Scalability and Optimization
Techniques for efficient and large-scale ML systems.
Chapter 24: Distributed Training Data parallelism, model parallelism Frameworks: PyTorch Distributed, TensorFlow TPU Chapter 25: Model Compression Pruning, quantization Knowledge distillation Tools: TensorRT, DeepSparse Chapter 26: Inference Optimization Batch inference, caching Early exiting Hardware accelerators: GPUs, TPUs Hardware-aware model design (New subtopic) Chapter 27: Hardware-Specific Optimizations (New) FPGA and ASIC deployment Specialized ML hardware (beyond GPUs/TPUs) [Tools: Vitis AI, Edge TPU; Custom silicon design] Chapter 28: Cost Optimization for ML Infrastructure (New) ML infrastructure cost modeling and budgeting Cost-aware model selection and deployment strategies [Tools: AWS Cost Explorer, Kubecost; TCO analysis]