I Ran a 284-Billion-Parameter Model on 3.2GB of RAM — By Streaming It Off an NVMe Drive
中文摘要
该C99项目通过NVMe流式传输，在3.2GB内存的电脑上成功运行了2840亿参数的DeepSeek-V4模型。
English Summary
A C99 project enables running the 284B-parameter DeepSeek-V4 model on 3.2GB RAM by streaming weights directly from an NVMe drive.
Original Excerpt
How a 6,700-line C99 project makes DeepSeek-V4-Flash runnable on a laptop, and what broke along the way Continue reading on Medium »