{"description":"\u4ee5\u524d\uff13\u6708\u306b4\u30d3\u30c3\u30c8\u306e\u6d6e\u52d5\u5c0f\u6570\u70b9\u6570\u306e\u8a71\u984c\u3092\u6319\u3052\u305f\u304c\uff0c\u516c\u5f0f\u306e\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u3082\u5c11\u306a\u304b\u3063\u305f\u306e\u3067\u65ad\u5b9a\u7684\u306a\u66f8\u304d\u65b9\u304c\u3067\u304d\u306a\u304b\u3063\u305f\u3002 fp4\u306e\u8a71\uff084\u30d3\u30c3\u30c8\u6d6e\u52d5\u5c0f\u6570\u70b9\u6570\uff09 - 48's diary \u3044\u304f\u3064\u304b\uff08\u7d20\u4eba\u5411\u3051\u306b\u3082\uff09\u516c\u306b\u306a\u3063\u3066\u304d\u305f\u306e\u3067\u7c21\u5358\u306b\u307e\u3068\u3081\u3066\u304a\u304f\u3002 6\u6708\u306eNVIDIA\u306e\u30c6\u30af\u30cb\u30ab\u30eb\u30d6\u30ed\u30b0\u306b\u3088\u308bNVFP4\u306e\u8aac\u660e Introducing NVFP4 for Efficient and Accurate Low-Precision Inference | NVIDIA Technical Blog FP4 (E2M1)\u306e\u5b9a\u7fa9\u306f\u79c1\u306e\u4e88\u60f3\u901a\u308a\u306e\u3088\u3046\u3067\u3042\u308b\u3002 \u3054\u4e01\u5be7\u306b\u300cFor example, the values\u2026","blog_url":"https://bleu48.hatenablog.com/","url":"https://bleu48.hatenablog.com/entry/2025/09/24/032358","author_name":"the48","provider_url":"https://hatena.blog","blog_title":"48's diary","categories":[],"version":"1.0","title":"MXFP4\u3084NVFP4\u306e\u8a71\uff08\u30d6\u30ed\u30c3\u30af\u6d6e\u52d5\u5c0f\u6570\u70b9\u6570\uff09","width":"100%","html":"<iframe src=\"https://hatenablog-parts.com/embed?url=https%3A%2F%2Fbleu48.hatenablog.com%2Fentry%2F2025%2F09%2F24%2F032358\" title=\"MXFP4\u3084NVFP4\u306e\u8a71\uff08\u30d6\u30ed\u30c3\u30af\u6d6e\u52d5\u5c0f\u6570\u70b9\u6570\uff09 - 48&#39;s diary\" class=\"embed-card embed-blogcard\" scrolling=\"no\" frameborder=\"0\" style=\"display: block; width: 100%; height: 190px; max-width: 500px; margin: 10px 0px;\"></iframe>","type":"rich","author_url":"https://blog.hatena.ne.jp/the48/","provider_name":"Hatena Blog","image_url":null,"published":"2025-09-24 03:23:58","height":"190"}