File size: 1,405 Bytes
adba853
 
 
095e160
adba853
 
 
 
ac2764c
 
 
adba853
ac2764c
53a01ee
ac2764c
 
 
9ac1f4d
ac2764c
271afe2
ac2764c
4e1ae22
b7b84d2
adba853
4e1ae22
 
 
 
f3a146b
ac2764c
 
 
 
 
 
 
adba853
ac2764c
53a01ee
ac2764c
 
 
 
9ac1f4d
ac2764c
271afe2
ac2764c
4e1ae22
adba853
 
4e1ae22
 
 
 
f3a146b
ac2764c
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
---
license: mit
tags:
- audio-generation
---

[Dance Diffusion](https://github.com/Harmonai-org/sample-generator) is now available in 🧨 Diffusers.

## FP32

```python
# !pip install diffusers[torch] accelerate scipy
from diffusers import DiffusionPipeline
import scipy.io

model_id = "harmonai/maestro-150k"
pipeline = DiffusionPipeline.from_pretrained(model_id)
pipeline = pipeline.to("cuda")

audios = pipeline(audio_length_in_s=4.0).audios

# To save locally
for i, audio in enumerate(audios):
    scipy.io.wavfile.write(f"maestro_test_{i}.wav", pipe.unet.sample_rate, audio.transpose())
    
# To dislay in google colab
import IPython.display as ipd
for audio in audios:
    display(ipd.Audio(audio, rate=pipe.unet.sample_rate))
```

## FP16

Faster at a small loss of quality

```python
# !pip install diffusers[torch] accelerate scipy
from diffusers import DiffusionPipeline
import scipy.io
import torch

model_id = "harmonai/maestro-150k"
pipeline = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16)
pipeline = pipeline.to("cuda")

audios = pipeline(audio_length_in_s=4.0).audios

# To save locally
for i, audio in enumerate(audios):
    scipy.io.wavfile.write(f"maestro_test_{i}.wav", pipe.unet.sample_rate, audio.transpose())
    
# To dislay in google colab
import IPython.display as ipd
for audio in audios:
    display(ipd.Audio(audio, rate=pipe.unet.sample_rate))
```