Sourcery Starbot ⭐ refactored manncodes/xeno

xeno/activations.py

@@ -59,7 +59,7 @@ def forward(self, input):
         return np.maximum(0.0, input)
 
     def derivative(self, input=None):
-        last_forward = input if input else self.last_forward
+        last_forward = input or self.last_forward
         res = np.zeros(last_forward.shape, dtype=get_dtype())
         res[last_forward > 0] = 1.
         return res
@@ -76,7 +76,7 @@ def forward(self, input):
         return input
 
     def derivative(self, input=None):
-        last_forward = input if input else self.last_forward
+        last_forward = input or self.last_forward
         return np.ones(last_forward.shape, dtype=get_dtype())
 
 
@@ -93,11 +93,10 @@ def forward(self, input):
         self.last_forward = input
         x = input - np.max(input, axis=1, keepdims=True)
         exp_x = np.exp(x)
-        s = exp_x / np.sum(exp_x, axis=1, keepdims=True)
-        return s
+        return exp_x / np.sum(exp_x, axis=1, keepdims=True)
 
     def derivative(self, input=None):
-        last_forward = input if input else self.last_forward
+        last_forward = input or self.last_forward
         return np.ones(last_forward.shape, dtype=get_dtype())
 
 

xeno/initializations.py

@@ -95,7 +95,7 @@ def decompose_size(size):
         fan_in = size[0]
         fan_out = size[1]
 
-    elif len(size) == 4 or len(size) == 5:
+    elif len(size) in [4, 5]:
         respective_field_size = np.prod(size[2:])
         fan_in = size[1] * respective_field_size
         fan_out = size[0] * respective_field_size

xeno/layers/core.py

@@ -86,8 +86,7 @@ def forward(self, input, *args, **kwargs):
 
         self.last_input = input
         linear_out = np.dot(input, self.W) + self.b
-        act_out = self.act_layer.forward(linear_out)
-        return act_out
+        return self.act_layer.forward(linear_out)
 
     def backward(self, pre_grad, *args, **kwargs):
 
@@ -128,11 +127,10 @@ def connect_to(self, prev_layer):
     def forward(self, input, train=True, *args, **kwargs):
 
         if 0. < self.p < 1.:
-            if train:
-                self.last_mask = get_rng().binomial(1, 1 - self.p, input.shape) / (1 - self.p)
-                return input * self.last_mask
-            else:
+            if not train:
                 return input * (1 - self.p)
+            self.last_mask = get_rng().binomial(1, 1 - self.p, input.shape) / (1 - self.p)
+            return input * self.last_mask
         else:
             return input
 

xeno/layers/normalization.py

@@ -92,10 +92,7 @@ def backward(self, pre_grad, *args, **kwargs):
         dmu = -1 * np.sum(dxmu1 + dxmu2, axis=0)
         dx2 = 1. / N * np.ones((N, D)) * dmu
 
-        # step0 done!
-        dx = dx1 + dx2
-
-        return dx
+        return dx1 + dx2
 
     @property
     def params(self):

xeno/model.py

@@ -64,10 +64,7 @@ def fit(self, X, Y, max_iter=100, batch_size=64, shuffle=True,
         else:
             valid_X, valid_Y = None, None
 
-        iter_idx = 0
-        while iter_idx < max_iter:
-            iter_idx += 1
-
+        for iter_idx in range(1, max_iter + 1):
             # shuffle
             if shuffle:
                 seed = get_rng().randint(111, 1111111)
@@ -142,8 +139,7 @@ def predict(self, X):
         x_next = X
         for layer in self.layers[:]:
             x_next = layer.forward(x_next)
-        y_pred = x_next
-        return y_pred
+        return x_next
 
     def accuracy(self, outputs, targets):
         y_predicts = np.argmax(outputs, axis=1)