模型介绍

本模型依托魔搭社区（ModelScope）AIGC专区模型训练环境与算力完成训练。

• 模型类型：LoRA
• 基础模型：ACE-Step/acestep-v15-xl-sft
• 训练代码：DiffSynth-Studio
• 训练数据量：13
• 总训练步数：4000
• 开源协议：Apache-2.0

推理代码

安装 DiffSynth-Studio：

pip install diffsynth

开始推理：

from diffsynth.pipelines.ace_step import AceStepPipeline, ModelConfig
from diffsynth.utils.data.audio import save_audio
import torch

pipe = AceStepPipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="ACE-Step/acestep-v15-xl-sft", origin_file_pattern="model-*.safetensors"),
        ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/model.safetensors"),
        ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
    ],
    text_tokenizer_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="Qwen3-Embedding-0.6B/"),
    silence_latent_config=ModelConfig(model_id="ACE-Step/Ace-Step1.5", origin_file_pattern="acestep-v15-turbo/silence_latent.pt"),
)
pipe.load_lora(pipe.dit, ModelConfig(model_id="yy3yy3/yy99", origin_file_pattern="yy99_c1-st4000.safetensors"))

prompt = "An explosive, high-energy pop-rock track with a strong anime theme song feel."
lyrics = "I Love you. Forever. I Love you. Forever."
audio = pipe(
    prompt=prompt,
    lyrics=lyrics,
    duration=160,
    bpm=100,
    keyscale="B minor",
    timesignature="4",
    vocal_language="zh",
    seed=1,
    num_inference_steps=50,
    cfg_scale=4.0,
)
save_audio(audio, pipe.vae.sampling_rate, "music.wav")