Merge branch 'master' into llama

CMakeLists.txt

@@ -33,7 +33,7 @@ endif()
 
 message(STATUS "CMAKE_CXX_FLAGS" ${CMAKE_CXX_FLAGS})
 set(FASTLLM_CXX_SOURCES src/fastllm.cpp src/device.cpp src/model.cpp src/executor.cpp
-        src/devices/cpu/cpudevice.cpp src/devices/cpu/cpudevicebatch.cpp
+        src/devices/cpu/cpudevice.cpp src/devices/cpu/cpudevicebatch.cpp src/models/minicpm.cpp
         src/models/chatglm.cpp src/models/moss.cpp src/models/llama.cpp src/models/qwen.cpp src/models/basellm.cpp src/models/glm.cpp)
 
 include_directories(include)

include/models/minicpm.h

@@ -0,0 +1,71 @@
+//
+// Created by huangyuyang on 6/1/23.
+//
+
+#ifndef FASTLLM_MINICPM_H
+#define FASTLLM_MINICPM_H
+
+#include "basellm.h"
+#include "cmath"
+
+#include <iostream>
+
+namespace fastllm {
+    class MiniCpmModel: public basellm {
+    public:
+        MiniCpmModel(); // 构造函数
+
+        // 推理
+        virtual int Forward(
+                const Data &inputIds,
+                const Data &attentionMask,
+                const Data &positionIds,
+                std::vector <std::pair <Data, Data> > &pastKeyValues,
+                const GenerationConfig &generationConfig = GenerationConfig(),
+                const LastTokensManager &lastTokens = LastTokensManager(),
+                std::vector <float> *logits = nullptr);
+
+        std::vector <int> ForwardBatch(
+                int batch,
+                const Data &inputIds,
+                const Data &attentionMask,
+                const Data &positionIds,
+                std::vector <std::pair <Data, Data> > &pastKeyValues,
+                const GenerationConfig &generationConfig = GenerationConfig(),
+                const LastTokensManager &lastTokens = LastTokensManager(),
+                std::vector <std::vector <float>*> *logits = nullptr);
+
+        std::vector <int> ForwardBatch(
+                int batch,
+                const Data &inputIds,
+                const std::vector <Data*> &attentionMask,
+                const std::vector <Data*> &positionIds,
+                const std::vector <int> &seqLens,
+                std::vector <std::pair <Data*, Data*> > &pastKeyValues,
+                const std::vector <GenerationConfig> &generationConfigs,
+                const LastTokensManager &lastTokens = LastTokensManager(),
+                std::vector <std::vector <float>*> *logits = nullptr);
+
+        virtual std::string Response(const std::string& input,
+                                     RuntimeResult retCb,
+                                     const GenerationConfig &generationConfig = GenerationConfig()); // 根据给出的内容回复
+
+        virtual void ResponseBatch(const std::vector <std::string> &inputs,
+                                   std::vector <std::string> &outputs,
+                                   RuntimeResultBatch retCb,
+                                   const GenerationConfig &generationConfig = GenerationConfig());
+
+        virtual int LaunchResponseTokens(const std::vector <int> &inputTokens,
+                                         const GenerationConfig &generationConfig = GenerationConfig()); // 启动一个response任务，返回分配的handleId
+
+        virtual int FetchResponseTokens(int handelId); // 获取指定handle的输出, -1代表输出结束了
+
+        virtual void WarmUp(); // 预热
+
+        virtual std::string MakeInput(const std::string &history, int round, const std::string &input); // 根据历史信息和当前输入生成prompt
+
+        virtual std::string MakeHistory(const std::string &history, int round, const std::string &input, const std::string &output); // 根据当前回复更新history
+    };
+}
+
+#endif //FASTLLM_MINICPM_H

src/model.cpp

@@ -9,6 +9,7 @@
 #include "llama.h"
 #include "qwen.h"
 #include "glm.h"
+#include "minicpm.h"
 
 namespace fastllm {
     void basellm::LoadFromFile(const std::string &fileName) {
@@ -106,6 +107,8 @@ namespace fastllm {
         } else if (modelType == "qwen2") {
             model = new LlamaModel();
             model->model_type = "qwen";
+        } else if (modelType=="minicpm") {
+	          model = (basellm*)(new MiniCpmModel());
         } else if (modelType == "qwen") {
             model = (basellm *) (new QWenModel());
             model->weight.tokenizer.type = Tokenizer::TokenizerType::QWEN;