Lightweight error mitigation strategies for post-training N:M activation sparsity in LLMs

Open in new window