add rocm docs

docs/Installation-Guide.rst

@@ -738,6 +738,65 @@ macOS
 
 The CUDA version is not supported on macOS.
 
+Build ROCm Version
+~~~~~~~~~~~~~~~~~~
+
+The `original GPU version <#build-gpu-version>`__ of LightGBM (``device_type=gpu``) is based on OpenCL.
+
+The ROCm-based version (``device_type=cuda``) is a separate implementation. Yes, the ROCm version reuses the ``device_type=cuda`` as a convenience for users.  Use this version in Linux environments with an AMD GPU.
+
+Windows
+^^^^^^^
+
+The ROCm version is not supported on Windows.
+Use the `GPU version <#build-gpu-version>`__ (``device_type=gpu``) for GPU acceleration on Windows.
+
+Linux
+^^^^^
+
+On Linux, a ROCm version of LightGBM can be built using
+
+- **CMake**, **gcc** and **ROCm**;
+- **CMake**, **Clang** and **ROCm**.
+
+Please refer to `the ROCm docs`_ for **ROCm** libraries installation.
+
+After compilation the executable and ``.so`` files will be in ``LightGBM/`` folder.
+
+gcc
+***
+
+1. Install `CMake`_, **gcc** and **ROCm**.
+
+2. Run the following commands:
+
+   .. code:: sh
+
+     git clone --recursive https://github.com/microsoft/LightGBM
+     cd LightGBM
+     cmake -B build -S . -DUSE_ROCM=ON
+     cmake --build build -j4
+
+Clang
+*****
+
+1. Install `CMake`_, **Clang**, **OpenMP** and **ROCm**.
+
+2. Run the following commands:
+
+   .. code:: sh
+
+     git clone --recursive https://github.com/microsoft/LightGBM
+     cd LightGBM
+     export CXX=clang++-14 CC=clang-14  # replace "14" with version of Clang installed on your machine
+     cmake -B build -S . -DUSE_ROCM=ON
+     cmake --build build -j4
+
+macOS
+^^^^^
+
+The ROCm version is not supported on macOS.
+
 Build Java Wrapper
 ~~~~~~~~~~~~~~~~~~
 
@@ -1047,6 +1106,8 @@ gcc
 
 .. _this detailed guide: https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html
 
+.. _the ROCm docs: https://rocm.docs.amd.com/projects/install-on-linux/en/latest/
+
 .. _following docs: https://github.com/google/sanitizers/wiki
 
 .. _Ninja: https://ninja-build.org

docs/Parameters.rst

@@ -264,7 +264,7 @@ Core Parameters
 
    -  ``cpu`` supports all LightGBM functionality and is portable across the widest range of operating systems and hardware
 
-   -  ``cuda`` offers faster training than ``gpu`` or ``cpu``, but only works on GPUs supporting CUDA
+   -  ``cuda`` offers faster training than ``gpu`` or ``cpu``, but only works on GPUs supporting CUDA or ROCm
 
    -  ``gpu`` can be faster than ``cpu`` and works on a wider range of GPUs than CUDA
 
@@ -272,7 +272,7 @@ Core Parameters
 
    -  **Note**: for the faster speed, GPU uses 32-bit float point to sum up by default, so this may affect the accuracy for some tasks. You can set ``gpu_use_dp=true`` to enable 64-bit float point, but it will slow down the training
 
-   -  **Note**: refer to `Installation Guide <./Installation-Guide.rst>`__ to build LightGBM with GPU or CUDA support
+   -  **Note**: refer to `Installation Guide <./Installation-Guide.rst>`__ to build LightGBM with GPU, CUDA, or ROCm support
 
 -  ``seed`` :raw-html:`<a id="seed" title="Permalink to this parameter" href="#seed">&#x1F517;&#xFE0E;</a>`, default = ``None``, type = int, aliases: ``random_seed``, ``random_state``
 

docs/_static/js/script.js

@@ -22,6 +22,7 @@ $(() => {
             "#build-mpi-version",
             "#build-gpu-version",
             "#build-cuda-version",
+            "#build-rocm-version",
             "#build-java-wrapper",
             "#build-python-package",
             "#build-r-package",