feat(llm): testing some local models
This commit is contained in:
@@ -157,6 +157,38 @@ in {
|
||||
];
|
||||
};
|
||||
})
|
||||
(lib.mkIf cfg.ollama.enable {
|
||||
intel-llm = {
|
||||
autoStart = true;
|
||||
image = "intelanalytics/ipex-llm-inference-cpp-xpu:latest";
|
||||
devices = [
|
||||
"/dev/dri:/dev/dri:rwm"
|
||||
];
|
||||
volumes = [
|
||||
"/var/lib/ollama:/models"
|
||||
];
|
||||
environment = {
|
||||
OLLAMA_ORIGINS = "http://192.168.*";
|
||||
SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS = "1";
|
||||
ONEAPI_DEVICE_SELECTOR = "level_zero:0";
|
||||
OLLAMA_HOST = "[::]:11434";
|
||||
no_proxy = "localhost,127.0.0.1";
|
||||
DEVICE = "Arc";
|
||||
OLLAMA_NUM_GPU = "999";
|
||||
ZES_ENABLE_SYSMAN = "1";
|
||||
};
|
||||
cmd = [
|
||||
"/bin/sh"
|
||||
"-c"
|
||||
"/llm/scripts/start-ollama.sh && echo 'Startup script finished, container is now idling.' && sleep infinity"
|
||||
];
|
||||
extraOptions = [
|
||||
"--net=host"
|
||||
"--memory=32G"
|
||||
"--shm-size=16g"
|
||||
];
|
||||
};
|
||||
})
|
||||
];
|
||||
};
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user