"Adaptive Hybrid Quantization Framework for deploying 7B+ LLMs on low-VRAM devices (e.g., GTX 1050). Features surgical block alignment and Numba-accelerated inference.
python machine-learning cuda inference transformer attention numba quantization deep-tech llm gguf low-vram qkv
-
Updated
Dec 24, 2025 - Python