A company wants to reduce the energy consumption and compute resources used by its training jobs to improve sustainability. Which actions will decrease energy use and computational cost associated with training? (Choose two.)

Using Amazon SageMaker Debugger to terminate non-converging training jobs saves energy and compute by preventing unnecessary resource consumption on models that won't improve. AWS Trainium instances are purpose-built for deep learning training, offering superior performance per watt and cost-efficiency compared to general-purpose GPUs, directly reducing energy use and computational cost. Amazon SageMaker Ground Truth is for data labeling, not training efficiency. Deploying models with AWS Lambda is for inference, not training, and doesn't directly reduce training costs. While distributed training with PyTorch or TensorFlow can speed up training, it often increases overall compute resource usage and energy consumption rather than decreasing it, as more machines are involved.