Tichigadzirira GPU servers neNVIDIA Tesla P40 (24 GB VRAM, Pascal architecture) yekushanda ne open-source LLMs senge Llama, Mistral uye Gemma neOllama. GPU servers hazvigone kupihwa; taura nesu kuti uzive kana zvave.
$ curl -fsSL https://ollama.com/install.sh | sh && ollama run llama3
# Mufananidzo wekuisa command
Ollama isiri- yakaoma chirongwa chekufambisa mamodheru ezvinyorwa zvakakura pakombuta yako. Pa GPU server, mamodheru anofamba pa GPU kwete pa CPU, izvo zvinoita kuti zvinyorwa zvive nyore kuongororwa.
Push Llama, Mistral, Gemma, Phi, Qwen uye mamwe mamodheru neimwe kodhi.
Fits 7B to 13B models comfortably and around 30B models when quantized. 70B models do not fit on one 24 GB card at common quantization levels.
Prompts uye data kugara panzvimbo yako server, pasina chitatu-bato API kufona.
Ollama anopa OpenAI-kuenderana endpoint kuti vakawanda varipo vanoshandisa vanogona kushandisa.
Ollama anogona kushandisa NVIDIA GPU kuti aitedzere basa rake. VPS.org GPU servers hazvina kuwanikwa kuti utore; taura nesu kuti uzive kana vachiri. Ollama ndeye LLM-inference / chat workload. Uchada kuti zvive nyore kurodha pasi kubva kudisk kuenda ku memory uye kukwanisa kushandisa VRAM yemufananidzo pamwe ne context window.
GPU servers hadzina kuwanikwa kuti utore. Nhasi unogona kuisa Ollama pane yakajairika CPU package, uko inoita basa rakakura kupfuura pane GPU: SSH in uye shandisa curl -fsSL https://ollama.com/install.sh | sh && ollama run llama3.
LLM inference VRAM inokwira nemamiriro ezvinhu emhando: 7B model inodiwa 5 kusvika 8 GB yeVRAM uye 13B model inodiwa 10 kusvika 14 GB, asi 70B-class models inodiwa GPUs dzakasiyana-siyana kana kuwanda kwechisimba.
GPU mutengo achaburitswa kana GPU servers anogona kupihwa. Tibate kana iwe uchida kuziva kana ichiitika.
Yega. Every VPS.org server inouya nefull root access, saka unogona kutamba zvimwe zvinhu pamwe Ollama mukati server's RAM uye kuchengetera.
Every VPS.org server inouya nepfupi root SSH kuwanikwa, saka unogona kuisa madhiraivha, mabhuku uye Ollama asi iwe unofanirwa.
Hapana server inouya neCUDA runtime kana NVIDIA mutyairi akaiswa, uye GPU servers hazvina kuwanikwa kuti aite kusvika parizvino. Unogona kuisa zvinyorwa Ollama zvinoda iwe pachako.
VPS.org servers ndezvemapurojekiti akaomarara, kwete ephemeral instances: Ollama models, configs uye data inogara paSSD storage yeserver kusvika wabvisa.
Tenga data rekushandisa pane SSD ye server yekuwedzera kuwanikwa kwenguva pfupi panguva ye Ollama runs, uye kopa zvakapera zviwanikwa (zviyero, zvizvarwa, embeddings) kune yekuchengetera kunze kwe VPS.org; snapshots uye backups vanogara pamwechete kuchengetera cluster.
Unogona kusimudzira chero VPS.org server kune yakakura chirongwa kubva kudzora panel; resize rinotangazve uye yako Ollama kuiswa inotora pamusoro. GPU servers hazvina kuwanikwa kusvika parizvino.
Yeah. Backups ndivo akabhadhara add-on (zuva nezuva + 30% yemari chirongwa, akachengetwa 7 mazuva; zuva nezuva + 20%, akachengetwa 28 mazuva) uye snapshots vakasununguka, izvo zvinokosha pamberi Ollama mazuva. Both stay on the same storage cluster as the server.
Contact support within 30 days of your first purchase to request a refund. GPU servers are not sold yet, so no GPU order is involved.
GPU kugona kuri kuwanikwa chete. Tiudzai nezvebasa rako uye tichakuudza kuti ripi rinowanikwa uye ripi risiri.