Demo julienne

.github/workflows/ci.yml

@@ -51,9 +51,9 @@ jobs:
         fpm-version: "v0.10.1"
     - uses: actions/checkout@v4
     - name: Compile
-      run: fpm build --profile debug
+      run: fpm build --profile debug --flag -ffree-line-length-none
     - name: Test
-      run: fpm test --profile debug
+      run: fpm test --profile debug --flag -ffree-line-length-none
 
   gnu-fpm-release:
     name: gnu-fpm-release
@@ -64,6 +64,6 @@ jobs:
         fpm-version: "v0.10.1"
     - uses: actions/checkout@v4
     - name: Compile
-      run: fpm build --profile release
+      run: fpm build --profile release --flag -ffree-line-length-none
     - name: Test
-      run: fpm test --profile release
+      run: fpm test --profile release --flag -ffree-line-length-none

README.md

@@ -93,6 +93,20 @@ in parallel, respectively:
 fpm build --compiler caf --profile release --flag "-cpp -DPARALLEL"
 ```
 
+An experimental capability exists for parallel runs when building with LLVM `flang-new`
+version 22 or later and [Caffeine](https://go.lbl.gov/caffeine).  Steps for installing
+LLVM 22.0.0git (the llvm-project main branch as of this writing) and Caffeine are
+outlined in [parallel-testing-with-flang.md].  Once installed, an `fpm` command of the
+following form should launch the neural-fortran test suite with two executing images:
+
+```
+GASNET_PSHM_NODES=2 \
+  fpm test \
+  --compiler flang-new \
+  --flag "-O3 -fcoarray -DPARALLEL" \
+  --link-flag "-lcaffeine -lgasnet-smp-seq -L<caffeine-install-prefix>/lib -L<gasnet-install-prefix>/lib"
+```
+
 #### Testing with fpm
 
 ```
@@ -305,3 +319,5 @@ group.
 Neural-fortran has been used successfully in over a dozen published studies.
 See all papers that cite it
 [here](https://scholar.google.com/scholar?cites=7315840714744905948).
+
+https://github.com/BerkeleyLab/julienne/blob/e9f7ea8069206bfc4abf6a9e6dbbd7d07bda075a/doc/parallel-testing-with-flang.md

fpm.toml

@@ -5,5 +5,8 @@ author = "Milan Curcic"
 maintainer = "[email protected]"
 copyright = "Copyright 2018-2025, neural-fortran contributors"
 
+[dev-dependencies]
+julienne = {git = "https://github.com/berkeleylab/julienne", tag = "3.2.1"}
+
 [preprocess]
 [preprocess.cpp]

test/driver.f90

@@ -0,0 +1,11 @@
+program test_suite_driver
+  use julienne_m, only : test_fixture_t, test_harness_t
+  use linear_2d_layer_test_m, only : linear_2d_layer_test_t
+  implicit none
+
+  associate(test_harness => test_harness_t([ &
+     test_fixture_t(linear_2d_layer_test_t()) &
+  ]))
+    call test_harness%report_results
+  end associate
+end program test_suite_driver

test/linear_2d_layer_test_m.f90

@@ -0,0 +1,120 @@
+#include "language-support.F90"
+  ! This include and the macros below are only required for gfortran versions older than 14.3
+  ! because those versions lacked a Fortran 2008 feature that facilitates more concise code.
+
+module linear_2d_layer_test_m
+  use nf_linear2d_layer, only: linear2d_layer
+  use julienne_m, only : &
+     test_t, test_description_t, test_diagnosis_t, test_result_t &
+    ,operator(.equalsExpected.), operator(//), operator(.approximates.), operator(.within.), operator(.also.), operator(.all.)
+#if ! HAVE_PROCEDURE_ACTUAL_FOR_POINTER_DUMMY
+  use julienne_m, only : diagnosis_function_i
+#endif
+  implicit none
+
+  type, extends(test_t) :: linear_2d_layer_test_t
+  contains
+    procedure, nopass :: subject
+    procedure, nopass :: results
+  end type
+
+contains
+
+  pure function subject() result(test_subject)
+    character(len=:), allocatable :: test_subject
+    test_subject = 'A linear_2d_layer'
+  end function
+
+#if HAVE_PROCEDURE_ACTUAL_FOR_POINTER_DUMMY
+
+  function results() result(test_results)
+    type(linear_2d_layer_test_t) linear_2d_layer_test
+    type(test_result_t), allocatable :: test_results(:)
+    test_results = linear_2d_layer_test%run( & 
+      [test_description_t('updating gradients', check_gradient_updates) &
+    ])
+  end function
+
+#else
+   ! Work around a missing Fortran 2008 feature that was added to gfortran in version 14.3
+
+  function results() result(test_results)
+    type(linear_2d_layer_test_t) linear_2d_layer_test
+    type(test_result_t), allocatable :: test_results(:)
+    procedure(diagnosis_function_i), pointer :: &
+      check_gradient_updates_ptr => check_gradient_updates
+
+    test_results = linear_2d_layer_test%run( &
+      [test_description_t('updating gradients', check_gradient_updates_ptr) &
+    ])
+  end function
+
+#endif
+
+
+  function check_gradient_updates() result(test_diagnosis)
+    type(test_diagnosis_t) test_diagnosis
+
+    real :: input(3, 4) = reshape([0.0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.11, 0.12], [3, 4])
+    real :: gradient(3, 2) = reshape([0.0, 10., 0.2, 3., 0.4, 1.], [3, 2])
+    type(linear2d_layer) :: linear
+    real, pointer :: w_ptr(:)
+    real, pointer :: b_ptr(:)
+
+    integer :: num_parameters
+    real, allocatable :: parameters(:)  ! Remove the fixed size
+    real :: expected_parameters(10) = [&
+        0.100000001, 0.100000001, 0.100000001, 0.100000001, 0.100000001, 0.100000001, 0.100000001, 0.100000001,&
+        0.109999999, 0.109999999&
+    ]
+    real :: gradients(10)
+    real :: expected_gradients(10) = [&
+        1.03999996, 4.09999990, 7.15999985, 1.12400007, 0.240000010, 1.56000006, 2.88000011, 2.86399961,&
+        10.1999998, 4.40000010&
+    ]
+    real :: updated_parameters(10)
+    real :: updated_weights(8)
+    real :: updated_biases(2)
+    real :: expected_weights(8) = [&
+        0.203999996, 0.509999990, 0.816000044, 0.212400019, 0.124000005, 0.256000012, 0.388000011, 0.386399955&
+    ]
+    real :: expected_biases(2) = [1.13000000, 0.550000012]
+
+    integer :: i
+    real, parameter :: tolerance = 0.
+
+    linear = linear2d_layer(out_features=2)
+    call linear % init([3, 4])
+    linear % weights = 0.1
+    linear % biases = 0.11
+    call linear % forward(input)
+    call linear % backward(input, gradient)
+    num_parameters = linear % get_num_params()
+
+    test_diagnosis = (num_parameters .equalsExpected. 10) // " (number of parameters)"
+
+    call linear % get_params_ptr(w_ptr, b_ptr)  ! Change this_layer to linear
+    allocate(parameters(size(w_ptr) + size(b_ptr)))
+    parameters(1:size(w_ptr)) = w_ptr
+    parameters(size(w_ptr)+1:) = b_ptr
+    test_diagnosis = test_diagnosis .also. (.all. (parameters .approximates. expected_parameters .within. tolerance) // " (parameters)")
+
+    gradients =  linear % get_gradients()
+    test_diagnosis = test_diagnosis .also. (.all. (gradients .approximates. expected_gradients .within. tolerance) // " (gradients)")
+
+    do i = 1, num_parameters
+      updated_parameters(i) = parameters(i) + 0.1 * gradients(i)
+    end do
+
+    call linear % get_params_ptr(w_ptr, b_ptr)  ! Change this_layer to linear
+    w_ptr = updated_parameters(1:size(w_ptr))
+    b_ptr = updated_parameters(size(w_ptr)+1:)
+    updated_weights = reshape(linear % weights, shape(expected_weights))
+    test_diagnosis = test_diagnosis .also. (.all. (updated_weights .approximates. expected_weights .within. tolerance) // " (updated weights)")
+
+    updated_biases = linear % biases
+    test_diagnosis = test_diagnosis .also. (.all. (updated_biases .approximates. expected_biases .within. tolerance) // " (updated biases)")
+
+  end function
+
+end module linear_2d_layer_test_m