We recommend you use conda to install this project and all required packages with the specific version to recurrent our experiments and results of them. The following commands can be used to install this project and all required packages.
conda env create -f requirements/conda.yml -n <env_name>
conda activate <env_name>
pip install -e .It is worth noting that you have to make sure that the cuda version of your machine is consistent with the version of the packages in `conda.yml`, since we have set the Pytorch version and cuda version in the `conda.yml`.
Notably, creating a new conda environment is the most recommended way to install our project and all the required packages to recurrent our results, since you can keep the same version of every package with ours. Installing our project with `pip` may result in different results with different versions of the packages. However, if you counter some problem in the installation process of `conda` method or you have to use a different cuda version, you can try to use the `pip` method to install this project and all required packages.
When installing this project with pip, you can choose the version of some packages you want to use, mainly some packages related to cuda, like Pytorch. See the following sections for details.
We recommend you use the latest version of Python, which works well generally and may provide a better performance. The minimum supported version of Python is 3.8.
Install Pytorch from their official site manually. You have to choose the version of Pytorch based on the cuda version on your machine. Similarly, we recommend you use the latest version of Pytorch, which works well generally and may provide a better performance. You can skip this step if it's fine to use the latest version of Pytorch and the pip will install it in the next section. The minimum supported version of Pytorch is 1.11.
Generally, you can just use the latest dependencies without a specific version, so you can use the command as follows to install this project and all required packages.
pip install -e ".[deps]"By default, we use WandbNamedLogger as the logger. To use Wandb, you have to create an account on their site and login following their doc.
You also can use any other logger you want to use. For example, you can use the TensorBoardLogger instead of the WandbNamedLogger. See the default_runtime.yaml under the configs folder for more details.