Mixed Sparsity Training: Achieving 4$\times$ FLOP Reduction for Transformer Pretraining

Re-assign community