Run 35B MoE models in ~2.5 GB of RAM. Weights stream from SSD; a trained router predicts loads a step ahead. Mac, iPhone, Android, Windows supported
Python
const CHAT_ERROR_KEYS: Record<string, string> = {
"E-SRV-NOTLOAD": "chat.err.notload",
"E-SRV-BUSY": "chat.err.busy",
"E-SRV-CANCEL": "chat.err.cancel",
"E-SRV-WORKER": "chat.err.worker",
"E-SRV-PARAM": "chat.err.param",
"E-MODEL-MISSING": "chat.err.missing",
"E-MEM-LOAD": "chat.err.memload",
"E-GPU-DEVICE": "chat.err.gpu",
"E-SRV-VERSION": "chat.err.version",
};
export function errorNextKey(code: string | null): string {
return (code && CHAT_ERROR_KEYS[code]) || "chat.err.generic";
}
export function hasKnownError(code: string | null): boolean {
return code !== null && code in CHAT_ERROR_KEYS;
}