webllm-llama-3.2-chat

Running

neetnestor commited on Sep 19, 2024

Commit

847e2f0

1 Parent(s): a297744

feat: switch to qwen-2.5 models

Files changed (6) hide show

README.md CHANGED Viewed

@@ -1,5 +1,5 @@
 ---
-title: Webllm Phi 3.5 Chat
 emoji: 💬
 colorFrom: gray
 colorTo: indigo
@@ -8,4 +8,4 @@ pinned: false
 license: apache-2.0
 ---
-A space to chat with Phi 3.5 running locally in your browser, powered by [WebLLM](https://github.com/mlc-ai/web-llm/).

 ---
+title: Webllm Qwen 2.5 Chat
 emoji: 💬
 colorFrom: gray
 colorTo: indigo
 license: apache-2.0
 ---
+A space to chat with Qwen 2.5 running locally in your browser, powered by [WebLLM](https://github.com/mlc-ai/web-llm/).

dist/index.js CHANGED Viewed

The diff for this file is too large to render. See raw diff

index.html CHANGED Viewed

@@ -3,7 +3,7 @@
   <head>
     <meta charset="utf-8" />
     <meta name="viewport" content="width=device-width" />
-    <title>WebLLM Phi 3.5 Chat</title>
     <link rel="stylesheet" href="styles/katex.min.css" />
     <link
       rel="stylesheet"
@@ -14,21 +14,32 @@
   <body>
     <main>
-      <h1>WebLLM Phi 3.5 Chat</h1>
       <p>
-        This space enables AI chat with Phi 3.5 models directly in your local
         browser, empowered by WebLLM.
       </p>
       <h2>Step 1: Configure And Download Model</h2>
       <div class="card vertical">
         <form class="configure-form">
           <!-- Quantization -->
           <div class="form-group">
             <label for="quantization">Quantization:</label>
             <select id="quantization" name="quantization">
-              <option value="q4f16">q4f16</option>
-              <option value="q4f32">q4f32</option>
             </select>
           </div>

   <head>
     <meta charset="utf-8" />
     <meta name="viewport" content="width=device-width" />
+    <title>WebLLM Qwen 2.5 Chat</title>
     <link rel="stylesheet" href="styles/katex.min.css" />
     <link
       rel="stylesheet"
   <body>
     <main>
+      <h1>WebLLM Qwen 2.5 Chat</h1>
       <p>
+        This space enables AI chat with Qwen 2.5 models directly in your local
         browser, empowered by WebLLM.
       </p>
       <h2>Step 1: Configure And Download Model</h2>
       <div class="card vertical">
         <form class="configure-form">
+          <!-- Model Size -->
+          <div class="form-group">
+            <label for="model_size">Model Size:</label>
+            <select id="model_size" name="model_size">
+              <option value="0.5B">0.5B</option>
+              <option value="1.5B">1.5B</option>
+              <option value="3B">3B</option>
+              <option value="7B">7B</option>
+            </select>
+          </div>
           <!-- Quantization -->
           <div class="form-group">
             <label for="quantization">Quantization:</label>
             <select id="quantization" name="quantization">
+              <option value="q4f16_1">q4f16</option>
+              <option value="q4f32_1">q4f32</option>
             </select>
           </div>

package.json CHANGED Viewed

@@ -8,7 +8,7 @@
   "dependencies": {
     "@babel/core": "^7.13.15",
     "@babel/preset-env": "^7.13.15",
-    "@mlc-ai/web-llm": "^0.2.62",
     "katex": "^0.16.11",
     "rehype-highlight": "^7.0.0",
     "rehype-katex": "^7.0.0",

   "dependencies": {
     "@babel/core": "^7.13.15",
     "@babel/preset-env": "^7.13.15",
+    "@mlc-ai/web-llm": "^0.2.63",
     "katex": "^0.16.11",
     "rehype-highlight": "^7.0.0",
     "rehype-katex": "^7.0.0",

src/index.js CHANGED Viewed

@@ -44,6 +44,7 @@ engine.setLogLevel("INFO");
 engine.setInitProgressCallback(updateEngineInitProgressCallback);
 async function initializeWebLLMEngine() {
   const quantization = document.getElementById("quantization").value;
   const context_window_size = parseInt(document.getElementById("context").value);
   const temperature = parseFloat(document.getElementById("temperature").value);
@@ -52,7 +53,7 @@ async function initializeWebLLMEngine() {
   const frequency_penalty = parseFloat(document.getElementById("frequency_penalty").value);
   document.getElementById("download-status").classList.remove("hidden");
-  const selectedModel = `Phi-3.5-mini-instruct-${quantization}_1-MLC`;
   const config = {
     temperature,
     top_p,

 engine.setInitProgressCallback(updateEngineInitProgressCallback);
 async function initializeWebLLMEngine() {
+  const model_size = document.getElementById("model_size").value;
   const quantization = document.getElementById("quantization").value;
   const context_window_size = parseInt(document.getElementById("context").value);
   const temperature = parseFloat(document.getElementById("temperature").value);
   const frequency_penalty = parseFloat(document.getElementById("frequency_penalty").value);
   document.getElementById("download-status").classList.remove("hidden");
+  const selectedModel = `Qwen2.5-${model_size}-Instruct-${quantization}-MLC`;
   const config = {
     temperature,
     top_p,

yarn.lock CHANGED Viewed

@@ -984,10 +984,10 @@
     "@jridgewell/resolve-uri" "^3.1.0"
     "@jridgewell/sourcemap-codec" "^1.4.14"
-"@mlc-ai/web-llm@^0.2.62":
-  version "0.2.62"
-  resolved "https://registry.npmjs.org/@mlc-ai/web-llm/-/web-llm-0.2.62.tgz"
-  integrity sha512-biEYs13nfNjRa4GRVXRsnXLk/yCTslxSH3OjW/4PjoyvZcCtcIe4Aks20WnrMpJMz63ezE6noJF2fkFWStsI0A==
   dependencies:
     loglevel "^1.9.1"

     "@jridgewell/resolve-uri" "^3.1.0"
     "@jridgewell/sourcemap-codec" "^1.4.14"
+"@mlc-ai/web-llm@^0.2.63":
+  version "0.2.63"
+  resolved "https://registry.yarnpkg.com/@mlc-ai/web-llm/-/web-llm-0.2.63.tgz#73d2871081fbcf088ce70ec29f08351b9889981e"
+  integrity sha512-X4xy3cS9Xh/UR//FU07eqcLdeGyMLTwLNF8pbIgJok8IJvzeigLqFXCGktGHf6HLlFRWE7A6+jVCjq1Iw4cGeA==
   dependencies:
     loglevel "^1.9.1"