Add doc string for each available model in fastai library.

The docs should help users to research details about the models.

fastai/torch_imports.py

@@ -31,17 +31,61 @@ def load_pre(pre, f, fn):
     if pre: load_model(m, f'{path}/weights/{fn}.pth')
     return m
 
-def inception_4(pre):
-    return children(load_pre(pre, InceptionV4, 'inceptionv4-97ef9c30'))[0]
+def _fastai_model(name, paper_title, paper_href):
+    def add_docs_wrapper(f):
+        f.__doc__ = f"""{name} model from
+        `"{paper_title}" <{paper_href}>`_
+
+        Args:
+           pre (bool): If True, returns a model pre-trained on ImageNet
+        """
+        return f
+    return add_docs_wrapper
+
+@_fastai_model('Inception 4', 'Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning',
+               'https://arxiv.org/pdf/1602.07261.pdf')
+def inception_4(pre): return children(load_pre(pre, InceptionV4, 'inceptionv4-97ef9c30'))[0]
+
+@_fastai_model('Inception 4', 'Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning',
+               'https://arxiv.org/pdf/1602.07261.pdf')
 def inceptionresnet_2(pre): return load_pre(pre, InceptionResnetV2, 'inceptionresnetv2-d579a627')
+
+@_fastai_model('ResNeXt 50', 'Aggregated Residual Transformations for Deep Neural Networks',
+               'https://arxiv.org/abs/1611.05431')
 def resnext50(pre): return load_pre(pre, resnext_50_32x4d, 'resnext_50_32x4d')
+
+@_fastai_model('ResNeXt 101_32', 'Aggregated Residual Transformations for Deep Neural Networks',
+               'https://arxiv.org/abs/1611.05431')
 def resnext101(pre): return load_pre(pre, resnext_101_32x4d, 'resnext_101_32x4d')
+
+@_fastai_model('ResNeXt 101_64', 'Aggregated Residual Transformations for Deep Neural Networks',
+               'https://arxiv.org/abs/1611.05431')
 def resnext101_64(pre): return load_pre(pre, resnext_101_64x4d, 'resnext_101_64x4d')
+
+@_fastai_model('Inception 4', 'Wide Residual Networks',
+               'https://arxiv.org/pdf/1605.07146.pdf')
 def wrn(pre): return load_pre(pre, wrn_50_2f, 'wrn_50_2f')
+
+@_fastai_model('Densenet-121', 'Densely Connected Convolutional Networks',
+               'https://arxiv.org/pdf/1608.06993.pdf')
 def dn121(pre): return children(densenet121(pre))[0]
+
+@_fastai_model('Densenet-169', 'Densely Connected Convolutional Networks',
+               'https://arxiv.org/pdf/1608.06993.pdf')
 def dn161(pre): return children(densenet161(pre))[0]
+
+@_fastai_model('Densenet-161', 'Densely Connected Convolutional Networks',
+               'https://arxiv.org/pdf/1608.06993.pdf')
 def dn169(pre): return children(densenet169(pre))[0]
+
+@_fastai_model('Densenet-201', 'Densely Connected Convolutional Networks',
+               'https://arxiv.org/pdf/1608.06993.pdf')
 def dn201(pre): return children(densenet201(pre))[0]
+
+@_fastai_model('Vgg-16 with batch norm added', 'Very Deep Convolutional Networks for Large-Scale Image Recognition',
+               'https://arxiv.org/pdf/1409.1556.pdf')
 def vgg16(pre): return children(vgg16_bn(pre))[0]
-def vgg19(pre): return children(vgg19_bn(pre))[0]
 
+@_fastai_model('Vgg-19 with batch norm added', 'Very Deep Convolutional Networks for Large-Scale Image Recognition',
+               'https://arxiv.org/pdf/1409.1556.pdf')
+def vgg19(pre): return children(vgg19_bn(pre))[0]