Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Vision] Support Phi-3.5-vision, the first VLM in WebLLM #563

Merged
merged 7 commits into from
Sep 23, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions examples/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -36,6 +36,7 @@ These examples demonstrate various capabilities via WebLLM's OpenAI-like API.
- [json-schema](json-schema): besides guaranteeing output to be in JSON, ensure output to adhere to a specific JSON schema specified the user
- [seed-to-reproduce](seed-to-reproduce): use seeding to ensure reproducible output with fields `seed`.
- [function-calling](function-calling) (WIP): function calling with fields `tools` and `tool_choice` (with preliminary support).
- [vision-model](vision-model): process request with image input using Vision Language Model (e.g. Phi3.5-vision)

#### Chrome Extension

Expand Down
14 changes: 14 additions & 0 deletions examples/vision-model/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,14 @@
# WebLLM Get Started App

This folder provides a minimum demo to show WebLLM API in a webapp setting.
To try it out, you can do the following steps under this folder

```bash
npm install
npm start
```

Note if you would like to hack WebLLM core package.
You can change web-llm dependencies as `"file:../.."`, and follow the build from source
instruction in the project to build webllm locally. This option is only recommended
if you would like to hack WebLLM core package.
20 changes: 20 additions & 0 deletions examples/vision-model/package.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,20 @@
{
"name": "get-started",
"version": "0.1.0",
"private": true,
"scripts": {
"start": "parcel src/vision_model.html --port 8888",
"build": "parcel build src/vision_model.html --dist-dir lib"
},
"devDependencies": {
"buffer": "^5.7.1",
"parcel": "^2.8.3",
"process": "^0.11.10",
"tslib": "^2.3.1",
"typescript": "^4.9.5",
"url": "^0.11.3"
},
"dependencies": {
"@mlc-ai/web-llm": "file:../.."
}
}
32 changes: 32 additions & 0 deletions examples/vision-model/src/utils.ts
Original file line number Diff line number Diff line change
@@ -0,0 +1,32 @@
export function getImageDataFromURL(url: string): Promise<ImageData> {
return new Promise((resolve, reject) => {
// Converts img to any, and later `as CanvasImageSource`, otherwise build complains
const img: any = new Image();
img.crossOrigin = "anonymous"; // Important for CORS
img.onload = () => {
const canvas: HTMLCanvasElement = document.createElement("canvas");
const ctx: CanvasRenderingContext2D = canvas.getContext("2d")!;
canvas.width = img.width;
canvas.height = img.height;
ctx.drawImage(img as CanvasImageSource, 0, 0);

const imageData = ctx.getImageData(0, 0, img.width, img.height);
resolve(imageData);
};
img.onerror = () => reject(new Error("Failed to load image"));
img.src = url;
});
}

export async function imageURLToBase64(url: string): Promise<string> {
const imageData: ImageData = await getImageDataFromURL(url);
const canvas = document.createElement("canvas");
const ctx = canvas.getContext("2d");

canvas.width = imageData.width;
canvas.height = imageData.height;

ctx!.putImageData(imageData, 0, 0);

return canvas.toDataURL();
}
23 changes: 23 additions & 0 deletions examples/vision-model/src/vision_model.html
Original file line number Diff line number Diff line change
@@ -0,0 +1,23 @@
<!doctype html>
<html>
<script>
webLLMGlobal = {};
</script>
<body>
<h2>WebLLM Test Page</h2>
Open console to see output
<br />
<br />
<label id="init-label"> </label>

<h3>Prompt</h3>
<label id="prompt-label"> </label>

<h3>Response</h3>
<label id="generate-label"> </label>
<br />
<label id="stats-label"> </label>

<script type="module" src="./vision_model.ts"></script>
</body>
</html>
104 changes: 104 additions & 0 deletions examples/vision-model/src/vision_model.ts
Original file line number Diff line number Diff line change
@@ -0,0 +1,104 @@
import * as webllm from "@mlc-ai/web-llm";
import { imageURLToBase64 } from "./utils";

function setLabel(id: string, text: string) {
const label = document.getElementById(id);
if (label == null) {
throw Error("Cannot find label " + id);
}
label.innerText = text;
}

const proxyUrl = "https://cors-anywhere.herokuapp.com/";
const url_https_street = "https://www.ilankelman.org/stopsigns/australia.jpg";
const url_https_tree = "https://www.ilankelman.org/sunset.jpg";
const url_https_sea =
"https://www.islandvulnerability.org/index/silhouette.jpg";

async function main() {
// can feed request with either base64 or http url
const url_base64_street = await imageURLToBase64(proxyUrl + url_https_street);

const initProgressCallback = (report: webllm.InitProgressReport) => {
setLabel("init-label", report.text);
};
const selectedModel = "Phi-3.5-vision-instruct-q4f16_1-MLC";
const engine: webllm.MLCEngineInterface = await webllm.CreateMLCEngine(
selectedModel,
{
initProgressCallback: initProgressCallback,
logLevel: "INFO", // specify the log level
},
{
context_window_size: 6144,
},
);

// 1. Single image input (with choices)
const messages: webllm.ChatCompletionMessageParam[] = [
{
role: "user",
content: [
{ type: "text", text: "List the items in each image concisely." },
{
type: "image_url",
image_url: {
url: url_base64_street,
},
},
{
type: "image_url",
image_url: {
url: proxyUrl + url_https_sea,
},
},
],
},
];
const request0: webllm.ChatCompletionRequest = {
stream: false, // can be streaming, same behavior
messages: messages,
};
const reply0 = await engine.chat.completions.create(request0);
const replyMessage0 = await engine.getMessage();
console.log(reply0);
console.log(replyMessage0);
console.log(reply0.usage);

// 2. A follow up text-only question
messages.push({ role: "assistant", content: replyMessage0 });
messages.push({ role: "user", content: "What is special about each image?" });
const request1: webllm.ChatCompletionRequest = {
stream: false, // can be streaming, same behavior
messages: messages,
};
const reply1 = await engine.chat.completions.create(request1);
const replyMessage1 = await engine.getMessage();
console.log(reply1);
console.log(replyMessage1);
console.log(reply1.usage);

// 3. A follow up multi-image question
messages.push({ role: "assistant", content: replyMessage1 });
messages.push({
role: "user",
content: [
{ type: "text", text: "What about this image? Answer concisely." },
{
type: "image_url",
image_url: { url: proxyUrl + url_https_tree },
},
],
});
const request2: webllm.ChatCompletionRequest = {
stream: false, // can be streaming, same behavior
messages: messages,
};
const reply2 = await engine.chat.completions.create(request2);
const replyMessage2 = await engine.getMessage();
console.log(reply2);
console.log(replyMessage2);
console.log(reply2.usage);
}

main();
32 changes: 32 additions & 0 deletions src/config.ts
Original file line number Diff line number Diff line change
Expand Up @@ -229,6 +229,7 @@ export function postInitAndCheckGenerationConfigValues(
export enum ModelType {
"LLM",
"embedding",
"VLM", // vision-language model
}

/**
Expand Down Expand Up @@ -512,6 +513,37 @@ export const prebuiltAppConfig: AppConfig = {
context_window_size: 1024,
},
},
// Phi-3.5-vision-instruct
{
model:
"https://huggingface.co/mlc-ai/Phi-3.5-vision-instruct-q4f16_1-MLC",
model_id: "Phi-3.5-vision-instruct-q4f16_1-MLC",
model_lib:
modelLibURLPrefix +
modelVersion +
"/Phi-3.5-vision-instruct-q4f16_1-ctx4k_cs2k-webgpu.wasm",
vram_required_MB: 3952.18,
low_resource_required: true,
overrides: {
context_window_size: 4096,
},
model_type: ModelType.VLM,
},
{
model:
"https://huggingface.co/mlc-ai/Phi-3.5-vision-instruct-q4f32_1-MLC",
model_id: "Phi-3.5-vision-instruct-q4f32_1-MLC",
model_lib:
modelLibURLPrefix +
modelVersion +
"/Phi-3.5-vision-instruct-q4f32_1-ctx4k_cs2k-webgpu.wasm",
vram_required_MB: 5879.84,
low_resource_required: true,
overrides: {
context_window_size: 4096,
},
model_type: ModelType.VLM,
},
// Mistral variants
{
model:
Expand Down
Loading
Loading