np.eye handling, np.dot and np.zeros fix, dataset update

Added np.eye for datatype forcing, Changed how the np.dot is being handled, fixed np.zeros issue, updated the datasets definitions

nnfs/__init__.py

@@ -1,3 +1,3 @@
-__version__ = '0.5.0'
+__version__ = '0.5.1'
 
 from .core import init

nnfs/core.py

@@ -7,8 +7,9 @@ def init(dot_precision_workaround=True, default_dtype='float32', random_seed=0):
 
     # Numpy methods to be replaced for a workaround
     methods_to_enclose = [
-        [np, 'zeros', False],
-        [np.random, 'randn', True],
+        [np, 'zeros', False, 1],
+        [np.random, 'randn', True, None],
+        [np, 'eye', False, 1],
     ]
 
     # https://github.com/numpy/numpy/issues/15591
@@ -21,7 +22,7 @@ def dot(*args, **kwargs):
             return orig_dot(*[a.astype('float64') for a in args], **kwargs).astype('float32')
         np.dot = dot
     else:
-        methods_to_enclose.append([np, 'dot', False])
+        methods_to_enclose.append([np, 'dot', True, None])
 
     # https://github.com/numpy/numpy/issues/6860
     # It is not possible to set default datatype with numpy
@@ -40,18 +41,18 @@ def dot(*args, **kwargs):
 def enclose(method, default_dtype):
 
     # Save a handler to original method
-    method.append(getattr(*method))
+    method.append(getattr(*method[:2]))
     def enclosed_method(*args, **kwargs):
 
         # If flag is True - use .astype()
         if method[2]:
-            return method[3](*args, **kwargs).astype(default_dtype)
+            return method[4](*args, **kwargs).astype(default_dtype)
 
         # Else pass dtype in kwargs
         else:
-            if 'dtype' not in kwargs:
+            if len(args) <= method[3] and 'dtype' not in kwargs:
                 kwargs['dtype'] = default_dtype
-            return method[3](*args, **kwargs)
+            return method[4](*args, **kwargs)
 
     # Replace numpy method with enclosed one
     setattr(*method[:2], enclosed_method)

nnfs/datasets/sine.py

@@ -2,7 +2,7 @@
 
 
 # Sine sample dataset
-def create_data(samples=10000):
+def create_data(samples=1000):
 
     X = np.arange(samples).reshape(-1, 1) / samples
     y = np.sin(2 * np.pi * X).reshape(-1, 1)

nnfs/datasets/spiral.py

@@ -4,13 +4,13 @@
 # Copyright (c) 2015 Andrej Karpathy
 # License: https://github.com/cs231n/cs231n.github.io/blob/master/LICENSE
 # Source: https://cs231n.github.io/neural-networks-case-study/
-def create_data(points, classes):
-    X = np.zeros((points*classes, 2))
-    y = np.zeros(points*classes, dtype='uint8')
+def create_data(samples, classes):
+    X = np.zeros((samples*classes, 2))
+    y = np.zeros(samples*classes, dtype='uint8')
     for class_number in range(classes):
-        ix = range(points*class_number, points*(class_number+1))
-        r = np.linspace(0.0, 1, points)
-        t = np.linspace(class_number*4, (class_number+1)*4, points) + np.random.randn(points)*0.2
+        ix = range(samples*class_number, samples*(class_number+1))
+        r = np.linspace(0.0, 1, samples)
+        t = np.linspace(class_number*4, (class_number+1)*4, samples) + np.random.randn(samples)*0.2
         X[ix] = np.c_[r*np.sin(t*2.5), r*np.cos(t*2.5)]
         y[ix] = class_number
     return X, y

nnfs/datasets/vertical.py

@@ -5,11 +5,11 @@
 # Copyright (c) 2015 Andrej Karpathy
 # License: https://github.com/cs231n/cs231n.github.io/blob/master/LICENSE
 # Source: https://cs231n.github.io/neural-networks-case-study/
-def create_data(points, classes):
-    X = np.zeros((points*classes, 2))
-    y = np.zeros(points*classes, dtype='uint8')
+def create_data(samples, classes):
+    X = np.zeros((samples*classes, 2))
+    y = np.zeros(samples*classes, dtype='uint8')
     for class_number in range(classes):
-        ix = range(points*class_number, points*(class_number+1))
-        X[ix] = np.c_[np.random.randn(points)*.1 + (class_number)/3, np.random.randn(points)*.1 + 0.5]
+        ix = range(samples*class_number, samples*(class_number+1))
+        X[ix] = np.c_[np.random.randn(samples)*.1 + (class_number)/3, np.random.randn(samples)*.1 + 0.5]
         y[ix] = class_number
     return X, y