|
--- |
|
inference: false |
|
tags: |
|
- musicgen |
|
license: cc-by-nc-4.0 |
|
pipeline_tag: text-to-audio |
|
--- |
|
|
|
# MeditationMusicGen |
|
|
|
This model is a fine-tuned version of facebooks MusicGen. |
|
Refer to https://huggingface.co/facebook/musicgen-small for more details. |
|
|
|
|
|
|
|
## π€ Transformers Usage |
|
|
|
You can run MeditationMusicGen locally with the π€ Transformers library from version 4.31.0 onwards. |
|
|
|
1. First install the π€ [Transformers library](https://github.com/huggingface/transformers) and scipy: |
|
|
|
``` |
|
pip install --upgrade pip |
|
pip install --upgrade transformers scipy |
|
``` |
|
|
|
```python |
|
from transformers import AutoProcessor, MusicgenForConditionalGeneration |
|
import scipy |
|
|
|
processor = AutoProcessor.from_pretrained("facebook/musicgen-small") |
|
model = MusicgenForConditionalGeneration.from_pretrained("bfh-genai/meditation-musicgen") |
|
|
|
relaxing_description = 'Peaceful meditation background sound' |
|
|
|
conditioned_text_input = processor(text=relaxing_description, padding=True, return_tensors="pt") |
|
|
|
audio_value = model.generate(**conditioned_text_input, |
|
do_sample=True, |
|
guidance_scale=3, # Value >1, best results achieved with 3. |
|
max_new_tokens=256 # 256 ^= 5 seconds of audio. |
|
) |
|
|
|
scipy.io.wavfile.write(f"my_audio.wav", rate=32_000, data=audio_values[0, 0].numpy()) |
|
``` |
|
|
|
**License:** Code is released under MIT, model weights are released under CC-BY-NC 4.0. |
|
|
|
|
|
|
|
|