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)