-
Notifications
You must be signed in to change notification settings - Fork 0
/
load.ts
51 lines (41 loc) · 1.45 KB
/
load.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
import * as fs from "fs";
import * as use from "@tensorflow-models/universal-sentence-encoder";
async function loadEmbeddingsFromFile(filePath: string) {
const data = await fs.promises.readFile(filePath);
return JSON.parse(data.toString());
}
async function searchEmbeddings(embeddings: any, query: string) {
const model = await use.load();
const queryEmbedding = await model.embed([query]);
const queryArray = queryEmbedding.arraySync()[0];
let closestMatch = "";
let closestDistance = Infinity;
for (const [key, embedding] of Object.entries(embeddings)) {
const distance = cosineDistance(queryArray, embedding as number[]);
// const distance = cosineDistance(queryArray, embedding);
if (distance < closestDistance) {
closestMatch = key;
closestDistance = distance;
}
}
return closestMatch;
}
function cosineDistance(arr1: number[], arr2: number[]) {
let dotProduct = 0;
let normA = 0;
let normB = 0;
for (let i = 0; i < arr1.length; i++) {
dotProduct += arr1[i] * arr2[i];
normA += arr1[i] ** 2;
normB += arr2[i] ** 2;
}
return 1 - dotProduct / (Math.sqrt(normA) * Math.sqrt(normB));
}
async function main() {
const embeddings = await loadEmbeddingsFromFile("embeddings.json");
// const query = prompt("Enter a search query:");
const query = "convertir a mayusculas"
const closestMatch = await searchEmbeddings(embeddings, query);
console.log("Closest match:", closestMatch);
}
main();