Spaces:
Running
Running
File size: 2,479 Bytes
539c560 |
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 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 |
//load Candle Bert Module wasm module
let init, ModelConditionalGeneration;
async function fetchArrayBuffer(url) {
const cacheName = "t5-candle-cache";
const cache = await caches.open(cacheName);
const cachedResponse = await cache.match(url);
if (cachedResponse) {
const data = await cachedResponse.arrayBuffer();
return new Uint8Array(data);
}
const res = await fetch(url, { cache: "force-cache" });
cache.put(url, res.clone());
return new Uint8Array(await res.arrayBuffer());
}
class ConditionalGeneration {
static instance = {};
static async getInstance(weightsURL, tokenizerURL, configURL, modelID) {
if (modelID.includes("quantized")) {
({ default: init, ModelConditionalGeneration } = await import(
"./build/m-quantized.js"
));
} else {
({ default: init, ModelConditionalGeneration } = await import(
"./build/m.js"
));
}
if (!this.instance[modelID]) {
await init();
self.postMessage({ status: "loading", message: "Loading Model" });
const [weightsArrayU8, tokenizerArrayU8, configArrayU8] =
await Promise.all([
fetchArrayBuffer(weightsURL),
fetchArrayBuffer(tokenizerURL),
fetchArrayBuffer(configURL),
]);
this.instance[modelID] = new ModelConditionalGeneration(
weightsArrayU8,
tokenizerArrayU8,
configArrayU8
);
} else {
self.postMessage({ status: "ready", message: "Model Already Loaded" });
}
return this.instance[modelID];
}
}
self.addEventListener("message", async (event) => {
const { weightsURL, tokenizerURL, configURL, modelID, prompt, params } =
event.data;
let {
temperature = 0.0,
seed = 299792458,
repeat_penalty = 1.1,
repeat_last_n = 64,
top_p = 1,
} = { ...params };
try {
self.postMessage({
status: "ready",
message: "Starting T5 Conditional Generation",
});
const model = await ConditionalGeneration.getInstance(
weightsURL,
tokenizerURL,
configURL,
modelID
);
self.postMessage({
status: "decoding",
message: "Decoding Prompt",
});
const output = model.decode({
prompt,
temperature,
seed,
top_p,
repeat_penalty,
repeat_last_n,
});
self.postMessage({
status: "complete",
message: "complete",
output: output,
});
} catch (e) {
self.postMessage({ error: e });
}
});
|