Hugging Face's logo

LJTSG
/
SmolLM2-360M-webgpu

Text Generation
smollm2
webgpu
browser-inference
strix-halo
amd
unified-memory
tiny-model
Model card Files
xet
 Community
SmolLM2-360M-webgpu
File size: 4,817 Bytes
fd6abf8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<title>SmolLM2-360M — WebGPU</title>
<style>
body { font-family: -apple-system, sans-serif; background: #0a0e14; color: #c9d1d9; max-width: 800px; margin: 0 auto; padding: 20px; }
h1 { color: #58a6ff; font-size: 20px; }
.card { background: #161b22; border: 1px solid #30363d; border-radius: 8px; padding: 16px; margin: 12px 0; }
button { background: #238636; color: white; border: none; border-radius: 6px; padding: 10px 20px; cursor: pointer; font-weight: bold; font-size: 14px; margin: 4px; }
button:disabled { opacity: 0.4; }
#status { color: #e8c87a; font-size: 13px; margin: 8px 0; }
#chat { background: #0d1117; border: 1px solid #30363d; border-radius: 8px; padding: 16px; margin: 12px 0; min-height: 300px; max-height: 500px; overflow-y: auto; }
.msg { margin: 8px 0; padding: 8px 12px; border-radius: 6px; white-space: pre-wrap; line-height: 1.5; }
.user { background: #1f3a5f; color: #e0e8f0; }
.assistant { background: #1a2332; color: #c9d1d9; }
#input-row { display: flex; gap: 8px; margin-top: 8px; }
#input { flex: 1; background: #0d1117; color: #c9d1d9; border: 1px solid #30363d; border-radius: 6px; padding: 10px; font-size: 14px; resize: none; }
.info { color: #8b949e; font-size: 12px; }
</style>
</head>
<body>
<h1>SmolLM2-360M on WebGPU</h1>
<p>HuggingFace's tiny but capable 360M parameter model. Q8_0 (369 MB). Loads in seconds.</p>

<div class="card">
  <button id="btn-load" onclick="doLoad()">Load Model (369 MB)</button>
  <div id="status">Click Load to start</div>
</div>

<div id="chat"></div>
<div id="input-row">
  <textarea id="input" rows="2" placeholder="Ask something..." disabled></textarea>
  <button id="btn-send" onclick="doSend()" disabled>Send</button>
</div>
<p class="info">SmolLM2-360M-Instruct via wllama WebGPU. Built for AMD Strix Halo unified memory.</p>

<script type="module">
import { Wllama } from './node_modules/@wllama/wllama/esm/index.js';

let wllama = null;
const statusEl = document.getElementById('status');
const chatEl = document.getElementById('chat');
const inputEl = document.getElementById('input');
let history = [];

function addMsg(role, text) {
  const div = document.createElement('div');
  div.className = `msg ${role}`;
  div.textContent = text || '';
  chatEl.appendChild(div);
  chatEl.scrollTop = chatEl.scrollHeight;
  return div;
}

window.doLoad = async function() {
  document.getElementById('btn-load').disabled = true;
  statusEl.textContent = 'Loading SmolLM2-360M...';

  wllama = new Wllama(
    { default: './node_modules/@wllama/wllama/esm/wasm/wllama.wasm' },
    { parallelDownloads: 3, logger: {
      debug: () => {},
      log: m => { statusEl.textContent = m; },
      warn: m => console.warn(m),
      error: m => console.error(m),
    }}
  );

  await wllama.loadModelFromUrl(
    window.location.origin + '/model/SmolLM2-360M-Instruct-Q8_0.gguf',
    {
      n_gpu_layers: 99,
      n_ctx: 2048,
      n_batch: 64,
      useCache: true,
      progressCallback: ({ loaded, total }) => {
        const p = Math.round(loaded / total * 100);
        if (p % 10 === 0) statusEl.textContent = `Downloading... ${p}%`;
      },
    }
  );

  statusEl.textContent = 'Ready — SmolLM2-360M on WebGPU';
  inputEl.disabled = false;
  document.getElementById('btn-send').disabled = false;
  inputEl.focus();
};

function buildPrompt() {
  let prompt = '<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n';
  for (const msg of history) {
    prompt += `<|im_start|>${msg.role}\n${msg.content}<|im_end|>\n`;
  }
  prompt += '<|im_start|>assistant\n';
  return prompt;
}

window.doSend = async function() {
  const text = inputEl.value.trim();
  if (!text || !wllama) return;
  inputEl.value = '';
  inputEl.disabled = true;
  document.getElementById('btn-send').disabled = true;

  history.push({ role: 'user', content: text });
  addMsg('user', text);

  const prompt = buildPrompt();
  const genStart = performance.now();

  statusEl.textContent = 'Generating...';

  const result = await wllama.createCompletion({
    prompt,
    max_tokens: 512,
    temperature: 0.7,
    top_k: 40,
    repeat_penalty: 1.1,
    stop: ['<|im_end|>', '<|im_start|>'],
  });

  const rawText = result?.choices?.[0]?.text || result?.text || '';
  const cleanText = rawText.replace('<|im_end|>', '').trim();

  addMsg('assistant', cleanText);
  history.push({ role: 'assistant', content: cleanText });

  const elapsed = (performance.now() - genStart) / 1000;
  statusEl.textContent = `Done — ${elapsed.toFixed(1)}s`;

  inputEl.disabled = false;
  document.getElementById('btn-send').disabled = false;
  inputEl.focus();
};

inputEl.addEventListener('keydown', (e) => {
  if (e.key === 'Enter' && !e.shiftKey) { e.preventDefault(); doSend(); }
});
</script>
</body>
</html>