{"id":4676,"date":"2023-11-20T16:12:07","date_gmt":"2023-11-20T08:12:07","guid":{"rendered":"http:\/\/101.37.83.176\/?p=4676"},"modified":"2024-01-18T15:06:38","modified_gmt":"2024-01-18T07:06:38","slug":"cuda-cudatoolkit-cudnn","status":"publish","type":"post","link":"http:\/\/101.37.83.176\/?p=4676","title":{"rendered":"CUDA\u3001cudatoolkit\u548ccuDNN\u7684\u5173\u7cfb"},"content":{"rendered":"

\u4ec0\u4e48\u662fCUDA\u3001cudatoolkit\u548ccuDNN<\/strong><\/p>\n

CUDA: Compute Unified Device Architecture. \u662fNVIDIA\u7528\u4e8eGPU\u7684\u5e76\u884c\u8ba1\u7b97\u5e73\u53f0\u548c\u7f16\u7a0b\u6a21\u578b\u3002Nvidia\u5b98\u65b9\u63d0\u4f9b\u7684CUDA \u5e93\u662f\u4e00\u4e2a\u5b8c\u6574\u7684\u5de5\u5177\u5b89\u88c5\u5305\uff0c\u5176\u4e2d\u63d0\u4f9b\u4e86 Nvidia\u9a71\u52a8\u7a0b\u5e8f\u3001\u5f00\u53d1 CUDA \u7a0b\u5e8f\u76f8\u5173\u7684\u5f00\u53d1\u5de5\u5177\u5305\u7b49\u53ef\u4f9b\u5b89\u88c5\u7684\u9009\u9879\u3002<\/p>\n

Anaconda \u5b89\u88c5\u7684 cudatoolkit\uff0c\u5176\u4e3b\u8981\u5305\u542b\u5e94\u7528\u7a0b\u5e8f\u5728\u4f7f\u7528 CUDA \u76f8\u5173\u7684\u529f\u80fd\u65f6\u6240\u4f9d\u8d56\u7684\u52a8\u6001\u94fe\u63a5\u5e93\u3002<\/p>\n

cuDNN: GPU-accelerated library of primitives for deep neural networks. \u662fNVIDIA\u7528\u4e8eDNN\u7684GPU\u52a0\u901f\u5e93\u3002<\/p>\n

CUDA \u4e0e cudatoolkit \u7684\u533a\u522b<\/strong><\/p>\n

Nvidia \u5b98\u65b9\u63d0\u4f9b\u5b89\u88c5\u7684 CUDA Toolkit \u5305\u542b\u4e86\u8fdb\u884c CUDA \u76f8\u5173\u7a0b\u5e8f\u5f00\u53d1\u7684\u7f16\u8bd1\u3001\u8c03\u8bd5\u7b49\u8fc7\u7a0b\u76f8\u5173\u7684\u6240\u6709\u7ec4\u4ef6\u3002\u4f46\u5bf9\u4e8e Pytorch \u4e4b\u7c7b\u7684\u6df1\u5ea6\u5b66\u4e60\u6846\u67b6\u800c\u8a00\uff0c\u5176\u5728\u5927\u591a\u6570\u9700\u8981\u4f7f\u7528 GPU \u7684\u60c5\u51b5\u4e2d\u53ea\u9700\u8981\u4f7f\u7528 CUDA \u7684\u52a8\u6001\u94fe\u63a5\u5e93\u652f\u6301\u7a0b\u5e8f\u7684\u8fd0\u884c\uff0c\u5c31\u50cf\u5e38\u89c1\u7684\u53ef\u6267\u884c\u7a0b\u5e8f\u4e00\u6837\uff0c\u4e0d\u9700\u8981\u91cd\u65b0\u8fdb\u884c\u7f16\u8bd1\u8fc7\u7a0b\uff0c\u53ea\u9700\u8981\u5176\u6240\u4f9d\u8d56\u7684\u52a8\u6001\u94fe\u63a5\u5e93\u5b58\u5728\u5373\u53ef\u6b63\u5e38\u8fd0\u884c\u3002\u5728\u5b89\u88c5\u4e86 cudatoolkit \u540e\uff0c\u53ea\u8981\u7cfb\u7edf\u4e0a\u5b58\u5728\u4e0e\u5f53\u524d\u7684 cudatoolkit \u6240\u517c\u5bb9\u7684 Nvidia \u9a71\u52a8\uff0c\u5219\u5df2\u7ecf\u7f16\u8bd1\u597d\u7684 CUDA \u76f8\u5173\u7684\u7a0b\u5e8f\u5c31\u53ef\u4ee5\u76f4\u63a5\u8fd0\u884c\uff0c\u800c\u4e0d\u9700\u8981\u5b89\u88c5\u5b8c\u6574\u7684 Nvidia \u5b98\u65b9\u63d0\u4f9b\u7684 CUDA Toolkit\u3002<\/p>\n

\"\"<\/p>\n

Tensorflow\u5404\u7248\u672c\u5bf9\u5e94\u7684cuda, cudatoolkit, cudnn<\/a><\/strong><\/p>\n\n\n\n\n\n\n\n\n\n\n\n\n\n
\u7248\u672c<\/strong><\/td>\nPython \u7248\u672c<\/strong><\/td>\n\u7f16\u8bd1\u5668<\/strong><\/td>\n\u6784\u5efa\u5de5\u5177<\/strong><\/td>\ncuDNN<\/strong><\/td>\nCUDA<\/strong><\/td>\n<\/tr>\n
tensorflow-2.6.0<\/td>\n3.6-3.9<\/td>\nGCC 7.3.1<\/td>\nBazel 3.7.2<\/td>\n8.1<\/td>\n11.2<\/td>\n<\/tr>\n
tensorflow-2.5.0<\/td>\n3.6-3.9<\/td>\nGCC 7.3.1<\/td>\nBazel 3.7.2<\/td>\n8.1<\/td>\n11.2<\/td>\n<\/tr>\n
tensorflow-2.4.0<\/td>\n3.6-3.8<\/td>\nGCC 7.3.1<\/td>\nBazel 3.1.0<\/td>\n8.0<\/td>\n11.0<\/td>\n<\/tr>\n
tensorflow-2.3.0<\/td>\n3.5-3.8<\/td>\nGCC 7.3.1<\/td>\nBazel 3.1.0<\/td>\n7.6<\/td>\n10.1<\/td>\n<\/tr>\n
tensorflow-2.2.0<\/td>\n3.5-3.8<\/td>\nGCC 7.3.1<\/td>\nBazel 2.0.0<\/td>\n7.6<\/td>\n10.1<\/td>\n<\/tr>\n
tensorflow-2.1.0<\/td>\n2.7\u30013.5-3.7<\/td>\nGCC 7.3.1<\/td>\nBazel 0.27.1<\/td>\n7.6<\/td>\n10.1<\/td>\n<\/tr>\n
tensorflow-2.0.0<\/td>\n2.7\u30013.3-3.7<\/td>\nGCC 7.3.1<\/td>\nBazel 0.26.1<\/td>\n7.4<\/td>\n10.0<\/td>\n<\/tr>\n
tensorflow_gpu-1.15.0<\/td>\n2.7\u30013.3-3.7<\/td>\nGCC 7.3.1<\/td>\nBazel 0.26.1<\/td>\n7.4<\/td>\n10.0<\/td>\n<\/tr>\n
tensorflow_gpu-1.14.0<\/td>\n2.7\u30013.3-3.7<\/td>\nGCC 4.8<\/td>\nBazel 0.24.1<\/td>\n7.4<\/td>\n10.0<\/td>\n<\/tr>\n
tensorflow_gpu-1.13.1<\/td>\n2.7\u30013.3-3.7<\/td>\nGCC 4.8<\/td>\nBazel 0.19.2<\/td>\n7.4<\/td>\n10.0<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n","protected":false},"excerpt":{"rendered":"

CUDA\u3001cudatoolkit\u548ccuDNN\u7684\u5173\u7cfb<\/p>\n","protected":false},"author":1,"featured_media":4678,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"ngg_post_thumbnail":0,"footnotes":""},"categories":[203],"tags":[328,324,325,326,327],"class_list":["post-4676","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-python","tag-ai","tag-cuda","tag-cudatoolkit","tag-cudnn","tag-tensorflow"],"_links":{"self":[{"href":"http:\/\/101.37.83.176\/index.php?rest_route=\/wp\/v2\/posts\/4676","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/101.37.83.176\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/101.37.83.176\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/101.37.83.176\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/101.37.83.176\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=4676"}],"version-history":[{"count":0,"href":"http:\/\/101.37.83.176\/index.php?rest_route=\/wp\/v2\/posts\/4676\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"http:\/\/101.37.83.176\/index.php?rest_route=\/wp\/v2\/media\/4678"}],"wp:attachment":[{"href":"http:\/\/101.37.83.176\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4676"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/101.37.83.176\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4676"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/101.37.83.176\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4676"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}