TSAIL group
- 1.1k followers
- FIT Building, Tsinghua University, Beijing, China
- https://ml.cs.tsinghua.edu.cn
Pinned Loading
Repositories
- SpargeAttn Public
SpargeAttention: A training-free sparse attention that can accelerate any model inference.
- SageAttention Public
Quantized Attention achieves speedup of 2-5x and 3-11x compared to FlashAttention and xformers, without lossing end-to-end metrics across language, image, and video models.
- EmbodiedActiveDefense Public
Official implementation for "Reinforced Embodied Active Defense: Exploiting Adaptive Interaction for Robust Visual Perception in Adversarial 3D Environments" (TPAMI 2025)
-
- Adaptive-Sparse-Trainer Public
Official implementation for "Pruning Large Language Models with Semi-Structural Adaptive Sparse Training" (AAAI 2025)
- MMTrustEval Public
A toolbox for benchmarking trustworthiness of multimodal large language models (MultiTrust, NeurIPS 2024 Track Datasets and Benchmarks)
- TetraJet-MXFP4Training Public
Pytorch implementation of "Oscillation-Reduced MXFP4 Training for Vision Transformers" on DeiT Model Pre-training