diff --git a/sbv2_core/Cargo.toml b/sbv2_core/Cargo.toml
index 37203d9..b3d5ed0 100644
--- a/sbv2_core/Cargo.toml
+++ b/sbv2_core/Cargo.toml
@@ -21,4 +21,5 @@ tokenizers = "0.20.0"
 [features]
 cuda = ["ort/cuda"]
 cuda_tf32 = []
-dynamic = ["ort/load-dynamic"]
\ No newline at end of file
+dynamic = ["ort/load-dynamic"]
+directml = ["ort/directml"]
\ No newline at end of file
diff --git a/sbv2_core/src/model.rs b/sbv2_core/src/model.rs
index 1598e9b..485941b 100644
--- a/sbv2_core/src/model.rs
+++ b/sbv2_core/src/model.rs
@@ -17,6 +17,10 @@ pub fn load_model<P: AsRef<[u8]>>(model_file: P) -> Result<Session> {
         }
         exp.push(cuda.build());
     }
+    #[cfg(feature = "directml")]
+    {
+        exp.push(ort::DirectMLExecutionProvider::default().build());
+    }
     exp.push(ort::CPUExecutionProvider::default().build());
     Ok(Session::builder()?
         .with_execution_providers(exp)?
@@ -26,6 +30,7 @@ pub fn load_model<P: AsRef<[u8]>>(model_file: P) -> Result<Session> {
         .with_inter_threads(num_cpus::get_physical())?
         .commit_from_memory(model_file.as_ref())?)
 }
+
 #[allow(clippy::too_many_arguments)]
 pub fn synthesize(
     session: &Session,