Ti n bọ: NVIDIA Tesla P40, 24 GB VRAM

Rọ́ọ̀nù LLaMA lórí GPU VPS

Àwọn ìṣàmúlò-ètò Llama Meta tí a fi pamọ́ fún ara wa nípa NVIDIA Tesla P40 (24 GB VRAM, ààyè Pascal). Àwọn ìṣàmúlò-ètò 8B nípa ìṣàmúlò-ètò ìpéwọ̀n àti àwọn ìṣàmúlò-ètò tí o jú jú lọ nígbà tí a bá fi pamọ́. Àwọn sáà GPU kò lè fi àwọn ààyè pamọ́ lọ́wọ́lọ́wọ́; kan si wa láti gba àṣẹ nígbà tí a bá fi pamọ́.

Ǹwá Sáfíìlì GPU Àwọn Àtòjọ-ẹ̀yàn GPU

$ curl -fsSL https://ollama.com/install.sh | sh && ollama run llama3
# Àṣẹ ìṣàmúlò-ètò àwọn ààtò

Àwọn àwọn ààyè-iṣẹ́ tí a fi pamọ́

Llama ni ẹ̀yàn Meta tí àwọn ìṣàmúlò-ètò ìtàn tí a lè lò. Ìrí Llama lórí àwọn sáà GPU rẹ̀ wẹ́ẹ̀bù àti àwọn àkọsílẹ̀ àìdárà.

LLaMA lórí ààtò GPU VPS.org

Flame 3 8B

Ń rọ́ọ̀nù àwọn àwòrán 8B ní pàtó 24 GB VRAM. Àwòrán 70B kò lè fi pàtó ní pàtó kaadì 24 GB ní àwọn ìpele ìṣàfarawé.

Àwọn Ìṣàfilọ́lẹ̀ Àìdá

Àwọn ìṣàmúlò-ètò rẹ̀ àti àwọn ìdálẹ̀ rẹ̀ tí wọ́n tí wa nínú sáà rẹ̀, láti inú àwọn ipe API àwọn ìṣàmúlò-ètò kẹtà.

Àwọn ìrísí-lẹ́tà ìṣàmúlò-ètò

Lo GGUF láti inú lama.cpp tàbí Ollama, tàbí àwọn ààyè PyTorch àìdájú.

API tí a ṣẹ̀dà

Fi àwọn ìṣàmúlò-ètò pamọ́ láti inú Ollama tàbí àwọn sáà llama.cpp fún àwọn ààtò ìparí iṣẹ́ tí a fẹ́.

Àwọn Ààyè Lòjútó LLaMA

Àwọn ìrànwọ́ AI ìtàn

Àwọn Àkọsílẹ̀

Àwọn ìṣàmúlò-ètò

Àwọn Àkọlé

Àwọn robotì ìrànwọ́

Àwọn Àṣẹ

Ǹwá Sáfíìlì GPU

Àwọn Àtòjọ-ẹ̀yàn GPU

GPUNVIDIA Tesla P40

VRAM24 GB GDDR5

Àwọn àwọ̀ CUDA3,840

FP3212 TFLOPS

INT847 TOPS

Àwọn àwọn àmì-ìwé346 GB/s

Àwọn Ìṣàmúlò-ètòPascal (GP102)

Àwọn Ìjánu-ìṣàmúlò-ètòPCIe

Àwọn Àtòjọ-ẹ̀yàn

Kini LLaMA nípa GPU VPS kan?

LLaMA lè lò NVIDIA GPU láti mú iṣẹ́ rẹ̀ lọ́wọ́lọ́wọ́. Àwọn sáàbù GPU VPS.org kò wa nígbà tí a bá fẹ́; kọ́kọ́ wà láti gbọ́ nígbà tí a bá jẹ́. LLaMA ní iṣẹ́ àìṣàmúlò-ètò GPU-ìṣàmúlò-ètò. Fi àwọn àwọn ìṣàmúlò-ètò rẹ̀ pamọ́ sínú ìfaramọ́ CUDA, àti pé àwọn ìṣàmúlò-ètò àwọn ìṣàmúlò-ètò àwọn ìṣàmúlò-ètò rẹ̀ ní ìṣàmúlò-ètò àwọn ìṣàmúlò-ètò fun LLaMA.

Bawo ni mo ṣe le fi LLaMA pamọ́ lórí GPU VPS kan?

Àwọn sáà GPU kò wà láti fi àwọn àṣẹ pamọ́. Ní ọjọ́ yìí, o lè ráǹdá LLaMA lórí àwọn ààyè-iṣẹ́ CPU tí a tì ní, níbí tí o tí n ṣiṣẹ́ lárá jú GPU lọ: SSH nínù àti ráǹdá curl -fsSL https://ollama.com/install.sh | sh && ollama run llama3.

Ìgbà wo nínú VRAM tí mò fẹ́ fún LLaMA?

VRAM needs for LLaMA depend on the model and batch size; check the project documentation for current figures.

Ìgbà wo nínú àwọn àwọn ààtò GPU fún LLaMA tí a fi kùnà?

Àwọn àwọn ìṣàmúlò-ètò GPU àwọn ìṣàmúlò-ètò GPU àwọn ìṣàmúlò-ètò GPU àwọn ìṣàmúlò-ètò GPU àwọn ìṣàmúlò-ètò GPU àwọn ìṣàmúlò-ètò GPU

Ń lè rọ́ọ̀nù àwọn àwọn ìrísí-lẹ́tà mìíràn nínú LLaMA?

Yes. Every VPS.org server comes with full root access, so you can run other tools alongside LLaMA within the server's RAM and storage.

Ń gbá ìṣẹ̀dà tòójútó nípa LLaMA GPU VPS?

Every VPS.org server comes with full root SSH access, so you can install drivers, libraries and LLaMA however you need.

Does VPS.org preinstall CUDA for LLaMA?

No. No server comes with a CUDA runtime or NVIDIA driver preinstalled, and GPU servers are not available to order yet. You install the versions LLaMA needs yourself.

Ń jẹ́ pé VPS GPU LLaMA mì tí n bọ́ nínú àwọn sáà?

Ya. Àwọn sáà VPS.org ní àwọn másí-lẹ́tà ìṣàfarawé, àti àwọn ààyè-iṣẹ́ àìdára: àwọn àwọn ààyè-iṣẹ́ LLaMA, àwọn ìṣàmúlò-ètò àti àwọn ààyè-iṣẹ́ tí wọ́n tí wa nínú àwọn ààyè-iṣẹ́ SSD sáà titi tí o fi pa wọn.

Njẹ́ mò yẹ́ ki mò fi àwọn ààyè pamọ́ fún iṣẹ́ mì LLaMA?

Keep working data on the server's SSD for fast access during LLaMA runs, and copy finished artifacts (weights, generations, embeddings) to storage outside VPS.org; snapshots and backups stay on the same storage cluster.

Ń lè fi LLaMA GPU VPS mì pọ̀ jú?

O lè yígbà ìṣàfarawé àwọn sáà VPS.org fún àwọn ààyè-iṣẹ́ tó ní ìṣàmúlò-ètò tòòjù láti inú pánẹ́ẹ̀lì ìṣakosó; ìṣàfarawé ìwọ̀n náà ń tún àwọn sáà LLaMA rẹ̀ ṣẹ̀dà. Sáà GPU kò wa níbẹ̀.

Àwọn ìpamọ́ àìpamọ́ tí wa fún VPS GPU mì?

Yes. Backups are a paid add-on (daily +30% of the plan price, kept 7 days; weekly +20%, kept 28 days) and snapshots are free, which is useful before long LLaMA runs. Both stay on the same storage cluster as the server.

Ń lè gbá ẹ̀yàn?

Fi igbẹkẹle ranṣẹ ninú ọjọ́ 30 tí o tí ráwọ́lẹ̀ láti fi beere fun ìgbàgbọ́. Àwọn sáà GPU kò tí a ràwọ́lẹ̀, bẹ́ẹ̀ kò si àwọn àṣẹ GPU tí a tí wa nínú.