search

wustl-cig / DOLCE

DOLCE, ICCV2023. Pytorch Implementation.
MIT License
36 stars 6 forks source link

About the environment.yaml #1

Open GreameLee opened 11 months ago

GreameLee commented 11 months ago

Could we have a requirement.txt to pip install? Using conda to install leap is very hard

JiamingLiu-Jeremy commented 11 months ago

Hi, thank you for interesting in our work. Have you tried install using "install_envs.sh". Once have a pytorch environment in conda, the leap is installed from source.

GreameLee commented 11 months ago

yes, I have tried it but the dependency has conflicts and it can not create the dolce in conda

JiamingLiu-Jeremy commented 11 months ago

Hi, can you please share your hardware/linux environment settings etc ? The current version tested well on NVIDIA V100, 3090 and A6000, with gcc 8.5 and CUDA 11.1.

dewan85 commented 11 months ago

Hello, is it possible to use this code in windows 10?

GreameLee commented 11 months ago

Hello, is it possible to use this code in windows 10?

I have tried but failed

JiamingLiu-Jeremy commented 11 months ago

Hi, the MPI might be the issue for environments inconsistency. Have you tried to run on singe GPU without using DDP at inference ?

wdayang commented 11 months ago

I also tried to create the environment using install_envs.sh, but it shows conflicts then fails. Actually, there are many packages to be installed, I believe only a few is used. By any chance, Is it possible to pick out the most key packages, so that the installation might be more friendly @JiamingLiu-Jeremy ? Thank you.