Remove unused variables and ET_EXPERIMENTAL from runner

Author: a-szymanska

packages/react-native-executorch/android/CMakeLists.txt

@@ -14,4 +14,7 @@ set(COMMON_CPP_DIR "${CMAKE_SOURCE_DIR}/../common")
 set(LIBS_DIR "${CMAKE_SOURCE_DIR}/../third-party/android/libs")
 set(INCLUDE_DIR "${CMAKE_SOURCE_DIR}/../third-party/include")
 
+# Treat third-party headers as system headers to suppress deprecation warnings
+include_directories(SYSTEM "${INCLUDE_DIR}")
+
 add_subdirectory("${ANDROID_CPP_DIR}")

packages/react-native-executorch/common/rnexecutorch/TokenizerModule.cpp

@@ -8,9 +8,9 @@ using namespace facebook;
 
 TokenizerModule::TokenizerModule(
     std::string source, std::shared_ptr<react::CallInvoker> callInvoker)
-    : memorySizeLowerBound(std::filesystem::file_size(source)),
-      tokenizer(tokenizers::Tokenizer::FromBlobJSON(
-          file_utils::loadBytesFromFile(source))) {}
+    : tokenizer(tokenizers::Tokenizer::FromBlobJSON(
+          file_utils::loadBytesFromFile(source))),
+      memorySizeLowerBound(std::filesystem::file_size(source)) {}
 
 void TokenizerModule::ensureTokenizerLoaded(
     const std::string &methodName) const {

packages/react-native-executorch/common/rnexecutorch/models/text_to_image/Scheduler.h

@@ -30,7 +30,6 @@ class Scheduler final {
   std::vector<float> tempFirstSample;
   std::vector<std::vector<float>> ets;
   float finalAlphaCumprod{1.0f};
-  float initNoiseSigma{1.0f};
 
   size_t numInferenceSteps{0};
 

packages/react-native-executorch/common/rnexecutorch/models/text_to_image/TextToImage.cpp

@@ -62,7 +62,6 @@ TextToImage::generate(std::string input, int32_t imageSize,
   auto embeddingsTensor =
       make_tensor_ptr(embeddingsShape, embeddings.data(), ScalarType::Float);
 
-  constexpr int32_t latentDownsample = 8;
   int32_t latentsSize = numChannels * latentImageSize * latentImageSize;
   std::vector<float> latents(latentsSize);
   std::mt19937 gen(seed);

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

@@ -21,7 +21,7 @@ namespace executorch {
 namespace extension {
 namespace llm {
 
-class ET_EXPERIMENTAL IRunner {
+class IRunner {
 public:
   virtual ~IRunner() = default;
 

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

@@ -39,7 +39,6 @@ std::string loadBytesFromFile(const std::string &path) {
 
 namespace {
 static constexpr auto kEnableDynamicShape = "enable_dynamic_shape";
-static constexpr auto kBosId = "get_bos_id";
 static constexpr auto kEosIds = "get_eos_ids";
 static constexpr auto kMaxSeqLen = "get_max_seq_len";
 static constexpr auto kMaxContextLen = "get_max_context_len";
@@ -175,10 +174,7 @@ Error Runner::generate(const std::string &prompt,
   stats_.inference_start_ms = llm::time_in_ms();
   shouldStop_ = false;
 
-  // Set the sequence length to the max seq length if not provided
-  int32_t seq_len = (seq_len > 0 && seq_len <= metadata_.at(kMaxSeqLen))
-                        ? seq_len
-                        : metadata_.at(kMaxSeqLen);
+  int32_t seq_len = metadata_.at(kMaxSeqLen);
 
   std::vector<int32_t> prompt_tokens = tokenizer_->Encode(prompt);
   std::vector<uint64_t> prompt_tokens_uint64(prompt_tokens.begin(),

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

@@ -27,7 +27,7 @@
 
 namespace example {
 
-class ET_EXPERIMENTAL Runner : public executorch::extension::llm::IRunner {
+class Runner : public executorch::extension::llm::IRunner {
 public:
   explicit Runner(const std::string &model_path,
                   const std::string &tokenizer_path,

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

@@ -26,12 +26,12 @@ namespace extension {
 namespace llm {
 // A simple llama2 sampler.
 
-template <typename T> struct ET_EXPERIMENTAL ProbIndex {
+template <typename T> struct ProbIndex {
   T prob;
   int32_t index;
 }; // struct used when sorting probabilities during top-p sampling
 
-class ET_EXPERIMENTAL Sampler {
+class Sampler {
 public:
   Sampler(int32_t vocab_size, float temperature, float topp,
           unsigned long long rng_seed);

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

@@ -18,7 +18,7 @@ namespace executorch {
 namespace extension {
 namespace llm {
 
-struct ET_EXPERIMENTAL Stats {
+struct Stats {
   // Scaling factor for timestamps - in this case, we use ms.
   const long SCALING_FACTOR_UNITS_PER_SECOND = 1000;
   // Time stamps for the different stages of the execution

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

@@ -20,7 +20,7 @@ namespace executorch {
 namespace extension {
 namespace llm {
 
-class ET_EXPERIMENTAL TextDecoderRunner {
+class TextDecoderRunner {
 public:
   TextDecoderRunner(Module *module, bool use_kv_cache, int32_t vocab_size,
                     float temperature);