{"id":111,"date":"2014-07-02T23:56:38","date_gmt":"2014-07-02T14:56:38","guid":{"rendered":"https:\/\/www.dogrow.net\/nnet\/?p=111"},"modified":"2025-06-11T22:49:48","modified_gmt":"2025-06-11T13:49:48","slug":"blog9","status":"publish","type":"post","link":"https:\/\/www.dogrow.net\/nnet\/blog9\/","title":{"rendered":"(9) cuda-convnet\u3092\u8a66\u3057\u3066\u307f\u308b"},"content":{"rendered":"<p><span class=\"my_fc_deeppinkBBig\">NVIDIA GTX-650<\/span>\u306e\u8f09\u3063\u305f Linux\u30de\u30b7\u30f3\u304c\u3042\u308b\u306e\u3067 <span class=\"my_fc_deeppinkBBig\">cuda-convnet<\/span> \u306a\u308b GPU\u5b9f\u88c5\u7248 NeuralNet\u30d7\u30ed\u30b0\u30e9\u30e0\u3092\u52d5\u304b\u3057\u3066\u307f\u308b\u3002\u30db\u30fc\u30e0\u30da\u30fc\u30b8\u306f\u3053\u3061\u3089(\u2193)<br \/>\n<a href=\"https:\/\/code.google.com\/p\/cuda-convnet\/\" target=\"_blank\" rel=\"noopener noreferrer\">https:\/\/code.google.com\/p\/cuda-convnet\/<\/a><\/p>\n<p>\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3068\u3057\u3066 <span class=\"my_fc_deeppinkB\">MNIST<\/span> \u306f\u63d0\u4f9b\u3055\u308c\u3066\u3044\u306a\u3044\u306e\u3067&#8230;<br \/>\n\u4eca\u56de\u306f\u540c\u30db\u30fc\u30e0\u30da\u30fc\u30b8\u3067\u63d0\u4f9b\u3055\u308c\u3066\u3044\u308b <span class=\"my_fc_deeppinkB\">CIFAR-10<\/span> \u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u4f7f\u7528\u3059\u308b\u3002<\/p>\n<h1 class=\"my_h\">1. cuda-convnet\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb<\/h1>\n<p>\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u624b\u9806\u306f\u3053\u3061\u3089\u306e\u30da\u30fc\u30b8\u306b\u66f8\u304b\u308c\u3066\u3044\u308b\u3002<br \/>\n<a href=\"https:\/\/code.google.com\/p\/cuda-convnet\/wiki\/Compiling\" target=\"_blank\" rel=\"noopener noreferrer\"><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-112\" src=\"https:\/\/www.dogrow.net\/nnet\/wp-content\/uploads\/2014\/07\/20140702_01.png\" alt=\"20140702_01\" width=\"320\" height=\"319\" srcset=\"https:\/\/www.dogrow.net\/nnet\/wp-content\/uploads\/2014\/07\/20140702_01.png 320w, https:\/\/www.dogrow.net\/nnet\/wp-content\/uploads\/2014\/07\/20140702_01-150x150.png 150w, https:\/\/www.dogrow.net\/nnet\/wp-content\/uploads\/2014\/07\/20140702_01-300x300.png 300w\" sizes=\"auto, (max-width: 320px) 100vw, 320px\" \/><\/a><\/p>\n<p>\u4eca\u56de\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3059\u308b\u30de\u30b7\u30f3\u306b\u306f <span class=\"my_fc_deeppinkB\">CUDA 6.0 <\/span>, <span class=\"my_fc_deeppinkB\">SDK 4.2 <\/span>\u3092\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3057\u3066\u3042\u308b\u3002<br \/>\n<span class=\"my_fc_crimsonB\"> \u3053\u306e\u8fba\u308a\u306e\u30d0\u30fc\u30b8\u30e7\u30f3\u9055\u3044\u306e\u30b4\u30bf\u30b4\u30bf\u306b\uff13\u6642\u9593\u4ee5\u4e0a\u3082\u8cbb\u3084\u3057\u3066\u3057\u307e\u3063\u305f&#8230;<\/span><\/p>\n<h1 class=\"my_h\">2. \u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9<\/h1>\n<p>\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306e\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u306f\u3053\u3061\u3089\u304b\u3089\u3002<br \/>\n<a href=\"https:\/\/code.google.com\/p\/cuda-convnet\/wiki\/Data\" target=\"_blank\" rel=\"noopener noreferrer\"><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-114\" src=\"https:\/\/www.dogrow.net\/nnet\/wp-content\/uploads\/2014\/07\/20140702_02.png\" alt=\"20140702_02\" width=\"320\" height=\"319\" srcset=\"https:\/\/www.dogrow.net\/nnet\/wp-content\/uploads\/2014\/07\/20140702_02.png 320w, https:\/\/www.dogrow.net\/nnet\/wp-content\/uploads\/2014\/07\/20140702_02-150x150.png 150w, https:\/\/www.dogrow.net\/nnet\/wp-content\/uploads\/2014\/07\/20140702_02-300x300.png 300w\" sizes=\"auto, (max-width: 320px) 100vw, 320px\" \/><\/a><\/p>\n<h1 class=\"my_h\">3. \u30d7\u30ed\u30b0\u30e9\u30e0\u3092\u5b9f\u884c<\/h1>\n<p>\u30d7\u30ed\u30b0\u30e9\u30e0\u306e\u5b9f\u884c\u624b\u9806\u306f\u3053\u3061\u3089\u306b\u66f8\u304b\u308c\u3066\u3044\u308b\u3002<br \/>\n<a href=\"https:\/\/code.google.com\/p\/cuda-convnet\/wiki\/TrainingNet\"><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-116\" src=\"https:\/\/www.dogrow.net\/nnet\/wp-content\/uploads\/2014\/07\/20140702_03.png\" alt=\"20140702_03\" width=\"320\" height=\"319\" srcset=\"https:\/\/www.dogrow.net\/nnet\/wp-content\/uploads\/2014\/07\/20140702_03.png 320w, https:\/\/www.dogrow.net\/nnet\/wp-content\/uploads\/2014\/07\/20140702_03-150x150.png 150w, https:\/\/www.dogrow.net\/nnet\/wp-content\/uploads\/2014\/07\/20140702_03-300x300.png 300w\" sizes=\"auto, (max-width: 320px) 100vw, 320px\" \/><\/a><\/p>\n<h1 class=\"my_h\">4. \u5b9f\u884c\u7d50\u679c<\/h1>\n<p>\u4eca\u56de\u306f <span class=\"my_fc_deeppinkB\">10 epochs<\/span> \u6307\u5b9a\u3067\u5b9f\u884c\u3057\u3066\u307f\u305f\u3002<\/p>\n<pre>\r\n[user@linux]$ python convnet.py --data-path=..\/data\/cifar-10\/ --save-path=..\/save --test-range=6 --train-range=1-5 --layer-def=.\/example-layers\/layers-19pct.cfg --layer-params=.\/example-layers\/layer-params-19pct.cfg --data-provider=cifar --test-freq=13\r\n<\/pre>\n<p>10epoch\u76ee\u306e\u7d50\u679c\u306f\u3053\u3093\u306a\u611f\u3058\u306b\u306a\u3063\u305f\u3002<br \/>\n\u30fb<span class=\"my_fc_deeppinkB\">error rate = 0.366100<\/span> (accuracy=73.4%)<br \/>\n\u30fb1epoch\u306e\u51e6\u7406\u306b\u8981\u3059\u308b\u6642\u9593\u306f\u304a\u3088\u305d <span class=\"my_fc_deeppinkB\">40\u79d2<\/span>\u524d\u5f8c <span class=\"my_fc_deeppinkB\">\u2192 octave\u3068\u6bd4\u8f03\u3059\u308b\u3068\u3081\u3061\u3083\u304f\u3061\u3083\u901f\u3044\uff01<\/span><\/p>\n<pre>------------------------------------------------------- \r\nLayer 'conv1' weights[0]: 2.856662e-02 [1.153782e-04] \r\nLayer 'conv1' biases: 9.223316e-03 [3.505556e-06] \r\nLayer 'conv2' weights[0]: 1.143599e-02 [2.939290e-05] \r\nLayer 'conv2' biases: 7.934627e-03 [3.020574e-06] \r\nLayer 'conv3' weights[0]: 1.032112e-02 [2.550705e-05] \r\nLayer 'conv3' biases: 7.986603e-03 [2.903473e-06] \r\nLayer 'fc10' weights[0]: 2.028455e-03 [1.244750e-04] \r\nLayer 'fc10' biases: 2.140154e-01 [1.096559e-04] \r\n-------------------------------------------------------\r\nSaved checkpoint to ..\/save\/ConvNet__2014-07-02_23.33.24\r\n======================================================= (8.310 sec)\r\n10.1... logprob:  1.094434, 0.383200 (6.475 sec)\r\n10.2... logprob:  1.132439, 0.391000 (6.473 sec)\r\n10.3... logprob:  1.077382, 0.378500 (6.475 sec)\r\n10.4... logprob:  1.107526, 0.386300 (6.474 sec)\r\n10.5... logprob:  1.072043, 0.371800 \r\n======================Test output======================\r\nlogprob:  1.077383, 0.366100 \r\n------------------------------------------------------- \r\n<\/pre>\n<p>\u3061\u306a\u307f\u306b\u3046\u3061\u306e GPU\u306e\u6027\u80fd\u306f\u3053\u3093\u306a\u611f\u3058\u3067\u3059\u3002CUDA\u30b3\u30a2\u6570\u306f 384\u500b\u3068\u3057\u3087\u307c\u3044\u3067\u3059&#8230;<\/p>\n<pre>Device 0: \"GeForce GTX 650\"\r\n  CUDA Driver Version \/ Runtime Version          6.0 \/ 6.0\r\n  CUDA Capability Major\/Minor version number:    3.0\r\n  Total amount of global memory:                 1023 MBytes (1073020928 bytes)\r\n  ( 2) Multiprocessors, (192) CUDA Cores\/MP:     384 CUDA Cores\r\n  GPU Clock rate:                                1058 MHz (1.06 GHz)\r\n  Memory Clock rate:                             2500 Mhz\r\n  Memory Bus Width:                              128-bit\r\n  L2 Cache Size:                                 262144 bytes\r\n  Maximum Texture Dimension Size (x,y,z)         1D=(65536), 2D=(65536, 65536), 3D=(4096, 4096, 4096)\r\n  Maximum Layered 1D Texture Size, (num) layers  1D=(16384), 2048 layers\r\n  Maximum Layered 2D Texture Size, (num) layers  2D=(16384, 16384), 2048 layers\r\n  Total amount of constant memory:               65536 bytes\r\n  Total amount of shared memory per block:       49152 bytes\r\n  Total number of registers available per block: 65536\r\n  Warp size:                                     32\r\n  Maximum number of threads per multiprocessor:  2048\r\n  Maximum number of threads per block:           1024\r\n  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)\r\n  Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)\r\n  Maximum memory pitch:                          2147483647 bytes\r\n  Texture alignment:                             512 bytes\r\n  Concurrent copy and kernel execution:          Yes with 1 copy engine(s)\r\n  Run time limit on kernels:                     No\r\n  Integrated GPU sharing Host Memory:            No\r\n  Support host page-locked memory mapping:       Yes\r\n  Alignment requirement for Surfaces:            Yes\r\n  Device has ECC support:                        Disabled\r\n  Device supports Unified Addressing (UVA):      Yes\r\n  Device PCI Bus ID \/ PCI location ID:           1 \/ 0\r\n<\/pre>\n<p>\u8fd1\u65e5\u4e2d\u306b cuda-convnet\u4e0a\u3067 MNIST\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3057\u3066\u307f\u3088\u3046\u3002<\/p>\n<hr class=\"my_hr_bottom\">\n","protected":false},"excerpt":{"rendered":"<p>NVIDIA GTX-650\u306e\u8f09\u3063\u305f Linux\u30de\u30b7\u30f3\u304c\u3042\u308b\u306e\u3067 cuda-convnet \u306a\u308b GPU\u5b9f\u88c5\u7248 NeuralNet\u30d7\u30ed\u30b0\u30e9\u30e0\u3092\u52d5\u304b\u3057\u3066\u307f\u308b\u3002\u30db\u30fc\u30e0\u30da\u30fc\u30b8\u306f\u3053\u3061\u3089(\u2193) https:\/\/code.googl\u2026 <span class=\"read-more\"><a href=\"https:\/\/www.dogrow.net\/nnet\/blog9\/\">\u7d9a\u304d\u3092\u8aad\u3080 &raquo;<\/a><\/span><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8,10,6],"tags":[],"class_list":["post-111","post","type-post","status-publish","format-standard","hentry","category-cuda","category-cuda-convnet","category-6"],"views":4113,"amp_enabled":true,"_links":{"self":[{"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/posts\/111","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/comments?post=111"}],"version-history":[{"count":27,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/posts\/111\/revisions"}],"predecessor-version":[{"id":2541,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/posts\/111\/revisions\/2541"}],"wp:attachment":[{"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/media?parent=111"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/categories?post=111"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/tags?post=111"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}