Installation

Prerequisites & Installation

conda create -n ezv2 python=3.8
conda activate ezv2
pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2
pip install -r requirements.txt

Before starting training, you need to build the c++/cython style external packages.

cd ez/mcts/ctree
bash make.sh
cd -

Install gym version 0.22.0 (previously 0.15.7, which caused the error: cannot import name ‘Monitor’ from 'gym.wrappers'):
```
pip install gym==0.22.0
```
Install wandb version 0.14.2 (the previous environment had no version restrictions, but needs downgrading to avoid TypeError):
```
pip install wandb==0.14.2
```
Install reference version 0.31.0 (version 0.30.2 fails to import referencing.jsonschema):
```
pip install reference==0.31.0
```

Install dmc2gym from GitHub:

pip install git+git://github.com/denisyarats/dmc2gym.git

Install imageio with ffmpeg and pyav to ensure eval_worker runs correctly:
```
pip install imageio[ffmpeg]
pip install imageio[pyav]
```

In addition to compiling ctree, also compile ctree_v2 (If you find the training process is stuck, this issue exists in some cases.):
```
cd ez/mcts/ctree_v2
sh make.sh
```
In addition to compiling ctree, also compile ori_ctree :
```
cd ez/mcts/ori_ctree
sh make.sh
```