backward incompatible change:

- nn modules now in `PyTorchAug.nn`, rather than in `PyTorchAug` directly

new functionality:
- anything in `nn` can now be used, without having to explicitly add to the program
- this also paves the way for being able to easily (and perhaps automatically) handle other types, like nnx etc

Author: hughperkins

setup.py

@@ -139,7 +139,7 @@ def get_file_datetime(filepath):
 
 setup(
     name='PyTorch',
-    version='2.10.0-SNAPSHOT',
+    version='3.0.0',
     author='Hugh Perkins',
     author_email='hughperkins@gmail.com',
     description=(

src/PyTorchAug.py

@@ -12,13 +12,6 @@ def getNextObjectId():
     nextObjectId += 1
     return res
 
-class Linear(object):
-    def __init__(self):
-        print('Linear.__init__')
-
-    def __attr__(self):
-        print('Linear.__attr__')
-
 def pushGlobal(lua, name1, name2=None, name3=None):
     lua.getGlobal(name1)
     if name2 is None:
@@ -88,6 +81,11 @@ def pushSomething(lua, something):
         pushObject(lua, something)
         return
 
+#    print('type(something)', type(something))
+#    print('str(type(something))', str(type(something)))
+#    if 'nn.' in str(type(something)):
+#        setupNnClass(
+
     raise Exception('pushing type ' + str(type(something)) + ' not implemented, value ', something)
 
 def popSomething(lua, self=None, name=None):
@@ -235,15 +233,42 @@ def __init__(self, *args, _fromLua=False, **kwargs):
     renamedClass = type(AnNnClass)(nnClassName, (AnNnClass,), {})
     return renamedClass
 
-
-luaClasses = {}
-nnClasses = [
-    'Linear', 'ClassNLLCriterion', 'MSECriterion', 'Sequential', 'LogSoftMax',
-    'Reshape', 'SpatialConvolutionMM', 'SpatialMaxPooling', 'ReLU', 'Tanh']
-for nnClassName in nnClasses:
+def setupNnClass(nnClassName):
     nnClass = loadNnClass(nnClassName)
     globals()[nnClassName] = nnClass
     luaClasses['nn.' + nnClassName] = nnClass
+    luaClassesReverse[nnClass] = 'nn.' + nnClassName
+    return nnClass
+
+#def mygetattr():
+#    print('mygetattr()')
+
+#def __getattr__():
+#    print('__getattr__')
+
+class Nn(object):
+    def __init__(self):
+        self.classes = {}
+
+    def __getattr__(self, name):
+#        print('Nn.__getattr__', name)
+        if name not in self.classes:
+            self.classes[name] = setupNnClass(name)
+        thisClass = self.classes[name]
+#        print('thisClass', thisClass)
+        return thisClass
+
+nn = Nn()
+
+luaClasses = {}
+#nnClasses = [
+#    'Linear', 'ClassNLLCriterion', 'MSECriterion', 'Sequential', 'LogSoftMax',
+#    'Reshape', 'SpatialConvolutionMM', 'SpatialMaxPooling', 'ReLU', 'Tanh']
+#for nnClassName in nnClasses:
+#    setupNnClass(nnClassName)
+#    nnClass = loadNnClass(nnClassName)
+#    globals()[nnClassName] = nnClass
+#    luaClasses['nn.' + nnClassName] = nnClass
 
 
 luaClassesReverse = {}
@@ -252,7 +277,7 @@ def populateLuaClassesReverse():
     for name in luaClasses:
         classtype = luaClasses[name]
         luaClassesReverse[classtype] = name
-populateLuaClassesReverse()
+#populateLuaClassesReverse()
 
 cythonClasses = {}
 cythonClasses['torch.FloatTensor'] = {'popFunction': PyTorch._popFloatTensor}

test/test_pynn.py

@@ -1,11 +1,11 @@
 from __future__ import print_function, division
 import PyTorch
 
-from PyTorchAug import *
+from PyTorchAug import nn
 
 def test_pynn():
     PyTorch.manualSeed(123)
-    linear = Linear(3, 5)
+    linear = nn.Linear(3, 5)
     linear
     print('linear', linear)
     print('linear.weight', linear.weight)
@@ -20,12 +20,12 @@ def test_pynn():
     gradInput = linear.updateGradInput(input, output)
     print('gradInput', gradInput)
 
-    criterion = ClassNLLCriterion()
+    criterion = nn.ClassNLLCriterion()
     print('criterion', criterion)
 
     print('dir(linear)', dir(linear))
 
-    mlp = Sequential()
+    mlp = nn.Sequential()
     mlp.add(linear)
 
     output = mlp.forward(input)
@@ -39,23 +39,23 @@ def test_pynn():
 
     numpy.random.seed(123)
 
-    mlp = Sequential()
+    mlp = nn.Sequential()
 
-    mlp.add(SpatialConvolutionMM(1, 16, 5, 5, 1, 1, 2, 2))
-    mlp.add(ReLU())
-    mlp.add(SpatialMaxPooling(3, 3, 3, 3))
+    mlp.add(nn.SpatialConvolutionMM(1, 16, 5, 5, 1, 1, 2, 2))
+    mlp.add(nn.ReLU())
+    mlp.add(nn.SpatialMaxPooling(3, 3, 3, 3))
 
-    mlp.add(SpatialConvolutionMM(16, 32, 3, 3, 1, 1, 1, 1))
-    mlp.add(ReLU())
-    mlp.add(SpatialMaxPooling(2, 2, 2, 2))
+    mlp.add(nn.SpatialConvolutionMM(16, 32, 3, 3, 1, 1, 1, 1))
+    mlp.add(nn.ReLU())
+    mlp.add(nn.SpatialMaxPooling(2, 2, 2, 2))
 
-    mlp.add(Reshape(32 * 4 * 4))
-    mlp.add(Linear(32 * 4 * 4, 150))
-    mlp.add(Tanh())
-    mlp.add(Linear(150, 10))
-    mlp.add(LogSoftMax())
+    mlp.add(nn.Reshape(32 * 4 * 4))
+    mlp.add(nn.Linear(32 * 4 * 4, 150))
+    mlp.add(nn.Tanh())
+    mlp.add(nn.Linear(150, 10))
+    mlp.add(nn.LogSoftMax())
 
-    criterion = ClassNLLCriterion()
+    criterion = nn.ClassNLLCriterion()
     print('got criterion')
 
     learningRate = 0.02