If you're looking at the anonymised repo then you can't clone/download it. However, you can just download the relevant files individually:
mkdir -p Orthogonal-Optimisers
cd Orthogonal-Optmiser
mkdir src/orth_optim
wget https://anonymous.4open.science/api/repo/Orthogonal-Optimisers/file/setup.py
wget https://anonymous.4open.science/api/repo/Orthogonal-Optimisers/file/src/orth_optim/__init__.py -O src/orth_optim/__init__.py
git clone https://github.com/MarkTuddenham/Orthogonal-Optimisers.git
cd Orthogonal-Optimisers
pip install .
or
pip install git+https://github.com/MarkTuddenham/Orthogonal-Optimisers.git#egg=orth_optim
And then at the top of your main python script:
from orth_optim import hook
hook()
Now the torch optimisers have an orthogonal option, e.g:
torch.optim.SGD(model.parameters(),
lr=1e-3,
momentum=0.9,
orth=True)
If you have a custom optimiser you can apply the orthogonalise
decorator.
from orth_optim import orthogonalise
@orthogonalise
class LARS(torch.optim.Optimizer):
...