yxlllc / DDSP-SVC

Real-time end-to-end singing voice conversion system based on DDSP (Differentiable Digital Signal Processing)
MIT License
1.87k stars 250 forks source link

怎么设置模型保存数量? #120

Closed zstthm closed 3 months ago

zstthm commented 3 months ago

模型怎么设置保存数量?保留最近的多少个模型

yxlllc commented 3 months ago

现在的机制是训练时只保留最新的一个模型,但设置 interval_force_save 参数可以控制每训练多少步强制保存一个模型(不会随着训练继续而删除)

zstthm commented 3 months ago

好的,谢谢了