fix(pipeline): correct hf model id to acestep-v15-xl-sft

The plan used 'ACE-Step/ACE-Step-v1.5-XL-SFT' as a placeholder per spec
§14.1 'verify each preset's actual canonical HF path before the preload
directive is finalised'. Confirmed via HF search: the real repo is
'ACE-Step/acestep-v15-xl-sft' (lowercase, no dot in v1.5).

Updated:
- ace_pipeline._DEFAULT_MODEL_ID
- tests/test_smoke_gpu.py ACE_MODEL_PATH default
- README.md preload_from_hub directive

README.md

@@ -10,7 +10,7 @@ python_version: "3.11"
 suggested_hardware: zero-a10g
 hf_oauth: false
 preload_from_hub:
-  - ACE-Step/ACE-Step-v1.5-XL-SFT *.safetensors,config.json,scheduler/*,vae/*,tokenizer/*
+  - ACE-Step/acestep-v15-xl-sft *.safetensors,config.json,scheduler/*,vae/*,tokenizer/*
   - Qwen/Qwen2.5-7B-Instruct *.safetensors,config.json,tokenizer*
   - facebook/htdemucs_ft *.th
 ---

ace_pipeline.py

@@ -44,7 +44,7 @@ def vram_limit_for(device: str) -> int | None:
 
 
 _PIPELINE = None  # module-level lazy singleton
-_DEFAULT_MODEL_ID = "ACE-Step/ACE-Step-v1.5-XL-SFT"
+_DEFAULT_MODEL_ID = "ACE-Step/acestep-v15-xl-sft"
 
 
 def _load_pipeline(device: str, model_path: str):

tests/test_smoke_gpu.py

@@ -23,7 +23,7 @@ pytestmark = pytest.mark.gpu
 
 def test_generate_minimum_song(tmp_path):
     """Smallest end-to-end: 5 s instrumental drone, seed=1."""
-    os.environ.setdefault("ACE_MODEL_PATH", "ACE-Step/ACE-Step-v1.5-XL-SFT")
+    os.environ.setdefault("ACE_MODEL_PATH", "ACE-Step/acestep-v15-xl-sft")
 
     from backend import ACEStepStudioBackend