guanming0717··on Show HN: InstinctFlash – High-Performance Serving Runtime for Robotics ModelsRuntime optimizations alone provide a 2–7× speedup with bitwise-exact outputs. SFew-step distillation provides another 4–5× speedup
guanming0717··on Show HN: InstinctFlash – Run 5B world-action models in real time on Jetson Thorayyy was great to meet you :)
guanming0717··on Launch HN: General Instinct (YC P26) – Frontier models on edge devicesYes I did, with other SOTA quant methods like HQQ, AWQ etc. You can find more info in our blog :) https://general-instinct.com/blog/frontier-moe-sub-4-bit
guanming0717··on Launch HN: General Instinct (YC P26) – Frontier models on edge devicesThanks for sharing! I'd love to chat with him. Would you be open to introducing us? :)