diff --git a/onnxruntime-sys/build.rs b/onnxruntime-sys/build.rs
index 1589555a..e0c46f2a 100644
--- a/onnxruntime-sys/build.rs
+++ b/onnxruntime-sys/build.rs
@@ -66,7 +66,18 @@ fn generate_bindings(_include_dir: &Path) {
 
 #[cfg(feature = "generate-bindings")]
 fn generate_bindings(include_dir: &Path) {
+    let os = env::var("CARGO_CFG_TARGET_OS").expect("Unable to get TARGET_OS");
     let clang_arg = format!("-I{}", include_dir.display());
+    let clang_cuda_arg = match env::var(ORT_ENV_GPU) {
+        Ok(cuda_env) => match cuda_env.to_lowercase().as_str() {
+            "1" | "yes" | "true" | "on" => match os.as_str() {
+                "linux" | "windows" => "-DORT_USE_CUDA",
+                _ => "",
+            },
+            _ => "",
+        },
+        Err(_) => "",
+    };
 
     // Tell cargo to invalidate the built crate whenever the wrapper changes
     println!("cargo:rerun-if-changed=wrapper.h");
@@ -81,6 +92,8 @@ fn generate_bindings(include_dir: &Path) {
         .header("wrapper.h")
         // The current working directory is 'onnxruntime-sys'
         .clang_arg(clang_arg)
+        // Add define ORT_USE_CUDA
+        .clang_arg(clang_cuda_arg)
         // Tell cargo to invalidate the built crate whenever any of the
         // included header files changed.
         .parse_callbacks(Box::new(bindgen::CargoCallbacks))
diff --git a/onnxruntime-sys/wrapper.h b/onnxruntime-sys/wrapper.h
index e63d3523..b11f5275 100644
--- a/onnxruntime-sys/wrapper.h
+++ b/onnxruntime-sys/wrapper.h
@@ -1 +1,5 @@
 #include "onnxruntime_c_api.h"
+#include "cpu_provider_factory.h"
+#ifdef ORT_USE_CUDA
+#include "cuda_provider_factory.h"
+#endif
diff --git a/onnxruntime/Cargo.toml b/onnxruntime/Cargo.toml
index 9ceec820..13d03a2b 100644
--- a/onnxruntime/Cargo.toml
+++ b/onnxruntime/Cargo.toml
@@ -36,6 +36,7 @@ tracing-subscriber = "0.2"
 ureq = "1.5.1"
 
 [features]
+cuda = []
 # Fetch model from ONNX Model Zoo (https://github.com/onnx/models)
 model-fetching = ["ureq"]
 # Disable build script; used for https://docs.rs
diff --git a/onnxruntime/examples/sample.rs b/onnxruntime/examples/sample.rs
index d16d08da..26543142 100644
--- a/onnxruntime/examples/sample.rs
+++ b/onnxruntime/examples/sample.rs
@@ -25,6 +25,15 @@ fn run() -> Result<(), Error> {
 
     tracing::subscriber::set_global_default(subscriber).expect("setting default subscriber failed");
 
+    #[cfg(feature = "cuda")]
+    let environment = Environment::builder()
+        .with_name("test")
+        .with_gpu(0)
+        // The ONNX Runtime's log level can be different than the one of the wrapper crate or the application.
+        .with_log_level(LoggingLevel::Info)
+        .build()?;
+
+    #[cfg(not(feature = "cuda"))]
     let environment = Environment::builder()
         .with_name("test")
         // The ONNX Runtime's log level can be different than the one of the wrapper crate or the application.
diff --git a/onnxruntime/src/session.rs b/onnxruntime/src/session.rs
index 04f9cf1c..e22b7ee0 100644
--- a/onnxruntime/src/session.rs
+++ b/onnxruntime/src/session.rs
@@ -124,6 +124,29 @@ impl<'a> SessionBuilder<'a> {
         Ok(self)
     }
 
+    /// Set the session use cpu provider
+    pub fn with_cpu(self, use_arena: bool) -> Result<SessionBuilder<'a>> {
+        unsafe {
+            sys::OrtSessionOptionsAppendExecutionProvider_CPU(
+                self.session_options_ptr,
+                use_arena.into(),
+            );
+        }
+        Ok(self)
+    }
+
+    /// Set the session use cuda provider
+    #[cfg(feature = "cuda")]
+    pub fn with_cuda(self, device_id: i32) -> Result<SessionBuilder<'a>> {
+        unsafe {
+            sys::OrtSessionOptionsAppendExecutionProvider_CUDA(
+                self.session_options_ptr,
+                device_id.into(),
+            );
+        }
+        Ok(self)
+    }
+
     /// Set the session's allocator
     ///
     /// Defaults to [`AllocatorType::Arena`](../enum.AllocatorType.html#variant.Arena)