Uqeqesho nokusebenza kweemodeli zePyTorch kwi-NVIDIA Tesla P40 (24 GB VRAM, uyilo lwePascal), ngokufikelela okupheleleyo kwengcambu. Abancedisi be-GPU abakwazi ukuthengwa; qhagamshelana nathi ukuze uqinisekiswe xa bekhona.
$ pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124
# Umzekelo wofakelo lomyalelo
I-PyTorch yinkqubo yokufunda enzulu esetyenziswa ngokubanzi. Umncedisi we-GPU uqhuba uqeqesho lwe-PyTorch kunye nokungaqondani kwi-GPU, ekhawulezayo kakhulu kune-CPU yeemodeli ezininzi.
Ufaka umqhubi we-NVIDIA, i-CUDA kunye nokwakha i-PyTorch ofunayo. Khangela ukuba ukwakha i-PyTorch yakho ixhasa i-Pascal GPUs (ukwazi ukulinganisa 6.1).
Indawo yemodeli ezincinci neziphakathi, iLoRA elungelelanisiweyo kunye nobungakanani obuphakathi beqela.
Run Jupyter notebooks on the server for interactive work on the GPU.
Qwalasela indawo engqongileyo ngokuchanekileyo njengoko iprojekti yakho ifuna.
PyTorch inokusebenzisa i NVIDIA GPU ukukhawulezisa umsebenzi wayo. VPS.org GPU abancedisi abafumanekanga ukucela ngoku; qhagamshelana nathi ukuva xa bekhona. PyTorch nguqeqesho / uxinzelelo lomsebenzi olunokulungiswa ngokulula. Icwangcisa imisebenzi ehamba ixesha elide: yenza i snapshot yakho ye VPS rhoqo kwaye ugcine ikopi yesisindo esiqeqeshiweyo ngaphandle kwe VPS.org.
Abancedisi be-GPU abafumanekanga ukucela ngoku. Namhlanje ungaqhuba i-PyTorch kwi-CPU eqhelekileyo, apho isebenza khona ngokucothayo kune-GPU: SSH ngaphakathi kwaye uqhube i-pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124.
Uqeqesho lwe-VRAM lulawulwa yimeko yokuphucula kunye nokuqhuba. Ugqibo olupheleleyo lwe-7B LLM lufuna malunga ne-24 ukuya kwi-48 GB; i-LoRA / i-QLoRA ilungele i-8 ukuya kwi-16 GB.
Ixabiso le-GPU liza kupapashwa xa iiseva ze-GPU zingafunyanwa. Xhulumanisa nathi ukuba ufuna ukuva xa oku kwenzeka.
Ewe. Isincedisi ngasinye se-VPS.org sifika sinee-root access ezipheleleyo, ngoko ungaqhuba ezinye izixhobo kunye ne-PyTorch ngaphakathi kwe-RAM yeseva kunye nokugcina.
Yonke i VPS.org umncedisi ifika nge root SSH unikezelo olupheleleyo, ngoko ungafaka abaqhubi, iincwadi ze PyTorch nokuba ufuna njani.
Hayi. Akukho mncedisi ufika nexesha lokuqhuba le-CUDA okanye umqhubi we-NVIDIA ofakwe ngaphambili, kwaye abancedisi be-GPU abafumanekanga ukucela ngoku. Ufaka uguqulelo lwe-PyTorch olufunayo ngokwakho.
Ewe. Abancedisi be-VPS.org basebenza njengezimatshini ezibonakalayo eziqhubekayo, azikho iziganeko ezikhawulezayo: iimodyuli ze-PyTorch, iinkqubo zokumiselwa kwenkqubo kunye nedata zihlala kwi-SSD yokugcina yeseva de uzicime.
Gcina i-data esebenza kwi-SSD yomncedisi ukufikelela ngokukhawuleza ngexesha lokusebenza kwe-PyTorch, kwaye ukope i-artifacts egqityiweyo (isisindo, u Generations, i-embeddings) ukugcinwa ngaphandle kwe-VPS.org; iifoto ezikhawulezayo kunye nokhuseleko luhlala kwiqela elinye lokugcina.
Unganyusa nayiphi na iseva ye-VPS.org ibe yinkqubo enkulu ukusuka kwiqela lenjongo yolawulo; ubungakanani obutsha buqala kwakhona kwaye ukufaka kwe-PyTorch kugqitha. Iiseva ze-GPU azifumanekanga ngoku.
Ewe. Iindawo zokugcina zihlawulwe zidityaniswe (ngosuku +30% yexabiso lenkqubo, igcinwe iintsuku ezi-7; ngeveki +20%, igcinwe iintsuku ezi-28) kwaye iifoto ezikhawulezayo zikhululekile, eziluncedo phambi kwe PyTorch ende iqhuba. Zonke zihlala kwiqela lokugcina elifanayo njengeseva.
Nxibelelana noncedo ngaphakathi kweentsuku ezingama-30 zokuthengiswa kwakho kokuqala ukuze ucele ubuyiselo. Iiseva ze-GPU azithengwanga, ngoko akukho mfuneko yokuthumela i-GPU.
Ubungakanani be-GPU bumiselwe. Sixelele malunga nobunzima bomsebenzi wakho kwaye siza kuphendula ngokufumaneka nokuthengiswa.