search

facebookresearch / vissl

VISSL is FAIR's library of extensible, modular and scalable components for SOTA Self-Supervised Learning with images.
https://vissl.ai
MIT License
3.25k stars 331 forks source link

Setup dependencies with python3.8 #481

Closed miriamrebekah closed 2 years ago

miriamrebekah commented 2 years ago

I am trying to set up an anaconda environment for VISSL with python3.8. From looking at other issue tickets, it looks like this is common for people to do. I'm having troubles getting the requirements and versions right though. Is there anywhere I can find a requirements.txt for VISSL for python3.8?

# Additionally, CUDA 10.2 is required if you'd like to run the programs with a GPU.
# If not, this is easily accomplished with the following commands:
#   wget -c https://repo.anaconda.com/archive/Anaconda3-2019.07-Linux-x86_64.sh
#   /bin/bash Anaconda3-2019.07-Linux-x86_64.sh -bfp /opt/conda
#   conda create --name latent-3812 python=3.8.12
#   conda activate latent-3812
#   conda install -c pytorch cuda-toolkit==10.2 
torch==1.7.1
torchvision==0.8.0a0
apex==0.1
cython==0.29.22
fairscale==0.3.8
fvcore==0.1.3.post20210317
iopath==0.1.9
hydra-core==1.1.1
numpy==1.21.2
parameterized==0.7.4
scikit-learn==0.24.1
submitit==1.3.3
tabulate==0.8.9
tensorboard==2.7.0
opencv-python==4.5.4.60
classy_vision==0.7.0.dev
fairscale==0.3.8
iseessel commented 2 years ago

Hi @miriamrebekah What's the error you are getting with 3.8? How are you installing vissl?

Can you please provide your environment information using the following command:

wget -nc -q https://github.com/facebookresearch/vissl/raw/main/vissl/utils/collect_env.py && python collect_env.py

I tried following instructions here: https://github.com/facebookresearch/vissl/blob/main/INSTALL.md#install-from-source-in-conda-environment, switch to python 3.8.

I did receive two errors like: 

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
pre-commit 2.15.0 requires toml, which is not installed.

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
tensorboard 2.6.0 requires requests<3,>=2.21.0, which is not installed.
tensorboard 2.6.0 requires werkzeug>=0.11.15, which is not installed.
pycocotools 2.0.2 requires matplotlib>=2.1.0, which is not installed.

However, I was still able to go through the flow and successfully import vissl.

miriamrebekah commented 2 years ago

@iseessel Thank you for the help! I tried again and now it's working! I really appreciate it.