Load weights and biases from file and test the accuracy of the pre-trained model

Author: milancurcic

src/nf/nf_network_submodule.f90

@@ -4,6 +4,7 @@
   use nf_flatten_layer, only: flatten_layer
   use nf_input1d_layer, only: input1d_layer
   use nf_input3d_layer, only: input3d_layer
+  use nf_io_hdf5, only: get_hdf5_dataset
   use nf_keras, only: get_keras_h5_layers, keras_layer
   use nf_layer, only: layer
   use nf_layer_constructors, only: dense, input
@@ -55,6 +56,8 @@ module function network_from_keras(filename) result(res)
     type(network) :: res
     type(keras_layer), allocatable :: keras_layers(:)
     type(layer), allocatable :: layers(:)
+    character(:), allocatable :: layer_name
+    character(:), allocatable :: object_name
     integer :: n
 
     keras_layers = get_keras_h5_layers(filename)
@@ -89,7 +92,44 @@ module function network_from_keras(filename) result(res)
 
     res = network(layers)
 
-    !TODO read weights and biases from Keras file and set here
+    ! Loop over layers and read weights and biases from the Keras h5 file
+    ! for each; currently only dense layers are implemented.
+    do n = 2, size(res % layers)
+
+      layer_name = keras_layers(n) % name
+
+      if (keras_layers(n) % class == 'Dense') then
+        select type(this_layer => res % layers(n) % p)
+
+          type is(dense_layer)
+
+            ! Read biases from file
+            object_name = '/model_weights/' // layer_name // '/' &
+              // layer_name // '/bias:0'
+            call get_hdf5_dataset(filename, object_name, this_layer % biases)
+
+            ! Read weights from file
+            object_name = '/model_weights/' // layer_name // '/' &
+              // layer_name // '/kernel:0'
+            call get_hdf5_dataset(filename, object_name, this_layer % weights)
+
+            ! TODO Multidimensional arrays are stored in HDF5 in C-order.
+            ! TODO Here we transpose the array to get to the Fortran order.
+            ! TODO There may be a way to do this without re-allocating.
+            ! TODO It probably doesn't matter much since we do this once.
+            ! TODO Figure it out later.
+            this_layer % weights = transpose(this_layer % weights)
+
+          class default
+            error stop 'Internal error in network_from_keras(); ' &
+              // 'mismatch in layer types between the Keras and ' &
+              // 'neural-fortran model layers.'
+
+        end select
+
+    end if
+
+  end do
 
   end function network_from_keras
 

test/test_dense_network_from_keras.f90

@@ -8,7 +8,6 @@ program test_dense_network_from_keras
 
   type(network) :: net
   character(*), parameter :: test_data_path = 'keras_dense_mnist.h5'
-
   logical :: file_exists
   logical :: ok = .true.
 
@@ -52,6 +51,29 @@ program test_dense_network_from_keras
     ok = .false.
   end if
 
+  block
+
+    use nf, only: load_mnist, label_digits
+
+    real, allocatable :: training_images(:,:), training_labels(:)
+    real, allocatable :: validation_images(:,:), validation_labels(:)
+    real, allocatable :: testing_images(:,:), testing_labels(:)
+    real :: acc
+
+    call load_mnist(training_images, training_labels, &
+                    validation_images, validation_labels, &
+                    testing_images, testing_labels)
+
+    acc = accuracy(net, testing_images, label_digits(testing_labels))
+
+    if (acc < 0.94) then
+      write(stderr, '(a)') &
+        'Pre-trained network accuracy should be > 0.94.. failed'
+      ok = .false.
+    end if
+
+  end block
+
   if (ok) then
     print '(a)', 'test_dense_network_from_keras: All tests passed.'
   else
@@ -60,4 +82,19 @@ program test_dense_network_from_keras
     stop 1
   end if
 
+contains
+
+  real function accuracy(net, x, y)
+    type(network), intent(in out) :: net
+    real, intent(in) :: x(:,:), y(:,:)
+    integer :: i, good
+    good = 0
+    do i = 1, size(x, dim=2)
+      if (all(maxloc(net % output(x(:,i))) == maxloc(y(:,i)))) then
+        good = good + 1
+      end if
+    end do
+    accuracy = real(good) / size(x, dim=2)
+  end function accuracy
+
 end program test_dense_network_from_keras