1. Create a new virtual environment and activate it. ```shell= conda env create -n anchor_pruning python=3.8 conda activate anchor_pruning ``` 2. Install pytorch and cuda. ```shell= conda install pytorch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0 cudatoolkit=11.1 -c pytorch -c conda-forge ``` 3. Test out pytorch installation. You can use the following python script to check if your GPU driver and CUDA is enabled and accessible by PyTorch. ```shell= import torch torch.cuda.is_available() d = torch.cuda.current_device() torch.cuda.get_device_name(d) ``` 4. [剩下的就照做](https://github.com/Mxbonn/anchor_pruning/tree/master#installation)