Kufulumizitsa kukulitsa gradient, matekinoloje a maphunziro ndi kuphunzitsa ma modeli ndi GPU pa NVIDIA Tesla P40 (24 GB VRAM, Pascal architecture). Ma seva a GPU sapezekapo; tiuzeni kuti mudziwe pamenepo.
$ pip install cuml-cu12 xgboost cudf-cu12
# Mfano wa kukhazikitsa lamulo
Mabuku monga XGBoost ndi CuPy angagwiritse ntchito GPU; NVIDIA RAPIDS amafuna GPUs zatsopano kuposa Pascal P40.
Kuyendetsa gradient kukulitsa mapangidwe a GPU.
NumPy-ogwirizana maudindo amene amayenda pa GPU.
Train neural mafano limodzi ndi classical amene pamodzi pa server.
Kukhazikitsa library zigawo zosowa zanu pipeline.
Machine Learning akhoza kugwiritsa ntchito NVIDIA GPU kuti afulumizitse ntchito yake. VPS.org GPU ma seva si angathe kulandira; tiuzeni kuti mudziwe pamene iwo ndi. Machine Learning ndi GPU-kufulumira ntchito ntchito. onetsetsani kuti mapulogalamu anu ali CUDA thandizo ndi kuti wanu woyendetsa / runtime zigawo kugwirizana zosowa ntchito ntchito kwa Machine Learning.
GPU ma seva si opezeka kuti aperekebe. Tsopano mungagwiritse ntchito Machine Learning pa CPU yachikhalidwe, komwe imagwira ntchito mofulumira kuposa pa GPU: SSH mu ndi kuyendetsa pip install cuml-cu12 xgboost cudf-cu12.
VRAM zosowa kwa Machine Learning zimadalira mtundu wa mtundu ndi gulu; onetsetsani ntchito zamakono za ntchitoyi za anthu amakono.
Kugulitsa kwa GPU kudzatulutsidwa pamene ma seva a GPU angagwiritsidwe ntchito. Tiuzeni ngati mukufuna kumvetsetsa pamene izi zikuchitika.
Yai. Kawirikawiri VPS.org server amachokera ndi wonse root kulowa, kotero inu mukhoza kuyendetsa ena zipangizo mbali Machine Learning m'ma RAM ndi kusungirako server.
Gawo lililonse VPS.org wothandizira amachokera ndi wonse root SSH kulowa, kotero inu mukhoza kukhazikitsa oyendetsa, mabuku ndi Machine Learning koma inu muyenera.
Sichoncho. Palibe wothandizira amafika ndi CUDA runtime kapena NVIDIA woyendetsa preinstalled, ndi GPU othandizira sizilipo kuti aperekebe. Mumakhazikitsa Machine Learning zojambula zofunikira nokha.
Ndikoyenera. VPS.org ma seva ndi olimba olimba, osati opanda pake: Machine Learning mapangidwe, zosankha ndi deta zimakhazikika pa SSD yosungira server mpaka mutachotsa.
Kusunga ntchito deta pa SSD seva kwa kulowa mosavuta pamene Machine Learning kunyamula, ndi kukopera kumaliza artifacts (mawuwo, chiyambi, embeddings) ku osungirako kunja VPS.org; snapshots ndi zosunga zobwezeretsera azikhala pamodzi osungirako gulu.
Mukhoza kuwonjezera iliyonse VPS.org wothandizira kuti malire lalikulu kuchokera ku gulu la kuwongolera; resize imayambanso ndi Machine Learning kukhazikitsa wanu amapita. GPU othandizira si opezekabe.
Yes. Backups ndi wolipira add-on (tsiku ndi tsiku + 30% ya mtengo wa ndondomeko, anasunga masiku 7; tsiku ndi tsiku + 20%, anasunga masiku 28) ndi snapshots ndi ufulu, amene ndi zothandiza pamaso Machine Learning m'mbuyo m'mbuyo.
Mutha kulumikizana ndi thandizo m'masiku 30 a kugula kwanu koyamba kuti mupemphe kuti mubwezere ndalama. GPU servers sizikugulitsidwabe, chifukwa chake palibe GPU order.
Kutha kwa GPU ndi kocheperako. Tiuzeni za ntchito yanu ndipo tidzakuyankha ndi kuchuluka kwa ndalama zomwe zilipo.