Is this project compatible with a recent GPU ?

[antoinersw]

I have a RTX 4080 (sm__89). It doesnt seem to be compatible with pytorch 1.6
https://pytorch.org/get-started/locally/

Do you believe I can use the nightly build with a more recent pytorch3d version ?

Thanks

[emlcpfx]

It works with my 3090 on Windows and WSL Ubuntu. Off the top of my head, you need to use a more recent version of PyTorch.

[zardam]

@emlcpfx can you help me im getting errors in running the code

[emlcpfx]

I’m a bit rusty, I barely remember this, but I’ll try. What are the errors?

[emlcpfx]

This is a guide to getting EMOCA to work on Windows. EMOCA is built on top of DECA, so maybe it’ll help— https://docs.google.com/document/d/13rbNHEtyFqbvgHoIrymD558wmTgd7A67zPE85ISe144/edit

[zardam]

im running the project on google colab

[zardam]

@emlcpfx can i get your email id or something else where i can contact you

[emlcpfx]

Leave yours, and I’ll email you.

[salvatore]

@zardamhussain how did you get this working on colab? the notebook provided is not compatible and i could not install any of the requirements

[zardam]

@erotavlas

1. download the miniconda in google-colab by using these command

! wget https://repo.anaconda.com/miniconda/Miniconda3-py37_4.8.2-Linux-x86_64.sh
! chmod +x Miniconda3-py38_4.8.3-Linux-x86_64.sh
! bash ./Miniconda3-py38_4.8.3-Linux-x86_64.sh -b -f -p /usr/local
import sys
sys.path.append('/usr/local/lib/python3.8/site-packages/')

2. then run these comamands

!conda install pytorch==1.7.1 torchvision==0.8.2 torchaudio==0.7.2 cudatoolkit=11.0 -c pytorch
!conda install -c fvcore -c iopath -c conda-forge fvcore iopath
!conda install pytorch3d -c pytorch3d

3. pip install -r requirements.txt

numpy
scipy
chumpy
scikit-image
opencv-python
scikit-image     #skimage
PyYAML       
# torch==1.6.0   # for compatible with pytorch3d
# torchvision==0.7.0
face-alignment
yacs
kornia==0.5.0
ninja
# fvcore
# pytorch3d