支持pagedattention , continuous batch 提供基础支持 , KVcachemanger 用于 KVcache管理 ，支持调度器调度

.vscode/c_cpp_properties.json

@@ -0,0 +1,21 @@
+{
+    "configurations": [
+        {
+            "name": "Linux",
+            "includePath": [
+                "${workspaceFolder}/**",
+                "${workspaceFolder}/kuiper/include",
+                "/usr/local/cuda/include",
+                "/usr/include/c++/**",
+                "/usr/include"
+            ],
+            "defines": [],
+            "compilerPath": "/usr/bin/g++",
+            "cStandard": "c17",
+            "cppStandard": "c++17",
+            "intelliSenseMode": "linux-gcc-x64",
+            "compileCommands": "${workspaceFolder}/build/compile_commands.json"
+        }
+    ],
+    "version": 4
+}

.vscode/launch.json

@@ -0,0 +1,30 @@
+{
+    // Use IntelliSense to learn about possible attributes.
+    // Hover to view descriptions of existing attributes.
+    // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
+    "version": "0.2.0",
+    "configurations": [
+        {
+            "name": "C++ Launch",
+            "type": "cppdbg",
+            "request": "launch",
+            "program": "/home/tangjin/KuiperLLama/build/demo/llama_infer",  // 可执行文件路径
+            "args": [],  // 可选参数
+            "stopAtEntry": false,
+            "cwd": "${workspaceFolder}",
+            "environment": [],                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                               
+            "externalConsole": false,
+            "MIMode": "gdb",
+            "miDebuggerPath": "/usr/bin/gdb",  // 调试器路径，可以是gdb或lldb
+            "setupCommands": [
+                {
+                "description": "Enable pretty-printing for gdb",
+                "text": "-enable-pretty-printing",
+                "ignoreFailures": true
+                }
+            ],
+            "preLaunchTask": "build",  // 构建任务名称, 可选
+            "internalConsoleOptions": "openOnSessionStart"
+        }
+    ]
+}

.vscode/settings.json

@@ -0,0 +1,118 @@
+
+{
+
+    "files.associations": {
+        "cctype": "cpp",
+        "cmath": "cpp",
+        "cstddef": "cpp",
+        "cstdio": "cpp",
+        "cstdlib": "cpp",
+        "cstring": "cpp",
+        "ctime": "cpp",
+        "cwchar": "cpp",
+        "cwctype": "cpp",
+        "array": "cpp",
+        "atomic": "cpp",
+        "bit": "cpp",
+        "*.tcc": "cpp",
+        "bitset": "cpp",
+        "chrono": "cpp",
+        "compare": "cpp",
+        "concepts": "cpp",
+        "cstdint": "cpp",
+        "unordered_map": "cpp",
+        "vector": "cpp",
+        "exception": "cpp",
+        "algorithm": "cpp",
+        "functional": "cpp",
+        "iterator": "cpp",
+        "memory": "cpp",
+        "memory_resource": "cpp",
+        "random": "cpp",
+        "ratio": "cpp",
+        "string": "cpp",
+        "string_view": "cpp",
+        "tuple": "cpp",
+        "type_traits": "cpp",
+        "utility": "cpp",
+        "initializer_list": "cpp",
+        "iosfwd": "cpp",
+        "istream": "cpp",
+        "limits": "cpp",
+        "new": "cpp",
+        "ostream": "cpp",
+        "ranges": "cpp",
+        "sstream": "cpp",
+        "stdexcept": "cpp",
+        "streambuf": "cpp",
+        "thread": "cpp",
+        "typeinfo": "cpp",
+        "__nullptr": "cpp",
+        "iostream": "cpp",
+        "cstdarg": "cpp",
+        "clocale": "cpp",
+        "complex": "cpp",
+        "condition_variable": "cpp",
+        "deque": "cpp",
+        "list": "cpp",
+        "map": "cpp",
+        "set": "cpp",
+        "fstream": "cpp",
+        "iomanip": "cpp",
+        "mutex": "cpp",
+        "numbers": "cpp",
+        "numeric": "cpp",
+        "optional": "cpp",
+        "semaphore": "cpp",
+        "stop_token": "cpp",
+        "system_error": "cpp",
+        "typeindex": "cpp",
+        "variant": "cpp",
+        "filesystem": "cpp",
+        "any": "cpp",
+        "regex": "cpp",
+        "unordered_set": "cpp",
+        "*.inc": "cpp",
+        "*.ipp": "cpp",
+        "span": "cpp",
+        "__node_handle": "cpp",
+        "__split_buffer": "cpp",
+        "queue": "cpp",
+        "stack": "cpp",
+        "shared_mutex": "cpp",
+        "__mutex_base": "cpp",
+        "valarray": "cpp",
+        "__hash_table": "cpp",
+        "__tree": "cpp",
+        "__locale": "cpp",
+        "__string": "cpp",
+        "forward_list": "cpp",
+        "__bit_reference": "cpp",
+        "csetjmp": "cpp",
+        "csignal": "cpp",
+        "strstream": "cpp",
+        "cfenv": "cpp",
+        "cinttypes": "cpp",
+        "codecvt": "cpp",
+        "source_location": "cpp",
+        "future": "cpp",
+        "barrier": "cpp",
+        "charconv": "cpp",
+        "coroutine": "cpp",
+        "cuchar": "cpp",
+        "latch": "cpp",
+        "scoped_allocator": "cpp",
+        "syncstream": "cpp",
+        "hash_map": "cpp",
+        "hash_set": "cpp",
+        "__config": "cpp",
+        "__tuple": "cpp",
+        "ios": "cpp"
+    },
+    "C_Cpp.errorSquiggles": "disabled",
+    "window.zoomLevel": 3,
+    "window.zoomPerWindow": false,
+    "testMate.cpp.test.executables": "/home/tangjin/KuiperLLama/build/*/*",
+    "testMate.cpp.test.workingDirectory": "${absDirpath}",
+    "cmake.ctest.testExplorerIntegrationEnabled": true
+}

.vscode/task.json

@@ -0,0 +1,16 @@
+{
+	"version": "2.0.0",
+	"tasks": [
+		{
+			"type": "cmake",
+			"label": "build",
+			"command": "build",
+			"targets": [
+				"all"
+			],
+			"group": "build",
+			"problemMatcher": [],
+			"detail": "CMake template build task"
+		}
+	]
+}

CMakeLists.txt

@@ -91,7 +91,8 @@ find_package(Armadillo REQUIRED)
 aux_source_directory(kuiper/source/tensor/ DIR_TENSOR)
 aux_source_directory(kuiper/source/base/ DIR_BASE)
 aux_source_directory(kuiper/source/op/ DIR_OP)
-aux_source_directory(kuiper/source/model/ DIR_MODEL)
+file(GLOB DIR_MODEL CONFIGURE_DEPENDS "${CMAKE_CURRENT_SOURCE_DIR}/kuiper/source/model/*.cpp")
+list(REMOVE_ITEM DIR_MODEL "${CMAKE_CURRENT_SOURCE_DIR}/kuiper/source/model/model_paged.cpp")
 aux_source_directory(kuiper/source/op/kernels/cpu DIR_KERNEL_CPU)
 aux_source_directory(kuiper/source/op/kernels/cuda DIR_KERNEL_CUDA)
 aux_source_directory(kuiper/source/op/kernels/ DIR_KERNEL)

demo/CMakeLists.txt

@@ -11,6 +11,39 @@ if (LLAMA3_SUPPORT)
 endif ()
 set_target_properties(llama_infer PROPERTIES CUDA_SEPARABLE_COMPILATION ON)
 
+add_executable(llama_infer_paged mainpaged.cpp)
+target_link_directories(llama_infer_paged PUBLIC ${PROJECT_SOURCE_DIR}/lib)
+target_link_libraries(llama_infer_paged llama)
+if (LLAMA3_SUPPORT)
+    find_package(absl REQUIRED)
+    find_package(re2 REQUIRED)
+    find_package(nlohmann_json REQUIRED)
+    target_link_libraries(llama_infer_paged absl::base re2::re2 nlohmann_json::nlohmann_json)
+endif ()
+set_target_properties(llama_infer_paged PROPERTIES CUDA_SEPARABLE_COMPILATION ON)
+
+add_executable(llama_infer_unified main_unified.cpp)
+target_link_directories(llama_infer_unified PUBLIC ${PROJECT_SOURCE_DIR}/lib)
+target_link_libraries(llama_infer_unified llama)
+if (LLAMA3_SUPPORT)
+    find_package(absl REQUIRED)
+    find_package(re2 REQUIRED)
+    find_package(nlohmann_json REQUIRED)
+    target_link_libraries(llama_infer_unified absl::base re2::re2 nlohmann_json::nlohmann_json)
+endif ()
+set_target_properties(llama_infer_unified PROPERTIES CUDA_SEPARABLE_COMPILATION ON)
+
+add_executable(llama_infer_continuous main_continuous.cpp)
+target_link_directories(llama_infer_continuous PUBLIC ${PROJECT_SOURCE_DIR}/lib)
+target_link_libraries(llama_infer_continuous llama)
+if (LLAMA3_SUPPORT)
+    find_package(absl REQUIRED)
+    find_package(re2 REQUIRED)
+    find_package(nlohmann_json REQUIRED)
+    target_link_libraries(llama_infer_continuous absl::base re2::re2 nlohmann_json::nlohmann_json)
+endif ()
+set_target_properties(llama_infer_continuous PROPERTIES CUDA_SEPARABLE_COMPILATION ON)
+
 if (QWEN2_SUPPORT)
     message(STATUS "LINK QWEN2 SUPPORT")
     add_executable(qwen_infer main_qwen.cpp)