Performant Deep-learning Inference framework from grounds-up in Nim
AirLLM enables 70B large language model inference on a single 4GB GPU, making large model inference accessible without expensive hardware.
Open-source C++/CUDA inference engine for ternary-weight LLMs with runtime LoRA. Solves the ternary merge problem: LoRA deltas (~1e-5 magnitude) are silently erased when merged into ternary weights...
volesti is the core C++ library of GeomScale for practical volume approximation and random sampling in high dimensions. Features: - Volume approximation for convex polytopes and related convex bod...
Anubhav Nain
GeomScale Project