Agnuxo commited on Sep 18

Commit

db3c893

verified ·

1 Parent(s): 95c13dc

Upload 36 files

Browse files

Files changed (37) hide show

.gitattributes +5 -0
ALL_BUILD.vcxproj +188 -0
ALL_BUILD.vcxproj.filters +8 -0
CMakeCache.txt +716 -0
CMakeLists.txt +45 -0
Structure.txt +90 -0
ZERO_CHECK.vcxproj +182 -0
ZERO_CHECK.vcxproj.filters +13 -0
cmake_install.cmake +56 -0
data_loader.cpp +35 -0
data_loader.hpp +14 -0
data_loader.obj +3 -0
fashion_mnist_trainer.device-link.obj +3 -0
fashion_mnist_trainer.exe +3 -0
fashion_mnist_trainer.exe.recipe +14 -0
fashion_mnist_trainer.sln +53 -0
fashion_mnist_trainer.vcxproj +444 -0
fashion_mnist_trainer.vcxproj.filters +56 -0
fungi.cu +615 -0
fungi.cu.cache +57 -0
fungi.cu1408063174.deps +181 -0
fungi.hpp +51 -0
fungi.obj +3 -0
fungi_Paremetres.hpp +29 -0
main.cpp +44 -0
main.obj +0 -0
optical_model.cu +937 -0
optical_model.cu-1057474305.deps +168 -0
optical_model.cu.cache +57 -0
optical_model.hpp +123 -0
optical_model.obj +3 -0
training.cpp +134 -0
training.hpp +22 -0
training.obj +0 -0
utils.cpp +28 -0
utils.hpp +9 -0
utils.obj +0 -0

.gitattributes CHANGED Viewed

@@ -33,3 +33,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text

 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+data_loader.obj filter=lfs diff=lfs merge=lfs -text
+fashion_mnist_trainer.device-link.obj filter=lfs diff=lfs merge=lfs -text
+fashion_mnist_trainer.exe filter=lfs diff=lfs merge=lfs -text
+fungi.obj filter=lfs diff=lfs merge=lfs -text
+optical_model.obj filter=lfs diff=lfs merge=lfs -text

ALL_BUILD.vcxproj ADDED Viewed

	@@ -0,0 +1,188 @@

+<?xml version="1.0" encoding="utf-8"?>
+<Project DefaultTargets="Build" ToolsVersion="17.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <PropertyGroup>
+    <PreferredToolArchitecture>x64</PreferredToolArchitecture>
+  </PropertyGroup>
+  <PropertyGroup>
+    <ResolveNugetPackages>false</ResolveNugetPackages>
+  </PropertyGroup>
+  <ItemGroup Label="ProjectConfigurations">
+    <ProjectConfiguration Include="Debug|x64">
+      <Configuration>Debug</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|x64">
+      <Configuration>Release</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="MinSizeRel|x64">
+      <Configuration>MinSizeRel</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="RelWithDebInfo|x64">
+      <Configuration>RelWithDebInfo</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+  </ItemGroup>
+  <PropertyGroup Label="Globals">
+    <ProjectGuid>{EF40E32F-9DB8-3F09-9CAA-07717AF7BF99}</ProjectGuid>
+    <Keyword>Win32Proj</Keyword>
+    <WindowsTargetPlatformVersion>10.0.26100.0</WindowsTargetPlatformVersion>
+    <Platform>x64</Platform>
+    <ProjectName>ALL_BUILD</ProjectName>
+    <CudaToolkitCustomDir>C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0\</CudaToolkitCustomDir>
+    <VCProjectUpgraderObjectName>NoUpgrade</VCProjectUpgraderObjectName>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'" Label="Configuration">
+    <ConfigurationType>Utility</ConfigurationType>
+    <CharacterSet>MultiByte</CharacterSet>
+    <PlatformToolset>v143</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'" Label="Configuration">
+    <ConfigurationType>Utility</ConfigurationType>
+    <CharacterSet>MultiByte</CharacterSet>
+    <PlatformToolset>v143</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'" Label="Configuration">
+    <ConfigurationType>Utility</ConfigurationType>
+    <CharacterSet>MultiByte</CharacterSet>
+    <PlatformToolset>v143</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'" Label="Configuration">
+    <ConfigurationType>Utility</ConfigurationType>
+    <CharacterSet>MultiByte</CharacterSet>
+    <PlatformToolset>v143</PlatformToolset>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
+  <ImportGroup Label="ExtensionSettings">
+    <Import Project="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0\extras\visual_studio_integration\MSBuildExtensions\CUDA 13.0.props" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <PropertyGroup Label="UserMacros" />
+  <PropertyGroup>
+    <_ProjectFileVersion>10.0.20506.1</_ProjectFileVersion>
+    <IntDir Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">$(Platform)\$(Configuration)\$(ProjectName)\</IntDir>
+    <IntDir Condition="'$(Configuration)|$(Platform)'=='Release|x64'">$(Platform)\$(Configuration)\$(ProjectName)\</IntDir>
+    <IntDir Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">$(Platform)\$(Configuration)\$(ProjectName)\</IntDir>
+    <IntDir Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">$(Platform)\$(Configuration)\$(ProjectName)\</IntDir>
+  </PropertyGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+    <Midl>
+      <AdditionalIncludeDirectories>%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <OutputDirectory>$(ProjectDir)/$(IntDir)</OutputDirectory>
+      <HeaderFileName>%(Filename).h</HeaderFileName>
+      <TypeLibraryName>%(Filename).tlb</TypeLibraryName>
+      <InterfaceIdentifierFileName>%(Filename)_i.c</InterfaceIdentifierFileName>
+      <ProxyFileName>%(Filename)_p.c</ProxyFileName>
+    </Midl>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+    <Midl>
+      <AdditionalIncludeDirectories>%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <OutputDirectory>$(ProjectDir)/$(IntDir)</OutputDirectory>
+      <HeaderFileName>%(Filename).h</HeaderFileName>
+      <TypeLibraryName>%(Filename).tlb</TypeLibraryName>
+      <InterfaceIdentifierFileName>%(Filename)_i.c</InterfaceIdentifierFileName>
+      <ProxyFileName>%(Filename)_p.c</ProxyFileName>
+    </Midl>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">
+    <Midl>
+      <AdditionalIncludeDirectories>%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <OutputDirectory>$(ProjectDir)/$(IntDir)</OutputDirectory>
+      <HeaderFileName>%(Filename).h</HeaderFileName>
+      <TypeLibraryName>%(Filename).tlb</TypeLibraryName>
+      <InterfaceIdentifierFileName>%(Filename)_i.c</InterfaceIdentifierFileName>
+      <ProxyFileName>%(Filename)_p.c</ProxyFileName>
+    </Midl>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">
+    <Midl>
+      <AdditionalIncludeDirectories>%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <OutputDirectory>$(ProjectDir)/$(IntDir)</OutputDirectory>
+      <HeaderFileName>%(Filename).h</HeaderFileName>
+      <TypeLibraryName>%(Filename).tlb</TypeLibraryName>
+      <InterfaceIdentifierFileName>%(Filename)_i.c</InterfaceIdentifierFileName>
+      <ProxyFileName>%(Filename)_p.c</ProxyFileName>
+    </Midl>
+  </ItemDefinitionGroup>
+  <ItemGroup>
+    <CustomBuild Include="E:\Fashion_MNIST_Optic_Evolution\CMakeLists.txt">
+      <UseUtf8Encoding>Always</UseUtf8Encoding>
+      <Message Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">Building Custom Rule E:/Fashion_MNIST_Optic_Evolution/CMakeLists.txt</Message>
+      <Command Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">setlocal
+"C:\Program Files\CMake\bin\cmake.exe" -SE:/Fashion_MNIST_Optic_Evolution -BE:/Fashion_MNIST_Optic_Evolution/build --check-stamp-file E:/Fashion_MNIST_Optic_Evolution/build/CMakeFiles/generate.stamp
+if %errorlevel% neq 0 goto :cmEnd
+:cmEnd
+endlocal &amp; call :cmErrorLevel %errorlevel% &amp; goto :cmDone
+:cmErrorLevel
+exit /b %1
+:cmDone
+if %errorlevel% neq 0 goto :VCEnd</Command>
+      <AdditionalInputs Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCUDAInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCXXInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCommonLanguageInclude.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeGenericSystem.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeInitializeConfigs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeLanguageInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeRCInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInitialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CheckLanguage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\CMakeCommonCompilerMacros.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindCUDAToolkit.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageHandleStandardArgs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageMessage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCUDALinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCXXLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCommonLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-Initialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\WindowsPaths.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCUDACompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCXXCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeRCCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeSystem.cmake;%(AdditionalInputs)</AdditionalInputs>
+      <Outputs Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\generate.stamp</Outputs>
+      <LinkObjects Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">false</LinkObjects>
+      <Message Condition="'$(Configuration)|$(Platform)'=='Release|x64'">Building Custom Rule E:/Fashion_MNIST_Optic_Evolution/CMakeLists.txt</Message>
+      <Command Condition="'$(Configuration)|$(Platform)'=='Release|x64'">setlocal
+"C:\Program Files\CMake\bin\cmake.exe" -SE:/Fashion_MNIST_Optic_Evolution -BE:/Fashion_MNIST_Optic_Evolution/build --check-stamp-file E:/Fashion_MNIST_Optic_Evolution/build/CMakeFiles/generate.stamp
+if %errorlevel% neq 0 goto :cmEnd
+:cmEnd
+endlocal &amp; call :cmErrorLevel %errorlevel% &amp; goto :cmDone
+:cmErrorLevel
+exit /b %1
+:cmDone
+if %errorlevel% neq 0 goto :VCEnd</Command>
+      <AdditionalInputs Condition="'$(Configuration)|$(Platform)'=='Release|x64'">C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCUDAInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCXXInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCommonLanguageInclude.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeGenericSystem.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeInitializeConfigs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeLanguageInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeRCInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInitialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CheckLanguage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\CMakeCommonCompilerMacros.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindCUDAToolkit.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageHandleStandardArgs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageMessage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCUDALinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCXXLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCommonLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-Initialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\WindowsPaths.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCUDACompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCXXCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeRCCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeSystem.cmake;%(AdditionalInputs)</AdditionalInputs>
+      <Outputs Condition="'$(Configuration)|$(Platform)'=='Release|x64'">E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\generate.stamp</Outputs>
+      <LinkObjects Condition="'$(Configuration)|$(Platform)'=='Release|x64'">false</LinkObjects>
+      <Message Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">Building Custom Rule E:/Fashion_MNIST_Optic_Evolution/CMakeLists.txt</Message>
+      <Command Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">setlocal
+"C:\Program Files\CMake\bin\cmake.exe" -SE:/Fashion_MNIST_Optic_Evolution -BE:/Fashion_MNIST_Optic_Evolution/build --check-stamp-file E:/Fashion_MNIST_Optic_Evolution/build/CMakeFiles/generate.stamp
+if %errorlevel% neq 0 goto :cmEnd
+:cmEnd
+endlocal &amp; call :cmErrorLevel %errorlevel% &amp; goto :cmDone
+:cmErrorLevel
+exit /b %1
+:cmDone
+if %errorlevel% neq 0 goto :VCEnd</Command>
+      <AdditionalInputs Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCUDAInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCXXInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCommonLanguageInclude.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeGenericSystem.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeInitializeConfigs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeLanguageInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeRCInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInitialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CheckLanguage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\CMakeCommonCompilerMacros.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindCUDAToolkit.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageHandleStandardArgs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageMessage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCUDALinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCXXLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCommonLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-Initialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\WindowsPaths.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCUDACompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCXXCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeRCCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeSystem.cmake;%(AdditionalInputs)</AdditionalInputs>
+      <Outputs Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\generate.stamp</Outputs>
+      <LinkObjects Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">false</LinkObjects>
+      <Message Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">Building Custom Rule E:/Fashion_MNIST_Optic_Evolution/CMakeLists.txt</Message>
+      <Command Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">setlocal
+"C:\Program Files\CMake\bin\cmake.exe" -SE:/Fashion_MNIST_Optic_Evolution -BE:/Fashion_MNIST_Optic_Evolution/build --check-stamp-file E:/Fashion_MNIST_Optic_Evolution/build/CMakeFiles/generate.stamp
+if %errorlevel% neq 0 goto :cmEnd
+:cmEnd
+endlocal &amp; call :cmErrorLevel %errorlevel% &amp; goto :cmDone
+:cmErrorLevel
+exit /b %1
+:cmDone
+if %errorlevel% neq 0 goto :VCEnd</Command>
+      <AdditionalInputs Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCUDAInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCXXInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCommonLanguageInclude.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeGenericSystem.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeInitializeConfigs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeLanguageInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeRCInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInitialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CheckLanguage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\CMakeCommonCompilerMacros.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindCUDAToolkit.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageHandleStandardArgs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageMessage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCUDALinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCXXLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCommonLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-Initialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\WindowsPaths.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCUDACompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCXXCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeRCCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeSystem.cmake;%(AdditionalInputs)</AdditionalInputs>
+      <Outputs Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\generate.stamp</Outputs>
+      <LinkObjects Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">false</LinkObjects>
+    </CustomBuild>
+  </ItemGroup>
+  <ItemGroup>
+  </ItemGroup>
+  <ItemGroup />
+  <ItemGroup>
+    <ProjectReference Include="E:\Fashion_MNIST_Optic_Evolution\build\ZERO_CHECK.vcxproj">
+      <Project>{B4CD8D8C-908F-3E91-878E-AFB61C98E407}</Project>
+      <Name>ZERO_CHECK</Name>
+      <ReferenceOutputAssembly>false</ReferenceOutputAssembly>
+      <CopyToOutputDirectory>Never</CopyToOutputDirectory>
+    </ProjectReference>
+    <ProjectReference Include="E:\Fashion_MNIST_Optic_Evolution\build\fashion_mnist_trainer.vcxproj">
+      <Project>{2B82790A-4403-3D6B-A562-F30C253F0098}</Project>
+      <Name>fashion_mnist_trainer</Name>
+    </ProjectReference>
+  </ItemGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
+  <ImportGroup Label="ExtensionTargets">
+    <Import Project="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0\extras\visual_studio_integration\MSBuildExtensions\CUDA 13.0.targets" />
+  </ImportGroup>
+</Project>

ALL_BUILD.vcxproj.filters ADDED Viewed

	@@ -0,0 +1,8 @@

+<?xml version="1.0" encoding="utf-8"?>
+<Project ToolsVersion="17.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <ItemGroup>
+    <CustomBuild Include="E:\Fashion_MNIST_Optic_Evolution\CMakeLists.txt" />
+  </ItemGroup>
+  <ItemGroup>
+  </ItemGroup>
+</Project>

CMakeCache.txt ADDED Viewed

	@@ -0,0 +1,716 @@

+# This is the CMakeCache file.
+# For build in directory: e:/Fashion_MNIST_Optic_Evolution/build
+# It was generated by CMake: C:/Program Files/CMake/bin/cmake.exe
+# You can edit this file to change values found and used by cmake.
+# If you do not want to change any of the values, simply exit the editor.
+# If you do want to change a value, simply edit, save, and exit the editor.
+# The syntax for the file is as follows:
+# KEY:TYPE=VALUE
+# KEY is the name of a variable in the cache.
+# TYPE is a hint to GUIs for the type of VALUE, DO NOT EDIT TYPE!.
+# VALUE is the current value for the KEY.
+########################
+# EXTERNAL cache entries
+########################
+//Path to a program.
+CMAKE_AR:FILEPATH=E:/VS2022/VC/Tools/MSVC/14.38.33130/bin/Hostx64/x64/lib.exe
+//No help, variable specified on the command line.
+CMAKE_BUILD_TYPE:UNINITIALIZED=Release
+//Semicolon separated list of supported configuration types, only
+// supports Debug, Release, MinSizeRel, and RelWithDebInfo, anything
+// else will be ignored.
+CMAKE_CONFIGURATION_TYPES:STRING=Debug;Release;MinSizeRel;RelWithDebInfo
+//CUDA architectures
+CMAKE_CUDA_ARCHITECTURES:STRING=75
+//CUDA compiler
+CMAKE_CUDA_COMPILER:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/bin/nvcc.exe
+//Flags used by the CUDA compiler during all build types.
+CMAKE_CUDA_FLAGS:STRING=-D_WINDOWS -Xcompiler=" /GR /EHsc"
+//Flags used by the CUDA compiler during DEBUG builds.
+CMAKE_CUDA_FLAGS_DEBUG:STRING=-Xcompiler=" -Zi -Ob0 -Od /RTC1"
+//Flags used by the CUDA compiler during MINSIZEREL builds.
+CMAKE_CUDA_FLAGS_MINSIZEREL:STRING=-Xcompiler="-O1 -Ob1" -DNDEBUG
+//Flags used by the CUDA compiler during RELEASE builds.
+CMAKE_CUDA_FLAGS_RELEASE:STRING=-Xcompiler="-O2 -Ob2" -DNDEBUG
+//Flags used by the CUDA compiler during RELWITHDEBINFO builds.
+CMAKE_CUDA_FLAGS_RELWITHDEBINFO:STRING=-Xcompiler=" -Zi -O2 -Ob1" -DNDEBUG
+//Libraries linked by default with all CUDA applications.
+CMAKE_CUDA_STANDARD_LIBRARIES:STRING=kernel32.lib user32.lib gdi32.lib winspool.lib shell32.lib ole32.lib oleaut32.lib uuid.lib comdlg32.lib advapi32.lib
+//Flags used by the CXX compiler during all build types.
+CMAKE_CXX_FLAGS:STRING=/DWIN32 /D_WINDOWS /GR /EHsc
+//Flags used by the CXX compiler during DEBUG builds.
+CMAKE_CXX_FLAGS_DEBUG:STRING=/Zi /Ob0 /Od /RTC1
+//Flags used by the CXX compiler during MINSIZEREL builds.
+CMAKE_CXX_FLAGS_MINSIZEREL:STRING=/O1 /Ob1 /DNDEBUG
+//Flags used by the CXX compiler during RELEASE builds.
+CMAKE_CXX_FLAGS_RELEASE:STRING=/O2 /Ob2 /DNDEBUG
+//Flags used by the CXX compiler during RELWITHDEBINFO builds.
+CMAKE_CXX_FLAGS_RELWITHDEBINFO:STRING=/Zi /O2 /Ob1 /DNDEBUG
+//Libraries linked by default with all C++ applications.
+CMAKE_CXX_STANDARD_LIBRARIES:STRING=kernel32.lib user32.lib gdi32.lib winspool.lib shell32.lib ole32.lib oleaut32.lib uuid.lib comdlg32.lib advapi32.lib
+//Flags used by the linker during all build types.
+CMAKE_EXE_LINKER_FLAGS:STRING=
+//Flags used by the linker during DEBUG builds.
+CMAKE_EXE_LINKER_FLAGS_DEBUG:STRING=/debug /INCREMENTAL
+//Flags used by the linker during MINSIZEREL builds.
+CMAKE_EXE_LINKER_FLAGS_MINSIZEREL:STRING=/INCREMENTAL:NO
+//Flags used by the linker during RELEASE builds.
+CMAKE_EXE_LINKER_FLAGS_RELEASE:STRING=/INCREMENTAL:NO
+//Flags used by the linker during RELWITHDEBINFO builds.
+CMAKE_EXE_LINKER_FLAGS_RELWITHDEBINFO:STRING=/debug /INCREMENTAL
+//Value Computed by CMake.
+CMAKE_FIND_PACKAGE_REDIRECTS_DIR:STATIC=E:/Fashion_MNIST_Optic_Evolution/build/CMakeFiles/pkgRedirects
+//Install path prefix, prepended onto install directories.
+CMAKE_INSTALL_PREFIX:PATH=C:/Program Files
+//Path to a program.
+CMAKE_LINKER:FILEPATH=C:/Program Files/Git/usr/bin/link.exe
+//Flags used by the linker during the creation of modules during
+// all build types.
+CMAKE_MODULE_LINKER_FLAGS:STRING=
+//Flags used by the linker during the creation of modules during
+// DEBUG builds.
+CMAKE_MODULE_LINKER_FLAGS_DEBUG:STRING=/debug /INCREMENTAL
+//Flags used by the linker during the creation of modules during
+// MINSIZEREL builds.
+CMAKE_MODULE_LINKER_FLAGS_MINSIZEREL:STRING=/INCREMENTAL:NO
+//Flags used by the linker during the creation of modules during
+// RELEASE builds.
+CMAKE_MODULE_LINKER_FLAGS_RELEASE:STRING=/INCREMENTAL:NO
+//Flags used by the linker during the creation of modules during
+// RELWITHDEBINFO builds.
+CMAKE_MODULE_LINKER_FLAGS_RELWITHDEBINFO:STRING=/debug /INCREMENTAL
+//Path to a program.
+CMAKE_MT:FILEPATH=CMAKE_MT-NOTFOUND
+//Value Computed by CMake
+CMAKE_PROJECT_DESCRIPTION:STATIC=
+//Value Computed by CMake
+CMAKE_PROJECT_HOMEPAGE_URL:STATIC=
+//Value Computed by CMake
+CMAKE_PROJECT_NAME:STATIC=fashion_mnist_trainer
+//RC compiler
+CMAKE_RC_COMPILER:FILEPATH=rc
+//Flags for Windows Resource Compiler during all build types.
+CMAKE_RC_FLAGS:STRING=-DWIN32
+//Flags for Windows Resource Compiler during DEBUG builds.
+CMAKE_RC_FLAGS_DEBUG:STRING=-D_DEBUG
+//Flags for Windows Resource Compiler during MINSIZEREL builds.
+CMAKE_RC_FLAGS_MINSIZEREL:STRING=
+//Flags for Windows Resource Compiler during RELEASE builds.
+CMAKE_RC_FLAGS_RELEASE:STRING=
+//Flags for Windows Resource Compiler during RELWITHDEBINFO builds.
+CMAKE_RC_FLAGS_RELWITHDEBINFO:STRING=
+//Flags used by the linker during the creation of shared libraries
+// during all build types.
+CMAKE_SHARED_LINKER_FLAGS:STRING=
+//Flags used by the linker during the creation of shared libraries
+// during DEBUG builds.
+CMAKE_SHARED_LINKER_FLAGS_DEBUG:STRING=/debug /INCREMENTAL
+//Flags used by the linker during the creation of shared libraries
+// during MINSIZEREL builds.
+CMAKE_SHARED_LINKER_FLAGS_MINSIZEREL:STRING=/INCREMENTAL:NO
+//Flags used by the linker during the creation of shared libraries
+// during RELEASE builds.
+CMAKE_SHARED_LINKER_FLAGS_RELEASE:STRING=/INCREMENTAL:NO
+//Flags used by the linker during the creation of shared libraries
+// during RELWITHDEBINFO builds.
+CMAKE_SHARED_LINKER_FLAGS_RELWITHDEBINFO:STRING=/debug /INCREMENTAL
+//If set, runtime paths are not added when installing shared libraries,
+// but are added when building.
+CMAKE_SKIP_INSTALL_RPATH:BOOL=NO
+//If set, runtime paths are not added when using shared libraries.
+CMAKE_SKIP_RPATH:BOOL=NO
+//Flags used by the linker during the creation of static libraries
+// during all build types.
+CMAKE_STATIC_LINKER_FLAGS:STRING=
+//Flags used by the linker during the creation of static libraries
+// during DEBUG builds.
+CMAKE_STATIC_LINKER_FLAGS_DEBUG:STRING=
+//Flags used by the linker during the creation of static libraries
+// during MINSIZEREL builds.
+CMAKE_STATIC_LINKER_FLAGS_MINSIZEREL:STRING=
+//Flags used by the linker during the creation of static libraries
+// during RELEASE builds.
+CMAKE_STATIC_LINKER_FLAGS_RELEASE:STRING=
+//Flags used by the linker during the creation of static libraries
+// during RELWITHDEBINFO builds.
+CMAKE_STATIC_LINKER_FLAGS_RELWITHDEBINFO:STRING=
+//If this value is on, makefiles will be generated without the
+// .SILENT directive, and all commands will be echoed to the console
+// during the make.  This is useful for debugging only. With Visual
+// Studio IDE projects all commands are done without /nologo.
+CMAKE_VERBOSE_MAKEFILE:BOOL=FALSE
+//Path to a file.
+CUDAToolkit_CUPTI_INCLUDE_DIR:PATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include/../extras/CUPTI/include
+//Path to a library.
+CUDA_CUDART:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/cudart.lib
+//Path to a library.
+CUDA_OpenCL_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/OpenCL.lib
+//Path to a library.
+CUDA_cuFile_LIBRARY:FILEPATH=CUDA_cuFile_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_cuFile_rdma_LIBRARY:FILEPATH=CUDA_cuFile_rdma_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_cuFile_rdma_static_LIBRARY:FILEPATH=CUDA_cuFile_rdma_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_cuFile_static_LIBRARY:FILEPATH=CUDA_cuFile_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_cublasLt_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/cublasLt.lib
+//Path to a library.
+CUDA_cublasLt_static_LIBRARY:FILEPATH=CUDA_cublasLt_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_cublas_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/cublas.lib
+//Path to a library.
+CUDA_cublas_static_LIBRARY:FILEPATH=CUDA_cublas_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_cuda_driver_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/cuda.lib
+//Path to a library.
+CUDA_cudart_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/cudart.lib
+//Path to a library.
+CUDA_cudart_static_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/cudart_static.lib
+//Path to a library.
+CUDA_cudla_LIBRARY:FILEPATH=CUDA_cudla_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_cufft_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/cufft.lib
+//Path to a library.
+CUDA_cufft_static_LIBRARY:FILEPATH=CUDA_cufft_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_cufft_static_nocallback_LIBRARY:FILEPATH=CUDA_cufft_static_nocallback_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_cufftw_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/cufftw.lib
+//Path to a library.
+CUDA_cufftw_static_LIBRARY:FILEPATH=CUDA_cufftw_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_culibos_LIBRARY:FILEPATH=CUDA_culibos_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_cupti_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/extras/CUPTI/lib64/cupti.lib
+//Path to a library.
+CUDA_cupti_static_LIBRARY:FILEPATH=CUDA_cupti_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_curand_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/curand.lib
+//Path to a library.
+CUDA_curand_static_LIBRARY:FILEPATH=CUDA_curand_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_cusolver_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/cusolver.lib
+//Path to a library.
+CUDA_cusolver_lapack_static_LIBRARY:FILEPATH=CUDA_cusolver_lapack_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_cusolver_metis_static_LIBRARY:FILEPATH=CUDA_cusolver_metis_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_cusolver_static_LIBRARY:FILEPATH=CUDA_cusolver_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_cusparse_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/cusparse.lib
+//Path to a library.
+CUDA_cusparse_static_LIBRARY:FILEPATH=CUDA_cusparse_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nppc_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nppc.lib
+//Path to a library.
+CUDA_nppc_static_LIBRARY:FILEPATH=CUDA_nppc_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nppial_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nppial.lib
+//Path to a library.
+CUDA_nppial_static_LIBRARY:FILEPATH=CUDA_nppial_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nppicc_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nppicc.lib
+//Path to a library.
+CUDA_nppicc_static_LIBRARY:FILEPATH=CUDA_nppicc_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nppicom_LIBRARY:FILEPATH=CUDA_nppicom_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nppicom_static_LIBRARY:FILEPATH=CUDA_nppicom_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nppidei_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nppidei.lib
+//Path to a library.
+CUDA_nppidei_static_LIBRARY:FILEPATH=CUDA_nppidei_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nppif_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nppif.lib
+//Path to a library.
+CUDA_nppif_static_LIBRARY:FILEPATH=CUDA_nppif_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nppig_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nppig.lib
+//Path to a library.
+CUDA_nppig_static_LIBRARY:FILEPATH=CUDA_nppig_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nppim_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nppim.lib
+//Path to a library.
+CUDA_nppim_static_LIBRARY:FILEPATH=CUDA_nppim_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nppist_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nppist.lib
+//Path to a library.
+CUDA_nppist_static_LIBRARY:FILEPATH=CUDA_nppist_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nppisu_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nppisu.lib
+//Path to a library.
+CUDA_nppisu_static_LIBRARY:FILEPATH=CUDA_nppisu_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nppitc_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nppitc.lib
+//Path to a library.
+CUDA_nppitc_static_LIBRARY:FILEPATH=CUDA_nppitc_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_npps_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/npps.lib
+//Path to a library.
+CUDA_npps_static_LIBRARY:FILEPATH=CUDA_npps_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nvJitLink_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nvJitLink.lib
+//Path to a library.
+CUDA_nvJitLink_static_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nvJitLink_static.lib
+//Path to a library.
+CUDA_nvToolsExt_LIBRARY:FILEPATH=CUDA_nvToolsExt_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nvfatbin_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nvfatbin.lib
+//Path to a library.
+CUDA_nvfatbin_static_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nvfatbin_static.lib
+//Path to a library.
+CUDA_nvgraph_LIBRARY:FILEPATH=CUDA_nvgraph_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nvgraph_static_LIBRARY:FILEPATH=CUDA_nvgraph_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nvjpeg_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nvjpeg.lib
+//Path to a library.
+CUDA_nvjpeg_static_LIBRARY:FILEPATH=CUDA_nvjpeg_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nvml_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nvml.lib
+//Path to a library.
+CUDA_nvml_static_LIBRARY:FILEPATH=CUDA_nvml_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nvperf_host_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/extras/CUPTI/lib64/nvperf_host.lib
+//Path to a library.
+CUDA_nvperf_host_static_LIBRARY:FILEPATH=CUDA_nvperf_host_static_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nvperf_target_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/extras/CUPTI/lib64/nvperf_target.lib
+//Path to a library.
+CUDA_nvptxcompiler_static_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nvptxcompiler_static.lib
+//Path to a library.
+CUDA_nvrtc_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nvrtc.lib
+//Path to a library.
+CUDA_nvrtc_builtins_LIBRARY:FILEPATH=CUDA_nvrtc_builtins_LIBRARY-NOTFOUND
+//Path to a library.
+CUDA_nvrtc_builtins_static_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nvrtc-builtins_static.lib
+//Path to a library.
+CUDA_nvrtc_static_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/nvrtc_static.lib
+//Path to a library.
+CUDA_pcsamplingutil_LIBRARY:FILEPATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/extras/CUPTI/lib64/pcsamplingutil.lib
+//Value Computed by CMake
+fashion_mnist_trainer_BINARY_DIR:STATIC=E:/Fashion_MNIST_Optic_Evolution/build
+//Value Computed by CMake
+fashion_mnist_trainer_IS_TOP_LEVEL:STATIC=ON
+//Value Computed by CMake
+fashion_mnist_trainer_SOURCE_DIR:STATIC=E:/Fashion_MNIST_Optic_Evolution
+########################
+# INTERNAL cache entries
+########################
+//ADVANCED property for variable: CMAKE_AR
+CMAKE_AR-ADVANCED:INTERNAL=1
+//This is the directory where this CMakeCache.txt was created
+CMAKE_CACHEFILE_DIR:INTERNAL=e:/Fashion_MNIST_Optic_Evolution/build
+//Major version of cmake used to create the current loaded cache
+CMAKE_CACHE_MAJOR_VERSION:INTERNAL=3
+//Minor version of cmake used to create the current loaded cache
+CMAKE_CACHE_MINOR_VERSION:INTERNAL=31
+//Patch version of cmake used to create the current loaded cache
+CMAKE_CACHE_PATCH_VERSION:INTERNAL=4
+//Path to CMake executable.
+CMAKE_COMMAND:INTERNAL=C:/Program Files/CMake/bin/cmake.exe
+//Path to cpack program executable.
+CMAKE_CPACK_COMMAND:INTERNAL=C:/Program Files/CMake/bin/cpack.exe
+//Path to ctest program executable.
+CMAKE_CTEST_COMMAND:INTERNAL=C:/Program Files/CMake/bin/ctest.exe
+//ADVANCED property for variable: CMAKE_CUDA_COMPILER
+CMAKE_CUDA_COMPILER-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_CUDA_FLAGS
+CMAKE_CUDA_FLAGS-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_CUDA_FLAGS_DEBUG
+CMAKE_CUDA_FLAGS_DEBUG-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_CUDA_FLAGS_MINSIZEREL
+CMAKE_CUDA_FLAGS_MINSIZEREL-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_CUDA_FLAGS_RELEASE
+CMAKE_CUDA_FLAGS_RELEASE-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_CUDA_FLAGS_RELWITHDEBINFO
+CMAKE_CUDA_FLAGS_RELWITHDEBINFO-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_CUDA_STANDARD_LIBRARIES
+CMAKE_CUDA_STANDARD_LIBRARIES-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_CXX_FLAGS
+CMAKE_CXX_FLAGS-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_CXX_FLAGS_DEBUG
+CMAKE_CXX_FLAGS_DEBUG-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_CXX_FLAGS_MINSIZEREL
+CMAKE_CXX_FLAGS_MINSIZEREL-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_CXX_FLAGS_RELEASE
+CMAKE_CXX_FLAGS_RELEASE-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_CXX_FLAGS_RELWITHDEBINFO
+CMAKE_CXX_FLAGS_RELWITHDEBINFO-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_CXX_STANDARD_LIBRARIES
+CMAKE_CXX_STANDARD_LIBRARIES-ADVANCED:INTERNAL=1
+//Executable file format
+CMAKE_EXECUTABLE_FORMAT:INTERNAL=Unknown
+//ADVANCED property for variable: CMAKE_EXE_LINKER_FLAGS
+CMAKE_EXE_LINKER_FLAGS-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_EXE_LINKER_FLAGS_DEBUG
+CMAKE_EXE_LINKER_FLAGS_DEBUG-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_EXE_LINKER_FLAGS_MINSIZEREL
+CMAKE_EXE_LINKER_FLAGS_MINSIZEREL-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_EXE_LINKER_FLAGS_RELEASE
+CMAKE_EXE_LINKER_FLAGS_RELEASE-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_EXE_LINKER_FLAGS_RELWITHDEBINFO
+CMAKE_EXE_LINKER_FLAGS_RELWITHDEBINFO-ADVANCED:INTERNAL=1
+//Name of external makefile project generator.
+CMAKE_EXTRA_GENERATOR:INTERNAL=
+//Name of generator.
+CMAKE_GENERATOR:INTERNAL=Visual Studio 17 2022
+//Generator instance identifier.
+CMAKE_GENERATOR_INSTANCE:INTERNAL=E:/VS2022
+//Name of generator platform.
+CMAKE_GENERATOR_PLATFORM:INTERNAL=x64
+//Name of generator toolset.
+CMAKE_GENERATOR_TOOLSET:INTERNAL=cuda=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0
+//Source directory with the top level CMakeLists.txt file for this
+// project
+CMAKE_HOME_DIRECTORY:INTERNAL=E:/Fashion_MNIST_Optic_Evolution
+//ADVANCED property for variable: CMAKE_LINKER
+CMAKE_LINKER-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_MODULE_LINKER_FLAGS
+CMAKE_MODULE_LINKER_FLAGS-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_MODULE_LINKER_FLAGS_DEBUG
+CMAKE_MODULE_LINKER_FLAGS_DEBUG-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_MODULE_LINKER_FLAGS_MINSIZEREL
+CMAKE_MODULE_LINKER_FLAGS_MINSIZEREL-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_MODULE_LINKER_FLAGS_RELEASE
+CMAKE_MODULE_LINKER_FLAGS_RELEASE-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_MODULE_LINKER_FLAGS_RELWITHDEBINFO
+CMAKE_MODULE_LINKER_FLAGS_RELWITHDEBINFO-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_MT
+CMAKE_MT-ADVANCED:INTERNAL=1
+//number of local generators
+CMAKE_NUMBER_OF_MAKEFILES:INTERNAL=1
+//Platform information initialized
+CMAKE_PLATFORM_INFO_INITIALIZED:INTERNAL=1
+//noop for ranlib
+CMAKE_RANLIB:INTERNAL=:
+//ADVANCED property for variable: CMAKE_RC_COMPILER
+CMAKE_RC_COMPILER-ADVANCED:INTERNAL=1
+CMAKE_RC_COMPILER_WORKS:INTERNAL=1
+//ADVANCED property for variable: CMAKE_RC_FLAGS
+CMAKE_RC_FLAGS-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_RC_FLAGS_DEBUG
+CMAKE_RC_FLAGS_DEBUG-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_RC_FLAGS_MINSIZEREL
+CMAKE_RC_FLAGS_MINSIZEREL-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_RC_FLAGS_RELEASE
+CMAKE_RC_FLAGS_RELEASE-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_RC_FLAGS_RELWITHDEBINFO
+CMAKE_RC_FLAGS_RELWITHDEBINFO-ADVANCED:INTERNAL=1
+//Path to CMake installation.
+CMAKE_ROOT:INTERNAL=C:/Program Files/CMake/share/cmake-3.31
+//ADVANCED property for variable: CMAKE_SHARED_LINKER_FLAGS
+CMAKE_SHARED_LINKER_FLAGS-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_SHARED_LINKER_FLAGS_DEBUG
+CMAKE_SHARED_LINKER_FLAGS_DEBUG-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_SHARED_LINKER_FLAGS_MINSIZEREL
+CMAKE_SHARED_LINKER_FLAGS_MINSIZEREL-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_SHARED_LINKER_FLAGS_RELEASE
+CMAKE_SHARED_LINKER_FLAGS_RELEASE-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_SHARED_LINKER_FLAGS_RELWITHDEBINFO
+CMAKE_SHARED_LINKER_FLAGS_RELWITHDEBINFO-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_SKIP_INSTALL_RPATH
+CMAKE_SKIP_INSTALL_RPATH-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_SKIP_RPATH
+CMAKE_SKIP_RPATH-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_STATIC_LINKER_FLAGS
+CMAKE_STATIC_LINKER_FLAGS-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_STATIC_LINKER_FLAGS_DEBUG
+CMAKE_STATIC_LINKER_FLAGS_DEBUG-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_STATIC_LINKER_FLAGS_MINSIZEREL
+CMAKE_STATIC_LINKER_FLAGS_MINSIZEREL-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_STATIC_LINKER_FLAGS_RELEASE
+CMAKE_STATIC_LINKER_FLAGS_RELEASE-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_STATIC_LINKER_FLAGS_RELWITHDEBINFO
+CMAKE_STATIC_LINKER_FLAGS_RELWITHDEBINFO-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CMAKE_VERBOSE_MAKEFILE
+CMAKE_VERBOSE_MAKEFILE-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDAToolkit_CUPTI_INCLUDE_DIR
+CUDAToolkit_CUPTI_INCLUDE_DIR-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_CUDART
+CUDA_CUDART-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_OpenCL_LIBRARY
+CUDA_OpenCL_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cuFile_LIBRARY
+CUDA_cuFile_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cuFile_rdma_LIBRARY
+CUDA_cuFile_rdma_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cuFile_rdma_static_LIBRARY
+CUDA_cuFile_rdma_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cuFile_static_LIBRARY
+CUDA_cuFile_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cublasLt_LIBRARY
+CUDA_cublasLt_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cublasLt_static_LIBRARY
+CUDA_cublasLt_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cublas_LIBRARY
+CUDA_cublas_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cublas_static_LIBRARY
+CUDA_cublas_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cuda_driver_LIBRARY
+CUDA_cuda_driver_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cudart_LIBRARY
+CUDA_cudart_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cudart_static_LIBRARY
+CUDA_cudart_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cudla_LIBRARY
+CUDA_cudla_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cufft_LIBRARY
+CUDA_cufft_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cufft_static_LIBRARY
+CUDA_cufft_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cufft_static_nocallback_LIBRARY
+CUDA_cufft_static_nocallback_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cufftw_LIBRARY
+CUDA_cufftw_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cufftw_static_LIBRARY
+CUDA_cufftw_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_culibos_LIBRARY
+CUDA_culibos_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cupti_LIBRARY
+CUDA_cupti_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cupti_static_LIBRARY
+CUDA_cupti_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_curand_LIBRARY
+CUDA_curand_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_curand_static_LIBRARY
+CUDA_curand_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cusolver_LIBRARY
+CUDA_cusolver_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cusolver_lapack_static_LIBRARY
+CUDA_cusolver_lapack_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cusolver_metis_static_LIBRARY
+CUDA_cusolver_metis_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cusolver_static_LIBRARY
+CUDA_cusolver_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cusparse_LIBRARY
+CUDA_cusparse_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_cusparse_static_LIBRARY
+CUDA_cusparse_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppc_LIBRARY
+CUDA_nppc_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppc_static_LIBRARY
+CUDA_nppc_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppial_LIBRARY
+CUDA_nppial_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppial_static_LIBRARY
+CUDA_nppial_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppicc_LIBRARY
+CUDA_nppicc_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppicc_static_LIBRARY
+CUDA_nppicc_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppicom_LIBRARY
+CUDA_nppicom_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppicom_static_LIBRARY
+CUDA_nppicom_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppidei_LIBRARY
+CUDA_nppidei_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppidei_static_LIBRARY
+CUDA_nppidei_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppif_LIBRARY
+CUDA_nppif_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppif_static_LIBRARY
+CUDA_nppif_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppig_LIBRARY
+CUDA_nppig_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppig_static_LIBRARY
+CUDA_nppig_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppim_LIBRARY
+CUDA_nppim_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppim_static_LIBRARY
+CUDA_nppim_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppist_LIBRARY
+CUDA_nppist_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppist_static_LIBRARY
+CUDA_nppist_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppisu_LIBRARY
+CUDA_nppisu_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppisu_static_LIBRARY
+CUDA_nppisu_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppitc_LIBRARY
+CUDA_nppitc_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nppitc_static_LIBRARY
+CUDA_nppitc_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_npps_LIBRARY
+CUDA_npps_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_npps_static_LIBRARY
+CUDA_npps_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvJitLink_LIBRARY
+CUDA_nvJitLink_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvJitLink_static_LIBRARY
+CUDA_nvJitLink_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvToolsExt_LIBRARY
+CUDA_nvToolsExt_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvfatbin_LIBRARY
+CUDA_nvfatbin_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvfatbin_static_LIBRARY
+CUDA_nvfatbin_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvgraph_LIBRARY
+CUDA_nvgraph_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvgraph_static_LIBRARY
+CUDA_nvgraph_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvjpeg_LIBRARY
+CUDA_nvjpeg_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvjpeg_static_LIBRARY
+CUDA_nvjpeg_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvml_LIBRARY
+CUDA_nvml_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvml_static_LIBRARY
+CUDA_nvml_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvperf_host_LIBRARY
+CUDA_nvperf_host_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvperf_host_static_LIBRARY
+CUDA_nvperf_host_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvperf_target_LIBRARY
+CUDA_nvperf_target_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvptxcompiler_static_LIBRARY
+CUDA_nvptxcompiler_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvrtc_LIBRARY
+CUDA_nvrtc_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvrtc_builtins_LIBRARY
+CUDA_nvrtc_builtins_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvrtc_builtins_static_LIBRARY
+CUDA_nvrtc_builtins_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_nvrtc_static_LIBRARY
+CUDA_nvrtc_static_LIBRARY-ADVANCED:INTERNAL=1
+//ADVANCED property for variable: CUDA_pcsamplingutil_LIBRARY
+CUDA_pcsamplingutil_LIBRARY-ADVANCED:INTERNAL=1
+//Details about finding CUDAToolkit
+FIND_PACKAGE_MESSAGE_DETAILS_CUDAToolkit:INTERNAL=[C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include;C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include/cccl][C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/cudart.lib][C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/bin][v13.0.48()]

CMakeLists.txt ADDED Viewed

	@@ -0,0 +1,45 @@

+cmake_minimum_required(VERSION 3.18)
+project(cancer_trainer LANGUAGES CXX CUDA)
+set(CMAKE_CXX_STANDARD 17)
+set(CMAKE_CUDA_STANDARD 17)
+if (MSVC)
+  add_compile_options(/O2 /W4)
+else()
+  add_compile_options(-O3 -ffast-math -fno-math-errno -fno-trapping-math -Wall -Wextra -Wno-unused-parameter)
+endif()
+enable_language(CUDA)
+set(CMAKE_CUDA_ARCHITECTURES 75 80 86) # Turing, Ampere
+find_package(CUDAToolkit REQUIRED)
+add_definitions(-D_FORCE_INLINES)
+# Tell CMake where to find stb_image.h
+target_include_directories(cancer_trainer PRIVATE vendor)
+file(GLOB SRC "src/*.cpp" "src/*.hpp" "src/*.cu")
+add_executable(cancer_trainer ${SRC})
+target_link_libraries(cancer_trainer PRIVATE CUDA::cudart CUDA::cufft)
+target_compile_options(cancer_trainer PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:-use_fast_math>)
+# Define STB_IMAGE_IMPLEMENTATION in only one compilation unit
+set_source_files_properties(src/data_loader.cpp PROPERTIES
+    COMPILE_DEFINITIONS STB_IMAGE_IMPLEMENTATION
+)```
+#### `vendor/stb_image.h`
+*Descarga y coloca el fichero `stb_image.h` de la [fuente oficial](https://github.com/nothings/stb/blob/master/stb_image.h) en esta carpeta.*
+#### `src/utils.hpp`
+```cpp
+#pragma once
+#include <string>
+#include <vector>
+#include <cuda_runtime.h>
+void check_cuda(cudaError_t st, const char* msg);
+void write_submission_csv(const std::string& path,
+                          const std::vector<std::string>& ids,
+                          const std::vector<float>& probabilities);

Structure.txt ADDED Viewed

	@@ -0,0 +1,90 @@

+cancer_detection/
+├─ CMakeLists.txt
+├─ README_RUN.md
+├─ vendor/
+│  └─ stb_image.h   <-- (Debes descargar y colocar este fichero aquí)
+├─ src/
+│  ├─ main.cpp
+│  ├─ data_loader.hpp
+│  ├─ data_loader.cpp
+│  ├─ optical_model.hpp
+│  ├─ optical_model.cu
+│  ├─ fungi.hpp
+│  ├─ fungi.cu
+│  ├─ training.hpp
+│  ├─ training.cpp
+│  └─ utils.hpp
+└─ data/
+   ├─ train_labels.csv
+   ├─ train/
+   │  ├─ abc...123.tif
+   │  └─ ...
+   └─ test/
+      ├─ def...456.tif
+      └─ ...
+Aquí tienes el programa completo en C++17/CUDA, diseñado para la competición Histopathologic Cancer Detection de Kaggle. Está totalmente en inglés, optimizado para una RTX 3090, y no utiliza ninguna capa convencional como CNNs o Transformers.
+Puntos Clave de la Adaptación
+Carga de Datos: El programa utiliza la librería stb_image.h (un solo fichero, sin dependencias) para leer las imágenes TIFF directamente desde el disco. Carga las etiquetas del train_labels.csv y asocia los IDs.
+Manejo de Color (RGB): Para mantener la simplicidad y la eficiencia del bloque óptico, las imágenes de 96x96x3 (RGB) se convierten a escala de grises (96x96x1) promediando los canales. Esto enfoca el modelo en la estructura y textura, que son clave en histopatología.
+Modelo de Clasificación: La salida es un único logit (un valor real), ya que es una clasificación binaria. Usamos una función de pérdida Binary Cross-Entropy (BCE), que es el estándar para esta tarea.
+Sistema Micelar: La población de "hongos" evoluciona para generar las máscaras ópticas de 96x96 que mejor discriminan entre tejido sano y tumoral. La "comida" se asigna en función de cómo cada hongo ayuda a reducir el error de clasificación.
+Generación de CSV por Época: Después de cada época de entrenamiento, el programa ejecuta la inferencia sobre todo el conjunto de test y genera un fichero submission_epoch_XX.csv listo para subir a Kaggle. Esto te permite monitorizar el progreso y elegir tu mejor modelo.
+📦 Estructura del Proyecto
+code
+Code
+cancer_detection/
+├─ CMakeLists.txt
+├─ README_RUN.md
+├─ vendor/
+│  └─ stb_image.h   <-- (Debes descargar y colocar este fichero aquí)
+├─ src/
+│  ├─ main.cpp
+│  ├─ data_loader.hpp
+│  ├─ data_loader.cpp
+│  ├─ optical_model.hpp
+│  ├─ optical_model.cu
+│  ├─ fungi.hpp
+│  ├─ fungi.cu
+│  ├─ training.hpp
+│  ├─ training.cpp
+│  └─ utils.hpp
+└─ data/
+   ├─ train_labels.csv
+   ├─ train/
+   │  ├─ abc...123.tif
+   │  └─ ...
+   └─ test/
+      ├─ def...456.tif
+      └─ ...
+🛠️ Build & Run (Instructions in English)
+(Contenido del fichero README_RUN.md)
+Mycelium-Optical Cancer Detection (C++/CUDA)
+This project implements a non-conventional optical/evolutionary model for the Kaggle Histopathologic Cancer Detection competition.
+Architecture:
+Input: 96x96 RGB images, converted to grayscale.
+Optical Block: A population of evolving "fungi" generates 96x96 amplitude and phase masks.
+Propagation: 2D FFT propagation using batched cuFFT.
+Detector: Intensity |U|^2 -> log1p non-linearity -> single logit output.
+Training: Custom backpropagation + Mycelial evolution (reward, gravity, reproduction).
+Loss: Binary Cross-Entropy with Logits.
+Requirements:
+Linux (recommended) or Windows with WSL2.
+CUDA Toolkit 11.4+ (optimized for Ampere, e.g., RTX 3090).
+A recent CMake (>= 3.18).
+Download stb_image.h from the official GitHub repo and place it in the vendor/ directory.
+Build:
+code
+Bash
+mkdir build && cd build
+cmake -DCMAKE_BUILD_TYPE=Release ..
+cmake --build . -j
+Run Training & Inference:
+Place the Kaggle data in the data/ directory as shown in the project layout.
+code
+Bash
+./cancer_trainer --data_dir ../data --epochs 50 --batch 256 --lr 1e-3 --fungi 256
+--data_dir: Path to the directory containing train/, test/, and train_labels.csv.
+--batch: Adjust based on VRAM. 256 works well on an RTX 3090 (24GB).
+--fungi: Number of fungi in the mycelial population.
+The program will save submission_epoch_01.csv, submission_epoch_02.csv, etc., in the execution directory after each epoch.

ZERO_CHECK.vcxproj ADDED Viewed

	@@ -0,0 +1,182 @@

+<?xml version="1.0" encoding="utf-8"?>
+<Project DefaultTargets="Build" ToolsVersion="17.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <PropertyGroup>
+    <PreferredToolArchitecture>x64</PreferredToolArchitecture>
+  </PropertyGroup>
+  <PropertyGroup>
+    <ResolveNugetPackages>false</ResolveNugetPackages>
+  </PropertyGroup>
+  <ItemGroup Label="ProjectConfigurations">
+    <ProjectConfiguration Include="Debug|x64">
+      <Configuration>Debug</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|x64">
+      <Configuration>Release</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="MinSizeRel|x64">
+      <Configuration>MinSizeRel</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="RelWithDebInfo|x64">
+      <Configuration>RelWithDebInfo</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+  </ItemGroup>
+  <PropertyGroup Label="Globals">
+    <ProjectGuid>{B4CD8D8C-908F-3E91-878E-AFB61C98E407}</ProjectGuid>
+    <Keyword>Win32Proj</Keyword>
+    <WindowsTargetPlatformVersion>10.0.26100.0</WindowsTargetPlatformVersion>
+    <Platform>x64</Platform>
+    <ProjectName>ZERO_CHECK</ProjectName>
+    <CudaToolkitCustomDir>C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0\</CudaToolkitCustomDir>
+    <VCProjectUpgraderObjectName>NoUpgrade</VCProjectUpgraderObjectName>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'" Label="Configuration">
+    <ConfigurationType>Utility</ConfigurationType>
+    <CharacterSet>MultiByte</CharacterSet>
+    <PlatformToolset>v143</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'" Label="Configuration">
+    <ConfigurationType>Utility</ConfigurationType>
+    <CharacterSet>MultiByte</CharacterSet>
+    <PlatformToolset>v143</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'" Label="Configuration">
+    <ConfigurationType>Utility</ConfigurationType>
+    <CharacterSet>MultiByte</CharacterSet>
+    <PlatformToolset>v143</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'" Label="Configuration">
+    <ConfigurationType>Utility</ConfigurationType>
+    <CharacterSet>MultiByte</CharacterSet>
+    <PlatformToolset>v143</PlatformToolset>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
+  <ImportGroup Label="ExtensionSettings">
+    <Import Project="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0\extras\visual_studio_integration\MSBuildExtensions\CUDA 13.0.props" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <PropertyGroup Label="UserMacros" />
+  <PropertyGroup>
+    <_ProjectFileVersion>10.0.20506.1</_ProjectFileVersion>
+    <IntDir Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">$(Platform)\$(Configuration)\$(ProjectName)\</IntDir>
+    <IntDir Condition="'$(Configuration)|$(Platform)'=='Release|x64'">$(Platform)\$(Configuration)\$(ProjectName)\</IntDir>
+    <IntDir Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">$(Platform)\$(Configuration)\$(ProjectName)\</IntDir>
+    <IntDir Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">$(Platform)\$(Configuration)\$(ProjectName)\</IntDir>
+  </PropertyGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+    <Midl>
+      <AdditionalIncludeDirectories>%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <OutputDirectory>$(ProjectDir)/$(IntDir)</OutputDirectory>
+      <HeaderFileName>%(Filename).h</HeaderFileName>
+      <TypeLibraryName>%(Filename).tlb</TypeLibraryName>
+      <InterfaceIdentifierFileName>%(Filename)_i.c</InterfaceIdentifierFileName>
+      <ProxyFileName>%(Filename)_p.c</ProxyFileName>
+    </Midl>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+    <Midl>
+      <AdditionalIncludeDirectories>%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <OutputDirectory>$(ProjectDir)/$(IntDir)</OutputDirectory>
+      <HeaderFileName>%(Filename).h</HeaderFileName>
+      <TypeLibraryName>%(Filename).tlb</TypeLibraryName>
+      <InterfaceIdentifierFileName>%(Filename)_i.c</InterfaceIdentifierFileName>
+      <ProxyFileName>%(Filename)_p.c</ProxyFileName>
+    </Midl>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">
+    <Midl>
+      <AdditionalIncludeDirectories>%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <OutputDirectory>$(ProjectDir)/$(IntDir)</OutputDirectory>
+      <HeaderFileName>%(Filename).h</HeaderFileName>
+      <TypeLibraryName>%(Filename).tlb</TypeLibraryName>
+      <InterfaceIdentifierFileName>%(Filename)_i.c</InterfaceIdentifierFileName>
+      <ProxyFileName>%(Filename)_p.c</ProxyFileName>
+    </Midl>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">
+    <Midl>
+      <AdditionalIncludeDirectories>%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <OutputDirectory>$(ProjectDir)/$(IntDir)</OutputDirectory>
+      <HeaderFileName>%(Filename).h</HeaderFileName>
+      <TypeLibraryName>%(Filename).tlb</TypeLibraryName>
+      <InterfaceIdentifierFileName>%(Filename)_i.c</InterfaceIdentifierFileName>
+      <ProxyFileName>%(Filename)_p.c</ProxyFileName>
+    </Midl>
+  </ItemDefinitionGroup>
+  <ItemGroup>
+    <CustomBuild Include="E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\6e114a7c9e7b8616a743eb7ed1a38c30\generate.stamp.rule">
+      <UseUtf8Encoding>Always</UseUtf8Encoding>
+      <BuildInParallel Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</BuildInParallel>
+      <Message Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">Checking Build System</Message>
+      <Command Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">setlocal
+"C:\Program Files\CMake\bin\cmake.exe" -SE:/Fashion_MNIST_Optic_Evolution -BE:/Fashion_MNIST_Optic_Evolution/build --check-stamp-list CMakeFiles/generate.stamp.list --vs-solution-file E:/Fashion_MNIST_Optic_Evolution/build/fashion_mnist_trainer.sln
+if %errorlevel% neq 0 goto :cmEnd
+:cmEnd
+endlocal &amp; call :cmErrorLevel %errorlevel% &amp; goto :cmDone
+:cmErrorLevel
+exit /b %1
+:cmDone
+if %errorlevel% neq 0 goto :VCEnd</Command>
+      <AdditionalInputs Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCUDAInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCXXInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCommonLanguageInclude.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeGenericSystem.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeInitializeConfigs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeLanguageInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeRCInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInitialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CheckLanguage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\CMakeCommonCompilerMacros.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindCUDAToolkit.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageHandleStandardArgs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageMessage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCUDALinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCXXLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCommonLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-Initialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\WindowsPaths.cmake;E:\Fashion_MNIST_Optic_Evolution\CMakeLists.txt;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCUDACompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCXXCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeRCCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeSystem.cmake;%(AdditionalInputs)</AdditionalInputs>
+      <Outputs Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\generate.stamp</Outputs>
+      <LinkObjects Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">false</LinkObjects>
+      <BuildInParallel Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</BuildInParallel>
+      <Message Condition="'$(Configuration)|$(Platform)'=='Release|x64'">Checking Build System</Message>
+      <Command Condition="'$(Configuration)|$(Platform)'=='Release|x64'">setlocal
+"C:\Program Files\CMake\bin\cmake.exe" -SE:/Fashion_MNIST_Optic_Evolution -BE:/Fashion_MNIST_Optic_Evolution/build --check-stamp-list CMakeFiles/generate.stamp.list --vs-solution-file E:/Fashion_MNIST_Optic_Evolution/build/fashion_mnist_trainer.sln
+if %errorlevel% neq 0 goto :cmEnd
+:cmEnd
+endlocal &amp; call :cmErrorLevel %errorlevel% &amp; goto :cmDone
+:cmErrorLevel
+exit /b %1
+:cmDone
+if %errorlevel% neq 0 goto :VCEnd</Command>
+      <AdditionalInputs Condition="'$(Configuration)|$(Platform)'=='Release|x64'">C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCUDAInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCXXInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCommonLanguageInclude.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeGenericSystem.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeInitializeConfigs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeLanguageInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeRCInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInitialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CheckLanguage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\CMakeCommonCompilerMacros.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindCUDAToolkit.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageHandleStandardArgs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageMessage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCUDALinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCXXLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCommonLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-Initialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\WindowsPaths.cmake;E:\Fashion_MNIST_Optic_Evolution\CMakeLists.txt;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCUDACompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCXXCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeRCCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeSystem.cmake;%(AdditionalInputs)</AdditionalInputs>
+      <Outputs Condition="'$(Configuration)|$(Platform)'=='Release|x64'">E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\generate.stamp</Outputs>
+      <LinkObjects Condition="'$(Configuration)|$(Platform)'=='Release|x64'">false</LinkObjects>
+      <BuildInParallel Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">true</BuildInParallel>
+      <Message Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">Checking Build System</Message>
+      <Command Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">setlocal
+"C:\Program Files\CMake\bin\cmake.exe" -SE:/Fashion_MNIST_Optic_Evolution -BE:/Fashion_MNIST_Optic_Evolution/build --check-stamp-list CMakeFiles/generate.stamp.list --vs-solution-file E:/Fashion_MNIST_Optic_Evolution/build/fashion_mnist_trainer.sln
+if %errorlevel% neq 0 goto :cmEnd
+:cmEnd
+endlocal &amp; call :cmErrorLevel %errorlevel% &amp; goto :cmDone
+:cmErrorLevel
+exit /b %1
+:cmDone
+if %errorlevel% neq 0 goto :VCEnd</Command>
+      <AdditionalInputs Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCUDAInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCXXInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCommonLanguageInclude.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeGenericSystem.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeInitializeConfigs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeLanguageInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeRCInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInitialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CheckLanguage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\CMakeCommonCompilerMacros.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindCUDAToolkit.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageHandleStandardArgs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageMessage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCUDALinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCXXLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCommonLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-Initialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\WindowsPaths.cmake;E:\Fashion_MNIST_Optic_Evolution\CMakeLists.txt;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCUDACompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCXXCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeRCCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeSystem.cmake;%(AdditionalInputs)</AdditionalInputs>
+      <Outputs Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\generate.stamp</Outputs>
+      <LinkObjects Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">false</LinkObjects>
+      <BuildInParallel Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">true</BuildInParallel>
+      <Message Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">Checking Build System</Message>
+      <Command Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">setlocal
+"C:\Program Files\CMake\bin\cmake.exe" -SE:/Fashion_MNIST_Optic_Evolution -BE:/Fashion_MNIST_Optic_Evolution/build --check-stamp-list CMakeFiles/generate.stamp.list --vs-solution-file E:/Fashion_MNIST_Optic_Evolution/build/fashion_mnist_trainer.sln
+if %errorlevel% neq 0 goto :cmEnd
+:cmEnd
+endlocal &amp; call :cmErrorLevel %errorlevel% &amp; goto :cmDone
+:cmErrorLevel
+exit /b %1
+:cmDone
+if %errorlevel% neq 0 goto :VCEnd</Command>
+      <AdditionalInputs Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCUDAInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCXXInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCommonLanguageInclude.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeGenericSystem.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeInitializeConfigs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeLanguageInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeRCInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInitialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CheckLanguage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\CMakeCommonCompilerMacros.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindCUDAToolkit.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageHandleStandardArgs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageMessage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCUDALinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCXXLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCommonLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-Initialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\WindowsPaths.cmake;E:\Fashion_MNIST_Optic_Evolution\CMakeLists.txt;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCUDACompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCXXCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeRCCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeSystem.cmake;%(AdditionalInputs)</AdditionalInputs>
+      <Outputs Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\generate.stamp</Outputs>
+      <LinkObjects Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">false</LinkObjects>
+    </CustomBuild>
+  </ItemGroup>
+  <ItemGroup>
+  </ItemGroup>
+  <ItemGroup />
+  <ItemGroup>
+  </ItemGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
+  <ImportGroup Label="ExtensionTargets">
+    <Import Project="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0\extras\visual_studio_integration\MSBuildExtensions\CUDA 13.0.targets" />
+  </ImportGroup>
+</Project>

ZERO_CHECK.vcxproj.filters ADDED Viewed

	@@ -0,0 +1,13 @@

+<?xml version="1.0" encoding="utf-8"?>
+<Project ToolsVersion="17.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <ItemGroup>
+    <CustomBuild Include="E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\6e114a7c9e7b8616a743eb7ed1a38c30\generate.stamp.rule">
+      <Filter>CMake Rules</Filter>
+    </CustomBuild>
+  </ItemGroup>
+  <ItemGroup>
+    <Filter Include="CMake Rules">
+      <UniqueIdentifier>{F8798282-D57C-369A-90EF-390B9340DAF2}</UniqueIdentifier>
+    </Filter>
+  </ItemGroup>
+</Project>

cmake_install.cmake ADDED Viewed

	@@ -0,0 +1,56 @@

+# Install script for directory: E:/Fashion_MNIST_Optic_Evolution
+# Set the install prefix
+if(NOT DEFINED CMAKE_INSTALL_PREFIX)
+  set(CMAKE_INSTALL_PREFIX "C:/Program Files")
+endif()
+string(REGEX REPLACE "/$" "" CMAKE_INSTALL_PREFIX "${CMAKE_INSTALL_PREFIX}")
+# Set the install configuration name.
+if(NOT DEFINED CMAKE_INSTALL_CONFIG_NAME)
+  if(BUILD_TYPE)
+    string(REGEX REPLACE "^[^A-Za-z0-9_]+" ""
+           CMAKE_INSTALL_CONFIG_NAME "${BUILD_TYPE}")
+  else()
+    set(CMAKE_INSTALL_CONFIG_NAME "Release")
+  endif()
+  message(STATUS "Install configuration: \"${CMAKE_INSTALL_CONFIG_NAME}\"")
+endif()
+# Set the component getting installed.
+if(NOT CMAKE_INSTALL_COMPONENT)
+  if(COMPONENT)
+    message(STATUS "Install component: \"${COMPONENT}\"")
+    set(CMAKE_INSTALL_COMPONENT "${COMPONENT}")
+  else()
+    set(CMAKE_INSTALL_COMPONENT)
+  endif()
+endif()
+# Is this installation the result of a crosscompile?
+if(NOT DEFINED CMAKE_CROSSCOMPILING)
+  set(CMAKE_CROSSCOMPILING "FALSE")
+endif()
+string(REPLACE ";" "\n" CMAKE_INSTALL_MANIFEST_CONTENT
+       "${CMAKE_INSTALL_MANIFEST_FILES}")
+if(CMAKE_INSTALL_LOCAL_ONLY)
+  file(WRITE "E:/Fashion_MNIST_Optic_Evolution/build/install_local_manifest.txt"
+     "${CMAKE_INSTALL_MANIFEST_CONTENT}")
+endif()
+if(CMAKE_INSTALL_COMPONENT)
+  if(CMAKE_INSTALL_COMPONENT MATCHES "^[a-zA-Z0-9_.+-]+$")
+    set(CMAKE_INSTALL_MANIFEST "install_manifest_${CMAKE_INSTALL_COMPONENT}.txt")
+  else()
+    string(MD5 CMAKE_INST_COMP_HASH "${CMAKE_INSTALL_COMPONENT}")
+    set(CMAKE_INSTALL_MANIFEST "install_manifest_${CMAKE_INST_COMP_HASH}.txt")
+    unset(CMAKE_INST_COMP_HASH)
+  endif()
+else()
+  set(CMAKE_INSTALL_MANIFEST "install_manifest.txt")
+endif()
+if(NOT CMAKE_INSTALL_LOCAL_ONLY)
+  file(WRITE "E:/Fashion_MNIST_Optic_Evolution/build/${CMAKE_INSTALL_MANIFEST}"
+     "${CMAKE_INSTALL_MANIFEST_CONTENT}")
+endif()

data_loader.cpp ADDED Viewed

	@@ -0,0 +1,35 @@

+#include "data_loader.hpp"
+#include <fstream>
+#include <stdexcept>
+#include <iostream>
+#include "optical_model.hpp" // For IMG_SIZE
+FashionMNISTSet load_fashion_mnist_data(const std::string& data_dir, bool is_train) {
+    FashionMNISTSet set;
+    const std::string prefix = is_train ? "train" : "test";
+    const std::string images_path = data_dir + "/" + prefix + "-images.bin";
+    const std::string labels_path = data_dir + "/" + prefix + "-labels.bin";
+    // Load images
+    std::ifstream f_images(images_path, std::ios::binary);
+    if (!f_images) throw std::runtime_error("Cannot open: " + images_path);
+    f_images.seekg(0, std::ios::end);
+    size_t num_bytes = f_images.tellg();
+    f_images.seekg(0, std::ios::beg);
+    set.N = num_bytes / (IMG_SIZE * sizeof(float));
+    set.images.resize(set.N * IMG_SIZE);
+    f_images.read(reinterpret_cast<char*>(set.images.data()), num_bytes);
+    // Load labels
+    std::ifstream f_labels(labels_path, std::ios::binary);
+    if (!f_labels) throw std::runtime_error("Cannot open: " + labels_path);
+    f_labels.seekg(0, std::ios::end);
+    num_bytes = f_labels.tellg();
+    f_labels.seekg(0, std::ios::beg);
+    if (set.N != num_bytes) throw std::runtime_error("Image and label count mismatch!");
+    set.labels.resize(set.N);
+    f_labels.read(reinterpret_cast<char*>(set.labels.data()), num_bytes);
+    std::cout << "[INFO] Loaded " << set.N << " " << prefix << " samples.\n";
+    return set;
+}

data_loader.hpp ADDED Viewed

	@@ -0,0 +1,14 @@

+#pragma once
+#include <string>
+#include <vector>
+struct FashionMNISTSet {
+    // Images as floats in [0,1], row-major, size = N * 784
+    std::vector<float> images;
+    // Labels in {0..9}, size = N
+    std::vector<uint8_t> labels;
+    size_t N = 0;
+};
+// Loads the entire Fashion-MNIST dataset from the binary files
+FashionMNISTSet load_fashion_mnist_data(const std::string& data_dir, bool is_train);

data_loader.obj ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:2b5725ee80384ae3d99cfaa5bbc8626576043e15368ec7c03d216d97547776a6
+size 111195

fashion_mnist_trainer.device-link.obj ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:af542334b1cebb545e362c72b5d6f5d83f6beec78fbdf04ac5f1e3bda06688ff
+size 1151471

fashion_mnist_trainer.exe ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:bae5556a38c217bef0d02bb369c6060c4d38265013bd82a4f6d8e58a090bc283
+size 2371584

fashion_mnist_trainer.exe.recipe ADDED Viewed

	@@ -0,0 +1,14 @@

+<?xml version="1.0" encoding="utf-8"?>
+<Project>
+  <ProjectOutputs>
+    <ProjectOutput>
+      <FullPath>E:\Fashion_MNIST_Optic_Evolution\build\x64\Release\ZERO_CHECK</FullPath>
+    </ProjectOutput>
+    <ProjectOutput>
+      <FullPath>E:\Fashion_MNIST_Optic_Evolution\build\Release\fashion_mnist_trainer.exe</FullPath>
+    </ProjectOutput>
+  </ProjectOutputs>
+  <ContentFiles />
+  <SatelliteDlls />
+  <NonRecipeFileRefs />
+</Project>

fashion_mnist_trainer.sln ADDED Viewed

	@@ -0,0 +1,53 @@

+Microsoft Visual Studio Solution File, Format Version 12.00
+# Visual Studio Version 17
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "ALL_BUILD", "ALL_BUILD.vcxproj", "{EF40E32F-9DB8-3F09-9CAA-07717AF7BF99}"
+	ProjectSection(ProjectDependencies) = postProject
+		{B4CD8D8C-908F-3E91-878E-AFB61C98E407} = {B4CD8D8C-908F-3E91-878E-AFB61C98E407}
+		{2B82790A-4403-3D6B-A562-F30C253F0098} = {2B82790A-4403-3D6B-A562-F30C253F0098}
+	EndProjectSection
+EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "ZERO_CHECK", "ZERO_CHECK.vcxproj", "{B4CD8D8C-908F-3E91-878E-AFB61C98E407}"
+	ProjectSection(ProjectDependencies) = postProject
+	EndProjectSection
+EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "fashion_mnist_trainer", "fashion_mnist_trainer.vcxproj", "{2B82790A-4403-3D6B-A562-F30C253F0098}"
+	ProjectSection(ProjectDependencies) = postProject
+		{B4CD8D8C-908F-3E91-878E-AFB61C98E407} = {B4CD8D8C-908F-3E91-878E-AFB61C98E407}
+	EndProjectSection
+EndProject
+Global
+	GlobalSection(SolutionConfigurationPlatforms) = preSolution
+		Debug|x64 = Debug|x64
+		Release|x64 = Release|x64
+		MinSizeRel|x64 = MinSizeRel|x64
+		RelWithDebInfo|x64 = RelWithDebInfo|x64
+	EndGlobalSection
+	GlobalSection(ProjectConfigurationPlatforms) = postSolution
+		{EF40E32F-9DB8-3F09-9CAA-07717AF7BF99}.Debug|x64.ActiveCfg = Debug|x64
+		{EF40E32F-9DB8-3F09-9CAA-07717AF7BF99}.Release|x64.ActiveCfg = Release|x64
+		{EF40E32F-9DB8-3F09-9CAA-07717AF7BF99}.MinSizeRel|x64.ActiveCfg = MinSizeRel|x64
+		{EF40E32F-9DB8-3F09-9CAA-07717AF7BF99}.RelWithDebInfo|x64.ActiveCfg = RelWithDebInfo|x64
+		{B4CD8D8C-908F-3E91-878E-AFB61C98E407}.Debug|x64.ActiveCfg = Debug|x64
+		{B4CD8D8C-908F-3E91-878E-AFB61C98E407}.Debug|x64.Build.0 = Debug|x64
+		{B4CD8D8C-908F-3E91-878E-AFB61C98E407}.Release|x64.ActiveCfg = Release|x64
+		{B4CD8D8C-908F-3E91-878E-AFB61C98E407}.Release|x64.Build.0 = Release|x64
+		{B4CD8D8C-908F-3E91-878E-AFB61C98E407}.MinSizeRel|x64.ActiveCfg = MinSizeRel|x64
+		{B4CD8D8C-908F-3E91-878E-AFB61C98E407}.MinSizeRel|x64.Build.0 = MinSizeRel|x64
+		{B4CD8D8C-908F-3E91-878E-AFB61C98E407}.RelWithDebInfo|x64.ActiveCfg = RelWithDebInfo|x64
+		{B4CD8D8C-908F-3E91-878E-AFB61C98E407}.RelWithDebInfo|x64.Build.0 = RelWithDebInfo|x64
+		{2B82790A-4403-3D6B-A562-F30C253F0098}.Debug|x64.ActiveCfg = Debug|x64
+		{2B82790A-4403-3D6B-A562-F30C253F0098}.Debug|x64.Build.0 = Debug|x64
+		{2B82790A-4403-3D6B-A562-F30C253F0098}.Release|x64.ActiveCfg = Release|x64
+		{2B82790A-4403-3D6B-A562-F30C253F0098}.Release|x64.Build.0 = Release|x64
+		{2B82790A-4403-3D6B-A562-F30C253F0098}.MinSizeRel|x64.ActiveCfg = MinSizeRel|x64
+		{2B82790A-4403-3D6B-A562-F30C253F0098}.MinSizeRel|x64.Build.0 = MinSizeRel|x64
+		{2B82790A-4403-3D6B-A562-F30C253F0098}.RelWithDebInfo|x64.ActiveCfg = RelWithDebInfo|x64
+		{2B82790A-4403-3D6B-A562-F30C253F0098}.RelWithDebInfo|x64.Build.0 = RelWithDebInfo|x64
+	EndGlobalSection
+	GlobalSection(ExtensibilityGlobals) = postSolution
+		SolutionGuid = {547978A7-D2DA-34ED-A106-C0E082D229F4}
+	EndGlobalSection
+	GlobalSection(ExtensibilityAddIns) = postSolution
+	EndGlobalSection
+EndGlobal

fashion_mnist_trainer.vcxproj ADDED Viewed

	@@ -0,0 +1,444 @@

+<?xml version="1.0" encoding="utf-8"?>
+<Project DefaultTargets="Build" ToolsVersion="17.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <PropertyGroup>
+    <PreferredToolArchitecture>x64</PreferredToolArchitecture>
+  </PropertyGroup>
+  <ItemGroup Label="ProjectConfigurations">
+    <ProjectConfiguration Include="Debug|x64">
+      <Configuration>Debug</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="Release|x64">
+      <Configuration>Release</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="MinSizeRel|x64">
+      <Configuration>MinSizeRel</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+    <ProjectConfiguration Include="RelWithDebInfo|x64">
+      <Configuration>RelWithDebInfo</Configuration>
+      <Platform>x64</Platform>
+    </ProjectConfiguration>
+  </ItemGroup>
+  <PropertyGroup Label="Globals">
+    <ProjectGuid>{2B82790A-4403-3D6B-A562-F30C253F0098}</ProjectGuid>
+    <Keyword>Win32Proj</Keyword>
+    <WindowsTargetPlatformVersion>10.0.26100.0</WindowsTargetPlatformVersion>
+    <Platform>x64</Platform>
+    <ProjectName>fashion_mnist_trainer</ProjectName>
+    <CudaToolkitCustomDir>C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0\</CudaToolkitCustomDir>
+    <VCProjectUpgraderObjectName>NoUpgrade</VCProjectUpgraderObjectName>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <CharacterSet>MultiByte</CharacterSet>
+    <PlatformToolset>v143</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <CharacterSet>MultiByte</CharacterSet>
+    <PlatformToolset>v143</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <CharacterSet>MultiByte</CharacterSet>
+    <PlatformToolset>v143</PlatformToolset>
+  </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'" Label="Configuration">
+    <ConfigurationType>Application</ConfigurationType>
+    <CharacterSet>MultiByte</CharacterSet>
+    <PlatformToolset>v143</PlatformToolset>
+  </PropertyGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
+  <ImportGroup Label="ExtensionSettings">
+    <Import Project="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0\extras\visual_studio_integration\MSBuildExtensions\CUDA 13.0.props" />
+  </ImportGroup>
+  <ImportGroup Label="PropertySheets">
+    <Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
+  </ImportGroup>
+  <PropertyGroup Label="UserMacros" />
+  <PropertyGroup>
+    <_ProjectFileVersion>10.0.20506.1</_ProjectFileVersion>
+    <OutDir Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">E:\Fashion_MNIST_Optic_Evolution\build\Debug\</OutDir>
+    <IntDir Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">fashion_mnist_trainer.dir\Debug\</IntDir>
+    <TargetName Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">fashion_mnist_trainer</TargetName>
+    <TargetExt Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">.exe</TargetExt>
+    <LinkIncremental Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</LinkIncremental>
+    <GenerateManifest Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</GenerateManifest>
+    <OutDir Condition="'$(Configuration)|$(Platform)'=='Release|x64'">E:\Fashion_MNIST_Optic_Evolution\build\Release\</OutDir>
+    <IntDir Condition="'$(Configuration)|$(Platform)'=='Release|x64'">fashion_mnist_trainer.dir\Release\</IntDir>
+    <TargetName Condition="'$(Configuration)|$(Platform)'=='Release|x64'">fashion_mnist_trainer</TargetName>
+    <TargetExt Condition="'$(Configuration)|$(Platform)'=='Release|x64'">.exe</TargetExt>
+    <LinkIncremental Condition="'$(Configuration)|$(Platform)'=='Release|x64'">false</LinkIncremental>
+    <GenerateManifest Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</GenerateManifest>
+    <OutDir Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">E:\Fashion_MNIST_Optic_Evolution\build\MinSizeRel\</OutDir>
+    <IntDir Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">fashion_mnist_trainer.dir\MinSizeRel\</IntDir>
+    <TargetName Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">fashion_mnist_trainer</TargetName>
+    <TargetExt Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">.exe</TargetExt>
+    <LinkIncremental Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">false</LinkIncremental>
+    <GenerateManifest Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">true</GenerateManifest>
+    <OutDir Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">E:\Fashion_MNIST_Optic_Evolution\build\RelWithDebInfo\</OutDir>
+    <IntDir Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">fashion_mnist_trainer.dir\RelWithDebInfo\</IntDir>
+    <TargetName Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">fashion_mnist_trainer</TargetName>
+    <TargetExt Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">.exe</TargetExt>
+    <LinkIncremental Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">true</LinkIncremental>
+    <GenerateManifest Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">true</GenerateManifest>
+  </PropertyGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
+    <ClCompile>
+      <AdditionalIncludeDirectories>E:\Fashion_MNIST_Optic_Evolution\src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalOptions>%(AdditionalOptions) /external:I "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include" /external:I "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include/cccl" -O3 -ffast-math</AdditionalOptions>
+      <AssemblerListingLocation>$(IntDir)</AssemblerListingLocation>
+      <BasicRuntimeChecks>EnableFastChecks</BasicRuntimeChecks>
+      <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
+      <ExceptionHandling>Sync</ExceptionHandling>
+      <ExternalWarningLevel>TurnOffAllWarnings</ExternalWarningLevel>
+      <InlineFunctionExpansion>Disabled</InlineFunctionExpansion>
+      <LanguageStandard>stdcpp17</LanguageStandard>
+      <MinimalRebuild></MinimalRebuild>
+      <Optimization>Disabled</Optimization>
+      <PrecompiledHeader>NotUsing</PrecompiledHeader>
+      <RuntimeLibrary>MultiThreadedDebugDLL</RuntimeLibrary>
+      <RuntimeTypeInfo>true</RuntimeTypeInfo>
+      <SupportJustMyCode></SupportJustMyCode>
+      <UseFullPaths>false</UseFullPaths>
+      <PreprocessorDefinitions>%(PreprocessorDefinitions);WIN32;_WINDOWS;CMAKE_INTDIR="Debug"</PreprocessorDefinitions>
+      <ObjectFileName>$(IntDir)</ObjectFileName>
+      <ScanSourceForModuleDependencies>false</ScanSourceForModuleDependencies>
+    </ClCompile>
+    <ResourceCompile>
+      <PreprocessorDefinitions>%(PreprocessorDefinitions);WIN32;_DEBUG;_WINDOWS;CMAKE_INTDIR=\"Debug\"</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>E:\Fashion_MNIST_Optic_Evolution\src;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cccl;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+    </ResourceCompile>
+    <CudaCompile>
+      <Include>E:\Fashion_MNIST_Optic_Evolution\src;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cccl;%(Include)</Include>
+      <Defines>%(Defines);_WINDOWS;CMAKE_INTDIR="Debug"</Defines>
+      <AdditionalOptions>%(AdditionalOptions) -std=c++17 --generate-code=arch=compute_75,code=[compute_75,sm_75] --generate-code=arch=compute_80,code=[compute_80,sm_80] --generate-code=arch=compute_86,code=[compute_86,sm_86] -Xcompiler="/EHsc -Zi -Ob0"</AdditionalOptions>
+      <CodeGeneration></CodeGeneration>
+      <CompileOut>$(IntDir)%(Filename).obj</CompileOut>
+      <CudaRuntime>Static</CudaRuntime>
+      <FastMath>true</FastMath>
+      <GPUDebugInfo>false</GPUDebugInfo>
+      <GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
+      <Optimization>Od</Optimization>
+      <Runtime>MDd</Runtime>
+      <RuntimeChecks>RTC1</RuntimeChecks>
+      <TargetMachinePlatform>64</TargetMachinePlatform>
+      <TypeInfo>true</TypeInfo>
+      <UseHostInclude>false</UseHostInclude>
+    </CudaCompile>
+    <Midl>
+      <AdditionalIncludeDirectories>E:\Fashion_MNIST_Optic_Evolution\src;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cccl;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <OutputDirectory>$(ProjectDir)/$(IntDir)</OutputDirectory>
+      <HeaderFileName>%(Filename).h</HeaderFileName>
+      <TypeLibraryName>%(Filename).tlb</TypeLibraryName>
+      <InterfaceIdentifierFileName>%(Filename)_i.c</InterfaceIdentifierFileName>
+      <ProxyFileName>%(Filename)_p.c</ProxyFileName>
+    </Midl>
+    <Link>
+      <AdditionalDependencies>C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\lib\x64\cudart.lib;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\lib\x64\cufft.lib;cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;comdlg32.lib;advapi32.lib</AdditionalDependencies>
+      <AdditionalLibraryDirectories>C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64;C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/$(Configuration);%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+      <IgnoreSpecificDefaultLibraries>%(IgnoreSpecificDefaultLibraries)</IgnoreSpecificDefaultLibraries>
+      <ImportLibrary>E:/Fashion_MNIST_Optic_Evolution/build/Debug/fashion_mnist_trainer.lib</ImportLibrary>
+      <ProgramDataBaseFile>E:/Fashion_MNIST_Optic_Evolution/build/Debug/fashion_mnist_trainer.pdb</ProgramDataBaseFile>
+      <SubSystem>Console</SubSystem>
+    </Link>
+    <ProjectReference>
+      <LinkLibraryDependencies>false</LinkLibraryDependencies>
+    </ProjectReference>
+    <CudaLink>
+      <AdditionalOptions>-forward-unknown-to-host-compiler -Wno-deprecated-gpu-targets -D_WINDOWS -Xcompiler=" /GR /EHsc" -use_fast_math -Xcompiler=" -Zi -Ob0 -Od /RTC1" "--generate-code=arch=compute_75,code=[compute_75,sm_75]" "--generate-code=arch=compute_80,code=[compute_80,sm_80]" "--generate-code=arch=compute_86,code=[compute_86,sm_86]" -Xcompiler=-MDd</AdditionalOptions>
+      <PerformDeviceLink>true</PerformDeviceLink>
+    </CudaLink>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
+    <ClCompile>
+      <AdditionalIncludeDirectories>E:\Fashion_MNIST_Optic_Evolution\src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalOptions>%(AdditionalOptions) /external:I "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include" /external:I "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include/cccl" -O3 -ffast-math</AdditionalOptions>
+      <AssemblerListingLocation>$(IntDir)</AssemblerListingLocation>
+      <BasicRuntimeChecks>Default</BasicRuntimeChecks>
+      <ExceptionHandling>Sync</ExceptionHandling>
+      <ExternalWarningLevel>TurnOffAllWarnings</ExternalWarningLevel>
+      <InlineFunctionExpansion>AnySuitable</InlineFunctionExpansion>
+      <LanguageStandard>stdcpp17</LanguageStandard>
+      <MinimalRebuild></MinimalRebuild>
+      <Optimization>MaxSpeed</Optimization>
+      <PrecompiledHeader>NotUsing</PrecompiledHeader>
+      <RuntimeLibrary>MultiThreadedDLL</RuntimeLibrary>
+      <RuntimeTypeInfo>true</RuntimeTypeInfo>
+      <SupportJustMyCode></SupportJustMyCode>
+      <UseFullPaths>false</UseFullPaths>
+      <PreprocessorDefinitions>%(PreprocessorDefinitions);WIN32;_WINDOWS;NDEBUG;CMAKE_INTDIR="Release"</PreprocessorDefinitions>
+      <ObjectFileName>$(IntDir)</ObjectFileName>
+      <DebugInformationFormat>
+      </DebugInformationFormat>
+      <ScanSourceForModuleDependencies>false</ScanSourceForModuleDependencies>
+    </ClCompile>
+    <ResourceCompile>
+      <PreprocessorDefinitions>%(PreprocessorDefinitions);WIN32;_WINDOWS;NDEBUG;CMAKE_INTDIR=\"Release\"</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>E:\Fashion_MNIST_Optic_Evolution\src;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cccl;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+    </ResourceCompile>
+    <CudaCompile>
+      <Include>E:\Fashion_MNIST_Optic_Evolution\src;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cccl;%(Include)</Include>
+      <Defines>%(Defines);_WINDOWS;NDEBUG;CMAKE_INTDIR="Release"</Defines>
+      <AdditionalOptions>%(AdditionalOptions) -std=c++17 --generate-code=arch=compute_75,code=[compute_75,sm_75] --generate-code=arch=compute_80,code=[compute_80,sm_80] --generate-code=arch=compute_86,code=[compute_86,sm_86] -Xcompiler="/EHsc -Ob2"</AdditionalOptions>
+      <CodeGeneration></CodeGeneration>
+      <CompileOut>$(IntDir)%(Filename).obj</CompileOut>
+      <CudaRuntime>Static</CudaRuntime>
+      <FastMath>true</FastMath>
+      <GPUDebugInfo>false</GPUDebugInfo>
+      <GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
+      <Optimization>O2</Optimization>
+      <Runtime>MD</Runtime>
+      <TargetMachinePlatform>64</TargetMachinePlatform>
+      <TypeInfo>true</TypeInfo>
+      <UseHostInclude>false</UseHostInclude>
+    </CudaCompile>
+    <Midl>
+      <AdditionalIncludeDirectories>E:\Fashion_MNIST_Optic_Evolution\src;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cccl;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <OutputDirectory>$(ProjectDir)/$(IntDir)</OutputDirectory>
+      <HeaderFileName>%(Filename).h</HeaderFileName>
+      <TypeLibraryName>%(Filename).tlb</TypeLibraryName>
+      <InterfaceIdentifierFileName>%(Filename)_i.c</InterfaceIdentifierFileName>
+      <ProxyFileName>%(Filename)_p.c</ProxyFileName>
+    </Midl>
+    <Link>
+      <AdditionalDependencies>C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\lib\x64\cudart.lib;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\lib\x64\cufft.lib;cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;comdlg32.lib;advapi32.lib</AdditionalDependencies>
+      <AdditionalLibraryDirectories>C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64;C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/$(Configuration);%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
+      <GenerateDebugInformation>false</GenerateDebugInformation>
+      <IgnoreSpecificDefaultLibraries>%(IgnoreSpecificDefaultLibraries)</IgnoreSpecificDefaultLibraries>
+      <ImportLibrary>E:/Fashion_MNIST_Optic_Evolution/build/Release/fashion_mnist_trainer.lib</ImportLibrary>
+      <ProgramDataBaseFile>E:/Fashion_MNIST_Optic_Evolution/build/Release/fashion_mnist_trainer.pdb</ProgramDataBaseFile>
+      <SubSystem>Console</SubSystem>
+    </Link>
+    <ProjectReference>
+      <LinkLibraryDependencies>false</LinkLibraryDependencies>
+    </ProjectReference>
+    <CudaLink>
+      <AdditionalOptions>-forward-unknown-to-host-compiler -Wno-deprecated-gpu-targets -D_WINDOWS -Xcompiler=" /GR /EHsc" -use_fast_math -Xcompiler="-O2 -Ob2" -DNDEBUG "--generate-code=arch=compute_75,code=[compute_75,sm_75]" "--generate-code=arch=compute_80,code=[compute_80,sm_80]" "--generate-code=arch=compute_86,code=[compute_86,sm_86]" -Xcompiler=-MD</AdditionalOptions>
+      <PerformDeviceLink>true</PerformDeviceLink>
+    </CudaLink>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">
+    <ClCompile>
+      <AdditionalIncludeDirectories>E:\Fashion_MNIST_Optic_Evolution\src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalOptions>%(AdditionalOptions) /external:I "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include" /external:I "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include/cccl" -O3 -ffast-math</AdditionalOptions>
+      <AssemblerListingLocation>$(IntDir)</AssemblerListingLocation>
+      <BasicRuntimeChecks>Default</BasicRuntimeChecks>
+      <ExceptionHandling>Sync</ExceptionHandling>
+      <ExternalWarningLevel>TurnOffAllWarnings</ExternalWarningLevel>
+      <InlineFunctionExpansion>OnlyExplicitInline</InlineFunctionExpansion>
+      <LanguageStandard>stdcpp17</LanguageStandard>
+      <MinimalRebuild></MinimalRebuild>
+      <Optimization>MinSpace</Optimization>
+      <PrecompiledHeader>NotUsing</PrecompiledHeader>
+      <RuntimeLibrary>MultiThreadedDLL</RuntimeLibrary>
+      <RuntimeTypeInfo>true</RuntimeTypeInfo>
+      <SupportJustMyCode></SupportJustMyCode>
+      <UseFullPaths>false</UseFullPaths>
+      <PreprocessorDefinitions>%(PreprocessorDefinitions);WIN32;_WINDOWS;NDEBUG;CMAKE_INTDIR="MinSizeRel"</PreprocessorDefinitions>
+      <ObjectFileName>$(IntDir)</ObjectFileName>
+      <DebugInformationFormat>
+      </DebugInformationFormat>
+      <ScanSourceForModuleDependencies>false</ScanSourceForModuleDependencies>
+    </ClCompile>
+    <ResourceCompile>
+      <PreprocessorDefinitions>%(PreprocessorDefinitions);WIN32;_WINDOWS;NDEBUG;CMAKE_INTDIR=\"MinSizeRel\"</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>E:\Fashion_MNIST_Optic_Evolution\src;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cccl;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+    </ResourceCompile>
+    <CudaCompile>
+      <Include>E:\Fashion_MNIST_Optic_Evolution\src;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cccl;%(Include)</Include>
+      <Defines>%(Defines);_WINDOWS;NDEBUG;CMAKE_INTDIR="MinSizeRel"</Defines>
+      <AdditionalOptions>%(AdditionalOptions) -std=c++17 --generate-code=arch=compute_75,code=[compute_75,sm_75] --generate-code=arch=compute_80,code=[compute_80,sm_80] --generate-code=arch=compute_86,code=[compute_86,sm_86] -Xcompiler="/EHsc -Ob1"</AdditionalOptions>
+      <CodeGeneration></CodeGeneration>
+      <CompileOut>$(IntDir)%(Filename).obj</CompileOut>
+      <CudaRuntime>Static</CudaRuntime>
+      <FastMath>true</FastMath>
+      <GPUDebugInfo>false</GPUDebugInfo>
+      <GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
+      <Optimization>O1</Optimization>
+      <Runtime>MD</Runtime>
+      <TargetMachinePlatform>64</TargetMachinePlatform>
+      <TypeInfo>true</TypeInfo>
+      <UseHostInclude>false</UseHostInclude>
+    </CudaCompile>
+    <Midl>
+      <AdditionalIncludeDirectories>E:\Fashion_MNIST_Optic_Evolution\src;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cccl;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <OutputDirectory>$(ProjectDir)/$(IntDir)</OutputDirectory>
+      <HeaderFileName>%(Filename).h</HeaderFileName>
+      <TypeLibraryName>%(Filename).tlb</TypeLibraryName>
+      <InterfaceIdentifierFileName>%(Filename)_i.c</InterfaceIdentifierFileName>
+      <ProxyFileName>%(Filename)_p.c</ProxyFileName>
+    </Midl>
+    <Link>
+      <AdditionalDependencies>C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\lib\x64\cudart.lib;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\lib\x64\cufft.lib;cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;comdlg32.lib;advapi32.lib</AdditionalDependencies>
+      <AdditionalLibraryDirectories>C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64;C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/$(Configuration);%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
+      <GenerateDebugInformation>false</GenerateDebugInformation>
+      <IgnoreSpecificDefaultLibraries>%(IgnoreSpecificDefaultLibraries)</IgnoreSpecificDefaultLibraries>
+      <ImportLibrary>E:/Fashion_MNIST_Optic_Evolution/build/MinSizeRel/fashion_mnist_trainer.lib</ImportLibrary>
+      <ProgramDataBaseFile>E:/Fashion_MNIST_Optic_Evolution/build/MinSizeRel/fashion_mnist_trainer.pdb</ProgramDataBaseFile>
+      <SubSystem>Console</SubSystem>
+    </Link>
+    <ProjectReference>
+      <LinkLibraryDependencies>false</LinkLibraryDependencies>
+    </ProjectReference>
+    <CudaLink>
+      <AdditionalOptions>-forward-unknown-to-host-compiler -Wno-deprecated-gpu-targets -D_WINDOWS -Xcompiler=" /GR /EHsc" -use_fast_math -Xcompiler="-O1 -Ob1" -DNDEBUG "--generate-code=arch=compute_75,code=[compute_75,sm_75]" "--generate-code=arch=compute_80,code=[compute_80,sm_80]" "--generate-code=arch=compute_86,code=[compute_86,sm_86]" -Xcompiler=-MD</AdditionalOptions>
+      <PerformDeviceLink>true</PerformDeviceLink>
+    </CudaLink>
+  </ItemDefinitionGroup>
+  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">
+    <ClCompile>
+      <AdditionalIncludeDirectories>E:\Fashion_MNIST_Optic_Evolution\src;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalOptions>%(AdditionalOptions) /external:I "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include" /external:I "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/include/cccl" -O3 -ffast-math</AdditionalOptions>
+      <AssemblerListingLocation>$(IntDir)</AssemblerListingLocation>
+      <BasicRuntimeChecks>Default</BasicRuntimeChecks>
+      <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
+      <ExceptionHandling>Sync</ExceptionHandling>
+      <ExternalWarningLevel>TurnOffAllWarnings</ExternalWarningLevel>
+      <InlineFunctionExpansion>OnlyExplicitInline</InlineFunctionExpansion>
+      <LanguageStandard>stdcpp17</LanguageStandard>
+      <MinimalRebuild></MinimalRebuild>
+      <Optimization>MaxSpeed</Optimization>
+      <PrecompiledHeader>NotUsing</PrecompiledHeader>
+      <RuntimeLibrary>MultiThreadedDLL</RuntimeLibrary>
+      <RuntimeTypeInfo>true</RuntimeTypeInfo>
+      <SupportJustMyCode></SupportJustMyCode>
+      <UseFullPaths>false</UseFullPaths>
+      <PreprocessorDefinitions>%(PreprocessorDefinitions);WIN32;_WINDOWS;NDEBUG;CMAKE_INTDIR="RelWithDebInfo"</PreprocessorDefinitions>
+      <ObjectFileName>$(IntDir)</ObjectFileName>
+      <ScanSourceForModuleDependencies>false</ScanSourceForModuleDependencies>
+    </ClCompile>
+    <ResourceCompile>
+      <PreprocessorDefinitions>%(PreprocessorDefinitions);WIN32;_WINDOWS;NDEBUG;CMAKE_INTDIR=\"RelWithDebInfo\"</PreprocessorDefinitions>
+      <AdditionalIncludeDirectories>E:\Fashion_MNIST_Optic_Evolution\src;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cccl;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+    </ResourceCompile>
+    <CudaCompile>
+      <Include>E:\Fashion_MNIST_Optic_Evolution\src;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cccl;%(Include)</Include>
+      <Defines>%(Defines);_WINDOWS;NDEBUG;CMAKE_INTDIR="RelWithDebInfo"</Defines>
+      <AdditionalOptions>%(AdditionalOptions) -std=c++17 --generate-code=arch=compute_75,code=[compute_75,sm_75] --generate-code=arch=compute_80,code=[compute_80,sm_80] --generate-code=arch=compute_86,code=[compute_86,sm_86] -Xcompiler="/EHsc -Zi -Ob1"</AdditionalOptions>
+      <CodeGeneration></CodeGeneration>
+      <CompileOut>$(IntDir)%(Filename).obj</CompileOut>
+      <CudaRuntime>Static</CudaRuntime>
+      <FastMath>true</FastMath>
+      <GPUDebugInfo>false</GPUDebugInfo>
+      <GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
+      <Optimization>O2</Optimization>
+      <Runtime>MD</Runtime>
+      <TargetMachinePlatform>64</TargetMachinePlatform>
+      <TypeInfo>true</TypeInfo>
+      <UseHostInclude>false</UseHostInclude>
+    </CudaCompile>
+    <Midl>
+      <AdditionalIncludeDirectories>E:\Fashion_MNIST_Optic_Evolution\src;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cccl;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <OutputDirectory>$(ProjectDir)/$(IntDir)</OutputDirectory>
+      <HeaderFileName>%(Filename).h</HeaderFileName>
+      <TypeLibraryName>%(Filename).tlb</TypeLibraryName>
+      <InterfaceIdentifierFileName>%(Filename)_i.c</InterfaceIdentifierFileName>
+      <ProxyFileName>%(Filename)_p.c</ProxyFileName>
+    </Midl>
+    <Link>
+      <AdditionalDependencies>C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\lib\x64\cudart.lib;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\lib\x64\cufft.lib;cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;comdlg32.lib;advapi32.lib</AdditionalDependencies>
+      <AdditionalLibraryDirectories>C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64;C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0/lib/x64/$(Configuration);%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
+      <GenerateDebugInformation>true</GenerateDebugInformation>
+      <IgnoreSpecificDefaultLibraries>%(IgnoreSpecificDefaultLibraries)</IgnoreSpecificDefaultLibraries>
+      <ImportLibrary>E:/Fashion_MNIST_Optic_Evolution/build/RelWithDebInfo/fashion_mnist_trainer.lib</ImportLibrary>
+      <ProgramDataBaseFile>E:/Fashion_MNIST_Optic_Evolution/build/RelWithDebInfo/fashion_mnist_trainer.pdb</ProgramDataBaseFile>
+      <SubSystem>Console</SubSystem>
+    </Link>
+    <ProjectReference>
+      <LinkLibraryDependencies>false</LinkLibraryDependencies>
+    </ProjectReference>
+    <CudaLink>
+      <AdditionalOptions>-forward-unknown-to-host-compiler -Wno-deprecated-gpu-targets -D_WINDOWS -Xcompiler=" /GR /EHsc" -use_fast_math -Xcompiler=" -Zi -O2 -Ob1" -DNDEBUG "--generate-code=arch=compute_75,code=[compute_75,sm_75]" "--generate-code=arch=compute_80,code=[compute_80,sm_80]" "--generate-code=arch=compute_86,code=[compute_86,sm_86]" -Xcompiler=-MD</AdditionalOptions>
+      <PerformDeviceLink>true</PerformDeviceLink>
+    </CudaLink>
+  </ItemDefinitionGroup>
+  <ItemGroup>
+    <CustomBuild Include="E:\Fashion_MNIST_Optic_Evolution\CMakeLists.txt">
+      <UseUtf8Encoding>Always</UseUtf8Encoding>
+      <Message Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">Building Custom Rule E:/Fashion_MNIST_Optic_Evolution/CMakeLists.txt</Message>
+      <Command Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">setlocal
+"C:\Program Files\CMake\bin\cmake.exe" -SE:/Fashion_MNIST_Optic_Evolution -BE:/Fashion_MNIST_Optic_Evolution/build --check-stamp-file E:/Fashion_MNIST_Optic_Evolution/build/CMakeFiles/generate.stamp
+if %errorlevel% neq 0 goto :cmEnd
+:cmEnd
+endlocal &amp; call :cmErrorLevel %errorlevel% &amp; goto :cmDone
+:cmErrorLevel
+exit /b %1
+:cmDone
+if %errorlevel% neq 0 goto :VCEnd</Command>
+      <AdditionalInputs Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCUDAInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCXXInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCommonLanguageInclude.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeGenericSystem.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeInitializeConfigs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeLanguageInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeRCInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInitialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CheckLanguage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\CMakeCommonCompilerMacros.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindCUDAToolkit.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageHandleStandardArgs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageMessage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCUDALinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCXXLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCommonLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-Initialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\WindowsPaths.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCUDACompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCXXCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeRCCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeSystem.cmake;%(AdditionalInputs)</AdditionalInputs>
+      <Outputs Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\generate.stamp</Outputs>
+      <LinkObjects Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">false</LinkObjects>
+      <Message Condition="'$(Configuration)|$(Platform)'=='Release|x64'">Building Custom Rule E:/Fashion_MNIST_Optic_Evolution/CMakeLists.txt</Message>
+      <Command Condition="'$(Configuration)|$(Platform)'=='Release|x64'">setlocal
+"C:\Program Files\CMake\bin\cmake.exe" -SE:/Fashion_MNIST_Optic_Evolution -BE:/Fashion_MNIST_Optic_Evolution/build --check-stamp-file E:/Fashion_MNIST_Optic_Evolution/build/CMakeFiles/generate.stamp
+if %errorlevel% neq 0 goto :cmEnd
+:cmEnd
+endlocal &amp; call :cmErrorLevel %errorlevel% &amp; goto :cmDone
+:cmErrorLevel
+exit /b %1
+:cmDone
+if %errorlevel% neq 0 goto :VCEnd</Command>
+      <AdditionalInputs Condition="'$(Configuration)|$(Platform)'=='Release|x64'">C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCUDAInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCXXInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCommonLanguageInclude.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeGenericSystem.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeInitializeConfigs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeLanguageInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeRCInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInitialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CheckLanguage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\CMakeCommonCompilerMacros.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindCUDAToolkit.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageHandleStandardArgs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageMessage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCUDALinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCXXLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCommonLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-Initialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\WindowsPaths.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCUDACompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCXXCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeRCCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeSystem.cmake;%(AdditionalInputs)</AdditionalInputs>
+      <Outputs Condition="'$(Configuration)|$(Platform)'=='Release|x64'">E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\generate.stamp</Outputs>
+      <LinkObjects Condition="'$(Configuration)|$(Platform)'=='Release|x64'">false</LinkObjects>
+      <Message Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">Building Custom Rule E:/Fashion_MNIST_Optic_Evolution/CMakeLists.txt</Message>
+      <Command Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">setlocal
+"C:\Program Files\CMake\bin\cmake.exe" -SE:/Fashion_MNIST_Optic_Evolution -BE:/Fashion_MNIST_Optic_Evolution/build --check-stamp-file E:/Fashion_MNIST_Optic_Evolution/build/CMakeFiles/generate.stamp
+if %errorlevel% neq 0 goto :cmEnd
+:cmEnd
+endlocal &amp; call :cmErrorLevel %errorlevel% &amp; goto :cmDone
+:cmErrorLevel
+exit /b %1
+:cmDone
+if %errorlevel% neq 0 goto :VCEnd</Command>
+      <AdditionalInputs Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCUDAInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCXXInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCommonLanguageInclude.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeGenericSystem.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeInitializeConfigs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeLanguageInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeRCInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInitialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CheckLanguage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\CMakeCommonCompilerMacros.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindCUDAToolkit.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageHandleStandardArgs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageMessage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCUDALinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCXXLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCommonLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-Initialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\WindowsPaths.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCUDACompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCXXCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeRCCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeSystem.cmake;%(AdditionalInputs)</AdditionalInputs>
+      <Outputs Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\generate.stamp</Outputs>
+      <LinkObjects Condition="'$(Configuration)|$(Platform)'=='MinSizeRel|x64'">false</LinkObjects>
+      <Message Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">Building Custom Rule E:/Fashion_MNIST_Optic_Evolution/CMakeLists.txt</Message>
+      <Command Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">setlocal
+"C:\Program Files\CMake\bin\cmake.exe" -SE:/Fashion_MNIST_Optic_Evolution -BE:/Fashion_MNIST_Optic_Evolution/build --check-stamp-file E:/Fashion_MNIST_Optic_Evolution/build/CMakeFiles/generate.stamp
+if %errorlevel% neq 0 goto :cmEnd
+:cmEnd
+endlocal &amp; call :cmErrorLevel %errorlevel% &amp; goto :cmDone
+:cmErrorLevel
+exit /b %1
+:cmDone
+if %errorlevel% neq 0 goto :VCEnd</Command>
+      <AdditionalInputs Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCUDAInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCXXInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeCommonLanguageInclude.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeGenericSystem.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeInitializeConfigs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeLanguageInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeRCInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CMakeSystemSpecificInitialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\CheckLanguage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\CMakeCommonCompilerMacros.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Compiler\NVIDIA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindCUDAToolkit.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageHandleStandardArgs.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\FindPackageMessage.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCUDALinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCXXLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Internal\CMakeCommonLinkerInformation.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Linker\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-Initialize.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC-CXX.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-MSVC.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows-NVIDIA-CUDA.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\Windows.cmake;C:\Program Files\CMake\share\cmake-3.31\Modules\Platform\WindowsPaths.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCUDACompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeCXXCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeRCCompiler.cmake;E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\3.31.4\CMakeSystem.cmake;%(AdditionalInputs)</AdditionalInputs>
+      <Outputs Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">E:\Fashion_MNIST_Optic_Evolution\build\CMakeFiles\generate.stamp</Outputs>
+      <LinkObjects Condition="'$(Configuration)|$(Platform)'=='RelWithDebInfo|x64'">false</LinkObjects>
+    </CustomBuild>
+  </ItemGroup>
+  <ItemGroup>
+    <ClCompile Include="E:\Fashion_MNIST_Optic_Evolution\src\data_loader.cpp" />
+    <ClCompile Include="E:\Fashion_MNIST_Optic_Evolution\src\main.cpp" />
+    <ClCompile Include="E:\Fashion_MNIST_Optic_Evolution\src\training.cpp" />
+    <ClCompile Include="E:\Fashion_MNIST_Optic_Evolution\src\utils.cpp" />
+    <CudaCompile Include="..\src\fungi.cu" />
+    <CudaCompile Include="..\src\optical_model.cu" />
+    <ClInclude Include="E:\Fashion_MNIST_Optic_Evolution\src\data_loader.hpp" />
+    <ClInclude Include="E:\Fashion_MNIST_Optic_Evolution\src\fungi.hpp" />
+    <ClInclude Include="E:\Fashion_MNIST_Optic_Evolution\src\fungi_Paremetres.hpp" />
+    <ClInclude Include="E:\Fashion_MNIST_Optic_Evolution\src\optical_model.hpp" />
+    <ClInclude Include="E:\Fashion_MNIST_Optic_Evolution\src\training.hpp" />
+    <ClInclude Include="E:\Fashion_MNIST_Optic_Evolution\src\utils.hpp" />
+  </ItemGroup>
+  <ItemGroup />
+  <ItemGroup>
+    <ProjectReference Include="E:\Fashion_MNIST_Optic_Evolution\build\ZERO_CHECK.vcxproj">
+      <Project>{B4CD8D8C-908F-3E91-878E-AFB61C98E407}</Project>
+      <Name>ZERO_CHECK</Name>
+      <ReferenceOutputAssembly>false</ReferenceOutputAssembly>
+      <CopyToOutputDirectory>Never</CopyToOutputDirectory>
+    </ProjectReference>
+  </ItemGroup>
+  <Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
+  <ImportGroup Label="ExtensionTargets">
+    <Import Project="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.0\extras\visual_studio_integration\MSBuildExtensions\CUDA 13.0.targets" />
+  </ImportGroup>
+</Project>

fashion_mnist_trainer.vcxproj.filters ADDED Viewed

	@@ -0,0 +1,56 @@

+<?xml version="1.0" encoding="utf-8"?>
+<Project ToolsVersion="17.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
+  <ItemGroup>
+    <ClCompile Include="E:\Fashion_MNIST_Optic_Evolution\src\data_loader.cpp">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="E:\Fashion_MNIST_Optic_Evolution\src\main.cpp">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="E:\Fashion_MNIST_Optic_Evolution\src\training.cpp">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+    <ClCompile Include="E:\Fashion_MNIST_Optic_Evolution\src\utils.cpp">
+      <Filter>Source Files</Filter>
+    </ClCompile>
+  </ItemGroup>
+  <ItemGroup>
+    <ClInclude Include="E:\Fashion_MNIST_Optic_Evolution\src\data_loader.hpp">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="E:\Fashion_MNIST_Optic_Evolution\src\fungi.hpp">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="E:\Fashion_MNIST_Optic_Evolution\src\fungi_Paremetres.hpp">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="E:\Fashion_MNIST_Optic_Evolution\src\optical_model.hpp">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="E:\Fashion_MNIST_Optic_Evolution\src\training.hpp">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+    <ClInclude Include="E:\Fashion_MNIST_Optic_Evolution\src\utils.hpp">
+      <Filter>Header Files</Filter>
+    </ClInclude>
+  </ItemGroup>
+  <ItemGroup>
+    <CudaCompile Include="..\src\fungi.cu">
+      <Filter>Source Files</Filter>
+    </CudaCompile>
+    <CudaCompile Include="..\src\optical_model.cu">
+      <Filter>Source Files</Filter>
+    </CudaCompile>
+  </ItemGroup>
+  <ItemGroup>
+    <CustomBuild Include="E:\Fashion_MNIST_Optic_Evolution\CMakeLists.txt" />
+  </ItemGroup>
+  <ItemGroup>
+    <Filter Include="Header Files">
+      <UniqueIdentifier>{44532C73-7FE8-32FC-B4E5-957A292A3CFC}</UniqueIdentifier>
+    </Filter>
+    <Filter Include="Source Files">
+      <UniqueIdentifier>{D89C3F0A-3ED6-31D3-BF76-113F82B1F136}</UniqueIdentifier>
+    </Filter>
+  </ItemGroup>
+</Project>

fungi.cu ADDED Viewed

	@@ -0,0 +1,615 @@

+// ======================= Mycelial Evolution System ==========================
+// Implements: reward->energy, gravity attraction, motion, growth/shrink,
+// death, sexual reproduction with recombination (3 children), mutation.
+// --- includes you likely already have ---
+#include "fungi.hpp"
+#include "fungi_Paremetres.hpp"
+#include <cuda_runtime.h>
+#include <curand_kernel.h>
+#include <algorithm>
+#include <random>
+#include <numeric>
+#include <cmath>
+#include <stdexcept>
+static inline void CK(cudaError_t st, const char* msg){
+  if (st != cudaSuccess) throw std::runtime_error(std::string("[CUDA] ")+msg+": "+cudaGetErrorString(st));
+}
+// ------------------------- 1) Reward from gradient --------------------------
+__device__ inline float d_phi_gauss(int x, int y, float cx, float cy,
+                                    float s, float a, float th){
+  // same φ as in build-masks (elliptical Gaussian)
+  float dx = x - cx, dy = y - cy;
+  float c = cosf(th), s0 = sinf(th);
+  float rx =  a/(s*s), ry = (1.f/a)/(s*s);
+  float xr =  c*dx + s0*dy;
+  float yr = -s0*dx + c*dy;
+  float q  = 0.5f*(xr*xr*rx + yr*yr*ry);
+  return __expf(-q);
+}
+/** R[h] = + Σ_{x,y} |grad(x,y)| * φ_h(x,y)  (FIX: reward HIGH gradients for feature detection) */
+__global__ void k_reward_map(const float* grad, int H,int W,
+                             const float* fx,const float* fy,const float* fs,
+                             const float* fa,const float* fth,
+                             float* R, int F){
+  int h = blockIdx.x*blockDim.x + threadIdx.x; if (h>=F) return;
+  float cx=fx[h], cy=fy[h], s=fs[h], a=fa[h], th=fth[h];
+  float acc = 0.f;
+  for(int y=0;y<H;y++){
+    for(int x=0;x<W;x++){
+      float ph = d_phi_gauss(x,y,cx,cy,s,a,th);
+      acc += ph * fabsf(grad[y*W + x]);
+    }
+  }
+  R[h] = +acc; // FIX: Reward HIGH gradients (removed negative sign)
+}
+// ------------------------- 2) Gravity & Motion ------------------------------
+/** Compute softened Newtonian attraction between fungi (O(F^2), fine up to ~1k). */
+__global__ void k_gravity_forces(const float* fx,const float* fy,const float* mass,
+                                 float* ax,float* ay, int F, float G, float eps2){
+  int i = blockIdx.x*blockDim.x + threadIdx.x; if (i>=F) return;
+  float xi=fx[i], yi=fy[i], mi = fmaxf(mass[i], 1e-3f);
+  float ax_i=0.f, ay_i=0.f;
+  for(int j=0;j<F;j++){
+    if (j==i) continue;
+    float dx=fx[j]-xi, dy=fy[j]-yi;
+    float r2 = dx*dx + dy*dy + eps2;
+    float invr = rsqrtf(r2);
+    float invr3 = invr*invr*invr;
+    float Fg = G * mi * mass[j] * invr3;
+    ax_i += Fg * dx / mi;
+    ay_i += Fg * dy / mi;
+  }
+  ax[i]=ax_i; ay[i]=ay_i;
+}
+/** Simple velocity Verlet / Euler update with damping and box bounds. */
+__global__ void k_apply_motion(float* fx,float* fy,
+                               float* vx,float* vy,
+                               const float* ax,const float* ay,
+                               int F, int W,int H, float dt, float damp){
+  int i=blockIdx.x*blockDim.x+threadIdx.x; if(i>=F) return;
+  float vxi = (vx?vx[i]:0.f), vyi=(vy?vy[i]:0.f);
+  vxi = (vxi + dt*ax[i]) * damp;
+  vyi = (vyi + dt*ay[i]) * damp;
+  float xi = fx[i] + dt*vxi;
+  float yi = fy[i] + dt*vyi;
+  // reflect at borders
+  if (xi<0){ xi=0; vxi=-vxi; } if (xi>W-1){ xi=W-1; vxi=-vxi; }
+  if (yi<0){ yi=0; vyi=-vyi; } if (yi>H-1){ yi=H-1; vyi=-vyi; }
+  fx[i]=xi; fy[i]=yi; if(vx) vx[i]=vxi; if(vy) vy[i]=vyi;
+}
+// -------------------- 3) Energy, growth/shrink, death -----------------------
+/**
+ * Update energy with decay/food/cost and adjust radius sigma via tanh(energy).
+ * Mark for death (flag=1) if energy < death_th.
+ */
+__global__ void k_energy_growth_mark(float* fx,float* fy,float* fs,float* fa,float* fth,
+                                     float* a0,float* p0,
+                                     float* energy,float* mass,int* age,
+                                     const float* R, int* deathFlag, int F,
+                                     float food,float decay,float death_th,float cost){
+  int h=blockIdx.x*blockDim.x+threadIdx.x; if(h>=F) return;
+  float e = energy[h]*decay + food*R[h] - cost*(1.f + 0.01f*fs[h]*fs[h]);
+  energy[h]=e; age[h] += 1;
+  // growth/shrink
+  float g=tanhf(e);
+  fs[h] = fmaxf(1.0f, fs[h]*(1.f + 0.05f*g));
+  // death mark
+  deathFlag[h] = (e<death_th) ? 1 : 0;
+}
+// -------------------- 4) Host-side reproduction (pairing etc.) --------------
+/** Helper: uniform crossover + Gaussian mutation of scalar gene. */
+static inline float _xover_mut(float a, float b, std::mt19937& rng, float mut_std){
+  std::uniform_real_distribution<float> U(0.f,1.f);
+  std::normal_distribution<float> N(0.f, mut_std);
+  float t = U(rng) < 0.5f ? a : b; // uniform crossover
+  return t + N(rng);               // mutation
+}
+static inline float _clip(float v, float lo, float hi){ return std::max(lo, std::min(hi, v)); }
+/** Create one child from parents i,j; append to population (HOST memory). */
+static void _spawn_child(const FungiSoA& P, int i, int j,
+                         FungiSoA& out, std::mt19937& rng,
+                         float mut_pos=0.5f, float mut_par=0.2f){
+  FungiSoA& R = out;
+  R.F += 1;
+  // Ensure capacity in vectors
+  auto reserveF = [&](std::vector<float>& v){ if((int)v.size()<R.F) v.resize(R.F); };
+  auto reserveI = [&](std::vector<int>& v){ if((int)v.size()<R.F) v.resize(R.F); };
+  reserveF(R.x); reserveF(R.y); reserveF(R.sigma); reserveF(R.alpha); reserveF(R.theta);
+  reserveF(R.a_base); reserveF(R.p_base); reserveF(R.energy); reserveF(R.mass); reserveI(R.age);
+  int k = R.F-1;
+  // genes
+  R.x[k]      = _clip(_xover_mut(P.x[i],      P.x[j],      rng, mut_pos), 0.f, (float)(R.W-1));
+  R.y[k]      = _clip(_xover_mut(P.y[i],      P.y[j],      rng, mut_pos), 0.f, (float)(R.H-1));
+  R.sigma[k]  = _clip(_xover_mut(P.sigma[i],  P.sigma[j],  rng, mut_par), 1.0f,  10.0f);
+  R.alpha[k]  = _clip(_xover_mut(P.alpha[i],  P.alpha[j],  rng, 0.1f),    0.3f,   3.0f);
+  R.theta[k]  = _xover_mut(P.theta[i],  P.theta[j],  rng, 0.2f);
+  R.a_base[k] = _xover_mut(P.a_base[i],P.a_base[j], rng, 0.1f);
+  R.p_base[k] = _xover_mut(P.p_base[i],P.p_base[j], rng, 0.1f);
+  // newborn state
+  R.energy[k] = 0.0f;
+  R.mass[k]   = 0.5f*(P.mass[i]+P.mass[j]) * 0.3f; // fraction of parents' mass
+  R.age[k]    = 0;
+}
+/**
+ * Pairing: pick adults by energy, find close pairs (distance < pair_dist),
+ * spawn `offspring_per_pair` children per pair (3 recommended).
+ * Parents donate part of mass/energy (they "shrink" a bit).
+ */
+static void _pair_and_reproduce(FungiSoA& P, const EvoParams& evo){
+  std::vector<int> ids(P.F); std::iota(ids.begin(), ids.end(), 0);
+  std::sort(ids.begin(), ids.end(), [&](int a,int b){ return P.energy[a] > P.energy[b]; });
+  std::mt19937 rng(evo.seed ^ 0xBADC0DEu);
+  FungiSoA out = P; // start with copy; children appended here
+  auto dist = [&](int i,int j){
+    float dx=P.x[i]-P.x[j], dy=P.y[i]-P.y[j];
+    return std::sqrt(dx*dx+dy*dy);
+  };
+  int maxPairs = std::max(1, P.F/8);
+  int made=0;
+  for (size_t idx=0; idx+1<ids.size() && made<maxPairs; ++idx){
+    int i = ids[idx], j = ids[idx+1];
+    if (i==j) continue;
+    // Consider as adults if energy positive
+    if (P.energy[i] <= 0.f || P.energy[j] <= 0.f) continue;
+    if (dist(i,j) > evo.pair_dist) continue;
+    // Spawn children
+    for (int c=0; c<evo.offspring_per_pair; ++c){
+      _spawn_child(P, i, j, out, rng);
+    }
+    // Parents donate some mass/energy → shrink a bit
+    P.mass[i] *= 0.85f;  P.mass[j] *= 0.85f;
+    P.sigma[i]*= 0.95f;  P.sigma[j]*= 0.95f;
+    P.energy[i]*= 0.7f;  P.energy[j]*= 0.7f;
+    made++;
+  }
+  // Compact deaths and cap population
+  // First remove completely "dead" individuals: energy << 0 or mass very small
+  std::vector<int> keep(out.F, 1);
+  for(int h=0; h<out.F; ++h){
+    if (out.energy[h] < -2.0f || out.mass[h] < 0.05f) keep[h]=0;
+  }
+  // Create list of survivors
+  std::vector<int> order; order.reserve(out.F);
+  for(int h=0; h<out.F; ++h) if(keep[h]) order.push_back(h);
+  // Cap by energy (keep best) if necessary
+  int cap = evo.max_population>0 ? evo.max_population : (int)std::floor(1.5 * (double)P.F);
+  if ((int)order.size() > cap){
+    std::sort(order.begin(), order.end(), [&](int a,int b){ return out.energy[a] > out.energy[b]; });
+    order.resize(cap);
+  }
+  // Rewrite P with survivors in `order`
+  auto pick = [&](std::vector<float>& v){
+    std::vector<float> nv; nv.reserve(order.size());
+    for(int id: order) nv.push_back(v[id]); v.swap(nv);
+  };
+  auto picki = [&](std::vector<int>& v){
+    std::vector<int> nv; nv.reserve(order.size());
+    for(int id: order) nv.push_back(v[id]); v.swap(nv);
+  };
+  pick(out.x); pick(out.y); pick(out.sigma); pick(out.alpha); pick(out.theta);
+  pick(out.a_base); pick(out.p_base); pick(out.energy); pick(out.mass); picki(out.age);
+  out.F = (int)order.size();
+  P = std::move(out);
+}
+// ---------------------- 5) Full ecology step (public) -----------------------
+void fungi_ecology_step(FungiSoA& pop, const float* d_grad_map, const EvoParams& evo){
+  const int F=pop.F, H=pop.H, W=pop.W;
+  if (F<=0) return;
+  // Upload SoA to device
+  float *fx,*fy,*fs,*fa,*fth,*a0,*p0,*E,*M; int *Age;
+  CK(cudaMalloc(&fx,sizeof(float)*F),"alloc fx");
+  CK(cudaMalloc(&fy,sizeof(float)*F),"alloc fy");
+  CK(cudaMalloc(&fs,sizeof(float)*F),"alloc fs");
+  CK(cudaMalloc(&fa,sizeof(float)*F),"alloc fa");
+  CK(cudaMalloc(&fth,sizeof(float)*F),"alloc th");
+  CK(cudaMalloc(&a0,sizeof(float)*F),"alloc a0");
+  CK(cudaMalloc(&p0,sizeof(float)*F),"alloc p0");
+  CK(cudaMalloc(&E, sizeof(float)*F),"alloc E");
+  CK(cudaMalloc(&M, sizeof(float)*F),"alloc M");
+  CK(cudaMalloc(&Age,sizeof(int)*F),"alloc Age");
+  CK(cudaMemcpy(fx, pop.x.data(),     sizeof(float)*F, cudaMemcpyHostToDevice),"H2D fx");
+  CK(cudaMemcpy(fy, pop.y.data(),     sizeof(float)*F, cudaMemcpyHostToDevice),"H2D fy");
+  CK(cudaMemcpy(fs, pop.sigma.data(), sizeof(float)*F, cudaMemcpyHostToDevice),"H2D fs");
+  CK(cudaMemcpy(fa, pop.alpha.data(), sizeof(float)*F, cudaMemcpyHostToDevice),"H2D fa");
+  CK(cudaMemcpy(fth,pop.theta.data(), sizeof(float)*F, cudaMemcpyHostToDevice),"H2D th");
+  CK(cudaMemcpy(a0, pop.a_base.data(),sizeof(float)*F, cudaMemcpyHostToDevice),"H2D a0");
+  CK(cudaMemcpy(p0, pop.p_base.data(),sizeof(float)*F, cudaMemcpyHostToDevice),"H2D p0");
+  CK(cudaMemcpy(E,  pop.energy.data(),sizeof(float)*F, cudaMemcpyHostToDevice),"H2D E");
+  CK(cudaMemcpy(M,  pop.mass.data(),  sizeof(float)*F, cudaMemcpyHostToDevice),"H2D M");
+  CK(cudaMemcpy(Age,pop.age.data(),   sizeof(int)*F,   cudaMemcpyHostToDevice),"H2D Age");
+  // 1) Reward
+  float* dR=nullptr; CK(cudaMalloc(&dR,sizeof(float)*F),"alloc R");
+  k_reward_map<<<(F+255)/256,256>>>(d_grad_map, H,W, fx,fy,fs,fa,fth, dR, F);
+  // 2) Gravity & motion (alloc vel/acc)
+  float *ax,*ay,*vx,*vy; CK(cudaMalloc(&ax,sizeof(float)*F),"alloc ax");
+  CK(cudaMalloc(&ay,sizeof(float)*F),"alloc ay");
+  CK(cudaMalloc(&vx,sizeof(float)*F),"alloc vx");
+  CK(cudaMalloc(&vy,sizeof(float)*F),"alloc vy");
+  CK(cudaMemset(vx,0,sizeof(float)*F),"zero vx");
+  CK(cudaMemset(vy,0,sizeof(float)*F),"zero vy");
+  k_gravity_forces<<<(F+255)/256,256>>>(fx,fy,M, ax,ay, F, evo.G, evo.eps2);
+  k_apply_motion<<<(F+255)/256,256>>>(fx,fy, vx,vy, ax,ay, F, W,H, evo.dt, evo.damp);
+  // 3) Energy/growth + death flags
+  int* dDead=nullptr; CK(cudaMalloc(&dDead,sizeof(int)*F),"alloc deadFlag");
+  CK(cudaMemset(dDead,0,sizeof(int)*F),"zero deadFlag");
+  k_energy_growth_mark<<<(F+255)/256,256>>>(fx,fy,fs,fa,fth, a0,p0, E,M,Age,
+                                            dR, dDead, F,
+                                            evo.food,evo.decay,evo.death_th,evo.cost);
+  // Download updated arrays
+  std::vector<int> hDead(F);
+  CK(cudaMemcpy(pop.x.data(), fx, sizeof(float)*F, cudaMemcpyDeviceToHost),"D2H fx");
+  CK(cudaMemcpy(pop.y.data(), fy, sizeof(float)*F, cudaMemcpyDeviceToHost),"D2H fy");
+  CK(cudaMemcpy(pop.sigma.data(), fs, sizeof(float)*F, cudaMemcpyDeviceToHost),"D2H fs");
+  CK(cudaMemcpy(pop.alpha.data(), fa, sizeof(float)*F, cudaMemcpyDeviceToHost),"D2H fa");
+  CK(cudaMemcpy(pop.theta.data(), fth,sizeof(float)*F, cudaMemcpyDeviceToHost),"D2H th");
+  CK(cudaMemcpy(pop.a_base.data(),a0, sizeof(float)*F, cudaMemcpyDeviceToHost),"D2H a0");
+  CK(cudaMemcpy(pop.p_base.data(),p0, sizeof(float)*F, cudaMemcpyDeviceToHost),"D2H p0");
+  CK(cudaMemcpy(pop.energy.data(),E,  sizeof(float)*F, cudaMemcpyDeviceToHost),"D2H E");
+  CK(cudaMemcpy(pop.mass.data(),  M,  sizeof(float)*F, cudaMemcpyDeviceToHost),"D2H M");
+  CK(cudaMemcpy(pop.age.data(), Age,  sizeof(int)*F,   cudaMemcpyDeviceToHost),"D2H Age");
+  CK(cudaMemcpy(hDead.data(), dDead,  sizeof(int)*F,   cudaMemcpyDeviceToHost),"D2H deadFlag");
+  // Free device temporaries
+  cudaFree(dR); cudaFree(ax); cudaFree(ay); cudaFree(vx); cudaFree(vy); cudaFree(dDead);
+  cudaFree(fx); cudaFree(fy); cudaFree(fs); cudaFree(fa); cudaFree(fth);
+  cudaFree(a0); cudaFree(p0); cudaFree(E); cudaFree(M); cudaFree(Age);
+  // Remove marked dead on HOST quickly
+  std::vector<int> keep; keep.reserve(pop.F);
+  for (int i=0;i<pop.F;i++) if(!hDead[i]) keep.push_back(i);
+  auto pick = [&](std::vector<float>& v){
+    std::vector<float> nv; nv.reserve(keep.size());
+    for(int id: keep) nv.push_back(v[id]); v.swap(nv);
+  };
+  auto picki=[&](std::vector<int>& v){
+    std::vector<int> nv; nv.reserve(keep.size());
+    for(int id: keep) nv.push_back(v[id]); v.swap(nv);
+  };
+  pick(pop.x); pick(pop.y); pick(pop.sigma); pick(pop.alpha); pick(pop.theta);
+  pick(pop.a_base); pick(pop.p_base); pick(pop.energy); pick(pop.mass); picki(pop.age);
+  pop.F = (int)keep.size();
+  // 4) Pairing & reproduction (3 children per pair by default)
+  _pair_and_reproduce(pop, evo);
+}
+#include "fungi.hpp"
+#include <cuda_runtime.h>
+#include <curand_kernel.h>
+#include <algorithm>
+#include <numeric>
+#include <cmath>
+#include <stdexcept>
+static inline void ck(cudaError_t st, const char* msg){
+  if (st != cudaSuccess) { throw std::runtime_error(std::string("[CUDA] ")+msg+": "+cudaGetErrorString(st)); }
+}
+void FungiSoA::resize(int F_, int H_, int W_) {
+  F = F_; H=H_; W=W_;
+  x.resize(F); y.resize(F); sigma.resize(F); alpha.resize(F); theta.resize(F);
+  a_base.resize(F); p_base.resize(F);
+  energy.assign(F, 0.f); mass.assign(F, 1.f); age.assign(F, 0);
+}
+void FungiSoA::init_random(unsigned seed, float sigma_min, float sigma_max) {
+  std::mt19937 rng(seed);
+  std::uniform_real_distribution<float> Ux(0.f, (float)(W-1));
+  std::uniform_real_distribution<float> Uy(0.f, (float)(H-1));
+  std::uniform_real_distribution<float> Us(sigma_min, sigma_max);
+  std::uniform_real_distribution<float> Ua(0.7f, 1.3f);
+  std::uniform_real_distribution<float> Ut(-3.1415926f, 3.1415926f);
+  std::normal_distribution<float> N01(0.f, 0.15f);
+  for (int i=0;i<F;i++){
+    x[i]=Ux(rng); y[i]=Uy(rng); sigma[i]=Us(rng);
+    alpha[i]=Ua(rng); theta[i]=Ut(rng);
+    a_base[i]=N01(rng); p_base[i]=N01(rng);
+  }
+}
+void FungiSoA::adjust_population(int newF, unsigned seed) {
+  if (newF < 1) newF = 1;
+  if (newF == F) return;
+  if (newF < F) {
+    std::vector<int> idx(F);
+    std::iota(idx.begin(), idx.end(), 0);
+    std::partial_sort(idx.begin(), idx.begin() + newF, idx.end(), [&](int a, int b){
+      return energy[a] > energy[b];
+    });
+    idx.resize(newF);
+    auto reorderF = [&](std::vector<float>& v){
+      std::vector<float> nv; nv.reserve(newF);
+      for (int id : idx) nv.push_back(v[id]);
+      v.swap(nv);
+    };
+    auto reorderI = [&](std::vector<int>& v){
+      std::vector<int> nv; nv.reserve(newF);
+      for (int id : idx) nv.push_back(v[id]);
+      v.swap(nv);
+    };
+    reorderF(x); reorderF(y); reorderF(sigma); reorderF(alpha); reorderF(theta);
+    reorderF(a_base); reorderF(p_base); reorderF(energy); reorderF(mass);
+    reorderI(age);
+    F = newF;
+    return;
+  }
+  int oldF = F;
+  x.resize(newF); y.resize(newF); sigma.resize(newF); alpha.resize(newF); theta.resize(newF);
+  a_base.resize(newF); p_base.resize(newF); energy.resize(newF); mass.resize(newF); age.resize(newF);
+  std::mt19937 rng(seed ^ (unsigned)newF);
+  std::uniform_real_distribution<float> Ux(0.f, (float)(W>0?W-1:0));
+  std::uniform_real_distribution<float> Uy(0.f, (float)(H>0?H-1:0));
+  std::uniform_real_distribution<float> Us(1.5f, 6.5f);
+  std::uniform_real_distribution<float> Ua(0.5f, 1.6f);
+  std::uniform_real_distribution<float> Ut(-3.1415926f, 3.1415926f);
+  std::normal_distribution<float> Namp(0.f, 0.3f);
+  for (int i = oldF; i < newF; ++i) {
+    x[i] = (W > 0) ? Ux(rng) : 0.f;
+    y[i] = (H > 0) ? Uy(rng) : 0.f;
+    sigma[i] = Us(rng);
+    alpha[i] = Ua(rng);
+    theta[i] = Ut(rng);
+    a_base[i] = Namp(rng);
+    p_base[i] = Namp(rng);
+    energy[i] = 0.f;
+    mass[i] = 1.f;
+    age[i] = 0;
+  }
+  F = newF;
+}
+// ------------------ GPU kernels ------------------
+__global__ void k_clear(float* A, int N, float v=0.f){ int i=blockIdx.x*blockDim.x+threadIdx.x; if(i<N) A[i]=v; }
+/** Gaussian elliptical basis φ_h(x,y). */
+__device__ inline float phi_gauss(int x, int y, float cx, float cy, float s, float a, float th){
+  float dx = x - cx, dy = y - cy;
+  float c = cosf(th), s0 = sinf(th);
+  float rx =  a/(s*s), ry = (1.f/a)/(s*s);
+  // Rotate
+  float xr =  c*dx + s0*dy;
+  float yr = -s0*dx + c*dy;
+  float q = 0.5f*(xr*xr*rx + yr*yr*ry);
+  return __expf(-q);
+}
+/** Accumulate masks from fungi (naive per-pixel; for 28x28 or small HxW it's fast). */
+__global__ void k_build_masks(const float* fx,const float* fy,const float* fs,const float* fa,const float* fth,
+                              const float* fa0,const float* fp0, int F,
+                              float* A, float* P, int H, int W){
+  int idx = blockIdx.x * blockDim.x + threadIdx.x;
+  int N = H*W; if (idx>=N) return;
+  int y = idx / W, x = idx % W;
+  float sa=0.f, sp=0.f;
+  #pragma unroll 4
+  for(int h=0; h<F; ++h){
+    float ph = phi_gauss(x,y, fx[h],fy[h],fs[h],fa[h],fth[h]);
+    sa += fa0[h] * ph;
+    sp += fp0[h] * ph;
+  }
+  // Map to physical domain with CLAMPING to prevent oversaturation
+  // A = clamped_softplus(sa)+eps ; P = pi * tanh(sp)
+  float Araw = fminf(log1pf(expf(fminf(sa, 2.0f))), 2.0f) + 1e-3f; // FIX: Clamp amplitude max 2.0
+  float Praw = 3.1415926f * tanhf(fminf(fmaxf(sp, -1.0f), 1.0f));  // FIX: Clamp phase input
+  A[idx] = Araw;
+  P[idx] = Praw;
+}
+void fungi_build_masks_GPU(const FungiSoA& pop,
+                           float* d_A, float* d_P,
+                           int /*tiles_y*/, int /*tiles_x*/) {
+  // Upload minimal arrays (F is small ~128-512)
+  float *dx,*dy,*ds,*da,*dt,*dab,*dpb;
+  ck(cudaMalloc(&dx, sizeof(float)*pop.F), "alloc fx");
+  ck(cudaMalloc(&dy, sizeof(float)*pop.F), "alloc fy");
+  ck(cudaMalloc(&ds, sizeof(float)*pop.F), "alloc fs");
+  ck(cudaMalloc(&da, sizeof(float)*pop.F), "alloc fa");
+  ck(cudaMalloc(&dt, sizeof(float)*pop.F), "alloc fth");
+  ck(cudaMalloc(&dab,sizeof(float)*pop.F), "alloc a0");
+  ck(cudaMalloc(&dpb,sizeof(float)*pop.F), "alloc p0");
+  ck(cudaMemcpy(dx, pop.x.data(), sizeof(float)*pop.F, cudaMemcpyHostToDevice), "H2D fx");
+  ck(cudaMemcpy(dy, pop.y.data(), sizeof(float)*pop.F, cudaMemcpyHostToDevice), "H2D fy");
+  ck(cudaMemcpy(ds, pop.sigma.data(), sizeof(float)*pop.F, cudaMemcpyHostToDevice), "H2D fs");
+  ck(cudaMemcpy(da, pop.alpha.data(), sizeof(float)*pop.F, cudaMemcpyHostToDevice), "H2D fa");
+  ck(cudaMemcpy(dt, pop.theta.data(), sizeof(float)*pop.F, cudaMemcpyHostToDevice), "H2D fth");
+  ck(cudaMemcpy(dab,pop.a_base.data(), sizeof(float)*pop.F, cudaMemcpyHostToDevice), "H2D a0");
+  ck(cudaMemcpy(dpb,pop.p_base.data(), sizeof(float)*pop.F, cudaMemcpyHostToDevice), "H2D p0");
+  int N = pop.H*pop.W;
+  k_clear<<<(N+255)/256,256>>>(d_A, N, 0.f);
+  k_clear<<<(N+255)/256,256>>>(d_P, N, 0.f);
+  k_build_masks<<<(N+255)/256,256>>>(dx,dy,ds,da,dt, dab,dpb,pop.F, d_A,d_P, pop.H,pop.W);
+  cudaFree(dx); cudaFree(dy); cudaFree(ds); cudaFree(da); cudaFree(dt); cudaFree(dab); cudaFree(dpb);
+}
+// --- simple RNG on device for evolution ---
+__global__ void k_rng_init(curandStatePhilox4_32_10_t* st, unsigned seed, int n){
+  int i=blockIdx.x*blockDim.x+threadIdx.x; if(i<n) curand_init(seed, i, 0, &st[i]);
+}
+// Reward accumulation: R_h = -sum φ_h * grad(x,y)
+__global__ void k_reward(const float* grad, int H, int W,
+                         const float* fx,const float* fy,const float* fs,const float* fa,const float* fth,
+                         float* R, int F){
+  int h = blockIdx.x*blockDim.x+threadIdx.x; if(h>=F) return;
+  float cx=fx[h], cy=fy[h], s=fs[h], a=fa[h], th=fth[h];
+  float acc=0.f;
+  for(int y=0;y<H;y++){
+    for(int x=0;x<W;x++){
+      float ph = phi_gauss(x,y,cx,cy,s,a,th);
+      acc += ph * fabsf(grad[y*W+x]); // magnitude proxy
+    }
+  }
+  R[h] = +acc; // FIX: Reward HIGH gradients (removed negative sign) // want to reduce gradient magnitude
+}
+__global__ void k_evolve(float* fx,float* fy,float* fs,float* fa,float* fth,
+                         float* a0,float* p0,float* energy,float* mass,int* age,
+                         const float* R, int F, int H, int W,
+                         float food,float decay,float death_th,float cost,
+                         curandStatePhilox4_32_10_t* rng){
+  int h = blockIdx.x*blockDim.x+threadIdx.x; if(h>=F) return;
+  curandStatePhilox4_32_10_t st = rng[h];
+  float e = energy[h]*decay + food*R[h] - cost*(1.f + 0.01f*fs[h]*fs[h]); // penalize large radius
+  energy[h] = e;
+  age[h] += 1;
+  // Growth/Shrink
+  float g = tanhf(e); // [-1,1]
+  fs[h] = fmaxf(1.0f, fs[h] * (1.f + 0.05f*g));
+  // Small random drift (Brownian)
+  float dx=(curand_uniform(&st)-0.5f)*1.0f, dy=(curand_uniform(&st)-0.5f)*1.0f;
+  fx[h] = fminf(fmaxf(fx[h] + dx, 0.f), (float)(W-1));
+  fy[h] = fminf(fmaxf(fy[h] + dy, 0.f), (float)(H-1));
+  // Rejuvenate bad fungi
+  if(e < death_th){
+    fx[h]=curand_uniform(&st)*(W-1.f);
+    fy[h]=curand_uniform(&st)*(H-1.f);
+    fs[h]=1.5f + 4.0f*curand_uniform(&st);
+    fa[h]=0.8f + 0.6f*curand_uniform(&st);
+    fth[h]=(curand_uniform(&st)*2.f-1.f)*3.1415926f;
+    a0[h]+= (curand_uniform(&st)-0.5f)*0.2f;
+    p0[h]+= (curand_uniform(&st)-0.5f)*0.2f;
+    energy[h]=0.f; mass[h]=1.f; age[h]=0;
+  }
+  rng[h]=st;
+}
+void fungi_evolve_GPU(FungiSoA& pop,
+                      const float* d_grad_map,
+                      int evo_pairs,
+                      float food, float decay, float death_th,
+                      float cost, unsigned seed){
+  (void)evo_pairs; // (pairing/attraction left minimal; reward-based dynamics are cheap & robust)
+  // Upload SoA
+  float *fx,*fy,*fs,*fa,*fth,*a0,*p0,*E,*M; int *Age;
+  ck(cudaMalloc(&fx,sizeof(float)*pop.F),"alloc fx"); ck(cudaMalloc(&fy,sizeof(float)*pop.F),"alloc fy");
+  ck(cudaMalloc(&fs,sizeof(float)*pop.F),"alloc fs"); ck(cudaMalloc(&fa,sizeof(float)*pop.F),"alloc fa");
+  ck(cudaMalloc(&fth,sizeof(float)*pop.F),"alloc th"); ck(cudaMalloc(&a0,sizeof(float)*pop.F),"alloc a0");
+  ck(cudaMalloc(&p0,sizeof(float)*pop.F),"alloc p0"); ck(cudaMalloc(&E,sizeof(float)*pop.F),"alloc E");
+  ck(cudaMalloc(&M,sizeof(float)*pop.F),"alloc M"); ck(cudaMalloc(&Age,sizeof(int)*pop.F),"alloc age");
+  ck(cudaMemcpy(fx, pop.x.data(), sizeof(float)*pop.F, cudaMemcpyHostToDevice),"H2D fx");
+  ck(cudaMemcpy(fy, pop.y.data(), sizeof(float)*pop.F, cudaMemcpyHostToDevice),"H2D fy");
+  ck(cudaMemcpy(fs, pop.sigma.data(), sizeof(float)*pop.F, cudaMemcpyHostToDevice),"H2D fs");
+  ck(cudaMemcpy(fa, pop.alpha.data(), sizeof(float)*pop.F, cudaMemcpyHostToDevice),"H2D fa");
+  ck(cudaMemcpy(fth,pop.theta.data(), sizeof(float)*pop.F, cudaMemcpyHostToDevice),"H2D th");
+  ck(cudaMemcpy(a0, pop.a_base.data(), sizeof(float)*pop.F, cudaMemcpyHostToDevice),"H2D a0");
+  ck(cudaMemcpy(p0, pop.p_base.data(), sizeof(float)*pop.F, cudaMemcpyHostToDevice),"H2D p0");
+  ck(cudaMemcpy(E,  pop.energy.data(),sizeof(float)*pop.F, cudaMemcpyHostToDevice),"H2D E");
+  ck(cudaMemcpy(M,  pop.mass.data(),  sizeof(float)*pop.F, cudaMemcpyHostToDevice),"H2D M");
+  ck(cudaMemcpy(Age,pop.age.data(),   sizeof(int)*pop.F,   cudaMemcpyHostToDevice),"H2D age");
+  // Reward
+  float* dR=nullptr; ck(cudaMalloc(&dR,sizeof(float)*pop.F),"alloc R");
+  k_reward<<<(pop.F+255)/256,256>>>(d_grad_map, pop.H,pop.W, fx,fy,fs,fa,fth, dR, pop.F);
+  // RNG & evolve
+  curandStatePhilox4_32_10_t* rng; ck(cudaMalloc(&rng,sizeof(curandStatePhilox4_32_10_t)*pop.F),"alloc rng");
+  k_rng_init<<<(pop.F+255)/256,256>>>(rng, seed, pop.F);
+  k_evolve<<<(pop.F+255)/256,256>>>(fx,fy,fs,fa,fth, a0,p0,E,M,Age, dR, pop.F, pop.H,pop.W, food,decay,death_th,cost, rng);
+  // Download
+  ck(cudaMemcpy(pop.x.data(), fx, sizeof(float)*pop.F, cudaMemcpyDeviceToHost),"D2H fx");
+  ck(cudaMemcpy(pop.y.data(), fy, sizeof(float)*pop.F, cudaMemcpyDeviceToHost),"D2H fy");
+  ck(cudaMemcpy(pop.sigma.data(), fs, sizeof(float)*pop.F, cudaMemcpyDeviceToHost),"D2H fs");
+  ck(cudaMemcpy(pop.alpha.data(), fa, sizeof(float)*pop.F, cudaMemcpyDeviceToHost),"D2H fa");
+  ck(cudaMemcpy(pop.theta.data(), fth,sizeof(float)*pop.F, cudaMemcpyDeviceToHost),"D2H th");
+  ck(cudaMemcpy(pop.a_base.data(), a0,sizeof(float)*pop.F, cudaMemcpyDeviceToHost),"D2H a0");
+  ck(cudaMemcpy(pop.p_base.data(), p0,sizeof(float)*pop.F, cudaMemcpyDeviceToHost),"D2H p0");
+  ck(cudaMemcpy(pop.energy.data(), E, sizeof(float)*pop.F, cudaMemcpyDeviceToHost),"D2H E");
+  ck(cudaMemcpy(pop.mass.data(),   M, sizeof(float)*pop.F, cudaMemcpyDeviceToHost),"D2H M");
+  ck(cudaMemcpy(pop.age.data(),  Age,sizeof(int)*pop.F,   cudaMemcpyDeviceToHost),"D2H age");
+  cudaFree(dR); cudaFree(rng);
+  cudaFree(fx); cudaFree(fy); cudaFree(fs); cudaFree(fa); cudaFree(fth);
+  cudaFree(a0); cudaFree(p0); cudaFree(E); cudaFree(M); cudaFree(Age);
+}
+void download_mask(float* h, const float* d, int HW){
+  ck(cudaMemcpy(h,d,sizeof(float)*HW,cudaMemcpyDeviceToHost),"D2H mask");
+}
+// ================== DIAGNOSTIC TOOLS ==================
+void fungi_create_test_pattern(float* h_pattern, int H, int W, int pattern_type) {
+    for (int y = 0; y < H; y++) {
+        for (int x = 0; x < W; x++) {
+            float val = 0.0f;
+            switch (pattern_type) {
+                case 0: // Checkerboard 4x4
+                    val = ((x/4 + y/4) % 2) ? 1.0f : 0.0f;
+                    break;
+                case 1: // Gradient horizontal
+                    val = (float)x / (W-1);
+                    break;
+                case 2: // Concentric circles
+                    {
+                        float cx = W/2.0f, cy = H/2.0f;
+                        float r = sqrtf((x-cx)*(x-cx) + (y-cy)*(y-cy));
+                        val = 0.5f + 0.5f * sinf(r * 0.5f);
+                    }
+                    break;
+                case 3: // Delta impulse (center)
+                    val = (x == W/2 && y == H/2) ? 1.0f : 0.0f;
+                    break;
+            }
+            h_pattern[y*W + x] = val;
+        }
+    }
+}
+void fungi_analyze_mask_statistics(const float* d_A, const float* d_P, int HW) {
+    std::vector<float> h_A(HW), h_P(HW);
+    ck(cudaMemcpy(h_A.data(), d_A, sizeof(float)*HW, cudaMemcpyDeviceToHost), "D2H A");
+    ck(cudaMemcpy(h_P.data(), d_P, sizeof(float)*HW, cudaMemcpyDeviceToHost), "D2H P");
+    // Estadísticas de Amplitud
+    float A_min = *std::min_element(h_A.begin(), h_A.end());
+    float A_max = *std::max_element(h_A.begin(), h_A.end());
+    float A_mean = std::accumulate(h_A.begin(), h_A.end(), 0.0f) / HW;
+    // Estadísticas de Fase
+    float P_min = *std::min_element(h_P.begin(), h_P.end());
+    float P_max = *std::max_element(h_P.begin(), h_P.end());
+    float P_mean = std::accumulate(h_P.begin(), h_P.end(), 0.0f) / HW;
+    printf("🔍 FUNGI MASK STATISTICS:\n");
+    printf("   Amplitude: min=%.3f, max=%.3f, mean=%.3f\n", A_min, A_max, A_mean);
+    printf("   Phase:     min=%.3f, max=%.3f, mean=%.3f\n", P_min, P_max, P_mean);
+    printf("   A oversaturation: %s\n", (A_max > 1.9f) ? "⚠️  YES" : "✅ NO");
+    printf("   P range check: %s\n", (P_max > 3.2f || P_min < -3.2f) ? "⚠️  OUT OF RANGE" : "✅ OK");
+}

fungi.cu.cache ADDED Viewed

	@@ -0,0 +1,57 @@

+Identity=..\src\fungi.cu
+AdditionalCompilerOptions=
+AdditionalCompilerOptions=
+AdditionalDependencies=
+AdditionalDeps=
+AdditionalLibraryDirectories=
+AdditionalOptions= -std=c++17 --generate-code=arch=compute_75,code=[compute_75,sm_75] --generate-code=arch=compute_80,code=[compute_80,sm_80] --generate-code=arch=compute_86,code=[compute_86,sm_86] -Xcompiler="/EHsc -Ob2"
+AdditionalOptions= -std=c++17 --generate-code=arch=compute_75,code=[compute_75,sm_75] --generate-code=arch=compute_80,code=[compute_80,sm_80] --generate-code=arch=compute_86,code=[compute_86,sm_86] -Xcompiler="/EHsc -Ob2"
+CodeGeneration=
+CodeGeneration=
+CompileOut=fashion_mnist_trainer.dir\Release\fungi.obj
+CudaRuntime=Static
+CudaToolkitCustomDir=
+DebugInformationFormat=None
+DebugInformationFormat=None
+Defines=;_WINDOWS;NDEBUG;CMAKE_INTDIR="Release";_MBCS;;WIN32;_WINDOWS;NDEBUG;CMAKE_INTDIR="Release"
+Emulation=false
+EnableVirtualArchInFatbin=true
+ExtensibleWholeProgramCompilation=false
+FastCompile=Off
+FastMath=true
+GenerateLineInfo=false
+GenerateRelocatableDeviceCode=true
+GPUDebugInfo=false
+GPUDebugInfo=false
+HostDebugInfo=false
+Include=E:\Fashion_MNIST_Optic_Evolution\src;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cccl
+Inputs=
+InterleaveSourceInPTX=false
+Keep=false
+KeepDir=fashion_.2B82790A\x64\Release
+LinkOut=
+MaxRegCount=0
+NvccCompilation=compile
+NvccPath=
+Optimization=O2
+Optimization=O2
+PerformDeviceLink=
+PerformDeviceLinkTimeOptimization=
+PtxAsOptionV=false
+RequiredIncludes=
+Runtime=MD
+Runtime=MD
+RuntimeChecks=Default
+RuntimeChecks=Default
+SplitCompile=Default
+SplitCompileCustomThreads=
+TargetMachinePlatform=64
+TargetMachinePlatform=64
+TypeInfo=true
+TypeInfo=true
+UseHostDefines=true
+UseHostInclude=false
+UseHostLibraryDependencies=
+UseHostLibraryDirectories=
+Warning=W1
+Warning=W1

fungi.cu1408063174.deps ADDED Viewed

	@@ -0,0 +1,181 @@

+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_malloc.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_math.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_memcpy_s.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_memory.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_search.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_share.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_stdio_config.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_terminate.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wconio.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wctype.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wdirect.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wio.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wprocess.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wstdio.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wstdlib.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wstring.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wtime.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\crtdbg.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\ctype.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\errno.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\float.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\locale.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\malloc.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\math.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\memory.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\share.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\stddef.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\stdio.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\stdlib.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\string.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\sys\stat.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\sys\types.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\time.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\wchar.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\builtin_types.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\channel_descriptor.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\common_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\cudacc_ext.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\device_double_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\device_double_functions.hpp
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\device_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\device_functions.hpp
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\host_config.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\host_defines.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\math_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\math_functions.hpp
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\sm_100_rt.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\sm_70_rt.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\sm_80_rt.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\sm_90_rt.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cuda_device_runtime_api.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cuda_runtime.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cuda_runtime_api.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand_discrete.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand_discrete2.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand_globals.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand_kernel.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand_lognormal.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand_mrg32k3a.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand_mtgp32.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand_mtgp32_kernel.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand_normal.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand_normal_static.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand_philox4x32_x.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand_poisson.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand_precalc.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\curand_uniform.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\device_atomic_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\device_launch_parameters.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\device_types.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\driver_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\driver_types.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\library_types.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\nv\detail\__preprocessor
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\nv\detail\__target_macros
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\nv\target
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_20_atomic_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_20_intrinsics.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_30_intrinsics.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_32_atomic_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_32_intrinsics.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_35_atomic_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_35_intrinsics.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_60_atomic_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_61_intrinsics.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\surface_indirect_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\surface_types.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\texture_indirect_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\texture_types.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\vector_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\vector_functions.hpp
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\vector_types.h
+E:\Fashion_MNIST_Optic_Evolution\src\fungi.hpp
+E:\Fashion_MNIST_Optic_Evolution\src\fungi_Paremetres.hpp
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\algorithm
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\ammintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\atomic
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cctype
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cerrno
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cfloat
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\climits
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\clocale
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cmath
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\concurrencysal.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\crtdefs.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cstddef
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cstdint
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cstdio
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cstdlib
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cstring
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\ctime
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cwchar
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\eh.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\emmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\exception
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\immintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\initializer_list
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\intrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\intrin0.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\intrin0.inl.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\ios
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\iosfwd
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\iterator
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\limits
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\limits.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\memory
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\mmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\new
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\nmmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\numeric
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\pmmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\random
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\sal.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\setjmp.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\smmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\stdexcept
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\stdint.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\streambuf
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\system_error
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\tmmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\typeinfo
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\type_traits
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\use_ansi.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\utility
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vadefs.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vcruntime.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vcruntime_exception.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vcruntime_new.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vcruntime_new_debug.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vcruntime_string.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vcruntime_typeinfo.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vector
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\wmmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xatomic.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xbit_ops.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xcall_once.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xerrc.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xfacet
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xiosbase
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xkeycheck.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xlocale
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xlocinfo
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xlocnum
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xmemory
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xmmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xstring
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xthreads.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xtimec.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xtr1common
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xutility
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\yvals.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\yvals_core.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\zmmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\__msvc_bit_utils.hpp
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\__msvc_int128.hpp
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\__msvc_iter_core.hpp
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\__msvc_sanitizer_annotate_container.hpp
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\__msvc_system_error_abi.hpp
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\__msvc_xlocinfo_types.hpp

fungi.hpp ADDED Viewed

	@@ -0,0 +1,51 @@

+#pragma once
+#include <vector>
+#include <cstdint>
+/**
+ * Fungal (mycelium) population in SoA layout for coalesced GPU access.
+ * Each fungus contributes a local basis φ_h(x,y) to build amplitude/phase masks.
+ */
+struct FungiSoA {
+  int F = 0;                  // number of fungi
+  int H = 0, W = 0;           // mask geometry
+  // Genome / state
+  std::vector<float> x, y;    // position in [0, W-1] & [0, H-1]
+  std::vector<float> sigma;   // radius
+  std::vector<float> alpha;   // anisotropy (ellipse)
+  std::vector<float> theta;   // orientation
+  std::vector<float> a_base;  // amplitude base coeff
+  std::vector<float> p_base;  // phase base coeff
+  // Energy & life-cycle
+  std::vector<float> energy;
+  std::vector<float> mass;
+  std::vector<int>   age;
+  void resize(int F_, int H_, int W_);
+  void init_random(unsigned seed, float sigma_min=1.5f, float sigma_max=5.5f);
+  void adjust_population(int newF, unsigned seed);
+};
+/** Build masks A(x,y), P(x,y) from fungi population (GPU). */
+void fungi_build_masks_GPU(const FungiSoA& pop,
+                           float* d_A, float* d_P, // [H*W]
+                           int tiles_y=7, int tiles_x=7);
+/** Evolution step (GPU): compute per-fungus reward and update energy, size, reproduction. */
+void fungi_evolve_GPU(FungiSoA& pop,
+                      const float* d_grad_map, // e.g., gradient magnitude per pixel [H*W]
+                      int evo_pairs,
+                      float food=0.05f, float decay=0.98f, float death_th=-0.5f,
+                      float cost=1e-3f, unsigned seed=1337);
+/** Host helpers: download masks (optional debug). */
+void download_mask(float* h, const float* d, int HW);
+/** DIAGNOSTIC TOOLS: Visual debugging for fungi evolution */
+void fungi_export_debug_images(const FungiSoA& pop,
+                               const float* d_A, const float* d_P,
+                               const float* d_grad_map,
+                               const char* prefix = "debug");
+void fungi_create_test_pattern(float* h_pattern, int H, int W, int pattern_type = 0);
+void fungi_analyze_mask_statistics(const float* d_A, const float* d_P, int HW);

fungi.obj ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:123d84b27a1106c530f09011a8086502de2325579bba75504f4bde07d9958e9a
+size 1392711

fungi_Paremetres.hpp ADDED Viewed

	@@ -0,0 +1,29 @@

+// --- Evolution parameters (tunable) ---
+struct EvoParams {
+  float food   = 0.05f;   // reward scale ("food" per unit merit)
+  float decay  = 0.98f;   // energy decay per step
+  float death_th = -0.5f; // energy threshold for death/reseed
+  float cost   = 5e-4f;   // FIX: Reduced metabolic cost for stability
+  // Gravity & motion
+  float G      = 1e-2f;   // gravitational constant
+  float eps2   = 1e-1f;   // softening (squared)
+  float dt     = 0.5f;    // integration step
+  float damp   = 0.90f;   // velocity damping
+  // Pairing & reproduction
+  float pair_dist = 6.0f; // FIX: Increased separation to reduce interference
+  int   offspring_per_pair = 3;
+  int   max_population = 0; // 0 = auto (cap at 1.5x current)
+  unsigned seed = 1337u;
+};
+/**
+ * Orchestrates a full ecology step:
+ *  1) Reward from gradient map (GPU)
+ *  2) Gravity forces & motion (GPU)
+ *  3) Energy update, growth/shrink, death marking (GPU)
+ *  4) Pairing & reproduction (HOST) with genetic recombination + mutation
+ *  5) Population capping and compaction (HOST)
+ */
+void fungi_ecology_step(FungiSoA& pop,
+                        const float* d_grad_map, // [H*W] (GPU pointer)
+                        const EvoParams& evo);

main.cpp ADDED Viewed

	@@ -0,0 +1,44 @@

+#include "data_loader.hpp"
+#include "training.hpp"
+#include <iostream>
+TrainConfig parse_cmd(int argc, char** argv) {
+    TrainConfig cfg;
+    for (int i = 1; i < argc; ++i) {
+        std::string a = argv[i];
+        auto next = [&](const char* name) -> std::string {
+            if (i + 1 >= argc) { std::cerr << "Missing value for " << name << "\n"; std::exit(1); }
+            return std::string(argv[++i]);
+        };
+        if      (a == "--data_dir") cfg.data_dir = next("--data_dir");
+        else if (a == "--epochs")   cfg.epochs = std::stoi(next("--epochs"));
+        else if (a == "--batch")    cfg.batch  = std::stoi(next("--batch"));
+        else if (a == "--lr")       cfg.lr     = std::stof(next("--lr"));
+        else if (a == "--fungi")    cfg.fungi_count = std::stoi(next("--fungi"));
+        else if (a == "--wd")       cfg.wd = std::stof(next("--wd"));
+        else if (a == "--accuracy_tolerance") cfg.accuracy_tolerance = std::stof(next("--accuracy_tolerance"));
+        else if (a == "--seed")      cfg.seed = std::stoul(next("--seed")); // Original default seed was 1337u
+        else std::cerr << "Unknown arg: " << a << "\n";
+    }
+    return cfg;
+}
+int main(int argc, char** argv) {
+    try {
+        TrainConfig cfg = parse_cmd(argc, argv);
+        std::cout << "[INFO] Loading Fashion-MNIST training data from: " << cfg.data_dir << "\n";
+        FashionMNISTSet train = load_fashion_mnist_data(cfg.data_dir, true);
+        std::cout << "[INFO] Loading Fashion-MNIST test data from: " << cfg.data_dir << "\n";
+        FashionMNISTSet test = load_fashion_mnist_data(cfg.data_dir, false);
+        train_model(train, test, cfg);
+        std::cout << "[DONE] Training finished.\n";
+        return 0;
+    } catch (const std::exception& e) {
+        std::cerr << "[FATAL] " << e.what() << "\n";
+        return 1;
+    }
+}

main.obj ADDED Viewed

Binary file (44.8 kB). View file

optical_model.cu ADDED Viewed

	@@ -0,0 +1,937 @@

+#include "optical_model.hpp"
+#include "utils.hpp"
+#include <iostream>
+#include <stdexcept>
+#include <cmath>
+#include <random>
+// --- Kernel Declarations ---
+__global__ void k_modulate(const float* x, const float* A, const float* P, cufftComplex* field, int N_pixels);
+__global__ void k_intensity_log1p(const cufftComplex* freq, float* y, int N_elements);
+// NEW: Two-layer MLP kernels
+__global__ void k_linear_relu_forward(const float* input, const float* W, const float* b, float* output, int B, int input_size, int output_size);
+__global__ void k_linear_forward_mlp(const float* input, const float* W, const float* b, float* output, int B, int input_size, int output_size);
+__global__ void k_relu_backward(const float* grad_output, const float* forward_output, float* grad_input, int N);
+__global__ void k_linear_backward_input(const float* grad_output, const float* W, float* grad_input, int B, int input_size, int output_size);
+__global__ void k_accum_linear_grads(const float* input, const float* grad_output, float* gW, float* gb, int B, int input_size, int output_size);
+__global__ void k_softmax_xent_loss_grad(const float* logits, const uint8_t* labels, float* grad_logits, float* total_loss, int B, int C);
+__global__ void k_reduce_grad_map(const float* grad_y, int B, int S, float* grad_map);
+__global__ void k_sigmoid(const float* logits, float* probs, int N);
+// NEW: Multi-scale optical processing kernels
+__global__ void k_downsample_2x2(const float* input, float* output, int input_h, int input_w, int B);
+__global__ void k_concatenate_features(const float* scale1, const float* scale2, const float* scale3,
+                                       float* multiscale, int B, int s1_size, int s2_size, int s3_size);
+// NEW: 6-scale mirror concatenation kernel
+__global__ void k_concatenate_6scale_mirror(const float* s1, const float* s2, const float* s3,
+                                            const float* s1_mir, const float* s2_mir, const float* s3_mir,
+                                            float* multiscale, int B, int s1_size, int s2_size, int s3_size);
+// NEW: Memory-efficient flip kernels for mirror architecture
+__global__ void k_flip_horizontal(const float* input, float* output, int height, int width, int B);
+__global__ void k_flip_vertical(const float* input, float* output, int height, int width, int B);
+// NEW: Diagnostic kernels for bottleneck analysis
+__global__ void k_analyze_activation_saturation(const float* activations, float* stats, int N);
+__global__ void k_analyze_gradient_flow(const float* gradients, float* stats, int N);
+__global__ void k_bottleneck_detector(const float* input_features, const float* hidden_act,
+                                     const float* logits, float* bottleneck_metrics,
+                                     int batch_size, int input_size, int hidden_size, int output_size);
+// BREAKTHROUGH: Rich FFT extraction preserving ALL complex information
+__global__ void k_intensity_magnitude_phase(const cufftComplex* freq, float* y, int N_elements);
+__global__ void k_rich_fft_extraction(const cufftComplex* freq, float* magnitude_out, float* phase_out, int N_elements);
+__global__ void k_concatenate_dual_channels(const float* magnitude_channel, const float* phase_channel,
+                                           float* rich_features, int B, int channel_size);
+// --- Device Memory Management ---
+// C++ OPTIMIZATION: Allocate persistent GPU buffers once
+void allocate_device_buffers(DeviceBuffers& db, int B) {
+    const size_t S = IMG_SIZE, H = HIDDEN_SIZE, C = NUM_CLASSES;
+    const size_t MS = MULTISCALE_SIZE; // Enhanced 6-scale mirror feature size = 2058
+    // Batch-dependent buffers
+    check_cuda(cudaMalloc(&db.d_batch_in, sizeof(float) * B * S), "alloc d_batch_in");
+    check_cuda(cudaMalloc(&db.d_batch_lbl, sizeof(uint8_t) * B), "alloc d_batch_lbl");
+    // Multi-scale optical processing buffers
+    check_cuda(cudaMalloc(&db.d_field_scale1, sizeof(cufftComplex) * B * SCALE_1_SIZE), "alloc d_field_scale1");
+    check_cuda(cudaMalloc(&db.d_freq_scale1, sizeof(cufftComplex) * B * SCALE_1_SIZE), "alloc d_freq_scale1");
+    check_cuda(cudaMalloc(&db.d_features_scale1, sizeof(float) * B * SCALE_1_SIZE), "alloc d_features_scale1");
+    check_cuda(cudaMalloc(&db.d_field_scale2, sizeof(cufftComplex) * B * SCALE_2_SIZE), "alloc d_field_scale2");
+    check_cuda(cudaMalloc(&db.d_freq_scale2, sizeof(cufftComplex) * B * SCALE_2_SIZE), "alloc d_freq_scale2");
+    check_cuda(cudaMalloc(&db.d_features_scale2, sizeof(float) * B * SCALE_2_SIZE), "alloc d_features_scale2");
+    check_cuda(cudaMalloc(&db.d_field_scale3, sizeof(cufftComplex) * B * SCALE_3_SIZE), "alloc d_field_scale3");
+    check_cuda(cudaMalloc(&db.d_freq_scale3, sizeof(cufftComplex) * B * SCALE_3_SIZE), "alloc d_freq_scale3");
+    check_cuda(cudaMalloc(&db.d_features_scale3, sizeof(float) * B * SCALE_3_SIZE), "alloc d_features_scale3");
+    // Mirror architecture: allocate mirrored feature buffers
+    check_cuda(cudaMalloc(&db.d_features_scale1_mirror, sizeof(float) * B * SCALE_1_SIZE), "alloc d_features_scale1_mirror");
+    check_cuda(cudaMalloc(&db.d_features_scale2_mirror, sizeof(float) * B * SCALE_2_SIZE), "alloc d_features_scale2_mirror");
+    check_cuda(cudaMalloc(&db.d_features_scale3_mirror, sizeof(float) * B * SCALE_3_SIZE), "alloc d_features_scale3_mirror");
+    // LEGACY: Rich dual-channel processing buffers (not used in intelligent solution)
+    // check_cuda(cudaMalloc(&db.d_magnitude_features, sizeof(float) * B * MS), "alloc d_magnitude_features");
+    // check_cuda(cudaMalloc(&db.d_phase_features, sizeof(float) * B * MS), "alloc d_phase_features");
+    check_cuda(cudaMalloc(&db.d_multiscale_features, sizeof(float) * B * MS), "alloc d_multiscale_features");
+    check_cuda(cudaMalloc(&db.d_hidden, sizeof(float) * B * H), "alloc d_hidden");
+    check_cuda(cudaMalloc(&db.d_logits, sizeof(float) * B * C), "alloc d_logits");
+    check_cuda(cudaMalloc(&db.d_probs, sizeof(float) * B * C), "alloc d_probs");
+    check_cuda(cudaMalloc(&db.d_grad_logits, sizeof(float) * B * C), "alloc d_grad_logits");
+    check_cuda(cudaMalloc(&db.d_grad_hidden, sizeof(float) * B * H), "alloc d_grad_hidden");
+    check_cuda(cudaMalloc(&db.d_grad_multiscale, sizeof(float) * B * MS), "alloc d_grad_multiscale");
+    // Fungi buffers
+    check_cuda(cudaMalloc(&db.d_A, sizeof(float) * S), "alloc d_A");
+    check_cuda(cudaMalloc(&db.d_P, sizeof(float) * S), "alloc d_P");
+    check_cuda(cudaMalloc(&db.d_grad_map, sizeof(float) * S), "alloc d_grad_map");
+    // C++ OPTIMIZATION: Persistent weight buffers (allocated once, updated in-place)
+    check_cuda(cudaMalloc(&db.d_W1, sizeof(float) * H * MS), "alloc persistent d_W1");
+    check_cuda(cudaMalloc(&db.d_b1, sizeof(float) * H), "alloc persistent d_b1");
+    check_cuda(cudaMalloc(&db.d_W2, sizeof(float) * C * H), "alloc persistent d_W2");
+    check_cuda(cudaMalloc(&db.d_b2, sizeof(float) * C), "alloc persistent d_b2");
+    check_cuda(cudaMalloc(&db.d_gW1, sizeof(float) * H * MS), "alloc persistent d_gW1");
+    check_cuda(cudaMalloc(&db.d_gb1, sizeof(float) * H), "alloc persistent d_gb1");
+    check_cuda(cudaMalloc(&db.d_gW2, sizeof(float) * C * H), "alloc persistent d_gW2");
+    check_cuda(cudaMalloc(&db.d_gb2, sizeof(float) * C), "alloc persistent d_gb2");
+    check_cuda(cudaMalloc(&db.d_loss_scalar, sizeof(float)), "alloc persistent d_loss_scalar");
+    // CRITICAL: Bottleneck detection buffer - [4] metrics array
+    check_cuda(cudaMalloc(&db.d_bottleneck_metrics, sizeof(float) * 4), "alloc bottleneck_metrics");
+}
+void free_device_buffers(DeviceBuffers& db) {
+    // Free batch-dependent buffers
+    if (db.d_batch_in) cudaFree(db.d_batch_in);
+    if (db.d_batch_lbl) cudaFree(db.d_batch_lbl);
+    // Free multi-scale optical processing buffers
+    if (db.d_field_scale1) cudaFree(db.d_field_scale1);
+    if (db.d_freq_scale1) cudaFree(db.d_freq_scale1);
+    if (db.d_features_scale1) cudaFree(db.d_features_scale1);
+    if (db.d_field_scale2) cudaFree(db.d_field_scale2);
+    if (db.d_freq_scale2) cudaFree(db.d_freq_scale2);
+    if (db.d_features_scale2) cudaFree(db.d_features_scale2);
+    if (db.d_field_scale3) cudaFree(db.d_field_scale3);
+    if (db.d_freq_scale3) cudaFree(db.d_freq_scale3);
+    if (db.d_features_scale3) cudaFree(db.d_features_scale3);
+    // Free mirror architecture buffers
+    if (db.d_features_scale1_mirror) cudaFree(db.d_features_scale1_mirror);
+    if (db.d_features_scale2_mirror) cudaFree(db.d_features_scale2_mirror);
+    if (db.d_features_scale3_mirror) cudaFree(db.d_features_scale3_mirror);
+    if (db.d_multiscale_features) cudaFree(db.d_multiscale_features);
+    if (db.d_hidden) cudaFree(db.d_hidden);
+    if (db.d_logits) cudaFree(db.d_logits);
+    if (db.d_probs) cudaFree(db.d_probs);
+    if (db.d_grad_logits) cudaFree(db.d_grad_logits);
+    if (db.d_grad_hidden) cudaFree(db.d_grad_hidden);
+    if (db.d_grad_multiscale) cudaFree(db.d_grad_multiscale);
+    // Free fungi buffers
+    if (db.d_A) cudaFree(db.d_A);
+    if (db.d_P) cudaFree(db.d_P);
+    if (db.d_grad_map) cudaFree(db.d_grad_map);
+    // Free persistent weight buffers
+    if (db.d_W1) cudaFree(db.d_W1);
+    if (db.d_b1) cudaFree(db.d_b1);
+    if (db.d_W2) cudaFree(db.d_W2);
+    if (db.d_b2) cudaFree(db.d_b2);
+    if (db.d_gW1) cudaFree(db.d_gW1);
+    if (db.d_gb1) cudaFree(db.d_gb1);
+    if (db.d_gW2) cudaFree(db.d_gW2);
+    if (db.d_gb2) cudaFree(db.d_gb2);
+    if (db.d_loss_scalar) cudaFree(db.d_loss_scalar);
+    // CRITICAL: Free bottleneck detection buffer
+    if (db.d_bottleneck_metrics) cudaFree(db.d_bottleneck_metrics);
+    // LEGACY: Free rich dual-channel buffers (not used in intelligent solution)
+    // if (db.d_magnitude_features) cudaFree(db.d_magnitude_features);
+    // if (db.d_phase_features) cudaFree(db.d_phase_features);
+}
+// --- Adam Updater (Host) ---
+static void adam_update(std::vector<float>& P, std::vector<float>& m, std::vector<float>& v,
+                        const float* g_dev, size_t n, float lr, float wd, int t) {
+    std::vector<float> g(n);
+    check_cuda(cudaMemcpy(g.data(), g_dev, sizeof(float) * n, cudaMemcpyDeviceToHost), "D2H grads for Adam");
+    float b1 = 0.9f, b2 = 0.999f, eps = 1e-8f;
+    float b1t = 1.f - std::pow(b1, (float)t);
+    float b2t = 1.f - std::pow(b2, (float)t);
+    for (size_t i = 0; i < n; ++i) {
+        m[i] = b1 * m[i] + (1 - b1) * g[i];
+        v[i] = b2 * v[i] + (1 - b2) * g[i] * g[i];
+        float mh = m[i] / b1t;
+        float vh = v[i] / b2t;
+        P[i] -= lr * (mh / (std::sqrt(vh) + eps) + wd * P[i]);
+    }
+}
+// C++ OPTIMIZATION: Efficient GPU weight management
+void upload_params_to_gpu(const OpticalParams& params, DeviceBuffers& db) {
+    const size_t MS = MULTISCALE_SIZE, H = HIDDEN_SIZE, C = NUM_CLASSES;
+    // Upload weights to persistent GPU buffers
+    check_cuda(cudaMemcpy(db.d_W1, params.W1.data(), sizeof(float) * H * MS, cudaMemcpyHostToDevice), "upload W1");
+    check_cuda(cudaMemcpy(db.d_b1, params.b1.data(), sizeof(float) * H, cudaMemcpyHostToDevice), "upload b1");
+    check_cuda(cudaMemcpy(db.d_W2, params.W2.data(), sizeof(float) * C * H, cudaMemcpyHostToDevice), "upload W2");
+    check_cuda(cudaMemcpy(db.d_b2, params.b2.data(), sizeof(float) * C, cudaMemcpyHostToDevice), "upload b2");
+}
+void download_params_from_gpu(OpticalParams& params, const DeviceBuffers& db) {
+    const size_t MS = MULTISCALE_SIZE, H = HIDDEN_SIZE, C = NUM_CLASSES;
+    // Download updated weights from GPU
+    check_cuda(cudaMemcpy(params.W1.data(), db.d_W1, sizeof(float) * H * MS, cudaMemcpyDeviceToHost), "download W1");
+    check_cuda(cudaMemcpy(params.b1.data(), db.d_b1, sizeof(float) * H, cudaMemcpyDeviceToHost), "download b1");
+    check_cuda(cudaMemcpy(params.W2.data(), db.d_W2, sizeof(float) * C * H, cudaMemcpyDeviceToHost), "download W2");
+    check_cuda(cudaMemcpy(params.b2.data(), db.d_b2, sizeof(float) * C, cudaMemcpyDeviceToHost), "download b2");
+}
+// --- Training Step for Multi-Scale Two-Layer MLP ---
+// CHANGE LOG: Multi-scale optical processing for 90%+ accuracy
+// FORWARD: multi_scale_features -> W1*features+b1 -> ReLU -> W2*hidden+b2 -> logits
+// BACKWARD: Full backpropagation through both layers with multi-scale features
+float train_batch(const float* h_batch_in, const uint8_t* h_batch_lbl,
+                  int B, FungiSoA& fungi, OpticalParams& params,
+                  DeviceBuffers& db, FFTPlan& fft,
+                  float lr, float wd, int t_adam) {
+    const int S = IMG_SIZE, H = HIDDEN_SIZE, C = NUM_CLASSES, MS = MULTISCALE_SIZE;
+    check_cuda(cudaMemcpy(db.d_batch_in, h_batch_in, sizeof(float) * B * S, cudaMemcpyHostToDevice), "H2D input");
+    check_cuda(cudaMemcpy(db.d_batch_lbl, h_batch_lbl, sizeof(uint8_t) * B, cudaMemcpyHostToDevice), "H2D labels");
+    // Multi-scale optical processing for 90%+ accuracy
+    fungi_build_masks_GPU(fungi, db.d_A, db.d_P);
+    // DIAGNOSTIC: Analyze mask statistics every few epochs
+    static int debug_counter = 0;
+    if (debug_counter % 10 == 0) { // Every 10 batches
+        fungi_analyze_mask_statistics(db.d_A, db.d_P, IMG_SIZE);
+    }
+    debug_counter++;
+    // Scale 1: Full resolution 28x28 = 784 features
+    k_modulate<<<(B * S + 255) / 256, 256>>>(db.d_batch_in, db.d_A, db.d_P, db.d_field_scale1, B * S);
+    cufftExecC2C(fft.plan_fwd_scale1, db.d_field_scale1, db.d_freq_scale1, CUFFT_FORWARD);
+    k_intensity_magnitude_phase<<<(B * SCALE_1_SIZE + 255) / 256, 256>>>(db.d_freq_scale1, db.d_features_scale1, B * SCALE_1_SIZE);
+    // Scale 2: Half resolution 14x14 = 196 features
+    k_downsample_2x2<<<(B * SCALE_2_SIZE + 255) / 256, 256>>>(db.d_batch_in, reinterpret_cast<float*>(db.d_field_scale2), IMG_H, IMG_W, B);
+    k_modulate<<<(B * SCALE_2_SIZE + 255) / 256, 256>>>(reinterpret_cast<float*>(db.d_field_scale2), db.d_A, db.d_P, db.d_field_scale2, B * SCALE_2_SIZE);
+    cufftExecC2C(fft.plan_fwd_scale2, db.d_field_scale2, db.d_freq_scale2, CUFFT_FORWARD);
+    k_intensity_magnitude_phase<<<(B * SCALE_2_SIZE + 255) / 256, 256>>>(db.d_freq_scale2, db.d_features_scale2, B * SCALE_2_SIZE);
+    // Scale 3: Quarter resolution 7x7 = 49 features (downsample from scale2)
+    k_downsample_2x2<<<(B * SCALE_3_SIZE + 255) / 256, 256>>>(db.d_features_scale2, reinterpret_cast<float*>(db.d_field_scale3), SCALE_2, SCALE_2, B);
+    k_modulate<<<(B * SCALE_3_SIZE + 255) / 256, 256>>>(reinterpret_cast<float*>(db.d_field_scale3), db.d_A, db.d_P, db.d_field_scale3, B * SCALE_3_SIZE);
+    cufftExecC2C(fft.plan_fwd_scale3, db.d_field_scale3, db.d_freq_scale3, CUFFT_FORWARD);
+    k_intensity_magnitude_phase<<<(B * SCALE_3_SIZE + 255) / 256, 256>>>(db.d_freq_scale3, db.d_features_scale3, B * SCALE_3_SIZE);
+    // Mirror processing: create horizontally flipped versions for enhanced features
+    k_flip_horizontal<<<(B * SCALE_1_SIZE + 255) / 256, 256>>>(db.d_features_scale1, db.d_features_scale1_mirror, SCALE_1, SCALE_1, B);
+    k_flip_horizontal<<<(B * SCALE_2_SIZE + 255) / 256, 256>>>(db.d_features_scale2, db.d_features_scale2_mirror, SCALE_2, SCALE_2, B);
+    k_flip_horizontal<<<(B * SCALE_3_SIZE + 255) / 256, 256>>>(db.d_features_scale3, db.d_features_scale3_mirror, SCALE_3, SCALE_3, B);
+    // INTELLIGENT SOLUTION: Enhanced 6-scale mirror with optimized FFT kernel
+    k_concatenate_6scale_mirror<<<(B * MS + 255) / 256, 256>>>(
+        db.d_features_scale1, db.d_features_scale2, db.d_features_scale3,
+        db.d_features_scale1_mirror, db.d_features_scale2_mirror, db.d_features_scale3_mirror,
+        db.d_multiscale_features, B, SCALE_1_SIZE, SCALE_2_SIZE, SCALE_3_SIZE);
+    // C++ OPTIMIZATION: Use persistent GPU buffers (NO malloc/free per batch!)
+    // Forward pass: Layer 1 (with ReLU) - Enhanced 2058 features with better FFT extraction
+    k_linear_relu_forward<<<(B*H+255)/256, 256>>>(db.d_multiscale_features, db.d_W1, db.d_b1, db.d_hidden, B, MS, H);
+    // Forward pass: Layer 2 (linear)
+    k_linear_forward_mlp<<<(B*C+255)/256, 256>>>(db.d_hidden, db.d_W2, db.d_b2, db.d_logits, B, H, C);
+    // CRITICAL: Real-time bottleneck detection - analyze information flow
+    static int bottleneck_counter = 0;
+    if (bottleneck_counter % 5 == 0) { // Every 5 batches for performance
+        cudaMemset(db.d_bottleneck_metrics, 0, sizeof(float) * 4);
+        int max_threads = fmaxf(fmaxf(MS, H), C);
+        k_bottleneck_detector<<<(max_threads + 255) / 256, 256>>>(
+            db.d_multiscale_features, db.d_hidden, db.d_logits, db.d_bottleneck_metrics,
+            B, MS, H, C);
+        // Download metrics and report critical bottlenecks
+        float h_metrics[4] = {0};
+        cudaMemcpy(h_metrics, db.d_bottleneck_metrics, sizeof(float) * 4, cudaMemcpyDeviceToHost);
+        float dead_features_pct = (h_metrics[0] / MS) * 100.0f;        // % dead input features
+        float dead_neurons_pct = (h_metrics[1] / H) * 100.0f;         // % dead hidden neurons
+        float saturated_neurons_pct = (h_metrics[2] / H) * 100.0f;    // % saturated hidden neurons
+        float poor_discrimination_pct = (h_metrics[3] / C) * 100.0f;  // % poor output discrimination
+        // ALERT: Critical bottleneck detection
+        if (dead_features_pct > 20.0f || dead_neurons_pct > 30.0f ||
+            saturated_neurons_pct > 30.0f || poor_discrimination_pct > 40.0f) {
+            printf("🚨 CRITICAL BOTTLENECK DETECTED:\n");
+            printf("   📉 Dead Features: %.1f%% | Dead Neurons: %.1f%% | Saturated: %.1f%% | Poor Discrim: %.1f%%\n",
+                   dead_features_pct, dead_neurons_pct, saturated_neurons_pct, poor_discrimination_pct);
+        }
+    }
+    bottleneck_counter++;
+    // Loss computation (using persistent buffer)
+    cudaMemset(db.d_loss_scalar, 0, sizeof(float));
+    k_softmax_xent_loss_grad<<<(B + 255) / 256, 256>>>(db.d_logits, (const uint8_t*)db.d_batch_lbl, db.d_grad_logits, db.d_loss_scalar, B, C);
+    // Backward pass: Layer 2 gradients (using persistent buffers)
+    cudaMemset(db.d_gW2, 0, sizeof(float)*C*H);
+    cudaMemset(db.d_gb2, 0, sizeof(float)*C);
+    k_accum_linear_grads<<<(C+255)/256, 256>>>(db.d_hidden, db.d_grad_logits, db.d_gW2, db.d_gb2, B, H, C);
+    // Backward pass: Hidden layer gradient
+    k_linear_backward_input<<<(B*H+255)/256, 256>>>(db.d_grad_logits, db.d_W2, db.d_grad_hidden, B, H, C);
+    // Backward pass: ReLU gradient
+    k_relu_backward<<<(B*H+255)/256, 256>>>(db.d_grad_hidden, db.d_hidden, db.d_grad_hidden, B*H);
+    // Backward pass: Layer 1 gradients (using persistent buffers with multi-scale)
+    cudaMemset(db.d_gW1, 0, sizeof(float)*H*MS);
+    cudaMemset(db.d_gb1, 0, sizeof(float)*H);
+    k_accum_linear_grads<<<(H+255)/256, 256>>>(db.d_multiscale_features, db.d_grad_hidden, db.d_gW1, db.d_gb1, B, MS, H);
+    // Backward pass: Multi-scale gradient for fungi (simplified - use scale 1 gradient)
+    k_linear_backward_input<<<(B*MS+255)/256, 256>>>(db.d_grad_hidden, db.d_W1, db.d_grad_multiscale, B, MS, H);
+    k_reduce_grad_map<<<(S + 255) / 256, 256>>>(db.d_grad_multiscale, B, S, db.d_grad_map); // Only use first S elements
+    // Adam updates for all parameters (using persistent buffers with multi-scale)
+    adam_update(params.W1, params.m_W1, params.v_W1, db.d_gW1, H * MS, lr, wd, t_adam);
+    adam_update(params.b1, params.m_b1, params.v_b1, db.d_gb1, H, lr, 0.0f, t_adam);
+    adam_update(params.W2, params.m_W2, params.v_W2, db.d_gW2, C * H, lr, wd, t_adam);
+    adam_update(params.b2, params.m_b2, params.v_b2, db.d_gb2, C, lr, 0.0f, t_adam);
+    // BUGFIX: Upload updated weights back to GPU after Adam
+    upload_params_to_gpu(params, db);
+    float h_loss;
+    check_cuda(cudaMemcpy(&h_loss, db.d_loss_scalar, sizeof(float), cudaMemcpyDeviceToHost), "D2H loss");
+    return h_loss / B;
+}
+// --- Inference for Multi-Scale Two-Layer MLP ---
+void infer_batch(const float* h_batch_in, int B,
+                 const FungiSoA& fungi, const OpticalParams& params,
+                 DeviceBuffers& db, FFTPlan& fft,
+                 std::vector<int>& out_predictions) {
+    const int S = IMG_SIZE, H = HIDDEN_SIZE, C = NUM_CLASSES, MS = MULTISCALE_SIZE;
+    check_cuda(cudaMemcpy(db.d_batch_in, h_batch_in, sizeof(float) * B * S, cudaMemcpyHostToDevice), "H2D infer input");
+    // Multi-scale optical processing for inference
+    fungi_build_masks_GPU(const_cast<FungiSoA&>(fungi), db.d_A, db.d_P);
+    // Scale 1: Full resolution 28x28 = 784 features
+    k_modulate<<<(B * S + 255) / 256, 256>>>(db.d_batch_in, db.d_A, db.d_P, db.d_field_scale1, B * S);
+    cufftExecC2C(fft.plan_fwd_scale1, db.d_field_scale1, db.d_freq_scale1, CUFFT_FORWARD);
+    k_intensity_magnitude_phase<<<(B * SCALE_1_SIZE + 255) / 256, 256>>>(db.d_freq_scale1, db.d_features_scale1, B * SCALE_1_SIZE);
+    // Scale 2: Half resolution 14x14 = 196 features
+    k_downsample_2x2<<<(B * SCALE_2_SIZE + 255) / 256, 256>>>(db.d_batch_in, reinterpret_cast<float*>(db.d_field_scale2), IMG_H, IMG_W, B);
+    k_modulate<<<(B * SCALE_2_SIZE + 255) / 256, 256>>>(reinterpret_cast<float*>(db.d_field_scale2), db.d_A, db.d_P, db.d_field_scale2, B * SCALE_2_SIZE);
+    cufftExecC2C(fft.plan_fwd_scale2, db.d_field_scale2, db.d_freq_scale2, CUFFT_FORWARD);
+    k_intensity_magnitude_phase<<<(B * SCALE_2_SIZE + 255) / 256, 256>>>(db.d_freq_scale2, db.d_features_scale2, B * SCALE_2_SIZE);
+    // Scale 3: Quarter resolution 7x7 = 49 features (downsample from scale2)
+    k_downsample_2x2<<<(B * SCALE_3_SIZE + 255) / 256, 256>>>(db.d_features_scale2, reinterpret_cast<float*>(db.d_field_scale3), SCALE_2, SCALE_2, B);
+    k_modulate<<<(B * SCALE_3_SIZE + 255) / 256, 256>>>(reinterpret_cast<float*>(db.d_field_scale3), db.d_A, db.d_P, db.d_field_scale3, B * SCALE_3_SIZE);
+    cufftExecC2C(fft.plan_fwd_scale3, db.d_field_scale3, db.d_freq_scale3, CUFFT_FORWARD);
+    k_intensity_magnitude_phase<<<(B * SCALE_3_SIZE + 255) / 256, 256>>>(db.d_freq_scale3, db.d_features_scale3, B * SCALE_3_SIZE);
+    // Mirror processing: create horizontally flipped versions for enhanced features
+    k_flip_horizontal<<<(B * SCALE_1_SIZE + 255) / 256, 256>>>(db.d_features_scale1, db.d_features_scale1_mirror, SCALE_1, SCALE_1, B);
+    k_flip_horizontal<<<(B * SCALE_2_SIZE + 255) / 256, 256>>>(db.d_features_scale2, db.d_features_scale2_mirror, SCALE_2, SCALE_2, B);
+    k_flip_horizontal<<<(B * SCALE_3_SIZE + 255) / 256, 256>>>(db.d_features_scale3, db.d_features_scale3_mirror, SCALE_3, SCALE_3, B);
+    // INTELLIGENT SOLUTION: Enhanced 6-scale mirror with optimized FFT kernel (INFERENCE)
+    k_concatenate_6scale_mirror<<<(B * MS + 255) / 256, 256>>>(
+        db.d_features_scale1, db.d_features_scale2, db.d_features_scale3,
+        db.d_features_scale1_mirror, db.d_features_scale2_mirror, db.d_features_scale3_mirror,
+        db.d_multiscale_features, B, SCALE_1_SIZE, SCALE_2_SIZE, SCALE_3_SIZE);
+    // C++ OPTIMIZATION: Use persistent GPU buffers for inference
+    // Forward pass: Layer 1 (with ReLU) - Enhanced 2058 features with better FFT extraction
+    k_linear_relu_forward<<<(B*H+255)/256, 256>>>(db.d_multiscale_features, db.d_W1, db.d_b1, db.d_hidden, B, MS, H);
+    // Forward pass: Layer 2 (linear)
+    k_linear_forward_mlp<<<(B*C+255)/256, 256>>>(db.d_hidden, db.d_W2, db.d_b2, db.d_logits, B, H, C);
+    std::vector<float> h_logits(B * C);
+    check_cuda(cudaMemcpy(h_logits.data(), db.d_logits, sizeof(float) * B * C, cudaMemcpyDeviceToHost), "D2H logits");
+    out_predictions.resize(B);
+    for (int b = 0; b < B; ++b) {
+        int best_class = 0;
+        float max_logit = h_logits[b * C];
+        for (int c = 1; c < C; ++c) {
+            if (h_logits[b * C + c] > max_logit) {
+                max_logit = h_logits[b * C + c];
+                best_class = c;
+            }
+        }
+        out_predictions[b] = best_class;
+    }
+}
+// --- Kernels ---
+__global__ void k_modulate(const float* x, const float* A, const float* P, cufftComplex* field, int N_pixels) {
+    int i = blockIdx.x * blockDim.x + threadIdx.x;
+    if (i >= N_pixels) return;
+    int pixel_idx = i % IMG_SIZE;
+    float input_val = x[i];
+    float amp = A[pixel_idx];
+    float phase = P[pixel_idx];
+    field[i].x = input_val * amp * cosf(phase);
+    field[i].y = input_val * amp * sinf(phase);
+}
+__global__ void k_intensity_log1p(const cufftComplex* freq, float* y, int N_elements) {
+    int i = blockIdx.x * blockDim.x + threadIdx.x;
+    if (i >= N_elements) return;
+    float intensity = freq[i].x * freq[i].x + freq[i].y * freq[i].y;
+    y[i] = log1pf(intensity);
+}
+// BOTTLENECK FIX: Enhanced extraction with magnitude and phase information
+__global__ void k_intensity_magnitude_phase(const cufftComplex* freq, float* y, int N_elements) {
+    int i = blockIdx.x * blockDim.x + threadIdx.x;
+    if (i >= N_elements) return;
+    float real = freq[i].x;
+    float imag = freq[i].y;
+    // Preserve both magnitude and phase information
+    float magnitude = sqrtf(real * real + imag * imag);
+    float phase = atan2f(imag, real);
+    // BREAKTHROUGH FIX: Instead of crushing to 1D, preserve rich information
+    // Method 1: Enhanced representation with multiple components
+    y[i] = log1pf(magnitude) + 0.5f * tanhf(phase) + 0.2f * (real / (fabsf(real) + 1e-6f)) + 0.1f * (imag / (fabsf(imag) + 1e-6f));
+}
+// REVOLUTIONARY: Rich FFT extraction - DOUBLES information capacity
+__global__ void k_rich_fft_extraction(const cufftComplex* freq, float* magnitude_out, float* phase_out, int N_elements) {
+    int i = blockIdx.x * blockDim.x + threadIdx.x;
+    if (i >= N_elements) return;
+    float real = freq[i].x;
+    float imag = freq[i].y;
+    // Preserve magnitude with enhanced dynamic range
+    float magnitude = sqrtf(real * real + imag * imag);
+    magnitude_out[i] = log1pf(magnitude) + 0.1f * atan2f(magnitude, 1.0f); // Enhanced magnitude
+    // Preserve phase with full resolution
+    float phase = atan2f(imag, real);
+    phase_out[i] = tanhf(2.0f * phase / 3.14159f); // Full phase preservation [-1,1]
+}
+// BREAKTHROUGH: Concatenate magnitude and phase channels into rich feature vector
+__global__ void k_concatenate_dual_channels(const float* magnitude_channel, const float* phase_channel,
+                                           float* rich_features, int B, int channel_size) {
+    int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    int total_size = 2 * channel_size; // magnitude + phase = double size
+    if (idx >= B * total_size) return;
+    int batch_idx = idx / total_size;
+    int feature_idx = idx % total_size;
+    if (feature_idx < channel_size) {
+        // First half: magnitude channel [0, channel_size)
+        rich_features[idx] = magnitude_channel[batch_idx * channel_size + feature_idx];
+    } else {
+        // Second half: phase channel [channel_size, 2*channel_size)
+        int phase_idx = feature_idx - channel_size;
+        rich_features[idx] = phase_channel[batch_idx * channel_size + phase_idx];
+    }
+}
+__global__ void k_linear_forward(const float* y, const float* W, const float* b, float* logits, int B, int S, int C) {
+    int batch_class = blockIdx.x * blockDim.x + threadIdx.x;
+    if (batch_class >= B * C) return;
+    int batch_idx = batch_class / C;
+    int class_idx = batch_class % C;
+    float sum = b[class_idx];
+    for (int s = 0; s < S; ++s) {
+        sum += W[class_idx * S + s] * y[batch_idx * S + s];
+    }
+    logits[batch_class] = sum;
+}
+__global__ void k_softmax_xent_loss_grad(const float* logits, const uint8_t* labels, float* grad_logits, float* total_loss, int B, int C) {
+    int b = blockIdx.x * blockDim.x + threadIdx.x;
+    if (b >= B) return;
+    const float* b_logits = logits + b * C;
+    float max_val = -1e20f;
+    for (int c = 0; c < C; ++c) {
+        if (b_logits[c] > max_val) max_val = b_logits[c];
+    }
+    float exp_sum = 0.f;
+    float exp_vals[10];
+    for (int c = 0; c < C; ++c) {
+        exp_vals[c] = expf(b_logits[c] - max_val);
+        exp_sum += exp_vals[c];
+    }
+    uint8_t true_label = labels[b];
+    float* b_grad = grad_logits + b * C;
+    float loss = 0.f;
+    for (int c = 0; c < C; ++c) {
+        float prob = exp_vals[c] / exp_sum;
+        b_grad[c] = prob - (c == true_label ? 1.f : 0.f);
+        if (c == true_label) {
+            loss = -logf(fmaxf(prob, 1e-9f));
+        }
+    }
+    atomicAdd(total_loss, loss);
+}
+__global__ void k_backprop_y(const float* grad_logits, const float* W, float* grad_y, int B, int S, int C) {
+    int batch_pixel = blockIdx.x * blockDim.x + threadIdx.x;
+    if (batch_pixel >= B * S) return;
+    int batch_idx = batch_pixel / S;
+    int pixel_idx = batch_pixel % S;
+    float sum = 0.f;
+    for (int c = 0; c < C; ++c) {
+        sum += grad_logits[batch_idx * C + c] * W[c * S + pixel_idx];
+    }
+    grad_y[batch_pixel] = sum;
+}
+__global__ void k_accum_grads_Wb(const float* y, const float* grad_logits, float* gW, float* gb, int B, int S, int C) {
+    int class_idx = blockIdx.x * blockDim.x + threadIdx.x;
+    if (class_idx >= C) return;
+    float gb_sum = 0.f;
+    for (int b = 0; b < B; ++b) {
+        gb_sum += grad_logits[b * C + class_idx];
+    }
+    gb[class_idx] = gb_sum;
+    for (int s = 0; s < S; ++s) {
+        float gw_sum = 0.f;
+        for (int b = 0; b < B; ++b) {
+            gw_sum += grad_logits[b * C + class_idx] * y[b * S + s];
+        }
+        gW[class_idx * S + s] = gw_sum;
+    }
+}
+__global__ void k_reduce_grad_map(const float* grad_y, int B, int S, float* grad_map) {
+    int pixel = blockIdx.x * blockDim.x + threadIdx.x;
+    if (pixel >= S) return;
+    float acc = 0.f;
+    for (int b = 0; b < B; ++b) {
+        acc += fabsf(grad_y[b * S + pixel]);
+    }
+    grad_map[pixel] = acc / static_cast<float>(B);
+}
+__global__ void k_sigmoid(const float* logits, float* probs, int N) {
+    int i = blockIdx.x * blockDim.x + threadIdx.x;
+    if (i >= N) return;
+    probs[i] = 1.f / (1.f + expf(-logits[i]));
+}
+void create_fft_plan(FFTPlan& fft, int batch) {
+    // Scale 1: 28x28 FFT plans
+    int n1[2] = {SCALE_1, SCALE_1};
+    cufftPlanMany(&fft.plan_fwd_scale1, 2, n1, nullptr, 1, SCALE_1_SIZE, nullptr, 1, SCALE_1_SIZE, CUFFT_C2C, batch);
+    cufftPlanMany(&fft.plan_inv_scale1, 2, n1, nullptr, 1, SCALE_1_SIZE, nullptr, 1, SCALE_1_SIZE, CUFFT_C2C, batch);
+    // Scale 2: 14x14 FFT plans
+    int n2[2] = {SCALE_2, SCALE_2};
+    cufftPlanMany(&fft.plan_fwd_scale2, 2, n2, nullptr, 1, SCALE_2_SIZE, nullptr, 1, SCALE_2_SIZE, CUFFT_C2C, batch);
+    cufftPlanMany(&fft.plan_inv_scale2, 2, n2, nullptr, 1, SCALE_2_SIZE, nullptr, 1, SCALE_2_SIZE, CUFFT_C2C, batch);
+    // Scale 3: 7x7 FFT plans
+    int n3[2] = {SCALE_3, SCALE_3};
+    cufftPlanMany(&fft.plan_fwd_scale3, 2, n3, nullptr, 1, SCALE_3_SIZE, nullptr, 1, SCALE_3_SIZE, CUFFT_C2C, batch);
+    cufftPlanMany(&fft.plan_inv_scale3, 2, n3, nullptr, 1, SCALE_3_SIZE, nullptr, 1, SCALE_3_SIZE, CUFFT_C2C, batch);
+}
+void destroy_fft_plan(FFTPlan& fft) {
+    if (fft.plan_fwd_scale1) cufftDestroy(fft.plan_fwd_scale1);
+    if (fft.plan_inv_scale1) cufftDestroy(fft.plan_inv_scale1);
+    if (fft.plan_fwd_scale2) cufftDestroy(fft.plan_fwd_scale2);
+    if (fft.plan_inv_scale2) cufftDestroy(fft.plan_inv_scale2);
+    if (fft.plan_fwd_scale3) cufftDestroy(fft.plan_fwd_scale3);
+    if (fft.plan_inv_scale3) cufftDestroy(fft.plan_inv_scale3);
+}
+// CHANGE LOG: Updated initialization for 6-scale mirror two-layer MLP
+// ORIGINAL: Single layer initialization (IMG_SIZE=784)
+// NEW: Xavier/Glorot initialization for both layers (MULTISCALE_SIZE=2058)
+void init_params(OpticalParams& p, unsigned seed) {
+    std::mt19937 gen(seed);
+    // Xavier initialization: std = sqrt(2 / (fan_in + fan_out))
+    float std_W1 = std::sqrt(2.0f / (MULTISCALE_SIZE + HIDDEN_SIZE));
+    float std_W2 = std::sqrt(2.0f / (HIDDEN_SIZE + NUM_CLASSES));
+    std::normal_distribution<float> dist_W1(0.f, std_W1);
+    std::normal_distribution<float> dist_W2(0.f, std_W2);
+    // First layer: MULTISCALE_SIZE -> HIDDEN_SIZE
+    size_t W1_size = HIDDEN_SIZE * MULTISCALE_SIZE;
+    p.W1.resize(W1_size);
+    p.b1.resize(HIDDEN_SIZE);
+    p.m_W1.resize(W1_size);
+    p.v_W1.resize(W1_size);
+    p.m_b1.resize(HIDDEN_SIZE);
+    p.v_b1.resize(HIDDEN_SIZE);
+    for (size_t i = 0; i < W1_size; ++i) {
+        p.W1[i] = dist_W1(gen);
+        p.m_W1[i] = 0.f;
+        p.v_W1[i] = 0.f;
+    }
+    for (size_t i = 0; i < HIDDEN_SIZE; ++i) {
+        p.b1[i] = 0.f;
+        p.m_b1[i] = 0.f;
+        p.v_b1[i] = 0.f;
+    }
+    // Second layer: HIDDEN_SIZE -> NUM_CLASSES
+    size_t W2_size = NUM_CLASSES * HIDDEN_SIZE;
+    p.W2.resize(W2_size);
+    p.b2.resize(NUM_CLASSES);
+    p.m_W2.resize(W2_size);
+    p.v_W2.resize(W2_size);
+    p.m_b2.resize(NUM_CLASSES);
+    p.v_b2.resize(NUM_CLASSES);
+    for (size_t i = 0; i < W2_size; ++i) {
+        p.W2[i] = dist_W2(gen);
+        p.m_W2[i] = 0.f;
+        p.v_W2[i] = 0.f;
+    }
+    for (size_t i = 0; i < NUM_CLASSES; ++i) {
+        p.b2[i] = 0.f;
+        p.m_b2[i] = 0.f;
+        p.v_b2[i] = 0.f;
+    }
+}
+// --- NEW KERNELS FOR TWO-LAYER MLP ---
+// Linear layer with ReLU activation: output = ReLU(W * input + b)
+__global__ void k_linear_relu_forward(const float* input, const float* W, const float* b, float* output, int B, int input_size, int output_size) {
+    int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    if (idx >= B * output_size) return;
+    int batch_idx = idx / output_size;
+    int out_idx = idx % output_size;
+    float sum = b[out_idx];
+    for (int i = 0; i < input_size; ++i) {
+        sum += W[out_idx * input_size + i] * input[batch_idx * input_size + i];
+    }
+    output[idx] = fmaxf(0.0f, sum); // ReLU activation
+}
+// Linear layer without activation: output = W * input + b
+__global__ void k_linear_forward_mlp(const float* input, const float* W, const float* b, float* output, int B, int input_size, int output_size) {
+    int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    if (idx >= B * output_size) return;
+    int batch_idx = idx / output_size;
+    int out_idx = idx % output_size;
+    float sum = b[out_idx];
+    for (int i = 0; i < input_size; ++i) {
+        sum += W[out_idx * input_size + i] * input[batch_idx * input_size + i];
+    }
+    output[idx] = sum;
+}
+// ReLU backward: grad_input = grad_output * (forward_output > 0)
+__global__ void k_relu_backward(const float* grad_output, const float* forward_output, float* grad_input, int N) {
+    int i = blockIdx.x * blockDim.x + threadIdx.x;
+    if (i >= N) return;
+    grad_input[i] = grad_output[i] * (forward_output[i] > 0.0f ? 1.0f : 0.0f);
+}
+// Linear backward (input gradients): grad_input = W^T * grad_output
+__global__ void k_linear_backward_input(const float* grad_output, const float* W, float* grad_input, int B, int input_size, int output_size) {
+    int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    if (idx >= B * input_size) return;
+    int batch_idx = idx / input_size;
+    int in_idx = idx % input_size;
+    float sum = 0.0f;
+    for (int o = 0; o < output_size; ++o) {
+        sum += W[o * input_size + in_idx] * grad_output[batch_idx * output_size + o];
+    }
+    grad_input[idx] = sum;
+}
+// Accumulate gradients for linear layer weights and biases
+__global__ void k_accum_linear_grads(const float* input, const float* grad_output, float* gW, float* gb, int B, int input_size, int output_size) {
+    int out_idx = blockIdx.x * blockDim.x + threadIdx.x;
+    if (out_idx >= output_size) return;
+    // Accumulate bias gradient
+    float gb_sum = 0.0f;
+    for (int b = 0; b < B; ++b) {
+        gb_sum += grad_output[b * output_size + out_idx];
+    }
+    gb[out_idx] = gb_sum;
+    // Accumulate weight gradients
+    for (int in_idx = 0; in_idx < input_size; ++in_idx) {
+        float gw_sum = 0.0f;
+        for (int b = 0; b < B; ++b) {
+            gw_sum += grad_output[b * output_size + out_idx] * input[b * input_size + in_idx];
+        }
+        gW[out_idx * input_size + in_idx] = gw_sum;
+    }
+}
+// --- NEW KERNELS FOR MULTI-SCALE OPTICAL PROCESSING ---
+// Downsample 28x28 to 14x14 using 2x2 average pooling
+__global__ void k_downsample_2x2(const float* input, float* output, int input_h, int input_w, int B) {
+    int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    int output_h = input_h / 2;
+    int output_w = input_w / 2;
+    int output_size = output_h * output_w;
+    if (idx >= B * output_size) return;
+    int batch_idx = idx / output_size;
+    int out_pixel = idx % output_size;
+    int out_y = out_pixel / output_w;
+    int out_x = out_pixel % output_w;
+    // Average 2x2 region
+    float sum = 0.0f;
+    for (int dy = 0; dy < 2; ++dy) {
+        for (int dx = 0; dx < 2; ++dx) {
+            int in_y = out_y * 2 + dy;
+            int in_x = out_x * 2 + dx;
+            int in_pixel = in_y * input_w + in_x;
+            sum += input[batch_idx * (input_h * input_w) + in_pixel];
+        }
+    }
+    output[idx] = sum * 0.25f; // Average of 4 pixels
+}
+// Concatenate multi-scale features: [scale1 | scale2 | scale3]
+__global__ void k_concatenate_features(const float* scale1, const float* scale2, const float* scale3,
+                                       float* multiscale, int B, int s1_size, int s2_size, int s3_size) {
+    int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    int total_size = s1_size + s2_size + s3_size;
+    if (idx >= B * total_size) return;
+    int batch_idx = idx / total_size;
+    int feature_idx = idx % total_size;
+    if (feature_idx < s1_size) {
+        // Copy from scale1
+        multiscale[idx] = scale1[batch_idx * s1_size + feature_idx];
+    } else if (feature_idx < s1_size + s2_size) {
+        // Copy from scale2
+        int s2_idx = feature_idx - s1_size;
+        multiscale[idx] = scale2[batch_idx * s2_size + s2_idx];
+    } else {
+        // Copy from scale3
+        int s3_idx = feature_idx - s1_size - s2_size;
+        multiscale[idx] = scale3[batch_idx * s3_size + s3_idx];
+    }
+}
+// Memory-efficient horizontal flip: flip left-right in place
+__global__ void k_flip_horizontal(const float* input, float* output, int height, int width, int B) {
+    int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    int total_pixels = B * height * width;
+    if (idx >= total_pixels) return;
+    int batch_idx = idx / (height * width);
+    int pixel_idx = idx % (height * width);
+    int row = pixel_idx / width;
+    int col = pixel_idx % width;
+    // Flip column: new_col = width - 1 - col
+    int flipped_col = width - 1 - col;
+    int flipped_idx = batch_idx * (height * width) + row * width + flipped_col;
+    output[idx] = input[flipped_idx];
+}
+// Memory-efficient vertical flip: flip top-bottom in place
+__global__ void k_flip_vertical(const float* input, float* output, int height, int width, int B) {
+    int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    int total_pixels = B * height * width;
+    if (idx >= total_pixels) return;
+    int batch_idx = idx / (height * width);
+    int pixel_idx = idx % (height * width);
+    int row = pixel_idx / width;
+    int col = pixel_idx % width;
+    // Flip row: new_row = height - 1 - row
+    int flipped_row = height - 1 - row;
+    int flipped_idx = batch_idx * (height * width) + flipped_row * width + col;
+    output[idx] = input[flipped_idx];
+}
+// 6-scale mirror concatenation: [s1 | s2 | s3 | s1_mir | s2_mir | s3_mir]
+__global__ void k_concatenate_6scale_mirror(const float* s1, const float* s2, const float* s3,
+                                            const float* s1_mir, const float* s2_mir, const float* s3_mir,
+                                            float* multiscale, int B, int s1_size, int s2_size, int s3_size) {
+    int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    int single_size = s1_size + s2_size + s3_size; // 1029
+    int total_size = 2 * single_size; // 2058
+    if (idx >= B * total_size) return;
+    int batch_idx = idx / total_size;
+    int feature_idx = idx % total_size;
+    if (feature_idx < single_size) {
+        // First half: normal features [s1 | s2 | s3]
+        if (feature_idx < s1_size) {
+            // Copy from scale1
+            multiscale[idx] = s1[batch_idx * s1_size + feature_idx];
+        } else if (feature_idx < s1_size + s2_size) {
+            // Copy from scale2
+            int s2_idx = feature_idx - s1_size;
+            multiscale[idx] = s2[batch_idx * s2_size + s2_idx];
+        } else {
+            // Copy from scale3
+            int s3_idx = feature_idx - s1_size - s2_size;
+            multiscale[idx] = s3[batch_idx * s3_size + s3_idx];
+        }
+    } else {
+        // Second half: mirrored features [s1_mir | s2_mir | s3_mir]
+        int mirror_idx = feature_idx - single_size;
+        if (mirror_idx < s1_size) {
+            // Copy from mirrored scale1
+            multiscale[idx] = s1_mir[batch_idx * s1_size + mirror_idx];
+        } else if (mirror_idx < s1_size + s2_size) {
+            // Copy from mirrored scale2
+            int s2_idx = mirror_idx - s1_size;
+            multiscale[idx] = s2_mir[batch_idx * s2_size + s2_idx];
+        } else {
+            // Copy from mirrored scale3
+            int s3_idx = mirror_idx - s1_size - s2_size;
+            multiscale[idx] = s3_mir[batch_idx * s3_size + s3_idx];
+        }
+    }
+}
+// ================= BOTTLENECK ANALYSIS KERNELS =================
+// Analyze activation saturation (ReLU dead neurons)
+__global__ void k_analyze_activation_saturation(const float* activations, float* stats, int N) {
+    int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    if (idx >= N) return;
+    float val = activations[idx];
+    // Use atomic operations to gather statistics
+    if (val <= 1e-6f) {
+        atomicAdd(&stats[0], 1.0f); // Dead neurons (ReLU=0)
+    } else if (val >= 0.99f) {
+        atomicAdd(&stats[1], 1.0f); // Saturated neurons
+    }
+    atomicAdd(&stats[2], val);      // Sum for mean
+    atomicAdd(&stats[3], val*val);  // Sum squares for variance
+}
+// Analyze gradient flow (vanishing/exploding gradients)
+__global__ void k_analyze_gradient_flow(const float* gradients, float* stats, int N) {
+    int idx = blockIdx.x * blockDim.x + threadIdx.x;
+    if (idx >= N) return;
+    float grad = gradients[idx];
+    float abs_grad = fabsf(grad);
+    if (abs_grad < 1e-6f) {
+        atomicAdd(&stats[0], 1.0f); // Vanishing gradients
+    } else if (abs_grad > 10.0f) {
+        atomicAdd(&stats[1], 1.0f); // Exploding gradients
+    }
+    atomicAdd(&stats[2], abs_grad); // Sum for mean
+}
+// CRITICAL: Real-time bottleneck detection - identifies where information is lost
+__global__ void k_bottleneck_detector(const float* input_features, const float* hidden_act,
+                                     const float* logits, float* bottleneck_metrics,
+                                     int batch_size, int input_size, int hidden_size, int output_size) {
+    int tid = blockIdx.x * blockDim.x + threadIdx.x;
+    // Feature diversity analysis (input layer) - detect information collapse
+    if (tid < input_size) {
+        float feature_sum = 0.0f, feature_var = 0.0f;
+        for (int b = 0; b < batch_size; b++) {
+            float val = input_features[b * input_size + tid];
+            feature_sum += val;
+        }
+        float mean = feature_sum / batch_size;
+        for (int b = 0; b < batch_size; b++) {
+            float val = input_features[b * input_size + tid];
+            feature_var += (val - mean) * (val - mean);
+        }
+        feature_var /= batch_size;
+        // Low variance = information loss (features all the same value)
+        if (feature_var < 1e-4f) atomicAdd(&bottleneck_metrics[0], 1.0f); // Dead features count
+    }
+    // Hidden activation analysis - detect neural saturation
+    if (tid < hidden_size) {
+        float hidden_sum = 0.0f;
+        for (int b = 0; b < batch_size; b++) {
+            hidden_sum += hidden_act[b * hidden_size + tid];
+        }
+        float hidden_mean = hidden_sum / batch_size;
+        // Saturation detection (critical bottleneck indicators)
+        if (hidden_mean < 0.01f) atomicAdd(&bottleneck_metrics[1], 1.0f);     // Dead neurons
+        if (hidden_mean > 0.99f) atomicAdd(&bottleneck_metrics[2], 1.0f);     // Saturated neurons
+    }
+    // Logits analysis (output bottleneck) - detect poor class discrimination
+    if (tid < output_size) {
+        float logit_range = 0.0f;
+        float min_logit = 1e10f, max_logit = -1e10f;
+        for (int b = 0; b < batch_size; b++) {
+            float val = logits[b * output_size + tid];
+            min_logit = fminf(min_logit, val);
+            max_logit = fmaxf(max_logit, val);
+        }
+        logit_range = max_logit - min_logit;
+        // Small range = poor discrimination ability (critical bottleneck)
+        if (logit_range < 0.1f) atomicAdd(&bottleneck_metrics[3], 1.0f); // Poor discrimination count
+    }
+}

optical_model.cu-1057474305.deps ADDED Viewed

	@@ -0,0 +1,168 @@

+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_malloc.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_math.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_memcpy_s.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_memory.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_search.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_share.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_stdio_config.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_terminate.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wconio.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wctype.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wdirect.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wio.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wprocess.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wstdio.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wstdlib.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wstring.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\corecrt_wtime.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\crtdbg.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\ctype.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\errno.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\float.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\locale.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\malloc.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\math.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\share.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\stddef.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\stdio.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\stdlib.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\string.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\sys\stat.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\sys\types.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\time.h
+C:\Program Files (x86)\Windows Kits\10\Include\10.0.26100.0\ucrt\wchar.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\builtin_types.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\channel_descriptor.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\common_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\cudacc_ext.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\device_double_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\device_double_functions.hpp
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\device_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\device_functions.hpp
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\host_config.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\host_defines.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\math_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\math_functions.hpp
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\sm_100_rt.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\sm_70_rt.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\sm_80_rt.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\crt\sm_90_rt.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cuComplex.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cuda_device_runtime_api.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cuda_runtime.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cuda_runtime_api.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cufft.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\device_atomic_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\device_launch_parameters.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\device_types.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\driver_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\driver_types.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\library_types.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_20_atomic_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_20_intrinsics.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_30_intrinsics.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_32_atomic_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_32_intrinsics.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_35_atomic_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_35_intrinsics.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_60_atomic_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\sm_61_intrinsics.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\surface_indirect_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\surface_types.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\texture_indirect_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\texture_types.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\vector_functions.h
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\vector_functions.hpp
+C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\vector_types.h
+E:\Fashion_MNIST_Optic_Evolution\src\fungi.hpp
+E:\Fashion_MNIST_Optic_Evolution\src\optical_model.hpp
+E:\Fashion_MNIST_Optic_Evolution\src\utils.hpp
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\algorithm
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\ammintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\atomic
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cctype
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cerrno
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cfloat
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\climits
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\clocale
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cmath
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\concurrencysal.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\crtdefs.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cstddef
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cstdint
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cstdio
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cstdlib
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cstring
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\ctime
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\cwchar
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\eh.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\emmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\exception
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\immintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\initializer_list
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\intrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\intrin0.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\intrin0.inl.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\ios
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\iosfwd
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\iostream
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\istream
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\iterator
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\limits
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\limits.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\memory
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\mmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\new
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\nmmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\ostream
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\pmmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\random
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\sal.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\setjmp.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\smmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\stdexcept
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\stdint.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\streambuf
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\string
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\system_error
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\tmmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\typeinfo
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\type_traits
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\use_ansi.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\utility
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vadefs.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vcruntime.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vcruntime_exception.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vcruntime_new.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vcruntime_new_debug.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vcruntime_string.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vcruntime_typeinfo.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\vector
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\wmmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xatomic.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xbit_ops.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xcall_once.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xerrc.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xfacet
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xiosbase
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xkeycheck.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xlocale
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xlocinfo
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xlocnum
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xmemory
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xmmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xstring
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xthreads.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xtimec.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xtr1common
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\xutility
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\yvals.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\yvals_core.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\zmmintrin.h
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\__msvc_bit_utils.hpp
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\__msvc_int128.hpp
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\__msvc_iter_core.hpp
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\__msvc_sanitizer_annotate_container.hpp
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\__msvc_system_error_abi.hpp
+E:\VS2022\VC\Tools\MSVC\14.38.33130\include\__msvc_xlocinfo_types.hpp

optical_model.cu.cache ADDED Viewed

	@@ -0,0 +1,57 @@

+Identity=..\src\optical_model.cu
+AdditionalCompilerOptions=
+AdditionalCompilerOptions=
+AdditionalDependencies=
+AdditionalDeps=
+AdditionalLibraryDirectories=
+AdditionalOptions= -std=c++17 --generate-code=arch=compute_75,code=[compute_75,sm_75] --generate-code=arch=compute_80,code=[compute_80,sm_80] --generate-code=arch=compute_86,code=[compute_86,sm_86] -Xcompiler="/EHsc -Ob2"
+AdditionalOptions= -std=c++17 --generate-code=arch=compute_75,code=[compute_75,sm_75] --generate-code=arch=compute_80,code=[compute_80,sm_80] --generate-code=arch=compute_86,code=[compute_86,sm_86] -Xcompiler="/EHsc -Ob2"
+CodeGeneration=
+CodeGeneration=
+CompileOut=fashion_mnist_trainer.dir\Release\optical_model.obj
+CudaRuntime=Static
+CudaToolkitCustomDir=
+DebugInformationFormat=None
+DebugInformationFormat=None
+Defines=;_WINDOWS;NDEBUG;CMAKE_INTDIR="Release";_MBCS;;WIN32;_WINDOWS;NDEBUG;CMAKE_INTDIR="Release"
+Emulation=false
+EnableVirtualArchInFatbin=true
+ExtensibleWholeProgramCompilation=false
+FastCompile=Off
+FastMath=true
+GenerateLineInfo=false
+GenerateRelocatableDeviceCode=true
+GPUDebugInfo=false
+GPUDebugInfo=false
+HostDebugInfo=false
+Include=E:\Fashion_MNIST_Optic_Evolution\src;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include\cccl
+Inputs=
+InterleaveSourceInPTX=false
+Keep=false
+KeepDir=fashion_.2B82790A\x64\Release
+LinkOut=
+MaxRegCount=0
+NvccCompilation=compile
+NvccPath=
+Optimization=O2
+Optimization=O2
+PerformDeviceLink=
+PerformDeviceLinkTimeOptimization=
+PtxAsOptionV=false
+RequiredIncludes=
+Runtime=MD
+Runtime=MD
+RuntimeChecks=Default
+RuntimeChecks=Default
+SplitCompile=Default
+SplitCompileCustomThreads=
+TargetMachinePlatform=64
+TargetMachinePlatform=64
+TypeInfo=true
+TypeInfo=true
+UseHostDefines=true
+UseHostInclude=false
+UseHostLibraryDependencies=
+UseHostLibraryDirectories=
+Warning=W1
+Warning=W1

optical_model.hpp ADDED Viewed

	@@ -0,0 +1,123 @@

+#pragma once
+#include <vector>
+#include <cuda_runtime.h>
+#include <cufft.h>
+#include "fungi.hpp"
+// Image geometry - Fashion-MNIST
+constexpr int IMG_H = 28;
+constexpr int IMG_W = 28;
+constexpr int IMG_SIZE = IMG_H * IMG_W;
+constexpr int NUM_CLASSES = 10; // Fashion-MNIST -> 10 classes
+// INTELLIGENT ROLLBACK: Enhanced FFT with stable 2058-feature architecture
+// ORIGINAL: Single scale FFT (28x28 = 784 features)
+// PREVIOUS: 3-scale FFT (28x28 + 14x14 + 7x7 = 1029 features)
+// CURRENT: Enhanced 6-scale mirror FFT (3 normal + 3 mirrored = 2058 features) - OPTIMIZED INFORMATION
+constexpr int SCALE_1 = 28; // Full resolution - fine details
+constexpr int SCALE_2 = 14; // Half resolution - texture patterns
+constexpr int SCALE_3 = 7;  // Quarter resolution - edge patterns
+constexpr int SCALE_1_SIZE = SCALE_1 * SCALE_1; // 784
+constexpr int SCALE_2_SIZE = SCALE_2 * SCALE_2; // 196
+constexpr int SCALE_3_SIZE = SCALE_3 * SCALE_3; // 49
+constexpr int SINGLE_SCALE_SIZE = SCALE_1_SIZE + SCALE_2_SIZE + SCALE_3_SIZE; // 1029 single features
+constexpr int MULTISCALE_SIZE = 2 * SINGLE_SCALE_SIZE; // 2058 total features (normal + mirrored)
+constexpr int HIDDEN_SIZE = 1800; // BALANCED: Optimal capacity for enhanced 2058-feature architecture
+struct OpticalParams {
+    // Multi-scale MLP: hidden = ReLU(W1 * multiscale_features + b1), logits = W2 * hidden + b2
+    std::vector<float> W1; // [HIDDEN_SIZE, MULTISCALE_SIZE] - First layer weights (2058 inputs)
+    std::vector<float> b1; // [HIDDEN_SIZE] - First layer bias
+    std::vector<float> W2; // [NUM_CLASSES, HIDDEN_SIZE] - Second layer weights
+    std::vector<float> b2; // [NUM_CLASSES] - Second layer bias
+    // Adam moments for all parameters
+    std::vector<float> m_W1, v_W1, m_b1, v_b1;
+    std::vector<float> m_W2, v_W2, m_b2, v_b2;
+};
+struct DeviceBuffers {
+    float* d_batch_in = nullptr;    // [B, IMG_SIZE]
+    uint8_t* d_batch_lbl = nullptr; // [B]
+    // Multi-scale optical processing buffers
+    cufftComplex* d_field_scale1 = nullptr;  // [B, SCALE_1_SIZE] - Full resolution field
+    cufftComplex* d_freq_scale1 = nullptr;   // [B, SCALE_1_SIZE] - Full resolution frequency
+    float* d_features_scale1 = nullptr;      // [B, SCALE_1_SIZE] - Full resolution features
+    cufftComplex* d_field_scale2 = nullptr;  // [B, SCALE_2_SIZE] - Half resolution field (14x14)
+    cufftComplex* d_freq_scale2 = nullptr;   // [B, SCALE_2_SIZE] - Half resolution frequency
+    float* d_features_scale2 = nullptr;      // [B, SCALE_2_SIZE] - Half resolution features
+    cufftComplex* d_field_scale3 = nullptr;  // [B, SCALE_3_SIZE] - Quarter resolution field (7x7)
+    cufftComplex* d_freq_scale3 = nullptr;   // [B, SCALE_3_SIZE] - Quarter resolution frequency
+    float* d_features_scale3 = nullptr;      // [B, SCALE_3_SIZE] - Quarter resolution features
+    // Mirror architecture: flipped versions for enhanced feature extraction
+    float* d_features_scale1_mirror = nullptr;  // [B, SCALE_1_SIZE] - Mirrored scale1 features
+    float* d_features_scale2_mirror = nullptr;  // [B, SCALE_2_SIZE] - Mirrored scale2 features
+    float* d_features_scale3_mirror = nullptr;  // [B, SCALE_3_SIZE] - Mirrored scale3 features
+    // BREAKTHROUGH: Rich dual-channel processing - separate magnitude and phase
+    float* d_magnitude_features = nullptr;       // [B, MIRROR_SCALE_SIZE] - All magnitude features (2058)
+    float* d_phase_features = nullptr;           // [B, MIRROR_SCALE_SIZE] - All phase features (2058)
+    float* d_multiscale_features = nullptr;  // [B, MULTISCALE_SIZE] - Enhanced mirror features (2058)
+    float* d_hidden = nullptr;               // [B, HIDDEN_SIZE] - Hidden layer activations
+    float* d_logits = nullptr;               // [B, NUM_CLASSES]
+    float* d_probs = nullptr;                // [B, NUM_CLASSES]
+    float* d_grad_logits = nullptr;          // [B, NUM_CLASSES]
+    float* d_grad_hidden = nullptr;          // [B, HIDDEN_SIZE] - Hidden layer gradients
+    float* d_grad_multiscale = nullptr;      // [B, MULTISCALE_SIZE] - Multi-scale gradients
+    float* d_A = nullptr;           // [IMG_SIZE]
+    float* d_P = nullptr;           // [IMG_SIZE]
+    float* d_grad_map = nullptr;    // [IMG_SIZE]
+    // C++ OPTIMIZATION: Persistent weight buffers in GPU memory
+    float* d_W1 = nullptr;          // [HIDDEN_SIZE, MULTISCALE_SIZE] - Persistent weights
+    float* d_b1 = nullptr;          // [HIDDEN_SIZE] - Persistent biases
+    float* d_W2 = nullptr;          // [NUM_CLASSES, HIDDEN_SIZE] - Persistent weights
+    float* d_b2 = nullptr;          // [NUM_CLASSES] - Persistent biases
+    float* d_gW1 = nullptr;         // [HIDDEN_SIZE, MULTISCALE_SIZE] - Persistent gradients
+    float* d_gb1 = nullptr;         // [HIDDEN_SIZE] - Persistent gradients
+    float* d_gW2 = nullptr;         // [NUM_CLASSES, HIDDEN_SIZE] - Persistent gradients
+    float* d_gb2 = nullptr;         // [NUM_CLASSES] - Persistent gradients
+    float* d_loss_scalar = nullptr; // [1] - Persistent loss buffer
+    // CRITICAL: Bottleneck detection buffers
+    float* d_bottleneck_metrics = nullptr; // [4] - Real-time bottleneck analysis
+};
+struct FFTPlan {
+    cufftHandle plan_fwd_scale1{};  // 28x28 FFT plan
+    cufftHandle plan_inv_scale1{};
+    cufftHandle plan_fwd_scale2{};  // 14x14 FFT plan
+    cufftHandle plan_inv_scale2{};
+    cufftHandle plan_fwd_scale3{};  // 7x7 FFT plan
+    cufftHandle plan_inv_scale3{};
+};
+void allocate_device_buffers(DeviceBuffers& db, int batch);
+void free_device_buffers(DeviceBuffers& db);
+void create_fft_plan(FFTPlan& fft, int batch);
+void destroy_fft_plan(FFTPlan& fft);
+void init_params(OpticalParams& p, unsigned seed);
+// C++ OPTIMIZATION: Initialize weights in GPU memory once
+void upload_params_to_gpu(const OpticalParams& params, DeviceBuffers& db);
+void download_params_from_gpu(OpticalParams& params, const DeviceBuffers& db);
+// C++ OPTIMIZATION: GPU-side Adam updates (no CPU transfers!)
+void adam_update_gpu(OpticalParams& params, const DeviceBuffers& db, float lr, float wd, int t_adam);
+float train_batch(const float* h_batch_in, const uint8_t* h_batch_lbl,
+                  int B, FungiSoA& fungi, OpticalParams& params,
+                  DeviceBuffers& db, FFTPlan& fft,
+                  float lr, float wd, int t_adam);
+void infer_batch(const float* h_batch_in, int B,
+                 const FungiSoA& fungi, const OpticalParams& params,
+                 DeviceBuffers& db, FFTPlan& fft,
+                 std::vector<int>& out_predictions);

optical_model.obj ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:3d4ebdb32d56b8f675ef86c38cf333d7b2ad4947c7dbdb21116df2c1c3cb8d7a
+size 326455

training.cpp ADDED Viewed

	@@ -0,0 +1,134 @@

+#include "training.hpp"
+#include "utils.hpp"
+#include "fungi_Paremetres.hpp"
+#include <iostream>
+#include <vector>
+#include <string>
+#include <numeric>
+#include <algorithm>
+#include <random>
+#include <iomanip>
+void train_model(const FashionMNISTSet& train, const FashionMNISTSet& test, TrainConfig& cfg) {
+    const int N_train = train.N;
+    const int N_test = test.N;
+    OpticalParams params;
+    init_params(params, cfg.seed);
+    FungiSoA fungi;
+    fungi.resize(cfg.fungi_count, IMG_H, IMG_W);
+    fungi.init_random(cfg.seed);
+    DeviceBuffers db;
+    allocate_device_buffers(db, cfg.batch);
+    // C++ OPTIMIZATION: Upload weights to GPU once at start
+    upload_params_to_gpu(params, db);
+    FFTPlan fft;
+    create_fft_plan(fft, cfg.batch);
+    std::vector<int> train_indices(N_train);
+    std::iota(train_indices.begin(), train_indices.end(), 0);
+    std::mt19937 rng(cfg.seed);
+    int adam_step = 0;
+    double prev_accuracy = -1.0;
+    for (int ep = 1; ep <= cfg.epochs; ++ep) {
+        std::shuffle(train_indices.begin(), train_indices.end(), rng);
+        double epoch_loss = 0.0;
+        int samples_seen = 0;
+        // --- Training Loop ---
+        for (int start = 0; start < N_train; start += cfg.batch) {
+            int current_B = std::min(cfg.batch, N_train - start);
+            std::vector<float> h_batch_in(current_B * IMG_SIZE);
+            std::vector<uint8_t> h_batch_lbl(current_B);
+            for (int i = 0; i < current_B; ++i) {
+                int idx = train_indices[start + i];
+                memcpy(&h_batch_in[i * IMG_SIZE], &train.images[idx * IMG_SIZE], IMG_SIZE * sizeof(float));
+                h_batch_lbl[i] = train.labels[idx];
+            }
+            adam_step++;
+            float loss = train_batch(h_batch_in.data(), h_batch_lbl.data(), current_B, fungi, params, db, fft, cfg.lr, cfg.wd, adam_step);
+            // Optional evolution step (disabled due to earlier memory issues)
+            /*
+            if (adam_step % 5 == 0) {
+                float* d_dummy_grad;
+                cudaMalloc(&d_dummy_grad, sizeof(float) * IMG_SIZE);
+                cudaMemset(d_dummy_grad, 0, sizeof(float) * IMG_SIZE);
+                EvoParams evo_cfg;
+                evo_cfg.seed = cfg.seed + adam_step;
+                fungi_ecology_step(fungi, d_dummy_grad, evo_cfg);
+                cudaFree(d_dummy_grad);
+            }
+            */
+            epoch_loss += loss * current_B;
+            samples_seen += current_B;
+            std::cout << "\r[Epoch " << ep << "] Progress: " << samples_seen << "/" << N_train
+                      << " Avg Loss: " << std::fixed << std::setprecision(5) << (epoch_loss / samples_seen)
+                      << std::flush;
+        }
+        std::cout << "\n";
+        // --- Evaluation ---
+        std::cout << "[INFO] Evaluating on test set for epoch " << ep << "...\n";
+        int correct_predictions = 0;
+        for (int start = 0; start < N_test; start += cfg.batch) {
+            int current_B = std::min(cfg.batch, N_test - start);
+            std::vector<float> h_batch_in(current_B * IMG_SIZE);
+            for (int i = 0; i < current_B; ++i) {
+                memcpy(&h_batch_in[i * IMG_SIZE], &test.images[(start + i) * IMG_SIZE], IMG_SIZE * sizeof(float));
+            }
+            std::vector<int> predictions;
+            infer_batch(h_batch_in.data(), current_B, fungi, params, db, fft, predictions);
+            for (int i = 0; i < current_B; ++i) {
+                if (predictions[i] == test.labels[start + i]) {
+                    correct_predictions++;
+                }
+            }
+        }
+        double accuracy = static_cast<double>(correct_predictions) / N_test;
+        std::cout << "[Epoch " << ep << " RESULT] Test Accuracy: "
+                  << std::fixed << std::setprecision(4) << (accuracy * 100.0) << "%\n";
+        if (prev_accuracy >= 0.0) {
+            double delta = accuracy - prev_accuracy;
+            if (delta > cfg.accuracy_tolerance) {
+                int target_fungi = static_cast<int>(std::ceil(static_cast<double>(fungi.F) * cfg.fungi_growth_rate));
+                target_fungi = std::max(cfg.fungi_min, std::min(cfg.fungi_max, target_fungi));
+                if (target_fungi > fungi.F) {
+                    fungi.adjust_population(target_fungi, cfg.seed + static_cast<unsigned>(ep * 17));
+                    cfg.fungi_count = fungi.F;
+                    std::cout << "[ADAPT] Accuracy improved by " << delta * 100.0
+                              << "% -> fungi population " << fungi.F << "\n";
+                }
+            } else if (delta < -cfg.accuracy_tolerance) {
+                int target_fungi = static_cast<int>(std::floor(static_cast<double>(fungi.F) * cfg.fungi_decay_rate));
+                target_fungi = std::max(cfg.fungi_min, std::min(cfg.fungi_max, target_fungi));
+                if (target_fungi < fungi.F) {
+                    fungi.adjust_population(target_fungi, cfg.seed + static_cast<unsigned>(ep * 23));
+                    cfg.fungi_count = fungi.F;
+                    std::cout << "[ADAPT] Accuracy decreased by " << -delta * 100.0
+                              << "% -> fungi population " << fungi.F << "\n";
+                }
+            }
+        }
+        prev_accuracy = accuracy;
+    }
+    free_device_buffers(db);
+    destroy_fft_plan(fft);
+}

training.hpp ADDED Viewed

	@@ -0,0 +1,22 @@

+#pragma once
+#include "data_loader.hpp"
+#include "optical_model.hpp"
+#include "fungi.hpp"
+struct TrainConfig {
+    std::string data_dir = "data";
+    int epochs = 100;
+    int batch = 256;
+    float lr = 1e-3f;
+    int fungi_count = 256;
+    int fungi_min = 128;
+    int fungi_max = 1024;
+    float fungi_growth_rate = 1.15f;
+    float fungi_decay_rate = 0.9f;
+    float accuracy_tolerance = 100.0f;
+    int smooth_accuracy_window = 5;
+    unsigned seed = 1337u;
+    float wd = 0.0f;
+};
+void train_model(const FashionMNISTSet& train, const FashionMNISTSet& test, TrainConfig& cfg);

training.obj ADDED Viewed

Binary file (54.6 kB). View file

utils.cpp ADDED Viewed

	@@ -0,0 +1,28 @@

+#include "utils.hpp"
+#include <iostream>
+#include <fstream>
+#include <stdexcept>
+#include <iomanip>
+void check_cuda(cudaError_t st, const char* msg) {
+    if (st != cudaSuccess) {
+        std::cerr << "[CUDA ERROR] " << msg << ": " << cudaGetErrorString(st) << "\n";
+        std::exit(1);
+    }
+}
+void write_submission_csv(const std::string& path,
+                          const std::vector<std::string>& ids,
+                          const std::vector<float>& probabilities) {
+    std::ofstream ofs(path);
+    if (!ofs.is_open()) {
+        throw std::runtime_error("Cannot open submission file: " + path);
+    }
+    ofs << "id,label\n";
+    ofs << std::fixed << std::setprecision(8);
+    for (size_t i = 0; i < ids.size(); ++i) {
+        ofs << ids[i] << "," << probabilities[i] << "\n";
+    }
+    ofs.close();
+    std::cout << "[INFO] Submission saved to: " << path << " (" << ids.size() << " rows)\n";
+}

utils.hpp ADDED Viewed

	@@ -0,0 +1,9 @@

+#pragma once
+#include <string>
+#include <vector>
+#include <cuda_runtime.h>
+void check_cuda(cudaError_t st, const char* msg);
+void write_submission_csv(const std::string& path,
+                          const std::vector<std::string>& ids,
+                          const std::vector<float>& probabilities);

utils.obj ADDED Viewed

Binary file (93 kB). View file