From b1ee2c5228ef8a88169f1e34205ef05a50410d01 Mon Sep 17 00:00:00 2001
From: Paul Zhang <paulzhan@meta.com>
Date: Wed, 23 Oct 2024 12:27:33 -0700
Subject: [PATCH] package version validation (#2513)

Summary:


We encounter a case where TorchRec package version does not match FBGEMM in release. Here, we check that in validating the binaries

Reviewed By: iamzainhuda

Differential Revision: D64840358
---
 .github/scripts/validate_binaries.sh | 71 ++++++++++++++++------------
 1 file changed, 41 insertions(+), 30 deletions(-)

diff --git a/.github/scripts/validate_binaries.sh b/.github/scripts/validate_binaries.sh
index 22009dac3..85ad0de47 100755
--- a/.github/scripts/validate_binaries.sh
+++ b/.github/scripts/validate_binaries.sh
@@ -40,40 +40,33 @@ else
     export CUDA_VERSION="cpu"
 fi
 
-if [[ ${MATRIX_CHANNEL} = 'pypi_release' ]]; then
-    echo "checking pypi release"
-    pip install torch
-    pip install fbgemm-gpu
-    pip install torchrec
-else
-    # figure out URL
-    if [[ ${MATRIX_CHANNEL} = 'nightly' ]]; then
-        export PYTORCH_URL="https://download.pytorch.org/whl/nightly/${CUDA_VERSION}"
-    elif [[ ${MATRIX_CHANNEL} = 'test' ]]; then
-        export PYTORCH_URL="https://download.pytorch.org/whl/test/${CUDA_VERSION}"
-    elif [[ ${MATRIX_CHANNEL} = 'release' ]]; then
-        export PYTORCH_URL="https://download.pytorch.org/whl/${CUDA_VERSION}"
-    fi
+# figure out URL
+if [[ ${MATRIX_CHANNEL} = 'nightly' ]]; then
+    export PYTORCH_URL="https://download.pytorch.org/whl/nightly/${CUDA_VERSION}"
+elif [[ ${MATRIX_CHANNEL} = 'test' ]]; then
+    export PYTORCH_URL="https://download.pytorch.org/whl/test/${CUDA_VERSION}"
+elif [[ ${MATRIX_CHANNEL} = 'release' ]]; then
+    export PYTORCH_URL="https://download.pytorch.org/whl/${CUDA_VERSION}"
+fi
 
-    # install pytorch
-    # switch back to conda once torch nightly is fixed
-    # if [[ ${MATRIX_GPU_ARCH_TYPE} = 'cuda' ]]; then
-    #     export PYTORCH_CUDA_PKG="pytorch-cuda=${MATRIX_GPU_ARCH_VERSION}"
-    # fi
-    conda run -n build_binary pip install torch --index-url "$PYTORCH_URL"
+# install pytorch
+# switch back to conda once torch nightly is fixed
+# if [[ ${MATRIX_GPU_ARCH_TYPE} = 'cuda' ]]; then
+#     export PYTORCH_CUDA_PKG="pytorch-cuda=${MATRIX_GPU_ARCH_VERSION}"
+# fi
+conda run -n build_binary pip install torch --index-url "$PYTORCH_URL"
 
-    # install fbgemm
-    conda run -n build_binary pip install fbgemm-gpu --index-url "$PYTORCH_URL"
+# install fbgemm
+conda run -n build_binary pip install fbgemm-gpu --index-url "$PYTORCH_URL"
 
-    # install requirements from pypi
-    conda run -n build_binary pip install torchmetrics==1.0.3
+# install requirements from pypi
+conda run -n build_binary pip install torchmetrics==1.0.3
 
-    # install torchrec
-    conda run -n build_binary pip install torchrec --index-url "$PYTORCH_URL"
+# install torchrec
+conda run -n build_binary pip install torchrec --index-url "$PYTORCH_URL"
 
-    # Run small import test
-    conda run -n build_binary python -c "import torch; import fbgemm_gpu; import torchrec"
-fi
+# Run small import test
+conda run -n build_binary python -c "import torch; import fbgemm_gpu; import torchrec"
 
 # check directory
 ls -R
@@ -98,13 +91,22 @@ fi
 
 if [[ ${MATRIX_CHANNEL} != 'release' ]]; then
     exit 0
+else
+    # Check version matches only for release binaries
+    torchrec_version=$(conda run -n build_binary pip show torchrec | grep Version | cut -d' ' -f2)
+    fbgemm_version=$(conda run -n build_binary pip show fbgemm_gpu | grep Version | cut -d' ' -f2)
+
+    if [ "$torchrec_version" != "$fbgemm_version" ]; then
+        echo "Error: TorchRec package version does not match FBGEMM package version"
+        exit 1
+    fi
 fi
 
 conda create -y -n build_binary python="${MATRIX_PYTHON_VERSION}"
 
 conda run -n build_binary python --version
 
-if [[ ${MATRIX_GPU_ARCH_VERSION} != '12.1' ]]; then
+if [[ ${MATRIX_GPU_ARCH_VERSION} != '12.4' ]]; then
     exit 0
 fi
 
@@ -113,6 +115,15 @@ conda run -n build_binary pip install torch
 conda run -n build_binary pip install fbgemm-gpu
 conda run -n build_binary pip install torchrec
 
+# Check version matching again for PyPI
+torchrec_version=$(conda run -n build_binary pip show torchrec | grep Version | cut -d' ' -f2)
+fbgemm_version=$(conda run -n build_binary pip show fbgemm_gpu | grep Version | cut -d' ' -f2)
+
+if [ "$torchrec_version" != "$fbgemm_version" ]; then
+    echo "Error: TorchRec package version does not match FBGEMM package version"
+    exit 1
+fi
+
 # check directory
 ls -R