models

README.md

@@ -1 +1,16 @@
 # ML
+
+
+needed packages for venv
+
+  302  sudo apt install python3-dev python3-pip
+  303  sudo pip3 install -U virtualenv 
+  331  pip list
+  332  pip install --upgrade tensorflow
+  382  pip3 --version
+  383  sudo apt install python3-pip
+  437  pip install --upgrade matplotlib
+  439  pip install --upgrade pandas
+  441  pip install --upgrade seaborn
+  443  pip install --upgrade tensorflow_docs
+  444  pip install git+https://github.com/tensorflow/docs

regression/ml1.py

@@ -0,0 +1,43 @@
+import tensorflow as tf
+
+mnist = tf.keras.datasets.mnist
+
+(x_train, y_train), (x_test, y_test) = mnist.load_data()
+x_train, x_test = x_train / 255.0, x_test / 255.0
+
+
+model = tf.keras.models.Sequential([
+  tf.keras.layers.Flatten(input_shape=(28, 28)),
+  tf.keras.layers.Dense(128, activation='relu'),
+  tf.keras.layers.Dropout(0.2),
+  tf.keras.layers.Dense(10)
+])
+
+
+predictions = model(x_train[:1]).numpy()
+predictions
+
+print ("predictions:")
+print(predictions)
+
+
+tf.nn.softmax(predictions).numpy()
+
+print (" softmax:")
+print(predictions)
+
+
+loss_fn = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)
+loss_fn(y_train[:1], predictions).numpy()
+
+
+model.compile(optimizer='adam',
+              loss=loss_fn,
+              metrics=['accuracy'])
+
+
+model.fit(x_train, y_train, epochs=5)
+
+
+model.evaluate(x_test,  y_test, verbose=2)
+

service/examples/coco/main.cpp

@@ -1,7 +1,3 @@
-//
-// Created by sergio on 16/05/19.
-//
-
 #include "../../include/Model.h"
 #include "../../include/Tensor.h"
 #include <opencv2/opencv.hpp>

service/examples/load_model/main.cpp

@@ -1,7 +1,3 @@
-//
-// Created by sergio on 16/05/19.
-//
-
 #include "../../include/Model.h"
 #include "../../include/Tensor.h"
 

service/examples/mnist/main.cpp

@@ -1,6 +1,3 @@
-//
-// Created by sergio on 12/05/19.
-//
 
 #include "../../include/Model.h"
 #include "../../include/Tensor.h"

service/include/Model.h

@@ -1,9 +1,8 @@
 //
-// Created by sergio on 12/05/19.
+// Model
 //
 
-#ifndef CPPFLOW_MODEL_H
-#define CPPFLOW_MODEL_H
+#pragma once
 
 #include <cstring>
 #include <algorithm>
@@ -12,11 +11,15 @@
 #include <iostream>
 #include <fstream>
 #include <tuple>
+#include <memory>
 #include <tensorflow/c/c_api.h>
 #include "Tensor.h"
 
 class Tensor;
 
+//using GraphResourcePtr = std::unique_ptr<TF_Graph> ;
+
+
 class Model {
 public:
     explicit Model(const std::string&);
@@ -48,8 +51,15 @@ class Model {
     void run(const std::vector<Tensor*>& inputs, Tensor* output);
     void run(Tensor* input, Tensor* output);
 
+    struct GraphCreate {
+        TF_Graph * operator()() { return TF_NewGraph(); }
+    };
+    struct GraphDeleter {
+        void operator()(TF_Graph* b) { TF_DeleteGraph(b); }
+    };
+
 private:
-    TF_Graph* graph;
+    std::unique_ptr<TF_Graph, GraphDeleter> graph;
     TF_Session* session;
     TF_Status* status;
 
@@ -64,4 +74,3 @@ class Model {
 };
 
 
-#endif //CPPFLOW_MODEL_H

service/include/Tensor.h

@@ -1,9 +1,4 @@
-//
-// Created by sergio on 13/05/19.
-//
-
-#ifndef CPPFLOW_TENSOR_H
-#define CPPFLOW_TENSOR_H
+#pragma once
 
 #include <vector>
 #include <memory>
@@ -63,4 +58,3 @@ class Tensor {
     friend class Model;
 };
 
-#endif //CPPFLOW_TENSOR_H

service/src/Model.cpp

@@ -1,33 +1,29 @@
-//
-// Created by sergio on 12/05/19.
-//
-
 #include "../include/Model.h"
 
-Model::Model(const std::string& model_filename) {
+Model::Model(const std::string& model_filename):graph(TF_NewGraph()){
 
     this->status = TF_NewStatus();
-    this->graph = TF_NewGraph();
+    //this->graph = std::make_unique<TF_Graph>(GraphCreate(), GraphDeleter); //TF_NewGraph();
 
     // Create the session.
     TF_SessionOptions* sess_opts = TF_NewSessionOptions();
 
-    this->session = TF_NewSession(this->graph, sess_opts, this->status);
+    this->session = TF_NewSession(graph.get(), sess_opts, this->status);
     TF_DeleteSessionOptions(sess_opts);
 
     // Check the status
     this->status_check(true);
 
     // Create the graph
-    TF_Graph* g = this->graph;
+    ////TF_Graph* g = this->graph;
 
 
     // Import the graph definition
     TF_Buffer* def = read(model_filename);
     this->error_check(def != nullptr, "An error occurred reading the model");
 
     TF_ImportGraphDefOptions* graph_opts = TF_NewImportGraphDefOptions();
-    TF_GraphImportGraphDef(g, def, graph_opts, this->status);
+    TF_GraphImportGraphDef(graph.get(), def, graph_opts, this->status);
     TF_DeleteImportGraphDefOptions(graph_opts);
     TF_DeleteBuffer(def);
 
@@ -37,14 +33,14 @@ Model::Model(const std::string& model_filename) {
 
 Model::~Model() {
     TF_DeleteSession(this->session, this->status);
-    TF_DeleteGraph(this->graph);
+    ///TF_DeleteGraph(this->graph);
     this->status_check(true);
     TF_DeleteStatus(this->status);
 }
 
 
 void Model::init() {
-    TF_Operation* init_op[1] = {TF_GraphOperationByName(this->graph, "init")};
+    TF_Operation* init_op[1] = {TF_GraphOperationByName(this->graph.get(), "init")};
 
     this->error_check(init_op[0]!= nullptr, "Error: No operation named \"init\" exists");
 
@@ -71,12 +67,12 @@ void Model::save(const std::string &ckpt) {
     }
 
     TF_Output output_file;
-    output_file.oper = TF_GraphOperationByName(this->graph, "save/Const");
+    output_file.oper = TF_GraphOperationByName(this->graph.get(), "save/Const");
     output_file.index = 0;
     TF_Output inputs[1] = {output_file};
 
     TF_Tensor* input_values[1] = {t};
-    const TF_Operation* restore_op[1] = {TF_GraphOperationByName(this->graph, "save/control_dependency")};
+    const TF_Operation* restore_op[1] = {TF_GraphOperationByName(this->graph.get(), "save/control_dependency")};
     if (!restore_op[0]) {
         TF_DeleteTensor(t);
         this->error_check(false, "Error: No operation named \"save/control_dependencyl\" exists");
@@ -106,12 +102,12 @@ void Model::restore(const std::string& ckpt) {
     }
 
     TF_Output output_file;
-    output_file.oper = TF_GraphOperationByName(this->graph, "save/Const");
+    output_file.oper = TF_GraphOperationByName(this->graph.get(), "save/Const");
     output_file.index = 0;
     TF_Output inputs[1] = {output_file};
 
     TF_Tensor* input_values[1] = {t};
-    const TF_Operation* restore_op[1] = {TF_GraphOperationByName(this->graph, "save/restore_all")};
+    const TF_Operation* restore_op[1] = {TF_GraphOperationByName(this->graph.get(), "save/restore_all")};
     if (!restore_op[0]) {
         TF_DeleteTensor(t);
         this->error_check(false, "Error: No operation named \"save/restore_all\" exists");
@@ -168,7 +164,7 @@ std::vector<std::string> Model::get_operations() const {
     TF_Operation* oper;
 
     // Iterate through the operations of a graph
-    while ((oper = TF_GraphNextOperation(this->graph, &pos)) != nullptr) {
+    while ((oper = TF_GraphNextOperation(this->graph.get(), &pos)) != nullptr) {
         result.emplace_back(TF_OperationName(oper));
     }
 

service/src/Tensor.cpp

@@ -1,15 +1,11 @@
-//
-// Created by sergio on 13/05/19.
-//
-
 #include "../include/Tensor.h"
 
 #include <utility>
 
 Tensor::Tensor(const Model& model, const std::string& operation) {
 
     // Get operation by the name
-    this->op.oper = TF_GraphOperationByName(model.graph, operation.c_str());
+    this->op.oper = TF_GraphOperationByName(model.graph.get(), operation.c_str());
     this->op.index = 0;
 
     // Operation did not exists
@@ -18,7 +14,7 @@ Tensor::Tensor(const Model& model, const std::string& operation) {
     // DIMENSIONS
 
     // Get number of dimensions
-    int n_dims = TF_GraphGetTensorNumDims(model.graph, this->op, model.status);
+    int n_dims = TF_GraphGetTensorNumDims(model.graph.get(), this->op, model.status);
 
     // DataType
     this->type = TF_OperationOutputType(this->op);
@@ -27,7 +23,7 @@ Tensor::Tensor(const Model& model, const std::string& operation) {
     if (n_dims > 0) {
         // Get dimensions
         auto *dims = new int64_t[n_dims];
-        TF_GraphGetTensorShape(model.graph, this->op, dims, n_dims, model.status);
+        TF_GraphGetTensorShape(model.graph.get(), this->op, dims, n_dims, model.status);
 
         // Check error on Model Status
         model.status_check(true);