diff --git a/README.md b/README.md
index fe777ba3ca..764608aa20 100644
--- a/README.md
+++ b/README.md
@@ -59,7 +59,7 @@ If you already have an ONNX model, just [install the runtime](#Installation) for
 ## APIs and Official Builds
 | API Documentation | CPU package | GPU package |
 |-----|-------------|-------------|
-| [Python](https://docs.microsoft.com/en-us/python/api/overview/azure/onnx/intro?view=azure-onnx-py) | [Windows/Linux/Mac](https://pypi.org/project/onnxruntime/)| [Windows/Linux](https://pypi.org/project/onnxruntime-gpu/) |
+| [Python](https://docs.microsoft.com/en-us/python/api/overview/azure/onnx/intro?view=azure-onnx-py) | [Windows/Linux/Mac](https://pypi.org/project/onnxruntime/)| [Windows/Linux](https://pypi.org/project/onnxruntime-gpu/)<br>(Compatible with CUDA9.1 and cuDNN 7.3) |
 | [C#](docs/CSharp_API.md) | [Windows/Linux](https://www.nuget.org/packages/Microsoft.ML.OnnxRuntime/)<br>Mac - Coming Soon| [Windows/Linux](https://www.nuget.org/packages/Microsoft.ML.OnnxRuntime.Gpu/)<br>Mac - Coming Soon|
 | [C](docs/C_API.md) | [Windows/Linux](https://www.nuget.org/packages/Microsoft.ML.OnnxRuntime/)<br>Mac - Coming Soon | [Windows/Linux](https://www.nuget.org/packages/Microsoft.ML.OnnxRuntime.Gpu/)<br>Mac - Coming Soon |
 | [C++](onnxruntime/core/session/inference_session.h) | TBD | TBD |
diff --git a/tools/ci_build/github/linux/docker/scripts/install_deps.sh b/tools/ci_build/github/linux/docker/scripts/install_deps.sh
index 7ca73605c0..13a77a7fa8 100755
--- a/tools/ci_build/github/linux/docker/scripts/install_deps.sh
+++ b/tools/ci_build/github/linux/docker/scripts/install_deps.sh
@@ -28,6 +28,9 @@ for build_type in 'Debug' 'Relwithdebinfo'; do
 done
 export ONNX_ML=1
 INSTALLED_PYTHON_VERSION=$(python3 -c 'import sys; version=sys.version_info[:2]; print("{0}.{1}".format(*version));')
+if [ "$INSTALLED_PYTHON_VERSION" = "3.7" ];then
+  pip3 install --upgrade setuptools
+fi
 if [ "$INSTALLED_PYTHON_VERSION" = "3.4" ];then
   echo "Python 3.5 and above is needed for running onnx tests!" 1>&2
 else