We provide some advice for installation all dependencies with conda.
- create a conda environement
conda create -n SiT python=3.7
- activate the environment
conda activate SiT
- install pytorch dependencies
Assuming GPU support, please check your CUDA version and select the appropriate installation command from PyTorch. This codebase works also for CPU only.
For CUDA 11.3 PyTorch version:
conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch -c nvidia
- install requierements
conda install -c conda-forge --file requirements.txt
- (Optional) Extra librairies
If you wish to use warm-up scheduler, you will need to pip install the following:
pip install git+https://github.com/ildoonet/pytorch-gradual-warmup-lr.git