Activation Compression of Graph Neural Networks using Block-wise Quantization with Improved Variance Minimization

GNN
Re-assign community