update dreamerv2 doc

dreamerv2/README.md

@@ -1 +1,40 @@
-TODO
+# Guide for Training DreamerV2 on CarDreaner
+
+This guide assumes you have installed ``car_dreamer``. If not, please follow the instructions in the [main README](../README.md).
+
+First, install the required dependencies for DreamerV2:
+```bash
+cd dreamerv2
+conda install -c "nvidia/label/cuda-11.8.0" cuda-toolkit
+pip install -r requirements.txt
+```
+
+Set up CARLA and environment variables:
+```bash
+export CUDNN_PATH=$(dirname $(python -c "import nvidia.cudnn;print(nvidia.cudnn.__file__)"))
+export CUSOLVER_PATH=$(dirname $(python -c "import nvidia.cusolver;print(nvidia.cusolver.__file__)"))
+export LD_LIBRARY_PATH=$CUDNN_PATH/lib:$CUSOLVER_PATH/lib:$CONDA_PREFIX/lib:$LD_LIBRARY_PATH
+```
+
+## Training
+
+Execute the training script with desired configurations:
+```bash
+cd ..
+bash train_dm2.sh 2000 0 --task carla_four_lane --dreamerv2.logdir ./logdir/carla_four_lane
+```
+
+``2000`` is the port number of the CARLA server. The script will automatically start the server so you don't need to start it manually.
+``0`` is the GPU number.
+``--task`` is the name of the task and ``--dreamerv2.logdir`` is the directory to save the training logs. For a complete list of tasks and their configurations, please refer to the [documentation](https://car-dreamer.readthedocs.io/en/latest/tasks.html).
+
+## Visualization
+
+Online data monitoring can be accessed on website on ``http://localhost:9000/``, where the port number should be changed to ``<carla-port> + 7000`` if you don't use the default port number ``2000`` for CARLA server.
+
+Offline data logging can be accessed through TensorBoard:
+```bash
+tensorboard --logdir ./logdir/carla_four_lane
+```
+
+Go to ``http://localhost:6006/`` in your browser to see the output.

dreamerv2/requirements.txt

@@ -0,0 +1,22 @@
+cloudpickle==3.0.0
+crafter==1.8.3
+gym==0.12.5
+dm_control==1.0.18
+jax==0.4.26
+jaxlib==0.4.26
+numpy==1.24.3
+optax==0.2.2
+rich==13.7.1
+ruamel.yaml==0.18.6
+tensorflow[and-cuda]==2.15.1
+tensorflow_probability==0.23.0
+scikit-image==0.23.2
+ffmpeg==1.4
+moviepy==1.0.3
+imageio==2.34.1
+wandb==0.16.6
+opencv-python>=4.9,<5.0
+shapely==1.8.4
+carla==0.9.15
+matplotlib==3.8.4
+tabulate==0.9.0

dreamerv3/README.md

@@ -43,9 +43,9 @@ Go to ``http://localhost:6006/`` in your browser to see the output.
 To use ``wandb`` for visualization, add the WandB logger to ``dreamerv3/train.py``:
 ```python
 logger = embodied.Logger(step, [
-        # ...
-        embodied.logger.WandBOutput(logdir.name, config),
-    ])
+    # ...
+    embodied.logger.WandBOutput(logdir.name, config),
+])
 ```
 
 Once you log in ``wandb``, put your project and entity name in ``dreamerv3/embodied/logger.py``:

dreamerv3/requirements.txt

@@ -3,7 +3,7 @@ crafter==1.8.3
 gym==0.12.5
 dm_control==1.0.18
 jax==0.4.26
-jaxlib==0.4.26+cuda12.cudnn89
+jaxlib==0.4.26
 numpy==1.24.3
 optax==0.2.2
 rich==13.7.1