add mwe for fypp preprocessor

.github/workflows/fypp.yml

@@ -0,0 +1,27 @@
+name: fypp-checks
+
+on:
+  # run on every push
+  push:
+
+jobs:
+  various:
+    name: FYPP checks - runs check on fypp and f90 files
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v3
+    - uses: actions/setup-python@v4
+      with:
+        python-version: "3.11"
+    - run: pip install fypp
+
+    - name: Check fypp matches f90
+      run: |
+        fypp src/ftorch.fypp src/temp.f90_temp
+        if ! diff -q src/ftorch.f90 src/temp.f90_temp; then
+          echo "Error: The code in ftorch.f90 does not match that expected from ftorch.fypp."
+          echo "Please re-run fypp on ftorch.fypp to ensure consistency and re-commit."
+          exit 1
+        else
+          exit 0
+        fi

README.md

@@ -128,11 +128,10 @@ To use the trained Torch model from within Fortran we need to import the `ftorch
 A very simple example is given below.
 For more detailed documentation please consult the API documentation, source code, and examples.
 
-This minimal snippet loads a saved Torch model, creates inputs consisting of two `10x10` matrices (one of ones, and one of zeros), and runs the model to infer output.
+This minimal snippet loads a saved Torch model, creates an input consisting of a `10x10` matrix of ones, and runs the model to infer output.  
+This is illustrative only, and we recommend following the [examples](examples/) before writing your own code to explore more features.
 
 ```fortran
-! Import any C bindings as required for this code
-use, intrinsic :: iso_c_binding, only: c_int, c_int64_t, c_loc
 ! Import library for interfacing with PyTorch
 use ftorch
 
@@ -141,34 +140,32 @@ implicit none
 ! Generate an object to hold the Torch model
 type(torch_module) :: model
 
-! Set up types of input and output data and the interface with C
-integer(c_int), parameter :: dims_input = 2
-integer(c_int64_t) :: shape_input(dims_input)
-integer(c_int), parameter :: n_inputs = 2
+! Set up types of input and output data
+integer, parameter :: n_inputs = 1
 type(torch_tensor), dimension(n_inputs) :: model_input_arr
-integer(c_int), parameter :: dims_output = 1
-integer(c_int64_t) :: shape_output(dims_output)
 type(torch_tensor) :: model_output
 
-! Set up the model inputs as Fortran arrays
-real, dimension(10,10), target  :: input_1, input_2
+! Set up the model input and output as Fortran arrays
+real, dimension(10,10), target  :: input
 real, dimension(5), target   :: output
 
+! Set up number of dimensions of input tensor and axis order
+integer, parameter :: in_dims = 2
+integer :: in_layout(in_dims) = [1,2]
+integer, parameter :: out_dims = 1
+integer :: out_layout(out_dims) = [1]
+
 ! Initialise the Torch model to be used
 model = torch_module_load("/path/to/saved/model.pt")
 
-! Initialise the inputs as Fortran
-input_1 = 0.0
-input_2 = 1.0
+! Initialise the inputs as Fortran array of ones
+input = 1.0
 
 ! Wrap Fortran data as no-copy Torch Tensors
 ! There may well be some reshaping required depending on the 
 ! structure of the model which is not covered here (see examples)
-shape_input = (/10, 10/)
-shape_output = (/5/)
-model_input_arr(1) = torch_tensor_from_blob(c_loc(input_1), dims_input, shape_input, torch_kFloat64, torch_kCPU)
-model_input_arr(2) = torch_tensor_from_blob(c_loc(input_2), dims_input, shape_input, torch_kFloat64, torch_kCPU)
-model_output = torch_tensor_from_blob(c_loc(output), dims_output, shape_output, torch_kFloat64, torch_kCPU)
+model_input_arr(1) = torch_tensor_from_array(input, in_layout, torch_kCPU)
+model_output = torch_tensor_from_array(output, out_layout, torch_kCPU)
 
 ! Run model and Infer
 ! Again, there may be some reshaping required depending on model design
@@ -180,7 +177,6 @@ write(*,*) output
 ! Clean up
 call torch_module_delete(model)
 call torch_tensor_delete(model_input_arr(1))
-call torch_tensor_delete(model_input_arr(2))
 call torch_tensor_delete(model_output)
 ```
 

examples/1_SimpleNet/simplenet_infer_fortran.f90

@@ -1,47 +1,45 @@
 program inference
 
-   ! Imports primitives used to interface with C
-   use, intrinsic :: iso_c_binding, only: c_int64_t, c_float, c_char, c_ptr, c_loc
+   ! Import precision info from iso
+   use, intrinsic :: iso_fortran_env, only : sp => real32
+
    ! Import our library for interfacing with PyTorch
    use ftorch
 
    implicit none
-
+
+   ! Set precision for reals
+   integer, parameter :: wp = sp
+
    integer :: num_args, ix
    character(len=128), dimension(:), allocatable :: args
 
-   ! Set up types of input and output data and the interface with C
+   ! Set up Fortran data structures
+   real(wp), dimension(5), target :: in_data
+   real(wp), dimension(5), target :: out_data
+   integer, parameter :: n_inputs = 1
+   integer :: tensor_layout(1) = [1]
+
+   ! Set up Torch data structures
    type(torch_module) :: model
    type(torch_tensor), dimension(1) :: in_tensor
    type(torch_tensor) :: out_tensor
 
-   real(c_float), dimension(:), allocatable, target :: in_data
-   integer(c_int), parameter :: n_inputs = 1
-   real(c_float), dimension(:), allocatable, target :: out_data
-
-   integer(c_int), parameter :: tensor_dims = 1
-   integer(c_int64_t) :: tensor_shape(tensor_dims) = [5]
-   integer(c_int) :: tensor_layout(tensor_dims) = [1]
-
    ! Get TorchScript model file as a command line argument
    num_args = command_argument_count()
    allocate(args(num_args))
    do ix = 1, num_args
        call get_command_argument(ix,args(ix))
    end do
 
-   ! Allocate one-dimensional input/output arrays, based on multiplication of all input/output dimension sizes
-   allocate(in_data(tensor_shape(1)))
-   allocate(out_data(tensor_shape(1)))
-
    ! Initialise data
    in_data = [0.0, 1.0, 2.0, 3.0, 4.0]
 
-   ! Create input/output tensors from the above arrays
-   in_tensor(1) = torch_tensor_from_blob(c_loc(in_data), tensor_dims, tensor_shape, torch_kFloat32, torch_kCPU, tensor_layout)
-   out_tensor = torch_tensor_from_blob(c_loc(out_data), tensor_dims, tensor_shape, torch_kFloat32, torch_kCPU, tensor_layout)
+   ! Create Torch input/output tensors from the above arrays
+   in_tensor(1) = torch_tensor_from_array(in_data, tensor_layout, torch_kCPU)
+   out_tensor = torch_tensor_from_array(out_data, tensor_layout, torch_kCPU)
 
-   ! Load ML model (edit this line to use different models)
+   ! Load ML model
    model = torch_module_load(args(1))
 
    ! Infer
@@ -52,7 +50,5 @@ program inference
    call torch_module_delete(model)
    call torch_tensor_delete(in_tensor(1))
    call torch_tensor_delete(out_tensor)
-   deallocate(in_data)
-   deallocate(out_data)
 
 end program inference

examples/2_ResNet18/resnet_infer_fortran.f90

@@ -1,18 +1,12 @@
 program inference
 
-   ! Imports primitives used to interface with C
-   use, intrinsic :: iso_c_binding, only: c_sp=>c_float, c_dp=>c_double, c_int64_t, c_loc
-   use, intrinsic :: iso_fortran_env, only : sp => real32, dp => real64
+   use, intrinsic :: iso_fortran_env, only : sp => real32
    ! Import our library for interfacing with PyTorch
    use :: ftorch
 
    implicit none
 
-   ! Define working precision for C primitives
-   ! Precision must match `wp` in resnet18.py and `wp_torch` in pt2ts.py
-   integer, parameter :: c_wp = c_sp
    integer, parameter :: wp = sp
-   integer, parameter :: torch_wp = torch_kFloat32
 
    call main()
 
@@ -25,21 +19,21 @@ subroutine main()
       integer :: num_args, ix
       character(len=128), dimension(:), allocatable :: args
 
-      ! Set up types of input and output data and the interface with C
+      ! Set up types of input and output data
       type(torch_module) :: model
       type(torch_tensor), dimension(1) :: in_tensor
       type(torch_tensor) :: out_tensor
 
-      real(c_wp), dimension(:,:,:,:), allocatable, target :: in_data
-      integer(c_int), parameter :: n_inputs = 1
-      real(c_wp), dimension(:,:), allocatable, target :: out_data
+      real(wp), dimension(:,:,:,:), allocatable, target :: in_data
+      real(wp), dimension(:,:), allocatable, target :: out_data
+      integer, parameter :: n_inputs = 1
 
-      integer(c_int), parameter :: in_dims = 4
-      integer(c_int64_t) :: in_shape(in_dims) = [1, 3, 224, 224]
-      integer(c_int) :: in_layout(in_dims) = [1,2,3,4]
-      integer(c_int), parameter :: out_dims = 2
-      integer(c_int64_t) :: out_shape(out_dims) = [1, 1000]
-      integer(c_int) :: out_layout(out_dims) = [1,2]
+      integer, parameter :: in_dims = 4
+      integer :: in_shape(in_dims) = [1, 3, 224, 224]
+      integer :: in_layout(in_dims) = [1,2,3,4]
+      integer, parameter :: out_dims = 2
+      integer :: out_shape(out_dims) = [1, 1000]
+      integer :: out_layout(out_dims) = [1,2]
 
       ! Binary file containing input tensor
       character(len=*), parameter :: filename = '../data/image_tensor.dat'
@@ -72,8 +66,9 @@ subroutine main()
       call load_data(filename, tensor_length, in_data)
 
       ! Create input/output tensors from the above arrays
-      in_tensor(1) = torch_tensor_from_blob(c_loc(in_data), in_dims, in_shape, torch_wp, torch_kCPU, in_layout)
-      out_tensor = torch_tensor_from_blob(c_loc(out_data), out_dims, out_shape, torch_wp, torch_kCPU, out_layout)
+      in_tensor(1) = torch_tensor_from_array(in_data, in_layout, torch_kCPU)
+
+      out_tensor = torch_tensor_from_array(out_data, out_layout, torch_kCPU)
 
       ! Load ML model (edit this line to use different models)
       model = torch_module_load(args(1))
@@ -113,9 +108,9 @@ subroutine load_data(filename, tensor_length, in_data)
 
       character(len=*), intent(in) :: filename
       integer, intent(in) :: tensor_length
-      real(c_wp), dimension(:,:,:,:), intent(out) :: in_data
+      real(wp), dimension(:,:,:,:), intent(out) :: in_data
 
-      real(c_wp) :: flat_data(tensor_length)
+      real(wp) :: flat_data(tensor_length)
       integer :: ios
       character(len=100) :: ioerrmsg
 
@@ -166,7 +161,7 @@ subroutine calc_probs(out_data, probabilities)
 
       implicit none
 
-      real(c_wp), dimension(:,:), intent(in) :: out_data
+      real(wp), dimension(:,:), intent(in) :: out_data
       real(wp), dimension(:,:), intent(out) :: probabilities
       real(wp) :: prob_sum
 

pre-commit

@@ -0,0 +1,69 @@
+#!/bin/sh
+#
+# A hook script to verify what is about to be committed.
+# Called by "git commit" with no arguments.  The hook should
+# exit with non-zero status after issuing an appropriate message if
+# it wants to stop the commit.
+
+# Fail immediately at first issue with the relevant exit status.
+set -eo pipefail
+
+# ===================================================================
+
+if git rev-parse --verify HEAD >/dev/null 2>&1
+then
+	against=HEAD
+else
+	# Initial commit: diff against an empty tree object
+	against=$(git hash-object -t tree /dev/null)
+fi
+
+# ===================================================================
+
+# Check that ftorch.90 is not modified and staged alone.
+git diff --cached --name-only | if grep --quiet "ftorch.f90"; then
+  git diff --cached --name-only | if ! grep --quiet "ftorch.fypp"; then
+    cat <<\EOF
+Error: File ftorch.f90 has been modified and staged without ftorch.fypp being changed.
+ftorch.90 should be generated from ftorch.fypp using fypp.
+Please restore ftorch.f90 and make your modifications to ftorch.fypp instead.
+EOF
+    exit 1
+  fi
+fi
+
+# Check to see if ftorch.fypp has been modified AND is staged.
+git diff --cached --name-only | if grep --quiet "ftorch.fypp"; then
+
+  # Check that ftorch.90 is also modified and staged.
+  git diff --cached --name-only | if ! grep --quiet "ftorch.f90"; then
+    cat <<\EOF
+Error: File ftorch.fypp has been modified and staged, but ftorch.f90 has not.
+ftorch.90 should be generated from ftorch.fypp and both committed together.
+Please run fypp on ftorch.fypp to generate ftorch.f90 and commit together.
+EOF
+    exit 1
+  else
+    # Check fypp available, and raise error and exit if not.
+    if ! command -v fypp &> /dev/null; then
+      cat <<\EOF
+echo "Error: Could not find fypp to run on ftorch.fypp.
+Please install fypp using "pip install fypp" and then try committing again.
+EOF
+      exit 1
+    fi
+
+    # If fypp is available and both .f90 and .fypp staged, check they match.
+    fypp src/ftorch.fypp src/ftorch.f90_tmp
+    if ! diff -q "src/ftorch.f90" "src/ftorch.f90_tmp" &> /dev/null; then
+      rm src/ftorch.f90_tmp
+      cat <<\EOF
+Error: The code in ftorch.f90 does not match that expected from ftorch.fypp.
+Please re-run fypp on ftorch.fypp to ensure consistency before committing.
+EOF
+      exit 1
+    else
+      rm src/ftorch.f90_tmp
+    fi
+  fi
+fi