Cleanup imports and move examples to new project

LICENSE

@@ -1,4 +1,4 @@
-Copyright (c) 2016, Huw Campbell
+Copyright (c) 2016-2017, Huw Campbell
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

README.md

@@ -20,8 +20,12 @@ specified and initialised with random weights in a few lines of code with
 ```haskell
 type MNIST
   = Network
-    '[ Convolution 1 10 5 5 1 1, Pooling 2 2 2 2, Relu, Convolution 10 16 5 5 1 1, Pooling 2 2 2 2, FlattenLayer, Relu, FullyConnected 256 80, Logit, FullyConnected 80 10, Logit]
+    '[ Convolution 1 10 5 5 1 1, Pooling 2 2 2 2, Relu
-    '[ 'D2 28 28, 'D3 24 24 10, 'D3 12 12 10, 'D3 12 12 10, 'D3 8 8 16, 'D3 4 4 16, 'D1 256, 'D1 256, 'D1 80, 'D1 80, 'D1 10, 'D1 10]
+     , Convolution 10 16 5 5 1 1, Pooling 2 2 2 2, FlattenLayer, Relu
+     , FullyConnected 256 80, Logit, FullyConnected 80 10, Logit]
+    '[ 'D2 28 28, 'D3 24 24 10, 'D3 12 12 10, 'D3 12 12 10
+     , 'D3 8 8 16, 'D3 4 4 16, 'D1 256, 'D1 256
+     , 'D1 80, 'D1 80, 'D1 10, 'D1 10]
 
 randomMnist :: MonadRandom m => m MNIST
 randomMnist = randomNetwork

examples/LICENSE

@@ -0,0 +1,10 @@
+Copyright (c) 2016-2017, Huw Campbell
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+
+1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

examples/grenade-examples.cabal

@@ -0,0 +1,108 @@
+name:                  grenade-examples
+version:               0.0.1
+license:               BSD2
+license-file:          LICENSE
+author:                Huw Campbell <huw.campbell@gmail.com>
+maintainer:            Huw Campbell <huw.campbell@gmail.com>
+copyright:             (c) 2016-2017 Huw Campbell.
+synopsis:              grenade-examples
+category:              System
+cabal-version:         >= 1.8
+build-type:            Simple
+description:           grenade-examples
+
+source-repository head
+  type: git
+  location: https://github.com/HuwCampbell/grenade.git
+
+library
+
+executable             feedforward
+    ghc-options:       -Wall -threaded -O2
+    main-is:           main/feedforward.hs
+    build-depends:     base
+                     , grenade
+                     , attoparsec
+                     , bytestring
+                     , cereal
+                     , either
+                     , optparse-applicative            == 0.13.*
+                     , text                            == 1.2.*
+                     , mtl                             >= 2.2.1      && < 2.3
+                     , hmatrix
+                     , transformers
+                     , singletons
+                     , semigroups
+                     , MonadRandom
+
+executable             mnist
+    ghc-options:       -Wall -threaded -O2
+    main-is:           main/mnist.hs
+    build-depends:     base
+                     , grenade
+                     , attoparsec
+                     , either
+                     , optparse-applicative            == 0.13.*
+                     , text                            == 1.2.*
+                     , mtl                             >= 2.2.1      && < 2.3
+                     , hmatrix                         >= 0.18       && < 0.19
+                     , transformers
+                     , semigroups
+                     , singletons
+                     , MonadRandom
+                     , vector
+
+executable             gan-mnist
+    ghc-options:       -Wall -threaded -O2
+    main-is:           main/gan-mnist.hs
+    build-depends:     base
+                     , grenade
+                     , attoparsec
+                     , bytestring
+                     , cereal
+                     , either
+                     , optparse-applicative            == 0.13.*
+                     , text                            == 1.2.*
+                     , mtl                             >= 2.2.1      && < 2.3
+                     , hmatrix                         >= 0.18       && < 0.19
+                     , transformers
+                     , semigroups
+                     , singletons
+                     , MonadRandom
+                     , vector
+
+executable             recurrent
+    ghc-options:       -Wall -threaded -O2
+    main-is:           main/recurrent.hs
+    build-depends:     base
+                     , grenade
+                     , attoparsec
+                     , either
+                     , optparse-applicative            == 0.13.*
+                     , text                            == 1.2.*
+                     , mtl                             >= 2.2.1      && < 2.3
+                     , hmatrix                         >= 0.18       && < 0.19
+                     , transformers
+                     , semigroups
+                     , singletons
+                     , MonadRandom
+
+executable             shakespeare
+    ghc-options:       -Wall -threaded -O2
+    main-is:           main/shakespeare.hs
+    build-depends:     base
+                     , grenade
+                     , attoparsec
+                     , bytestring
+                     , cereal
+                     , either
+                     , optparse-applicative            == 0.13.*
+                     , text                            == 1.2.*
+                     , mtl                             >= 2.2.1      && < 2.3
+                     , hmatrix                         >= 0.18       && < 0.19
+                     , transformers
+                     , semigroups
+                     , singletons
+                     , vector
+                     , MonadRandom
+                     , containers

examples/main/gan-mnist.hs

@@ -77,9 +77,10 @@ trainExample rate discriminator generator realExample noiseSource
        (discriminatorTapeFake, guessFake) = runNetwork discriminator fakeExample
 
        (discriminator'real, _)            = runGradient discriminator discriminatorTapeReal ( guessReal - 1 )
-       (discriminator'fake, push)         = runGradient discriminator discriminatorTapeFake guessFake
+       (discriminator'fake, _)            = runGradient discriminator discriminatorTapeFake guessFake
+       (_, push)                          = runGradient discriminator discriminatorTapeFake ( guessFake - 1)
 
-       (generator', _)                    = runGradient generator generatorTape (-push)
+       (generator', _)                    = runGradient generator generatorTape push
 
        newDiscriminator                   = foldl' (applyUpdate rate { learningRegulariser = learningRegulariser rate * 10}) discriminator [ discriminator'real, discriminator'fake ]
        newGenerator                       = applyUpdate rate generator generator'

examples/main/mnist.hs

@@ -34,8 +34,24 @@ import           Grenade.Utils.OneHot
 
 -- With the mnist data from Kaggle normalised to doubles between 0 and 1, learning rate of 0.01 and 15 iterations,
 -- this network should get down to about a 1.3% error rate.
-type MNIST = Network '[ Convolution 1 10 5 5 1 1, Pooling 2 2 2 2, Relu, Convolution 10 16 5 5 1 1, Pooling 2 2 2 2, Reshape, Relu, FullyConnected 256 80, Logit, FullyConnected 80 10, Logit]
+--
-                     '[ 'D2 28 28, 'D3 24 24 10, 'D3 12 12 10, 'D3 12 12 10, 'D3 8 8 16, 'D3 4 4 16, 'D1 256, 'D1 256, 'D1 80, 'D1 80, 'D1 10, 'D1 10]
+-- /NOTE:/ This model is actually too complex for MNIST, and one should use the type given in the readme instead.
+--         This one is just here to demonstrate Inception layers in use.
+--
+type MNIST =
+  Network
+    '[ Reshape
+     , Inception 28 28 1 5 5 5, Pooling 2 2 2 2, Relu
+     , Inception 14 14 15 5 5 5, Pooling 2 2 2 2, Relu
+     , Reshape
+     , FullyConnected 735 80, Logit
+     , FullyConnected 80 10, Logit]
+    '[ 'D2 28 28, 'D3 28 28 1
+     , 'D3 28 28 15, 'D3 14 14 15, 'D3 14 14 15
+     , 'D3 14 14 15, 'D3 7 7 15, 'D3 7 7 15
+     , 'D1 735
+     , 'D1 80, 'D1 80
+     , 'D1 10, 'D1 10]
 
 randomMnist :: MonadRandom m => m MNIST
 randomMnist = randomNetwork

grenade.cabal

@@ -4,7 +4,7 @@ license:               BSD2
 license-file:          LICENSE
 author:                Huw Campbell <huw.campbell@gmail.com>
 maintainer:            Huw Campbell <huw.campbell@gmail.com>
-copyright:             (c) 2015 Huw Campbell.
+copyright:             (c) 2016-2017 Huw Campbell.
 synopsis:              grenade
 category:              System
 cabal-version:         >= 1.8
@@ -27,25 +27,26 @@ library
   build-depends:
                        base                            >= 4.8         && < 5
                      , bytestring                      == 0.10.*
-                     , containers
+                     , containers                      >= 0.5         && < 0.6
-                     , deepseq
+                     , cereal                          >= 0.5         && < 0.6
-                     , either                          == 4.4.*
+                     , deepseq                         >= 1.4         && < 1.5
-                     , cereal
                      , exceptions                      == 0.8.*
                      , hmatrix                         == 0.18.*
-                     , MonadRandom
+                     , MonadRandom                     >= 0.4         && < 0.6
                      , mtl                             >= 2.2.1       && < 2.3
-                     , primitive
+                     , primitive                       >= 0.6         && < 0.7
                      , text                            == 1.2.*
-                     , transformers
                      , singletons                      >= 2.1         && < 2.3
-                     , vector                          == 0.11.*
+                     , vector                          >= 0.11        && < 0.13
 
   ghc-options:
                        -Wall
   hs-source-dirs:
                        src
 
+  if impl(ghc < 8.0)
+    ghc-options:       -fno-warn-incomplete-patterns
+
 
   exposed-modules:
                        Grenade
@@ -55,19 +56,24 @@ library
                        Grenade.Core.Network
                        Grenade.Core.Runner
                        Grenade.Core.Shape
-                       Grenade.Layers.Crop
+
+                       Grenade.Layers
                        Grenade.Layers.Concat
                        Grenade.Layers.Convolution
+                       Grenade.Layers.Crop
                        Grenade.Layers.Dropout
+                       Grenade.Layers.Elu
                        Grenade.Layers.FullyConnected
-                       Grenade.Layers.Reshape
+                       Grenade.Layers.Inception
                        Grenade.Layers.Logit
                        Grenade.Layers.Merge
-                       Grenade.Layers.Relu
-                       Grenade.Layers.Elu
-                       Grenade.Layers.Tanh
                        Grenade.Layers.Pad
                        Grenade.Layers.Pooling
+                       Grenade.Layers.Relu
+                       Grenade.Layers.Reshape
+                       Grenade.Layers.Softmax
+                       Grenade.Layers.Tanh
+                       Grenade.Layers.Trivial
 
                        Grenade.Layers.Internal.Convolution
                        Grenade.Layers.Internal.Pad
@@ -81,6 +87,7 @@ library
                        Grenade.Recurrent.Core.Network
                        Grenade.Recurrent.Core.Runner
 
+                       Grenade.Recurrent.Layers
                        Grenade.Recurrent.Layers.BasicRecurrent
                        Grenade.Recurrent.Layers.LSTM
 
@@ -95,98 +102,6 @@ library
 
   cc-options:          -std=c99 -O3 -msse4.2 -Wall -Werror -DCABAL=1
 
-executable             feedforward
-    ghc-options:       -Wall -threaded -O2
-    main-is:           main/feedforward.hs
-    build-depends:     base
-                     , grenade
-                     , attoparsec
-                     , bytestring
-                     , cereal
-                     , either
-                     , optparse-applicative            == 0.13.*
-                     , text                            == 1.2.*
-                     , mtl                             >= 2.2.1      && < 2.3
-                     , hmatrix
-                     , transformers
-                     , singletons
-                     , semigroups
-                     , MonadRandom
-
-executable             mnist
-    ghc-options:       -Wall -threaded -O2
-    main-is:           main/mnist.hs
-    build-depends:     base
-                     , grenade
-                     , attoparsec
-                     , either
-                     , optparse-applicative            == 0.13.*
-                     , text                            == 1.2.*
-                     , mtl                             >= 2.2.1      && < 2.3
-                     , hmatrix                         >= 0.18       && < 0.19
-                     , transformers
-                     , semigroups
-                     , singletons
-                     , MonadRandom
-                     , vector
-
-executable             gan-mnist
-    ghc-options:       -Wall -threaded -O2
-    main-is:           main/gan-mnist.hs
-    build-depends:     base
-                     , grenade
-                     , attoparsec
-                     , bytestring
-                     , cereal
-                     , either
-                     , optparse-applicative            == 0.13.*
-                     , text                            == 1.2.*
-                     , mtl                             >= 2.2.1      && < 2.3
-                     , hmatrix                         >= 0.18       && < 0.19
-                     , transformers
-                     , semigroups
-                     , singletons
-                     , MonadRandom
-                     , vector
-
-executable             recurrent
-    ghc-options:       -Wall -threaded -O2
-    main-is:           main/recurrent.hs
-    build-depends:     base
-                     , grenade
-                     , attoparsec
-                     , either
-                     , optparse-applicative            == 0.13.*
-                     , text                            == 1.2.*
-                     , mtl                             >= 2.2.1      && < 2.3
-                     , hmatrix                         >= 0.18       && < 0.19
-                     , transformers
-                     , semigroups
-                     , singletons
-                     , MonadRandom
-
-
-executable             shakespeare
-    ghc-options:       -Wall -threaded -O2
-    main-is:           main/shakespeare.hs
-    build-depends:     base
-                     , grenade
-                     , attoparsec
-                     , bytestring
-                     , cereal
-                     , either
-                     , optparse-applicative            == 0.13.*
-                     , text                            == 1.2.*
-                     , mtl                             >= 2.2.1      && < 2.3
-                     , hmatrix                         >= 0.18       && < 0.19
-                     , transformers
-                     , semigroups
-                     , singletons
-                     , vector
-                     , MonadRandom
-                     , containers
-
-
 test-suite test
   type:                exitcode-stdio-1.0
 

src/Grenade.hs

@@ -1,23 +1,52 @@
 module Grenade (
-    module X
+  -- | This is an empty module which simply re-exports public definitions
+  --   for machine learning with Grenade.
+
+  -- * Exported modules
+  --
+  -- | The core types and runners for Grenade.
+    module Grenade.Core
+
+  -- | The neural network layer zoo
+  , module Grenade.Layers
+
+
+    -- * Overview of the library
+    -- $library
+
+    -- * Example usage
+    -- $example
+
   ) where
 
-import           Grenade.Core.LearningParameters as X
+import           Grenade.Core
-import           Grenade.Core.Layer as X
+import           Grenade.Layers
-import           Grenade.Core.Network as X
+
-import           Grenade.Core.Runner  as X
+{- $library
-import           Grenade.Core.Shape  as X
+Grenade is a purely functional deep learning library.
-import           Grenade.Layers.Concat as X
+
-import           Grenade.Layers.Crop as X
+It provides an expressive type level API for the construction
-import           Grenade.Layers.Dropout as X
+of complex neural network architectures. Backing this API is and
-import           Grenade.Layers.Pad as X
+implementation written using BLAS and LAPACK, mostly provided by
-import           Grenade.Layers.Pooling as X
+the hmatrix library.
-import           Grenade.Layers.Reshape as X
+-}
-import           Grenade.Layers.FullyConnected as X
+
-import           Grenade.Layers.Logit as X
+{- $example
-import           Grenade.Layers.Merge as X
+A few examples are provided at https://github.com/HuwCampbell/grenade
-import           Grenade.Layers.Convolution as X
+under the examples folder.
-import           Grenade.Layers.Relu as X
+
-import           Grenade.Layers.Elu as X
+The starting place is to write your neural network type and a
-import           Grenade.Layers.Tanh as X
+function to create a random layer of that type. The following
-import           Grenade.Layers.Softmax as X
+is a simple example which runs a logistic regression.
+
+> type MyNet = Network '[ FullyConnected 10 1, Logit ] '[ 'D1 10, 'D1 1, 'D1 1 ]
+>
+> randomMyNet :: MonadRandom MyNet
+> randomMyNet = randomNetwork
+
+The function `randomMyNet` witnesses the `CreatableNetwork`
+constraint of the neural network, that is it ensures the network
+can be built, and hence, that the architecture is sound.
+-}
+
+

src/Grenade/Core.hs

@@ -1,8 +1,13 @@
 module Grenade.Core (
-    module X
+    module Grenade.Core.Layer
+  , module Grenade.Core.LearningParameters
+  , module Grenade.Core.Network
+  , module Grenade.Core.Runner
+  , module Grenade.Core.Shape
   ) where
 
-import           Grenade.Core.Layer as X
+import           Grenade.Core.Layer
-import           Grenade.Core.LearningParameters as X
+import           Grenade.Core.LearningParameters
-import           Grenade.Core.Shape as X
+import           Grenade.Core.Network
-import           Grenade.Core.Network as X
+import           Grenade.Core.Runner
+import           Grenade.Core.Shape

src/Grenade/Core/LearningParameters.hs

@@ -1,4 +1,8 @@
 module Grenade.Core.LearningParameters (
+  -- | This module contains learning algorithm specific
+  --   code. Currently, this module should be consifered
+  --   unstable, due to issue #26.
+
     LearningParameters (..)
   ) where
 

src/Grenade/Core/Network.hs

@@ -1,4 +1,3 @@
-{-# LANGUAGE CPP                   #-}
 {-# LANGUAGE DataKinds             #-}
 {-# LANGUAGE BangPatterns          #-}
 {-# LANGUAGE GADTs                 #-}
@@ -19,10 +18,6 @@ This module defines the core data types and functions
 for non-recurrent neural networks.
 -}
 
-#if __GLASGOW_HASKELL__ < 800
-{-# OPTIONS_GHC -fno-warn-incomplete-patterns #-}
-#endif
-
 module Grenade.Core.Network (
     Network (..)
   , Gradients (..)
@@ -47,9 +42,9 @@ import           Grenade.Core.Shape
 
 -- | Type of a network.
 --
---   The [*] type specifies the types of the layers.
+--   The @[*]@ type specifies the types of the layers.
 --
---   The [Shape] type specifies the shapes of data passed between the layers.
+--   The @[Shape]@ type specifies the shapes of data passed between the layers.
 --
 --   Can be considered to be a heterogeneous list of layers which are able to
 --   transform the data shapes of the network.

src/Grenade/Core/Shape.hs

@@ -1,4 +1,3 @@
-{-# LANGUAGE CPP                   #-}
 {-# LANGUAGE DataKinds             #-}
 {-# LANGUAGE GADTs                 #-}
 {-# LANGUAGE KindSignatures        #-}
@@ -8,11 +7,6 @@
 {-# LANGUAGE FlexibleContexts      #-}
 {-# LANGUAGE ScopedTypeVariables   #-}
 {-# LANGUAGE RankNTypes            #-}
-
--- Ghc 7.10 fails to recognise n2 is complete.
-#if __GLASGOW_HASKELL__ < 800
-{-# OPTIONS_GHC -fno-warn-incomplete-patterns #-}
-#endif
 {-|
 Module      : Grenade.Core.Shape
 Description : Core definition of the Shapes of data we understand
@@ -65,17 +59,14 @@ data Shape
 --   All shapes are held in contiguous memory.
 --   3D is held in a matrix (usually row oriented) which has height depth * rows.
 data S (n :: Shape) where
-  -- | One dimensional data
   S1D :: ( KnownNat len )
       => R len
       -> S ('D1 len)
 
-  -- | Two dimensional data
   S2D :: ( KnownNat rows, KnownNat columns )
       => L rows columns
       -> S ('D2 rows columns)
 
-  -- | Three dimensional data
   S3D :: ( KnownNat rows
          , KnownNat columns
          , KnownNat depth

src/Grenade/Layers.hs

@@ -0,0 +1,33 @@
+module Grenade.Layers (
+    module Grenade.Layers.Concat
+  , module Grenade.Layers.Convolution
+  , module Grenade.Layers.Crop
+  , module Grenade.Layers.Elu
+  , module Grenade.Layers.FullyConnected
+  , module Grenade.Layers.Inception
+  , module Grenade.Layers.Logit
+  , module Grenade.Layers.Merge
+  , module Grenade.Layers.Pad
+  , module Grenade.Layers.Pooling
+  , module Grenade.Layers.Reshape
+  , module Grenade.Layers.Relu
+  , module Grenade.Layers.Softmax
+  , module Grenade.Layers.Tanh
+  , module Grenade.Layers.Trivial
+  ) where
+
+import           Grenade.Layers.Concat
+import           Grenade.Layers.Convolution
+import           Grenade.Layers.Crop
+import           Grenade.Layers.Elu
+import           Grenade.Layers.Pad
+import           Grenade.Layers.FullyConnected
+import           Grenade.Layers.Inception
+import           Grenade.Layers.Logit
+import           Grenade.Layers.Merge
+import           Grenade.Layers.Pooling
+import           Grenade.Layers.Reshape
+import           Grenade.Layers.Relu
+import           Grenade.Layers.Softmax
+import           Grenade.Layers.Tanh
+import           Grenade.Layers.Trivial

src/Grenade/Layers/Concat.hs

@@ -9,11 +9,13 @@
 {-# LANGUAGE ScopedTypeVariables   #-}
 {-# LANGUAGE StandaloneDeriving    #-}
 {-|
-Module      : Grenade.Core.Network
+Module      : Grenade.Layers.Concat
-Description : Core definition a simple neural etwork
+Description : Concatenation layer
 Copyright   : (c) Huw Campbell, 2016-2017
 License     : BSD2
 Stability   : experimental
+
+This module provides the concatenation layer, whic used to run two separate layers in parallel and combine their outputs.
 -}
 module Grenade.Layers.Concat (
     Concat (..)

src/Grenade/Layers/Inception.hs

@@ -9,10 +9,13 @@
 {-# LANGUAGE ScopedTypeVariables   #-}
 {-|
 Module      : Grenade.Core.Network
-Description : Core definition a simple neural etwork
+Description : Inception style parallel convolutional network composition.
 Copyright   : (c) Huw Campbell, 2016-2017
 License     : BSD2
 Stability   : experimental
+
+Export an Inception style type, which can be used to build up
+complex multiconvolution size networks.
 -}
 module Grenade.Layers.Inception (
     Inception
@@ -25,24 +28,35 @@ import           Grenade.Layers.Convolution
 import           Grenade.Layers.Pad
 import           Grenade.Layers.Concat
 
-
+-- | Type of an inception layer.
+--
+--   It looks like a bit of a handful, but is actually pretty easy to use.
+--
+--   The first three type parameters are the size of the (3D) data the
+--   inception layer will take. It will emit 3D data with the number of
+--   channels being the sum of @chx@, @chy@, @chz@, which are the number
+--   of convolution filters in the 3x3, 5x5, and 7x7 convolutions Layers
+--   respectively.
+--
+--   The network get padded effectively before each convolution filters
+--   such that the output dimension is the same x and y as the input.
 type Inception rows cols channels chx chy chz
-  = Network '[ Concat ('D3 (rows - 2) (cols - 2) (chx + chy)) (InceptionS rows cols channels chx chy) ('D3 (rows - 2) (cols - 2) chz) (Inception7x7 rows cols channels chz) ]
+  = Network '[ Concat ('D3 rows cols (chx + chy)) (InceptionS rows cols channels chx chy) ('D3 rows cols chz) (Inception7x7 rows cols channels chz) ]
-            '[ 'D3 rows cols channels, 'D3 (rows -2) (cols -2) (chx + chy + chz) ]
+            '[ 'D3 rows cols channels, 'D3 rows cols (chx + chy + chz) ]
 
 type InceptionS rows cols channels chx chy
-  = Network '[ Concat ('D3 (rows - 2) (cols - 2) chx) (Inception3x3 rows cols channels chx) ('D3 (rows - 2) (cols - 2) chy) (Inception5x5 rows cols channels chy) ]
+  = Network '[ Concat ('D3 rows cols chx) (Inception3x3 rows cols channels chx) ('D3 rows cols chy) (Inception5x5 rows cols channels chy) ]
-            '[ 'D3 rows cols channels, 'D3 (rows -2) (cols -2) (chx + chy) ]
+            '[ 'D3 rows cols channels, 'D3 rows cols (chx + chy) ]
 
 type Inception3x3 rows cols channels chx
-  = Network '[ Convolution channels chx 3 3 1 1 ]
+  = Network '[ Pad 1 1 1 1, Convolution channels chx 3 3 1 1 ]
-            '[ 'D3 rows cols channels, 'D3 (rows -2) (cols -2) chx ]
+            '[ 'D3 rows cols channels, 'D3 (rows + 2) (cols + 2) channels, 'D3 rows cols chx ]
 
 type Inception5x5 rows cols channels chx
-  = Network '[ Pad 1 1 1 1, Convolution channels chx 5 5 1 1 ]
+  = Network '[ Pad 2 2 2 2, Convolution channels chx 5 5 1 1 ]
-            '[ 'D3 rows cols channels, 'D3 (rows + 2) (cols + 2) channels, 'D3 (rows - 2) (cols - 2) chx ]
+            '[ 'D3 rows cols channels, 'D3 (rows + 4) (cols + 4) channels, 'D3 rows cols chx ]
 
 type Inception7x7 rows cols channels chx
-  = Network '[ Pad 2 2 2 2, Convolution channels chx 7 7 1 1 ]
+  = Network '[ Pad 3 3 3 3, Convolution channels chx 7 7 1 1 ]
-            '[ 'D3 rows cols channels, 'D3 (rows + 4) (cols + 4) channels, 'D3 (rows - 2) (cols - 2) chx ]
+            '[ 'D3 rows cols channels, 'D3 (rows + 6) (cols + 6) channels, 'D3 rows cols chx ]
 

src/Grenade/Layers/Trivial.hs

@@ -9,7 +9,7 @@ module Grenade.Layers.Trivial (
 
 import           Data.Serialize
 
-import           Grenade.Core.Network
+import           Grenade.Core
 
 -- | A trivial layer.
 data Trivial = Trivial
@@ -25,5 +25,6 @@ instance UpdateLayer Trivial where
   createRandom = return Trivial
 
 instance (a ~ b) => Layer Trivial a b where
-  runForwards _ = id
+  type Tape Trivial a b = ()
+  runForwards _ a = ((), a)
   runBackwards _ _ y = ((), y)

src/Grenade/Recurrent.hs

@@ -1,9 +1,7 @@
 module Grenade.Recurrent (
-    module X
+    module Grenade.Recurrent.Core
+  , module Grenade.Recurrent.Layers
   ) where
 
-import           Grenade.Recurrent.Core.Layer as X
+import           Grenade.Recurrent.Core
-import           Grenade.Recurrent.Core.Network as X
+import           Grenade.Recurrent.Layers
-import           Grenade.Recurrent.Core.Runner  as X
-import           Grenade.Recurrent.Layers.BasicRecurrent as X
-import           Grenade.Recurrent.Layers.LSTM as X

src/Grenade/Recurrent/Core.hs

@@ -1,6 +1,9 @@
 module Grenade.Recurrent.Core (
-    module X
+    module Grenade.Recurrent.Core.Layer
+  , module Grenade.Recurrent.Core.Network
+  , module Grenade.Recurrent.Core.Runner
   ) where
 
-import           Grenade.Recurrent.Core.Layer as X
+import           Grenade.Recurrent.Core.Layer
-import           Grenade.Recurrent.Core.Network as X
+import           Grenade.Recurrent.Core.Network
+import           Grenade.Recurrent.Core.Runner

src/Grenade/Recurrent/Core/Network.hs

@@ -1,4 +1,3 @@
-{-# LANGUAGE CPP                   #-}
 {-# LANGUAGE DataKinds             #-}
 {-# LANGUAGE GADTs                 #-}
 {-# LANGUAGE TypeOperators         #-}
@@ -10,10 +9,7 @@
 {-# LANGUAGE RankNTypes            #-}
 {-# LANGUAGE BangPatterns          #-}
 {-# LANGUAGE ScopedTypeVariables   #-}
-
+{-# LANGUAGE UndecidableInstances  #-}
-#if __GLASGOW_HASKELL__ < 800
-{-# OPTIONS_GHC -fno-warn-incomplete-patterns #-}
-#endif
 
 module Grenade.Recurrent.Core.Network (
     Recurrent

src/Grenade/Recurrent/Core/Runner.hs

@@ -3,16 +3,11 @@
 {-# LANGUAGE DataKinds             #-}
 {-# LANGUAGE ScopedTypeVariables   #-}
 {-# LANGUAGE TypeOperators         #-}
-{-# LANGUAGE CPP                   #-}
 {-# LANGUAGE TypeFamilies          #-}
 {-# LANGUAGE FlexibleContexts      #-}
 {-# LANGUAGE RankNTypes            #-}
 {-# LANGUAGE RecordWildCards       #-}
 
-#if __GLASGOW_HASKELL__ < 800
-{-# OPTIONS_GHC -fno-warn-incomplete-patterns #-}
-#endif
-
 module Grenade.Recurrent.Core.Runner (
     trainRecurrent
   , runRecurrent

src/Grenade/Recurrent/Layers.hs

@@ -0,0 +1,7 @@
+module Grenade.Recurrent.Layers (
+    module Grenade.Recurrent.Layers.BasicRecurrent
+  , module Grenade.Recurrent.Layers.LSTM
+  ) where
+
+import           Grenade.Recurrent.Layers.BasicRecurrent
+import           Grenade.Recurrent.Layers.LSTM

src/Grenade/Recurrent/Layers/BasicRecurrent.hs

@@ -1,4 +1,3 @@
-{-# LANGUAGE CPP                   #-}
 {-# LANGUAGE DataKinds             #-}
 {-# LANGUAGE GADTs                 #-}
 {-# LANGUAGE RecordWildCards       #-}
@@ -8,10 +7,6 @@
 {-# LANGUAGE FlexibleContexts      #-}
 {-# LANGUAGE UndecidableInstances  #-}
 
--- GHC 7.10 doesn't see recurrent run functions as total.
-#if __GLASGOW_HASKELL__ < 800
-{-# OPTIONS_GHC -fno-warn-incomplete-patterns #-}
-#endif
 module Grenade.Recurrent.Layers.BasicRecurrent (
     BasicRecurrent (..)
   , randomBasicRecurrent

src/Grenade/Recurrent/Layers/LSTM.hs

@@ -1,5 +1,4 @@
 {-# LANGUAGE BangPatterns          #-}
-{-# LANGUAGE CPP                   #-}
 {-# LANGUAGE DataKinds             #-}
 {-# LANGUAGE GADTs                 #-}
 {-# LANGUAGE RankNTypes            #-}
@@ -11,11 +10,6 @@
 {-# LANGUAGE ViewPatterns          #-}
 {-# LANGUAGE ScopedTypeVariables   #-}
 
--- GHC 7.10 doesn't see recurrent run functions as total.
-#if __GLASGOW_HASKELL__ < 800
-{-# OPTIONS_GHC -fno-warn-incomplete-patterns #-}
-#endif
-
 module Grenade.Recurrent.Layers.LSTM (
     LSTM (..)
   , LSTMWeights (..)

test/Test/Grenade/Layers/PadCrop.hs

@@ -30,6 +30,15 @@ prop_pad_crop =
           (_    , grad) = runBackwards net tapes d
       in  d ~~~ res .&&. grad ~~~ d
 
+prop_pad_crop_2d :: Property
+prop_pad_crop_2d =
+  let net :: Network '[Pad 2 3 4 6, Crop 2 3 4 6] '[ 'D2 7 9, 'D2 16 15, 'D2 7 9 ]
+      net = Pad :~> Crop :~> NNil
+  in  gamble genOfShape $ \(d :: S ('D2 7 9)) ->
+      let (tapes, res)  = runForwards  net d
+          (_    , grad) = runBackwards net tapes d
+      in  d ~~~ res .&&. grad ~~~ d
+
 (~~~) :: S x -> S x -> Bool
 (S1D x) ~~~ (S1D y) = norm_Inf (x - y) < 0.00001
 (S2D x) ~~~ (S2D y) = norm_Inf (x - y) < 0.00001