diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 0df37f11..e7ec789a 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -14,52 +14,50 @@ on:
       - "fpm.toml"
       - "**.f90"
 
-env:
-  FPM_FFLAGS: -I/usr/include/hdf5/serial
-  FPM_LDFLAGS: -L/usr/lib/x86_64-linux-gnu/hdf5/serial
-
 jobs:
 
-  build_and_test_debug_profile:
-    name: Build and test in debug mode
+  gnu-cmake-debug:
+    name: gnu-cmake-debug
     runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v4
+    - name: Compile
+      run: cmake -DCMAKE_BUILD_TYPE=Debug -DSERIAL=1 . && make
+    - name: Test
+      run: make test
 
+  gnu-cmake-release:
+    name: gnu-cmake-release
+    runs-on: ubuntu-latest
     steps:
+    - uses: actions/checkout@v4
+    - name: Compile
+      run: cmake -DCMAKE_BUILD_TYPE=Release -DSERIAL=1 . && make
+    - name: Test
+      run: make test
 
-    - uses: fortran-lang/setup-fpm@v4
+  gnu-fpm-debug:
+    name: gnu-fpm-debug
+    runs-on: ubuntu-latest
+    steps:
+    - uses: fortran-lang/setup-fpm@v5
       with:
-        fpm-version: "v0.6.0"
-
-    - name: Install HDF5
-      run: |
-        sudo apt update
-        sudo apt install --no-install-recommends libhdf5-dev
-    - uses: actions/checkout@v2
-
+        fpm-version: "v0.10.1"
+    - uses: actions/checkout@v4
     - name: Compile
       run: fpm build --profile debug
-
     - name: Test
       run: fpm test --profile debug
 
-  build_and_test_release_profile:
-    name: Build and test in release mode
+  gnu-fpm-release:
+    name: gnu-fpm-release
     runs-on: ubuntu-latest
-
     steps:
-
-    - uses: fortran-lang/setup-fpm@v4
+    - uses: fortran-lang/setup-fpm@v5
       with:
-        fpm-version: "v0.6.0"
-
-    - name: Install HDF5
-      run: |
-        sudo apt update
-        sudo apt install --no-install-recommends libhdf5-dev
-    - uses: actions/checkout@v2
-
+        fpm-version: "v0.10.1"
+    - uses: actions/checkout@v4
     - name: Compile
       run: fpm build --profile release
-
     - name: Test
       run: fpm test --profile release
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 59f86fff..0340d6b2 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -6,21 +6,13 @@ if(NOT CMAKE_BUILD_TYPE)
   set(CMAKE_BUILD_TYPE "Release" CACHE STRING "Default build Release")
 endif()
 
-project(neural-fortran
-LANGUAGES C Fortran
-)
+project(neural-fortran LANGUAGES Fortran)
 
 enable_testing()
 
-include(FetchContent)
-
 include(cmake/options.cmake)
 include(cmake/compilers.cmake)
 
-include(cmake/functional.cmake)
-include(cmake/h5fortran.cmake)
-include(cmake/json.cmake)
-
 # library to archive (libneural-fortran.a)
 add_library(neural-fortran
   src/nf.f90
@@ -40,8 +32,6 @@ add_library(neural-fortran
   src/nf/nf_input1d_layer_submodule.f90
   src/nf/nf_input3d_layer.f90
   src/nf/nf_input3d_layer_submodule.f90
-  src/nf/nf_keras.f90
-  src/nf/nf_keras_submodule.f90
   src/nf/nf_layer_constructors.f90
   src/nf/nf_layer_constructors_submodule.f90
   src/nf/nf_layer.f90
@@ -61,16 +51,9 @@ add_library(neural-fortran
   src/nf/nf_reshape_layer_submodule.f90
   src/nf/io/nf_io_binary.f90
   src/nf/io/nf_io_binary_submodule.f90
-  src/nf/io/nf_io_hdf5.f90
-  src/nf/io/nf_io_hdf5_submodule.f90
 )
 
-target_link_libraries(neural-fortran PRIVATE
-  functional::functional
-  h5fortran::h5fortran
-  HDF5::HDF5
-  jsonfortran::jsonfortran
-)
+target_link_libraries(neural-fortran PRIVATE)
 
 install(TARGETS neural-fortran)
 
diff --git a/README.md b/README.md
index 054582b2..d512f91d 100644
--- a/README.md
+++ b/README.md
@@ -21,8 +21,9 @@ Read the paper [here](https://arxiv.org/abs/1902.06714).
   RMSProp, Adagrad, Adam, AdamW
 * More than a dozen activation functions and their derivatives
 * Loss functions and metrics: Quadratic, Mean Squared Error, Pearson Correlation etc.
-* Loading dense and convolutional models from Keras HDF5 (.h5) files
 * Data-based parallelism
+* Loading dense and convolutional models from Keras HDF5 (.h5) files
+(see the [nf-keras-hdf5](https://github.com/neural-fortran/nf-keras-hdf5) add-on)
 
 ### Available layers
 
@@ -51,14 +52,8 @@ cd neural-fortran
 Required dependencies are:
 
 * A Fortran compiler
-* [HDF5](https://www.hdfgroup.org/downloads/hdf5/)
-  (must be provided by the OS package manager or your own build from source)
-* [functional-fortran](https://github.com/wavebitscientific/functional-fortran),
-  [h5fortran](https://github.com/geospace-code/h5fortran),
-  [json-fortran](https://github.com/jacobwilliams/json-fortran)
-  (all handled by neural-fortran's build systems, no need for a manual install)
 * [fpm](https://github.com/fortran-lang/fpm) or
-  [CMake](https://cmake.org) for building the code
+  [CMake](https://cmake.org) to build the code
 
 Optional dependencies are:
 
@@ -79,23 +74,7 @@ Compilers tested include:
 With gfortran, the following will create an optimized build of neural-fortran:
 
 ```
-fpm build \
-  --profile release \
-  --flag "-I$HDF5INC -L$HDF5LIB"
-```
-
-HDF5 is now a required dependency, so you have to provide it to fpm.
-The above command assumes that the `HDF5INC` and `HDF5LIB` environment
-variables are set to the include and library paths, respectively, of your
-HDF5 install.
-
-If you use Conda, the following instructions work:
-
-```
-conda create -n nf hdf5
-conda activate nf
-fpm build --profile release --flag "-I$CONDA_PREFIX/include -L$CONDA_PREFIX/lib -Wl,-rpath -Wl,$CONDA_PREFIX/lib"
-fpm test --profile release --flag "-I$CONDA_PREFIX/include -L$CONDA_PREFIX/lib -Wl,-rpath -Wl,$CONDA_PREFIX/lib"
+fpm build --profile release
 ```
 
 #### Building in parallel mode
@@ -106,25 +85,20 @@ Once installed, use the compiler wrappers `caf` and `cafrun` to build and execut
 in parallel, respectively:
 
 ```
-fpm build \
-  --compiler caf \
-  --profile release \
-  --flag "-I$HDF5INC -L$HDF5LIB"
+fpm build --compiler caf --profile release
 ```
 
 #### Testing with fpm
 
 ```
-fpm test \
-  --profile release \
-  --flag "-I$HDF5INC -L$HDF5LIB"
+fpm test --profile release
 ```
 
 For the time being, you need to specify the same compiler flags to `fpm test`
 as you did in `fpm build` so that fpm knows it should use the same build
 profile.
 
-See [Fortran Package Manager](https://github.com/fortran-lang/fpm) for more info on fpm.
+See the [Fortran Package Manager](https://github.com/fortran-lang/fpm) for more info on fpm.
 
 ### Building with CMake
 
@@ -156,8 +130,7 @@ cafrun -n 4 bin/mnist # run MNIST example on 4 cores
 #### Building with a different compiler
 
 If you want to build with a different compiler, such as Intel Fortran,
-set the `HDF5_ROOT` environment variable to the root path of your
-Intel HDF5 build, and specify `FC` when issuing `cmake`:
+specify `FC` when issuing `cmake`:
 
 ```
 FC=ifort cmake ..
@@ -213,6 +186,7 @@ You can configure neural-fortran by setting the appropriate options before
 including the subproject.
 
 The following should be added in the CMake file of your directory:
+
 ```cmake
 if(NOT TARGET "neural-fortran::neural-fortran")
   find_package("neural-fortran" REQUIRED)
@@ -230,11 +204,7 @@ examples, in increasing level of complexity:
 3. [dense_mnist](example/dense_mnist.f90): Hand-written digit recognition
   (MNIST dataset) using a dense (fully-connected) network
 4. [cnn_mnist](example/cnn_mnist.f90): Training a CNN on the MNIST dataset
-5. [dense_from_keras](example/dense_from_keras.f90): Creating a pre-trained
-  dense model from a Keras HDF5 file and running the inference.
-6. [cnn_from_keras](example/cnn_from_keras.f90): Creating a pre-trained
-  convolutional model from a Keras HDF5 file and running the inference.
-7. [get_set_network_params](example/get_set_network_params.f90): Getting and
+5. [get_set_network_params](example/get_set_network_params.f90): Getting and
   setting hyperparameters of a network.
 
 The examples also show you the extent of the public API that's meant to be
diff --git a/cmake/functional.cmake b/cmake/functional.cmake
deleted file mode 100644
index d8f21200..00000000
--- a/cmake/functional.cmake
+++ /dev/null
@@ -1,18 +0,0 @@
-FetchContent_Declare(functional
-  GIT_REPOSITORY https://github.com/wavebitscientific/functional-fortran
-  GIT_TAG 0.6.1
-  GIT_SHALLOW true
-)
-
-FetchContent_Populate(functional)
-
-add_library(functional ${functional_SOURCE_DIR}/src/functional.f90)
-target_include_directories(functional PUBLIC
-$<BUILD_INTERFACE:${CMAKE_BINARY_DIR}/include>
-$<INSTALL_INTERFACE:include>
-)
-
-add_library(functional::functional INTERFACE IMPORTED GLOBAL)
-target_link_libraries(functional::functional INTERFACE functional)
-
-install(TARGETS functional)
diff --git a/cmake/h5fortran.cmake b/cmake/h5fortran.cmake
deleted file mode 100644
index 6351e63a..00000000
--- a/cmake/h5fortran.cmake
+++ /dev/null
@@ -1,15 +0,0 @@
-set(h5fortran_BUILD_TESTING false)
-
-FetchContent_Declare(h5fortran
-  GIT_REPOSITORY https://github.com/geospace-code/h5fortran
-  GIT_TAG v4.6.3
-  GIT_SHALLOW true
-)
-
-FetchContent_MakeAvailable(h5fortran)
-
-file(MAKE_DIRECTORY ${h5fortran_BINARY_DIR}/include)
-
-
-list(APPEND CMAKE_MODULE_PATH ${h5fortran_SOURCE_DIR}/cmake/Modules)
-find_package(HDF5 COMPONENTS Fortran REQUIRED)
diff --git a/cmake/json.cmake b/cmake/json.cmake
deleted file mode 100644
index 5713382e..00000000
--- a/cmake/json.cmake
+++ /dev/null
@@ -1,35 +0,0 @@
-# use our own CMake script to build jsonfortran instead of jsonfortran/CMakelists.txt
-
-FetchContent_Declare(jsonfortran
-  GIT_REPOSITORY https://github.com/jacobwilliams/json-fortran
-  GIT_TAG 8.3.0
-  GIT_SHALLOW true
-)
-
-FetchContent_Populate(jsonfortran)
-
-SET(JSON_REAL_KIND "REAL64")
-SET(JSON_INT_KIND "INT32")
-
-set(_src ${jsonfortran_SOURCE_DIR}/src)
-
-set (JF_LIB_SRCS
-${_src}/json_kinds.F90
-${_src}/json_parameters.F90
-${_src}/json_string_utilities.F90
-${_src}/json_value_module.F90
-${_src}/json_file_module.F90
-${_src}/json_module.F90
-)
-
-add_library(jsonfortran ${JF_LIB_SRCS})
-target_compile_definitions(jsonfortran PRIVATE ${JSON_REAL_KIND} ${JSON_INT_KIND})
-target_include_directories(jsonfortran PUBLIC
-$<BUILD_INTERFACE:${CMAKE_BINARY_DIR}/include>
-$<INSTALL_INTERFACE:include>
-)
-
-add_library(jsonfortran::jsonfortran INTERFACE IMPORTED GLOBAL)
-target_link_libraries(jsonfortran::jsonfortran INTERFACE jsonfortran)
-
-install(TARGETS jsonfortran)
diff --git a/example/CMakeLists.txt b/example/CMakeLists.txt
index 48bcc91e..28cf71a7 100644
--- a/example/CMakeLists.txt
+++ b/example/CMakeLists.txt
@@ -1,8 +1,6 @@
 foreach(execid
   cnn_mnist
-  cnn_from_keras
   dense_mnist
-  dense_from_keras
   get_set_network_params
   network_parameters
   simple
@@ -12,8 +10,6 @@ foreach(execid
   add_executable(${execid} ${execid}.f90)
   target_link_libraries(${execid} PRIVATE
     neural-fortran
-    h5fortran::h5fortran
-    jsonfortran::jsonfortran
     ${LIBS}
   )
 endforeach()
diff --git a/example/cnn_from_keras.f90 b/example/cnn_from_keras.f90
deleted file mode 100644
index 6b21f894..00000000
--- a/example/cnn_from_keras.f90
+++ /dev/null
@@ -1,58 +0,0 @@
-program cnn_from_keras
-
-  ! This example demonstrates loading a convolutional model
-  ! pre-trained on the MNIST dataset from a Keras HDF5
-  ! file and running an inferrence on the testing dataset.
-
-  use nf, only: network, label_digits, load_mnist
-  use nf_datasets, only: download_and_unpack, keras_cnn_mnist_url
-
-  implicit none
-
-  type(network) :: net
-  real, allocatable :: training_images(:,:), training_labels(:)
-  real, allocatable :: validation_images(:,:), validation_labels(:)
-  real, allocatable :: testing_images(:,:), testing_labels(:)
-  character(*), parameter :: keras_cnn_path = 'keras_cnn_mnist.h5'
-  logical :: file_exists
-  real :: acc
-
-  inquire(file=keras_cnn_path, exist=file_exists)
-  if (.not. file_exists) call download_and_unpack(keras_cnn_mnist_url)
-
-  call load_mnist(training_images, training_labels, &
-                  validation_images, validation_labels, &
-                  testing_images, testing_labels)
-
-  print '("Loading a pre-trained CNN model from Keras")'
-  print '(60("="))'
-
-  net = network(keras_cnn_path)
-
-  call net % print_info()
-
-  if (this_image() == 1) then
-    acc = accuracy( &
-      net, &
-      reshape(testing_images(:,:), shape=[1,28,28,size(testing_images,2)]), &
-      label_digits(testing_labels) &
-    )
-    print '(a,f5.2,a)', 'Accuracy: ', acc * 100, ' %'
-  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=4)
-      if (all(maxloc(net % predict(x(:,:,:,i))) == maxloc(y(:,i)))) then
-        good = good + 1
-      end if
-    end do
-    accuracy = real(good) / size(x, dim=4)
-  end function accuracy
-
-end program cnn_from_keras
diff --git a/example/dense_from_keras.f90 b/example/dense_from_keras.f90
deleted file mode 100644
index 4fc332d1..00000000
--- a/example/dense_from_keras.f90
+++ /dev/null
@@ -1,52 +0,0 @@
-program dense_from_keras
-
-  ! This example demonstrates loading a dense model
-  ! pre-trained on the MNIST dataset from a Keras HDF5
-  ! file and running an inferrence on the testing dataset.
-
-  use nf, only: network, label_digits, load_mnist
-  use nf_datasets, only: download_and_unpack, keras_dense_mnist_url
-
-  implicit none
-
-  type(network) :: net
-  real, allocatable :: training_images(:,:), training_labels(:)
-  real, allocatable :: validation_images(:,:), validation_labels(:)
-  real, allocatable :: testing_images(:,:), testing_labels(:)
-  character(*), parameter :: keras_dense_path = 'keras_dense_mnist.h5'
-  logical :: file_exists
-
-  inquire(file=keras_dense_path, exist=file_exists)
-  if (.not. file_exists) call download_and_unpack(keras_dense_mnist_url)
-
-  call load_mnist(training_images, training_labels, &
-                  validation_images, validation_labels, &
-                  testing_images, testing_labels)
-
-  print '("Loading a pre-trained dense model from Keras")'
-  print '(60("="))'
-
-  net = network(keras_dense_path)
-
-  call net % print_info()
-
-  if (this_image() == 1) &
-    print '(a,f5.2,a)', 'Accuracy: ', accuracy( &
-      net, testing_images, label_digits(testing_labels)) * 100, ' %'
-
-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 % predict(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 dense_from_keras
diff --git a/fpm.toml b/fpm.toml
index b638d215..5f68f8f6 100644
--- a/fpm.toml
+++ b/fpm.toml
@@ -1,15 +1,6 @@
 name = "neural-fortran"
-version = "0.17.0"
+version = "0.18.0"
 license = "MIT"
 author = "Milan Curcic"
 maintainer = "milancurcic@hey.com"
 copyright = "Copyright 2018-2024, neural-fortran contributors"
-
-[build]
-external-modules = "hdf5"
-link = ["hdf5", "hdf5_fortran"]
-
-[dependencies]
-functional = { git = "https://github.com/wavebitscientific/functional-fortran" }
-h5fortran = { git = "https://github.com/geospace-code/h5fortran" }
-json-fortran = { git = "https://github.com/jacobwilliams/json-fortran" }
diff --git a/src/nf/io/nf_io_hdf5.f90 b/src/nf/io/nf_io_hdf5.f90
deleted file mode 100644
index ac74524e..00000000
--- a/src/nf/io/nf_io_hdf5.f90
+++ /dev/null
@@ -1,60 +0,0 @@
-module nf_io_hdf5
-
-  !! This module provides convenience functions to read HDF5 files.
-
-  use iso_fortran_env, only: real32
-
-  implicit none
-
-  private
-  public :: get_hdf5_dataset, hdf5_attribute_string
-
-  interface
-    module function hdf5_attribute_string( &
-      filename, object_name, attribute_name) result(res)
-      !! Read and return an HDF5 variable-length UTF-8 string attribute.
-      character(*), intent(in) :: filename
-        !! HDF5 file name
-      character(*), intent(in) :: object_name
-        !! Object (group, dataset) name
-      character(*), intent(in) :: attribute_name
-        !! Name of the attribute to read
-      character(:), allocatable :: res
-    end function hdf5_attribute_string
-  end interface
-
-  interface get_hdf5_dataset
-
-    module subroutine get_hdf5_dataset_real32_1d(filename, object_name, values)
-      !! Read a 1-d real32 array from an HDF5 dataset.
-      character(*), intent(in) :: filename
-        !! HDF5 file name
-      character(*), intent(in) :: object_name
-        !! Object (dataset) name
-      real(real32), allocatable, intent(out) :: values(:)
-        !! Array to store the dataset values into
-    end subroutine get_hdf5_dataset_real32_1d
-
-    module subroutine get_hdf5_dataset_real32_2d(filename, object_name, values)
-      !! Read a 2-d real32 array from an HDF5 dataset.
-      character(*), intent(in) :: filename
-        !! HDF5 file name
-      character(*), intent(in) :: object_name
-        !! Object (dataset) name
-      real(real32), allocatable, intent(out) :: values(:,:)
-        !! Array to store the dataset values into
-    end subroutine get_hdf5_dataset_real32_2d
-  
-    module subroutine get_hdf5_dataset_real32_4d(filename, object_name, values)
-      !! Read a 4-d real32 array from an HDF5 dataset.
-      character(*), intent(in) :: filename
-        !! HDF5 file name
-      character(*), intent(in) :: object_name
-        !! Object (dataset) name
-      real(real32), allocatable, intent(out) :: values(:,:,:,:)
-        !! Array to store the dataset values into
-    end subroutine get_hdf5_dataset_real32_4d
-
-  end interface get_hdf5_dataset
-
-end module nf_io_hdf5
diff --git a/src/nf/io/nf_io_hdf5_submodule.f90 b/src/nf/io/nf_io_hdf5_submodule.f90
deleted file mode 100644
index 8cb6f760..00000000
--- a/src/nf/io/nf_io_hdf5_submodule.f90
+++ /dev/null
@@ -1,111 +0,0 @@
-submodule(nf_io_hdf5) nf_io_hdf5_submodule
-
-  use iso_fortran_env, only: int64, real32, stderr => error_unit
-  use h5fortran, only: hdf5_file
-  use hdf5, only: H5F_ACC_RDONLY_F, HID_T, &
-                  h5aget_type_f, h5aopen_by_name_f, h5aread_f, &
-                  h5fclose_f, h5fopen_f
-  use iso_c_binding, only: c_char, c_f_pointer, c_loc, c_null_char, c_ptr
-
-  implicit none
-
-contains
-
-  module function hdf5_attribute_string( &
-    filename, object_name, attribute_name) result(res)
-
-    character(*), intent(in) :: filename
-    character(*), intent(in) :: object_name
-    character(*), intent(in) :: attribute_name
-    character(:), allocatable :: res
-
-    ! Make sufficiently large to hold most attributes
-    integer, parameter :: BUFLEN = 10000
-
-    type(c_ptr) :: f_ptr
-    type(c_ptr), target :: buffer
-    character(len=BUFLEN, kind=c_char), pointer :: string => null()
-    integer(HID_T) :: fid, aid, atype
-    integer :: hdferr
-
-    ! Open the file and get the type of the attribute
-    call h5fopen_f(filename, H5F_ACC_RDONLY_F, fid, hdferr)
-    call h5aopen_by_name_f(fid, object_name, attribute_name, aid, hdferr)
-    call h5aget_type_f(aid, atype, hdferr)
-
-    ! Read the data
-    f_ptr = c_loc(buffer)
-    call h5aread_f(aid, atype, f_ptr, hdferr)
-    call c_f_pointer(buffer, string)
-
-    ! Close the file 
-    call h5fclose_f(fid, hdferr)
-    
-    res = string(:index(string, c_null_char) - 1)
-
-  end function hdf5_attribute_string
-
-
-  module subroutine get_hdf5_dataset_real32_1d(filename, object_name, values)
-
-    character(*), intent(in) :: filename
-    character(*), intent(in) :: object_name
-    real(real32), allocatable, intent(out) :: values(:)
-
-    type(hdf5_file) :: f
-    integer(int64), allocatable :: dims(:)
-
-    call f % open(filename, 'r')
-    call f % shape(object_name, dims)
-
-    allocate(values(dims(1)))
-
-    call f % read(object_name, values)
-    call f % close()
-
-  end subroutine get_hdf5_dataset_real32_1d
-
-
-  module subroutine get_hdf5_dataset_real32_2d(filename, object_name, values)
-
-    character(*), intent(in) :: filename
-    character(*), intent(in) :: object_name
-    real(real32), allocatable, intent(out) :: values(:,:)
-
-    type(hdf5_file) :: f
-    integer(int64), allocatable :: dims(:)
-
-    call f % open(filename, 'r')
-    call f % shape(object_name, dims)
-
-    allocate(values(dims(1), dims(2)))
-
-    call f % read(object_name, values)
-    call f % close()
-
-    ! Transpose the array to respect Keras's storage order
-    values = transpose(values)
-
-  end subroutine get_hdf5_dataset_real32_2d
-
-
-  module subroutine get_hdf5_dataset_real32_4d(filename, object_name, values)
-
-    character(*), intent(in) :: filename
-    character(*), intent(in) :: object_name
-    real(real32), allocatable, intent(out) :: values(:,:,:,:)
-
-    type(hdf5_file) :: f
-    integer(int64), allocatable :: dims(:)
-
-    call f % open(filename, 'r')
-    call f % shape(object_name, dims)
-
-    allocate(values(dims(1), dims(2), dims(3), dims(4)))
-
-    call f % read(object_name, values)
-    call f % close()
-
-  end subroutine get_hdf5_dataset_real32_4d
-
-end submodule nf_io_hdf5_submodule
diff --git a/src/nf/nf_activation.f90 b/src/nf/nf_activation.f90
index e413243c..237ce266 100644
--- a/src/nf/nf_activation.f90
+++ b/src/nf/nf_activation.f90
@@ -7,6 +7,7 @@ module nf_activation
   private
 
   public :: activation_function
+  public :: get_activation_by_name
   public :: elu
   public :: exponential
   public :: gaussian
@@ -556,6 +557,59 @@ pure function eval_3d_celu_prime(self, x) result(res)
     end where
   end function
 
+  pure function get_activation_by_name(activation_name) result(res)
+  ! Workaround to get activation_function with some
+  ! hardcoded default parameters by its name.
+  ! Need this function since we get only activation name
+  ! from keras files.
+    character(len=*), intent(in) :: activation_name
+    class(activation_function), allocatable :: res
+
+    select case(trim(activation_name))
+    case('elu')
+     allocate ( res, source = elu(alpha = 0.1) )
+
+    case('exponential')
+      allocate ( res, source = exponential() )
+
+    case('gaussian')
+      allocate ( res, source = gaussian() )
+
+    case('linear')
+      allocate ( res, source = linear() )
+
+    case('relu')
+      allocate ( res, source = relu() )
+
+    case('leaky_relu')
+      allocate ( res, source = leaky_relu(alpha = 0.1) )
+
+    case('sigmoid')
+      allocate ( res, source = sigmoid() )
+
+    case('softmax')
+      allocate ( res, source = softmax() )
+
+    case('softplus')
+      allocate ( res, source = softplus() )
+
+    case('step')
+      allocate ( res, source = step() )
+
+    case('tanh')
+      allocate ( res, source = tanhf() )
+
+    case('celu')
+      allocate ( res, source = celu() )
+
+    case default
+        error stop 'activation_name must be one of: ' // &
+          '"elu", "exponential", "gaussian", "linear", "relu", ' // &
+          '"leaky_relu", "sigmoid", "softmax", "softplus", "step", "tanh" or "celu".'
+    end select
+
+  end function get_activation_by_name
+
   pure function get_name(self) result(name)
     !! Return the name of the activation function.
     !!
diff --git a/src/nf/nf_keras.f90 b/src/nf/nf_keras.f90
deleted file mode 100644
index efe2905a..00000000
--- a/src/nf/nf_keras.f90
+++ /dev/null
@@ -1,47 +0,0 @@
-module nf_keras
-
-  !! This module provides procedures to read and parse Keras models
-  !! from HDF5 files.
-
-  implicit none
-
-  private
-  public :: get_keras_h5_layers, keras_layer
-
-  type :: keras_layer
-
-    !! Intermediate container to convey the Keras layer
-    !! information to neural-fortran layer constructors.
-
-    ! General metadata that applies to any (or most) layers
-    character(:), allocatable :: class
-    character(:), allocatable :: name
-    character(:), allocatable :: activation
-
-    ! Dense
-    integer, allocatable :: units(:)
-
-    ! Conv2D
-    integer :: filters
-    integer, allocatable :: kernel_size(:)
-
-    ! MaxPooling2D
-    integer, allocatable :: pool_size(:)
-    integer, allocatable :: strides(:)
-
-    ! Reshape
-    integer, allocatable :: target_shape(:)
-
-  end type keras_layer
-
-  interface
-
-    module function get_keras_h5_layers(filename) result(res)
-      character(*), intent(in) :: filename
-        !! HDF5 file name
-      type(keras_layer), allocatable :: res(:)
-    end function get_keras_h5_layers
-
-  end interface
-
-end module nf_keras
diff --git a/src/nf/nf_keras_submodule.f90 b/src/nf/nf_keras_submodule.f90
deleted file mode 100644
index b5c0d292..00000000
--- a/src/nf/nf_keras_submodule.f90
+++ /dev/null
@@ -1,104 +0,0 @@
-submodule(nf_keras) nf_keras_submodule
-
-  use functional, only: reverse
-  use json_module, only: json_core, json_value
-  use nf_io_hdf5, only: hdf5_attribute_string
-
-  implicit none
-
-contains
-
-  module function get_keras_h5_layers(filename) result(res)
-    character(*), intent(in) :: filename
-    type(keras_layer), allocatable :: res(:)
-
-    character(:), allocatable :: model_config_string
-
-    type(json_core) :: json
-    type(json_value), pointer :: &
-      model_config_json, layers_json, layer_json, layer_config_json
-
-    real, allocatable :: tmp_array(:)
-    integer :: n, num_layers, units
-    logical :: found
-
-    model_config_string = hdf5_attribute_string(filename, '.', 'model_config')
-
-    call json % parse(model_config_json, model_config_string)
-    call json % get(model_config_json, 'config.layers', layers_json)
-
-    num_layers = json % count(layers_json)
-
-    allocate(res(num_layers))
-
-    ! Iterate over layers
-    layers: do n = 1, num_layers
-
-      ! Get pointer to the layer
-      call json % get_child(layers_json, n, layer_json)
-
-      ! Get type of layer as a string
-      call json % get(layer_json, 'class_name', res(n) % class)
-
-      ! Get pointer to the layer config
-      call json % get(layer_json, 'config', layer_config_json)
-
-      ! Get layer name
-      call json % get(layer_config_json, 'name', res(n) % name)
-
-      ! Get size of layer and activation if applicable;
-      ! Instantiate neural-fortran layers at this time.
-      select case(res(n) % class)
-
-        case('InputLayer')
-          call json % get(layer_config_json, 'batch_input_shape', tmp_array)
-          res(n) % units = reverse(tmp_array(2:)) ! skip the 1st (batch) dim
-      
-        case('Dense')
-          call json % get(layer_config_json, 'units', units, found)
-          res(n) % units = [units]
-          call json % get(layer_config_json, 'activation', res(n) % activation)
-
-        case('Flatten')
-          ! Nothing to read here; merely a placeholder.
-          continue
-
-        case('Conv2D')
-          call json % get(layer_config_json, &
-            'filters', res(n) % filters, found)
-          call json % get(layer_config_json, &
-            'kernel_size', res(n) % kernel_size, found)
-          call json % get(layer_config_json, &
-            'activation', res(n) % activation)
-          ! Reverse to account for C -> Fortran order
-          res(n) % kernel_size = reverse(res(n) % kernel_size)
-
-        case('MaxPooling2D')
-          call json % get(layer_config_json, &
-            'pool_size',  res(n) % pool_size, found)
-          call json % get(layer_config_json, &
-            'strides',  res(n) % strides, found)
-          ! Reverse to account for C -> Fortran order
-          res(n) % pool_size = reverse(res(n) % pool_size)
-          res(n) % strides = reverse(res(n) % strides)
-
-        case('Reshape')
-          ! Only read target shape
-          call json % get(layer_config_json, &
-            'target_shape',  res(n) % target_shape, found)
-          ! Reverse to account for C -> Fortran order
-          res(n) % target_shape = reverse(res(n) % target_shape)
-
-        case default
-          error stop 'This Keras layer is not supported'
-
-      end select
-
-    end do layers
-
-    ! free the memory:
-    call json % destroy(model_config_json)
-
-  end function get_keras_h5_layers
-
-end submodule nf_keras_submodule
diff --git a/src/nf/nf_network_submodule.f90 b/src/nf/nf_network_submodule.f90
index ff8ba78c..e52051aa 100644
--- a/src/nf/nf_network_submodule.f90
+++ b/src/nf/nf_network_submodule.f90
@@ -7,8 +7,6 @@
   use nf_input3d_layer, only: input3d_layer
   use nf_maxpool2d_layer, only: maxpool2d_layer
   use nf_reshape_layer, only: reshape3d_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: conv2d, dense, flatten, input, maxpool2d, reshape
   use nf_loss, only: quadratic
@@ -95,191 +93,6 @@ module function network_from_layers(layers) result(res)
   end function network_from_layers
 
 
-  module function network_from_keras(filename) result(res)
-    character(*), intent(in) :: filename
-    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)
-
-    allocate(layers(size(keras_layers)))
-
-    do n = 1, size(layers)
-
-      select case(keras_layers(n) % class)
-
-        case('Conv2D')
-
-          if (keras_layers(n) % kernel_size(1) &
-            /= keras_layers(n) % kernel_size(2)) &
-            error stop 'Non-square kernel in conv2d layer not supported.'
-
-          layers(n) = conv2d( &
-            keras_layers(n) % filters, &
-            !FIXME add support for non-square kernel
-            keras_layers(n) % kernel_size(1), &
-            get_activation_by_name(keras_layers(n) % activation) &
-          )
-
-        case('Dense')
-
-          layers(n) = dense( &
-            keras_layers(n) % units(1), &
-            get_activation_by_name(keras_layers(n) % activation) &
-          )
-
-        case('Flatten')
-          layers(n) = flatten()
-
-        case('InputLayer')
-          if (size(keras_layers(n) % units) == 1) then
-            ! input1d
-            layers(n) = input(keras_layers(n) % units(1))
-          else
-            ! input3d
-            layers(n) = input(keras_layers(n) % units)
-          end if
-
-        case('MaxPooling2D')
-
-          if (keras_layers(n) % pool_size(1) &
-            /= keras_layers(n) % pool_size(2)) &
-            error stop 'Non-square pool in maxpool2d layer not supported.'
-
-          if (keras_layers(n) % strides(1) &
-            /= keras_layers(n) % strides(2)) &
-            error stop 'Unequal strides in maxpool2d layer are not supported.'
-
-          layers(n) = maxpool2d( &
-            !FIXME add support for non-square pool and stride
-            keras_layers(n) % pool_size(1), &
-            keras_layers(n) % strides(1) &
-          )
-
-        case('Reshape')
-          layers(n) = reshape(keras_layers(n) % target_shape)
-
-        case default
-          error stop 'This Keras layer is not supported'
-
-      end select
-
-    end do
-
-    res = network(layers)
-
-    ! 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
-
-      select type(this_layer => res % layers(n) % p)
-
-        type is(conv2d_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 % kernel)
-
-        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)
-
-        type is(flatten_layer)
-          ! Nothing to do
-          continue
-
-        type is(maxpool2d_layer)
-          ! Nothing to do
-          continue
-
-        type is(reshape3d_layer)
-          ! Nothing to do
-          continue
-
-        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 do
-
-  end function network_from_keras
-
-
-  pure function get_activation_by_name(activation_name) result(res)
-  ! Workaround to get activation_function with some
-  ! hardcoded default parameters by its name.
-  ! Need this function since we get only activation name
-  ! from keras files.
-    character(len=*), intent(in) :: activation_name
-    class(activation_function), allocatable :: res
-
-    select case(trim(activation_name))
-    case('elu')
-     allocate ( res, source = elu(alpha = 0.1) )
-
-    case('exponential')
-      allocate ( res, source = exponential() )
-
-    case('gaussian')
-      allocate ( res, source = gaussian() )
-
-    case('linear')
-      allocate ( res, source = linear() )
-
-    case('relu')
-      allocate ( res, source = relu() )
-
-    case('leaky_relu')
-      allocate ( res, source = leaky_relu(alpha = 0.1) )
-
-    case('sigmoid')
-      allocate ( res, source = sigmoid() )
-
-    case('softmax')
-      allocate ( res, source = softmax() )
-
-    case('softplus')
-      allocate ( res, source = softplus() )
-
-    case('step')
-      allocate ( res, source = step() )
-
-    case('tanh')
-      allocate ( res, source = tanhf() )
-
-    case('celu')
-      allocate ( res, source = celu() )
-
-    case default
-        error stop 'activation_name must be one of: ' // &
-          '"elu", "exponential", "gaussian", "linear", "relu", ' // &
-          '"leaky_relu", "sigmoid", "softmax", "softplus", "step", "tanh" or "celu".'
-    end select
-
-  end function get_activation_by_name
-
   pure module subroutine backward(self, output, loss)
     class(network), intent(in out) :: self
     real, intent(in) :: output(:)
diff --git a/test/CMakeLists.txt b/test/CMakeLists.txt
index 1be8bb8d..bfd3538a 100644
--- a/test/CMakeLists.txt
+++ b/test/CMakeLists.txt
@@ -10,21 +10,13 @@ foreach(execid
   reshape_layer
   dense_network
   get_set_network_params
-  io_hdf5
-  keras_read_model
-  dense_network_from_keras
-  cnn_from_keras
   conv2d_network
   optimizers
   loss
   metrics
   )
   add_executable(test_${execid} test_${execid}.f90)
-  target_link_libraries(test_${execid} PRIVATE neural-fortran h5fortran::h5fortran jsonfortran::jsonfortran ${LIBS})
+  target_link_libraries(test_${execid} PRIVATE neural-fortran ${LIBS})
 
   add_test(NAME test_${execid} COMMAND test_${execid})
 endforeach()
-
-set_tests_properties(test_dense_network_from_keras test_io_hdf5 test_keras_read_model PROPERTIES
-RESOURCE_LOCK download
-)
diff --git a/test/test_cnn_from_keras.f90 b/test/test_cnn_from_keras.f90
deleted file mode 100644
index 4d93dc51..00000000
--- a/test/test_cnn_from_keras.f90
+++ /dev/null
@@ -1,69 +0,0 @@
-program test_cnn_from_keras
-
-  use iso_fortran_env, only: stderr => error_unit
-  use nf, only: network
-  use nf_datasets, only: download_and_unpack, keras_cnn_mnist_url
-
-  implicit none
-
-  type(network) :: net
-  character(*), parameter :: test_data_path = 'keras_cnn_mnist.h5'
-  logical :: file_exists
-  logical :: ok = .true.
-
-  inquire(file=test_data_path, exist=file_exists)
-  if (.not. file_exists) call download_and_unpack(keras_cnn_mnist_url)
-
-  net = network(test_data_path)
-
-  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, allocatable :: input_reshaped(:,:,:,:)
-    real :: acc
-
-    call load_mnist(training_images, training_labels, &
-                    validation_images, validation_labels, &
-                    testing_images, testing_labels)
-
-    ! Use only the first 1000 images to make the test short
-    input_reshaped = reshape(testing_images(:,:1000), shape=[1,28,28,1000])
-
-    acc = accuracy(net, input_reshaped, label_digits(testing_labels(:1000)))
-
-    if (acc < 0.97) then
-      write(stderr, '(a)') &
-        'Pre-trained network accuracy should be > 0.97.. failed'
-      ok = .false.
-    end if
-
-  end block
-
-  if (ok) then
-    print '(a)', 'test_cnn_from_keras: All tests passed.'
-  else
-    write(stderr, '(a)') &
-      'test_cnn_from_keras: One or more tests failed.'
-    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=4)
-      if (all(maxloc(net % predict(x(:,:,:,i))) == maxloc(y(:,i)))) then
-        good = good + 1
-      end if
-    end do
-    accuracy = real(good) / size(x, dim=4)
-  end function accuracy
-
-end program test_cnn_from_keras
diff --git a/test/test_dense_network_from_keras.f90 b/test/test_dense_network_from_keras.f90
deleted file mode 100644
index ba247e38..00000000
--- a/test/test_dense_network_from_keras.f90
+++ /dev/null
@@ -1,100 +0,0 @@
-program test_dense_network_from_keras
-
-  use iso_fortran_env, only: stderr => error_unit
-  use nf, only: network
-  use nf_datasets, only: download_and_unpack, keras_dense_mnist_url
-
-  implicit none
-
-  type(network) :: net
-  character(*), parameter :: test_data_path = 'keras_dense_mnist.h5'
-  logical :: file_exists
-  logical :: ok = .true.
-
-  inquire(file=test_data_path, exist=file_exists)
-  if (.not. file_exists) call download_and_unpack(keras_dense_mnist_url)
-
-  net = network(test_data_path)
-
-  if (.not. size(net % layers) == 3) then
-    write(stderr, '(a)') 'dense network should have 3 layers.. failed'
-    ok = .false.
-  end if
-
-  if (.not. net % layers(1) % name == 'input') then
-    write(stderr, '(a)') 'First layer should be an input layer.. failed'
-    ok = .false.
-  end if
-
-  if (.not. all(net % layers(1) % layer_shape == [784])) then
-    write(stderr, '(a)') 'First layer should have shape [784].. failed'
-    ok = .false.
-  end if
-
-  if (.not. net % layers(2) % name == 'dense') then
-    write(stderr, '(a)') 'Second layer should be a dense layer.. failed'
-    ok = .false.
-  end if
-
-  if (.not. all(net % layers(2) % layer_shape == [30])) then
-    write(stderr, '(a)') 'Second layer should have shape [30].. failed'
-    ok = .false.
-  end if
-
-  if (.not. net % layers(3) % name == 'dense') then
-    write(stderr, '(a)') 'Third layer should be a dense layer.. failed'
-    ok = .false.
-  end if
-
-  if (.not. all(net % layers(3) % layer_shape == [10])) then
-    write(stderr, '(a)') 'Third layer should have shape [10].. failed'
-    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
-    write(stderr, '(a)') &
-      'test_dense_network_from_keras: One or more tests failed.'
-    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 % predict(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
diff --git a/test/test_io_hdf5.f90 b/test/test_io_hdf5.f90
deleted file mode 100644
index e7e0fc88..00000000
--- a/test/test_io_hdf5.f90
+++ /dev/null
@@ -1,57 +0,0 @@
-program test_io_hdf5
-
-  use iso_fortran_env, only: stderr => error_unit
-  use nf_datasets, only: download_and_unpack, keras_dense_mnist_url
-  use nf_io_hdf5, only: hdf5_attribute_string, get_hdf5_dataset
-
-  implicit none
-
-  character(:), allocatable :: attr
-  character(*), parameter :: test_data_path = 'keras_dense_mnist.h5'
-  real, allocatable :: bias(:), weights(:,:)
-
-  logical :: file_exists
-  logical :: ok = .true.
-
-  inquire(file=test_data_path, exist=file_exists)
-  if (.not. file_exists) call download_and_unpack(keras_dense_mnist_url)
-
-  attr = hdf5_attribute_string(test_data_path, '.', 'backend')
-
-  if (.not. attr == 'tensorflow') then
-    ok = .false.
-    write(stderr, '(a)') &
-      'HDF5 variable length string attribute was read correctly.. failed'
-  end if
-
-  ! Read 1-d real32 dataset
-  call get_hdf5_dataset( &
-    test_data_path, &
-    '/model_weights/dense/dense/bias:0', &
-    bias &
-  )
-
-  ! Read 2-d real32 dataset
-  call get_hdf5_dataset(test_data_path, &
-    '/model_weights/dense/dense/kernel:0', &
-    weights &
-  )
-
-  if (.not. all(shape(bias) == [30])) then
-    ok = .false.
-    write(stderr, '(a)') 'HDF5 1-d dataset dims inquiry is correct.. failed'
-  end if
-
-  if (.not. all(shape(weights) == [784, 30])) then
-    ok = .false.
-    write(stderr, '(a)') 'HDF5 2-d dataset dims inquiry is correct.. failed'
-  end if
-
-  if (ok) then
-    print '(a)', 'test_io_hdf5: All tests passed.'
-  else
-    write(stderr, '(a)') 'test_io_hdf5: One or more tests failed.'
-    stop 1
-  end if
-
-end program test_io_hdf5
diff --git a/test/test_keras_read_model.f90 b/test/test_keras_read_model.f90
deleted file mode 100644
index ba021a5a..00000000
--- a/test/test_keras_read_model.f90
+++ /dev/null
@@ -1,108 +0,0 @@
-program test_keras_read_model
-
-  use iso_fortran_env, only: stderr => error_unit
-  use nf_datasets, only: download_and_unpack, keras_dense_mnist_url, &
-    keras_cnn_mnist_url
-  use nf_keras, only: get_keras_h5_layers, keras_layer
-
-  implicit none
-
-  character(*), parameter :: keras_dense_path = 'keras_dense_mnist.h5'
-  character(*), parameter :: keras_cnn_path = 'keras_cnn_mnist.h5'
-
-  type(keras_layer), allocatable :: keras_layers(:)
-
-  logical :: file_exists
-  logical :: ok = .true.
-
-  ! First test the dense model
-
-  inquire(file=keras_dense_path, exist=file_exists)
-  if (.not. file_exists) call download_and_unpack(keras_dense_mnist_url)
-
-  keras_layers = get_keras_h5_layers(keras_dense_path)
-
-  if (size(keras_layers) /= 3) then
-    ok = .false.
-    write(stderr, '(a)') 'Keras dense MNIST model has 3 layers.. failed'
-  end if
-
-  if (keras_layers(1) % class /= 'InputLayer') then
-    ok = .false.
-    write(stderr, '(a)') 'Keras first layer should be InputLayer.. failed'
-  end if
-
-  if (.not. all(keras_layers(1) % units == [784])) then
-    ok = .false.
-    write(stderr, '(a)') 'Keras first layer should have 784 elements.. failed'
-  end if
-
-  if (allocated(keras_layers(1) % activation)) then
-    ok = .false.
-    write(stderr, '(a)') &
-      'Keras first layer activation should not be allocated.. failed'
-  end if
-
-  if (.not. keras_layers(2) % class == 'Dense') then
-    ok = .false.
-    write(stderr, '(a)') &
-      'Keras second and third layers should be dense.. failed'
-  end if
-
-  ! Now testing for the CNN model
-
-  inquire(file=keras_cnn_path, exist=file_exists)
-  if (.not. file_exists) call download_and_unpack(keras_cnn_mnist_url)
-
-  keras_layers = get_keras_h5_layers(keras_cnn_path)
-
-  if (.not. all(keras_layers(1) % units == [1, 28, 28])) then
-    ok = .false.
-    write(stderr, '(a)') &
-      'Keras CNN input layer shape is expected.. failed'
-  end if
-
-  if (.not. keras_layers(2) % class == 'Conv2D') then
-    ok = .false.
-    write(stderr, '(a)') &
-      'Keras CNN second layer is Conv2D.. failed'
-  end if
-
-  if (.not. keras_layers(2) % filters == 8) then
-    ok = .false.
-    write(stderr, '(a)') &
-      'Keras CNN second layer number of filters is expected.. failed'
-  end if
-
-  if (.not. all(keras_layers(2) % kernel_size == [3, 3])) then
-    ok = .false.
-    write(stderr, '(a)') &
-      'Keras CNN second layer kernel_size is expected.. failed'
-  end if
-
-  if (.not. keras_layers(3) % class == 'MaxPooling2D') then
-    ok = .false.
-    write(stderr, '(a)') &
-      'Keras CNN third layer is MaxPooling2D.. failed'
-  end if
-
-  if (.not. all(keras_layers(3) % pool_size == [2, 2])) then
-    ok = .false.
-    write(stderr, '(a)') &
-      'Keras CNN second layer pool_size is expected.. failed'
-  end if
-
-  if (.not. all(keras_layers(3) % strides == [2, 2])) then
-    ok = .false.
-    write(stderr, '(a)') &
-      'Keras CNN second layer strides are expected.. failed'
-  end if
-
-  if (ok) then
-    print '(a)', 'test_keras_read_model: All tests passed.'
-  else
-    write(stderr, '(a)') 'test_keras_read_model: One or more tests failed.'
-    stop 1
-  end if
-
-end program test_keras_read_model
diff --git a/test/test_reshape_layer.f90 b/test/test_reshape_layer.f90
index 405afd98..a448fc12 100644
--- a/test/test_reshape_layer.f90
+++ b/test/test_reshape_layer.f90
@@ -43,31 +43,6 @@ program test_reshape_layer
     ok = .false.
   end if
 
-  ! Now test reading the reshape layer from a Keras h5 model.
-  inquire(file=keras_reshape_path, exist=file_exists)
-  if (.not. file_exists) call download_and_unpack(keras_reshape_url)
-
-  net = network(keras_reshape_path)
-
-  if (.not. size(net % layers) == 2) then
-    write(stderr, '(a)') 'the reshape network from Keras has the correct size.. failed'
-    ok = .false.
-  end if
-
-  if (.not. net % layers(2) % name == 'reshape') then
-    write(stderr, '(a)') 'the 2nd layer of the reshape network from Keras is a reshape layer.. failed'
-    ok = .false.
-  end if
-
-  ! Test that the output shape checks out
-  call net % layers(1) % get_output(sample_input)
-  call net % layers(2) % get_output(output)
-
-  if (.not. all(shape(output) == [1, 28, 28])) then
-    write(stderr, '(a)') 'the target shape of the reshape layer is correct.. failed'
-    ok = .false.
-  end if
-
   if (ok) then
     print '(a)', 'test_reshape_layer: All tests passed.'
   else