Enabling exporting symbols on windows (#1035)

* Enabling exporting symbols on windows

Small fix to allow for the built library to be used in windows #728

* added macro to allow for exported symbols on windows

* added macro to allow for exported symbols on windows

* removed cmake command

* added dllimport using torchvision_EXPORTS preprocessor

torchvision/csrc/models/alexnet.h

@@ -2,12 +2,13 @@
 #define ALEXNET_H
 
 #include <torch/torch.h>
+#include "general.h"
 
 namespace vision {
 namespace models {
 // AlexNet model architecture from the
 // "One weird trick..." <https://arxiv.org/abs/1404.5997> paper.
-struct AlexNetImpl : torch::nn::Module {
+struct VISION_API AlexNetImpl : torch::nn::Module {
   torch::nn::Sequential features{nullptr}, classifier{nullptr};
 
   AlexNetImpl(int64_t num_classes = 1000);

torchvision/csrc/models/densenet.h

@@ -2,6 +2,7 @@
 #define DENSENET_H
 
 #include <torch/torch.h>
+#include "general.h"
 
 namespace vision {
 namespace models {
@@ -18,7 +19,7 @@ namespace models {
 //     bn_size (int) - multiplicative factor for number of bottle neck layers
 //         (i.e. bn_size * k features in the bottleneck layer)
 //     drop_rate (float) - dropout rate after each dense layer
-struct DenseNetImpl : torch::nn::Module {
+struct VISION_API DenseNetImpl : torch::nn::Module {
   torch::nn::Sequential features{nullptr};
   torch::nn::Linear classifier{nullptr};
 
@@ -33,7 +34,7 @@ struct DenseNetImpl : torch::nn::Module {
   torch::Tensor forward(torch::Tensor x);
 };
 
-struct DenseNet121Impl : DenseNetImpl {
+struct VISION_API DenseNet121Impl : DenseNetImpl {
   DenseNet121Impl(
       int64_t num_classes = 1000,
       int64_t growth_rate = 32,
@@ -43,7 +44,7 @@ struct DenseNet121Impl : DenseNetImpl {
       double drop_rate = 0);
 };
 
-struct DenseNet169Impl : DenseNetImpl {
+struct VISION_API DenseNet169Impl : DenseNetImpl {
   DenseNet169Impl(
       int64_t num_classes = 1000,
       int64_t growth_rate = 32,
@@ -53,7 +54,7 @@ struct DenseNet169Impl : DenseNetImpl {
       double drop_rate = 0);
 };
 
-struct DenseNet201Impl : DenseNetImpl {
+struct VISION_API DenseNet201Impl : DenseNetImpl {
   DenseNet201Impl(
       int64_t num_classes = 1000,
       int64_t growth_rate = 32,
@@ -63,7 +64,7 @@ struct DenseNet201Impl : DenseNetImpl {
       double drop_rate = 0);
 };
 
-struct DenseNet161Impl : DenseNetImpl {
+struct VISION_API DenseNet161Impl : DenseNetImpl {
   DenseNet161Impl(
       int64_t num_classes = 1000,
       int64_t growth_rate = 48,

torchvision/csrc/models/general.h

+#ifndef VISION_GENERAL_H
+#define VISION_GENERAL_H
+
+#ifdef _WIN32
+#if defined(torchvision_EXPORTS)
+#define VISION_API __declspec(dllexport)
+#else
+#define VISION_API __declspec(dllimport)
+#endif
+#else
+#define VISION_API
+#endif
+
+#endif // VISION_GENERAL_H
\ No newline at end of file

torchvision/csrc/models/googlenet.h

@@ -2,12 +2,13 @@
 #define GOOGLENET_H
 
 #include <torch/torch.h>
+#include "general.h"
 
 namespace vision {
 namespace models {
 
 namespace _googlenetimpl {
-struct BasicConv2dImpl : torch::nn::Module {
+struct VISION_API BasicConv2dImpl : torch::nn::Module {
   torch::nn::Conv2d conv{nullptr};
   torch::nn::BatchNorm bn{nullptr};
 
@@ -18,7 +19,7 @@ struct BasicConv2dImpl : torch::nn::Module {
 
 TORCH_MODULE(BasicConv2d);
 
-struct InceptionImpl : torch::nn::Module {
+struct VISION_API InceptionImpl : torch::nn::Module {
   BasicConv2d branch1{nullptr};
   torch::nn::Sequential branch2, branch3, branch4;
 
@@ -36,7 +37,7 @@ struct InceptionImpl : torch::nn::Module {
 
 TORCH_MODULE(Inception);
 
-struct InceptionAuxImpl : torch::nn::Module {
+struct VISION_API InceptionAuxImpl : torch::nn::Module {
   BasicConv2d conv{nullptr};
   torch::nn::Linear fc1{nullptr}, fc2{nullptr};
 
@@ -49,13 +50,13 @@ TORCH_MODULE(InceptionAux);
 
 } // namespace _googlenetimpl
 
-struct GoogLeNetOutput {
+struct VISION_API GoogLeNetOutput {
   torch::Tensor output;
   torch::Tensor aux1;
   torch::Tensor aux2;
 };
 
-struct GoogLeNetImpl : torch::nn::Module {
+struct VISION_API GoogLeNetImpl : torch::nn::Module {
   bool aux_logits, transform_input;
 
   _googlenetimpl::BasicConv2d conv1{nullptr}, conv2{nullptr}, conv3{nullptr};

torchvision/csrc/models/inception.h

@@ -2,11 +2,12 @@
 #define INCEPTION_H
 
 #include <torch/torch.h>
+#include "general.h"
 
 namespace vision {
 namespace models {
 namespace _inceptionimpl {
-struct BasicConv2dImpl : torch::nn::Module {
+struct VISION_API BasicConv2dImpl : torch::nn::Module {
   torch::nn::Conv2d conv{nullptr};
   torch::nn::BatchNorm bn{nullptr};
 
@@ -17,7 +18,7 @@ struct BasicConv2dImpl : torch::nn::Module {
 
 TORCH_MODULE(BasicConv2d);
 
-struct InceptionAImpl : torch::nn::Module {
+struct VISION_API InceptionAImpl : torch::nn::Module {
   BasicConv2d branch1x1, branch5x5_1, branch5x5_2, branch3x3dbl_1,
       branch3x3dbl_2, branch3x3dbl_3, branch_pool;
 
@@ -26,7 +27,7 @@ struct InceptionAImpl : torch::nn::Module {
   torch::Tensor forward(torch::Tensor x);
 };
 
-struct InceptionBImpl : torch::nn::Module {
+struct VISION_API InceptionBImpl : torch::nn::Module {
   BasicConv2d branch3x3, branch3x3dbl_1, branch3x3dbl_2, branch3x3dbl_3;
 
   InceptionBImpl(int64_t in_channels);
@@ -34,7 +35,7 @@ struct InceptionBImpl : torch::nn::Module {
   torch::Tensor forward(torch::Tensor x);
 };
 
-struct InceptionCImpl : torch::nn::Module {
+struct VISION_API InceptionCImpl : torch::nn::Module {
   BasicConv2d branch1x1{nullptr}, branch7x7_1{nullptr}, branch7x7_2{nullptr},
       branch7x7_3{nullptr}, branch7x7dbl_1{nullptr}, branch7x7dbl_2{nullptr},
       branch7x7dbl_3{nullptr}, branch7x7dbl_4{nullptr}, branch7x7dbl_5{nullptr},
@@ -45,7 +46,7 @@ struct InceptionCImpl : torch::nn::Module {
   torch::Tensor forward(torch::Tensor x);
 };
 
-struct InceptionDImpl : torch::nn::Module {
+struct VISION_API InceptionDImpl : torch::nn::Module {
   BasicConv2d branch3x3_1, branch3x3_2, branch7x7x3_1, branch7x7x3_2,
       branch7x7x3_3, branch7x7x3_4;
 
@@ -54,7 +55,7 @@ struct InceptionDImpl : torch::nn::Module {
   torch::Tensor forward(torch::Tensor x);
 };
 
-struct InceptionEImpl : torch::nn::Module {
+struct VISION_API InceptionEImpl : torch::nn::Module {
   BasicConv2d branch1x1, branch3x3_1, branch3x3_2a, branch3x3_2b,
       branch3x3dbl_1, branch3x3dbl_2, branch3x3dbl_3a, branch3x3dbl_3b,
       branch_pool;
@@ -64,7 +65,7 @@ struct InceptionEImpl : torch::nn::Module {
   torch::Tensor forward(torch::Tensor x);
 };
 
-struct InceptionAuxImpl : torch::nn::Module {
+struct VISION_API InceptionAuxImpl : torch::nn::Module {
   BasicConv2d conv0;
   BasicConv2d conv1;
   torch::nn::Linear fc;
@@ -83,7 +84,7 @@ TORCH_MODULE(InceptionAux);
 
 } // namespace _inceptionimpl
 
-struct InceptionV3Output {
+struct VISION_API InceptionV3Output {
   torch::Tensor output;
   torch::Tensor aux;
 };
@@ -91,7 +92,7 @@ struct InceptionV3Output {
 // Inception v3 model architecture from
 //"Rethinking the Inception Architecture for Computer Vision"
 //<http://arxiv.org/abs/1512.00567>
-struct InceptionV3Impl : torch::nn::Module {
+struct VISION_API InceptionV3Impl : torch::nn::Module {
   bool aux_logits, transform_input;
 
   _inceptionimpl::BasicConv2d Conv2d_1a_3x3{nullptr}, Conv2d_2a_3x3{nullptr},

torchvision/csrc/models/mobilenet.h

@@ -2,10 +2,11 @@
 #define MOBILENET_H
 
 #include <torch/torch.h>
+#include "general.h"
 
 namespace vision {
 namespace models {
-struct MobileNetV2Impl : torch::nn::Module {
+struct VISION_API MobileNetV2Impl : torch::nn::Module {
   int64_t last_channel;
   torch::nn::Sequential features, classifier;
 

torchvision/csrc/models/resnet.h

@@ -2,6 +2,7 @@
 #define RESNET_H
 
 #include <torch/torch.h>
+#include "general.h"
 
 namespace vision {
 namespace models {
@@ -19,7 +20,7 @@ torch::nn::Conv2d conv3x3(
 // 1x1 convolution
 torch::nn::Conv2d conv1x1(int64_t in, int64_t out, int64_t stride = 1);
 
-struct BasicBlock : torch::nn::Module {
+struct VISION_API BasicBlock : torch::nn::Module {
   template <typename Block>
   friend struct vision::models::ResNetImpl;
 
@@ -42,7 +43,7 @@ struct BasicBlock : torch::nn::Module {
   torch::Tensor forward(torch::Tensor x);
 };
 
-struct Bottleneck : torch::nn::Module {
+struct VISION_API Bottleneck : torch::nn::Module {
   template <typename Block>
   friend struct vision::models::ResNetImpl;
 
@@ -184,40 +185,40 @@ torch::Tensor ResNetImpl<Block>::forward(torch::Tensor x) {
   return x;
 }
 
-struct ResNet18Impl : ResNetImpl<_resnetimpl::BasicBlock> {
+struct VISION_API ResNet18Impl : ResNetImpl<_resnetimpl::BasicBlock> {
   ResNet18Impl(int64_t num_classes = 1000, bool zero_init_residual = false);
 };
 
-struct ResNet34Impl : ResNetImpl<_resnetimpl::BasicBlock> {
+struct VISION_API ResNet34Impl : ResNetImpl<_resnetimpl::BasicBlock> {
   ResNet34Impl(int64_t num_classes = 1000, bool zero_init_residual = false);
 };
 
-struct ResNet50Impl : ResNetImpl<_resnetimpl::Bottleneck> {
+struct VISION_API ResNet50Impl : ResNetImpl<_resnetimpl::Bottleneck> {
   ResNet50Impl(int64_t num_classes = 1000, bool zero_init_residual = false);
 };
 
-struct ResNet101Impl : ResNetImpl<_resnetimpl::Bottleneck> {
+struct VISION_API ResNet101Impl : ResNetImpl<_resnetimpl::Bottleneck> {
   ResNet101Impl(int64_t num_classes = 1000, bool zero_init_residual = false);
 };
 
-struct ResNet152Impl : ResNetImpl<_resnetimpl::Bottleneck> {
+struct VISION_API ResNet152Impl : ResNetImpl<_resnetimpl::Bottleneck> {
   ResNet152Impl(int64_t num_classes = 1000, bool zero_init_residual = false);
 };
 
-struct ResNext50_32x4dImpl : ResNetImpl<_resnetimpl::Bottleneck> {
+struct VISION_API ResNext50_32x4dImpl : ResNetImpl<_resnetimpl::Bottleneck> {
   ResNext50_32x4dImpl(
       int64_t num_classes = 1000,
       bool zero_init_residual = false);
 };
 
-struct ResNext101_32x8dImpl : ResNetImpl<_resnetimpl::Bottleneck> {
+struct VISION_API ResNext101_32x8dImpl : ResNetImpl<_resnetimpl::Bottleneck> {
   ResNext101_32x8dImpl(
       int64_t num_classes = 1000,
       bool zero_init_residual = false);
 };
 
 template <typename Block>
-struct ResNet : torch::nn::ModuleHolder<ResNetImpl<Block>> {
+struct VISION_API ResNet : torch::nn::ModuleHolder<ResNetImpl<Block>> {
   using torch::nn::ModuleHolder<ResNetImpl<Block>>::ModuleHolder;
 };
 

torchvision/csrc/models/shufflenetv2.h

@@ -2,11 +2,12 @@
 #define SHUFFLENETV2_H
 
 #include <torch/torch.h>
+#include "general.h"
 
 namespace vision {
 namespace models {
 
-struct ShuffleNetV2Impl : torch::nn::Module {
+struct VISION_API ShuffleNetV2Impl : torch::nn::Module {
   std::vector<int64_t> _stage_out_channels;
   torch::nn::Sequential conv1{nullptr}, stage2, stage3, stage4, conv5{nullptr};
   torch::nn::Linear fc{nullptr};
@@ -19,19 +20,19 @@ struct ShuffleNetV2Impl : torch::nn::Module {
   torch::Tensor forward(torch::Tensor x);
 };
 
-struct ShuffleNetV2_x0_5Impl : ShuffleNetV2Impl {
+struct VISION_API ShuffleNetV2_x0_5Impl : ShuffleNetV2Impl {
   ShuffleNetV2_x0_5Impl(int64_t num_classes = 1000);
 };
 
-struct ShuffleNetV2_x1_0Impl : ShuffleNetV2Impl {
+struct VISION_API ShuffleNetV2_x1_0Impl : ShuffleNetV2Impl {
   ShuffleNetV2_x1_0Impl(int64_t num_classes = 1000);
 };
 
-struct ShuffleNetV2_x1_5Impl : ShuffleNetV2Impl {
+struct VISION_API ShuffleNetV2_x1_5Impl : ShuffleNetV2Impl {
   ShuffleNetV2_x1_5Impl(int64_t num_classes = 1000);
 };
 
-struct ShuffleNetV2_x2_0Impl : ShuffleNetV2Impl {
+struct VISION_API ShuffleNetV2_x2_0Impl : ShuffleNetV2Impl {
   ShuffleNetV2_x2_0Impl(int64_t num_classes = 1000);
 };
 

torchvision/csrc/models/squeezenet.h

@@ -2,10 +2,11 @@
 #define SQUEEZENET_H
 
 #include <torch/torch.h>
+#include "general.h"
 
 namespace vision {
 namespace models {
-struct SqueezeNetImpl : torch::nn::Module {
+struct VISION_API SqueezeNetImpl : torch::nn::Module {
   int64_t num_classes;
   torch::nn::Sequential features{nullptr}, classifier{nullptr};
 
@@ -17,7 +18,7 @@ struct SqueezeNetImpl : torch::nn::Module {
 // SqueezeNet model architecture from the "SqueezeNet: AlexNet-level
 // accuracy with 50x fewer parameters and <0.5MB model size"
 // <https://arxiv.org/abs/1602.07360> paper.
-struct SqueezeNet1_0Impl : SqueezeNetImpl {
+struct VISION_API SqueezeNet1_0Impl : SqueezeNetImpl {
   SqueezeNet1_0Impl(int64_t num_classes = 1000);
 };
 
@@ -25,7 +26,7 @@ struct SqueezeNet1_0Impl : SqueezeNetImpl {
 // <https://github.com/DeepScale/SqueezeNet/tree/master/SqueezeNet_v1.1>.
 // SqueezeNet 1.1 has 2.4x less computation and slightly fewer parameters
 // than SqueezeNet 1.0, without sacrificing accuracy.
-struct SqueezeNet1_1Impl : SqueezeNetImpl {
+struct VISION_API SqueezeNet1_1Impl : SqueezeNetImpl {
   SqueezeNet1_1Impl(int64_t num_classes = 1000);
 };
 

torchvision/csrc/models/vgg.h

@@ -2,10 +2,11 @@
 #define VGG_H
 
 #include <torch/torch.h>
+#include "general.h"
 
 namespace vision {
 namespace models {
-struct VGGImpl : torch::nn::Module {
+struct VISION_API VGGImpl : torch::nn::Module {
   torch::nn::Sequential features{nullptr}, classifier{nullptr};
 
   void _initialize_weights();
@@ -19,42 +20,42 @@ struct VGGImpl : torch::nn::Module {
 };
 
 // VGG 11-layer model (configuration "A")
-struct VGG11Impl : VGGImpl {
+struct VISION_API VGG11Impl : VGGImpl {
   VGG11Impl(int64_t num_classes = 1000, bool initialize_weights = true);
 };
 
 // VGG 13-layer model (configuration "B")
-struct VGG13Impl : VGGImpl {
+struct VISION_API VGG13Impl : VGGImpl {
   VGG13Impl(int64_t num_classes = 1000, bool initialize_weights = true);
 };
 
 // VGG 16-layer model (configuration "D")
-struct VGG16Impl : VGGImpl {
+struct VISION_API VGG16Impl : VGGImpl {
   VGG16Impl(int64_t num_classes = 1000, bool initialize_weights = true);
 };
 
 // VGG 19-layer model (configuration "E")
-struct VGG19Impl : VGGImpl {
+struct VISION_API VGG19Impl : VGGImpl {
   VGG19Impl(int64_t num_classes = 1000, bool initialize_weights = true);
 };
 
 // VGG 11-layer model (configuration "A") with batch normalization
-struct VGG11BNImpl : VGGImpl {
+struct VISION_API VGG11BNImpl : VGGImpl {
   VGG11BNImpl(int64_t num_classes = 1000, bool initialize_weights = true);
 };
 
 // VGG 13-layer model (configuration "B") with batch normalization
-struct VGG13BNImpl : VGGImpl {
+struct VISION_API VGG13BNImpl : VGGImpl {
   VGG13BNImpl(int64_t num_classes = 1000, bool initialize_weights = true);
 };
 
 // VGG 16-layer model (configuration "D") with batch normalization
-struct VGG16BNImpl : VGGImpl {
+struct VISION_API VGG16BNImpl : VGGImpl {
   VGG16BNImpl(int64_t num_classes = 1000, bool initialize_weights = true);
 };
 
 // VGG 19-layer model (configuration 'E') with batch normalization
-struct VGG19BNImpl : VGGImpl {
+struct VISION_API VGG19BNImpl : VGGImpl {
   VGG19BNImpl(int64_t num_classes = 1000, bool initialize_weights = true);
 };