jacobian

commit 1f3f753317f0f5e2699b27a64fbff9e32d5e82af
parent 576b6092063be05f05b8adb9c828702c1ba0b4fc
Author: David Freifeld <freifeld.david@gmail.com>
Date:   Wed,  1 Jul 2020 14:55:55 -0700

Using wandb to measure things more properly

Diffstat:
M
bpnn.cpp
 | 
7
+++++++
M
bpnn.hpp
 | 
3
+++
M
example.cpp
 | 
11
++---------
M
example.py
 | 
71
++++++++++++++++++-----------------------------------------------------
M
mr_bpnn_2.cpp
 | 
3
++-

5 files changed, 32 insertions(+), 63 deletions(-)
diff --git a/bpnn.cpp b/bpnn.cpp
@@ -294,6 +294,7 @@ void Network::train(int total_epochs)
       //t++;
     }
     epoch_accuracy = 1.0/((float) instances/batch_size) * acc_sum;
+    epoch_acc = epoch_accuracy;
     epoch_cost = 1.0/((float) instances/batch_size) * cost_sum;
     printf("Epoch %i/%i - cost %f - acc %f\n", epochs+1, total_epochs, epoch_cost, epoch_accuracy);
     batches=1;
@@ -301,3 +302,9 @@ void Network::train(int total_epochs)
     rewind(data);
   }
 }
+
+float Network::get_info()
+{
+  return epoch_acc;
+}
+  
diff --git a/bpnn.hpp b/bpnn.hpp
@@ -38,6 +38,7 @@ public:
   int length;
   int t;
 
+  float epoch_acc;
   float learning_rate;
   float bias_lr;
   int batch_size;
@@ -60,6 +61,8 @@ public:
   int next_batch();
   float test(char* path);
   void train(int total_epochs);
+
+  float get_info();
 };
 
 void demo(int total_epochs);
diff --git a/example.cpp b/example.cpp
@@ -18,13 +18,6 @@ double bench(int batch_sz)
 
 int main()
 {
-  double x[1340];
-  for(int i = 0; i < 3; i++) {
-    x[i] = bench(i);
-  }
-  printf("[");
-  for(int i = 0; i < 3; i++) {
-    printf("%d, ", x[i]);
-  }
-  printf("]");
+  sleep(10);
+  bench(10);
 }
diff --git a/example.py b/example.py
@@ -3,56 +3,21 @@ import matplotlib.pyplot as plt
 import numpy
 import time
 
-def bench(batch_sz, layers):
-    init = time.time()
-    net = mrbpnn.Network("./data_banknote_authentication.txt", batch_sz, 0.0155, 0.03)
-    net.add_layer(4, "linear")
-    for i in range(layers):
-        net.add_layer(5, "relu")
-    net.add_layer(1, "resig")
-    net.initialize()
-#    net.list_net()
-    net.train(50)
-    end = time.time()
-    return (end-init)
-
-# timesum=0
-# trials = 20
-# for i in range(trials):
-#     timesum+=bench(10,1)
-# print("Averages over %s trials\n--------------\nTime: %s seconds.\n" % (trials, timesum/trials))
-x = []
-y = []
-i = 1
-while(i < 1340):
-    y.append(bench(i, 1))
-    x.append(i)
-    i+=1
-plt.plot(x,y, label = "Jacobian (Sequential)")
-i = 20
-otherlist = []
-while(i < 1340):
-    otherlist.append(i)
-    i += 20
-
-plt.plot(otherlist, [2.3994078636169434, 1.2735769748687744, 1.0030598640441895, 0.8972160816192627, 0.801548957824707, 0.752018928527832, 0.7073678970336914, 0.6857280731201172, 0.6707980632781982, 0.6421489715576172, 0.6614980697631836, 0.6403779983520508, 0.7251319885253906, 0.6796879768371582, 0.6601080894470215, 0.6711599826812744, 0.6432759761810303, 0.6491389274597168, 0.6762490272521973, 0.6859049797058105, 0.7067179679870605, 0.7142889499664307, 0.7258059978485107, 0.7868969440460205, 0.7326970100402832, 0.7365641593933105, 0.7576079368591309, 0.7772500514984131, 0.8062641620635986, 0.7768490314483643, 0.8253629207611084, 0.8264601230621338, 0.8459320068359375, 0.9670729637145996, 0.8388969898223877, 0.9129719734191895, 0.9009649753570557, 0.8916170597076416, 0.8926799297332764, 0.9171609878540039, 0.9242072105407715, 0.9534740447998047, 0.947465181350708, 0.9723358154296875, 1.018247127532959, 1.1208629608154297, 1.014026165008545, 1.034980058670044, 1.0626468658447266, 1.080394983291626, 1.0627479553222656, 1.0839190483093262, 1.0938241481781006, 1.127730131149292, 1.1265759468078613, 1.136888027191162, 1.140428066253662, 1.1712510585784912, 1.206390142440796, 1.2087180614471436, 1.4066569805145264, 1.2425589561462402, 1.280066967010498, 1.2891559600830078, 1.3243582248687744, 1.3152379989624023], label="Keras")
-plt.legend()
-plt.show()
-
-# # sum = 0
-# # for i in range(10):
-# #     sum += bench(10, 1)
-# # print(sum/10)
-
-# # x = ['Keras', 'MIP-Sequential']
-# # speed = [4.71297559738, 0.043680644035339354]
-
-# # x_pos = [i for i, _ in enumerate(x)]
-
-# # plt.bar(x_pos, speed, color='green')
-# # plt.ylabel("Time (s)")
-# # plt.title("Average Runtime (10 trials)")
-
-# # plt.xticks(x_pos, x)
-
-# # plt.show()
+batch_sz = 10
+layers = 1
+
+net = mrbpnn.Network("./data_banknote_authentication.txt", batch_sz, 0.0155, 0.03)
+net.add_layer(4, "linear")
+for i in range(layers):
+    net.add_layer(5, "relu")
+net.add_layer(1, "resig")
+net.initialize()
+
+import wandb
+wandb.init(project="jacobian")
+wandb.config.update({"epochs": 50, "batch_size": batch_sz, "hidden_layers": layers})
+for i in range(50):
+    net.train(1)
+    wandb.log({'accuracy': net.get_info()})
+
+wandb.save('jacobian.h5')
diff --git a/mr_bpnn_2.cpp b/mr_bpnn_2.cpp
@@ -118,5 +118,6 @@ PYBIND11_MODULE(mrbpnn, m) {
     .def("accuracy", &Network::accuracy)
     .def("update_layer", &Network::update_layer, py::arg("vals"), py::arg("len"), py::arg("index"))
     .def("next_batch", &Network::next_batch)
-    .def("train", &Network::train, py::arg("epochs"));
+    .def("train", &Network::train, py::arg("epochs"))
+    .def("get_info", &Network::get_info);
 }