Spaces:
Running
Running
File size: 2,963 Bytes
e176226 | 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 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 | // Web Worker: owns the model. Single in-flight inference (main thread enforces it, worker double-checks).
import { pipeline, env, RawImage } from '@huggingface/transformers';
import { MODEL_ID, MODEL_REVISION, MODEL_DTYPE } from './config.js';
env.allowLocalModels = false;
env.allowRemoteModels = true;
env.useBrowserCache = true; // model cached by the browser Cache API after first download
let detector = null;
let device = null;
let busy = false;
function configureWasm(wasmBase, wantWebGPU) {
// Self-hosted ONNX Runtime Web binaries (copied from the pinned npm package) instead of a third-party CDN.
// Plain build for WASM/CPU; asyncify (JSEP) build only when trying WebGPU.
const suffix = wantWebGPU ? '.asyncify' : '';
env.backends.onnx.wasm.wasmPaths = {
mjs: `${wasmBase}ort-wasm-simd-threaded${suffix}.mjs`,
wasm: `${wasmBase}ort-wasm-simd-threaded${suffix}.wasm`,
};
const cores = self.navigator?.hardwareConcurrency || 2;
env.backends.onnx.wasm.numThreads = self.crossOriginIsolated ? Math.min(4, cores) : 1;
}
async function load({ wantWebGPU, wasmBase, fallbackReason = null }) {
configureWasm(wasmBase, wantWebGPU);
const progress_callback = (p) => self.postMessage({ type: 'progress', data: { status: p.status, file: p.file, loaded: p.loaded, total: p.total } });
const tryDevice = async (dev) =>
pipeline('zero-shot-object-detection', MODEL_ID, { revision: MODEL_REVISION, dtype: MODEL_DTYPE, device: dev, progress_callback });
if (wantWebGPU) {
try {
detector = await tryDevice('webgpu');
device = 'webgpu';
} catch (e) {
// ORT's wasm binary cannot be swapped inside this worker; ask the main thread for a fresh WASM worker.
self.postMessage({ type: 'webgpu-failed', reason: String(e?.message || e).slice(0, 200) });
return;
}
} else {
detector = await tryDevice('wasm');
device = 'wasm';
}
self.postMessage({ type: 'ready', device, fallbackReason, threads: env.backends.onnx.wasm.numThreads });
}
async function detect({ id, width, height, buffer, labels, threshold }) {
if (!detector) throw new Error('El modelo no está cargado.');
if (busy) throw new Error('Ya hay una inferencia en curso.');
busy = true;
try {
const rgba = new Uint8ClampedArray(buffer);
const image = new RawImage(rgba, width, height, 4).rgb();
const t0 = performance.now();
const out = await detector(image, labels, { threshold, percentage: false });
const latencyMs = performance.now() - t0;
self.postMessage({ type: 'result', id, detections: out, latencyMs, device });
} finally {
busy = false;
}
}
self.addEventListener('message', async (ev) => {
const msg = ev.data || {};
try {
if (msg.type === 'load') await load(msg);
else if (msg.type === 'detect') await detect(msg);
} catch (e) {
self.postMessage({ type: 'error', id: msg.id, stage: msg.type, message: String(e?.message || e).slice(0, 400) });
}
});
|