New method speeds up AI model compression using math shortcuts

Original: BaKron: Efficient Quantization with Kronecker-Factored Hessians

Source: arxiv.org ↗