Sambert-HifiGAN TTS - Ascend NPU Adaptation

# 安装 PyTorch NPU
pip install torch==2.9.0 torch_npu==2.9.0.post1

# 安装其他依赖
pip install numpy pyyaml

pip install modelscope
modelscope download --model iic/speech_sambert-hifigan_tts_zhiya_zh-cn_16k

# 精度验证：CPU vs NPU 对比
python inference.py --mode validate

# 性能基准测试：NPU 推理性能
python inference.py --mode benchmark

# 端到端验证（精度 + 性能）
python inference.py --mode all

# 快速测试：CPU 前向传播
python inference.py --mode test

from inference import load_models, run_inference

# 在 NPU 上加载模型
am, voc = load_models('npu:0')

# 运行推理，seq_len 控制合成文本长度
audio = run_inference(am, voc, seq_len=30, device='npu:0')
print(f"输出音频: {audio.shape}, {audio.size} 采样点")

@inproceedings{li2020robutrans,
  title={Robutrans: A robust transformer-based text-to-speech model},
  author={Li, Naihan and Liu, Yanqing and Wu, Yu and Liu, Shujie and Zhao, Sheng and Liu, Ming},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={34},
  number={05},
  pages={8228--8235},
  year={2020}
}

@article{kong2020hifi,
  title={Hifi-gan: Generative adversarial networks for efficient and high fidelity speech synthesis},
  author={Kong, Jungil and Kim, Jaehyeon and Bae, Jaekyoung},
  journal={Advances in Neural Information Processing Systems},
  volume={33},
  pages={17022--17033},
  year={2020}
}