{"id":2289,"date":"2019-12-16T23:27:18","date_gmt":"2019-12-16T20:27:18","guid":{"rendered":"http:\/\/caneroglu.com\/?p=2289"},"modified":"2019-12-17T14:30:53","modified_gmt":"2019-12-17T11:30:53","slug":"cuda-ile-deep-learning-kullanimi","status":"publish","type":"post","link":"https:\/\/caneroglu.com\/?p=2289","title":{"rendered":"CUDA ile Deep Learning Kullan\u0131m\u0131"},"content":{"rendered":"\n

TensorFlow ile Derin \u00d6\u011frenmeye Giri\u015f TensorFlow, Google\u2019\u0131n a\u00e7\u0131k kaynak kodlu makina \u00f6\u011frenmesi k\u00fct\u00fcphanesi ve \u00f6zellikle derin \u00f6\u011frenme i\u00e7in kullan\u0131l\u0131yor.<\/p>\n\n\n\n

GPU\u2019lar\u0131n CPU\u2019nun yapaca\u011f\u0131 i\u015fleri yapmas\u0131 g\u00f6zle g\u00f6r\u00fcl\u00fcr derecede h\u0131z art\u0131\u015f\u0131 sa\u011flar. Bu deste\u011fi Nvidia CUDA ismini verdi\u011fi GPU \u00fczerinde \u00e7al\u0131\u015fmas\u0131n\u0131 sa\u011flayan geli\u015ftirme ara\u00e7lar\u0131 k\u00fcmesi (Toolkit) sayesinde ger\u00e7ekle\u015ftirir. CPU \u00fczerinde ger\u00e7eklemesi zor olan b\u00fcy\u00fck i\u015flemlerde CUDA i\u015flemi daha k\u00fc\u00e7\u00fck par\u00e7alara ay\u0131r\u0131p paralel olarak yapt\u0131\u011f\u0131 i\u00e7in b\u00fcy\u00fck avantaj sa\u011flar.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

CPU vs GPU Hesaplama<\/strong><\/p>\n\n\n\n

CPU (Central Processing Unit)<\/strong><\/h5>\n\n\n\n

CPU, bir bilgisayar program\u0131n\u0131n talimatlar\u0131nda belirtilen temel aritmetik, mant\u0131ksal, kontrol ve giri\u015f\/\u00e7\u0131k\u0131\u015f i\u015flemlerini yapan bilgisayar\u0131n beyni olarak \u00e7al\u0131\u015fan elektronik devredir. Kullanm\u0131\u015f oldu\u011fumuz bilgisayar\u0131n beyni olarak ifade edebiliriz.<\/p>\n\n\n\n

GPU (Graphical Processing Unit)<\/strong><\/h5>\n\n\n\n

GPU, 2D ve 3D grafikleri bir CPU ile birlikte olu\u015fturmak i\u00e7in tasarlanm\u0131\u015f \u00f6zel bir elektronik devredir. GPU, Gamer\u2019lar taraf\u0131ndan oyun k\u00fclt\u00fcr\u00fc i\u00e7in s\u0131kl\u0131kla tercih edilmekte ve kullan\u0131lmaktad\u0131r. G\u00fcn\u00fcm\u00fczde GPU, finansal modelleme, son teknoloji bilimsel ara\u015ft\u0131rma, derin \u00f6\u011frenme, analitik ve petrol ve gaz aramalar\u0131 gibi alanlarda hesaplamal\u0131 i\u015f y\u00fcklerini h\u0131zland\u0131rmak i\u00e7in daha geni\u015f bir \u015fekilde kullan\u0131l\u0131yor.<\/p>\n\n\n\n

G\u00fcn\u00fcm\u00fczde GPU kartlar\u0131n\u0131 NVIDIA, AMD gibi teknoloji devleri \u00fcretmekte ve piyasay\u0131 domine etmektedirler. Intel’de GPU \u00fcretimi ile birlikte son d\u00f6nemde bu paydan par\u00e7a koparmak i\u00e7in \u00e7ok b\u00fcy\u00fck \u00e7aba sarfetmeye ba\u015flad\u0131. <\/p>\n\n\n\n

Bunlar\u0131n d\u0131\u015f\u0131nda TPU (Tensor Processing Unit) <\/strong>ve FPGA (Field Programmable Gate Array) <\/strong>kartlar\u0131n\u0131n y\u00fcksek hesaplamal\u0131 i\u015flemler i\u00e7in kullan\u0131ld\u0131\u011f\u0131n\u0131 bilmenizde fayda var.<\/p>\n\n\n\n

<\/p>\n\n\n\n

2019 Aral\u0131k Ay\u0131 itibari ile Cuda Destekleyen NVIDIA Ekran Kartlar\u0131 Listesi<\/p>\n\n\n\n

Tesla Workstation GPU’lar<\/h3>\n\n\n\n
GPU<\/th>Compute Capability<\/th><\/tr><\/thead>
Tesla K80<\/a><\/td>3.7<\/td><\/tr>
Tesla K40<\/a><\/td>3.5<\/td><\/tr>
Tesla K20<\/a><\/td>3.5<\/td><\/tr>
Tesla C2075<\/a><\/td>2.0<\/td><\/tr>
Tesla C2050\/C2070<\/a><\/td>2.0<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

Tesla Data Center GPU’lar<\/h3>\n\n\n\n
GPU<\/th>Compute Capability<\/th><\/tr><\/thead>
Tesla T4<\/a><\/td>7.5<\/td><\/tr>
Tesla V100<\/a><\/td>7.0<\/td><\/tr>
Tesla P100<\/a><\/td>6.0<\/td><\/tr>
Tesla P40<\/a><\/td>6.1<\/td><\/tr>
Tesla P4<\/a><\/td>6.1<\/td><\/tr>
Tesla M60<\/a><\/td>5.2<\/td><\/tr>
Tesla M40<\/a><\/td>5.2<\/td><\/tr>
Tesla K80<\/a><\/td>3.7<\/td><\/tr>
Tesla K40<\/a><\/td>3.5<\/td><\/tr>
Tesla K20<\/a><\/td>3.5<\/td><\/tr>
Tesla K10<\/a><\/td>3.0<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

Quadro GPU’lar<\/h3>\n\n\n\n
<\/tr>
GPU<\/th>Compute Capability<\/th><\/tr><\/thead>
Quadro RTX 8000<\/a><\/td>7.5<\/td><\/tr>
Quadro RTX 6000<\/a><\/td>7.5<\/td><\/tr>
Quadro RTX 5000<\/a><\/td>7.5<\/td><\/tr>
Quadro RTX 4000<\/a><\/td>7.5<\/td><\/tr>
Quadro GV100<\/a><\/td>7.0<\/td><\/tr>
Quadro GP100<\/a><\/td>6.0<\/td><\/tr>
Quadro P6000<\/a><\/td>6.1<\/td><\/tr>
Quadro P5000<\/a><\/td>6.1<\/td><\/tr>
Quadro P4000<\/a><\/td>6.1<\/td><\/tr>
Quadro P2200<\/a><\/td>6.1<\/td><\/tr>
Quadro P2000<\/a><\/td>6.1<\/td><\/tr>
Quadro P1000<\/a><\/td>6.1<\/td><\/tr>
Quadro P620<\/a><\/td>6.1<\/td><\/tr>
Quadro P600<\/a><\/td>6.1<\/td><\/tr>
Quadro P400<\/a><\/td>6.1<\/td><\/tr>
Quadro M6000 24GB<\/a><\/td>5.2<\/td><\/tr>
Quadro M6000<\/a><\/td>5.2<\/td><\/tr>
Quadro K6000<\/a><\/td>3.5<\/td><\/tr>
Quadro M5000<\/a><\/td>5.2<\/td><\/tr>
Quadro K5200<\/a><\/td>3.5<\/td><\/tr>
Quadro K5000<\/a><\/td>3.0<\/td><\/tr>
Quadro M4000<\/a><\/td>5.2<\/td><\/tr>
Quadro K4200<\/a><\/td>3.0<\/td><\/tr>
Quadro K4000<\/a><\/td>3.0<\/td><\/tr>
Quadro M2000<\/a><\/td>5.2<\/td><\/tr>
Quadro K2200<\/a><\/td>3.0<\/td><\/tr>
Quadro K2000<\/a><\/td>3.0<\/td><\/tr>
Quadro K2000D<\/a><\/td>3.0<\/td><\/tr>
Quadro K1200<\/a><\/td>5.0<\/td><\/tr>
Quadro K620<\/a><\/td>5.0<\/td><\/tr>
Quadro K600<\/a><\/td>3.0<\/td><\/tr>
Quadro K420<\/a><\/td>3.0<\/td><\/tr>
Quadro 410<\/a><\/td>3.0<\/td><\/tr>
Quadro Plex 7000<\/a><\/td>2.0<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

Quadro Mobile GPU’lar<\/h3>\n\n\n\n
<\/tr><\/tr>
GPU<\/th>Compute Capability<\/th><\/tr><\/thead>
RTX 5000<\/a><\/td>7.5<\/td><\/tr>
RTX 4000<\/a><\/td>7.5<\/td><\/tr>
RTX 3000<\/a><\/td>7.5<\/td><\/tr>
T2000<\/a><\/td>7.5<\/td><\/tr>
T1000<\/a><\/td>7.5<\/td><\/tr>
P620<\/a><\/td>6.1<\/td><\/tr>
P520<\/a><\/td>6.1<\/td><\/tr>
Quadro P5200<\/a><\/td>6.1<\/td><\/tr>
Quadro P4200<\/a><\/td>6.1<\/td><\/tr>
Quadro P3200<\/a><\/td>6.1<\/td><\/tr>
Quadro P5000<\/a><\/td>6.1<\/td><\/tr>
Quadro P4000<\/a><\/td>6.1<\/td><\/tr>
Quadro P3000<\/a><\/td>6.1<\/td><\/tr>
Quadro P2000<\/a><\/td>6.1<\/td><\/tr>
Quadro P1000<\/a><\/td>6.1<\/td><\/tr>
Quadro P600<\/a><\/td>6.1<\/td><\/tr>
Quadro P500<\/a><\/td>6.1<\/td><\/tr>
Quadro M5500M<\/a><\/td>5.2<\/td><\/tr>
Quadro M2200<\/a><\/td>5.2<\/td><\/tr>
Quadro M1200<\/a><\/td>5.0<\/td><\/tr>
Quadro M620<\/a><\/td>5.2<\/td><\/tr>
Quadro M520<\/a><\/td>5.0<\/td><\/tr>
Quadro K6000M<\/a><\/td>3.0<\/td><\/tr>
Quadro K5200M<\/a><\/td>3.0<\/td><\/tr>
Quadro K5100M<\/a><\/td>3.0<\/td><\/tr>
Quadro M5000M<\/a><\/td>5.0<\/td><\/tr>
Quadro K500M<\/a><\/td>3.0<\/td><\/tr>
Quadro K4200M<\/a><\/td>3.0<\/td><\/tr>
Quadro K4100M<\/a><\/td>3.0<\/td><\/tr>
Quadro M4000M<\/a><\/td>5.0<\/td><\/tr>
Quadro K3100M<\/a><\/td>3.0<\/td><\/tr>
Quadro M3000M<\/a><\/td>5.0<\/td><\/tr>
Quadro K2200M<\/a><\/td>3.0<\/td><\/tr>
Quadro K2100M<\/a><\/td>3.0<\/td><\/tr>
Quadro M2000M<\/a><\/td>5.0<\/td><\/tr>
Quadro K1100M<\/a><\/td>3.0<\/td><\/tr>
Quadro M1000M<\/a><\/td>5.0<\/td><\/tr>
Quadro K620M<\/a><\/td>5.0<\/td><\/tr>
Quadro K610M<\/a><\/td>3.5<\/td><\/tr>
Quadro M600M<\/a><\/td>5.0<\/td><\/tr>
Quadro K510M<\/a><\/td>3.5<\/td><\/tr>
Quadro M500M<\/a><\/td>5.0<\/td><\/tr><\/tbody><\/table><\/figure>\n","protected":false},"excerpt":{"rendered":"

TensorFlow ile Derin \u00d6\u011frenmeye Giri\u015f TensorFlow, Google\u2019\u0131n a\u00e7\u0131k kaynak kodlu makina \u00f6\u011frenmesi k\u00fct\u00fcphanesi ve \u00f6zellikle derin \u00f6\u011frenme i\u00e7in kullan\u0131l\u0131yor. GPU\u2019lar\u0131n CPU\u2019nun yapaca\u011f\u0131 i\u015fleri yapmas\u0131 g\u00f6zle…<\/p>\n

Devam\u0131n\u0131 okuyunCUDA ile Deep Learning Kullan\u0131m\u0131<\/span><\/a><\/div>\n","protected":false},"author":1,"featured_media":2292,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[1,5],"tags":[85,83,89,84,87,88,86],"class_list":["post-2289","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-genel","category-teknoloji","tag-cuda","tag-deeplearning","tag-gpu","tag-nvidia","tag-quadro","tag-tesla","tag-titan","entry"],"jetpack_featured_media_url":"https:\/\/caneroglu.com\/wp-content\/uploads\/2019\/12\/TitanRTX_KV_Hero_02_v008_HC_2000px.png","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/caneroglu.com\/index.php?rest_route=\/wp\/v2\/posts\/2289"}],"collection":[{"href":"https:\/\/caneroglu.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/caneroglu.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/caneroglu.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/caneroglu.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=2289"}],"version-history":[{"count":8,"href":"https:\/\/caneroglu.com\/index.php?rest_route=\/wp\/v2\/posts\/2289\/revisions"}],"predecessor-version":[{"id":2346,"href":"https:\/\/caneroglu.com\/index.php?rest_route=\/wp\/v2\/posts\/2289\/revisions\/2346"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/caneroglu.com\/index.php?rest_route=\/wp\/v2\/media\/2292"}],"wp:attachment":[{"href":"https:\/\/caneroglu.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=2289"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/caneroglu.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=2289"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/caneroglu.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=2289"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}