Skip to content

Latest commit

 

History

History
10 lines (7 loc) · 736 Bytes

File metadata and controls

10 lines (7 loc) · 736 Bytes

FAQ

  1. Q: How to select models?
    A: Please refer to docs/model_zoo.md

  2. Q: Can face_enhance be used for anime images/animation videos?
    A: No, it can only be used for real faces. It is recommended not to use this option for anime images/animation videos to save GPU memory.

  3. Q: Error "slow_conv2d_cpu" not implemented for 'Half'
    A: In order to save GPU memory consumption and speed up inference, Real-ESRGAN uses half precision (fp16) during inference by default. However, some operators for half inference are not implemented in CPU mode. You need to add --fp32 option for the commands. For example, python inference_realesrgan.py -n RealESRGAN_x4plus.pth -i inputs --fp32.