Running 8 TurboQuant on Consumer GPUs — 100K Context on RTX 3090, 64K on RTX 4070 🚀 8 Extend LLM context to 100K tokens on consumer GPUs