all model weights or only a small added subset. The data, learning rate, and stopping point are chosen to gain specialization without unnecessarily erasing useful pretrained behavior.
Fine-tuning continues training a pretrained model on a narrower dataset or objective to change its behavior for a target use.
Updates may affect all model weights or only a small added subset. The data, learning rate, and stopping point are chosen to gain specialization without unnecessarily erasing useful pretrained behavior.