added cuda explanation

Author: benja263

docs/quickstart.rst

@@ -7,11 +7,11 @@ Install GBRL via pip:
    
    pip install gbrl
 
-CPU only version is installed with the following command:
+CPU only version despite having a valid CUDA installation by setting `CUDA_ONLY=1` as an environment variable:
 
 .. code-block:: console
-
-   CPU_ONLY=1 pip install gbrl
+   export CPU_ONLY=1 
+   pip install gbrl
 
 Dependencies 
 ============ 
@@ -42,15 +42,31 @@ Once installed make sure that the appropriate environment variables are set:
 CUDA
 ~~~~ 
 
+Make sure that ``CUDA_HOME`` is set. 
+For integration with Microsoft Visual Studio make sure to copy the following files:
+
+.. code-block:: console
+   CUDA <cuda_version>.props
+   CUDA <cuda_version>.targets
+   CUDA <cuda_version>.xml
+   Nvda.Build.CudaTasks.v<cuda_version>.dll
+   cudart.lib
+
+
+into ``<visual studio path>\BuildTools\MSBuild\Microsoft\VC\v160\BuildCustomizations``.
+
+Once GBRL is installed, verify that CUDA is enabled by running
+ 
 .. code-block:: python
    # Verify that GPU is visible by running
    import gbrl
 
    print(gbrl.cuda_available())
 
 
-Graphviz
-~~~~~~~~
+Graphviz (optional)
+~~~~~~~~~~~~~~~~~~~
+For tree visualization make sure graphviz is installed before compilation. 
+
 
-*OPTIONAL*  
-For tree visualization make sure graphviz is installed before compilation. 
+ 

gbrl/gbt.py

@@ -281,7 +281,7 @@ def print_tree(self, tree_idx: int) -> None:
         self._model.print_tree(tree_idx)
 
     def plot_tree(self, tree_idx: int, filename: str) -> None:
-        """Plots tree using graphviz 
+        """Plots tree using (only works if GBRL was compiled with graphviz)
 
         Args:
             tree_idx (int): tree index to plot