Lossless LLM compression for efficient GPU inference via dynamic-length float

📅 2025-04-25    ⚓ Hacker News    🌐 Source    🖼️ Load Image