Fix card: AUSteerSteeringController quickstart

README.md

@@ -1,24 +1,58 @@
 ---
 library_name: audio-interv
 tags:
-  - ace-step
+  - activation-steering
   - audio
   - austeer
-  - electronic_music
+  - diffusion
+  - genre
   - interpretability
   - music
+  - steering
+  - tf6tf7
 ---
 
-# AUSteer Scores — `electronic_music` (layers=tf6tf7) — ACE-Step
+# AUSteer — `genre` (ACE-Step)
 
-AUSteer activation-momentum scores for the **electronic_music** concept on ACE-Step. Consumed by `AuSteerController.from_pretrained`. The controller does global top-k feature selection at runtime over the active layer set.
+Per-(step, layer) sparse activation-momentum scores for the **genre** concept on ACE-Step. At inference, `AUSteerSteeringController` adds `alpha` along the top-`k` most concept-discriminative bins.
 
 ## Quickstart
 
 ```python
-from src.steering import SteerableACEModel, get_method
-import src.steering.methods.austeer
+from src.steering import SteerableACEModel, AUSteerSteeringController
 
-model = SteerableACEModel(device="cuda"); model.pipeline.load()
-ctrl = get_method("austeer").from_pretrained("lukasz-staniszewski/ace-step-austeer-electronic-music-tf6tf7", alpha=20)
+model = SteerableACEModel(device="cuda")
+model.pipeline.load()
+ctrl = AUSteerSteeringController.from_pretrained(
+    "lukasz-staniszewski/ace-step-austeer-electronic-music-tf6tf7", 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": "genre",
+  "lyrics": "[inst]",
+  "layers": "tf6tf7",
+  "layers_collected": [
+    "tf6",
+    "tf7"
+  ],
+  "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
+}
 ```