Merge branch 'mm_runner_ext' of https://github.com/pytorch/executorch into mm_runner_ext

Author: larryliu0820

CMakeLists.txt

@@ -650,6 +650,13 @@ if(EXECUTORCH_BUILD_EXTENSION_LLM)
   list(APPEND _executorch_extensions tokenizers)
 endif()
 
+<<<<<<< HEAD
+=======
+if(EXECUTORCH_BUILD_EXTENSION_LLM_APPLE)
+  add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/extension/llm/apple)
+endif()
+
+>>>>>>> 13d8d946c0edc7e0f8df38194406c874f4e2fbbb
 if(EXECUTORCH_BUILD_EXTENSION_RUNNER_UTIL)
   add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/extension/runner_util)
   install(
@@ -900,10 +907,13 @@ if(EXECUTORCH_BUILD_EXTENSION_LLM_RUNNER)
   list(APPEND _executorch_extensions extension_llm_runner)
 endif()
 
+<<<<<<< HEAD
 if(EXECUTORCH_BUILD_EXTENSION_LLM_APPLE)
   add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/extension/llm/apple)
 endif()
 
+=======
+>>>>>>> 13d8d946c0edc7e0f8df38194406c874f4e2fbbb
 if(EXECUTORCH_BUILD_KERNELS_LLM)
   # TODO: move all custom kernels to ${CMAKE_CURRENT_SOURCE_DIR}/kernels/custom
   add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/extension/llm/custom_ops)

extension/llm/runner/_llm_runner.pyi

@@ -34,6 +34,7 @@ class GenerationConfig:
     num_eos: int
     """Number of EOS tokens to add to the prompt."""
 
+<<<<<<< HEAD
     def __init__(
         self,
         *,
@@ -46,6 +47,10 @@ class GenerationConfig:
         num_eos: int = 0,
     ) -> None:
         """Initialize GenerationConfig with optional keyword arguments for all fields."""
+=======
+    def __init__(self) -> None:
+        """Initialize GenerationConfig with default values."""
+>>>>>>> 13d8d946c0edc7e0f8df38194406c874f4e2fbbb
         ...
 
     def resolve_max_new_tokens(
@@ -368,6 +373,7 @@ class MultimodalRunner:
         Raises:
             RuntimeError: If generation fails
         """
+<<<<<<< HEAD
     ...
 
     def generate(
@@ -406,6 +412,8 @@ class MultimodalRunner:
         Raises:
             RuntimeError: If prefill fails
         """
+=======
+>>>>>>> 13d8d946c0edc7e0f8df38194406c874f4e2fbbb
         ...
 
     def generate_text(
@@ -424,6 +432,7 @@ class MultimodalRunner:
         Raises:
             RuntimeError: If generation fails
         """
+<<<<<<< HEAD
     ...
 
     def generate_text(self, inputs: dict, config: GenerationConfig) -> str:
@@ -432,6 +441,8 @@ class MultimodalRunner:
 
         See generate(inputs: dict, ...) for expected keys and constraints.
         """
+=======
+>>>>>>> 13d8d946c0edc7e0f8df38194406c874f4e2fbbb
         ...
 
     def stop(self) -> None:

extension/llm/runner/pybindings.cpp

@@ -165,7 +165,6 @@ PYBIND11_MODULE(_llm_runner, m) {
   py::class_<GenerationConfig>(m, "GenerationConfig")
       // Constructor with keyword arguments for all fields (all optional via
       // defaults)
-      .def(
           py::init([](bool echo,
                       int32_t max_new_tokens,
                       bool warming,
@@ -190,9 +189,6 @@ PYBIND11_MODULE(_llm_runner, m) {
           py::arg("temperature") = 0.8f,
           py::arg("num_bos") = 0,
           py::arg("num_eos") = 0)
-      .def_readwrite("echo", &GenerationConfig::echo)
-      .def_readwrite("max_new_tokens", &GenerationConfig::max_new_tokens)
-      .def_readwrite("warming", &GenerationConfig::warming)
       .def_readwrite("seq_len", &GenerationConfig::seq_len)
       .def_readwrite("temperature", &GenerationConfig::temperature)
       .def_readwrite("num_bos", &GenerationConfig::num_bos)