[fbgemm_gpu][docs] Move general documentation into Sphinx

- Move general build, installation, and test documentation into Sphinx
pytorch · Dec 20, 2023 · c5c6217 · c5c6217
1 parent 74808cd
commit c5c6217
Show file tree

Hide file tree

Showing 51 changed files with 1,013 additions and 862 deletions.
diff --git a/.github/scripts/fbgemm_gpu_build.bash b/.github/scripts/fbgemm_gpu_build.bash
@@ -77,6 +77,8 @@ __configure_fbgemm_gpu_build_rocm () {
       echo "[BUILD] Architectures list from rocminfo: ${arch_list}"
 
       if [ "$arch_list" == "" ]; then
+        # It is possible to build FBGEMM_GPU-ROCm on a machine without AMD
+        # cards, in which case the arch_list will be empty.
         echo "[BUILD] rocminfo did not return anything valid!"
 
         # By default, we build just for MI100 and MI250 to save time.  This list

diff --git a/.github/workflows/fbgemm_gpu_docs.yml b/.github/workflows/fbgemm_gpu_docs.yml
@@ -37,7 +37,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: [ "3.11" ]
+        python-version: [ "3.12" ]
 
     steps:
     - name: Setup Build Container

diff --git a/fbgemm_gpu/README.md b/fbgemm_gpu/README.md
@@ -20,15 +20,15 @@ See our [Documentation](docs/README.md) for more information.
 
 The full installation instructions
 for the CUDA, ROCm, and CPU-only variants of FBGEMM_GPU can be found
-[here](docs/InstallationInstructions.md).  In addition, instructions for running
-example tests and benchmarks can be found [here](docs/TestInstructions.md).
+[here](docs/src/general/InstallationInstructions.rst).  In addition, instructions for running
+example tests and benchmarks can be found [here](docs/src/general/TestInstructions.rst).
 
 
 ## Build Instructions
 
 This section is intended for FBGEMM_GPU developers only.  The full build
 instructions for the CUDA, ROCm, and CPU-only variants of FBGEMM_GPU can be
-found [here](docs/BuildInstructions.md).
+found [here](docs/src/general/BuildInstructions.rst).
 
 
 ## Join the FBGEMM_GPU Community

diff --git a/fbgemm_gpu/codegen/embedding_bounds_check_host.cpp b/fbgemm_gpu/codegen/embedding_bounds_check_host.cpp
@@ -16,7 +16,7 @@
 
 using Tensor = at::Tensor;
 
-///@defgroup embedding-cuda Embedding CUDA Operators
+/// @defgroup embedding-cuda Embedding CUDA Operators
 
 ///@ingroup embedding-cuda
 void bounds_check_indices_cuda(

diff --git a/fbgemm_gpu/codegen/embedding_bounds_check_host_cpu.cpp b/fbgemm_gpu/codegen/embedding_bounds_check_host_cpu.cpp
@@ -17,7 +17,7 @@
 using Tensor = at::Tensor;
 using namespace fbgemm_gpu;
 
-///@defgroup embedding-cpu Embedding CPU Operators
+/// @defgroup embedding-cpu Embedding CPU Operators
 ///
 
 namespace {

diff --git a/fbgemm_gpu/codegen/embedding_forward_quantized_host.cpp b/fbgemm_gpu/codegen/embedding_forward_quantized_host.cpp
@@ -184,7 +184,7 @@ void process_uvm_cache_stats(
 } // namespace
 #endif
 
-///@defgroup embedding-cuda Embedding CUDA Operators
+/// @defgroup embedding-cuda Embedding CUDA Operators
 ///
 
 Tensor int_nbit_split_embedding_codegen_forward_unweighted_cuda(

diff --git a/fbgemm_gpu/codegen/embedding_forward_quantized_host_cpu.cpp b/fbgemm_gpu/codegen/embedding_forward_quantized_host_cpu.cpp
@@ -24,7 +24,7 @@
 using Tensor = at::Tensor;
 using namespace fbgemm_gpu;
 
-///@defgroup embedding-cpu Embedding CPU Operators
+/// @defgroup embedding-cpu Embedding CPU Operators
 ///
 
 Tensor int_nbit_split_embedding_codegen_forward_unweighted_cpu(