Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,85 @@
|
|
1 |
-
---
|
2 |
-
license: mit
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
base_model:
|
4 |
+
- deepseek-ai/Janus-Pro-1B
|
5 |
+
pipeline_tag: any-to-any
|
6 |
+
library_name: transformers.js
|
7 |
+
tags:
|
8 |
+
- text-to-image
|
9 |
+
- image-to-text
|
10 |
+
- image-text-to-text
|
11 |
+
---
|
12 |
+
|
13 |
+
https://huggingface.co/deepseek-ai/Janus-Pro-1B with ONNX weights to be compatible with Transformers.js.
|
14 |
+
|
15 |
+
## Usage (Transformers.js)
|
16 |
+
|
17 |
+
If you haven't already, you can install the [Transformers.js](https://huggingface.co/docs/transformers.js) JavaScript library from [NPM](https://www.npmjs.com/package/@huggingface/transformers) using:
|
18 |
+
```bash
|
19 |
+
npm i @huggingface/transformers
|
20 |
+
```
|
21 |
+
|
22 |
+
**Example:** Image+text to text
|
23 |
+
|
24 |
+
```js
|
25 |
+
import { AutoProcessor, MultiModalityCausalLM } from "@huggingface/transformers";
|
26 |
+
|
27 |
+
// Load processor and model
|
28 |
+
const model_id = "onnx-community/Janus-Pro-1B-ONNX";
|
29 |
+
const processor = await AutoProcessor.from_pretrained(model_id);
|
30 |
+
const model = await MultiModalityCausalLM.from_pretrained(model_id);
|
31 |
+
|
32 |
+
// Prepare inputs
|
33 |
+
const conversation = [
|
34 |
+
{
|
35 |
+
role: "<|User|>",
|
36 |
+
content: "<image_placeholder>\nConvert the formula into latex code.",
|
37 |
+
images: ["https://huggingface.co/datasets/Xenova/transformers.js-docs/resolve/main/quadratic_formula.png"],
|
38 |
+
},
|
39 |
+
];
|
40 |
+
const inputs = await processor(conversation);
|
41 |
+
|
42 |
+
// Generate response
|
43 |
+
const outputs = await model.generate({
|
44 |
+
...inputs,
|
45 |
+
max_new_tokens: 150,
|
46 |
+
do_sample: false,
|
47 |
+
});
|
48 |
+
|
49 |
+
// Decode output
|
50 |
+
const new_tokens = outputs.slice(null, [inputs.input_ids.dims.at(-1), null]);
|
51 |
+
const decoded = processor.batch_decode(new_tokens, { skip_special_tokens: true });
|
52 |
+
console.log(decoded[0]);
|
53 |
+
```
|
54 |
+
|
55 |
+
**Example:** Text to image
|
56 |
+
|
57 |
+
```js
|
58 |
+
import { AutoProcessor, MultiModalityCausalLM } from "@huggingface/transformers";
|
59 |
+
|
60 |
+
// Load processor and model
|
61 |
+
const model_id = "onnx-community/Janus-Pro-1B-ONNX";
|
62 |
+
const processor = await AutoProcessor.from_pretrained(model_id);
|
63 |
+
const model = await MultiModalityCausalLM.from_pretrained(model_id);
|
64 |
+
|
65 |
+
// Prepare inputs
|
66 |
+
const conversation = [
|
67 |
+
{
|
68 |
+
role: "<|User|>",
|
69 |
+
content: "A stunning princess from kabul in red, white traditional clothing, blue eyes, brown hair",
|
70 |
+
},
|
71 |
+
];
|
72 |
+
const inputs = await processor(conversation, { chat_template: "text_to_image" });
|
73 |
+
|
74 |
+
// Generate response
|
75 |
+
const num_image_tokens = processor.num_image_tokens;
|
76 |
+
const outputs = await model.generate_images({
|
77 |
+
...inputs,
|
78 |
+
min_new_tokens: num_image_tokens,
|
79 |
+
max_new_tokens: num_image_tokens,
|
80 |
+
do_sample: true,
|
81 |
+
});
|
82 |
+
|
83 |
+
// Save the generated image
|
84 |
+
await outputs[0].save("test.png");
|
85 |
+
```
|