From 361b295d92824ba8349d73564bad190803d22955 Mon Sep 17 00:00:00 2001
From: zhenyan-zhang-meta <zhenyanzhang@meta.com>
Date: Tue, 29 Apr 2025 22:43:51 -0700
Subject: [PATCH 1/3] [ExecuTorch] Separate `extension.Module` Namespaces from
 Aten and non-Aten

Separate `extension.Module` Namespaces to be `executorch::extension::module` and `executorch::extension::module_aten`, otherwise in the future there will be issues like P1799454769.

It's similar to what we already did for bundled_program in https://github.com/pytorch/executorch/pull/10307

Differential Revision: [D73903870](https://our.internmc.facebook.com/intern/diff/D73903870/)

[ghstack-poisoned]
---
 docs/source/extension-module.md               |  2 ++
 docs/source/extension-tensor.md               |  2 ++
 docs/source/getting-started.md                |  1 +
 docs/source/llm/getting-started.md            |  2 +-
 docs/source/using-executorch-cpp.md           |  1 +
 examples/llm_manual/main.cpp                  |  2 +-
 examples/models/llama/runner/runner.cpp       |  2 +-
 examples/models/llama/runner/runner.h         |  2 +-
 .../llava/runner/llava_image_prefiller.h      |  3 ++-
 .../llava/runner/llava_text_decoder_runner.h  |  3 ++-
 examples/models/phi-3-mini/runner.cpp         |  2 +-
 examples/models/phi-3-mini/runner.h           |  2 +-
 .../oss_scripts/llama/runner/io_manager.cpp   |  2 +-
 .../oss_scripts/llama/runner/io_manager.h     | 16 +++++++++++-----
 .../oss_scripts/llama/runner/runner.cpp       |  2 +-
 .../oss_scripts/llama/runner/runner.h         |  4 +++-
 .../qaihub_scripts/llama/runner/io_memory.cpp |  2 +-
 .../qaihub_scripts/llama/runner/io_memory.h   | 16 ++++++++++++----
 .../qaihub_scripts/llama/runner/runner.cpp    |  2 +-
 .../qaihub_scripts/llama/runner/runner.h      |  4 +++-
 .../stable_diffusion/runner/runner.cpp        |  2 +-
 .../stable_diffusion/runner/runner.h          |  4 +++-
 extension/android/jni/jni_layer.cpp           |  1 +
 .../ExecuTorch/Exported/ExecuTorchModule.mm   |  1 +
 extension/llm/runner/image_prefiller.h        |  5 ++++-
 extension/llm/runner/multimodal_runner.h      |  2 ++
 extension/llm/runner/text_decoder_runner.h    |  2 ++
 extension/module/module.cpp                   |  2 ++
 extension/module/module.h                     | 10 +++++++++-
 extension/module/test/module_test.cpp         |  1 +
 extension/training/module/training_module.cpp | 19 ++++++++++++-------
 extension/training/module/training_module.h   |  4 ++--
 32 files changed, 89 insertions(+), 36 deletions(-)

diff --git a/docs/source/extension-module.md b/docs/source/extension-module.md
index 835c5c12e27..d5bcddf9a65 100644
--- a/docs/source/extension-module.md
+++ b/docs/source/extension-module.md
@@ -13,6 +13,7 @@ Let's see how we can run the `SimpleConv` model generated from the [Exporting to
 #include <executorch/extension/tensor/tensor.h>
 
 using namespace ::executorch::extension;
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 // Create a Module.
 Module module("/path/to/model.pte");
@@ -221,6 +222,7 @@ Use [ExecuTorch Dump](etdump.md) to trace model execution. Create an `ETDumpGen`
 #include <executorch/devtools/etdump/etdump_flatcc.h>
 
 using namespace ::executorch::extension;
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 Module module("/path/to/model.pte", Module::LoadMode::MmapUseMlock, std::make_unique<ETDumpGen>());
 
diff --git a/docs/source/extension-tensor.md b/docs/source/extension-tensor.md
index 910c06053ed..fdfc72a2227 100644
--- a/docs/source/extension-tensor.md
+++ b/docs/source/extension-tensor.md
@@ -11,6 +11,7 @@ Imagine you’re working with a [`Module`](extension-module.md) interface, and y
 
 using namespace executorch::aten;
 using namespace executorch::extension;
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 SizesType sizes[] = {2, 3};
 DimOrderType dim_order[] = {0, 1};
@@ -42,6 +43,7 @@ Here’s how you can use it:
 #include <executorch/extension/tensor/tensor.h>
 
 using namespace executorch::extension;
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 auto tensor = make_tensor_ptr(
     {2, 3},                                // sizes
diff --git a/docs/source/getting-started.md b/docs/source/getting-started.md
index b7a97190b49..a2c64d07663 100644
--- a/docs/source/getting-started.md
+++ b/docs/source/getting-started.md
@@ -194,6 +194,7 @@ Loading and running a model using the high-level API can be done as follows:
 #include <executorch/extension/tensor/tensor.h>
 
 using namespace ::executorch::extension;
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 // Load the model.
 Module module("/path/to/model.pte");
diff --git a/docs/source/llm/getting-started.md b/docs/source/llm/getting-started.md
index 152162841e4..522529a1eac 100644
--- a/docs/source/llm/getting-started.md
+++ b/docs/source/llm/getting-started.md
@@ -209,7 +209,7 @@ Create a file called main.cpp with the following contents:
 using executorch::aten::ScalarType;
 using executorch::aten::Tensor;
 using executorch::extension::from_blob;
-using executorch::extension::Module;
+using executorch::extension::ET_MODULE_NAMESPACE::Module;
 using executorch::runtime::EValue;
 using executorch::runtime::Result;
 ```
diff --git a/docs/source/using-executorch-cpp.md b/docs/source/using-executorch-cpp.md
index d64dad97da9..46c0c15c39b 100644
--- a/docs/source/using-executorch-cpp.md
+++ b/docs/source/using-executorch-cpp.md
@@ -13,6 +13,7 @@ In addition the Module class, the tensor extension provides an encapsulated inte
 #include <executorch/extension/tensor/tensor.h>
 
 using namespace ::executorch::extension;
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 // Load the model.
 Module module("/path/to/model.pte");
diff --git a/examples/llm_manual/main.cpp b/examples/llm_manual/main.cpp
index 76492513f91..a1ae207db96 100644
--- a/examples/llm_manual/main.cpp
+++ b/examples/llm_manual/main.cpp
@@ -20,7 +20,7 @@
 using executorch::aten::ScalarType;
 using executorch::aten::Tensor;
 using executorch::extension::from_blob;
-using executorch::extension::Module;
+using executorch::extension::ET_MODULE_NAMESPACE::Module;
 using executorch::runtime::EValue;
 using executorch::runtime::Result;
 
diff --git a/examples/models/llama/runner/runner.cpp b/examples/models/llama/runner/runner.cpp
index 186c2013616..c6df62658fd 100644
--- a/examples/models/llama/runner/runner.cpp
+++ b/examples/models/llama/runner/runner.cpp
@@ -21,7 +21,7 @@
 
 namespace example {
 
-using ::executorch::extension::Module;
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 using ::executorch::runtime::Error;
 using ::executorch::runtime::Result;
 
diff --git a/examples/models/llama/runner/runner.h b/examples/models/llama/runner/runner.h
index 97ffe4b98b7..9cc9deaf586 100644
--- a/examples/models/llama/runner/runner.h
+++ b/examples/models/llama/runner/runner.h
@@ -60,7 +60,7 @@ class ET_EXPERIMENTAL Runner : public executorch::extension::llm::IRunner {
   bool shouldStop_{false};
 
   // model
-  std::unique_ptr<::executorch::extension::Module> module_;
+  std::unique_ptr<::executorch::extension::ET_MODULE_NAMESPACE::Module> module_;
   std::string tokenizer_path_;
   std::unique_ptr<::tokenizers::Tokenizer> tokenizer_;
   std::unordered_map<std::string, int64_t> metadata_;
diff --git a/examples/models/llava/runner/llava_image_prefiller.h b/examples/models/llava/runner/llava_image_prefiller.h
index 762a28d0d07..aa377d1528d 100644
--- a/examples/models/llava/runner/llava_image_prefiller.h
+++ b/examples/models/llava/runner/llava_image_prefiller.h
@@ -18,7 +18,8 @@ namespace example {
 class ET_EXPERIMENTAL LlavaImagePrefiller
     : public ::executorch::extension::llm::ImagePrefiller {
  public:
-  explicit LlavaImagePrefiller(::executorch::extension::Module* module)
+  explicit LlavaImagePrefiller(
+      ::executorch::extension::ET_MODULE_NAMESPACE::Module* module)
       : ImagePrefiller(module){};
   /**
    * Prefill an LLM Module with the given image input.
diff --git a/examples/models/llava/runner/llava_text_decoder_runner.h b/examples/models/llava/runner/llava_text_decoder_runner.h
index 3de418b57ea..cbec7dd927d 100644
--- a/examples/models/llava/runner/llava_text_decoder_runner.h
+++ b/examples/models/llava/runner/llava_text_decoder_runner.h
@@ -17,7 +17,8 @@ namespace example {
 class ET_EXPERIMENTAL LlavaTextDecoderRunner
     : public executorch::extension::llm::TextDecoderRunner {
  public:
-  explicit LlavaTextDecoderRunner(executorch::extension::Module* module)
+  explicit LlavaTextDecoderRunner(
+      executorch::extension::ET_MODULE_NAMESPACE::Module* module)
       : TextDecoderRunner(module, true) {}
 
   inline executorch::runtime::Result<executorch::aten::Tensor> step(
diff --git a/examples/models/phi-3-mini/runner.cpp b/examples/models/phi-3-mini/runner.cpp
index 15f76e9522c..8b5457c62ec 100644
--- a/examples/models/phi-3-mini/runner.cpp
+++ b/examples/models/phi-3-mini/runner.cpp
@@ -16,7 +16,7 @@
 #include <pytorch/tokenizers/llama2c_tokenizer.h>
 
 using executorch::aten::ScalarType;
-using executorch::extension::Module;
+using executorch::extension::ET_MODULE_NAMESPACE::Module;
 using executorch::extension::llm::Sampler;
 using executorch::runtime::Error;
 using tokenizers::Llama2cTokenizer;
diff --git a/examples/models/phi-3-mini/runner.h b/examples/models/phi-3-mini/runner.h
index 2f0042a57ea..fcb6cd71efa 100644
--- a/examples/models/phi-3-mini/runner.h
+++ b/examples/models/phi-3-mini/runner.h
@@ -42,7 +42,7 @@ class Runner {
   uint64_t prefill(std::vector<uint64_t>& tokens);
   uint64_t run_model_step(uint64_t token);
 
-  std::unique_ptr<executorch::extension::Module> module_;
+  std::unique_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module> module_;
   std::unique_ptr<tokenizers::Tokenizer> tokenizer_;
   std::unique_ptr<executorch::extension::llm::Sampler> sampler_;
 };
diff --git a/examples/qualcomm/oss_scripts/llama/runner/io_manager.cpp b/examples/qualcomm/oss_scripts/llama/runner/io_manager.cpp
index c2bf7b04fbb..8bdcc462923 100644
--- a/examples/qualcomm/oss_scripts/llama/runner/io_manager.cpp
+++ b/examples/qualcomm/oss_scripts/llama/runner/io_manager.cpp
@@ -12,7 +12,7 @@
 
 using executorch::aten::Tensor;
 using executorch::aten::TensorImpl;
-using executorch::extension::Module;
+using executorch::extension::ET_MODULE_NAMESPACE::Module;
 using executorch::runtime::Error;
 using executorch::runtime::MemoryAllocator;
 using executorch::runtime::MethodMeta;
diff --git a/examples/qualcomm/oss_scripts/llama/runner/io_manager.h b/examples/qualcomm/oss_scripts/llama/runner/io_manager.h
index 0f10eef8ddc..5a305fa7406 100644
--- a/examples/qualcomm/oss_scripts/llama/runner/io_manager.h
+++ b/examples/qualcomm/oss_scripts/llama/runner/io_manager.h
@@ -20,6 +20,8 @@
 #include <executorch/extension/module/module.h>
 #include <executorch/runtime/executor/method_meta.h>
 
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
+
 namespace example {
 
 enum EvalMode {
@@ -29,8 +31,8 @@ enum EvalMode {
 };
 class IoMgrBase {
  public:
-  IoMgrBase(
-      std::vector<std::shared_ptr<executorch::extension::Module>>& modules);
+  IoMgrBase(std::vector<std::shared_ptr<
+                executorch::extension::ET_MODULE_NAMESPACE::Module>>& modules);
   virtual ~IoMgrBase();
   virtual void init_io() = 0;
   virtual void reset_io(
@@ -81,13 +83,16 @@ class IoMgrBase {
       std::string,
       std::vector<std::vector<executorch::aten::TensorImpl*>>>
       output_tensors_;
-  std::vector<std::shared_ptr<executorch::extension::Module>> modules_;
+  std::vector<
+      std::shared_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>
+      modules_;
 };
 
 class ShiftPointerIoMgr : public IoMgrBase {
  public:
   ShiftPointerIoMgr(
-      std::vector<std::shared_ptr<executorch::extension::Module>>& modules,
+      std::vector<std::shared_ptr<
+          executorch::extension::ET_MODULE_NAMESPACE::Module>>& modules,
       int32_t context_len,
       int32_t prefill_ar_len,
       int32_t prefill_cache_len,
@@ -199,7 +204,8 @@ class ShiftPointerIoMgr : public IoMgrBase {
 class SmartMaskIoMgr : public IoMgrBase {
  public:
   SmartMaskIoMgr(
-      std::vector<std::shared_ptr<executorch::extension::Module>>& modules,
+      std::vector<std::shared_ptr<
+          executorch::extension::ET_MODULE_NAMESPACE::Module>>& modules,
       int32_t context_len,
       int32_t prefill_ar_len,
       int32_t prefill_cache_len,
diff --git a/examples/qualcomm/oss_scripts/llama/runner/runner.cpp b/examples/qualcomm/oss_scripts/llama/runner/runner.cpp
index dafc911a172..7e7c48a517e 100644
--- a/examples/qualcomm/oss_scripts/llama/runner/runner.cpp
+++ b/examples/qualcomm/oss_scripts/llama/runner/runner.cpp
@@ -23,7 +23,7 @@
 #include <sstream>
 
 using executorch::aten::Tensor;
-using executorch::extension::Module;
+using executorch::extension::ET_MODULE_NAMESPACE::Module;
 using executorch::extension::llm::Sampler;
 using executorch::extension::llm::time_in_ms;
 using executorch::runtime::Error;
diff --git a/examples/qualcomm/oss_scripts/llama/runner/runner.h b/examples/qualcomm/oss_scripts/llama/runner/runner.h
index e693bcd7077..93c3440a503 100644
--- a/examples/qualcomm/oss_scripts/llama/runner/runner.h
+++ b/examples/qualcomm/oss_scripts/llama/runner/runner.h
@@ -101,7 +101,9 @@ class Runner {
   std::unordered_set<uint64_t> eos_id_;
   const int32_t n_bos_;
   const int32_t n_eos_;
-  std::vector<std::shared_ptr<executorch::extension::Module>> modules_;
+  std::vector<
+      std::shared_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>
+      modules_;
   std::string tokenizer_path_;
   std::string performance_output_path_;
   float logits_scale_;
diff --git a/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.cpp b/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.cpp
index 9ee7551650a..8a5ddcd7ded 100644
--- a/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.cpp
+++ b/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.cpp
@@ -14,7 +14,7 @@
 
 using executorch::aten::Tensor;
 using executorch::aten::TensorImpl;
-using executorch::extension::Module;
+using executorch::extension::ET_MODULE_NAMESPACE::Module;
 using executorch::runtime::Error;
 using executorch::runtime::MethodMeta;
 using executorch::runtime::Result;
diff --git a/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.h b/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.h
index 445be2ed21a..012f40404b4 100644
--- a/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.h
+++ b/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.h
@@ -32,7 +32,9 @@ class Memory {
  public:
   Memory(
       const std::vector<std::string>& pos_embs_path,
-      std::vector<std::shared_ptr<executorch::extension::Module>>& modules);
+      std::vector<
+          std::shared_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>&
+          modules);
   virtual ~Memory();
   virtual void prepare_io(
       const std::vector<
@@ -51,7 +53,9 @@ class Memory {
   std::vector<std::vector<executorch::aten::TensorImpl*>> input_tensors_;
   std::vector<std::vector<executorch::aten::TensorImpl*>> output_tensors_;
   std::vector<std::string> pos_embs_path_;
-  std::vector<std::shared_ptr<executorch::extension::Module>> modules_;
+  std::vector<
+      std::shared_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>
+      modules_;
   std::vector<std::string> method_names_;
 };
 
@@ -59,7 +63,9 @@ class BertMemory : public Memory {
  public:
   BertMemory(
       const std::vector<std::string>& pos_embs_path,
-      std::vector<std::shared_ptr<executorch::extension::Module>>& modules,
+      std::vector<
+          std::shared_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>&
+          modules,
       std::vector<int> shard_layers);
   void prepare_io(const std::vector<executorch::runtime::Result<
                       executorch::runtime::MethodMeta>>& methods_meta) override;
@@ -121,7 +127,9 @@ class KVCachedMemory : public Memory {
  public:
   KVCachedMemory(
       const std::vector<std::string>& pos_embs_path,
-      std::vector<std::shared_ptr<executorch::extension::Module>>& modules,
+      std::vector<
+          std::shared_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>&
+          modules,
       std::vector<int> shard_layers);
   void prepare_io(const std::vector<executorch::runtime::Result<
                       executorch::runtime::MethodMeta>>& methods_meta) override;
diff --git a/examples/qualcomm/qaihub_scripts/llama/runner/runner.cpp b/examples/qualcomm/qaihub_scripts/llama/runner/runner.cpp
index 06ea324ef6f..cb98fcf04f3 100644
--- a/examples/qualcomm/qaihub_scripts/llama/runner/runner.cpp
+++ b/examples/qualcomm/qaihub_scripts/llama/runner/runner.cpp
@@ -30,7 +30,7 @@
 #endif
 
 using executorch::aten::Tensor;
-using executorch::extension::Module;
+using executorch::extension::ET_MODULE_NAMESPACE::Module;
 using executorch::extension::llm::Sampler;
 using executorch::extension::llm::time_in_ms;
 using executorch::runtime::Error;
diff --git a/examples/qualcomm/qaihub_scripts/llama/runner/runner.h b/examples/qualcomm/qaihub_scripts/llama/runner/runner.h
index 9672d6a3586..effd2ca8c50 100644
--- a/examples/qualcomm/qaihub_scripts/llama/runner/runner.h
+++ b/examples/qualcomm/qaihub_scripts/llama/runner/runner.h
@@ -97,7 +97,9 @@ class Runner {
   const int32_t vocab_size_;
   const int32_t max_seq_len_;
   int32_t eval_mode_;
-  std::vector<std::shared_ptr<executorch::extension::Module>> modules_;
+  std::vector<
+      std::shared_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>
+      modules_;
   std::vector<std::string> method_names_;
   std::string tokenizer_path_;
   float temperature_;
diff --git a/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.cpp b/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.cpp
index 585d58b21ee..90091c3b870 100644
--- a/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.cpp
+++ b/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.cpp
@@ -23,8 +23,8 @@
 #include <executorch/runtime/platform/log.h>
 
 using executorch::extension::from_blob;
-using executorch::extension::Module;
 using executorch::extension::TensorPtr;
+using executorch::extension::ET_MODULE_NAMESPACE::Module;
 using executorch::extension::llm::time_in_ms;
 using executorch::runtime::Error;
 using executorch::runtime::MethodMeta;
diff --git a/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.h b/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.h
index e49201bca25..df37e4829bd 100644
--- a/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.h
+++ b/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.h
@@ -111,7 +111,9 @@ class Runner {
 
  private:
   Stats stats_;
-  std::vector<std::unique_ptr<executorch::extension::Module>> modules_;
+  std::vector<
+      std::unique_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>
+      modules_;
   std::vector<std::string> method_names_;
   std::vector<std::vector<uint16_t>> time_emb_list_;
   std::unordered_map<std::string, int32_t> vocab_to_token_map_;
diff --git a/extension/android/jni/jni_layer.cpp b/extension/android/jni/jni_layer.cpp
index f3c62e1d70f..ebbd3780c39 100644
--- a/extension/android/jni/jni_layer.cpp
+++ b/extension/android/jni/jni_layer.cpp
@@ -36,6 +36,7 @@
 
 using namespace executorch::extension;
 using namespace torch::executor;
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 namespace executorch::extension {
 class TensorHybrid : public facebook::jni::HybridClass<TensorHybrid> {
diff --git a/extension/apple/ExecuTorch/Exported/ExecuTorchModule.mm b/extension/apple/ExecuTorch/Exported/ExecuTorchModule.mm
index 98f0f555991..414fba7dc13 100644
--- a/extension/apple/ExecuTorch/Exported/ExecuTorchModule.mm
+++ b/extension/apple/ExecuTorch/Exported/ExecuTorchModule.mm
@@ -15,6 +15,7 @@
 
 using namespace executorch::extension;
 using namespace executorch::runtime;
+using namespace executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 static inline EValue toEValue(ExecuTorchValue *value) {
   if (value.isTensor) {
diff --git a/extension/llm/runner/image_prefiller.h b/extension/llm/runner/image_prefiller.h
index 826cea24db9..aaca29fb5f0 100644
--- a/extension/llm/runner/image_prefiller.h
+++ b/extension/llm/runner/image_prefiller.h
@@ -14,6 +14,8 @@
 #include <executorch/extension/module/module.h>
 #include <executorch/runtime/platform/compiler.h>
 
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
+
 namespace executorch {
 namespace extension {
 namespace llm {
@@ -21,7 +23,8 @@ namespace llm {
 // Assuming kv cache and parallel prefill are enabled.
 class ET_EXPERIMENTAL ImagePrefiller {
  public:
-  explicit ImagePrefiller(::executorch::extension::Module* module)
+  explicit ImagePrefiller(
+      ::executorch::extension::ET_MODULE_NAMESPACE::Module* module)
       : module_(module) {}
 
   /**
diff --git a/extension/llm/runner/multimodal_runner.h b/extension/llm/runner/multimodal_runner.h
index c17e039c11b..fc66e56553d 100644
--- a/extension/llm/runner/multimodal_runner.h
+++ b/extension/llm/runner/multimodal_runner.h
@@ -29,6 +29,8 @@
 #include <executorch/extension/module/module.h>
 #include <pytorch/tokenizers/tokenizer.h>
 
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
+
 namespace executorch {
 namespace extension {
 namespace llm {
diff --git a/extension/llm/runner/text_decoder_runner.h b/extension/llm/runner/text_decoder_runner.h
index 6c1256c6b90..7f40a7efae0 100644
--- a/extension/llm/runner/text_decoder_runner.h
+++ b/extension/llm/runner/text_decoder_runner.h
@@ -15,6 +15,8 @@
 #include <executorch/extension/tensor/tensor.h>
 #include <executorch/runtime/platform/compiler.h>
 
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
+
 namespace executorch {
 namespace extension {
 namespace llm {
diff --git a/extension/module/module.cpp b/extension/module/module.cpp
index ec01323edc7..2bab588569f 100644
--- a/extension/module/module.cpp
+++ b/extension/module/module.cpp
@@ -36,6 +36,7 @@
 
 namespace executorch {
 namespace extension {
+namespace ET_MODULE_NAMESPACE {
 
 using ET_RUNTIME_NAMESPACE::MethodMeta;
 using ET_RUNTIME_NAMESPACE::Program;
@@ -312,5 +313,6 @@ ET_NODISCARD inline runtime::Result<Method*> Module::get_method(
   return methods_[method_name].method.get();
 }
 
+} // namespace ET_MODULE_NAMESPACE
 } // namespace extension
 } // namespace executorch
diff --git a/extension/module/module.h b/extension/module/module.h
index 201887b9ccc..46e35dd7ca5 100644
--- a/extension/module/module.h
+++ b/extension/module/module.h
@@ -16,8 +16,15 @@
 
 #include <executorch/runtime/executor/program.h>
 
+#ifdef USE_ATEN_LIB
+#define ET_MODULE_NAMESPACE module::aten
+#else // !USE_ATEN_LIB
+#define ET_MODULE_NAMESPACE module
+#endif // USE_ATEN_LIB
+
 namespace executorch {
 namespace extension {
+namespace ET_MODULE_NAMESPACE {
 
 using ET_RUNTIME_NAMESPACE::Method;
 using ET_RUNTIME_NAMESPACE::MethodMeta;
@@ -506,6 +513,7 @@ class Module {
   friend class ExecuTorchJni;
 };
 
+} // namespace ET_MODULE_NAMESPACE
 } // namespace extension
 } // namespace executorch
 
@@ -513,6 +521,6 @@ namespace torch {
 namespace executor {
 // TODO(T197294990): Remove these deprecated aliases once all users have moved
 // to the new `::executorch` namespaces.
-using ::executorch::extension::Module;
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 } // namespace executor
 } // namespace torch
diff --git a/extension/module/test/module_test.cpp b/extension/module/test/module_test.cpp
index a82e257a703..472af2cecdb 100644
--- a/extension/module/test/module_test.cpp
+++ b/extension/module/test/module_test.cpp
@@ -18,6 +18,7 @@
 
 using namespace ::executorch::extension;
 using namespace ::executorch::runtime;
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 class ModuleTest : public ::testing::Test {
  protected:
diff --git a/extension/training/module/training_module.cpp b/extension/training/module/training_module.cpp
index d119738715e..6b3015a9d79 100644
--- a/extension/training/module/training_module.cpp
+++ b/extension/training/module/training_module.cpp
@@ -25,7 +25,8 @@ TrainingModule::execute_forward_backward(
   // Find where the user outputs end.
   const std::string gradients_method_name =
       gradients_method_prefix + method_name;
-  auto res = executorch::extension::Module::execute(gradients_method_name);
+  auto res = executorch::extension::ET_MODULE_NAMESPACE::Module::execute(
+      gradients_method_name);
   if (!res.ok()) {
     return res.error();
   }
@@ -34,8 +35,8 @@ TrainingModule::execute_forward_backward(
   const std::string parameters_method_name =
       parameters_method_prefix + method_name;
   // get params start.
-  auto param_res =
-      executorch::extension::Module::execute(parameters_method_name);
+  auto param_res = executorch::extension::ET_MODULE_NAMESPACE::Module::execute(
+      parameters_method_name);
   if (!param_res.ok()) {
     return param_res.error();
   }
@@ -67,7 +68,8 @@ TrainingModule::execute_forward_backward(
 
     // Get names if we havent seen this method before.
     const std::string fqn_method_name = fqn_method_prefix + method_name;
-    auto fqn_res = executorch::extension::Module::execute(fqn_method_name);
+    auto fqn_res = executorch::extension::ET_MODULE_NAMESPACE::Module::execute(
+        fqn_method_name);
     if (!fqn_res.ok()) {
       return fqn_res.error();
     }
@@ -100,7 +102,8 @@ TrainingModule::named_parameters(const std::string& method_name) {
     method_named_parameters_.insert({method_name, {}});
 
     // get names.
-    auto fqn_res = executorch::extension::Module::execute(fqn_method_name);
+    auto fqn_res = executorch::extension::ET_MODULE_NAMESPACE::Module::execute(
+        fqn_method_name);
     if (!fqn_res.ok()) {
       return fqn_res.error();
     }
@@ -108,7 +111,8 @@ TrainingModule::named_parameters(const std::string& method_name) {
 
     // get params start.
     auto param_res =
-        executorch::extension::Module::execute(parameters_method_name);
+        executorch::extension::ET_MODULE_NAMESPACE::Module::execute(
+            parameters_method_name);
     if (!param_res.ok()) {
       return param_res.error();
     }
@@ -116,7 +120,8 @@ TrainingModule::named_parameters(const std::string& method_name) {
     uint64_t param_start = param_res.get()[0].toInt();
 
     // Load the method if it is not already loaded.
-    auto e = executorch::extension::Module::load_method(method_name);
+    auto e = executorch::extension::ET_MODULE_NAMESPACE::Module::load_method(
+        method_name);
     if (e != runtime::Error::Ok) {
       return e;
     }
diff --git a/extension/training/module/training_module.h b/extension/training/module/training_module.h
index 7bf81623c04..1d48355b7ba 100644
--- a/extension/training/module/training_module.h
+++ b/extension/training/module/training_module.h
@@ -27,7 +27,7 @@ namespace training {
  * methods within them.
  */
 class ET_EXPERIMENTAL TrainingModule final
-    : public executorch::extension::Module {
+    : public executorch::extension::ET_MODULE_NAMESPACE::Module {
  public:
   explicit TrainingModule(
       std::unique_ptr<runtime::DataLoader> data_loader,
@@ -35,7 +35,7 @@ class ET_EXPERIMENTAL TrainingModule final
       std::unique_ptr<runtime::MemoryAllocator> temp_allocator = nullptr,
       std::unique_ptr<runtime::EventTracer> event_tracer = nullptr,
       std::unique_ptr<runtime::DataLoader> data_map_data_loader = nullptr)
-      : executorch::extension::Module(
+      : executorch::extension::ET_MODULE_NAMESPACE::Module(
             std::move(data_loader),
             std::move(memory_allocator),
             std::move(temp_allocator),

From 89c020d154abf27e7ade0c4d864965b527af08e0 Mon Sep 17 00:00:00 2001
From: zhenyan-zhang-meta <zhenyanzhang@meta.com>
Date: Wed, 30 Apr 2025 00:29:39 -0700
Subject: [PATCH 2/3] Update on "[ExecuTorch] Separate `extension.Module`
 Namespaces from Aten and non-Aten"

Separate `extension.Module` Namespaces to be `executorch::extension::module` and `executorch::extension::module_aten`, otherwise in the future there will be issues like P1799454769.

It's similar to what we already did for bundled_program in https://github.com/pytorch/executorch/pull/10307

Differential Revision: [D73903870](https://our.internmc.facebook.com/intern/diff/D73903870/)

[ghstack-poisoned]
---
 docs/source/extension-module.md               |  2 --
 docs/source/extension-tensor.md               |  2 --
 docs/source/getting-started.md                |  1 -
 docs/source/llm/getting-started.md            |  2 +-
 docs/source/using-executorch-cpp.md           |  1 -
 examples/llm_manual/main.cpp                  |  2 +-
 examples/models/llama/runner/runner.cpp       |  2 +-
 examples/models/llama/runner/runner.h         |  2 +-
 .../llava/runner/llava_image_prefiller.h      |  3 +--
 .../llava/runner/llava_text_decoder_runner.h  |  3 +--
 examples/models/phi-3-mini/runner.cpp         |  2 +-
 examples/models/phi-3-mini/runner.h           |  2 +-
 .../oss_scripts/llama/runner/io_manager.cpp   |  2 +-
 .../oss_scripts/llama/runner/io_manager.h     | 16 +++++-----------
 .../oss_scripts/llama/runner/runner.cpp       |  2 +-
 .../oss_scripts/llama/runner/runner.h         |  4 +---
 .../qaihub_scripts/llama/runner/io_memory.cpp |  2 +-
 .../qaihub_scripts/llama/runner/io_memory.h   | 16 ++++------------
 .../qaihub_scripts/llama/runner/runner.cpp    |  2 +-
 .../qaihub_scripts/llama/runner/runner.h      |  4 +---
 .../stable_diffusion/runner/runner.cpp        |  2 +-
 .../stable_diffusion/runner/runner.h          |  4 +---
 extension/android/jni/jni_layer.cpp           |  1 -
 .../ExecuTorch/Exported/ExecuTorchModule.mm   |  1 -
 extension/llm/runner/image_prefiller.h        |  5 +----
 extension/llm/runner/multimodal_runner.h      |  2 --
 extension/llm/runner/text_decoder_runner.h    |  2 --
 extension/module/module.cpp                   |  2 --
 extension/module/module.h                     |  7 +++++++
 extension/module/test/module_test.cpp         |  1 -
 extension/training/module/training_module.cpp | 19 +++++++------------
 extension/training/module/training_module.h   |  4 ++--
 32 files changed, 42 insertions(+), 80 deletions(-)

diff --git a/docs/source/extension-module.md b/docs/source/extension-module.md
index d5bcddf9a65..835c5c12e27 100644
--- a/docs/source/extension-module.md
+++ b/docs/source/extension-module.md
@@ -13,7 +13,6 @@ Let's see how we can run the `SimpleConv` model generated from the [Exporting to
 #include <executorch/extension/tensor/tensor.h>
 
 using namespace ::executorch::extension;
-using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 // Create a Module.
 Module module("/path/to/model.pte");
@@ -222,7 +221,6 @@ Use [ExecuTorch Dump](etdump.md) to trace model execution. Create an `ETDumpGen`
 #include <executorch/devtools/etdump/etdump_flatcc.h>
 
 using namespace ::executorch::extension;
-using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 Module module("/path/to/model.pte", Module::LoadMode::MmapUseMlock, std::make_unique<ETDumpGen>());
 
diff --git a/docs/source/extension-tensor.md b/docs/source/extension-tensor.md
index fdfc72a2227..910c06053ed 100644
--- a/docs/source/extension-tensor.md
+++ b/docs/source/extension-tensor.md
@@ -11,7 +11,6 @@ Imagine you’re working with a [`Module`](extension-module.md) interface, and y
 
 using namespace executorch::aten;
 using namespace executorch::extension;
-using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 SizesType sizes[] = {2, 3};
 DimOrderType dim_order[] = {0, 1};
@@ -43,7 +42,6 @@ Here’s how you can use it:
 #include <executorch/extension/tensor/tensor.h>
 
 using namespace executorch::extension;
-using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 auto tensor = make_tensor_ptr(
     {2, 3},                                // sizes
diff --git a/docs/source/getting-started.md b/docs/source/getting-started.md
index a2c64d07663..b7a97190b49 100644
--- a/docs/source/getting-started.md
+++ b/docs/source/getting-started.md
@@ -194,7 +194,6 @@ Loading and running a model using the high-level API can be done as follows:
 #include <executorch/extension/tensor/tensor.h>
 
 using namespace ::executorch::extension;
-using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 // Load the model.
 Module module("/path/to/model.pte");
diff --git a/docs/source/llm/getting-started.md b/docs/source/llm/getting-started.md
index 522529a1eac..152162841e4 100644
--- a/docs/source/llm/getting-started.md
+++ b/docs/source/llm/getting-started.md
@@ -209,7 +209,7 @@ Create a file called main.cpp with the following contents:
 using executorch::aten::ScalarType;
 using executorch::aten::Tensor;
 using executorch::extension::from_blob;
-using executorch::extension::ET_MODULE_NAMESPACE::Module;
+using executorch::extension::Module;
 using executorch::runtime::EValue;
 using executorch::runtime::Result;
 ```
diff --git a/docs/source/using-executorch-cpp.md b/docs/source/using-executorch-cpp.md
index 46c0c15c39b..d64dad97da9 100644
--- a/docs/source/using-executorch-cpp.md
+++ b/docs/source/using-executorch-cpp.md
@@ -13,7 +13,6 @@ In addition the Module class, the tensor extension provides an encapsulated inte
 #include <executorch/extension/tensor/tensor.h>
 
 using namespace ::executorch::extension;
-using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 // Load the model.
 Module module("/path/to/model.pte");
diff --git a/examples/llm_manual/main.cpp b/examples/llm_manual/main.cpp
index a1ae207db96..76492513f91 100644
--- a/examples/llm_manual/main.cpp
+++ b/examples/llm_manual/main.cpp
@@ -20,7 +20,7 @@
 using executorch::aten::ScalarType;
 using executorch::aten::Tensor;
 using executorch::extension::from_blob;
-using executorch::extension::ET_MODULE_NAMESPACE::Module;
+using executorch::extension::Module;
 using executorch::runtime::EValue;
 using executorch::runtime::Result;
 
diff --git a/examples/models/llama/runner/runner.cpp b/examples/models/llama/runner/runner.cpp
index c6df62658fd..186c2013616 100644
--- a/examples/models/llama/runner/runner.cpp
+++ b/examples/models/llama/runner/runner.cpp
@@ -21,7 +21,7 @@
 
 namespace example {
 
-using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
+using ::executorch::extension::Module;
 using ::executorch::runtime::Error;
 using ::executorch::runtime::Result;
 
diff --git a/examples/models/llama/runner/runner.h b/examples/models/llama/runner/runner.h
index 9cc9deaf586..97ffe4b98b7 100644
--- a/examples/models/llama/runner/runner.h
+++ b/examples/models/llama/runner/runner.h
@@ -60,7 +60,7 @@ class ET_EXPERIMENTAL Runner : public executorch::extension::llm::IRunner {
   bool shouldStop_{false};
 
   // model
-  std::unique_ptr<::executorch::extension::ET_MODULE_NAMESPACE::Module> module_;
+  std::unique_ptr<::executorch::extension::Module> module_;
   std::string tokenizer_path_;
   std::unique_ptr<::tokenizers::Tokenizer> tokenizer_;
   std::unordered_map<std::string, int64_t> metadata_;
diff --git a/examples/models/llava/runner/llava_image_prefiller.h b/examples/models/llava/runner/llava_image_prefiller.h
index aa377d1528d..762a28d0d07 100644
--- a/examples/models/llava/runner/llava_image_prefiller.h
+++ b/examples/models/llava/runner/llava_image_prefiller.h
@@ -18,8 +18,7 @@ namespace example {
 class ET_EXPERIMENTAL LlavaImagePrefiller
     : public ::executorch::extension::llm::ImagePrefiller {
  public:
-  explicit LlavaImagePrefiller(
-      ::executorch::extension::ET_MODULE_NAMESPACE::Module* module)
+  explicit LlavaImagePrefiller(::executorch::extension::Module* module)
       : ImagePrefiller(module){};
   /**
    * Prefill an LLM Module with the given image input.
diff --git a/examples/models/llava/runner/llava_text_decoder_runner.h b/examples/models/llava/runner/llava_text_decoder_runner.h
index cbec7dd927d..3de418b57ea 100644
--- a/examples/models/llava/runner/llava_text_decoder_runner.h
+++ b/examples/models/llava/runner/llava_text_decoder_runner.h
@@ -17,8 +17,7 @@ namespace example {
 class ET_EXPERIMENTAL LlavaTextDecoderRunner
     : public executorch::extension::llm::TextDecoderRunner {
  public:
-  explicit LlavaTextDecoderRunner(
-      executorch::extension::ET_MODULE_NAMESPACE::Module* module)
+  explicit LlavaTextDecoderRunner(executorch::extension::Module* module)
       : TextDecoderRunner(module, true) {}
 
   inline executorch::runtime::Result<executorch::aten::Tensor> step(
diff --git a/examples/models/phi-3-mini/runner.cpp b/examples/models/phi-3-mini/runner.cpp
index 8b5457c62ec..15f76e9522c 100644
--- a/examples/models/phi-3-mini/runner.cpp
+++ b/examples/models/phi-3-mini/runner.cpp
@@ -16,7 +16,7 @@
 #include <pytorch/tokenizers/llama2c_tokenizer.h>
 
 using executorch::aten::ScalarType;
-using executorch::extension::ET_MODULE_NAMESPACE::Module;
+using executorch::extension::Module;
 using executorch::extension::llm::Sampler;
 using executorch::runtime::Error;
 using tokenizers::Llama2cTokenizer;
diff --git a/examples/models/phi-3-mini/runner.h b/examples/models/phi-3-mini/runner.h
index fcb6cd71efa..2f0042a57ea 100644
--- a/examples/models/phi-3-mini/runner.h
+++ b/examples/models/phi-3-mini/runner.h
@@ -42,7 +42,7 @@ class Runner {
   uint64_t prefill(std::vector<uint64_t>& tokens);
   uint64_t run_model_step(uint64_t token);
 
-  std::unique_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module> module_;
+  std::unique_ptr<executorch::extension::Module> module_;
   std::unique_ptr<tokenizers::Tokenizer> tokenizer_;
   std::unique_ptr<executorch::extension::llm::Sampler> sampler_;
 };
diff --git a/examples/qualcomm/oss_scripts/llama/runner/io_manager.cpp b/examples/qualcomm/oss_scripts/llama/runner/io_manager.cpp
index 8bdcc462923..c2bf7b04fbb 100644
--- a/examples/qualcomm/oss_scripts/llama/runner/io_manager.cpp
+++ b/examples/qualcomm/oss_scripts/llama/runner/io_manager.cpp
@@ -12,7 +12,7 @@
 
 using executorch::aten::Tensor;
 using executorch::aten::TensorImpl;
-using executorch::extension::ET_MODULE_NAMESPACE::Module;
+using executorch::extension::Module;
 using executorch::runtime::Error;
 using executorch::runtime::MemoryAllocator;
 using executorch::runtime::MethodMeta;
diff --git a/examples/qualcomm/oss_scripts/llama/runner/io_manager.h b/examples/qualcomm/oss_scripts/llama/runner/io_manager.h
index 5a305fa7406..0f10eef8ddc 100644
--- a/examples/qualcomm/oss_scripts/llama/runner/io_manager.h
+++ b/examples/qualcomm/oss_scripts/llama/runner/io_manager.h
@@ -20,8 +20,6 @@
 #include <executorch/extension/module/module.h>
 #include <executorch/runtime/executor/method_meta.h>
 
-using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
-
 namespace example {
 
 enum EvalMode {
@@ -31,8 +29,8 @@ enum EvalMode {
 };
 class IoMgrBase {
  public:
-  IoMgrBase(std::vector<std::shared_ptr<
-                executorch::extension::ET_MODULE_NAMESPACE::Module>>& modules);
+  IoMgrBase(
+      std::vector<std::shared_ptr<executorch::extension::Module>>& modules);
   virtual ~IoMgrBase();
   virtual void init_io() = 0;
   virtual void reset_io(
@@ -83,16 +81,13 @@ class IoMgrBase {
       std::string,
       std::vector<std::vector<executorch::aten::TensorImpl*>>>
       output_tensors_;
-  std::vector<
-      std::shared_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>
-      modules_;
+  std::vector<std::shared_ptr<executorch::extension::Module>> modules_;
 };
 
 class ShiftPointerIoMgr : public IoMgrBase {
  public:
   ShiftPointerIoMgr(
-      std::vector<std::shared_ptr<
-          executorch::extension::ET_MODULE_NAMESPACE::Module>>& modules,
+      std::vector<std::shared_ptr<executorch::extension::Module>>& modules,
       int32_t context_len,
       int32_t prefill_ar_len,
       int32_t prefill_cache_len,
@@ -204,8 +199,7 @@ class ShiftPointerIoMgr : public IoMgrBase {
 class SmartMaskIoMgr : public IoMgrBase {
  public:
   SmartMaskIoMgr(
-      std::vector<std::shared_ptr<
-          executorch::extension::ET_MODULE_NAMESPACE::Module>>& modules,
+      std::vector<std::shared_ptr<executorch::extension::Module>>& modules,
       int32_t context_len,
       int32_t prefill_ar_len,
       int32_t prefill_cache_len,
diff --git a/examples/qualcomm/oss_scripts/llama/runner/runner.cpp b/examples/qualcomm/oss_scripts/llama/runner/runner.cpp
index 7e7c48a517e..dafc911a172 100644
--- a/examples/qualcomm/oss_scripts/llama/runner/runner.cpp
+++ b/examples/qualcomm/oss_scripts/llama/runner/runner.cpp
@@ -23,7 +23,7 @@
 #include <sstream>
 
 using executorch::aten::Tensor;
-using executorch::extension::ET_MODULE_NAMESPACE::Module;
+using executorch::extension::Module;
 using executorch::extension::llm::Sampler;
 using executorch::extension::llm::time_in_ms;
 using executorch::runtime::Error;
diff --git a/examples/qualcomm/oss_scripts/llama/runner/runner.h b/examples/qualcomm/oss_scripts/llama/runner/runner.h
index 93c3440a503..e693bcd7077 100644
--- a/examples/qualcomm/oss_scripts/llama/runner/runner.h
+++ b/examples/qualcomm/oss_scripts/llama/runner/runner.h
@@ -101,9 +101,7 @@ class Runner {
   std::unordered_set<uint64_t> eos_id_;
   const int32_t n_bos_;
   const int32_t n_eos_;
-  std::vector<
-      std::shared_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>
-      modules_;
+  std::vector<std::shared_ptr<executorch::extension::Module>> modules_;
   std::string tokenizer_path_;
   std::string performance_output_path_;
   float logits_scale_;
diff --git a/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.cpp b/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.cpp
index 8a5ddcd7ded..9ee7551650a 100644
--- a/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.cpp
+++ b/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.cpp
@@ -14,7 +14,7 @@
 
 using executorch::aten::Tensor;
 using executorch::aten::TensorImpl;
-using executorch::extension::ET_MODULE_NAMESPACE::Module;
+using executorch::extension::Module;
 using executorch::runtime::Error;
 using executorch::runtime::MethodMeta;
 using executorch::runtime::Result;
diff --git a/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.h b/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.h
index 012f40404b4..445be2ed21a 100644
--- a/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.h
+++ b/examples/qualcomm/qaihub_scripts/llama/runner/io_memory.h
@@ -32,9 +32,7 @@ class Memory {
  public:
   Memory(
       const std::vector<std::string>& pos_embs_path,
-      std::vector<
-          std::shared_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>&
-          modules);
+      std::vector<std::shared_ptr<executorch::extension::Module>>& modules);
   virtual ~Memory();
   virtual void prepare_io(
       const std::vector<
@@ -53,9 +51,7 @@ class Memory {
   std::vector<std::vector<executorch::aten::TensorImpl*>> input_tensors_;
   std::vector<std::vector<executorch::aten::TensorImpl*>> output_tensors_;
   std::vector<std::string> pos_embs_path_;
-  std::vector<
-      std::shared_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>
-      modules_;
+  std::vector<std::shared_ptr<executorch::extension::Module>> modules_;
   std::vector<std::string> method_names_;
 };
 
@@ -63,9 +59,7 @@ class BertMemory : public Memory {
  public:
   BertMemory(
       const std::vector<std::string>& pos_embs_path,
-      std::vector<
-          std::shared_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>&
-          modules,
+      std::vector<std::shared_ptr<executorch::extension::Module>>& modules,
       std::vector<int> shard_layers);
   void prepare_io(const std::vector<executorch::runtime::Result<
                       executorch::runtime::MethodMeta>>& methods_meta) override;
@@ -127,9 +121,7 @@ class KVCachedMemory : public Memory {
  public:
   KVCachedMemory(
       const std::vector<std::string>& pos_embs_path,
-      std::vector<
-          std::shared_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>&
-          modules,
+      std::vector<std::shared_ptr<executorch::extension::Module>>& modules,
       std::vector<int> shard_layers);
   void prepare_io(const std::vector<executorch::runtime::Result<
                       executorch::runtime::MethodMeta>>& methods_meta) override;
diff --git a/examples/qualcomm/qaihub_scripts/llama/runner/runner.cpp b/examples/qualcomm/qaihub_scripts/llama/runner/runner.cpp
index cb98fcf04f3..06ea324ef6f 100644
--- a/examples/qualcomm/qaihub_scripts/llama/runner/runner.cpp
+++ b/examples/qualcomm/qaihub_scripts/llama/runner/runner.cpp
@@ -30,7 +30,7 @@
 #endif
 
 using executorch::aten::Tensor;
-using executorch::extension::ET_MODULE_NAMESPACE::Module;
+using executorch::extension::Module;
 using executorch::extension::llm::Sampler;
 using executorch::extension::llm::time_in_ms;
 using executorch::runtime::Error;
diff --git a/examples/qualcomm/qaihub_scripts/llama/runner/runner.h b/examples/qualcomm/qaihub_scripts/llama/runner/runner.h
index effd2ca8c50..9672d6a3586 100644
--- a/examples/qualcomm/qaihub_scripts/llama/runner/runner.h
+++ b/examples/qualcomm/qaihub_scripts/llama/runner/runner.h
@@ -97,9 +97,7 @@ class Runner {
   const int32_t vocab_size_;
   const int32_t max_seq_len_;
   int32_t eval_mode_;
-  std::vector<
-      std::shared_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>
-      modules_;
+  std::vector<std::shared_ptr<executorch::extension::Module>> modules_;
   std::vector<std::string> method_names_;
   std::string tokenizer_path_;
   float temperature_;
diff --git a/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.cpp b/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.cpp
index 90091c3b870..585d58b21ee 100644
--- a/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.cpp
+++ b/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.cpp
@@ -23,8 +23,8 @@
 #include <executorch/runtime/platform/log.h>
 
 using executorch::extension::from_blob;
+using executorch::extension::Module;
 using executorch::extension::TensorPtr;
-using executorch::extension::ET_MODULE_NAMESPACE::Module;
 using executorch::extension::llm::time_in_ms;
 using executorch::runtime::Error;
 using executorch::runtime::MethodMeta;
diff --git a/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.h b/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.h
index df37e4829bd..e49201bca25 100644
--- a/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.h
+++ b/examples/qualcomm/qaihub_scripts/stable_diffusion/runner/runner.h
@@ -111,9 +111,7 @@ class Runner {
 
  private:
   Stats stats_;
-  std::vector<
-      std::unique_ptr<executorch::extension::ET_MODULE_NAMESPACE::Module>>
-      modules_;
+  std::vector<std::unique_ptr<executorch::extension::Module>> modules_;
   std::vector<std::string> method_names_;
   std::vector<std::vector<uint16_t>> time_emb_list_;
   std::unordered_map<std::string, int32_t> vocab_to_token_map_;
diff --git a/extension/android/jni/jni_layer.cpp b/extension/android/jni/jni_layer.cpp
index ebbd3780c39..f3c62e1d70f 100644
--- a/extension/android/jni/jni_layer.cpp
+++ b/extension/android/jni/jni_layer.cpp
@@ -36,7 +36,6 @@
 
 using namespace executorch::extension;
 using namespace torch::executor;
-using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 namespace executorch::extension {
 class TensorHybrid : public facebook::jni::HybridClass<TensorHybrid> {
diff --git a/extension/apple/ExecuTorch/Exported/ExecuTorchModule.mm b/extension/apple/ExecuTorch/Exported/ExecuTorchModule.mm
index 414fba7dc13..98f0f555991 100644
--- a/extension/apple/ExecuTorch/Exported/ExecuTorchModule.mm
+++ b/extension/apple/ExecuTorch/Exported/ExecuTorchModule.mm
@@ -15,7 +15,6 @@
 
 using namespace executorch::extension;
 using namespace executorch::runtime;
-using namespace executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 static inline EValue toEValue(ExecuTorchValue *value) {
   if (value.isTensor) {
diff --git a/extension/llm/runner/image_prefiller.h b/extension/llm/runner/image_prefiller.h
index aaca29fb5f0..826cea24db9 100644
--- a/extension/llm/runner/image_prefiller.h
+++ b/extension/llm/runner/image_prefiller.h
@@ -14,8 +14,6 @@
 #include <executorch/extension/module/module.h>
 #include <executorch/runtime/platform/compiler.h>
 
-using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
-
 namespace executorch {
 namespace extension {
 namespace llm {
@@ -23,8 +21,7 @@ namespace llm {
 // Assuming kv cache and parallel prefill are enabled.
 class ET_EXPERIMENTAL ImagePrefiller {
  public:
-  explicit ImagePrefiller(
-      ::executorch::extension::ET_MODULE_NAMESPACE::Module* module)
+  explicit ImagePrefiller(::executorch::extension::Module* module)
       : module_(module) {}
 
   /**
diff --git a/extension/llm/runner/multimodal_runner.h b/extension/llm/runner/multimodal_runner.h
index fc66e56553d..c17e039c11b 100644
--- a/extension/llm/runner/multimodal_runner.h
+++ b/extension/llm/runner/multimodal_runner.h
@@ -29,8 +29,6 @@
 #include <executorch/extension/module/module.h>
 #include <pytorch/tokenizers/tokenizer.h>
 
-using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
-
 namespace executorch {
 namespace extension {
 namespace llm {
diff --git a/extension/llm/runner/text_decoder_runner.h b/extension/llm/runner/text_decoder_runner.h
index 7f40a7efae0..6c1256c6b90 100644
--- a/extension/llm/runner/text_decoder_runner.h
+++ b/extension/llm/runner/text_decoder_runner.h
@@ -15,8 +15,6 @@
 #include <executorch/extension/tensor/tensor.h>
 #include <executorch/runtime/platform/compiler.h>
 
-using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
-
 namespace executorch {
 namespace extension {
 namespace llm {
diff --git a/extension/module/module.cpp b/extension/module/module.cpp
index 2bab588569f..ec01323edc7 100644
--- a/extension/module/module.cpp
+++ b/extension/module/module.cpp
@@ -36,7 +36,6 @@
 
 namespace executorch {
 namespace extension {
-namespace ET_MODULE_NAMESPACE {
 
 using ET_RUNTIME_NAMESPACE::MethodMeta;
 using ET_RUNTIME_NAMESPACE::Program;
@@ -313,6 +312,5 @@ ET_NODISCARD inline runtime::Result<Method*> Module::get_method(
   return methods_[method_name].method.get();
 }
 
-} // namespace ET_MODULE_NAMESPACE
 } // namespace extension
 } // namespace executorch
diff --git a/extension/module/module.h b/extension/module/module.h
index 46e35dd7ca5..4f8d4ea784d 100644
--- a/extension/module/module.h
+++ b/extension/module/module.h
@@ -524,3 +524,10 @@ namespace executor {
 using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 } // namespace executor
 } // namespace torch
+
+namespace executorch {
+namespace extension {
+// backward compatible namespace alias
+using namespace ::executorch::extension::ET_MODULE_NAMESPACE;
+} // namespace extension
+} // namespace executorch
diff --git a/extension/module/test/module_test.cpp b/extension/module/test/module_test.cpp
index 472af2cecdb..a82e257a703 100644
--- a/extension/module/test/module_test.cpp
+++ b/extension/module/test/module_test.cpp
@@ -18,7 +18,6 @@
 
 using namespace ::executorch::extension;
 using namespace ::executorch::runtime;
-using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 
 class ModuleTest : public ::testing::Test {
  protected:
diff --git a/extension/training/module/training_module.cpp b/extension/training/module/training_module.cpp
index 6b3015a9d79..d119738715e 100644
--- a/extension/training/module/training_module.cpp
+++ b/extension/training/module/training_module.cpp
@@ -25,8 +25,7 @@ TrainingModule::execute_forward_backward(
   // Find where the user outputs end.
   const std::string gradients_method_name =
       gradients_method_prefix + method_name;
-  auto res = executorch::extension::ET_MODULE_NAMESPACE::Module::execute(
-      gradients_method_name);
+  auto res = executorch::extension::Module::execute(gradients_method_name);
   if (!res.ok()) {
     return res.error();
   }
@@ -35,8 +34,8 @@ TrainingModule::execute_forward_backward(
   const std::string parameters_method_name =
       parameters_method_prefix + method_name;
   // get params start.
-  auto param_res = executorch::extension::ET_MODULE_NAMESPACE::Module::execute(
-      parameters_method_name);
+  auto param_res =
+      executorch::extension::Module::execute(parameters_method_name);
   if (!param_res.ok()) {
     return param_res.error();
   }
@@ -68,8 +67,7 @@ TrainingModule::execute_forward_backward(
 
     // Get names if we havent seen this method before.
     const std::string fqn_method_name = fqn_method_prefix + method_name;
-    auto fqn_res = executorch::extension::ET_MODULE_NAMESPACE::Module::execute(
-        fqn_method_name);
+    auto fqn_res = executorch::extension::Module::execute(fqn_method_name);
     if (!fqn_res.ok()) {
       return fqn_res.error();
     }
@@ -102,8 +100,7 @@ TrainingModule::named_parameters(const std::string& method_name) {
     method_named_parameters_.insert({method_name, {}});
 
     // get names.
-    auto fqn_res = executorch::extension::ET_MODULE_NAMESPACE::Module::execute(
-        fqn_method_name);
+    auto fqn_res = executorch::extension::Module::execute(fqn_method_name);
     if (!fqn_res.ok()) {
       return fqn_res.error();
     }
@@ -111,8 +108,7 @@ TrainingModule::named_parameters(const std::string& method_name) {
 
     // get params start.
     auto param_res =
-        executorch::extension::ET_MODULE_NAMESPACE::Module::execute(
-            parameters_method_name);
+        executorch::extension::Module::execute(parameters_method_name);
     if (!param_res.ok()) {
       return param_res.error();
     }
@@ -120,8 +116,7 @@ TrainingModule::named_parameters(const std::string& method_name) {
     uint64_t param_start = param_res.get()[0].toInt();
 
     // Load the method if it is not already loaded.
-    auto e = executorch::extension::ET_MODULE_NAMESPACE::Module::load_method(
-        method_name);
+    auto e = executorch::extension::Module::load_method(method_name);
     if (e != runtime::Error::Ok) {
       return e;
     }
diff --git a/extension/training/module/training_module.h b/extension/training/module/training_module.h
index 1d48355b7ba..7bf81623c04 100644
--- a/extension/training/module/training_module.h
+++ b/extension/training/module/training_module.h
@@ -27,7 +27,7 @@ namespace training {
  * methods within them.
  */
 class ET_EXPERIMENTAL TrainingModule final
-    : public executorch::extension::ET_MODULE_NAMESPACE::Module {
+    : public executorch::extension::Module {
  public:
   explicit TrainingModule(
       std::unique_ptr<runtime::DataLoader> data_loader,
@@ -35,7 +35,7 @@ class ET_EXPERIMENTAL TrainingModule final
       std::unique_ptr<runtime::MemoryAllocator> temp_allocator = nullptr,
       std::unique_ptr<runtime::EventTracer> event_tracer = nullptr,
       std::unique_ptr<runtime::DataLoader> data_map_data_loader = nullptr)
-      : executorch::extension::ET_MODULE_NAMESPACE::Module(
+      : executorch::extension::Module(
             std::move(data_loader),
             std::move(memory_allocator),
             std::move(temp_allocator),

From 064462747307f191fc90c32fe002d3ea15b344c0 Mon Sep 17 00:00:00 2001
From: zhenyan-zhang-meta <zhenyanzhang@meta.com>
Date: Wed, 30 Apr 2025 11:18:35 -0700
Subject: [PATCH 3/3] Update on "[ExecuTorch] Separate `extension.Module`
 Namespaces from Aten and non-Aten"

# Context

Separate `extension.Module` Namespaces to be `executorch::extension::module` and `executorch::extension::module::aten`, otherwise if a package relies on both aten and non-aten of the same implementation and the namespace is the same, there will be duplicate symbol issue like:

```
ld.lld: error: duplicate symbol: vtable for executorch::extension::Module
>>> defined at {redacted}/executorch/extension/module/__module__/__stripped__/module.cpp.pic.stripped.o:(vtable for executorch::extension::Module)
>>> defined at {redacted}/executorch/extension/module/__module_aten__/__stripped__/module.cpp.pic.stripped.o:
```

# Proposal

Doing something similar to what we already did for `bundled_program` in https://github.com/pytorch/executorch/pull/10307

Since `extension.Module` is a public API, we introduce a namespace alias, so that existing use cases won't get affected. Since namespace alias doesn't create additional symbols, there won't be duplicate symbol issue.

Differential Revision: [D73903870](https://our.internmc.facebook.com/intern/diff/D73903870/)

[ghstack-poisoned]
---
 extension/module/module.h | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/extension/module/module.h b/extension/module/module.h
index 4f8d4ea784d..afd2fa47973 100644
--- a/extension/module/module.h
+++ b/extension/module/module.h
@@ -24,13 +24,16 @@
 
 namespace executorch {
 namespace extension {
-namespace ET_MODULE_NAMESPACE {
 
 using ET_RUNTIME_NAMESPACE::Method;
 using ET_RUNTIME_NAMESPACE::MethodMeta;
 using ET_RUNTIME_NAMESPACE::NamedDataMap;
 using ET_RUNTIME_NAMESPACE::Program;
 
+class ExecuTorchJni;
+
+namespace ET_MODULE_NAMESPACE {
+
 /**
  * A facade class for loading programs and executing methods within them.
  */
@@ -510,7 +513,7 @@ class Module {
       const std::string& method_name);
   std::unordered_map<std::string, MethodHolder> methods_;
 
-  friend class ExecuTorchJni;
+  friend class executorch::extension::ExecuTorchJni;
 };
 
 } // namespace ET_MODULE_NAMESPACE
@@ -528,6 +531,6 @@ using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 namespace executorch {
 namespace extension {
 // backward compatible namespace alias
-using namespace ::executorch::extension::ET_MODULE_NAMESPACE;
+using ::executorch::extension::ET_MODULE_NAMESPACE::Module;
 } // namespace extension
 } // namespace executorch