Upload app.js with huggingface_hub

app.js

@@ -1,7 +1,9 @@
 import * as ort from "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.20.1/dist/ort.webgpu.min.js";
 
-const MODEL_URL = "https://huggingface.co/jaswanthsanjay88/mara-small/resolve/main/model_fp16.onnx";
-const TOKENIZER_URL = "https://huggingface.co/jaswanthsanjay88/mara-small/resolve/main/tokenizer.json";
+const BASE_URL = "https://huggingface.co/jaswanthsanjay88/mara-small/resolve/main/";
+const TOKENIZER_URL = BASE_URL + "tokenizer.json";
+const HAS_WEBGPU = typeof navigator.gpu !== "undefined";
+const MODEL_URL = BASE_URL + (HAS_WEBGPU ? "model_fp16.onnx" : "model_int8.onnx");
 const MAX_SEQ = 512;
 const EOS_ID = 0;
 
@@ -110,10 +112,10 @@ async function init() {
   try {
     tokenizer = await loadTokenizer();
     session = await ort.InferenceSession.create(MODEL_URL, {
-      executionProviders: ["webgpu", "wasm"],
+      executionProviders: HAS_WEBGPU ? ["webgpu"] : ["wasm"],
     });
-    provider = session.inputNames && ort.env.webgpu?.available ? "webgpu" : "wasm";
-    statusEl.innerHTML = `ready · backend: <b>${provider}</b>`;
+    provider = HAS_WEBGPU ? "webgpu (fp16)" : "wasm (int8)";
+    statusEl.innerHTML = `ready · backend: <b>${provider}</b> · model: ${HAS_WEBGPU ? "fp16" : "int8"}`;
     goBtn.disabled = false;
   } catch (e) {
     statusEl.textContent = "failed to load model: " + e.message;