I-GPU iyithuluzi elinamandla lokusheshisa ukubala kumasethi amakhulu wedatha. Kodwa-ke, ukukhubaza i-GPU kwesinye isikhathi kungadingeka ukuze kuthuthukiswe ukusebenza. Isizathu esisodwa esijwayelekile sokukhubaza i-GPU yilapho usebenzisa i-TensorFlow kukhompuyutha ephathekayo noma yedeskithophu enezinsiza ezilinganiselwe. Ukukhubaza i-GPU kungase futhi kuthuthukise ukusebenza uma usebenzisa i-TensorFlow kudivayisi yeselula.
import os os.environ["CUDA_VISIBLE_DEVICES"]="-1"
Umugqa wokuqala ungenisa imojuli ye-os. Umugqa wesibili usetha okuguquguqukayo kwemvelo CUDA_VISIBLE_DEVICES ukuze kube -1. Lokhu kutshela i-TensorFlow ukuthi ingasebenzisi noma yimaphi ama-GPU.
Incwadi kaJupyter
I-Jupyter Notebook iyisicelo sewebhu esikuvumela ukuthi udale futhi wabelane ngamadokhumenti nabanye endaweni yokusebenzisana. Kuhlanganisa izici ezifana nokufometha kombhalo onothile, ukuhlela amakhodi abukhoma, nokukhiqizwa kwamadokhumenti okuzenzakalelayo.
I-Tensorflow
I-TensorFlow ilabhulali yokuhlaziya idatha nokufunda ngomshini. Ikuvumela ukuthi udale amamodeli wangokwezifiso nama-algorithms ukuze usheshise ukuqeqeshwa kwalawo mamodeli. I-TensorFlow ingasetshenziswa kuPython, okwenza kube lula ukuqalisa.
GPU
I-GPU kuPython iyithuluzi elinamandla elingasetshenziswa ukwenza imisebenzi eyahlukene, okuhlanganisa ukucutshungulwa kwezithombe, ukufunda ngomshini, kanye nekhompyutha yesayensi. Ama-GPU ama-microprocessors akhethekile adizayinelwe ukunikezwa kwezithombe ezisebenza kahle kakhulu nokufunda okujulile. I-Python ihlinzeka nge-wrapper elula yomtapo wezincwadi we-CUDA, okwenza kube lula ukufinyelela izinsiza ze-GPU.
