Closed samsullivandelgobbo closed 1 year ago
Hi, your error is the file you are importing. It should be the nerfstudio_models
folder.
Try this:
ns-train in2n --data outputs/office --load-dir outputs/office/nerfacto/2023-04-12_102834/nerfstudio_models/ --pipeline.prompt 'make the floor lava' --pipeline.guidance-scale 7.5 --pipeline.image-guidance-scale 1.5
Thanks for the help. Sorry for the oversight
System:
Windows 10 Version 10.0.19044 Build 190 Ubuntu 20.04 running in WSL2 CUDA 11.7 installed using miniconda in ~/miniconda3/envs/nerfstudio/ Torch 1.13.1+cu117
Problem:
Trying to train instruct2nerf model on an freshly trained and processed model using the following command
ns-train in2n --data outputs/office --load-dir outputs/office/nerfacto/2023-04-12_102834 --pipeline.prompt 'make the floor lava' --pipeline.guidance-scale 7.5 --pipeline.image-guidance-scale 1.5
fails and returns
ValueError: invalid literal for int() with base 10: 'config'
I can't seem to fix it, tried reinstalling both using the basic pip command and through directly cloning the repo. Seems like the config file is being read incorrectly. Has anyone run into this?
Full log:
And here is the config file it seems to be erroring out on: