So far we've covered training models from scratch with data. But training an image classifier or text generation model from the ground up every time requires massive data, time, and cost. Often it's far more sensible to use an already well-trained model and adjust it slightly for your problem. SageMaker JumpStart opens this path. Today we cover JumpStart and transfer learning, plus comprehensive training cost optimization.
In the MLA-C01 exam, this topic appears as keywords: "fast deploy of pre-trained models", "transfer learning to build models with little data and cost", "multiple ways to reduce training cost".