| --- |
| library_name: audio-interv |
| tags: |
| - activation-steering |
| - all |
| - audio |
| - austeer |
| - diffusion |
| - interpretability |
| - music |
| - steering |
| - tempo |
| --- |
| |
| # AUSteer — `tempo` (ACE-Step) |
|
|
| Per-(step, layer) sparse activation-momentum scores for the **tempo** concept on ACE-Step. At inference, `AUSteerSteeringController` adds `alpha` along the top-`k` most concept-discriminative bins. |
|
|
| ## Paper |
|
|
| TADA! Tuning Audio Diffusion Models through Activation Steering — [https://huggingface.co/papers/2602.11910](https://huggingface.co/papers/2602.11910) |
|
|
| ## Quickstart |
|
|
| ```python |
| from src.steering import SteerableACEModel, AUSteerSteeringController |
| |
| model = SteerableACEModel(device="cuda") |
| model.pipeline.load() |
| ctrl = AUSteerSteeringController.from_pretrained( |
| "lukasz-staniszewski/ace-step-austeer-tempo-all", alpha=15.0, k=256, mode="additive", |
| ) |
| |
| with model.steer(ctrl): |
| audio = model.generate( |
| prompt="instrumental music", lyrics="[inst]", |
| audio_duration=10.0, infer_step=30, manual_seed=0, |
| ) |
| ``` |
|
|
| ## Generation config |
|
|
| ```json |
| { |
| "method": "austeer", |
| "concept": "tempo", |
| "lyrics": "[inst]", |
| "layers": "all", |
| "layers_collected": [ |
| "tf0", |
| "tf1", |
| "tf2", |
| "tf3", |
| "tf4", |
| "tf5", |
| "tf6", |
| "tf7", |
| "tf8", |
| "tf9", |
| "tf10", |
| "tf11", |
| "tf12", |
| "tf13", |
| "tf14", |
| "tf15", |
| "tf16", |
| "tf17", |
| "tf18", |
| "tf19", |
| "tf20", |
| "tf21", |
| "tf22", |
| "tf23" |
| ], |
| "num_inference_steps": 30, |
| "audio_duration": 30.0, |
| "seed": 10, |
| "guidance_scale": 5.0, |
| "guidance_scale_text": 0.0, |
| "guidance_scale_lyric": 0.0, |
| "guidance_interval": 1.0, |
| "guidance_interval_decay": 0.0 |
| } |
| ``` |
|
|
