Unify order of echo parameter to be last in all layers

Author: cmodi-meta

examples/demo-apps/android/LlamaDemo/app/src/main/java/com/example/executorchllamademo/MainActivity.java

@@ -624,8 +624,8 @@ public void run() {
                           ModelUtils.VISION_MODEL_IMAGE_CHANNELS,
                           prompt,
                           ModelUtils.VISION_MODEL_SEQ_LEN,
-                          false,
-                          MainActivity.this);
+                          MainActivity.this,
+                          false);
                     } else {
                       // no image selected, we pass in empty int array
                       mModule.generate(
@@ -635,12 +635,12 @@ public void run() {
                           ModelUtils.VISION_MODEL_IMAGE_CHANNELS,
                           prompt,
                           ModelUtils.VISION_MODEL_SEQ_LEN,
-                          false,
-                          MainActivity.this);
+                          MainActivity.this,
+                          false);
                     }
                   } else {
                     mModule.generate(
-                        prompt, ModelUtils.TEXT_MODEL_SEQ_LEN, false, MainActivity.this);
+                        prompt, ModelUtils.TEXT_MODEL_SEQ_LEN, MainActivity.this, false);
                   }
 
                   long generateDuration = System.currentTimeMillis() - generateStartTime;

extension/android/jni/jni_layer_llama.cpp

@@ -150,8 +150,8 @@ class ExecuTorchLlamaJni
       jint channels,
       facebook::jni::alias_ref<jstring> prompt,
       jint seq_len,
-      jboolean echo,
-      facebook::jni::alias_ref<ExecuTorchLlamaCallbackJni> callback) {
+      facebook::jni::alias_ref<ExecuTorchLlamaCallbackJni> callback,
+      jboolean echo) {
     if (model_type_category_ == MODEL_TYPE_CATEGORY_MULTIMODAL) {
       auto image_size = image->size();
       std::vector<Image> images;
@@ -249,8 +249,8 @@ class ExecuTorchLlamaJni
       facebook::jni::alias_ref<jstring> prompt,
       jint seq_len,
       jlong start_pos,
-      jboolean echo,
-      facebook::jni::alias_ref<ExecuTorchLlamaCallbackJni> callback) {
+      facebook::jni::alias_ref<ExecuTorchLlamaCallbackJni> callback,
+      jboolean echo) {
     if (model_type_category_ != MODEL_TYPE_CATEGORY_MULTIMODAL) {
       return static_cast<jint>(Error::NotSupported);
     }

extension/android/src/main/java/org/pytorch/executorch/LlamaModule.java

@@ -60,7 +60,7 @@ public void resetNative() {
    * @param llamaCallback callback object to receive results.
    */
   public int generate(String prompt, LlamaCallback llamaCallback) {
-    return generate(prompt, DEFAULT_SEQ_LEN, DEFAULT_ECHO, llamaCallback);
+    return generate(prompt, DEFAULT_SEQ_LEN, llamaCallback, DEFAULT_ECHO);
   }
 
   /**
@@ -71,30 +71,30 @@ public int generate(String prompt, LlamaCallback llamaCallback) {
    * @param llamaCallback callback object to receive results.
    */
   public int generate(String prompt, int seqLen, LlamaCallback llamaCallback) {
-    return generate(null, 0, 0, 0, prompt, seqLen, DEFAULT_ECHO, llamaCallback);
+    return generate(null, 0, 0, 0, prompt, seqLen, llamaCallback, DEFAULT_ECHO);
   }
 
   /**
    * Start generating tokens from the module.
    *
    * @param prompt Input prompt
+   * @param llamaCallback callback object to receive results
    * @param echo indicate whether to echo the input prompt or not (text completion vs chat)
-   * @param llamaCallback callback object to receive results.
    */
-  public int generate(String prompt, boolean echo, LlamaCallback llamaCallback) {
-    return generate(null, 0, 0, 0, prompt, DEFAULT_SEQ_LEN, echo, llamaCallback);
+  public int generate(String prompt, LlamaCallback llamaCallback, boolean echo) {
+    return generate(null, 0, 0, 0, prompt, DEFAULT_SEQ_LEN, llamaCallback, echo);
   }
 
   /**
    * Start generating tokens from the module.
    *
    * @param prompt Input prompt
    * @param seqLen sequence length
+   * @param llamaCallback callback object to receive results
    * @param echo indicate whether to echo the input prompt or not (text completion vs chat)
-   * @param llamaCallback callback object to receive results.
    */
-  public int generate(String prompt, int seqLen, boolean echo, LlamaCallback llamaCallback) {
-    return generate(null, 0, 0, 0, prompt, seqLen, echo, llamaCallback);
+  public int generate(String prompt, int seqLen, LlamaCallback llamaCallback, boolean echo) {
+    return generate(null, 0, 0, 0, prompt, seqLen, llamaCallback, echo);
   }
 
   /**
@@ -106,8 +106,8 @@ public int generate(String prompt, int seqLen, boolean echo, LlamaCallback llama
    * @param channels Input image number of channels
    * @param prompt Input prompt
    * @param seqLen sequence length
-   * @param echo indicate whether to echo the input prompt or not (text completion vs chat)
    * @param llamaCallback callback object to receive results.
+   * @param echo indicate whether to echo the input prompt or not (text completion vs chat)
    */
   @DoNotStrip
   public native int generate(
@@ -117,8 +117,8 @@ public native int generate(
       int channels,
       String prompt,
       int seqLen,
-      boolean echo,
-      LlamaCallback llamaCallback);
+      LlamaCallback llamaCallback,
+      boolean echo);
 
   /**
    * Prefill an LLaVA Module with the given images input.
@@ -171,12 +171,12 @@ public long prefillPrompt(String prompt, long startPos, int bos, int eos) {
    * @param prompt The text prompt to LLaVA.
    * @param seqLen The total sequence length, including the prompt tokens and new tokens.
    * @param startPos The starting position in KV cache of the input in the LLM.
-   * @param echo indicate whether to echo the input prompt or not
    * @param llamaCallback callback object to receive results.
+   * @param echo indicate whether to echo the input prompt or not.
    * @return The error code.
    */
   public native int generateFromPos(
-      String prompt, int seqLen, long startPos, boolean echo, LlamaCallback callback);
+      String prompt, int seqLen, long startPos, LlamaCallback callback, boolean echo);
 
   /** Stop current generate() before it finishes. */
   @DoNotStrip