LLM inference in C/C++
C++ ggml
import { nameOf } from '$lib/utils/model-names';
/** One family of a model list, and the entries it covers. */
export interface ModelFamilyGroup<T> {
entries: T[];
key: string;
label: string;
}
/** Leading letters of a repo name, e.g. the `Qwen` of `Qwen3.8-27B`. */
const FAMILY_LETTERS = /^[A-Za-z]+/;
/** Separators between the segments of a repo name, e.g. the `Qwen3` of `Qwen3-8B`. */
const FAMILY_SEPARATORS = /[-_.]/;
/**
* Family a repo belongs to: the leading letters of its name, so `Qwen3.5` and
* `Qwen3.8-27B` both read as `Qwen`.
*/
export function modelFamilyKey(repo: string): string {
const name = nameOf(repo);
const letters = name.match(FAMILY_LETTERS);
return letters ? letters[0] : name.split(FAMILY_SEPARATORS)[0] || name;
}
/** Fold entries into families, so `Qwen` collects its sizes and variants. */
export function groupModelFamilies<T>(
entries: T[],
modelOf: (entry: T) => string
): ModelFamilyGroup<T>[] {
const families = new Map<string, ModelFamilyGroup<T>>();
for (const entry of entries) {
const key = modelFamilyKey(modelOf(entry));
const family = families.get(key);
if (family) {
family.entries.push(entry);
continue;
}
families.set(key, { entries: [entry], key, label: key });
}
return Array.from(families.values());
}