Optimizing ResNet18: FPGA And Quantization's Synergy
Tests neural network hyperparameters optimization (hpo) on cifar10 with quantized resnet18 network from a full precision state. There are many options for quantization: Small integers (weights and activations quantized to scaled integer values, e. g. Int8 ), minifloats (e. g. More details regarding this thesis can be found in the research report under hardware acceleration for ai.
Quantization, the process of reducing the neural network parameters into lower precision, optimizes models for the limited computational resources of fpgas. Models like lenet and. Efficient acceleration of deep convolutional neural networks is currently a major focus in edge computing research.
