Merge pull request #54 from ROCmSoftwarePlatform/IFU-2024-01-05

Ifu 2024 01 05

.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
@@ -418,6 +420,8 @@ build_fbgemm_gpu_install () {
   # fbgemm_gpu/ subdirectory present
   cd - || return 1
   (test_python_import_package "${env_name}" fbgemm_gpu) || return 1
+  (test_python_import_symbol "${env_name}" fbgemm_gpu __version__) || return 1
+  cd - || return 1
 
   echo "[BUILD] FBGEMM-GPU build + install completed"
 }

.github/scripts/fbgemm_gpu_docs.bash

@@ -34,13 +34,20 @@ install_docs_tools () {
   # shellcheck disable=SC2155
   local env_prefix=$(env_name_or_prefix "${env_name}")
 
-  echo "[INSTALL] Installing docs tools ..."
+  echo "[INSTALL] Installing Doxygen ..."
+
   # shellcheck disable=SC2086
   (exec_with_retries 3 conda install ${env_prefix} -c conda-forge -y \
-    doxygen) || return 1
+    doxygen \
+    make) || return 1
 
   # Check binaries are visible in the PATH
   (test_binpath "${env_name}" doxygen) || return 1
+  (test_binpath "${env_name}" make) || return 1
+
+  echo "[BUILD] Installing docs-build dependencies ..."
+  # shellcheck disable=SC2086
+  (exec_with_retries 3 conda run ${env_prefix} python -m pip install -r requirements.txt) || return 1
 
   echo "[INSTALL] Successfully installed all the docs tools"
 }
@@ -69,10 +76,6 @@ build_fbgemm_gpu_docs () {
   # shellcheck disable=SC2155
   local env_prefix=$(env_name_or_prefix "${env_name}")
 
-  echo "[BUILD] Installing docs-build dependencies ..."
-  # shellcheck disable=SC2086
-  (exec_with_retries 3 conda run ${env_prefix} python -m pip install -r requirements.txt) || return 1
-
   echo "[BUILD] Running Doxygen build ..."
   # shellcheck disable=SC2086
   (exec_with_retries 3 conda run ${env_prefix} doxygen Doxyfile.in) || return 1

.github/scripts/fbgemm_gpu_install.bash

@@ -26,8 +26,8 @@ __fbgemm_gpu_post_install_checks () {
   (test_python_import_symbol "${env_name}" fbgemm_gpu __version__) || return 1
 
   echo "[CHECK] Printing out the FBGEMM-GPU version ..."
-  # shellcheck disable=SC2086
-  installed_fbgemm_gpu_version=$(conda run ${env_prefix} python -c "import fbgemm_gpu; print(fbgemm_gpu.__version__)")
+  # shellcheck disable=SC2086,SC2155
+  local installed_fbgemm_gpu_version=$(conda run ${env_prefix} python -c "import fbgemm_gpu; print(fbgemm_gpu.__version__)")
   echo "[CHECK] The installed version is: ${installed_fbgemm_gpu_version}"
 }
 
@@ -67,13 +67,15 @@ install_fbgemm_gpu_wheel () {
 
 install_fbgemm_gpu_pip () {
   local env_name="$1"
-  local fbgemm_gpu_version="$2"
-  local fbgemm_gpu_variant_type="$3"
-  local fbgemm_gpu_variant_version="$4"
-  if [ "$fbgemm_gpu_variant_type" == "" ]; then
-    echo "Usage: ${FUNCNAME[0]} ENV_NAME FBGEMM_GPU_VERSION FBGEMM_GPU_VARIANT_TYPE [FBGEMM_GPU_VARIANT_VERSION]"
+  local fbgemm_gpu_channel_version="$2"
+  local fbgemm_gpu_variant_type_version="$3"
+  if [ "$fbgemm_gpu_variant_type_version" == "" ]; then
+    echo "Usage: ${FUNCNAME[0]} ENV_NAME FBGEMM_GPU_CHANNEL[/VERSION] FBGEMM_GPU_VARIANT_TYPE[/VARIANT_VERSION]"
     echo "Example(s):"
-    echo "    ${FUNCNAME[0]} build_env 0.5.0rc2 cuda 12.1.1        # Install a specific version of the package (PyPI)"
+    echo "    ${FUNCNAME[0]} build_env 0.5.0 cpu                  # Install the CPU variant, specific version from release channel"
+    echo "    ${FUNCNAME[0]} build_env release cuda 12.1.1        # Install the CUDA variant, latest version from release channel"
+    echo "    ${FUNCNAME[0]} build_env test/0.6.0rc0 cuda 12.1.0  # Install the CUDA 12.1 variant, specific version from test channel"
+    echo "    ${FUNCNAME[0]} build_env nightly rocm 5.3           # Install the ROCM 5.3 variant, latest version from nightly channel"
     return 1
   else
     echo "################################################################################"
@@ -86,7 +88,7 @@ install_fbgemm_gpu_pip () {
 
   # Install the package from PyTorch PIP (not PyPI)
   # The package's canonical name is 'fbgemm-gpu' (hyphen, not underscore)
-  install_from_pytorch_pip "${env_name}" fbgemm_gpu "${fbgemm_gpu_version}" "${fbgemm_gpu_variant_type}" "${fbgemm_gpu_variant_version}" || return 1
+  install_from_pytorch_pip "${env_name}" fbgemm_gpu "${fbgemm_gpu_channel_version}" "${fbgemm_gpu_variant_type_version}" || return 1
 
   # Run post-installation checks
   __fbgemm_gpu_post_install_checks "${env_name}" || return 1

.github/scripts/fbgemm_gpu_test.bash

@@ -182,7 +182,7 @@ test_setup_conda_environment () {
   if [ "$pytorch_installer" == "conda" ]; then
     install_pytorch_conda     "${env_name}" "${pytorch_version}" "${pytorch_variant_type}" "${pytorch_variant_version}" || return 1
   else
-    install_pytorch_pip       "${env_name}" "${pytorch_version}" "${pytorch_variant_type}" "${pytorch_variant_version}" || return 1
+    install_pytorch_pip       "${env_name}" "${pytorch_version}" "${pytorch_variant_type}"/"${pytorch_variant_version}" || return 1
   fi
 }
 

.github/scripts/nova_postscript.bash

@@ -28,13 +28,22 @@ install_fbgemm_gpu_wheel "${BUILD_ENV_NAME}" fbgemm_gpu/dist/*.whl
 
 # Test with PyTest
 echo "[NOVA] Current working directory: $(pwd)"
-CPU_GPU="${CU_VERSION}"
-if [ "${CU_VERSION}" != 'cpu' ]; then
-    CPU_GPU=""
+if [[ $CU_VERSION = cu* ]]; then
+  echo "[NOVA] Testing the CUDA variant of FBGEMM_GPU ..."
+  export fbgemm_variant="cuda"
+
+elif [[ $CU_VERSION = rocm* ]]; then
+  echo "[NOVA] Testing the ROCm variant of FBGEMM_GPU ..."
+  export fbgemm_variant="rocm"
+
+else
+  echo "[NOVA] Testing the CPU variant of FBGEMM_GPU ..."
+  export fbgemm_variant="cpu"
 fi
+
 $CONDA_RUN python3 -c "import torch; print('cuda.is_available() ', torch.cuda.is_available()); print ('device_count() ',torch.cuda.device_count());"
 cd "${FBGEMM_REPO}/fbgemm_gpu/test" || { echo "[NOVA] Failed to cd to fbgemm_gpu/test from $(pwd)"; };
-run_fbgemm_gpu_tests "${BUILD_ENV_NAME}" "${CPU_GPU}"
+run_fbgemm_gpu_tests "${BUILD_ENV_NAME}" "${fbgemm_variant}"
 
 # Workaround EACCES: permission denied error at checkout step
 chown -R 1000:1000 /__w/FBGEMM/FBGEMM/ || echo "Unable to chown 1000:1000 from $USER, uid: $(id -u)"

.github/scripts/utils_build.bash

@@ -166,12 +166,14 @@ install_build_tools () {
     cmake \
     hypothesis \
     jinja2 \
+    make \
     ninja \
     numpy \
     scikit-build \
     wheel) || return 1
 
   # Check binaries are visible in the PAATH
+  (test_binpath "${env_name}" make) || return 1
   (test_binpath "${env_name}" cmake) || return 1
   (test_binpath "${env_name}" ninja) || return 1