From 2be628552507f6b7413ac7360fd28157c3af7a53 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Aur=C3=A9lien=20Hamy?= <aunetx@users.noreply.github.com>
Date: Sun, 10 Feb 2019 20:52:19 +0100
Subject: [PATCH] Added support for json load of params in train Updated
 gitignore

---
 .gitignore       |  3 +++
 loulou.py        |  2 ++
 scripts/ffl.py   |  2 +-
 scripts/train.py | 22 ++++++++++++++++++----
 4 files changed, 24 insertions(+), 5 deletions(-)
 create mode 100644 loulou.py

diff --git a/.gitignore b/.gitignore
index 510c73d..6f1e0b0 100644
--- a/.gitignore
+++ b/.gitignore
@@ -112,3 +112,6 @@ dmypy.json
 
 # Pyre type checker
 .pyre/
+
+# Others
+testimage.png
diff --git a/loulou.py b/loulou.py
new file mode 100644
index 0000000..a94446f
--- /dev/null
+++ b/loulou.py
@@ -0,0 +1,2 @@
+import scripts.train as train
+import scripts.run as run
diff --git a/scripts/ffl.py b/scripts/ffl.py
index 056dd9e..5e7cf03 100644
--- a/scripts/ffl.py
+++ b/scripts/ffl.py
@@ -18,7 +18,7 @@ def grads(X, Y, weights):
         grads[i-1] = a[i-1].T.dot(delta) # calculating errors of weights and storing onto |grads|
     return grads / len(X)
 
-def train(weights, trX, trY, teX, teY, filename=False, epochs=30, batch=20, learning_rate=0.03):
+def train(weights, trX, trY, teX, teY, filename, epochs, batch, learning_rate):
     prediction = np.argmax(feed_forward(teX, weights)[-1], axis=1)
     print(0, np.mean(prediction == np.argmax(teY, axis=1)))
     for i in range(epochs):
diff --git a/scripts/train.py b/scripts/train.py
index 16fabd0..6fbe3e9 100644
--- a/scripts/train.py
+++ b/scripts/train.py
@@ -1,9 +1,23 @@
 import mnist
 from ffl import *
+import json
 
-if __name__ == '__main__':
-    filename = 'four layers.npy'
-    filename = '../trains/' + filename
+def runTrain(params, file='trained.npy'):
+    params = json.loads(params)
+    epochs = params['epochs']
+    batch = params['batch']
+    learning_rate = params['learning_rate']
+    file = '../trains/' + file
     trX, trY, teX, teY = mnist.load_data()
     weights = [np.random.randn(*w) * 0.1 for w in [(784, 400), (400,200), (200,100), (100, 10)]]
-    train(weights, trX, trY, teX, teY, filename=filename, epochs=0)
+    train(weights, trX, trY, teX, teY, 'trained.npy', epochs, batch, learning_rate)
+
+if __name__ == '__main__':
+    params = {}
+    params['epochs'] = 30
+    params['batch'] = 20
+    params['learning_rate'] = 0.03
+    params = json.dumps(params)
+
+    filename = 'trained.npy'
+    runTrain(params, file=filename)