{"provider_url":"https://hatena.blog","version":"1.0","categories":["MoE"],"blog_url":"https://tadaoyamaoka.hatenablog.com/","type":"rich","height":"190","provider_name":"Hatena Blog","url":"https://tadaoyamaoka.hatenablog.com/entry/2026/05/29/235122","description":"\u524d\u56de\u3001Python\u3067\u5b9f\u88c5\u3057\u305f\u30a8\u30f3\u30b8\u30f3\u30d3\u30eb\u30c9\u30b9\u30af\u30ea\u30d7\u30c8\u3067\u4fdd\u5b58\u3057\u305f.engine\u3092\u8aad\u307f\u8fbc\u3093\u3067\u3001TensorRT\u3067\u63a8\u8ad6\u3059\u308b\u51e6\u7406\u3092C++\u3067\u5b9f\u88c5\u3059\u308b\u3002 \u63a8\u8ad6\u51e6\u7406 TensorRT\u306e\u30e9\u30a4\u30d6\u30e9\u30ea\u306e\u4f7f\u7528\u304c\u30e1\u30a4\u30f3\u3067\u3042\u308b\u3002 \u30c7\u30d5\u30a9\u30eb\u30c8\u3067\u3001CUDA Graph\u3092\u6709\u52b9\u306b\u3057\u3066\u3044\u308b\u3002 CUDA Graph\u306f\u3001\u30d7\u30e9\u30b0\u30a4\u30f3\u3067CUDA\u30ab\u30fc\u30cd\u30eb\u3092\u547c\u3073\u51fa\u3059\u969b\u306e\u30aa\u30fc\u30d0\u30fc\u30d8\u30c3\u30c9\u3092\u524a\u6e1b\u3059\u308b\u4ed5\u7d44\u307f\u3067\u3042\u308b\u3002 infer.cpp #include <NvInfer.h> #include <NvInferPlugin.h> #include <cuda_runtime_api.h> #include <dlfcn.h> #include <a\u2026","width":"100%","published":"2026-05-29 23:51:22","author_url":"https://blog.hatena.ne.jp/TadaoYamaoka/","author_name":"TadaoYamaoka","html":"<iframe src=\"https://hatenablog-parts.com/embed?url=https%3A%2F%2Ftadaoyamaoka.hatenablog.com%2Fentry%2F2026%2F05%2F29%2F235122\" title=\"dropless MoE(Mixture of Experts)\u3092\u8a66\u3059 \u305d\u306e5(\u63a8\u8ad6\u51e6\u7406) - TadaoYamaoka\u306e\u958b\u767a\u65e5\u8a18\" class=\"embed-card embed-blogcard\" scrolling=\"no\" frameborder=\"0\" style=\"display: block; width: 100%; height: 190px; max-width: 500px; margin: 10px 0px;\"></iframe>","image_url":null,"blog_title":"TadaoYamaoka\u306e\u958b\u767a\u65e5\u8a18","title":"dropless MoE(Mixture of Experts)\u3092\u8a66\u3059 \u305d\u306e5(\u63a8\u8ad6\u51e6\u7406)"}