Popular repositories Loading

1. slim slim Public

   SLiM: One-shot Quantized Sparse Plus Low-rank Approximation of LLMs (ICML 2025)

   Python 35 5

2. slope slope Public

   SLoPe: Double-Pruned Sparse Plus Lazy Low-rank Adapter Pretraining - Published at ICLR 2025

   Python 13 1

3. mkor mkor Public

   MKOR: Momentum-Enabled Kronecker-Factor-Based Optimizer Using Rank-1 Updates - Accepted at NeurIPS 2023

   Python 12 1

4. stoicc stoicc Public

   C++ 5

5. patch patch Public

   Officiel code for PATCH: Learnable Tile-level Hybrid Sparsity for LLMs

   Python 5

6. beam beam Public

   BEAM: Blockwise Error Minimization for One-shot Compression of LLMs

   Python 4