diff --git a/extension/android/jni/jni_layer_llama.cpp b/extension/android/jni/jni_layer_llama.cpp
index 5c58c89ee91..e71323da201 100644
--- a/extension/android/jni/jni_layer_llama.cpp
+++ b/extension/android/jni/jni_layer_llama.cpp
@@ -132,16 +132,22 @@ class ExecuTorchLlamaJni
       jint model_type_category,
       facebook::jni::alias_ref<jstring> model_path,
       facebook::jni::alias_ref<jstring> tokenizer_path,
-      jfloat temperature) {
+      jfloat temperature,
+      facebook::jni::alias_ref<jstring> data_path) {
     return makeCxxInstance(
-        model_type_category, model_path, tokenizer_path, temperature);
+        model_type_category,
+        model_path,
+        tokenizer_path,
+        temperature,
+        data_path);
   }
 
   ExecuTorchLlamaJni(
       jint model_type_category,
       facebook::jni::alias_ref<jstring> model_path,
       facebook::jni::alias_ref<jstring> tokenizer_path,
-      jfloat temperature) {
+      jfloat temperature,
+      facebook::jni::alias_ref<jstring> data_path = nullptr) {
 #if defined(ET_USE_THREADPOOL)
     // Reserve 1 thread for the main thread.
     uint32_t num_performant_cores =
@@ -160,10 +166,18 @@ class ExecuTorchLlamaJni
           tokenizer_path->toStdString().c_str(),
           temperature);
     } else if (model_type_category == MODEL_TYPE_CATEGORY_LLM) {
-      runner_ = std::make_unique<example::Runner>(
-          model_path->toStdString().c_str(),
-          tokenizer_path->toStdString().c_str(),
-          temperature);
+      if (data_path != nullptr) {
+        runner_ = std::make_unique<example::Runner>(
+            model_path->toStdString().c_str(),
+            tokenizer_path->toStdString().c_str(),
+            temperature,
+            data_path->toStdString().c_str());
+      } else {
+        runner_ = std::make_unique<example::Runner>(
+            model_path->toStdString().c_str(),
+            tokenizer_path->toStdString().c_str(),
+            temperature);
+      }
 #if defined(EXECUTORCH_BUILD_MEDIATEK)
     } else if (model_type_category == MODEL_TYPE_MEDIATEK_LLAMA) {
       runner_ = std::make_unique<MTKLlamaRunner>(
diff --git a/extension/android/src/main/java/org/pytorch/executorch/LlamaModule.java b/extension/android/src/main/java/org/pytorch/executorch/LlamaModule.java
index 6de26bc7fe8..d64c7d0bf89 100644
--- a/extension/android/src/main/java/org/pytorch/executorch/LlamaModule.java
+++ b/extension/android/src/main/java/org/pytorch/executorch/LlamaModule.java
@@ -39,16 +39,24 @@ public class LlamaModule {
 
   @DoNotStrip
   private static native HybridData initHybrid(
-      int modelType, String modulePath, String tokenizerPath, float temperature);
+      int modelType, String modulePath, String tokenizerPath, float temperature, String dataPath);
 
-  /** Constructs a LLAMA Module for a model with given path, tokenizer, and temperature. */
+  /** Constructs a LLAMA Module for a model with given model path, tokenizer, temperature. */
   public LlamaModule(String modulePath, String tokenizerPath, float temperature) {
-    mHybridData = initHybrid(MODEL_TYPE_TEXT, modulePath, tokenizerPath, temperature);
+    mHybridData = initHybrid(MODEL_TYPE_TEXT, modulePath, tokenizerPath, temperature, null);
+  }
+
+  /**
+   * Constructs a LLAMA Module for a model with given model path, tokenizer, temperature and data
+   * path.
+   */
+  public LlamaModule(String modulePath, String tokenizerPath, float temperature, String dataPath) {
+    mHybridData = initHybrid(MODEL_TYPE_TEXT, modulePath, tokenizerPath, temperature, dataPath);
   }
 
   /** Constructs a LLM Module for a model with given path, tokenizer, and temperature. */
   public LlamaModule(int modelType, String modulePath, String tokenizerPath, float temperature) {
-    mHybridData = initHybrid(modelType, modulePath, tokenizerPath, temperature);
+    mHybridData = initHybrid(modelType, modulePath, tokenizerPath, temperature, null);
   }
 
   public void resetNative() {