能否提供多卡推理

BadToBest / EchoMimic

Lifelike Audio-Driven Portrait Animations through Editable Landmark Conditioning

https://badtobest.github.io/echomimic.html

Apache License 2.0

2.26k stars 263 forks source link

Closed jinghong-6 closed 1 month ago

jinghong-6 commented 1 month ago

能否提供自适应多卡推理功能，我在四卡机器上运行，只使用了第一张显卡，速度不够理想

yuange250 commented 1 month ago

抱歉这个与现有算法基础有冲突，改造量较大，短期内可能很难支持

JoeFannie commented 1 month ago

加速的模型马上就要发布了，速度会有明显提升，预计加速10倍