Muhammadreza commited on
Commit
e8af36e
1 Parent(s): 8c3b7b9

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +2 -56
README.md CHANGED
@@ -3,60 +3,6 @@ library_name: transformers.js
3
  license: cc-by-nc-4.0
4
  ---
5
 
6
- https://huggingface.co/facebook/musicgen-small with ONNX weights to be compatible with Transformers.js.
7
 
8
- ## Usage (Transformers.js)
9
-
10
- > [!IMPORTANT]
11
- > NOTE: MusicGen support is experimental and requires you to install Transformers.js [v3](https://github.com/xenova/transformers.js/tree/v3) from source.
12
-
13
- If you haven't already, you can install the [Transformers.js](https://huggingface.co/docs/transformers.js) JavaScript library from [GitHub](https://github.com/xenova/transformers.js/tree/v3) using:
14
- ```bash
15
- npm install xenova/transformers.js#v3
16
- ```
17
-
18
- **Example:** Generate music with `Xenova/musicgen-small`.
19
- ```js
20
- import { AutoTokenizer, MusicgenForConditionalGeneration } from '@xenova/transformers';
21
-
22
- // Load tokenizer and model
23
- const tokenizer = await AutoTokenizer.from_pretrained('Xenova/musicgen-small');
24
- const model = await MusicgenForConditionalGeneration.from_pretrained('Xenova/musicgen-small', {
25
- dtype: {
26
- text_encoder: 'q8',
27
- decoder_model_merged: 'q8',
28
- encodec_decode: 'fp32',
29
- },
30
- });
31
-
32
- // Prepare text input
33
- const prompt = 'a light and cheerly EDM track, with syncopated drums, aery pads, and strong emotions bpm: 130';
34
- const inputs = tokenizer(prompt);
35
-
36
- // Generate audio
37
- const audio_values = await model.generate({
38
- ...inputs,
39
- max_new_tokens: 500,
40
- do_sample: true,
41
- guidance_scale: 3,
42
- });
43
-
44
- // (Optional) Write the output to a WAV file
45
- import wavefile from 'wavefile';
46
- import fs from 'fs';
47
-
48
- const wav = new wavefile.WaveFile();
49
- wav.fromScratch(1, model.config.audio_encoder.sampling_rate, '32f', audio_values.data);
50
- fs.writeFileSync('musicgen.wav', wav.toBuffer());
51
- ```
52
-
53
- <audio controls src="https://cdn-uploads.huggingface.co/production/uploads/61b253b7ac5ecaae3d1efe0c/cJ9HFIDstOqJN0eGyJwlS.wav"></audio>
54
-
55
- We also released an online demo, which you can try yourself: https://huggingface.co/spaces/Xenova/musicgen-web
56
-
57
-
58
- <video controls src="https://cdn-uploads.huggingface.co/production/uploads/61b253b7ac5ecaae3d1efe0c/zc43B_VuUVJm4kOJPOHNh.mp4"></video>
59
-
60
- ---
61
-
62
- Note: Having a separate repo for ONNX weights is intended to be a temporary solution until WebML gains more traction. If you would like to make your models web-ready, we recommend converting to ONNX using [🤗 Optimum](https://huggingface.co/docs/optimum/index) and structuring your repo like this one (with ONNX weights located in a subfolder named `onnx`).
 
3
  license: cc-by-nc-4.0
4
  ---
5
 
6
+ # Avasaz ONNX (In Browser model)
7
 
8
+ Based on [Xenova](https://huggingface.co/xeonva/musicgen-small)'s ONNX quantization.