FoldQuantVLA: Native Low-Bit Quantization of Vision-Language-Action Models via Consistent Folding
Paper • 2609.24433 • Published • 2
FoldQuantVLA
W4A4/W8A8 quantization of VLA models via consistent folding: paper, checkpoints, calibration data. github.com/cair-vinuni/FoldQuantVLA
-
-
vrfai/GR00T-N1.7-LIBERO-w4a4
Robotics • Updated -
vrfai/GR00T-N1.7-LIBERO-w4a4-res8
Robotics • Updated • 1 -
vrfai/GR00T-N1.6-LIBERO-w4a4
Robotics • 2B • Updated • 31 -
vrfai/GR00T-N1.6-LIBERO-w4a4-res8
Robotics • 2B • Updated • 29 • 1 -
vrfai/GR00T-N1.5-LIBERO-4suite-combined-w4a4
Robotics • 2B • Updated • 28 -
vrfai/GR00T-N1.5-LIBERO-4suite-combined-w4a4-res8
Robotics • 2B • Updated • 31 -
vrfai/pi05-libero-w4a4
Robotics • 2B • Updated • 38 -
vrfai/pi05-libero-w4a4-res8
Robotics • 3B • Updated • 37
hungho77/libero-4suites-calib
Franka Panda • Updated • 339 episodes • 136 • 1Note LIBERO calibration pool (128 obs, seed 0)
hungho77/so101-multitask-calib
SO-101 • Updated • 143 episodes • 146Note SO101 real-robot calibration set