Merge branch 'main' of https://huggingface.co/Yi3852/MuFun-ACEStep into main
Browse files
README.md
CHANGED
|
@@ -9,6 +9,8 @@ base_model:
|
|
| 9 |
---
|
| 10 |
a prompt generator for the [ACE-Step](https://huggingface.co/ACE-Step/ACE-Step-v1-3.5B) music generation model, fintuned from the MuFun model proposed in [Advancing the Foundation Model for Music Understanding](https://arxiv.org/abs/2508.01178)
|
| 11 |
|
|
|
|
|
|
|
| 12 |
## Usage
|
| 13 |
some audio processing packages like mutagen, torchaudio are needed to be installed
|
| 14 |
```python
|
|
@@ -19,7 +21,7 @@ device='cuda'
|
|
| 19 |
model = AutoModelForCausalLM.from_pretrained(hf_path, trust_remote_code=True, torch_dtype="bfloat16")
|
| 20 |
model.to(device)
|
| 21 |
|
| 22 |
-
aud="/path/to/your/song.
|
| 23 |
inp='<audio>\nDeconstruct this song, listing its tags and lyrics. Directly output a JSON object with prompt and lyrics fields, without any additional explanations or text.'
|
| 24 |
res=model.chat(prompt=inp, audio_files=aud, segs=None, tokenizer=tokenizer)
|
| 25 |
print(res)
|
|
|
|
| 9 |
---
|
| 10 |
a prompt generator for the [ACE-Step](https://huggingface.co/ACE-Step/ACE-Step-v1-3.5B) music generation model, fintuned from the MuFun model proposed in [Advancing the Foundation Model for Music Understanding](https://arxiv.org/abs/2508.01178)
|
| 11 |
|
| 12 |
+
train code: https://github.com/laitselec/MuFun
|
| 13 |
+
|
| 14 |
## Usage
|
| 15 |
some audio processing packages like mutagen, torchaudio are needed to be installed
|
| 16 |
```python
|
|
|
|
| 21 |
model = AutoModelForCausalLM.from_pretrained(hf_path, trust_remote_code=True, torch_dtype="bfloat16")
|
| 22 |
model.to(device)
|
| 23 |
|
| 24 |
+
aud="/path/to/your/song.wav"
|
| 25 |
inp='<audio>\nDeconstruct this song, listing its tags and lyrics. Directly output a JSON object with prompt and lyrics fields, without any additional explanations or text.'
|
| 26 |
res=model.chat(prompt=inp, audio_files=aud, segs=None, tokenizer=tokenizer)
|
| 27 |
print(res)
|