rlx-web 0.2.18

WebAssembly entry point for RLX — run models in the browser (CPU backend; WebGPU device bring-up)
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

<!doctype html>
<!--
  RLX - versatile ML compiler + runtime.
  Copyright (C) 2026 Eugene Hauptmann, Nataliya Kosmyna.
  SPDX-License-Identifier: MIT OR Apache-2.0

  Browser harness for rlx-web: forward + backward on CPU and (if built with
  --webgpu) WebGPU. Serve with `just serve-web` (ES modules + wasm need http://).
-->
<html lang="en">
  <head>
    <meta charset="utf-8" />
    <meta name="viewport" content="width=device-width, initial-scale=1" />
    <title>RLX in the browser — forward & backward</title>
    <style>
      body { font: 15px/1.5 system-ui, sans-serif; max-width: 52rem; margin: 2.5rem auto; padding: 0 1rem; }
      h1 { font-size: 1.4rem; } h2 { font-size: 1.05rem; margin-top: 1.6rem; }
      .row { margin: 0.35rem 0; }
      .k { color: #666; display: inline-block; min-width: 10rem; vertical-align: top; }
      .v { font-family: ui-monospace, monospace; }
      .ok { color: #0a7; } .warn { color: #c70; } .err { color: #c33; }
      pre { background: #f5f5f5; padding: 0.7rem; border-radius: 6px; overflow-x: auto; white-space: pre-wrap; }
    </style>
  </head>
  <body>
    <nav>
      <a href="./index.html"><b>MLP demo</b></a>
      <a href="./vision-bench.html">Vision bench</a>
      <a href="./cluster.html">Tab cluster</a>
    </nav>
    <h1>RLX in WebAssembly — forward &amp; backward</h1>
    <p>A two-layer MLP (<code>relu(x·W1 + b1)·W2 + b2</code>, MSE loss) is built
       as an RLX IR graph. The <b>backward</b> pass is the autodiff gradient
       graph. The same graphs run on every backend — only the executor differs.</p>

    <div class="row"><span class="k">Status</span><span class="v" id="status">loading…</span></div>
    <div class="row"><span class="k">CPU backend</span><span class="v" id="backend">—</span></div>
    <div class="row"><span class="k">WebGPU</span><span class="v" id="webgpu">—</span></div>
    <div class="row"><span class="k">WebGL2</span><span class="v" id="webgl">—</span></div>

    <h2>CPU — forward</h2>
    <pre id="cpu-fwd">—</pre>
    <h2>CPU — backward (loss + gradients)</h2>
    <pre id="cpu-bwd">—</pre>
    <h2>CPU — training (SGD)</h2>
    <pre id="cpu-train">—</pre>
    <h2>Real transformer — decoder-only (RMSNorm · RoPE · causal attention · SwiGLU)</h2>
    <pre id="tf">—</pre>
    <h2>WebGPU — forward &amp; backward (compute shaders)</h2>
    <pre id="gpu">—</pre>
    <h2>WebGL2 — forward &amp; backward (render-to-texture)</h2>
    <pre id="glout">—</pre>

    <script type="module" src="./index.js"></script>
  </body>
</html>