2020-03-01ãã1ã¶æéã®è¨äºä¸è¦§
ãããããã¨ã¯è²ã ã¨ããã®ã ãã©ãæ¤åºå¨ã¯æ²¼ãããã®ã§ããã¨ã«ãããã æ¬å½ã¯semantic segmentationã§ãªã«ããããããã¯ã¼ã¯çµãã§ãã¢ãã«ãä½ãããããé¢åããããOpenCVã§ã§ããç¨åº¦ã§æ¤åºå¨ãä½ã£ã¦ããããããã¼ã¿ã»ãããä½ã£ã¦ããããå¦ç¿â¦
ã©ãã«ãããã«ãWindowsã§"deep"ãªããããã¯é¬¼éã¨ããããè¨ã®éã®ããã§ãCUDA 10.1ã§Tensorflow-gpuã使ããã¨ããã¨ãåã¯ãããªãMãããªãã§ãè¦ãããªããæ¹ã§ãããªãã¨ãããªãã https://dev.infohub.cc/tensorflow2-alpha-gpu/#CUDA_101 ä»æ¹ãâ¦
æããããã¨ã«gccãå ¥ã£ã¦ãªãã£ãã®ã§ãMinGWããå ¥ããã yohshiy.blog.fc2.com ãã¨å ¥ããã¤ããã«ãªã£ã¦ããCUDAãå ¥ã£ã¦ãªãã£ããã®ã§ãä½ãèããã«ã¾ãã¯10.2ãå ¥ãã¦ããã qiita.com docs.nvidia.com www.kkaneko.jp
Windowsã§éçºããªããã°ãªããªããªã£ããã©ãã¨ãã£ã¿ã©ãããã£ã¦ãªã£ã¦ãä»æ´emacsã®ãã¼ãã¤ã³ãããæãåºããªãããemacs使ããããã ãã©ãWindowsã§emacsã¯è¨ã®éãªã®ã§VSCã«emacsã®ãã¼ãã¤ã³ã使ããã¨ã«ãããæ¬å½ã¯VSCãªãã¦å°å»ºçãªãã®ã¯ä½¿ãâ¦
ã£ã¦ããããWSLã£ã¦Linuxã®ã«ã¼ãã«å ¥ã£ã¦ãããããLunixç¨ã®NVIDIAã®ãã©ã¤ãå ¥ãã¦ãæå³ãªããï¼ããããWSLã£ã¦GPUèªèããªããã ã£ã¦ã ãããããå èµSSDè²·ã£ã¦ãã¦ãLinuxã¨ãã«ããã¼ãã«ãã¤ã¤ããã°ããWindowsã§tensorflow使ãæãã§ããããé¢â¦
WSLã¯ãªãã¨ãªã使ããããã«ãªã£ãã WSLã§CUDA使ã£ã¦Kerasåããããã®ã§ãCUDAã¨ããWSLã«å ¥ããã Pythonç°å¢èªä½ã¯ä»®æ³ç°å¢ã§åãããªãã¨å¾ã§çæ»ããã®ã§ãä»®æ³ç°å¢ä½¿ããããã«ããã qiita.com WSLã«CUDAãå ¥ããããæ¹ã¯ãããããªäººãããããªãâ¦
ä»æ¥ã¯NVIDIAã®GPUã®ãã©ã¤ããå ¥ãã¦ãCUDAãã¤ã³ã¹ãã¼ã«ãããPythonã®ç°å¢ä½ãã®ã¯ææ¥ã âã®ãã¼ã¸è¦ãã°ãªãã¨ãªãã§ãããRTX2070Superã§ããã©ã¤ãã¯geme readyã«ãã¨ããããã¨NVIDIAã®ãã©ã¤ã使ãã®ã®ã¦ã¼ã¶ç»é²ããªãããã¦ãªãã¨ãã³ã¼ããã²ãâ¦
10å¹´ã»ã©ä½¿ã£ãMacBookProãããããããããªã£ã¦ããããMacã®ããªã·ã¼ããªããæè¿ã¯çª®å±ã«ãªã£ã¦ããã®ã§ãWindowsã§WSLã¨ãããã®ã§ãããã使ã£ã¦ã¿ããã¨ãããã¨ã§ããã¯ã¢ã§ããããã®GPUã®å ¥ã£ã¦ããã½ã³ã³ãè²·ã£ã¦ãã¾ãããæåã¯GPUã®å ¥ã£ããã½â¦