chore: apply review suggestions

Author: IgorSwat

packages/react-native-executorch/common/rnexecutorch/models/BaseModel.h

@@ -55,4 +55,4 @@ class BaseModel {
 
 REGISTER_CONSTRUCTOR(models::BaseModel, std::string,
                      std::shared_ptr<react::CallInvoker>);
-} // namespace rnexecutorch
+} // namespace rnexecutorch

packages/react-native-executorch/common/rnexecutorch/models/llm/LLM.cpp

@@ -6,6 +6,7 @@
 #include <rnexecutorch/threads/GlobalThreadPool.h>
 
 namespace rnexecutorch::models::llm {
+namespace fs = std::filesystem;
 using namespace facebook;
 using executorch::extension::TensorPtr;
 using executorch::extension::module::Module;
@@ -22,13 +23,13 @@ LLM::LLM(const std::string &modelSource, const std::string &tokenizerSource,
                              std::to_string(static_cast<int>(loadResult)));
   }
 
-  memorySizeLowerBound =
-      std::filesystem::file_size(std::filesystem::path(modelSource)) +
-      std::filesystem::file_size(std::filesystem::path(tokenizerSource));
+  memorySizeLowerBound = fs::file_size(fs::path(modelSource)) +
+                         fs::file_size(fs::path(tokenizerSource));
 
   // Determine the input mode
-  auto tokensTensorShape = getInputShape("forward", 0);
-  auto positionsTensorShape = getInputShape("forward", 1);
+  auto inputShapes = getAllInputShapes("forward");
+  auto &tokensTensorShape = inputShapes[0];
+  auto &positionsTensorShape = inputShapes[1];
   if (tokensTensorShape.size() != 2 || positionsTensorShape.size() != 1) {
     throw std::runtime_error("Unsupported LLM input format");
   }

packages/react-native-executorch/common/rnexecutorch/models/llm/LLM.h

@@ -39,4 +39,4 @@ class LLM : public BaseModel {
 
 REGISTER_CONSTRUCTOR(models::llm::LLM, std::string, std::string,
                      std::shared_ptr<react::CallInvoker>);
-} // namespace rnexecutorch
+} // namespace rnexecutorch

packages/react-native-executorch/common/runner/runner.cpp

@@ -47,8 +47,7 @@ static constexpr auto kUseKVCache = "use_kv_cache";
 static constexpr auto kUseSDPAWithKVCache = "use_sdpa_with_kv_cache";
 } // namespace
 
-Runner::Runner(Module *module, const std::string &model_path,
-               const std::string &tokenizer_path,
+Runner::Runner(Module *module, const std::string &tokenizer_path,
                const bool extended_input_mode, const float temperature,
                std::optional<const std::string> data_path)
     : module_(module), temperature_(temperature),
@@ -58,10 +57,7 @@ Runner::Runner(Module *module, const std::string &model_path,
                                            {kMaxContextLen, 128},
                                            {kUseKVCache, true},
                                            {kUseSDPAWithKVCache, false},
-                                       }) {
-  ET_LOG(Info, "Creating LLM runner: model_path=%s, tokenizer_path=%s",
-         model_path.c_str(), tokenizer_path.c_str());
-}
+                                       }) {}
 
 bool Runner::is_loaded() const {
   return module_->is_loaded() && tokenizer_ && text_decoder_runner_ &&

packages/react-native-executorch/common/runner/runner.h

@@ -29,13 +29,10 @@ namespace example {
 
 class Runner : public executorch::extension::llm::IRunner {
 public:
-  explicit Runner(
-      ::executorch::extension::Module *module,
-      const std::string &model_path, // TODO: consider removing this arg since
-                                     // it is only used for debug purposes
-      const std::string &tokenizer_path, const bool extended_input_mode = false,
-      const float temperature = 0.8f,
-      std::optional<const std::string> data_path = std::nullopt);
+  explicit Runner(::executorch::extension::Module *module,
+                  const std::string &tokenizer_path,
+                  bool extended_input_mode = false, float temperature = 0.8f,
+                  std::optional<const std::string> data_path = std::nullopt);
 
   bool is_loaded() const;
   ::executorch::runtime::Error load();
@@ -46,7 +43,7 @@ class Runner : public executorch::extension::llm::IRunner {
                stats_callback = {},
            bool echo = true, bool warming = false);
   ::executorch::runtime::Error warmup(const std::string &prompt);
-  void set_extended_input_mode(bool extend_position_input);
+  void set_extended_input_mode(bool extend_position_input) noexcept;
   void set_count_interval(size_t count_interval);
   void set_time_interval(size_t time_interval);
   void stop();
@@ -72,4 +69,4 @@ class Runner : public executorch::extension::llm::IRunner {
       text_token_generator_;
 };
 
-} // namespace example
+} // namespace example