feat: switch example models to 16f variants

gsuuon · Mar 12, 2024 · 132be17 · 132be17
1 parent 39c3cb9
commit 132be17
Show file tree

Hide file tree

Showing 8 changed files with 461 additions and 442 deletions.
diff --git a/example/vite-demo/colorgen/app.tsx b/example/vite-demo/colorgen/app.tsx
@@ -122,7 +122,7 @@ const Load = () => {
       when={model()}
       fallback={
         <Loading
-          llamaModel='Llama-2-7b-chat-hf-q4f32_1'
+          llamaModel='Llama-2-7b-chat-hf-q4f16_1'
           onLoad={setModel}
         />
       }>{

diff --git a/example/vite-demo/hn/main.ts b/example/vite-demo/hn/main.ts
@@ -33,7 +33,7 @@ const hnApiGetRandomWhosHiring = async (tries = 2): Promise<string> => {
 
 const result = await renderTemplate(app, async () => {
   const model = await loadModel(
-    'Llama-2-7b-chat-hf-q4f32_1',
+    'Llama-2-7b-chat-hf-q4f16_1',
     report => app.innerHTML = `<pre id='progress'><code>${JSON.stringify(report, null, 2)}</code></pre>`,
     new URLSearchParams(window.location.search).get('cpu') === null
       ? TargetDevice.GPU

diff --git a/example/vite-demo/murder/main.ts b/example/vite-demo/murder/main.ts
@@ -9,7 +9,7 @@ if (import.meta.hot) { import.meta.hot.accept() }
 const infer = document.getElementById('infer')!
 
 const model = await loadModel(
-  'Llama-2-7b-chat-hf-q4f32_1',
+  'Llama-2-7b-chat-hf-q4f16_1',
   report => infer.innerHTML = `<pre id='progress'><code>${JSON.stringify(report, null, 2)}</code></pre>`,
   new URLSearchParams(window.location.search).get('cpu') === null
     ? TargetDevice.GPU

diff --git a/example/vite-demo/package-lock.json b/example/vite-demo/package-lock.json
diff --git a/example/vite-demo/package.json b/example/vite-demo/package.json
@@ -6,7 +6,7 @@
   "scripts": {
     "dev": "vite",
     "build": "tsc && vite build",
-    "preview": "vite preview"
+    "preview": "npm run build && vite preview"
   },
   "devDependencies": {
     "typescript": "^5.0.2",

diff --git a/example/vite-demo/playground/main.ts b/example/vite-demo/playground/main.ts
@@ -45,7 +45,7 @@ onresize.observe(editorContainer)
 
 const model =
   await loadModel(
-    'Llama-2-7b-chat-hf-q4f32_1',
+    'Llama-2-7b-chat-hf-q4f16_1',
     report => inference.innerHTML = `<pre id='progress'><code>${JSON.stringify(report, null, 2)}</code></pre>`,
     new URLSearchParams(window.location.search).get('cpu') === null
       ? TargetDevice.GPU