Search papers, labs, and topics across Lattice.
State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, A4 SmoothQuant 69.2% 73.2% 69.6% 40.9% 63.2% -13.3% 4.7 1.
2
0
4
3
Forget retraining for every compression ratio: this auto-regressive feature compressor lets you dial in any ratio you want on the fly.
VLA models can be compressed to 29% of their original VRAM with minimal performance loss by intelligently quantizing different channels based on their impact on action execution.
