Remove some unused variables and a broadcast (#180)

* Remove unused variables

* remove batch_end variable as unused

---------

Co-authored-by: Vandenplas, Jeremie <jeremie.vandenplas@wur.nl>

Author: jvdp1

example/cnn_mnist.f90

@@ -11,9 +11,6 @@ program cnn_mnist
   real, allocatable :: training_images(:,:), training_labels(:)
   real, allocatable :: validation_images(:,:), validation_labels(:)
   real, allocatable :: testing_images(:,:), testing_labels(:)
-  real, allocatable :: input_reshaped(:,:,:,:)
-  real :: acc
-  logical :: ok
   integer :: n
   integer, parameter :: num_epochs = 10
 

example/get_set_network_params.f90

@@ -9,8 +9,7 @@ program get_set_network_params
   integer, parameter :: test_size = 30
   real :: xtest(test_size), ytest(test_size)
   real :: ypred1(test_size), ypred2(test_size)
-  integer :: i, n, nparam
-  real, allocatable :: parameters(:)
+  integer :: i, n
 
   print '("Getting and setting network parameters")'
   print '(60("="))'

example/quadratic.f90

@@ -24,7 +24,7 @@ program quadratic_fit
   real, allocatable :: x(:), y(:) ! training data
   real, allocatable :: xtest(:), ytest(:) ! testing data
 
-  integer :: i, n
+  integer :: i
 
   print '("Fitting quadratic function")'
   print '(60("="))'
@@ -277,7 +277,7 @@ subroutine rmsprop_optimizer( &
     real, intent(in) :: xtest(:), ytest(:)
     real, intent(in) :: learning_rate, decay_rate
     integer, intent(in) :: num_epochs
-    integer :: i, j, n
+    integer :: i, n
     real, allocatable :: ypred(:)
 
     print '(a)', 'RMSProp optimizer'
@@ -446,4 +446,4 @@ subroutine shuffle(arr)
     end do
   end subroutine shuffle
 
-end program quadratic_fit
+end program quadratic_fit

src/nf/nf_network_submodule.f90

@@ -596,11 +596,10 @@ module subroutine train(self, input_data, output_data, batch_size, &
     integer, intent(in) :: epochs
     class(optimizer_base_type), intent(in), optional :: optimizer
     class(loss_type), intent(in), optional :: loss
-    class(optimizer_base_type), allocatable :: optimizer_
 
     real :: pos
     integer :: dataset_size
-    integer :: batch_start, batch_end
+    integer :: batch_start
     integer :: i, j, n
     integer :: istart, iend, indices(2)
 
@@ -630,11 +629,9 @@ module subroutine train(self, input_data, output_data, batch_size, &
         ! Pull a random mini-batch from the dataset
         call random_number(pos)
         batch_start = int(pos * (dataset_size - batch_size + 1)) + 1
-        batch_end = batch_start + batch_size - 1
 
         ! FIXME shuffle in a way that doesn't require co_broadcast
         call co_broadcast(batch_start, 1)
-        call co_broadcast(batch_end, 1)
 
         ! Distribute the batch in nearly equal pieces to all images
         indices = tile_indices(batch_size)
@@ -658,7 +655,6 @@ module subroutine update(self, optimizer, batch_size)
     class(network), intent(in out) :: self
     class(optimizer_base_type), intent(in), optional :: optimizer
     integer, intent(in), optional :: batch_size
-    class(optimizer_base_type), allocatable :: optimizer_
     integer :: batch_size_
     real, allocatable :: params(:)
     integer :: n

test/test_flatten_layer.f90

@@ -9,7 +9,7 @@ program test_flatten_layer
 
   type(layer) :: test_layer, input_layer
   type(network) :: net
-  real, allocatable :: input_data(:,:,:), gradient(:,:,:)
+  real, allocatable :: gradient(:,:,:)
   real, allocatable :: output(:)
   logical :: ok = .true.
 

test/test_get_set_network_params.f90

@@ -3,7 +3,6 @@ program test_get_set_network_params
   use nf, only: conv2d, dense, flatten, input, maxpool2d, network
   implicit none
   type(network) :: net
-  integer :: n
   logical :: ok = .true.
   real :: test_params_dense(8) = [1, 2, 3, 4, 5, 6, 7, 8]
   real :: test_params_conv2d(10) = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]

test/test_optimizers.f90

@@ -8,7 +8,7 @@ program test_optimizers
   real, allocatable :: x(:), y(:)
   real, allocatable :: ypred(:)
   integer, parameter :: num_iterations = 1000
-  integer :: n, i
+  integer :: n
   logical :: ok = .true.
   logical :: converged = .false.