[C++] Update reflection block for C++ objects

cpp/serve/config.cc

@@ -20,6 +20,11 @@ namespace mlc {
 namespace llm {
 namespace serve {
 
+TVM_FFI_STATIC_INIT_BLOCK() {
+  GenerationConfigNode::RegisterReflection();
+  EngineConfigNode::RegisterReflection();
+}
+
 uint64_t TotalDetectGlobalMemory(DLDevice device) {
   // Get single-card GPU size.
   tvm::ffi::Any rv;

cpp/serve/config.h

@@ -7,6 +7,7 @@
 
 #include <picojson.h>
 #include <tvm/ffi/container/array.h>
+#include <tvm/ffi/reflection/registry.h>
 #include <tvm/ffi/string.h>
 #include <tvm/runtime/device_api.h>
 #include <tvm/runtime/int_tuple.h>
@@ -135,6 +136,11 @@ class GenerationConfigNode : public Object {
 
   picojson::object AsJSON() const;
 
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<GenerationConfigNode>();
+  }
+
   static constexpr const bool _type_has_method_sequal_reduce = false;
   static constexpr const bool _type_has_method_shash_reduce = false;
   TVM_FFI_DECLARE_OBJECT_INFO("mlc.serve.GenerationConfig", GenerationConfigNode, Object);
@@ -302,6 +308,11 @@ class EngineConfigNode : public Object {
 
   String AsJSONString() const;
 
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<EngineConfigNode>();
+  }
+
   static constexpr const bool _type_has_method_sequal_reduce = false;
   static constexpr const bool _type_has_method_shash_reduce = false;
   static constexpr const bool _type_mutable = true;

cpp/serve/data.cc

@@ -13,6 +13,14 @@ namespace mlc {
 namespace llm {
 namespace serve {
 
+TVM_FFI_STATIC_INIT_BLOCK() {
+  DataNode::RegisterReflection();
+  TextDataNode::RegisterReflection();
+  TokenDataNode::RegisterReflection();
+  ImageDataNode::RegisterReflection();
+  RequestStreamOutputObj::RegisterReflection();
+}
+
 /****************** Data ******************/
 
 std::pair<Array<Data>, Array<Data>> SplitData(const Array<Data>& original_data, int total_length,

cpp/serve/data.h

@@ -8,6 +8,7 @@
 #include <tvm/ffi/container/array.h>
 #include <tvm/ffi/container/shape.h>
 #include <tvm/ffi/optional.h>
+#include <tvm/ffi/reflection/registry.h>
 #include <tvm/ffi/string.h>
 #include <tvm/node/cast.h>
 #include <tvm/runtime/int_tuple.h>
@@ -50,8 +51,14 @@ class DataNode : public Object {
    */
   virtual ObjectRef GetEmbedding(Model model, ObjectRef* dst = nullptr, int offset = 0) const = 0;
 
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<DataNode>();
+  }
+
   static constexpr const bool _type_has_method_sequal_reduce = false;
   static constexpr const bool _type_has_method_shash_reduce = false;
+  static constexpr const uint32_t _type_child_slots = 3;
   TVM_FFI_DECLARE_OBJECT_INFO("mlc.serve.Data", DataNode, Object);
 };
 
@@ -75,7 +82,12 @@ class TextDataNode : public DataNode {
   int GetLength() const final;
   ObjectRef GetEmbedding(Model model, ObjectRef* dst = nullptr, int offset = 0) const final;
 
-  TVM_FFI_DECLARE_OBJECT_INFO("mlc.serve.TextData", TextDataNode, DataNode);
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<TextDataNode>();
+  }
+
+  TVM_FFI_DECLARE_OBJECT_INFO_FINAL("mlc.serve.TextData", TextDataNode, DataNode);
 };
 
 class TextData : public Data {
@@ -96,7 +108,12 @@ class TokenDataNode : public DataNode {
   int GetLength() const final;
   ObjectRef GetEmbedding(Model model, ObjectRef* dst = nullptr, int offset = 0) const final;
 
-  TVM_FFI_DECLARE_OBJECT_INFO("mlc.serve.TokenData", TokenDataNode, DataNode);
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<TokenDataNode>();
+  }
+
+  TVM_FFI_DECLARE_OBJECT_INFO_FINAL("mlc.serve.TokenData", TokenDataNode, DataNode);
 };
 
 class TokenData : public Data {
@@ -120,7 +137,12 @@ class ImageDataNode : public DataNode {
   int GetLength() const final;
   ObjectRef GetEmbedding(Model model, ObjectRef* dst = nullptr, int offset = 0) const final;
 
-  TVM_FFI_DECLARE_OBJECT_INFO("mlc.serve.ImageData", ImageDataNode, DataNode);
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<ImageDataNode>();
+  }
+
+  TVM_FFI_DECLARE_OBJECT_INFO_FINAL("mlc.serve.ImageData", ImageDataNode, DataNode);
 };
 
 class ImageData : public Data {
@@ -197,6 +219,11 @@ class RequestStreamOutputObj : public Object {
 
   std::atomic<bool> unpacked = false;
 
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<RequestStreamOutputObj>();
+  }
+
   static constexpr const bool _type_has_method_sequal_reduce = false;
   static constexpr const bool _type_has_method_shash_reduce = false;
   static constexpr const bool _type_mutable = true;

cpp/serve/draft_token_workspace_manager.cc

@@ -11,6 +11,8 @@ namespace mlc {
 namespace llm {
 namespace serve {
 
+TVM_FFI_STATIC_INIT_BLOCK() { DraftTokenWorkspaceManagerObj::RegisterReflection(); }
+
 DraftTokenWorkspaceManagerObj::DraftTokenWorkspaceManagerObj(int max_num_tokens, int vocab_size,
                                                              int hidden_size,
                                                              DLDataType hidden_states_dtype,

cpp/serve/draft_token_workspace_manager.h

@@ -5,6 +5,7 @@
 
 #ifndef MLC_LLM_SERVE_DRAFT_TOKEN_WORKSPACE_MANAGER_H_
 #define MLC_LLM_SERVE_DRAFT_TOKEN_WORKSPACE_MANAGER_H_
+#include <tvm/ffi/reflection/registry.h>
 #include <tvm/runtime/device_api.h>
 
 #include <numeric>
@@ -73,6 +74,11 @@ class DraftTokenWorkspaceManagerObj : public Object {
    */
   void FreeSlots(const std::vector<int>& slots);
 
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<DraftTokenWorkspaceManagerObj>();
+  }
+
   static constexpr const bool _type_has_method_sequal_reduce = false;
   static constexpr const bool _type_has_method_shash_reduce = false;
   static constexpr const bool _type_mutable = true;

cpp/serve/engine_actions/action.cc

@@ -7,6 +7,10 @@
 
 namespace mlc {
 namespace llm {
-namespace serve {}  // namespace serve
+namespace serve {
+
+TVM_FFI_STATIC_INIT_BLOCK() { EngineActionObj::RegisterReflection(); }
+
+}  // namespace serve
 }  // namespace llm
 }  // namespace mlc

cpp/serve/engine_actions/action.h

@@ -38,6 +38,11 @@ class EngineActionObj : public Object {
    */
   virtual Array<Request> Step(EngineState estate) = 0;
 
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<EngineActionObj>();
+  }
+
   static constexpr const bool _type_has_method_sequal_reduce = false;
   static constexpr const bool _type_has_method_shash_reduce = false;
   static constexpr const bool _type_mutable = true;

cpp/serve/engine_state.cc

@@ -8,6 +8,8 @@ namespace mlc {
 namespace llm {
 namespace serve {
 
+TVM_FFI_STATIC_INIT_BLOCK() { EngineStateObj::RegisterReflection(); }
+
 EngineState::EngineState() { data_ = tvm::ffi::make_object<EngineStateObj>(); }
 
 void EngineStateObj::Reset() {

cpp/serve/engine_state.h

@@ -98,6 +98,11 @@ class EngineStateObj : public Object {
   /*! \brief Return the running request state entries*/
   const std::vector<RequestStateEntry>& GetRunningRequestStateEntries();
 
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<EngineStateObj>();
+  }
+
   static constexpr const bool _type_has_method_sequal_reduce = false;
   static constexpr const bool _type_has_method_shash_reduce = false;
   static constexpr const bool _type_mutable = true;

cpp/serve/event_trace_recorder.cc

@@ -149,6 +149,7 @@ EventTraceRecorder EventTraceRecorder::Create() {
 
 TVM_FFI_STATIC_INIT_BLOCK() {
   namespace refl = tvm::ffi::reflection;
+  EventTraceRecorderImpl::RegisterReflection();
   refl::GlobalDef()
       .def("mlc.serve.EventTraceRecorder", []() { return EventTraceRecorder::Create(); })
       .def("mlc.serve.EventTraceRecorderAddEvent",

cpp/serve/event_trace_recorder.h

@@ -7,6 +7,7 @@
 #define MLC_LLM_SERVE_EVENT_TRACE_RECORDER_H_
 
 #include <tvm/ffi/container/array.h>
+#include <tvm/ffi/reflection/registry.h>
 #include <tvm/ffi/string.h>
 #include <tvm/runtime/object.h>
 
@@ -41,6 +42,11 @@ class EventTraceRecorderObj : public Object {
   /*! \brief Dump the logged events in Chrome Trace Event Format in JSON string. */
   virtual std::string DumpJSON() = 0;
 
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<EventTraceRecorderObj>();
+  }
+
   static constexpr const bool _type_has_method_sequal_reduce = false;
   static constexpr const bool _type_has_method_shash_reduce = false;
   static constexpr const bool _type_mutable = true;

cpp/serve/logit_processor.cc

@@ -32,6 +32,8 @@ inline void SyncCopyStream(Device device, TVMStreamHandle compute_stream,
 
 /***************** LogitProcessor Implementation *****************/
 
+TVM_FFI_STATIC_INIT_BLOCK() { LogitProcessorObj::RegisterReflection(); }
+
 class LogitProcessorImpl : public LogitProcessorObj {
  public:
   /*! * \brief Constructor of LogitProcessorImpl. */

cpp/serve/logit_processor.h

@@ -68,6 +68,11 @@ class LogitProcessorObj : public Object {
                                         const Array<String>& request_ids,
                                         const std::vector<int>* cum_num_token = nullptr) = 0;
 
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<LogitProcessorObj>();
+  }
+
   static constexpr const bool _type_has_method_sequal_reduce = false;
   static constexpr const bool _type_has_method_shash_reduce = false;
   static constexpr const bool _type_mutable = true;

cpp/serve/model.cc

@@ -23,6 +23,8 @@ namespace serve {
 
 /*********************** Model Implementation ***********************/
 
+TVM_FFI_STATIC_INIT_BLOCK() { ModelObj::RegisterReflection(); }
+
 class ModelImpl;
 
 Model Model::Create(String reload_lib_path, String model_path, const picojson::object& model_config,

cpp/serve/model.h

@@ -364,6 +364,11 @@ class ModelObj : public Object {
   /*! \brief Call the given global function on all workers. Only for debug purpose. */
   virtual void DebugCallFuncOnAllAllWorker(const String& func_name, Optional<String> func_args) = 0;
 
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<ModelObj>();
+  }
+
   static constexpr const bool _type_has_method_sequal_reduce = false;
   static constexpr const bool _type_has_method_shash_reduce = false;
   static constexpr const bool _type_mutable = true;

cpp/serve/prefix_cache.cc

@@ -13,6 +13,8 @@ namespace serve {
 
 using namespace tvm::runtime;
 
+TVM_FFI_STATIC_INIT_BLOCK() { PrefixCacheObj::RegisterReflection(); }
+
 /*!
  * \brief The implementation of prefix cache.
  */

cpp/serve/prefix_cache.h

@@ -5,6 +5,7 @@
 #ifndef MLC_LLM_SERVE_PREFIX_CACHE_H_
 #define MLC_LLM_SERVE_PREFIX_CACHE_H_
 #include <tvm/ffi/container/shape.h>
+#include <tvm/ffi/reflection/registry.h>
 #include <tvm/runtime/object.h>
 
 #include <functional>
@@ -122,6 +123,11 @@ class PrefixCacheObj : public Object {
   /*! \brief Return the prefix cache mode. */
   virtual PrefixCacheMode Mode() = 0;
 
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<PrefixCacheObj>();
+  }
+
   static constexpr const bool _type_mutable = true;
   TVM_FFI_DECLARE_OBJECT_INFO("mlc.serve.PrefixCache", PrefixCacheObj, Object);
 };

cpp/serve/radix_tree.cc

@@ -14,6 +14,8 @@ namespace serve {
 
 using namespace tvm::runtime;
 
+TVM_FFI_STATIC_INIT_BLOCK() { PagedRadixTreeObj::RegisterReflection(); }
+
 /*!
  * \brief The sequence ID linked list structure in paged radix tree node.
  */

cpp/serve/radix_tree.h

@@ -5,6 +5,7 @@
 #ifndef MLC_LLM_SERVE_RADIX_TREE_H_
 #define MLC_LLM_SERVE_RADIX_TREE_H_
 #include <tvm/ffi/container/shape.h>
+#include <tvm/ffi/reflection/registry.h>
 #include <tvm/runtime/int_tuple.h>
 #include <tvm/runtime/object.h>
 
@@ -107,6 +108,11 @@ class PagedRadixTreeObj : public Object {
    */
   virtual void Reset() = 0;
 
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<PagedRadixTreeObj>();
+  }
+
   static constexpr const bool _type_mutable = true;
   TVM_FFI_DECLARE_OBJECT_INFO("mlc.serve.PagedRadixTree", PagedRadixTreeObj, Object);
 };

cpp/serve/request.cc

@@ -16,6 +16,8 @@ namespace serve {
 
 /****************** Request ******************/
 
+TVM_FFI_STATIC_INIT_BLOCK() { RequestNode::RegisterReflection(); }
+
 Request::Request(String id, Array<Data> inputs, GenerationConfig generation_cfg) {
   if (generation_cfg->debug_config.special_request == SpecialRequestKind::kNone) {
     CHECK(!inputs.empty()) << "No input data is given.";

cpp/serve/request.h

@@ -7,6 +7,7 @@
 #define MLC_LLM_SERVE_REQUEST_H_
 
 #include <tvm/ffi/container/array.h>
+#include <tvm/ffi/reflection/registry.h>
 #include <tvm/ffi/string.h>
 #include <tvm/runtime/object.h>
 
@@ -55,6 +56,11 @@ class RequestNode : public Object {
   /*! \brief Backward reference to the request state. */
   Object* rstate = nullptr;
 
+  static void RegisterReflection() {
+    namespace refl = tvm::ffi::reflection;
+    refl::ObjectDef<RequestNode>();
+  }
+
   static constexpr const bool _type_has_method_sequal_reduce = false;
   static constexpr const bool _type_has_method_shash_reduce = false;
   static constexpr const bool _type_mutable = true;

cpp/serve/request_state.cc

@@ -9,6 +9,12 @@ namespace mlc {
 namespace llm {
 namespace serve {
 
+TVM_FFI_STATIC_INIT_BLOCK() {
+  RequestModelStateNode::RegisterReflection();
+  RequestStateEntryNode::RegisterReflection();
+  RequestStateNode::RegisterReflection();
+}
+
 /****************** RequestModelState ******************/
 
 RequestModelState::RequestModelState(