search

cassiePython / CLIPNeRF

CLIP-NeRF: Text-and-Image Driven Manipulation of Neural Radiance Fields
Apache License 2.0
260 stars 15 forks source link

Very slow training #10

Closed LiquidAmmonia closed 2 years ago

LiquidAmmonia commented 2 years ago

Hi, I have encountered a very slow training, it took me about 10 seconds to finish an iteration. The clip process costs the most of the time. My training script is directly copied from your README.md without any changes, and I checked the CLIP model, input image and text are all on cuda. Is this speed normal? Is there any way to speed up this process?

cassiePython commented 2 years ago

What is your GPU platform？

LiquidAmmonia commented 2 years ago

A Nvidia Titan X with cuda

cassiePython commented 2 years ago

Another researcher also encounters the same question on a TitanX, but obtains good results after training on a 3090ti. As the code is not fully tested on Titan X, you are advised to use a GPU with higher performance.