MMODELYST
People/Xu Han
PPL

Xu Han

Papers (4)
DashAttention: Differentiable and Adaptive Sparse Hierarchical AttentionGAMMA: Global Bit Allocation for Mixed-Precision Models under Arbitrary BudgetsDECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side DevicesFitting Is Not Enough: Smoothness in Extremely Quantized LLMs