├── .gitignore ├── Data ├── ['DUTS-TR']_statistics.pth ├── __init__.py ├── dataloader.py └── preprocess.py ├── Models ├── PCSA.py ├── PCSA │ ├── .cproject │ ├── .project │ ├── PCSA_Module │ │ ├── .cproject │ │ ├── .project │ │ ├── CMakeLists.txt │ │ ├── cmakebuild │ │ │ ├── CMakeCache.txt │ │ │ ├── CMakeFiles │ │ │ │ ├── 3.10.2 │ │ │ │ │ ├── CMakeCCompiler.cmake │ │ │ │ │ ├── CMakeCXXCompiler.cmake │ │ │ │ │ ├── CMakeDetermineCompilerABI_C.bin │ │ │ │ │ ├── CMakeDetermineCompilerABI_CXX.bin │ │ │ │ │ ├── CMakeSystem.cmake │ │ │ │ │ ├── CompilerIdC │ │ │ │ │ │ ├── CMakeCCompilerId.c │ │ │ │ │ │ └── a.out │ │ │ │ │ └── CompilerIdCXX │ │ │ │ │ │ ├── CMakeCXXCompilerId.cpp │ │ │ │ │ │ └── a.out │ │ │ │ ├── CMakeDirectoryInformation.cmake │ │ │ │ ├── CMakeError.log │ │ │ │ ├── CMakeOutput.log │ │ │ │ ├── CMakeRuleHashes.txt │ │ │ │ ├── Makefile.cmake │ │ │ │ ├── Makefile2 │ │ │ │ ├── SA.dir │ │ │ │ │ ├── CXX.includecache │ │ │ │ │ ├── DependInfo.cmake │ │ │ │ │ ├── SA_generated_sa.cu.o │ │ │ │ │ ├── SA_generated_sa.cu.o.cmake │ │ │ │ │ ├── SA_generated_sa.cu.o.cmake.pre-gen │ │ │ │ │ ├── SA_generated_sa.cu.o.depend │ │ │ │ │ ├── build.make │ │ │ │ │ ├── cmake_clean.cmake │ │ │ │ │ ├── depend.internal │ │ │ │ │ ├── depend.make │ │ │ │ │ ├── flags.make │ │ │ │ │ ├── link.txt │ │ │ │ │ ├── progress.make │ │ │ │ │ └── sa_ext.cpp.o │ │ │ │ ├── TargetDirectories.txt │ │ │ │ ├── cmake.check_cache │ │ │ │ ├── feature_tests.bin │ │ │ │ ├── feature_tests.c │ │ │ │ ├── feature_tests.cxx │ │ │ │ └── progress.marks │ │ │ ├── Makefile │ │ │ ├── SA │ │ │ ├── cmake_install.cmake │ │ │ ├── detect_cuda_compute_capabilities.cpp │ │ │ └── detect_cuda_version.cc │ │ ├── reference.cpp │ │ ├── reference.h │ │ ├── sa.cu │ │ ├── sa.cu.bak │ │ ├── sa_ext.cpp │ │ ├── self_cuda │ │ │ └── function.py │ │ ├── timer.h │ │ └── utils.h │ ├── build │ │ ├── lib.linux-x86_64-3.6 │ │ │ └── self_cuda_backend.cpython-36m-x86_64-linux-gnu.so │ │ └── temp.linux-x86_64-3.6 │ │ │ └── SelfAttention_Module │ │ │ ├── reference.o │ │ │ ├── sa.o │ │ │ └── sa_ext.o │ ├── dist │ │ └── self_cuda-0.0.0-py3.6-linux-x86_64.egg │ ├── self_cuda.egg-info │ │ ├── PKG-INFO │ │ ├── SOURCES.txt │ │ ├── dependency_links.txt │ │ └── top_level.txt │ └── setup.py ├── __init__.py ├── lightrfb.py ├── mobilenetv3_pretrain.py ├── mobilenetv3temporal_PCSA.py └── statedict │ └── mobilenetv3-large.pth ├── README.md ├── build.sh ├── config.py ├── finetune.sh ├── finetune_temporal_distribute.py ├── pretrain.sh ├── pretrain_distribute.py ├── speed.py └── utils ├── Distribute └── engine.py └── utils.py /.gitignore: -------------------------------------------------------------------------------- 1 | **/__pycache__ 2 | .idea/ 3 | 4 | -------------------------------------------------------------------------------- /Data/['DUTS-TR']_statistics.pth: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Data/['DUTS-TR']_statistics.pth -------------------------------------------------------------------------------- /Data/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Data/__init__.py -------------------------------------------------------------------------------- /Data/dataloader.py: -------------------------------------------------------------------------------- 1 | import os 2 | from torch.utils.data import Dataset,DataLoader 3 | from Data.preprocess import * 4 | from PIL import Image 5 | import torch 6 | from config import config 7 | import glob 8 | ##pretrain_dataset 9 | class Pretrain(Dataset): 10 | def __init__(self,img_dataset_list,video_dataset_list,transform): 11 | self.file_list=[] 12 | for dataset in img_dataset_list: 13 | listfile=os.path.join(config.img_dataset_root,dataset+'.lst') 14 | with open(listfile,"r") as f: 15 | file_list=f.readlines() 16 | file_list=[filename.strip() for filename in file_list] 17 | self.file_list.extend([(os.path.join(config.img_dataset_root,filename.split(' ')[0][1:]),os.path.join(config.img_dataset_root,filename.split(' ')[1][1:])) for filename in file_list]) 18 | 19 | for dataset in video_dataset_list: 20 | path=os.path.join(config.video_dataset_root,dataset) 21 | video_list=glob.glob(path+"/**/*.jpg",recursive=True) 22 | self.file_list.extend([(filepath,filepath.replace("Imgs","ground-truth").replace("jpg","png")) for filepath in video_list]) 23 | 24 | self.img_label_transform = transform 25 | 26 | def __getitem__(self, idx): 27 | img_path, label_path = self.file_list[idx] 28 | img = Image.open(img_path).convert('RGB') 29 | label = Image.open(label_path).convert('L') 30 | img, label = self._process(img, label) 31 | return img,label 32 | 33 | def _process(self, img, label): 34 | img, label = self.img_label_transform(img, label) 35 | return img, label 36 | 37 | def __len__(self): 38 | return len(self.file_list) 39 | 40 | def get_pretrain_loader(): 41 | statistics = torch.load(config.data_statistics) 42 | trsf_main=Compose_imglabel([ 43 | Resize(config.size[0],config.size[1]), 44 | Random_crop_Resize(15), 45 | Random_horizontal_flip(0.5), 46 | toTensor(), 47 | Normalize(statistics["mean"],statistics["std"]) 48 | ]) 49 | trsf_scale1 = Compose_imglabel([ 50 | Resize(int(config.size[0] * 1.5), int(config.size[1] * 1.5)), 51 | # ColorAug(), 52 | Random_crop_Resize(50), 53 | Random_horizontal_flip(0.5), 54 | toTensor(), 55 | Normalize(statistics["mean"], statistics["std"]) 56 | ]) 57 | trsf_scale2 = Compose_imglabel([ 58 | Resize(int(config.size[0] * 1.25), int(config.size[1] * 1.25)), 59 | Random_crop_Resize(25), 60 | Random_horizontal_flip(0.5), 61 | toTensor(), 62 | Normalize(statistics["mean"], statistics["std"]) 63 | ]) 64 | trsf_scale3 = Compose_imglabel([ 65 | Resize(int(config.size[0] * 0.75), int(config.size[1] * 0.75)), 66 | Random_crop_Resize(15), 67 | Random_horizontal_flip(0.5), 68 | toTensor(), 69 | Normalize(statistics["mean"], statistics["std"]) 70 | ]) 71 | 72 | trsf_scale4 = Compose_imglabel([ 73 | Resize(int(config.size[0] * 0.5), int(config.size[1] * 0.5)), 74 | Random_horizontal_flip(0.5), 75 | toTensor(), 76 | Normalize(statistics["mean"], statistics["std"]) 77 | ]) 78 | train_loader=DataLoader(Pretrain(config.img_dataset_list,config.video_dataset_list,transform=trsf_main),batch_size=config.pretrain_batchsize,shuffle=True,drop_last=False,num_workers=8) 79 | multicale_loader=[ 80 | DataLoader(Pretrain(config.img_dataset_list, config.video_dataset_list, transform=trsf_scale1), 81 | batch_size=config.pretrain_batchsize, shuffle=True, drop_last=False,num_workers=8), 82 | DataLoader(Pretrain(config.img_dataset_list, config.video_dataset_list, transform=trsf_scale2), 83 | batch_size=config.pretrain_batchsize, shuffle=True, drop_last=False,num_workers=8), 84 | DataLoader(Pretrain(config.img_dataset_list, config.video_dataset_list, transform=trsf_scale3), 85 | batch_size=config.pretrain_batchsize, shuffle=True, drop_last=False,num_workers=8), 86 | DataLoader(Pretrain(config.img_dataset_list, config.video_dataset_list, transform=trsf_scale4), 87 | batch_size=config.pretrain_batchsize, shuffle=True, drop_last=False,num_workers=8) 88 | ] 89 | return train_loader,multicale_loader,statistics 90 | 91 | def get_pretrain_dataset(): 92 | statistics = torch.load(config.data_statistics) 93 | trsf_main=Compose_imglabel([ 94 | Resize(config.size[0],config.size[1]), 95 | Random_crop_Resize(15), 96 | Random_horizontal_flip(0.5), 97 | toTensor(), 98 | Normalize(statistics["mean"],statistics["std"]) 99 | ]) 100 | trsf_scale1 = Compose_imglabel([ 101 | Resize(int(config.size[0] * 1.5), int(config.size[1] * 1.5)), 102 | # ColorAug(), 103 | Random_crop_Resize(50), 104 | Random_horizontal_flip(0.5), 105 | toTensor(), 106 | Normalize(statistics["mean"], statistics["std"]) 107 | ]) 108 | trsf_scale2 = Compose_imglabel([ 109 | Resize(int(config.size[0] * 1.25), int(config.size[1] * 1.25)), 110 | Random_crop_Resize(25), 111 | Random_horizontal_flip(0.5), 112 | toTensor(), 113 | Normalize(statistics["mean"], statistics["std"]) 114 | ]) 115 | trsf_scale3 = Compose_imglabel([ 116 | Resize(int(config.size[0] * 0.75), int(config.size[1] * 0.75)), 117 | Random_crop_Resize(15), 118 | Random_horizontal_flip(0.5), 119 | toTensor(), 120 | Normalize(statistics["mean"], statistics["std"]) 121 | ]) 122 | 123 | trsf_scale4 = Compose_imglabel([ 124 | Resize(int(config.size[0] * 0.5), int(config.size[1] * 0.5)), 125 | Random_horizontal_flip(0.5), 126 | toTensor(), 127 | Normalize(statistics["mean"], statistics["std"]) 128 | ]) 129 | train_loader=Pretrain(config.img_dataset_list,config.video_dataset_list,transform=trsf_main) 130 | multicale_loader=[ 131 | Pretrain(config.img_dataset_list, config.video_dataset_list, transform=trsf_scale1), 132 | Pretrain(config.img_dataset_list, config.video_dataset_list, transform=trsf_scale2), 133 | Pretrain(config.img_dataset_list, config.video_dataset_list, transform=trsf_scale3), 134 | Pretrain(config.img_dataset_list, config.video_dataset_list, transform=trsf_scale4) 135 | ] 136 | return train_loader,multicale_loader,statistics 137 | 138 | class VideoDataset(Dataset): 139 | def __init__(self,video_dataset_list,transform=None,time_interval=1): 140 | super(VideoDataset, self).__init__() 141 | self.video_filelist=video_dataset_list 142 | self.time_clips=config.video_time_clips 143 | self.video_train_list = [] 144 | 145 | for video_name in video_dataset_list: 146 | video_root=os.path.join(config.video_dataset_root,video_name) 147 | cls_list=os.listdir(video_root) 148 | self.video_filelist={} 149 | for cls in cls_list: 150 | self.video_filelist[cls]=[] 151 | cls_path=os.path.join(video_root,cls) 152 | cls_img_path=os.path.join(cls_path,"Imgs") 153 | cls_label_path=os.path.join(cls_path,"ground-truth") 154 | tmp_list=os.listdir(cls_img_path) 155 | tmp_list.sort() 156 | for filename in tmp_list: 157 | self.video_filelist[cls].append(( 158 | os.path.join(cls_img_path,filename), 159 | os.path.join(cls_label_path,filename.replace(".jpg",".png")) 160 | )) 161 | #emsemble 162 | for cls in cls_list: 163 | li=self.video_filelist[cls] 164 | for begin in range(len(li)-(self.time_clips-1)*time_interval): 165 | batch_clips=[] 166 | for t in range(self.time_clips): 167 | batch_clips.append(li[begin+time_interval*t]) 168 | self.video_train_list.append(batch_clips) 169 | self.img_label_transform=transform 170 | 171 | def __getitem__(self, idx): 172 | img_label_li = self.video_train_list[idx] 173 | IMG = None 174 | LABEL = None 175 | img_li=[] 176 | label_li=[] 177 | for idx, (img_path, label_path) in enumerate(img_label_li): 178 | img = Image.open(img_path).convert('RGB') 179 | label = Image.open(label_path).convert('L') 180 | img_li.append(img) 181 | label_li.append(label) 182 | img_li,label_li=self.img_label_transform(img_li,label_li) 183 | for idx,(img,label) in enumerate(zip(img_li,label_li)): 184 | if IMG is not None: 185 | IMG[idx,:,:,:]=img 186 | LABEL[idx,:,:,:]=label 187 | else: 188 | IMG=torch.zeros(len(img_li),*(img.shape)) 189 | LABEL=torch.zeros(len(img_li),*(label.shape)) 190 | IMG[idx, :, :, :] = img 191 | LABEL[idx, :, :, :] = label 192 | return IMG, LABEL 193 | 194 | def __len__(self): 195 | return len(self.video_train_list) 196 | 197 | def get_video_loader(): 198 | statistics = torch.load(config.data_statistics) 199 | trsf_main = Compose_imglabel([ 200 | Resize_video(config.size[0], config.size[1]), 201 | Random_crop_Resize_Video(7), 202 | Random_horizontal_flip_video(0.5), 203 | toTensor_video(), 204 | Normalize_video(statistics["mean"], statistics["std"]) 205 | ]) 206 | 207 | trsf_scale1 = Compose_imglabel([ 208 | Resize_video(int(config.size[0]), int(config.size[1])), 209 | Random_crop_Resize_Video(22), 210 | Random_horizontal_flip_video(0.5), 211 | toTensor_video(), 212 | Normalize_video(statistics["mean"], statistics["std"]) 213 | ]) 214 | 215 | train_loader = DataLoader(VideoDataset(config.video_dataset_list, transform=trsf_main, time_interval=1), 216 | batch_size=config.video_batchsize, shuffle=True, num_workers=8) 217 | multiscale_loader = [ 218 | DataLoader(VideoDataset(config.video_dataset_list, transform=trsf_scale1, time_interval=1), 219 | batch_size=config.video_batchsize, 220 | shuffle=True, num_workers=8), 221 | DataLoader(VideoDataset(config.video_dataset_list, transform=trsf_scale1, time_interval=2), 222 | batch_size=config.video_batchsize, 223 | shuffle=True, num_workers=8), 224 | DataLoader(VideoDataset(config.video_dataset_list, transform=trsf_scale1, time_interval=3), 225 | batch_size=config.video_batchsize, 226 | shuffle=True, num_workers=8), 227 | DataLoader(VideoDataset(config.video_dataset_list, transform=trsf_scale1, time_interval=4), 228 | batch_size=config.video_batchsize, 229 | shuffle=True, num_workers=8), 230 | DataLoader(VideoDataset(config.video_dataset_list, transform=trsf_scale1, time_interval=5), 231 | batch_size=config.video_batchsize, 232 | shuffle=True, num_workers=8), 233 | DataLoader(VideoDataset(config.video_dataset_list, transform=trsf_scale1, time_interval=6), 234 | batch_size=config.video_batchsize, 235 | shuffle=True, num_workers=8)] 236 | return train_loader,multiscale_loader,statistics 237 | 238 | def get_video_dataset(): 239 | statistics = torch.load(config.data_statistics) 240 | trsf_main = Compose_imglabel([ 241 | Resize_video(config.size[0], config.size[1]), 242 | Random_crop_Resize_Video(7), 243 | Random_horizontal_flip_video(0.5), 244 | toTensor_video(), 245 | Normalize_video(statistics["mean"], statistics["std"]) 246 | ]) 247 | 248 | trsf_scale1 = Compose_imglabel([ 249 | Resize_video(int(config.size[0]), int(config.size[1])), 250 | Random_crop_Resize_Video(22), 251 | Random_horizontal_flip_video(0.5), 252 | toTensor_video(), 253 | Normalize_video(statistics["mean"], statistics["std"]) 254 | ]) 255 | 256 | train_loader = VideoDataset(config.video_dataset_list, transform=trsf_main, time_interval=1) 257 | multiscale_loader = [ 258 | VideoDataset(config.video_dataset_list, transform=trsf_scale1, time_interval=1), 259 | VideoDataset(config.video_dataset_list, transform=trsf_scale1, time_interval=2), 260 | VideoDataset(config.video_dataset_list, transform=trsf_scale1, time_interval=3), 261 | VideoDataset(config.video_dataset_list, transform=trsf_scale1, time_interval=4), 262 | VideoDataset(config.video_dataset_list, transform=trsf_scale1, time_interval=5), 263 | VideoDataset(config.video_dataset_list, transform=trsf_scale1, time_interval=6)] 264 | return train_loader,multiscale_loader,statistics 265 | 266 | if __name__=="__main__": 267 | pass -------------------------------------------------------------------------------- /Data/preprocess.py: -------------------------------------------------------------------------------- 1 | from PIL import Image 2 | import random 3 | from torchvision.transforms import Normalize as torchnorm 4 | from torchvision.transforms import ToTensor as torchtotensor 5 | class Compose_imglabel(object): 6 | def __init__(self, transforms): 7 | self.transforms = transforms 8 | 9 | def __call__(self, img,label): 10 | for t in self.transforms: 11 | img,label = t(img,label) 12 | return img,label 13 | 14 | class Random_vertical_flip(object): 15 | def _vertical_flip(self,img,label): 16 | return img.transpose(Image.FLIP_TOP_BOTTOM),label.transpose(Image.FLIP_TOP_BOTTOM) 17 | def __init__(self,prob): 18 | ''' 19 | :param prob: should be (0,1) 20 | ''' 21 | assert prob>=0 and prob<=1,"prob should be [0,1]" 22 | self.prob=prob 23 | def __call__(self, img,label): 24 | ''' 25 | flip img and label simultaneously 26 | :param img:should be PIL image 27 | :param label:should be PIL image 28 | :return: 29 | ''' 30 | assert isinstance(img, Image.Image),"should be PIL image" 31 | assert isinstance(label, Image.Image),"should be PIL image" 32 | if random.random()=0 and prob<=1,"prob should be [0,1]" 46 | self.prob=prob 47 | 48 | def __call__(self, img,label): 49 | ''' 50 | flip img and label simultaneously 51 | :param img:should be PIL image 52 | :param label:should be PIL image 53 | :return: 54 | ''' 55 | assert isinstance(img, Image.Image),"should be PIL image" 56 | assert isinstance(label, Image.Image),"should be PIL image" 57 | if random.random()=0 and prob<=1,"prob should be [0,1]" 156 | self.prob=prob 157 | 158 | def __call__(self, imgs,labels): 159 | ''' 160 | flip img and label simultaneously 161 | :param img:should be PIL image 162 | :param label:should be PIL image 163 | :return: 164 | ''' 165 | if random.random() 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | -------------------------------------------------------------------------------- /Models/PCSA/.project: -------------------------------------------------------------------------------- 1 | 2 | 3 | plocaltnonlocal 4 | 5 | 6 | 7 | 8 | 9 | org.eclipse.cdt.managedbuilder.core.genmakebuilder 10 | clean,full,incremental, 11 | 12 | 13 | 14 | 15 | org.eclipse.cdt.managedbuilder.core.ScannerConfigBuilder 16 | full,incremental, 17 | 18 | 19 | 20 | 21 | 22 | org.eclipse.cdt.core.cnature 23 | org.eclipse.cdt.core.ccnature 24 | org.eclipse.cdt.managedbuilder.core.managedBuildNature 25 | org.eclipse.cdt.managedbuilder.core.ScannerConfigNature 26 | 27 | 28 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/.cproject: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/.project: -------------------------------------------------------------------------------- 1 | 2 | 3 | selfatt 4 | 5 | 6 | 7 | 8 | 9 | org.eclipse.cdt.managedbuilder.core.genmakebuilder 10 | clean,full,incremental, 11 | 12 | 13 | 14 | 15 | org.eclipse.cdt.managedbuilder.core.ScannerConfigBuilder 16 | full,incremental, 17 | 18 | 19 | 20 | 21 | 22 | org.eclipse.cdt.core.cnature 23 | org.eclipse.cdt.core.ccnature 24 | org.eclipse.cdt.managedbuilder.core.managedBuildNature 25 | org.eclipse.cdt.managedbuilder.core.ScannerConfigNature 26 | 27 | 28 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/CMakeLists.txt: -------------------------------------------------------------------------------- 1 | ############################################################################ 2 | # 2019-08-05 3 | # Guyuchao 4 | ############################################################################ 5 | cmake_minimum_required(VERSION 3.0 FATAL_ERROR) 6 | project(SA) 7 | 8 | set(Torch_DIR /home/guyuchao/software/libtorch/share/cmake/Torch) 9 | 10 | find_package(Torch REQUIRED) 11 | find_package(CUDA REQUIRED) 12 | 13 | set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fPIC") 14 | set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fPIC") 15 | 16 | if(CUDA_FOUND) 17 | # add -Wextra compiler flag for gcc compilations 18 | if (UNIX) 19 | set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-Xcompiler -D_GLIBCXX_USE_CXX11_ABI=0") 20 | endif (UNIX) 21 | 22 | # add debugging to CUDA NVCC flags. For NVidia's NSight tools. 23 | set(CUDA_NVCC_FLAGS_DEBUG ${CUDA_NVCC_FLAGS_DEBUG} "-G") 24 | 25 | file( GLOB cu *.cu) 26 | file( GLOB hdr *.hpp *.h ) 27 | SET (CPP_FILES sa_ext.cpp) 28 | CUDA_ADD_EXECUTABLE(SA ${CPP_FILES} ${cu} ${hdr}) 29 | target_link_libraries(SA "${TORCH_LIBRARIES}") 30 | set_property(TARGET SA PROPERTY CXX_STANDARD 11) 31 | else(CUDA_FOUND) 32 | message("CUDA is not found!") 33 | endif() 34 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/3.10.2/CMakeCCompiler.cmake: -------------------------------------------------------------------------------- 1 | set(CMAKE_C_COMPILER "/usr/bin/cc") 2 | set(CMAKE_C_COMPILER_ARG1 "") 3 | set(CMAKE_C_COMPILER_ID "GNU") 4 | set(CMAKE_C_COMPILER_VERSION "7.4.0") 5 | set(CMAKE_C_COMPILER_VERSION_INTERNAL "") 6 | set(CMAKE_C_COMPILER_WRAPPER "") 7 | set(CMAKE_C_STANDARD_COMPUTED_DEFAULT "11") 8 | set(CMAKE_C_COMPILE_FEATURES "c_std_90;c_function_prototypes;c_std_99;c_restrict;c_variadic_macros;c_std_11;c_static_assert") 9 | set(CMAKE_C90_COMPILE_FEATURES "c_std_90;c_function_prototypes") 10 | set(CMAKE_C99_COMPILE_FEATURES "c_std_99;c_restrict;c_variadic_macros") 11 | set(CMAKE_C11_COMPILE_FEATURES "c_std_11;c_static_assert") 12 | 13 | set(CMAKE_C_PLATFORM_ID "Linux") 14 | set(CMAKE_C_SIMULATE_ID "") 15 | set(CMAKE_C_SIMULATE_VERSION "") 16 | 17 | 18 | 19 | set(CMAKE_AR "/usr/bin/ar") 20 | set(CMAKE_C_COMPILER_AR "/usr/bin/gcc-ar-7") 21 | set(CMAKE_RANLIB "/usr/bin/ranlib") 22 | set(CMAKE_C_COMPILER_RANLIB "/usr/bin/gcc-ranlib-7") 23 | set(CMAKE_LINKER "/usr/bin/ld") 24 | set(CMAKE_COMPILER_IS_GNUCC 1) 25 | set(CMAKE_C_COMPILER_LOADED 1) 26 | set(CMAKE_C_COMPILER_WORKS TRUE) 27 | set(CMAKE_C_ABI_COMPILED TRUE) 28 | set(CMAKE_COMPILER_IS_MINGW ) 29 | set(CMAKE_COMPILER_IS_CYGWIN ) 30 | if(CMAKE_COMPILER_IS_CYGWIN) 31 | set(CYGWIN 1) 32 | set(UNIX 1) 33 | endif() 34 | 35 | set(CMAKE_C_COMPILER_ENV_VAR "CC") 36 | 37 | if(CMAKE_COMPILER_IS_MINGW) 38 | set(MINGW 1) 39 | endif() 40 | set(CMAKE_C_COMPILER_ID_RUN 1) 41 | set(CMAKE_C_SOURCE_FILE_EXTENSIONS c;m) 42 | set(CMAKE_C_IGNORE_EXTENSIONS h;H;o;O;obj;OBJ;def;DEF;rc;RC) 43 | set(CMAKE_C_LINKER_PREFERENCE 10) 44 | 45 | # Save compiler ABI information. 46 | set(CMAKE_C_SIZEOF_DATA_PTR "8") 47 | set(CMAKE_C_COMPILER_ABI "ELF") 48 | set(CMAKE_C_LIBRARY_ARCHITECTURE "x86_64-linux-gnu") 49 | 50 | if(CMAKE_C_SIZEOF_DATA_PTR) 51 | set(CMAKE_SIZEOF_VOID_P "${CMAKE_C_SIZEOF_DATA_PTR}") 52 | endif() 53 | 54 | if(CMAKE_C_COMPILER_ABI) 55 | set(CMAKE_INTERNAL_PLATFORM_ABI "${CMAKE_C_COMPILER_ABI}") 56 | endif() 57 | 58 | if(CMAKE_C_LIBRARY_ARCHITECTURE) 59 | set(CMAKE_LIBRARY_ARCHITECTURE "x86_64-linux-gnu") 60 | endif() 61 | 62 | set(CMAKE_C_CL_SHOWINCLUDES_PREFIX "") 63 | if(CMAKE_C_CL_SHOWINCLUDES_PREFIX) 64 | set(CMAKE_CL_SHOWINCLUDES_PREFIX "${CMAKE_C_CL_SHOWINCLUDES_PREFIX}") 65 | endif() 66 | 67 | 68 | 69 | 70 | 71 | set(CMAKE_C_IMPLICIT_LINK_LIBRARIES "gcc;gcc_s;c;gcc;gcc_s") 72 | set(CMAKE_C_IMPLICIT_LINK_DIRECTORIES "/usr/lib/gcc/x86_64-linux-gnu/7;/usr/lib/x86_64-linux-gnu;/usr/lib;/lib/x86_64-linux-gnu;/lib") 73 | set(CMAKE_C_IMPLICIT_LINK_FRAMEWORK_DIRECTORIES "") 74 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/3.10.2/CMakeCXXCompiler.cmake: -------------------------------------------------------------------------------- 1 | set(CMAKE_CXX_COMPILER "/usr/bin/c++") 2 | set(CMAKE_CXX_COMPILER_ARG1 "") 3 | set(CMAKE_CXX_COMPILER_ID "GNU") 4 | set(CMAKE_CXX_COMPILER_VERSION "7.4.0") 5 | set(CMAKE_CXX_COMPILER_VERSION_INTERNAL "") 6 | set(CMAKE_CXX_COMPILER_WRAPPER "") 7 | set(CMAKE_CXX_STANDARD_COMPUTED_DEFAULT "14") 8 | set(CMAKE_CXX_COMPILE_FEATURES "cxx_std_98;cxx_template_template_parameters;cxx_std_11;cxx_alias_templates;cxx_alignas;cxx_alignof;cxx_attributes;cxx_auto_type;cxx_constexpr;cxx_decltype;cxx_decltype_incomplete_return_types;cxx_default_function_template_args;cxx_defaulted_functions;cxx_defaulted_move_initializers;cxx_delegating_constructors;cxx_deleted_functions;cxx_enum_forward_declarations;cxx_explicit_conversions;cxx_extended_friend_declarations;cxx_extern_templates;cxx_final;cxx_func_identifier;cxx_generalized_initializers;cxx_inheriting_constructors;cxx_inline_namespaces;cxx_lambdas;cxx_local_type_template_args;cxx_long_long_type;cxx_noexcept;cxx_nonstatic_member_init;cxx_nullptr;cxx_override;cxx_range_for;cxx_raw_string_literals;cxx_reference_qualified_functions;cxx_right_angle_brackets;cxx_rvalue_references;cxx_sizeof_member;cxx_static_assert;cxx_strong_enums;cxx_thread_local;cxx_trailing_return_types;cxx_unicode_literals;cxx_uniform_initialization;cxx_unrestricted_unions;cxx_user_literals;cxx_variadic_macros;cxx_variadic_templates;cxx_std_14;cxx_aggregate_default_initializers;cxx_attribute_deprecated;cxx_binary_literals;cxx_contextual_conversions;cxx_decltype_auto;cxx_digit_separators;cxx_generic_lambdas;cxx_lambda_init_captures;cxx_relaxed_constexpr;cxx_return_type_deduction;cxx_variable_templates;cxx_std_17") 9 | set(CMAKE_CXX98_COMPILE_FEATURES "cxx_std_98;cxx_template_template_parameters") 10 | set(CMAKE_CXX11_COMPILE_FEATURES "cxx_std_11;cxx_alias_templates;cxx_alignas;cxx_alignof;cxx_attributes;cxx_auto_type;cxx_constexpr;cxx_decltype;cxx_decltype_incomplete_return_types;cxx_default_function_template_args;cxx_defaulted_functions;cxx_defaulted_move_initializers;cxx_delegating_constructors;cxx_deleted_functions;cxx_enum_forward_declarations;cxx_explicit_conversions;cxx_extended_friend_declarations;cxx_extern_templates;cxx_final;cxx_func_identifier;cxx_generalized_initializers;cxx_inheriting_constructors;cxx_inline_namespaces;cxx_lambdas;cxx_local_type_template_args;cxx_long_long_type;cxx_noexcept;cxx_nonstatic_member_init;cxx_nullptr;cxx_override;cxx_range_for;cxx_raw_string_literals;cxx_reference_qualified_functions;cxx_right_angle_brackets;cxx_rvalue_references;cxx_sizeof_member;cxx_static_assert;cxx_strong_enums;cxx_thread_local;cxx_trailing_return_types;cxx_unicode_literals;cxx_uniform_initialization;cxx_unrestricted_unions;cxx_user_literals;cxx_variadic_macros;cxx_variadic_templates") 11 | set(CMAKE_CXX14_COMPILE_FEATURES "cxx_std_14;cxx_aggregate_default_initializers;cxx_attribute_deprecated;cxx_binary_literals;cxx_contextual_conversions;cxx_decltype_auto;cxx_digit_separators;cxx_generic_lambdas;cxx_lambda_init_captures;cxx_relaxed_constexpr;cxx_return_type_deduction;cxx_variable_templates") 12 | set(CMAKE_CXX17_COMPILE_FEATURES "cxx_std_17") 13 | 14 | set(CMAKE_CXX_PLATFORM_ID "Linux") 15 | set(CMAKE_CXX_SIMULATE_ID "") 16 | set(CMAKE_CXX_SIMULATE_VERSION "") 17 | 18 | 19 | 20 | set(CMAKE_AR "/usr/bin/ar") 21 | set(CMAKE_CXX_COMPILER_AR "/usr/bin/gcc-ar-7") 22 | set(CMAKE_RANLIB "/usr/bin/ranlib") 23 | set(CMAKE_CXX_COMPILER_RANLIB "/usr/bin/gcc-ranlib-7") 24 | set(CMAKE_LINKER "/usr/bin/ld") 25 | set(CMAKE_COMPILER_IS_GNUCXX 1) 26 | set(CMAKE_CXX_COMPILER_LOADED 1) 27 | set(CMAKE_CXX_COMPILER_WORKS TRUE) 28 | set(CMAKE_CXX_ABI_COMPILED TRUE) 29 | set(CMAKE_COMPILER_IS_MINGW ) 30 | set(CMAKE_COMPILER_IS_CYGWIN ) 31 | if(CMAKE_COMPILER_IS_CYGWIN) 32 | set(CYGWIN 1) 33 | set(UNIX 1) 34 | endif() 35 | 36 | set(CMAKE_CXX_COMPILER_ENV_VAR "CXX") 37 | 38 | if(CMAKE_COMPILER_IS_MINGW) 39 | set(MINGW 1) 40 | endif() 41 | set(CMAKE_CXX_COMPILER_ID_RUN 1) 42 | set(CMAKE_CXX_IGNORE_EXTENSIONS inl;h;hpp;HPP;H;o;O;obj;OBJ;def;DEF;rc;RC) 43 | set(CMAKE_CXX_SOURCE_FILE_EXTENSIONS C;M;c++;cc;cpp;cxx;mm;CPP) 44 | set(CMAKE_CXX_LINKER_PREFERENCE 30) 45 | set(CMAKE_CXX_LINKER_PREFERENCE_PROPAGATES 1) 46 | 47 | # Save compiler ABI information. 48 | set(CMAKE_CXX_SIZEOF_DATA_PTR "8") 49 | set(CMAKE_CXX_COMPILER_ABI "ELF") 50 | set(CMAKE_CXX_LIBRARY_ARCHITECTURE "x86_64-linux-gnu") 51 | 52 | if(CMAKE_CXX_SIZEOF_DATA_PTR) 53 | set(CMAKE_SIZEOF_VOID_P "${CMAKE_CXX_SIZEOF_DATA_PTR}") 54 | endif() 55 | 56 | if(CMAKE_CXX_COMPILER_ABI) 57 | set(CMAKE_INTERNAL_PLATFORM_ABI "${CMAKE_CXX_COMPILER_ABI}") 58 | endif() 59 | 60 | if(CMAKE_CXX_LIBRARY_ARCHITECTURE) 61 | set(CMAKE_LIBRARY_ARCHITECTURE "x86_64-linux-gnu") 62 | endif() 63 | 64 | set(CMAKE_CXX_CL_SHOWINCLUDES_PREFIX "") 65 | if(CMAKE_CXX_CL_SHOWINCLUDES_PREFIX) 66 | set(CMAKE_CL_SHOWINCLUDES_PREFIX "${CMAKE_CXX_CL_SHOWINCLUDES_PREFIX}") 67 | endif() 68 | 69 | 70 | 71 | 72 | 73 | set(CMAKE_CXX_IMPLICIT_LINK_LIBRARIES "stdc++;m;gcc_s;gcc;c;gcc_s;gcc") 74 | set(CMAKE_CXX_IMPLICIT_LINK_DIRECTORIES "/usr/lib/gcc/x86_64-linux-gnu/7;/usr/lib/x86_64-linux-gnu;/usr/lib;/lib/x86_64-linux-gnu;/lib") 75 | set(CMAKE_CXX_IMPLICIT_LINK_FRAMEWORK_DIRECTORIES "") 76 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/3.10.2/CMakeDetermineCompilerABI_C.bin: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/3.10.2/CMakeDetermineCompilerABI_C.bin -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/3.10.2/CMakeDetermineCompilerABI_CXX.bin: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/3.10.2/CMakeDetermineCompilerABI_CXX.bin -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/3.10.2/CMakeSystem.cmake: -------------------------------------------------------------------------------- 1 | set(CMAKE_HOST_SYSTEM "Linux-5.0.0-23-generic") 2 | set(CMAKE_HOST_SYSTEM_NAME "Linux") 3 | set(CMAKE_HOST_SYSTEM_VERSION "5.0.0-23-generic") 4 | set(CMAKE_HOST_SYSTEM_PROCESSOR "x86_64") 5 | 6 | 7 | 8 | set(CMAKE_SYSTEM "Linux-5.0.0-23-generic") 9 | set(CMAKE_SYSTEM_NAME "Linux") 10 | set(CMAKE_SYSTEM_VERSION "5.0.0-23-generic") 11 | set(CMAKE_SYSTEM_PROCESSOR "x86_64") 12 | 13 | set(CMAKE_CROSSCOMPILING "FALSE") 14 | 15 | set(CMAKE_SYSTEM_LOADED 1) 16 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/3.10.2/CompilerIdC/a.out: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/3.10.2/CompilerIdC/a.out -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/3.10.2/CompilerIdCXX/a.out: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/3.10.2/CompilerIdCXX/a.out -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/CMakeDirectoryInformation.cmake: -------------------------------------------------------------------------------- 1 | # CMAKE generated file: DO NOT EDIT! 2 | # Generated by "Unix Makefiles" Generator, CMake Version 3.10 3 | 4 | # Relative path conversion top directories. 5 | set(CMAKE_RELATIVE_PATH_TOP_SOURCE "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module") 6 | set(CMAKE_RELATIVE_PATH_TOP_BINARY "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild") 7 | 8 | # Force unix paths in dependencies. 9 | set(CMAKE_FORCE_UNIX_PATHS 1) 10 | 11 | 12 | # The C and CXX include file regular expressions for this directory. 13 | set(CMAKE_C_INCLUDE_REGEX_SCAN "^.*$") 14 | set(CMAKE_C_INCLUDE_REGEX_COMPLAIN "^$") 15 | set(CMAKE_CXX_INCLUDE_REGEX_SCAN ${CMAKE_C_INCLUDE_REGEX_SCAN}) 16 | set(CMAKE_CXX_INCLUDE_REGEX_COMPLAIN ${CMAKE_C_INCLUDE_REGEX_COMPLAIN}) 17 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/CMakeError.log: -------------------------------------------------------------------------------- 1 | Determining if the pthread_create exist failed with the following output: 2 | Change Dir: /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/CMakeTmp 3 | 4 | Run Build Command:"/usr/bin/make" "cmTC_38ddb/fast" 5 | /usr/bin/make -f CMakeFiles/cmTC_38ddb.dir/build.make CMakeFiles/cmTC_38ddb.dir/build 6 | make[1]: 进入目录“/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/CMakeTmp” 7 | Building C object CMakeFiles/cmTC_38ddb.dir/CheckSymbolExists.c.o 8 | /usr/bin/cc -o CMakeFiles/cmTC_38ddb.dir/CheckSymbolExists.c.o -c /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/CMakeTmp/CheckSymbolExists.c 9 | Linking C executable cmTC_38ddb 10 | /usr/bin/cmake -E cmake_link_script CMakeFiles/cmTC_38ddb.dir/link.txt --verbose=1 11 | /usr/bin/cc -rdynamic CMakeFiles/cmTC_38ddb.dir/CheckSymbolExists.c.o -o cmTC_38ddb 12 | CMakeFiles/cmTC_38ddb.dir/CheckSymbolExists.c.o：在函数‘main’中： 13 | CheckSymbolExists.c:(.text+0x1b)：对‘pthread_create’未定义的引用 14 | collect2: error: ld returned 1 exit status 15 | CMakeFiles/cmTC_38ddb.dir/build.make:97: recipe for target 'cmTC_38ddb' failed 16 | make[1]: *** [cmTC_38ddb] Error 1 17 | make[1]: 离开目录“/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/CMakeTmp” 18 | Makefile:126: recipe for target 'cmTC_38ddb/fast' failed 19 | make: *** [cmTC_38ddb/fast] Error 2 20 | 21 | File /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/CMakeTmp/CheckSymbolExists.c: 22 | /* */ 23 | #include 24 | 25 | int main(int argc, char** argv) 26 | { 27 | (void)argv; 28 | #ifndef pthread_create 29 | return ((int*)(&pthread_create))[argc]; 30 | #else 31 | (void)argc; 32 | return 0; 33 | #endif 34 | } 35 | 36 | Determining if the function pthread_create exists in the pthreads failed with the following output: 37 | Change Dir: /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/CMakeTmp 38 | 39 | Run Build Command:"/usr/bin/make" "cmTC_8df24/fast" 40 | /usr/bin/make -f CMakeFiles/cmTC_8df24.dir/build.make CMakeFiles/cmTC_8df24.dir/build 41 | make[1]: 进入目录“/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/CMakeTmp” 42 | Building C object CMakeFiles/cmTC_8df24.dir/CheckFunctionExists.c.o 43 | /usr/bin/cc -DCHECK_FUNCTION_EXISTS=pthread_create -o CMakeFiles/cmTC_8df24.dir/CheckFunctionExists.c.o -c /usr/share/cmake-3.10/Modules/CheckFunctionExists.c 44 | Linking C executable cmTC_8df24 45 | /usr/bin/cmake -E cmake_link_script CMakeFiles/cmTC_8df24.dir/link.txt --verbose=1 46 | /usr/bin/cc -DCHECK_FUNCTION_EXISTS=pthread_create -rdynamic CMakeFiles/cmTC_8df24.dir/CheckFunctionExists.c.o -o cmTC_8df24 -lpthreads 47 | /usr/bin/ld: 找不到 -lpthreads 48 | collect2: error: ld returned 1 exit status 49 | CMakeFiles/cmTC_8df24.dir/build.make:97: recipe for target 'cmTC_8df24' failed 50 | make[1]: *** [cmTC_8df24] Error 1 51 | make[1]: 离开目录“/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/CMakeTmp” 52 | Makefile:126: recipe for target 'cmTC_8df24/fast' failed 53 | make: *** [cmTC_8df24/fast] Error 2 54 | 55 | 56 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/CMakeRuleHashes.txt: -------------------------------------------------------------------------------- 1 | # Hashes of file build rules. 2 | 51619ed7518b028b471246c895ad49fc CMakeFiles/SA.dir/SA_generated_sa.cu.o 3 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/Makefile.cmake: -------------------------------------------------------------------------------- 1 | # CMAKE generated file: DO NOT EDIT! 2 | # Generated by "Unix Makefiles" Generator, CMake Version 3.10 3 | 4 | # The generator used is: 5 | set(CMAKE_DEPENDS_GENERATOR "Unix Makefiles") 6 | 7 | # The top level Makefile was generated from the following files: 8 | set(CMAKE_MAKEFILE_DEPENDS 9 | "CMakeCache.txt" 10 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Caffe2Config.cmake" 11 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Caffe2ConfigVersion.cmake" 12 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Caffe2Targets-release.cmake" 13 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Caffe2Targets.cmake" 14 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Modules_CUDA_fix/FindCUDA.cmake" 15 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Modules_CUDA_fix/upstream/CMakeInitializeConfigs.cmake" 16 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Modules_CUDA_fix/upstream/FindCUDA.cmake" 17 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Modules_CUDA_fix/upstream/FindCUDA/run_nvcc.cmake" 18 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Modules_CUDA_fix/upstream/FindCUDA/select_compute_arch.cmake" 19 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Modules_CUDA_fix/upstream/FindPackageHandleStandardArgs.cmake" 20 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Modules_CUDA_fix/upstream/FindPackageMessage.cmake" 21 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/public/cuda.cmake" 22 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/public/mkl.cmake" 23 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/public/mkldnn.cmake" 24 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/public/threads.cmake" 25 | "/home/guyuchao/software/libtorch/share/cmake/Caffe2/public/utils.cmake" 26 | "/home/guyuchao/software/libtorch/share/cmake/Torch/TorchConfig.cmake" 27 | "/home/guyuchao/software/libtorch/share/cmake/Torch/TorchConfigVersion.cmake" 28 | "../CMakeLists.txt" 29 | "CMakeFiles/3.10.2/CMakeCCompiler.cmake" 30 | "CMakeFiles/3.10.2/CMakeCXXCompiler.cmake" 31 | "CMakeFiles/3.10.2/CMakeSystem.cmake" 32 | "CMakeFiles/SA.dir/SA_generated_sa.cu.o.cmake.pre-gen" 33 | "CMakeFiles/SA.dir/SA_generated_sa.cu.o.depend" 34 | "detect_cuda_compute_capabilities.cpp" 35 | "detect_cuda_version.cc" 36 | "/usr/share/cmake-3.10/Modules/CMakeCInformation.cmake" 37 | "/usr/share/cmake-3.10/Modules/CMakeCXXInformation.cmake" 38 | "/usr/share/cmake-3.10/Modules/CMakeCommonLanguageInclude.cmake" 39 | "/usr/share/cmake-3.10/Modules/CMakeGenericSystem.cmake" 40 | "/usr/share/cmake-3.10/Modules/CMakeLanguageInformation.cmake" 41 | "/usr/share/cmake-3.10/Modules/CMakeSystemSpecificInformation.cmake" 42 | "/usr/share/cmake-3.10/Modules/CMakeSystemSpecificInitialize.cmake" 43 | "/usr/share/cmake-3.10/Modules/CheckIncludeFile.cmake" 44 | "/usr/share/cmake-3.10/Modules/CheckLibraryExists.cmake" 45 | "/usr/share/cmake-3.10/Modules/CheckSymbolExists.cmake" 46 | "/usr/share/cmake-3.10/Modules/Compiler/CMakeCommonCompilerMacros.cmake" 47 | "/usr/share/cmake-3.10/Modules/Compiler/GNU-C.cmake" 48 | "/usr/share/cmake-3.10/Modules/Compiler/GNU-CXX.cmake" 49 | "/usr/share/cmake-3.10/Modules/Compiler/GNU.cmake" 50 | "/usr/share/cmake-3.10/Modules/FindPackageHandleStandardArgs.cmake" 51 | "/usr/share/cmake-3.10/Modules/FindPackageMessage.cmake" 52 | "/usr/share/cmake-3.10/Modules/FindThreads.cmake" 53 | "/usr/share/cmake-3.10/Modules/Platform/Linux-GNU-C.cmake" 54 | "/usr/share/cmake-3.10/Modules/Platform/Linux-GNU-CXX.cmake" 55 | "/usr/share/cmake-3.10/Modules/Platform/Linux-GNU.cmake" 56 | "/usr/share/cmake-3.10/Modules/Platform/Linux.cmake" 57 | "/usr/share/cmake-3.10/Modules/Platform/UnixPaths.cmake" 58 | ) 59 | 60 | # The corresponding makefile is: 61 | set(CMAKE_MAKEFILE_OUTPUTS 62 | "Makefile" 63 | "CMakeFiles/cmake.check_cache" 64 | ) 65 | 66 | # Byproducts of CMake generate step: 67 | set(CMAKE_MAKEFILE_PRODUCTS 68 | "CMakeFiles/SA.dir/SA_generated_sa.cu.o.cmake.pre-gen" 69 | "CMakeFiles/SA.dir/SA_generated_sa.cu.o.cmake" 70 | "CMakeFiles/CMakeDirectoryInformation.cmake" 71 | ) 72 | 73 | # Dependency information for all targets: 74 | set(CMAKE_DEPEND_INFO_FILES 75 | "CMakeFiles/SA.dir/DependInfo.cmake" 76 | ) 77 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/Makefile2: -------------------------------------------------------------------------------- 1 | # CMAKE generated file: DO NOT EDIT! 2 | # Generated by "Unix Makefiles" Generator, CMake Version 3.10 3 | 4 | # Default target executed when no arguments are given to make. 5 | default_target: all 6 | 7 | .PHONY : default_target 8 | 9 | # The main recursive all target 10 | all: 11 | 12 | .PHONY : all 13 | 14 | # The main recursive preinstall target 15 | preinstall: 16 | 17 | .PHONY : preinstall 18 | 19 | #============================================================================= 20 | # Special targets provided by cmake. 21 | 22 | # Disable implicit rules so canonical targets will work. 23 | .SUFFIXES: 24 | 25 | 26 | # Remove some rules from gmake that .SUFFIXES does not remove. 27 | SUFFIXES = 28 | 29 | .SUFFIXES: .hpux_make_needs_suffix_list 30 | 31 | 32 | # Suppress display of executed commands. 33 | $(VERBOSE).SILENT: 34 | 35 | 36 | # A target that is always out of date. 37 | cmake_force: 38 | 39 | .PHONY : cmake_force 40 | 41 | #============================================================================= 42 | # Set environment variables for the build. 43 | 44 | # The shell in which to execute make rules. 45 | SHELL = /bin/sh 46 | 47 | # The CMake executable. 48 | CMAKE_COMMAND = /usr/bin/cmake 49 | 50 | # The command to remove a file. 51 | RM = /usr/bin/cmake -E remove -f 52 | 53 | # Escaping for special characters. 54 | EQUALS = = 55 | 56 | # The top-level source directory on which CMake was run. 57 | CMAKE_SOURCE_DIR = /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module 58 | 59 | # The top-level build directory on which CMake was run. 60 | CMAKE_BINARY_DIR = /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild 61 | 62 | #============================================================================= 63 | # Target rules for target CMakeFiles/SA.dir 64 | 65 | # All Build rule for target. 66 | CMakeFiles/SA.dir/all: 67 | $(MAKE) -f CMakeFiles/SA.dir/build.make CMakeFiles/SA.dir/depend 68 | $(MAKE) -f CMakeFiles/SA.dir/build.make CMakeFiles/SA.dir/build 69 | @$(CMAKE_COMMAND) -E cmake_echo_color --switch=$(COLOR) --progress-dir=/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles --progress-num=1,2,3 "Built target SA" 70 | .PHONY : CMakeFiles/SA.dir/all 71 | 72 | # Include target in all. 73 | all: CMakeFiles/SA.dir/all 74 | 75 | .PHONY : all 76 | 77 | # Build rule for subdir invocation for target. 78 | CMakeFiles/SA.dir/rule: cmake_check_build_system 79 | $(CMAKE_COMMAND) -E cmake_progress_start /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles 3 80 | $(MAKE) -f CMakeFiles/Makefile2 CMakeFiles/SA.dir/all 81 | $(CMAKE_COMMAND) -E cmake_progress_start /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles 0 82 | .PHONY : CMakeFiles/SA.dir/rule 83 | 84 | # Convenience name for target. 85 | SA: CMakeFiles/SA.dir/rule 86 | 87 | .PHONY : SA 88 | 89 | # clean rule for target. 90 | CMakeFiles/SA.dir/clean: 91 | $(MAKE) -f CMakeFiles/SA.dir/build.make CMakeFiles/SA.dir/clean 92 | .PHONY : CMakeFiles/SA.dir/clean 93 | 94 | # clean rule for target. 95 | clean: CMakeFiles/SA.dir/clean 96 | 97 | .PHONY : clean 98 | 99 | #============================================================================= 100 | # Special targets to cleanup operation of make. 101 | 102 | # Special rule to run CMake to check the build system integrity. 103 | # No rule that depends on this can have commands that come from listfiles 104 | # because they might be regenerated. 105 | cmake_check_build_system: 106 | $(CMAKE_COMMAND) -H$(CMAKE_SOURCE_DIR) -B$(CMAKE_BINARY_DIR) --check-build-system CMakeFiles/Makefile.cmake 0 107 | .PHONY : cmake_check_build_system 108 | 109 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/SA.dir/DependInfo.cmake: -------------------------------------------------------------------------------- 1 | # The set of languages for which implicit dependencies are needed: 2 | set(CMAKE_DEPENDS_LANGUAGES 3 | "CXX" 4 | ) 5 | # The set of files for implicit dependencies of each language: 6 | set(CMAKE_DEPENDS_CHECK_CXX 7 | "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/sa_ext.cpp" "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/SA.dir/sa_ext.cpp.o" 8 | ) 9 | set(CMAKE_CXX_COMPILER_ID "GNU") 10 | 11 | # The include file search paths: 12 | set(CMAKE_CXX_TARGET_INCLUDE_PATH 13 | "/usr/local/cuda-10.0/include" 14 | "/home/guyuchao/software/libtorch/include" 15 | "/home/guyuchao/software/libtorch/include/torch/csrc/api/include" 16 | ) 17 | 18 | # Targets to which this target links. 19 | set(CMAKE_TARGET_LINKED_INFO_FILES 20 | ) 21 | 22 | # Fortran module output directory. 23 | set(CMAKE_Fortran_TARGET_MODULE_DIR "") 24 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/SA.dir/SA_generated_sa.cu.o: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/SA.dir/SA_generated_sa.cu.o -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/SA.dir/SA_generated_sa.cu.o.cmake: -------------------------------------------------------------------------------- 1 | # James Bigler, NVIDIA Corp (nvidia.com - jbigler) 2 | # 3 | # Copyright (c) 2008 - 2009 NVIDIA Corporation. All rights reserved. 4 | # 5 | # This code is licensed under the MIT License. See the FindCUDA.cmake script 6 | # for the text of the license. 7 | 8 | # The MIT License 9 | # 10 | # License for the specific language governing rights and limitations under 11 | # Permission is hereby granted, free of charge, to any person obtaining a 12 | # copy of this software and associated documentation files (the "Software"), 13 | # to deal in the Software without restriction, including without limitation 14 | # the rights to use, copy, modify, merge, publish, distribute, sublicense, 15 | # and/or sell copies of the Software, and to permit persons to whom the 16 | # Software is furnished to do so, subject to the following conditions: 17 | # 18 | # The above copyright notice and this permission notice shall be included 19 | # in all copies or substantial portions of the Software. 20 | # 21 | # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS 22 | # OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 23 | # FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 24 | # THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 25 | # LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 26 | # FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER 27 | # DEALINGS IN THE SOFTWARE. 28 | 29 | 30 | ########################################################################## 31 | # This file runs the nvcc commands to produce the desired output file along with 32 | # the dependency file needed by CMake to compute dependencies. In addition the 33 | # file checks the output of each command and if the command fails it deletes the 34 | # output files. 35 | 36 | # Input variables 37 | # 38 | # verbose:BOOL=<> OFF: Be as quiet as possible (default) 39 | # ON : Describe each step 40 | # 41 | # build_configuration:STRING=<> Typically one of Debug, MinSizeRel, Release, or 42 | # RelWithDebInfo, but it should match one of the 43 | # entries in CUDA_HOST_FLAGS. This is the build 44 | # configuration used when compiling the code. If 45 | # blank or unspecified Debug is assumed as this is 46 | # what CMake does. 47 | # 48 | # generated_file:STRING=<> File to generate. This argument must be passed in. 49 | # 50 | # generated_cubin_file:STRING=<> File to generate. This argument must be passed 51 | # in if build_cubin is true. 52 | 53 | cmake_policy(PUSH) 54 | cmake_policy(SET CMP0007 NEW) 55 | if(NOT generated_file) 56 | message(FATAL_ERROR "You must specify generated_file on the command line") 57 | endif() 58 | 59 | # Set these up as variables to make reading the generated file easier 60 | set(CMAKE_COMMAND "/usr/bin/cmake") # path 61 | set(source_file "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/sa.cu") # path 62 | set(NVCC_generated_dependency_file "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/SA.dir//SA_generated_sa.cu.o.NVCC-depend") # path 63 | set(cmake_dependency_file "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/SA.dir//SA_generated_sa.cu.o.depend") # path 64 | set(CUDA_make2cmake "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Modules_CUDA_fix/upstream/FindCUDA/make2cmake.cmake") # path 65 | set(CUDA_parse_cubin "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Modules_CUDA_fix/upstream/FindCUDA/parse_cubin.cmake") # path 66 | set(build_cubin OFF) # bool 67 | set(CUDA_HOST_COMPILER "/usr/bin/cc") # path 68 | # We won't actually use these variables for now, but we need to set this, in 69 | # order to force this file to be run again if it changes. 70 | set(generated_file_path "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/SA.dir//.") # path 71 | set(generated_file_internal "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/SA.dir//./SA_generated_sa.cu.o") # path 72 | set(generated_cubin_file_internal "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/SA.dir//./SA_generated_sa.cu.o.cubin.txt") # path 73 | 74 | set(CUDA_NVCC_EXECUTABLE "/usr/local/cuda-10.0/bin/nvcc") # path 75 | set(CUDA_NVCC_FLAGS -DONNX_NAMESPACE=onnx_c2;-gencode;arch=compute_61,code=sm_61;-Xcudafe;--diag_suppress=cc_clobber_ignored;-Xcudafe;--diag_suppress=integer_sign_change;-Xcudafe;--diag_suppress=useless_using_declaration;-Xcudafe;--diag_suppress=set_but_not_used;-std=c++11;-Xcompiler;-fPIC;--expt-relaxed-constexpr;--expt-extended-lambda;-Xcompiler -D_GLIBCXX_USE_CXX11_ABI=0 ;; ) # list 76 | # Build specific configuration flags 77 | set(CUDA_NVCC_FLAGS_DEBUG -G ; ) 78 | set(CUDA_NVCC_FLAGS_MINSIZEREL ; ) 79 | set(CUDA_NVCC_FLAGS_RELEASE ; ) 80 | set(CUDA_NVCC_FLAGS_RELWITHDEBINFO ; ) 81 | set(nvcc_flags -m64) # list 82 | set(CUDA_NVCC_INCLUDE_DIRS "/usr/local/cuda-10.0/include;/usr/local/cuda-10.0/include;/home/guyuchao/software/libtorch/include;/home/guyuchao/software/libtorch/include/torch/csrc/api/include;/home/guyuchao/software/libtorch/include;/home/guyuchao/software/libtorch/include;/home/guyuchao/software/libtorch/include;/usr/local/cuda-10.0/include;/home/guyuchao/software/libtorch/include;/usr/local/cuda-10.0/include;/usr/local/cuda-10.0/include;/usr/include;/usr/local/cuda-10.0/include") # list (needs to be in quotes to handle spaces properly). 83 | set(CUDA_NVCC_COMPILE_DEFINITIONS [==[]==]) # list (needs to be in lua quotes see #16510 ). 84 | set(format_flag "-c") # string 85 | set(cuda_language_flag ) # list 86 | 87 | # Clean up list of include directories and add -I flags 88 | list(REMOVE_DUPLICATES CUDA_NVCC_INCLUDE_DIRS) 89 | set(CUDA_NVCC_INCLUDE_ARGS) 90 | foreach(dir ${CUDA_NVCC_INCLUDE_DIRS}) 91 | # Extra quotes are added around each flag to help nvcc parse out flags with spaces. 92 | list(APPEND CUDA_NVCC_INCLUDE_ARGS "-I${dir}") 93 | endforeach() 94 | 95 | # Clean up list of compile definitions, add -D flags, and append to nvcc_flags 96 | list(REMOVE_DUPLICATES CUDA_NVCC_COMPILE_DEFINITIONS) 97 | foreach(def ${CUDA_NVCC_COMPILE_DEFINITIONS}) 98 | list(APPEND nvcc_flags "-D${def}") 99 | endforeach() 100 | 101 | if(build_cubin AND NOT generated_cubin_file) 102 | message(FATAL_ERROR "You must specify generated_cubin_file on the command line") 103 | endif() 104 | 105 | # This is the list of host compilation flags. It C or CXX should already have 106 | # been chosen by FindCUDA.cmake. 107 | set(CMAKE_HOST_FLAGS -fPIC ) 108 | set(CMAKE_HOST_FLAGS_DEBUG -g) 109 | set(CMAKE_HOST_FLAGS_MINSIZEREL -Os -DNDEBUG) 110 | set(CMAKE_HOST_FLAGS_RELEASE -O3 -DNDEBUG) 111 | set(CMAKE_HOST_FLAGS_RELWITHDEBINFO -O2 -g -DNDEBUG) 112 | 113 | # Take the compiler flags and package them up to be sent to the compiler via -Xcompiler 114 | set(nvcc_host_compiler_flags "") 115 | # If we weren't given a build_configuration, use Debug. 116 | if(NOT build_configuration) 117 | set(build_configuration Debug) 118 | endif() 119 | string(TOUPPER "${build_configuration}" build_configuration) 120 | #message("CUDA_NVCC_HOST_COMPILER_FLAGS = ${CUDA_NVCC_HOST_COMPILER_FLAGS}") 121 | foreach(flag ${CMAKE_HOST_FLAGS} ${CMAKE_HOST_FLAGS_${build_configuration}}) 122 | # Extra quotes are added around each flag to help nvcc parse out flags with spaces. 123 | string(APPEND nvcc_host_compiler_flags ",\"${flag}\"") 124 | endforeach() 125 | if (nvcc_host_compiler_flags) 126 | set(nvcc_host_compiler_flags "-Xcompiler" ${nvcc_host_compiler_flags}) 127 | endif() 128 | #message("nvcc_host_compiler_flags = \"${nvcc_host_compiler_flags}\"") 129 | # Add the build specific configuration flags 130 | list(APPEND CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS_${build_configuration}}) 131 | 132 | # Any -ccbin existing in CUDA_NVCC_FLAGS gets highest priority 133 | list( FIND CUDA_NVCC_FLAGS "-ccbin" ccbin_found0 ) 134 | list( FIND CUDA_NVCC_FLAGS "--compiler-bindir" ccbin_found1 ) 135 | if( ccbin_found0 LESS 0 AND ccbin_found1 LESS 0 AND CUDA_HOST_COMPILER ) 136 | if (CUDA_HOST_COMPILER STREQUAL "" AND DEFINED CCBIN) 137 | set(CCBIN -ccbin "${CCBIN}") 138 | else() 139 | set(CCBIN -ccbin "${CUDA_HOST_COMPILER}") 140 | endif() 141 | endif() 142 | 143 | # cuda_execute_process - Executes a command with optional command echo and status message. 144 | # 145 | # status - Status message to print if verbose is true 146 | # command - COMMAND argument from the usual execute_process argument structure 147 | # ARGN - Remaining arguments are the command with arguments 148 | # 149 | # CUDA_result - return value from running the command 150 | # 151 | # Make this a macro instead of a function, so that things like RESULT_VARIABLE 152 | # and other return variables are present after executing the process. 153 | macro(cuda_execute_process status command) 154 | set(_command ${command}) 155 | if(NOT "x${_command}" STREQUAL "xCOMMAND") 156 | message(FATAL_ERROR "Malformed call to cuda_execute_process. Missing COMMAND as second argument. (command = ${command})") 157 | endif() 158 | if(verbose) 159 | execute_process(COMMAND "${CMAKE_COMMAND}" -E echo -- ${status}) 160 | # Now we need to build up our command string. We are accounting for quotes 161 | # and spaces, anything else is left up to the user to fix if they want to 162 | # copy and paste a runnable command line. 163 | set(cuda_execute_process_string) 164 | foreach(arg ${ARGN}) 165 | # If there are quotes, excape them, so they come through. 166 | string(REPLACE "\"" "\\\"" arg ${arg}) 167 | # Args with spaces need quotes around them to get them to be parsed as a single argument. 168 | if(arg MATCHES " ") 169 | list(APPEND cuda_execute_process_string "\"${arg}\"") 170 | else() 171 | list(APPEND cuda_execute_process_string ${arg}) 172 | endif() 173 | endforeach() 174 | # Echo the command 175 | execute_process(COMMAND ${CMAKE_COMMAND} -E echo ${cuda_execute_process_string}) 176 | endif() 177 | # Run the command 178 | execute_process(COMMAND ${ARGN} RESULT_VARIABLE CUDA_result ) 179 | endmacro() 180 | 181 | # Delete the target file 182 | cuda_execute_process( 183 | "Removing ${generated_file}" 184 | COMMAND "${CMAKE_COMMAND}" -E remove "${generated_file}" 185 | ) 186 | 187 | # For CUDA 2.3 and below, -G -M doesn't work, so remove the -G flag 188 | # for dependency generation and hope for the best. 189 | set(depends_CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS}") 190 | set(CUDA_VERSION 10.0) 191 | if(CUDA_VERSION VERSION_LESS "3.0") 192 | # Note that this will remove all occurrences of -G. 193 | list(REMOVE_ITEM depends_CUDA_NVCC_FLAGS "-G") 194 | endif() 195 | 196 | # nvcc doesn't define __CUDACC__ for some reason when generating dependency files. This 197 | # can cause incorrect dependencies when #including files based on this macro which is 198 | # defined in the generating passes of nvcc invocation. We will go ahead and manually 199 | # define this for now until a future version fixes this bug. 200 | set(CUDACC_DEFINE -D__CUDACC__) 201 | 202 | # Generate the dependency file 203 | cuda_execute_process( 204 | "Generating dependency file: ${NVCC_generated_dependency_file}" 205 | COMMAND "${CUDA_NVCC_EXECUTABLE}" 206 | -M 207 | ${CUDACC_DEFINE} 208 | "${source_file}" 209 | -o "${NVCC_generated_dependency_file}" 210 | ${CCBIN} 211 | ${nvcc_flags} 212 | ${nvcc_host_compiler_flags} 213 | ${depends_CUDA_NVCC_FLAGS} 214 | -DNVCC 215 | ${CUDA_NVCC_INCLUDE_ARGS} 216 | ) 217 | 218 | if(CUDA_result) 219 | message(FATAL_ERROR "Error generating ${generated_file}") 220 | endif() 221 | 222 | # Generate the cmake readable dependency file to a temp file. Don't put the 223 | # quotes just around the filenames for the input_file and output_file variables. 224 | # CMake will pass the quotes through and not be able to find the file. 225 | cuda_execute_process( 226 | "Generating temporary cmake readable file: ${cmake_dependency_file}.tmp" 227 | COMMAND "${CMAKE_COMMAND}" 228 | -D "input_file:FILEPATH=${NVCC_generated_dependency_file}" 229 | -D "output_file:FILEPATH=${cmake_dependency_file}.tmp" 230 | -D "verbose=${verbose}" 231 | -P "${CUDA_make2cmake}" 232 | ) 233 | 234 | if(CUDA_result) 235 | message(FATAL_ERROR "Error generating ${generated_file}") 236 | endif() 237 | 238 | # Copy the file if it is different 239 | cuda_execute_process( 240 | "Copy if different ${cmake_dependency_file}.tmp to ${cmake_dependency_file}" 241 | COMMAND "${CMAKE_COMMAND}" -E copy_if_different "${cmake_dependency_file}.tmp" "${cmake_dependency_file}" 242 | ) 243 | 244 | if(CUDA_result) 245 | message(FATAL_ERROR "Error generating ${generated_file}") 246 | endif() 247 | 248 | # Delete the temporary file 249 | cuda_execute_process( 250 | "Removing ${cmake_dependency_file}.tmp and ${NVCC_generated_dependency_file}" 251 | COMMAND "${CMAKE_COMMAND}" -E remove "${cmake_dependency_file}.tmp" "${NVCC_generated_dependency_file}" 252 | ) 253 | 254 | if(CUDA_result) 255 | message(FATAL_ERROR "Error generating ${generated_file}") 256 | endif() 257 | 258 | # Generate the code 259 | cuda_execute_process( 260 | "Generating ${generated_file}" 261 | COMMAND "${CUDA_NVCC_EXECUTABLE}" 262 | "${source_file}" 263 | ${cuda_language_flag} 264 | ${format_flag} -o "${generated_file}" 265 | ${CCBIN} 266 | ${nvcc_flags} 267 | ${nvcc_host_compiler_flags} 268 | ${CUDA_NVCC_FLAGS} 269 | -DNVCC 270 | ${CUDA_NVCC_INCLUDE_ARGS} 271 | ) 272 | 273 | if(CUDA_result) 274 | # Since nvcc can sometimes leave half done files make sure that we delete the output file. 275 | cuda_execute_process( 276 | "Removing ${generated_file}" 277 | COMMAND "${CMAKE_COMMAND}" -E remove "${generated_file}" 278 | ) 279 | message(FATAL_ERROR "Error generating file ${generated_file}") 280 | else() 281 | if(verbose) 282 | message("Generated ${generated_file} successfully.") 283 | endif() 284 | endif() 285 | 286 | # Cubin resource report commands. 287 | if( build_cubin ) 288 | # Run with -cubin to produce resource usage report. 289 | cuda_execute_process( 290 | "Generating ${generated_cubin_file}" 291 | COMMAND "${CUDA_NVCC_EXECUTABLE}" 292 | "${source_file}" 293 | ${CUDA_NVCC_FLAGS} 294 | ${nvcc_flags} 295 | ${CCBIN} 296 | ${nvcc_host_compiler_flags} 297 | -DNVCC 298 | -cubin 299 | -o "${generated_cubin_file}" 300 | ${CUDA_NVCC_INCLUDE_ARGS} 301 | ) 302 | 303 | # Execute the parser script. 304 | cuda_execute_process( 305 | "Executing the parser script" 306 | COMMAND "${CMAKE_COMMAND}" 307 | -D "input_file:STRING=${generated_cubin_file}" 308 | -P "${CUDA_parse_cubin}" 309 | ) 310 | 311 | endif() 312 | 313 | cmake_policy(POP) 314 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/SA.dir/SA_generated_sa.cu.o.cmake.pre-gen: -------------------------------------------------------------------------------- 1 | # James Bigler, NVIDIA Corp (nvidia.com - jbigler) 2 | # 3 | # Copyright (c) 2008 - 2009 NVIDIA Corporation. All rights reserved. 4 | # 5 | # This code is licensed under the MIT License. See the FindCUDA.cmake script 6 | # for the text of the license. 7 | 8 | # The MIT License 9 | # 10 | # License for the specific language governing rights and limitations under 11 | # Permission is hereby granted, free of charge, to any person obtaining a 12 | # copy of this software and associated documentation files (the "Software"), 13 | # to deal in the Software without restriction, including without limitation 14 | # the rights to use, copy, modify, merge, publish, distribute, sublicense, 15 | # and/or sell copies of the Software, and to permit persons to whom the 16 | # Software is furnished to do so, subject to the following conditions: 17 | # 18 | # The above copyright notice and this permission notice shall be included 19 | # in all copies or substantial portions of the Software. 20 | # 21 | # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS 22 | # OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 23 | # FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 24 | # THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 25 | # LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 26 | # FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER 27 | # DEALINGS IN THE SOFTWARE. 28 | 29 | 30 | ########################################################################## 31 | # This file runs the nvcc commands to produce the desired output file along with 32 | # the dependency file needed by CMake to compute dependencies. In addition the 33 | # file checks the output of each command and if the command fails it deletes the 34 | # output files. 35 | 36 | # Input variables 37 | # 38 | # verbose:BOOL=<> OFF: Be as quiet as possible (default) 39 | # ON : Describe each step 40 | # 41 | # build_configuration:STRING=<> Typically one of Debug, MinSizeRel, Release, or 42 | # RelWithDebInfo, but it should match one of the 43 | # entries in CUDA_HOST_FLAGS. This is the build 44 | # configuration used when compiling the code. If 45 | # blank or unspecified Debug is assumed as this is 46 | # what CMake does. 47 | # 48 | # generated_file:STRING=<> File to generate. This argument must be passed in. 49 | # 50 | # generated_cubin_file:STRING=<> File to generate. This argument must be passed 51 | # in if build_cubin is true. 52 | 53 | cmake_policy(PUSH) 54 | cmake_policy(SET CMP0007 NEW) 55 | if(NOT generated_file) 56 | message(FATAL_ERROR "You must specify generated_file on the command line") 57 | endif() 58 | 59 | # Set these up as variables to make reading the generated file easier 60 | set(CMAKE_COMMAND "/usr/bin/cmake") # path 61 | set(source_file "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/sa.cu") # path 62 | set(NVCC_generated_dependency_file "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/SA.dir//SA_generated_sa.cu.o.NVCC-depend") # path 63 | set(cmake_dependency_file "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/SA.dir//SA_generated_sa.cu.o.depend") # path 64 | set(CUDA_make2cmake "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Modules_CUDA_fix/upstream/FindCUDA/make2cmake.cmake") # path 65 | set(CUDA_parse_cubin "/home/guyuchao/software/libtorch/share/cmake/Caffe2/Modules_CUDA_fix/upstream/FindCUDA/parse_cubin.cmake") # path 66 | set(build_cubin OFF) # bool 67 | set(CUDA_HOST_COMPILER "/usr/bin/cc") # path 68 | # We won't actually use these variables for now, but we need to set this, in 69 | # order to force this file to be run again if it changes. 70 | set(generated_file_path "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/SA.dir//.") # path 71 | set(generated_file_internal "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/SA.dir//./SA_generated_sa.cu.o") # path 72 | set(generated_cubin_file_internal "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/SA.dir//./SA_generated_sa.cu.o.cubin.txt") # path 73 | 74 | set(CUDA_NVCC_EXECUTABLE "/usr/local/cuda-10.0/bin/nvcc") # path 75 | set(CUDA_NVCC_FLAGS -DONNX_NAMESPACE=onnx_c2;-gencode;arch=compute_61,code=sm_61;-Xcudafe;--diag_suppress=cc_clobber_ignored;-Xcudafe;--diag_suppress=integer_sign_change;-Xcudafe;--diag_suppress=useless_using_declaration;-Xcudafe;--diag_suppress=set_but_not_used;-std=c++11;-Xcompiler;-fPIC;--expt-relaxed-constexpr;--expt-extended-lambda;-Xcompiler -D_GLIBCXX_USE_CXX11_ABI=0 ;; ) # list 76 | # Build specific configuration flags 77 | set(CUDA_NVCC_FLAGS_DEBUG -G ; ) 78 | set(CUDA_NVCC_FLAGS_MINSIZEREL ; ) 79 | set(CUDA_NVCC_FLAGS_RELEASE ; ) 80 | set(CUDA_NVCC_FLAGS_RELWITHDEBINFO ; ) 81 | set(nvcc_flags -m64) # list 82 | set(CUDA_NVCC_INCLUDE_DIRS "/usr/local/cuda-10.0/include;$") # list (needs to be in quotes to handle spaces properly). 83 | set(CUDA_NVCC_COMPILE_DEFINITIONS [==[$]==]) # list (needs to be in lua quotes see #16510 ). 84 | set(format_flag "-c") # string 85 | set(cuda_language_flag ) # list 86 | 87 | # Clean up list of include directories and add -I flags 88 | list(REMOVE_DUPLICATES CUDA_NVCC_INCLUDE_DIRS) 89 | set(CUDA_NVCC_INCLUDE_ARGS) 90 | foreach(dir ${CUDA_NVCC_INCLUDE_DIRS}) 91 | # Extra quotes are added around each flag to help nvcc parse out flags with spaces. 92 | list(APPEND CUDA_NVCC_INCLUDE_ARGS "-I${dir}") 93 | endforeach() 94 | 95 | # Clean up list of compile definitions, add -D flags, and append to nvcc_flags 96 | list(REMOVE_DUPLICATES CUDA_NVCC_COMPILE_DEFINITIONS) 97 | foreach(def ${CUDA_NVCC_COMPILE_DEFINITIONS}) 98 | list(APPEND nvcc_flags "-D${def}") 99 | endforeach() 100 | 101 | if(build_cubin AND NOT generated_cubin_file) 102 | message(FATAL_ERROR "You must specify generated_cubin_file on the command line") 103 | endif() 104 | 105 | # This is the list of host compilation flags. It C or CXX should already have 106 | # been chosen by FindCUDA.cmake. 107 | set(CMAKE_HOST_FLAGS -fPIC ) 108 | set(CMAKE_HOST_FLAGS_DEBUG -g) 109 | set(CMAKE_HOST_FLAGS_MINSIZEREL -Os -DNDEBUG) 110 | set(CMAKE_HOST_FLAGS_RELEASE -O3 -DNDEBUG) 111 | set(CMAKE_HOST_FLAGS_RELWITHDEBINFO -O2 -g -DNDEBUG) 112 | 113 | # Take the compiler flags and package them up to be sent to the compiler via -Xcompiler 114 | set(nvcc_host_compiler_flags "") 115 | # If we weren't given a build_configuration, use Debug. 116 | if(NOT build_configuration) 117 | set(build_configuration Debug) 118 | endif() 119 | string(TOUPPER "${build_configuration}" build_configuration) 120 | #message("CUDA_NVCC_HOST_COMPILER_FLAGS = ${CUDA_NVCC_HOST_COMPILER_FLAGS}") 121 | foreach(flag ${CMAKE_HOST_FLAGS} ${CMAKE_HOST_FLAGS_${build_configuration}}) 122 | # Extra quotes are added around each flag to help nvcc parse out flags with spaces. 123 | string(APPEND nvcc_host_compiler_flags ",\"${flag}\"") 124 | endforeach() 125 | if (nvcc_host_compiler_flags) 126 | set(nvcc_host_compiler_flags "-Xcompiler" ${nvcc_host_compiler_flags}) 127 | endif() 128 | #message("nvcc_host_compiler_flags = \"${nvcc_host_compiler_flags}\"") 129 | # Add the build specific configuration flags 130 | list(APPEND CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS_${build_configuration}}) 131 | 132 | # Any -ccbin existing in CUDA_NVCC_FLAGS gets highest priority 133 | list( FIND CUDA_NVCC_FLAGS "-ccbin" ccbin_found0 ) 134 | list( FIND CUDA_NVCC_FLAGS "--compiler-bindir" ccbin_found1 ) 135 | if( ccbin_found0 LESS 0 AND ccbin_found1 LESS 0 AND CUDA_HOST_COMPILER ) 136 | if (CUDA_HOST_COMPILER STREQUAL "" AND DEFINED CCBIN) 137 | set(CCBIN -ccbin "${CCBIN}") 138 | else() 139 | set(CCBIN -ccbin "${CUDA_HOST_COMPILER}") 140 | endif() 141 | endif() 142 | 143 | # cuda_execute_process - Executes a command with optional command echo and status message. 144 | # 145 | # status - Status message to print if verbose is true 146 | # command - COMMAND argument from the usual execute_process argument structure 147 | # ARGN - Remaining arguments are the command with arguments 148 | # 149 | # CUDA_result - return value from running the command 150 | # 151 | # Make this a macro instead of a function, so that things like RESULT_VARIABLE 152 | # and other return variables are present after executing the process. 153 | macro(cuda_execute_process status command) 154 | set(_command ${command}) 155 | if(NOT "x${_command}" STREQUAL "xCOMMAND") 156 | message(FATAL_ERROR "Malformed call to cuda_execute_process. Missing COMMAND as second argument. (command = ${command})") 157 | endif() 158 | if(verbose) 159 | execute_process(COMMAND "${CMAKE_COMMAND}" -E echo -- ${status}) 160 | # Now we need to build up our command string. We are accounting for quotes 161 | # and spaces, anything else is left up to the user to fix if they want to 162 | # copy and paste a runnable command line. 163 | set(cuda_execute_process_string) 164 | foreach(arg ${ARGN}) 165 | # If there are quotes, excape them, so they come through. 166 | string(REPLACE "\"" "\\\"" arg ${arg}) 167 | # Args with spaces need quotes around them to get them to be parsed as a single argument. 168 | if(arg MATCHES " ") 169 | list(APPEND cuda_execute_process_string "\"${arg}\"") 170 | else() 171 | list(APPEND cuda_execute_process_string ${arg}) 172 | endif() 173 | endforeach() 174 | # Echo the command 175 | execute_process(COMMAND ${CMAKE_COMMAND} -E echo ${cuda_execute_process_string}) 176 | endif() 177 | # Run the command 178 | execute_process(COMMAND ${ARGN} RESULT_VARIABLE CUDA_result ) 179 | endmacro() 180 | 181 | # Delete the target file 182 | cuda_execute_process( 183 | "Removing ${generated_file}" 184 | COMMAND "${CMAKE_COMMAND}" -E remove "${generated_file}" 185 | ) 186 | 187 | # For CUDA 2.3 and below, -G -M doesn't work, so remove the -G flag 188 | # for dependency generation and hope for the best. 189 | set(depends_CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS}") 190 | set(CUDA_VERSION 10.0) 191 | if(CUDA_VERSION VERSION_LESS "3.0") 192 | # Note that this will remove all occurrences of -G. 193 | list(REMOVE_ITEM depends_CUDA_NVCC_FLAGS "-G") 194 | endif() 195 | 196 | # nvcc doesn't define __CUDACC__ for some reason when generating dependency files. This 197 | # can cause incorrect dependencies when #including files based on this macro which is 198 | # defined in the generating passes of nvcc invocation. We will go ahead and manually 199 | # define this for now until a future version fixes this bug. 200 | set(CUDACC_DEFINE -D__CUDACC__) 201 | 202 | # Generate the dependency file 203 | cuda_execute_process( 204 | "Generating dependency file: ${NVCC_generated_dependency_file}" 205 | COMMAND "${CUDA_NVCC_EXECUTABLE}" 206 | -M 207 | ${CUDACC_DEFINE} 208 | "${source_file}" 209 | -o "${NVCC_generated_dependency_file}" 210 | ${CCBIN} 211 | ${nvcc_flags} 212 | ${nvcc_host_compiler_flags} 213 | ${depends_CUDA_NVCC_FLAGS} 214 | -DNVCC 215 | ${CUDA_NVCC_INCLUDE_ARGS} 216 | ) 217 | 218 | if(CUDA_result) 219 | message(FATAL_ERROR "Error generating ${generated_file}") 220 | endif() 221 | 222 | # Generate the cmake readable dependency file to a temp file. Don't put the 223 | # quotes just around the filenames for the input_file and output_file variables. 224 | # CMake will pass the quotes through and not be able to find the file. 225 | cuda_execute_process( 226 | "Generating temporary cmake readable file: ${cmake_dependency_file}.tmp" 227 | COMMAND "${CMAKE_COMMAND}" 228 | -D "input_file:FILEPATH=${NVCC_generated_dependency_file}" 229 | -D "output_file:FILEPATH=${cmake_dependency_file}.tmp" 230 | -D "verbose=${verbose}" 231 | -P "${CUDA_make2cmake}" 232 | ) 233 | 234 | if(CUDA_result) 235 | message(FATAL_ERROR "Error generating ${generated_file}") 236 | endif() 237 | 238 | # Copy the file if it is different 239 | cuda_execute_process( 240 | "Copy if different ${cmake_dependency_file}.tmp to ${cmake_dependency_file}" 241 | COMMAND "${CMAKE_COMMAND}" -E copy_if_different "${cmake_dependency_file}.tmp" "${cmake_dependency_file}" 242 | ) 243 | 244 | if(CUDA_result) 245 | message(FATAL_ERROR "Error generating ${generated_file}") 246 | endif() 247 | 248 | # Delete the temporary file 249 | cuda_execute_process( 250 | "Removing ${cmake_dependency_file}.tmp and ${NVCC_generated_dependency_file}" 251 | COMMAND "${CMAKE_COMMAND}" -E remove "${cmake_dependency_file}.tmp" "${NVCC_generated_dependency_file}" 252 | ) 253 | 254 | if(CUDA_result) 255 | message(FATAL_ERROR "Error generating ${generated_file}") 256 | endif() 257 | 258 | # Generate the code 259 | cuda_execute_process( 260 | "Generating ${generated_file}" 261 | COMMAND "${CUDA_NVCC_EXECUTABLE}" 262 | "${source_file}" 263 | ${cuda_language_flag} 264 | ${format_flag} -o "${generated_file}" 265 | ${CCBIN} 266 | ${nvcc_flags} 267 | ${nvcc_host_compiler_flags} 268 | ${CUDA_NVCC_FLAGS} 269 | -DNVCC 270 | ${CUDA_NVCC_INCLUDE_ARGS} 271 | ) 272 | 273 | if(CUDA_result) 274 | # Since nvcc can sometimes leave half done files make sure that we delete the output file. 275 | cuda_execute_process( 276 | "Removing ${generated_file}" 277 | COMMAND "${CMAKE_COMMAND}" -E remove "${generated_file}" 278 | ) 279 | message(FATAL_ERROR "Error generating file ${generated_file}") 280 | else() 281 | if(verbose) 282 | message("Generated ${generated_file} successfully.") 283 | endif() 284 | endif() 285 | 286 | # Cubin resource report commands. 287 | if( build_cubin ) 288 | # Run with -cubin to produce resource usage report. 289 | cuda_execute_process( 290 | "Generating ${generated_cubin_file}" 291 | COMMAND "${CUDA_NVCC_EXECUTABLE}" 292 | "${source_file}" 293 | ${CUDA_NVCC_FLAGS} 294 | ${nvcc_flags} 295 | ${CCBIN} 296 | ${nvcc_host_compiler_flags} 297 | -DNVCC 298 | -cubin 299 | -o "${generated_cubin_file}" 300 | ${CUDA_NVCC_INCLUDE_ARGS} 301 | ) 302 | 303 | # Execute the parser script. 304 | cuda_execute_process( 305 | "Executing the parser script" 306 | COMMAND "${CMAKE_COMMAND}" 307 | -D "input_file:STRING=${generated_cubin_file}" 308 | -P "${CUDA_parse_cubin}" 309 | ) 310 | 311 | endif() 312 | 313 | cmake_policy(POP) 314 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/SA.dir/cmake_clean.cmake: -------------------------------------------------------------------------------- 1 | file(REMOVE_RECURSE 2 | "CMakeFiles/SA.dir/SA_generated_sa.cu.o" 3 | "CMakeFiles/SA.dir/sa_ext.cpp.o" 4 | "SA.pdb" 5 | "SA" 6 | ) 7 | 8 | # Per-language clean rules from dependency scanning. 9 | foreach(lang CXX) 10 | include(CMakeFiles/SA.dir/cmake_clean_${lang}.cmake OPTIONAL) 11 | endforeach() 12 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/SA.dir/flags.make: -------------------------------------------------------------------------------- 1 | # CMAKE generated file: DO NOT EDIT! 2 | # Generated by "Unix Makefiles" Generator, CMake Version 3.10 3 | 4 | # compile CXX with /usr/bin/c++ 5 | CXX_FLAGS = -fPIC -D_GLIBCXX_USE_CXX11_ABI=0 -std=c++11 -std=gnu++11 6 | 7 | CXX_DEFINES = 8 | 9 | CXX_INCLUDES = -isystem /usr/local/cuda-10.0/include -isystem /home/guyuchao/software/libtorch/include -isystem /home/guyuchao/software/libtorch/include/torch/csrc/api/include 10 | 11 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/SA.dir/link.txt: -------------------------------------------------------------------------------- 1 | /usr/bin/c++ -fPIC -rdynamic CMakeFiles/SA.dir/sa_ext.cpp.o CMakeFiles/SA.dir/SA_generated_sa.cu.o -o SA -Wl,-rpath,/usr/local/cuda-10.0/lib64:/home/guyuchao/software/libtorch/lib /usr/local/cuda-10.0/lib64/libcudart.so /home/guyuchao/software/libtorch/lib/libtorch.so /home/guyuchao/software/libtorch/lib/libc10.so -lcuda /usr/local/cuda-10.0/lib64/libnvrtc.so /usr/local/cuda-10.0/lib64/libnvToolsExt.so /usr/local/cuda-10.0/lib64/libcudart.so /home/guyuchao/software/libtorch/lib/libc10_cuda.so -Wl,--no-as-needed,/home/guyuchao/software/libtorch/lib/libcaffe2.so -Wl,--as-needed -Wl,--no-as-needed,/home/guyuchao/software/libtorch/lib/libcaffe2_gpu.so -Wl,--as-needed /home/guyuchao/software/libtorch/lib/libc10_cuda.so /home/guyuchao/software/libtorch/lib/libcaffe2.so /home/guyuchao/software/libtorch/lib/libc10.so -lpthread /usr/local/cuda-10.0/lib64/libcufft.so /usr/local/cuda-10.0/lib64/libcurand.so /usr/lib/x86_64-linux-gnu/libcudnn.so /usr/local/cuda/lib64/libculibos.a -ldl /usr/local/cuda/lib64/libculibos.a -ldl /usr/local/cuda-10.0/lib64/libcublas.so /usr/local/cuda-10.0/lib64/libcudart.so 2 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/SA.dir/progress.make: -------------------------------------------------------------------------------- 1 | CMAKE_PROGRESS_1 = 1 2 | CMAKE_PROGRESS_2 = 2 3 | CMAKE_PROGRESS_3 = 3 4 | 5 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/SA.dir/sa_ext.cpp.o: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/SA.dir/sa_ext.cpp.o -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/TargetDirectories.txt: -------------------------------------------------------------------------------- 1 | /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/rebuild_cache.dir 2 | /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/edit_cache.dir 3 | /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/SA.dir 4 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/cmake.check_cache: -------------------------------------------------------------------------------- 1 | # This file is generated by cmake for dependency checking of the CMakeCache.txt file 2 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/feature_tests.bin: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/feature_tests.bin -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/feature_tests.c: -------------------------------------------------------------------------------- 1 | 2 | const char features[] = {"\n" 3 | "C_FEATURE:" 4 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 304 5 | "1" 6 | #else 7 | "0" 8 | #endif 9 | "c_function_prototypes\n" 10 | "C_FEATURE:" 11 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 304 && defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L 12 | "1" 13 | #else 14 | "0" 15 | #endif 16 | "c_restrict\n" 17 | "C_FEATURE:" 18 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 406 && defined(__STDC_VERSION__) && __STDC_VERSION__ >= 201000L 19 | "1" 20 | #else 21 | "0" 22 | #endif 23 | "c_static_assert\n" 24 | "C_FEATURE:" 25 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 304 && defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L 26 | "1" 27 | #else 28 | "0" 29 | #endif 30 | "c_variadic_macros\n" 31 | 32 | }; 33 | 34 | int main(int argc, char** argv) { (void)argv; return features[argc]; } 35 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/feature_tests.cxx: -------------------------------------------------------------------------------- 1 | 2 | const char features[] = {"\n" 3 | "CXX_FEATURE:" 4 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 500 && __cplusplus >= 201402L 5 | "1" 6 | #else 7 | "0" 8 | #endif 9 | "cxx_aggregate_default_initializers\n" 10 | "CXX_FEATURE:" 11 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 407 && __cplusplus >= 201103L 12 | "1" 13 | #else 14 | "0" 15 | #endif 16 | "cxx_alias_templates\n" 17 | "CXX_FEATURE:" 18 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 408 && __cplusplus >= 201103L 19 | "1" 20 | #else 21 | "0" 22 | #endif 23 | "cxx_alignas\n" 24 | "CXX_FEATURE:" 25 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 408 && __cplusplus >= 201103L 26 | "1" 27 | #else 28 | "0" 29 | #endif 30 | "cxx_alignof\n" 31 | "CXX_FEATURE:" 32 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 408 && __cplusplus >= 201103L 33 | "1" 34 | #else 35 | "0" 36 | #endif 37 | "cxx_attributes\n" 38 | "CXX_FEATURE:" 39 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 409 && __cplusplus > 201103L 40 | "1" 41 | #else 42 | "0" 43 | #endif 44 | "cxx_attribute_deprecated\n" 45 | "CXX_FEATURE:" 46 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 47 | "1" 48 | #else 49 | "0" 50 | #endif 51 | "cxx_auto_type\n" 52 | "CXX_FEATURE:" 53 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 409 && __cplusplus > 201103L 54 | "1" 55 | #else 56 | "0" 57 | #endif 58 | "cxx_binary_literals\n" 59 | "CXX_FEATURE:" 60 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 406 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 61 | "1" 62 | #else 63 | "0" 64 | #endif 65 | "cxx_constexpr\n" 66 | "CXX_FEATURE:" 67 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 409 && __cplusplus > 201103L 68 | "1" 69 | #else 70 | "0" 71 | #endif 72 | "cxx_contextual_conversions\n" 73 | "CXX_FEATURE:" 74 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 75 | "1" 76 | #else 77 | "0" 78 | #endif 79 | "cxx_decltype\n" 80 | "CXX_FEATURE:" 81 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 409 && __cplusplus > 201103L 82 | "1" 83 | #else 84 | "0" 85 | #endif 86 | "cxx_decltype_auto\n" 87 | "CXX_FEATURE:" 88 | #if ((__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__) >= 40801) && __cplusplus >= 201103L 89 | "1" 90 | #else 91 | "0" 92 | #endif 93 | "cxx_decltype_incomplete_return_types\n" 94 | "CXX_FEATURE:" 95 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 96 | "1" 97 | #else 98 | "0" 99 | #endif 100 | "cxx_default_function_template_args\n" 101 | "CXX_FEATURE:" 102 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 103 | "1" 104 | #else 105 | "0" 106 | #endif 107 | "cxx_defaulted_functions\n" 108 | "CXX_FEATURE:" 109 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 406 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 110 | "1" 111 | #else 112 | "0" 113 | #endif 114 | "cxx_defaulted_move_initializers\n" 115 | "CXX_FEATURE:" 116 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 407 && __cplusplus >= 201103L 117 | "1" 118 | #else 119 | "0" 120 | #endif 121 | "cxx_delegating_constructors\n" 122 | "CXX_FEATURE:" 123 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 124 | "1" 125 | #else 126 | "0" 127 | #endif 128 | "cxx_deleted_functions\n" 129 | "CXX_FEATURE:" 130 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 409 && __cplusplus > 201103L 131 | "1" 132 | #else 133 | "0" 134 | #endif 135 | "cxx_digit_separators\n" 136 | "CXX_FEATURE:" 137 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 406 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 138 | "1" 139 | #else 140 | "0" 141 | #endif 142 | "cxx_enum_forward_declarations\n" 143 | "CXX_FEATURE:" 144 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 405 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 145 | "1" 146 | #else 147 | "0" 148 | #endif 149 | "cxx_explicit_conversions\n" 150 | "CXX_FEATURE:" 151 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 407 && __cplusplus >= 201103L 152 | "1" 153 | #else 154 | "0" 155 | #endif 156 | "cxx_extended_friend_declarations\n" 157 | "CXX_FEATURE:" 158 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 159 | "1" 160 | #else 161 | "0" 162 | #endif 163 | "cxx_extern_templates\n" 164 | "CXX_FEATURE:" 165 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 407 && __cplusplus >= 201103L 166 | "1" 167 | #else 168 | "0" 169 | #endif 170 | "cxx_final\n" 171 | "CXX_FEATURE:" 172 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 173 | "1" 174 | #else 175 | "0" 176 | #endif 177 | "cxx_func_identifier\n" 178 | "CXX_FEATURE:" 179 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 180 | "1" 181 | #else 182 | "0" 183 | #endif 184 | "cxx_generalized_initializers\n" 185 | "CXX_FEATURE:" 186 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 409 && __cplusplus > 201103L 187 | "1" 188 | #else 189 | "0" 190 | #endif 191 | "cxx_generic_lambdas\n" 192 | "CXX_FEATURE:" 193 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 408 && __cplusplus >= 201103L 194 | "1" 195 | #else 196 | "0" 197 | #endif 198 | "cxx_inheriting_constructors\n" 199 | "CXX_FEATURE:" 200 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 201 | "1" 202 | #else 203 | "0" 204 | #endif 205 | "cxx_inline_namespaces\n" 206 | "CXX_FEATURE:" 207 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 405 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 208 | "1" 209 | #else 210 | "0" 211 | #endif 212 | "cxx_lambdas\n" 213 | "CXX_FEATURE:" 214 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 409 && __cplusplus > 201103L 215 | "1" 216 | #else 217 | "0" 218 | #endif 219 | "cxx_lambda_init_captures\n" 220 | "CXX_FEATURE:" 221 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 405 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 222 | "1" 223 | #else 224 | "0" 225 | #endif 226 | "cxx_local_type_template_args\n" 227 | "CXX_FEATURE:" 228 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 229 | "1" 230 | #else 231 | "0" 232 | #endif 233 | "cxx_long_long_type\n" 234 | "CXX_FEATURE:" 235 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 406 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 236 | "1" 237 | #else 238 | "0" 239 | #endif 240 | "cxx_noexcept\n" 241 | "CXX_FEATURE:" 242 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 407 && __cplusplus >= 201103L 243 | "1" 244 | #else 245 | "0" 246 | #endif 247 | "cxx_nonstatic_member_init\n" 248 | "CXX_FEATURE:" 249 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 406 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 250 | "1" 251 | #else 252 | "0" 253 | #endif 254 | "cxx_nullptr\n" 255 | "CXX_FEATURE:" 256 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 407 && __cplusplus >= 201103L 257 | "1" 258 | #else 259 | "0" 260 | #endif 261 | "cxx_override\n" 262 | "CXX_FEATURE:" 263 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 406 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 264 | "1" 265 | #else 266 | "0" 267 | #endif 268 | "cxx_range_for\n" 269 | "CXX_FEATURE:" 270 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 405 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 271 | "1" 272 | #else 273 | "0" 274 | #endif 275 | "cxx_raw_string_literals\n" 276 | "CXX_FEATURE:" 277 | #if ((__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__) >= 40801) && __cplusplus >= 201103L 278 | "1" 279 | #else 280 | "0" 281 | #endif 282 | "cxx_reference_qualified_functions\n" 283 | "CXX_FEATURE:" 284 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 500 && __cplusplus >= 201402L 285 | "1" 286 | #else 287 | "0" 288 | #endif 289 | "cxx_relaxed_constexpr\n" 290 | "CXX_FEATURE:" 291 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 409 && __cplusplus > 201103L 292 | "1" 293 | #else 294 | "0" 295 | #endif 296 | "cxx_return_type_deduction\n" 297 | "CXX_FEATURE:" 298 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 299 | "1" 300 | #else 301 | "0" 302 | #endif 303 | "cxx_right_angle_brackets\n" 304 | "CXX_FEATURE:" 305 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 306 | "1" 307 | #else 308 | "0" 309 | #endif 310 | "cxx_rvalue_references\n" 311 | "CXX_FEATURE:" 312 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 313 | "1" 314 | #else 315 | "0" 316 | #endif 317 | "cxx_sizeof_member\n" 318 | "CXX_FEATURE:" 319 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 320 | "1" 321 | #else 322 | "0" 323 | #endif 324 | "cxx_static_assert\n" 325 | "CXX_FEATURE:" 326 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 327 | "1" 328 | #else 329 | "0" 330 | #endif 331 | "cxx_strong_enums\n" 332 | "CXX_FEATURE:" 333 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && __cplusplus 334 | "1" 335 | #else 336 | "0" 337 | #endif 338 | "cxx_template_template_parameters\n" 339 | "CXX_FEATURE:" 340 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 408 && __cplusplus >= 201103L 341 | "1" 342 | #else 343 | "0" 344 | #endif 345 | "cxx_thread_local\n" 346 | "CXX_FEATURE:" 347 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 348 | "1" 349 | #else 350 | "0" 351 | #endif 352 | "cxx_trailing_return_types\n" 353 | "CXX_FEATURE:" 354 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 355 | "1" 356 | #else 357 | "0" 358 | #endif 359 | "cxx_unicode_literals\n" 360 | "CXX_FEATURE:" 361 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 362 | "1" 363 | #else 364 | "0" 365 | #endif 366 | "cxx_uniform_initialization\n" 367 | "CXX_FEATURE:" 368 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 406 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 369 | "1" 370 | #else 371 | "0" 372 | #endif 373 | "cxx_unrestricted_unions\n" 374 | "CXX_FEATURE:" 375 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 407 && __cplusplus >= 201103L 376 | "1" 377 | #else 378 | "0" 379 | #endif 380 | "cxx_user_literals\n" 381 | "CXX_FEATURE:" 382 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 500 && __cplusplus >= 201402L 383 | "1" 384 | #else 385 | "0" 386 | #endif 387 | "cxx_variable_templates\n" 388 | "CXX_FEATURE:" 389 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 390 | "1" 391 | #else 392 | "0" 393 | #endif 394 | "cxx_variadic_macros\n" 395 | "CXX_FEATURE:" 396 | #if (__GNUC__ * 100 + __GNUC_MINOR__) >= 404 && (__cplusplus >= 201103L || (defined(__GXX_EXPERIMENTAL_CXX0X__) && __GXX_EXPERIMENTAL_CXX0X__)) 397 | "1" 398 | #else 399 | "0" 400 | #endif 401 | "cxx_variadic_templates\n" 402 | 403 | }; 404 | 405 | int main(int argc, char** argv) { (void)argv; return features[argc]; } 406 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/CMakeFiles/progress.marks: -------------------------------------------------------------------------------- 1 | 3 2 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/Makefile: -------------------------------------------------------------------------------- 1 | # CMAKE generated file: DO NOT EDIT! 2 | # Generated by "Unix Makefiles" Generator, CMake Version 3.10 3 | 4 | # Default target executed when no arguments are given to make. 5 | default_target: all 6 | 7 | .PHONY : default_target 8 | 9 | # Allow only one "make -f Makefile2" at a time, but pass parallelism. 10 | .NOTPARALLEL: 11 | 12 | 13 | #============================================================================= 14 | # Special targets provided by cmake. 15 | 16 | # Disable implicit rules so canonical targets will work. 17 | .SUFFIXES: 18 | 19 | 20 | # Remove some rules from gmake that .SUFFIXES does not remove. 21 | SUFFIXES = 22 | 23 | .SUFFIXES: .hpux_make_needs_suffix_list 24 | 25 | 26 | # Suppress display of executed commands. 27 | $(VERBOSE).SILENT: 28 | 29 | 30 | # A target that is always out of date. 31 | cmake_force: 32 | 33 | .PHONY : cmake_force 34 | 35 | #============================================================================= 36 | # Set environment variables for the build. 37 | 38 | # The shell in which to execute make rules. 39 | SHELL = /bin/sh 40 | 41 | # The CMake executable. 42 | CMAKE_COMMAND = /usr/bin/cmake 43 | 44 | # The command to remove a file. 45 | RM = /usr/bin/cmake -E remove -f 46 | 47 | # Escaping for special characters. 48 | EQUALS = = 49 | 50 | # The top-level source directory on which CMake was run. 51 | CMAKE_SOURCE_DIR = /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module 52 | 53 | # The top-level build directory on which CMake was run. 54 | CMAKE_BINARY_DIR = /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild 55 | 56 | #============================================================================= 57 | # Targets provided globally by CMake. 58 | 59 | # Special rule for the target rebuild_cache 60 | rebuild_cache: 61 | @$(CMAKE_COMMAND) -E cmake_echo_color --switch=$(COLOR) --cyan "Running CMake to regenerate build system..." 62 | /usr/bin/cmake -H$(CMAKE_SOURCE_DIR) -B$(CMAKE_BINARY_DIR) 63 | .PHONY : rebuild_cache 64 | 65 | # Special rule for the target rebuild_cache 66 | rebuild_cache/fast: rebuild_cache 67 | 68 | .PHONY : rebuild_cache/fast 69 | 70 | # Special rule for the target edit_cache 71 | edit_cache: 72 | @$(CMAKE_COMMAND) -E cmake_echo_color --switch=$(COLOR) --cyan "Running CMake cache editor..." 73 | /usr/bin/ccmake -H$(CMAKE_SOURCE_DIR) -B$(CMAKE_BINARY_DIR) 74 | .PHONY : edit_cache 75 | 76 | # Special rule for the target edit_cache 77 | edit_cache/fast: edit_cache 78 | 79 | .PHONY : edit_cache/fast 80 | 81 | # The main all target 82 | all: cmake_check_build_system 83 | $(CMAKE_COMMAND) -E cmake_progress_start /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles/progress.marks 84 | $(MAKE) -f CMakeFiles/Makefile2 all 85 | $(CMAKE_COMMAND) -E cmake_progress_start /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/CMakeFiles 0 86 | .PHONY : all 87 | 88 | # The main clean target 89 | clean: 90 | $(MAKE) -f CMakeFiles/Makefile2 clean 91 | .PHONY : clean 92 | 93 | # The main clean target 94 | clean/fast: clean 95 | 96 | .PHONY : clean/fast 97 | 98 | # Prepare targets for installation. 99 | preinstall: all 100 | $(MAKE) -f CMakeFiles/Makefile2 preinstall 101 | .PHONY : preinstall 102 | 103 | # Prepare targets for installation. 104 | preinstall/fast: 105 | $(MAKE) -f CMakeFiles/Makefile2 preinstall 106 | .PHONY : preinstall/fast 107 | 108 | # clear depends 109 | depend: 110 | $(CMAKE_COMMAND) -H$(CMAKE_SOURCE_DIR) -B$(CMAKE_BINARY_DIR) --check-build-system CMakeFiles/Makefile.cmake 1 111 | .PHONY : depend 112 | 113 | #============================================================================= 114 | # Target rules for targets named SA 115 | 116 | # Build rule for target. 117 | SA: cmake_check_build_system 118 | $(MAKE) -f CMakeFiles/Makefile2 SA 119 | .PHONY : SA 120 | 121 | # fast build rule for target. 122 | SA/fast: 123 | $(MAKE) -f CMakeFiles/SA.dir/build.make CMakeFiles/SA.dir/build 124 | .PHONY : SA/fast 125 | 126 | sa_ext.o: sa_ext.cpp.o 127 | 128 | .PHONY : sa_ext.o 129 | 130 | # target to build an object file 131 | sa_ext.cpp.o: 132 | $(MAKE) -f CMakeFiles/SA.dir/build.make CMakeFiles/SA.dir/sa_ext.cpp.o 133 | .PHONY : sa_ext.cpp.o 134 | 135 | sa_ext.i: sa_ext.cpp.i 136 | 137 | .PHONY : sa_ext.i 138 | 139 | # target to preprocess a source file 140 | sa_ext.cpp.i: 141 | $(MAKE) -f CMakeFiles/SA.dir/build.make CMakeFiles/SA.dir/sa_ext.cpp.i 142 | .PHONY : sa_ext.cpp.i 143 | 144 | sa_ext.s: sa_ext.cpp.s 145 | 146 | .PHONY : sa_ext.s 147 | 148 | # target to generate assembly for a file 149 | sa_ext.cpp.s: 150 | $(MAKE) -f CMakeFiles/SA.dir/build.make CMakeFiles/SA.dir/sa_ext.cpp.s 151 | .PHONY : sa_ext.cpp.s 152 | 153 | # Help Target 154 | help: 155 | @echo "The following are some of the valid targets for this Makefile:" 156 | @echo "... all (the default if no target is provided)" 157 | @echo "... clean" 158 | @echo "... depend" 159 | @echo "... rebuild_cache" 160 | @echo "... edit_cache" 161 | @echo "... SA" 162 | @echo "... sa_ext.o" 163 | @echo "... sa_ext.i" 164 | @echo "... sa_ext.s" 165 | .PHONY : help 166 | 167 | 168 | 169 | #============================================================================= 170 | # Special targets to cleanup operation of make. 171 | 172 | # Special rule to run CMake to check the build system integrity. 173 | # No rule that depends on this can have commands that come from listfiles 174 | # because they might be regenerated. 175 | cmake_check_build_system: 176 | $(CMAKE_COMMAND) -H$(CMAKE_SOURCE_DIR) -B$(CMAKE_BINARY_DIR) --check-build-system CMakeFiles/Makefile.cmake 0 177 | .PHONY : cmake_check_build_system 178 | 179 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/SA: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/PCSA/PCSA_Module/cmakebuild/SA -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/cmake_install.cmake: -------------------------------------------------------------------------------- 1 | # Install script for directory: /home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module 2 | 3 | # Set the install prefix 4 | if(NOT DEFINED CMAKE_INSTALL_PREFIX) 5 | set(CMAKE_INSTALL_PREFIX "/usr/local") 6 | endif() 7 | string(REGEX REPLACE "/$" "" CMAKE_INSTALL_PREFIX "${CMAKE_INSTALL_PREFIX}") 8 | 9 | # Set the install configuration name. 10 | if(NOT DEFINED CMAKE_INSTALL_CONFIG_NAME) 11 | if(BUILD_TYPE) 12 | string(REGEX REPLACE "^[^A-Za-z0-9_]+" "" 13 | CMAKE_INSTALL_CONFIG_NAME "${BUILD_TYPE}") 14 | else() 15 | set(CMAKE_INSTALL_CONFIG_NAME "") 16 | endif() 17 | message(STATUS "Install configuration: \"${CMAKE_INSTALL_CONFIG_NAME}\"") 18 | endif() 19 | 20 | # Set the component getting installed. 21 | if(NOT CMAKE_INSTALL_COMPONENT) 22 | if(COMPONENT) 23 | message(STATUS "Install component: \"${COMPONENT}\"") 24 | set(CMAKE_INSTALL_COMPONENT "${COMPONENT}") 25 | else() 26 | set(CMAKE_INSTALL_COMPONENT) 27 | endif() 28 | endif() 29 | 30 | # Install shared libraries without execute permission? 31 | if(NOT DEFINED CMAKE_INSTALL_SO_NO_EXE) 32 | set(CMAKE_INSTALL_SO_NO_EXE "1") 33 | endif() 34 | 35 | # Is this installation the result of a crosscompile? 36 | if(NOT DEFINED CMAKE_CROSSCOMPILING) 37 | set(CMAKE_CROSSCOMPILING "FALSE") 38 | endif() 39 | 40 | if(CMAKE_INSTALL_COMPONENT) 41 | set(CMAKE_INSTALL_MANIFEST "install_manifest_${CMAKE_INSTALL_COMPONENT}.txt") 42 | else() 43 | set(CMAKE_INSTALL_MANIFEST "install_manifest.txt") 44 | endif() 45 | 46 | string(REPLACE ";" "\n" CMAKE_INSTALL_MANIFEST_CONTENT 47 | "${CMAKE_INSTALL_MANIFEST_FILES}") 48 | file(WRITE "/home/guyuchao/ssd/PycharmProjects/videofastsal/Models/plocaltnonlocal/SelfAttention_Module/cmakebuild/${CMAKE_INSTALL_MANIFEST}" 49 | "${CMAKE_INSTALL_MANIFEST_CONTENT}") 50 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/detect_cuda_compute_capabilities.cpp: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | int main() 4 | { 5 | int count = 0; 6 | if (cudaSuccess != cudaGetDeviceCount(&count)) return -1; 7 | if (count == 0) return -1; 8 | for (int device = 0; device < count; ++device) 9 | { 10 | cudaDeviceProp prop; 11 | if (cudaSuccess == cudaGetDeviceProperties(&prop, device)) 12 | std::printf("%d.%d ", prop.major, prop.minor); 13 | } 14 | return 0; 15 | } 16 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/cmakebuild/detect_cuda_version.cc: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | int main() { 4 | printf("%d.%d", CUDA_VERSION / 1000, (CUDA_VERSION / 10) % 100); 5 | return 0; 6 | } 7 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/reference.cpp: -------------------------------------------------------------------------------- 1 | #include 2 | //#include 3 | #include 4 | #include 5 | 6 | //串行比对 7 | void sa_weight_forward_Ref(const torch::Tensor& query,const torch::Tensor& key,torch::Tensor& weight,int B,int T,int C,int H,int W,int radius,int dilation){ 8 | int diameter=2*radius+1; 9 | 10 | for(int batch=0;batch=0&&dw+w>=0){ 22 | for(int c=0;c=0&&h+dh=0&&w+dw=0&&h+dh=0&&w+dw=0&&h+dh=0&&w+dw=0&&dw+w>=0){ 141 | for(int c=0;c=0&&h+dh=0&&w+dw 2 | #include 3 | #include 4 | #include 5 | #include 6 | 7 | #define TensorAccessor5D torch::PackedTensorAccessor 8 | /* 9 | #if !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 600 10 | #else 11 | static __inline__ __device__ double atomicAdd(double *address, double val) { 12 | unsigned long long int* address_as_ull = (unsigned long long int*)address; 13 | unsigned long long int old = *address_as_ull, assumed; 14 | if (val==0.0) 15 | return __longlong_as_double(old); 16 | do { 17 | assumed = old; 18 | old = atomicCAS(address_as_ull, assumed, __double_as_longlong(val +__longlong_as_double(assumed))); 19 | } while (assumed != old); 20 | return __longlong_as_double(old); 21 | } 22 | #endif 23 | */ 24 | template 25 | __global__ 26 | void sa_weight_forward_kernel( 27 | const TensorAccessor5D query, 28 | const TensorAccessor5D key, 29 | TensorAccessor5D weight,int B,int T,int C,int H,int W,int radius,int dilation){ 30 | int w = blockIdx.x * blockDim.x + threadIdx.x;//col 31 | int h = blockIdx.y * blockDim.y + threadIdx.y;//row 32 | int time = blockIdx.z;//time 33 | int diameter=2*radius+1; 34 | 35 | //query B*T*C*H*W 36 | //key B*T*C*H*W 37 | //weight B*T*9T*H*W 38 | if(w=0&&w+dw=0){ 45 | for(int c=0;c 60 | __global__ 61 | void sa_map_forward_kernel( 62 | const TensorAccessor5D weight, 63 | const TensorAccessor5D proj, 64 | TensorAccessor5D out,int B,int T,int C,int H,int W,int radius,int dilation){ 65 | int w = blockIdx.x * blockDim.x + threadIdx.x;//col 66 | int h = blockIdx.y * blockDim.y + threadIdx.y;//row 67 | int time = blockIdx.z;//time 68 | int diameter=2*radius+1; 69 | 70 | //weight B*T*9T*H*W 71 | //proj B*T*C*H*W 72 | //out B*T*C*H*W 73 | if(w=0&&w+dw=0){ 81 | scalar_t weight_temp=weight[batch][time][cal_time*diameter*diameter+(dh/dilation+radius)*(2*radius+1)+(dw/dilation+radius)][h][w]; 82 | scalar_t proj_value=proj[batch][cal_time][c][h+dh][w+dw]; 83 | sum+=weight_temp*proj_value; 84 | } 85 | } 86 | } 87 | } 88 | out[batch][time][c][h][w]=sum; 89 | } 90 | } 91 | } 92 | } 93 | 94 | template 95 | __global__ 96 | void sa_weight_backward_kernel_query( 97 | const TensorAccessor5D dweight, 98 | const TensorAccessor5D key, 99 | TensorAccessor5D dquery,int B,int T,int C,int H,int W,int radius,int dilation){ 100 | int w = blockIdx.x * blockDim.x + threadIdx.x;//col 101 | int h = blockIdx.y * blockDim.y + threadIdx.y;//row 102 | int time = blockIdx.z;//time 103 | int diameter=2*radius+1; 104 | 105 | //weight B*T*9T*H*W 106 | //proj B*T*C*H*W 107 | //out B*T*C*H*W 108 | if(w=0&&w+dw=0){ 116 | scalar_t _dweight=dweight[batch][time][cal_time*diameter*diameter+(dh/dilation+radius)*(2*radius+1)+(dw/dilation+radius)][h][w]; 117 | scalar_t _key=key[batch][cal_time][c][h+dh][w+dw]; 118 | sum+=_dweight*_key; 119 | } 120 | } 121 | } 122 | } 123 | dquery[batch][time][c][h][w]=sum; 124 | } 125 | } 126 | } 127 | } 128 | 129 | template 130 | __global__ 131 | void sa_weight_backward_kernel_key( 132 | const TensorAccessor5D dweight, 133 | const TensorAccessor5D query, 134 | TensorAccessor5D dkey,int B,int T,int C,int H,int W,int radius,int dilation){ 135 | int w = blockIdx.x * blockDim.x + threadIdx.x;//col 136 | int h = blockIdx.y * blockDim.y + threadIdx.y;//row 137 | int time = blockIdx.z;//time 138 | int diameter=2*radius+1; 139 | 140 | //weight B*T*9T*H*W 141 | //proj B*T*C*H*W 142 | //out B*T*C*H*W 143 | if(w=0&&w+dw=0){ 150 | scalar_t _dweight=dweight[batch][time][cal_time*diameter*diameter+(dh/dilation+radius)*(2*radius+1)+(dw/dilation+radius)][h][w]; 151 | scalar_t _query=query[batch][time][c][h][w]; 152 | atomicAdd(&dkey[batch][cal_time][c][h+dh][w+dw],_dweight*_query); 153 | } 154 | } 155 | } 156 | } 157 | } 158 | } 159 | } 160 | } 161 | 162 | template 163 | __global__ 164 | void sa_map_backward_kernel_weight( 165 | const TensorAccessor5D dout, 166 | const TensorAccessor5D proj, 167 | TensorAccessor5D dweight,int B,int T,int C,int H,int W,int radius,int dilation){ 168 | int w = blockIdx.x * blockDim.x + threadIdx.x;//col 169 | int h = blockIdx.y * blockDim.y + threadIdx.y;//row 170 | int time = blockIdx.z;//time 171 | int diameter=2*radius+1; 172 | 173 | //weight B*T*9T*H*W 174 | //proj B*T*C*H*W 175 | //out B*T*C*H*W 176 | if(w=0&&w+dw=0){ 184 | scalar_t _proj=proj[batch][cal_time][c][h+dh][w+dw]; 185 | scalar_t _dout=dout[batch][time][c][h][w]; 186 | sum+=_dout*_proj; 187 | } 188 | } 189 | dweight[batch][time][cal_time*diameter*diameter+(dh/dilation+radius)*(2*radius+1)+(dw/dilation+radius)][h][w]=sum; 190 | } 191 | } 192 | } 193 | } 194 | } 195 | } 196 | 197 | template 198 | __global__ 199 | void sa_map_backward_kernel_proj( 200 | const TensorAccessor5D dout, 201 | const TensorAccessor5D weight, 202 | TensorAccessor5D dproj,int B,int T,int C,int H,int W,int radius,int dilation){ 203 | int w = blockIdx.x * blockDim.x + threadIdx.x;//col 204 | int h = blockIdx.y * blockDim.y + threadIdx.y;//row 205 | int time = blockIdx.z;//time 206 | int diameter=2*radius+1; 207 | //weight B*T*9T*H*W 208 | //proj B*T*C*H*W 209 | //out B*T*C*H*W 210 | if(w=0&&w+dw=0){ 217 | scalar_t weight_temp=weight[batch][time][cal_time*diameter*diameter+(dh/dilation+radius)*(2*radius+1)+(dw/dilation+radius)][h][w]; 218 | scalar_t _dout=dout[batch][time][c][h][w]; 219 | atomicAdd(&dproj[batch][cal_time][c][h+dh][w+dw],_dout*weight_temp); 220 | } 221 | } 222 | } 223 | } 224 | } 225 | } 226 | } 227 | } 228 | 229 | void _sa_weight_forward_cuda(const torch::Tensor& query,const torch::Tensor& key,torch::Tensor& weight,int B,int T,int C,int H,int W,int radius,int dilation){ 230 | dim3 threads(16,16); 231 | dim3 blocks((W+threads.x-1)/threads.x,(H+threads.y-1)/threads.y,T); 232 | 233 | AT_DISPATCH_FLOATING_TYPES(weight.scalar_type(), "sa_weight_forward_cuda", ([&] { 234 | sa_weight_forward_kernel<<>>( 235 | query.packed_accessor(), 236 | key.packed_accessor(), 237 | weight.packed_accessor(),B,T,C,H,W,radius,dilation); 238 | })); 239 | } 240 | 241 | void _sa_map_forward_cuda(const torch::Tensor& weight,const torch::Tensor& proj,torch::Tensor& out,int B,int T,int C,int H,int W,int radius,int dilation){ 242 | dim3 threads(16,16); 243 | dim3 blocks((W+threads.x-1)/threads.x,(H+threads.y-1)/threads.y,T); 244 | AT_DISPATCH_FLOATING_TYPES(weight.scalar_type(), "sa_map_forward_cuda", ([&] { 245 | sa_map_forward_kernel<<>>( 246 | weight.packed_accessor(), 247 | proj.packed_accessor(), 248 | out.packed_accessor(),B,T,C,H,W,radius,dilation); 249 | })); 250 | } 251 | 252 | void _sa_weight_backward_cuda(const torch::Tensor& dw,const torch::Tensor& query, 253 | const torch::Tensor& key,torch::Tensor& dquery,torch::Tensor& dkey, 254 | int B,int T,int C,int H,int W,int radius,int dilation){ 255 | dim3 threads(16,16); 256 | dim3 blocks((W+threads.x-1)/threads.x,(H+threads.y-1)/threads.y,T); 257 | AT_DISPATCH_FLOATING_TYPES(dw.scalar_type(), "sa_weight_backward_cuda", ([&] { 258 | const TensorAccessor5D dw_acc=dw.packed_accessor(); 259 | const TensorAccessor5D query_acc=query.packed_accessor(); 260 | const TensorAccessor5D key_acc=key.packed_accessor(); 261 | TensorAccessor5D dquery_acc=dquery.packed_accessor(); 262 | TensorAccessor5D dkey_acc=dkey.packed_accessor(); 263 | sa_weight_backward_kernel_query<<>>(dw_acc,key_acc,dquery_acc,B,T,C,H,W,radius,dilation); 264 | sa_weight_backward_kernel_key<<>>(dw_acc,query_acc,dkey_acc,B,T,C,H,W,radius,dilation); 265 | })); 266 | } 267 | 268 | void _sa_map_backward_cuda(const torch::Tensor& dout, const torch::Tensor& weight, 269 | const torch::Tensor& proj,torch::Tensor& dweight,torch::Tensor& dproj, 270 | int B,int T,int C,int H,int W,int radius,int dilation){ 271 | dim3 threads(16,16); 272 | dim3 blocks((W+threads.x-1)/threads.x,(H+threads.y-1)/threads.y,T); 273 | 274 | AT_DISPATCH_FLOATING_TYPES(dout.scalar_type(), "sa_map_backward_cuda", ([&] { 275 | const TensorAccessor5D dout_acc=dout.packed_accessor(); 276 | const TensorAccessor5D weight_acc=weight.packed_accessor(); 277 | const TensorAccessor5D proj_acc=proj.packed_accessor(); 278 | TensorAccessor5D dweight_acc=dweight.packed_accessor(); 279 | TensorAccessor5D dproj_acc=dproj.packed_accessor(); 280 | sa_map_backward_kernel_weight<<>>(dout_acc,proj_acc,dweight_acc,B,T,C,H,W,radius,dilation); 281 | sa_map_backward_kernel_proj<<>>(dout_acc,weight_acc,dproj_acc,B,T,C,H,W,radius,dilation); 282 | })); 283 | } 284 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/sa.cu.bak: -------------------------------------------------------------------------------- 1 | #include 2 | //#include 3 | #include 4 | #include 5 | #include 6 | 7 | #define TensorAccessor5D torch::PackedTensorAccessor 8 | 9 | template 10 | __global__ 11 | void sa_weight_forward_kernel( 12 | const TensorAccessor5D query, 13 | const TensorAccessor5D key, 14 | TensorAccessor5D weight,int B,int T,int C,int H,int W){ 15 | int col = blockIdx.x * blockDim.x + threadIdx.x;//col 16 | int row = blockIdx.y * blockDim.y + threadIdx.y;//row 17 | int time = blockIdx.z;//time 18 | //query B*T*C*H*W 19 | //key B*T*C*H*W 20 | //weight B*T*9T*H*W 21 | if(col=0&&col+w=0){ 28 | for(int channel=0;channel 43 | __global__ 44 | void sa_map_forward_kernel( 45 | const TensorAccessor5D weight, 46 | const TensorAccessor5D proj, 47 | TensorAccessor5D out,int B,int T,int C,int H,int W){ 48 | int col = blockIdx.x * blockDim.x + threadIdx.x;//col 49 | int row = blockIdx.y * blockDim.y + threadIdx.y;//row 50 | int time = blockIdx.z;//time 51 | //weight B*T*9T*H*W 52 | //proj B*T*C*H*W 53 | //out B*T*C*H*W 54 | if(col=0&&col+w=0){ 60 | float weight_temp=weight[batch][time][t*9+(h+1)*3+(w+1)][row][col]; 61 | for(int channel=0;channel 74 | __global__ 75 | void sa_weight_backward_kernel_query( 76 | const TensorAccessor5D dw, 77 | const TensorAccessor5D key, 78 | TensorAccessor5D dquery,int B,int T,int C,int H,int W){ 79 | int col = blockIdx.x * blockDim.x + threadIdx.x;//col 80 | int row = blockIdx.y * blockDim.y + threadIdx.y;//row 81 | int time = blockIdx.z;//time 82 | //weight B*T*9T*H*W 83 | //proj B*T*C*H*W 84 | //out B*T*C*H*W 85 | if(col=0&&col+w=0){ 91 | float _dw=dw[batch][time][t*9+(h+1)*3+(w+1)][row][col]; 92 | for(int channel=0;channel 105 | __global__ 106 | void sa_weight_backward_kernel_key( 107 | const TensorAccessor5D dw, 108 | const TensorAccessor5D query, 109 | TensorAccessor5D dkey,int B,int T,int C,int H,int W){ 110 | int col = blockIdx.x * blockDim.x + threadIdx.x;//col 111 | int row = blockIdx.y * blockDim.y + threadIdx.y;//row 112 | int time = blockIdx.z;//time 113 | //weight B*T*9T*H*W 114 | //proj B*T*C*H*W 115 | //out B*T*C*H*W 116 | if(col=0&&col+w=0){ 122 | float _dw=dw[batch][time][t*9+(h+1)*3+(w+1)][row][col]; 123 | for(int channel=0;channel 136 | __global__ 137 | void sa_map_backward_kernel_weight( 138 | const TensorAccessor5D dout, 139 | const TensorAccessor5D proj, 140 | TensorAccessor5D dweight,int B,int T,int C,int H,int W){ 141 | int col = blockIdx.x * blockDim.x + threadIdx.x;//col 142 | int row = blockIdx.y * blockDim.y + threadIdx.y;//row 143 | int time = blockIdx.z;//time 144 | //weight B*T*9T*H*W 145 | //proj B*T*C*H*W 146 | //out B*T*C*H*W 147 | if(col=0&&col+w=0){ 155 | float _proj=proj[batch][t][channel][row+h][col+w]; 156 | dweight[batch][time][t*9+(h+1)*3+(w+1)][row][col]+=_dout*_proj; 157 | } 158 | } 159 | } 160 | } 161 | } 162 | } 163 | } 164 | } 165 | 166 | template 167 | __global__ 168 | void sa_map_backward_kernel_proj( 169 | const TensorAccessor5D dout, 170 | const TensorAccessor5D weight, 171 | TensorAccessor5D dproj,int B,int T,int C,int H,int W){ 172 | int col = blockIdx.x * blockDim.x + threadIdx.x;//col 173 | int row = blockIdx.y * blockDim.y + threadIdx.y;//row 174 | int time = blockIdx.z;//time 175 | //weight B*T*9T*H*W 176 | //proj B*T*C*H*W 177 | //out B*T*C*H*W 178 | if(col=0&&col+w=0){ 184 | float weight_temp=weight[batch][time][t*9+(h+1)*3+(w+1)][row][col]; 185 | for(int channel=0;channel<<>>( 202 | query.packed_accessor(), 203 | key.packed_accessor(), 204 | weight.packed_accessor(),B,T,C,H,W); 205 | })); 206 | } 207 | 208 | void _sa_map_forward_cuda(const torch::Tensor& weight,const torch::Tensor& proj,torch::Tensor& out,int B,int T,int C,int H,int W){ 209 | dim3 threads(32,32); 210 | dim3 blocks((W+threads.x-1)/threads.x,(H+threads.y-1)/threads.y,T); 211 | AT_DISPATCH_FLOATING_TYPES(weight.scalar_type(), "sa_weight_forward_cuda", ([&] { 212 | sa_map_forward_kernel<<>>( 213 | weight.packed_accessor(), 214 | proj.packed_accessor(), 215 | out.packed_accessor(),B,T,C,H,W); 216 | })); 217 | } 218 | 219 | void _sa_weight_backward_cuda(const torch::Tensor& dw,const torch::Tensor& query, 220 | const torch::Tensor& key,torch::Tensor& dquery,torch::Tensor& dkey, 221 | int B,int T,int C,int H,int W){ 222 | dim3 threads(32,32); 223 | dim3 blocks((W+threads.x-1)/threads.x,(H+threads.y-1)/threads.y,T); 224 | AT_DISPATCH_FLOATING_TYPES(dw.scalar_type(), "sa_weight_forward_cuda", ([&] { 225 | const TensorAccessor5D dw_acc=dw.packed_accessor(); 226 | const TensorAccessor5D query_acc=query.packed_accessor(); 227 | const TensorAccessor5D key_acc=key.packed_accessor(); 228 | TensorAccessor5D dquery_acc=dquery.packed_accessor(); 229 | TensorAccessor5D dkey_acc=dkey.packed_accessor(); 230 | sa_weight_backward_kernel_query<<>>(dw_acc,key_acc,dquery_acc,B,T,C,H,W); 231 | sa_weight_backward_kernel_key<<>>(dw_acc,query_acc,dkey_acc,B,T,C,H,W); 232 | })); 233 | } 234 | 235 | void _sa_map_backward_cuda(const torch::Tensor& dout,const torch::Tensor& weight, 236 | const torch::Tensor& proj,torch::Tensor& dweight,torch::Tensor& dproj, 237 | int B,int T,int C,int H,int W){ 238 | dim3 threads(32,32); 239 | dim3 blocks((W+threads.x-1)/threads.x,(H+threads.y-1)/threads.y,T); 240 | AT_DISPATCH_FLOATING_TYPES(dout.scalar_type(), "sa_weight_forward_cuda", ([&] { 241 | const TensorAccessor5D dout_acc=dout.packed_accessor(); 242 | const TensorAccessor5D weight_acc=weight.packed_accessor(); 243 | const TensorAccessor5D proj_acc=proj.packed_accessor(); 244 | TensorAccessor5D dweight_acc=dweight.packed_accessor(); 245 | TensorAccessor5D dproj_acc=dproj.packed_accessor(); 246 | sa_map_backward_kernel_weight<<>>(dout_acc,proj_acc,dweight_acc,B,T,C,H,W); 247 | sa_map_backward_kernel_proj<<>>(dout_acc,weight_acc,dproj_acc,B,T,C,H,W); 248 | })); 249 | } 250 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/sa_ext.cpp: -------------------------------------------------------------------------------- 1 | //#include 2 | #include 3 | #include"utils.h" 4 | #include"timer.h" 5 | #include"reference.h" 6 | 7 | void get_sizes(const torch::Tensor& t,int *B,int *T,int *C,int *H,int *W){ 8 | *B=t.size(0); 9 | *T=t.size(1); 10 | *C=t.size(2); 11 | *H=t.size(3); 12 | *W=t.size(4); 13 | } 14 | 15 | void _sa_weight_forward_cuda(const torch::Tensor& query,const torch::Tensor& key,torch::Tensor& weight,int B,int T,int C,int H,int W,int radius,int dilation); 16 | void _sa_map_forward_cuda(const torch::Tensor& weight,const torch::Tensor& proj,torch::Tensor& out,int B,int T,int C,int H,int W,int radius,int dilation); 17 | void _sa_weight_backward_cuda(const torch::Tensor& dw,const torch::Tensor& query, 18 | const torch::Tensor& key,torch::Tensor& dquery,torch::Tensor& dkey, 19 | int B,int T,int C,int H,int W,int radius,int dilation); 20 | void _sa_map_backward_cuda(const torch::Tensor& dout,const torch::Tensor& weight, 21 | const torch::Tensor& proj,torch::Tensor& dweight,torch::Tensor& dproj, 22 | int B,int T,int C,int H,int W,int radius,int dilaiton); 23 | 24 | 25 | //forward declarations-------python pass information here 26 | void sa_weight_forward(const torch::Tensor& query,const torch::Tensor& key,torch::Tensor& weight,int radius,int dilation){ 27 | int B,T,C,H,W; 28 | get_sizes(query,&B,&T,&C,&H,&W); 29 | //GpuTimer timer; 30 | //timer.Start(); 31 | _sa_weight_forward_cuda(query,key,weight,B,T,C,H,W,radius,dilation); 32 | //timer.Stop(); 33 | //cudaDeviceSynchronize(); 34 | checkCudaErrors(cudaGetLastError()); 35 | //printf("Your code ran in: %f msecs.\n", timer.Elapsed()); 36 | } 37 | 38 | void sa_map_forward(const torch::Tensor& weight,const torch::Tensor& proj,torch::Tensor& out,int radius,int dilation){ 39 | int B,T,C,H,W; 40 | get_sizes(proj,&B,&T,&C,&H,&W); 41 | //GpuTimer timer; 42 | //timer.Start(); 43 | _sa_map_forward_cuda(weight,proj,out,B,T,C,H,W,radius,dilation); 44 | //timer.Stop(); 45 | //cudaDeviceSynchronize(); 46 | checkCudaErrors(cudaGetLastError()); 47 | //printf("Your code ran in: %f msecs.\n", timer.Elapsed()); 48 | } 49 | 50 | void sa_weight_backward(const torch::Tensor& dw,const torch::Tensor& query,const torch::Tensor& key,torch::Tensor& dquery,torch::Tensor& dkey,int radius,int dilation){ 51 | int B,T,C,H,W; 52 | get_sizes(query,&B,&T,&C,&H,&W); 53 | //GpuTimer timer; 54 | //timer.Start(); 55 | _sa_weight_backward_cuda(dw,query,key,dquery,dkey,B,T,C,H,W,radius,dilation); 56 | //timer.Stop(); 57 | //cudaDeviceSynchronize(); 58 | checkCudaErrors(cudaGetLastError()); 59 | //printf("Your code ran in: %f msecs.\n", timer.Elapsed()); 60 | } 61 | 62 | void sa_map_backward(const torch::Tensor& dout,const torch::Tensor& weight,const torch::Tensor& proj,torch::Tensor& dweight,torch::Tensor& dproj,int radius,int dilation){ 63 | int B,T,C,H,W; 64 | get_sizes(proj,&B,&T,&C,&H,&W); 65 | //GpuTimer timer; 66 | //timer.Start(); 67 | _sa_map_backward_cuda(dout,weight,proj,dweight,dproj,B,T,C,H,W,radius,dilation); 68 | //timer.Stop(); 69 | //cudaDeviceSynchronize(); 70 | checkCudaErrors(cudaGetLastError()); 71 | //printf("Your code ran in: %f msecs.\n", timer.Elapsed()); 72 | } 73 | 74 | void sa_weight_forward_ref(const torch::Tensor& query,const torch::Tensor& key,torch::Tensor& weight,int radius,int dilation){ 75 | int B,T,C,H,W; 76 | get_sizes(query,&B,&T,&C,&H,&W); 77 | sa_weight_forward_Ref(query,key,weight,B,T,C,H,W,radius,dilation); 78 | } 79 | 80 | void sa_weight_backward_ref(const torch::Tensor& dw,const torch::Tensor& query,const torch::Tensor& key,torch::Tensor& dquery,torch::Tensor& dkey,int radius,int dilation){ 81 | int B,T,C,H,W; 82 | get_sizes(query,&B,&T,&C,&H,&W); 83 | sa_weight_backward_query_Ref(dw,query,key,dquery,B,T,C,H,W,radius,dilation); 84 | sa_weight_backward_key_Ref(dw,query,key,dkey,B,T,C,H,W,radius,dilation); 85 | } 86 | 87 | void sa_map_forward_ref(const torch::Tensor& weight,const torch::Tensor& proj,torch::Tensor& out,int radius,int dilation){ 88 | int B,T,C,H,W; 89 | get_sizes(proj,&B,&T,&C,&H,&W); 90 | sa_map_forward_Ref(weight,proj,out,B,T,C,H,W,radius,dilation); 91 | } 92 | 93 | void sa_map_backward_ref(const torch::Tensor& dout,const torch::Tensor& weight,const torch::Tensor& proj,torch::Tensor& dweight,torch::Tensor& dproj,int radius,int dilation){ 94 | int B,T,C,H,W; 95 | get_sizes(proj,&B,&T,&C,&H,&W); 96 | sa_map_backward_weight_Ref(dout,weight,proj,dweight,B,T,C,H,W,radius,dilation); 97 | sa_map_backward_proj_Ref(dout,weight,proj,dproj,B,T,C,H,W,radius,dilation); 98 | } 99 | 100 | 101 | PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) { 102 | m.def("weight_forward", &sa_weight_forward, "weight forward (CUDA)"); 103 | m.def("weight_backward", &sa_weight_backward, "weight backward (CUDA)"); 104 | m.def("map_forward", &sa_map_forward, "map forward (CUDA)"); 105 | m.def("map_backward", &sa_map_backward, "map backward (CUDA)"); 106 | m.def("weight_forward_ref", &sa_weight_forward_ref, "weight forward ref (CUDA)"); 107 | m.def("weight_backward_ref", &sa_weight_backward_ref, "weight backward ref (CUDA)"); 108 | m.def("map_forward_ref", &sa_map_forward_ref, "map forward ref (CUDA)"); 109 | m.def("map_backward_ref", &sa_map_backward_ref, "map backward ref (CUDA)"); 110 | } 111 | /* 112 | int main() { 113 | //torch::Tensor weight=torch::ones({2,5,5*9,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 114 | //torch::Tensor query=torch::ones({2,5,8,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 115 | //torch::Tensor key=torch::ones({2,5,8,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 116 | //sa_weight_forward(query,key,weight); 117 | /* 118 | torch::Tensor weight=torch::ones({1,5,5*9,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 119 | torch::Tensor proj=torch::ones({1,5,8,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 120 | torch::Tensor out=torch::zeros({1,5,8,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 121 | sa_map_forward(weight,proj,out); 122 | 123 | /* 124 | torch::Tensor dw=torch::ones({1,5,5*9,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 125 | torch::Tensor query=torch::ones({1,5,8,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 126 | torch::Tensor key=torch::ones({1,5,8,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 127 | torch::Tensor dquery=torch::zeros({1,5,8,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 128 | torch::Tensor dkey=torch::zeros({1,5,8,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 129 | sa_weight_backward(dw,query,key,dquery,dkey); 130 | 131 | torch::Tensor dout=torch::ones({1,5,8,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 132 | 133 | torch::Tensor weight=torch::ones({1,5,5*9,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 134 | torch::Tensor proj=torch::ones({1,5,8,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 135 | torch::Tensor dweight=torch::zeros({1,5,5*9,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 136 | torch::Tensor dproj=torch::zeros({1,5,8,28,42}, at::kFloat).to(torch::Device(torch::kCUDA, 0)); 137 | sa_map_backward(dout,weight,proj,dweight,dproj); 138 | std::cout< 5 | 6 | struct GpuTimer 7 | { 8 | cudaEvent_t start; 9 | cudaEvent_t stop; 10 | 11 | GpuTimer() 12 | { 13 | cudaEventCreate(&start); 14 | cudaEventCreate(&stop); 15 | } 16 | 17 | ~GpuTimer() 18 | { 19 | cudaEventDestroy(start); 20 | cudaEventDestroy(stop); 21 | } 22 | 23 | void Start() 24 | { 25 | cudaEventRecord(start, 0); 26 | } 27 | 28 | void Stop() 29 | { 30 | cudaEventRecord(stop, 0); 31 | } 32 | 33 | float Elapsed() 34 | { 35 | float elapsed; 36 | cudaEventSynchronize(stop); 37 | cudaEventElapsedTime(&elapsed, start, stop); 38 | return elapsed; 39 | } 40 | }; 41 | 42 | #endif /* GPU_TIMER_H__ */ 43 | -------------------------------------------------------------------------------- /Models/PCSA/PCSA_Module/utils.h: -------------------------------------------------------------------------------- 1 | #ifndef UTILS_H__ 2 | #define UTILS_H__ 3 | 4 | #include 5 | #include 6 | #include 7 | #include 8 | #include 9 | #include 10 | #include 11 | 12 | #define checkCudaErrors(val) check( (val), #val, __FILE__, __LINE__) 13 | 14 | template 15 | void check(T err, const char* const func, const char* const file, const int line) { 16 | if (err != cudaSuccess) { 17 | std::cerr << "CUDA error at: " << file << ":" << line << std::endl; 18 | std::cerr << cudaGetErrorString(err) << " " << func << std::endl; 19 | exit(1); 20 | } 21 | } 22 | 23 | template 24 | void checkResultsExact(const T* const ref, const T* const gpu, size_t numElem) { 25 | //check that the GPU result matches the CPU result 26 | for (size_t i = 0; i < numElem; ++i) { 27 | if (ref[i] != gpu[i]) { 28 | std::cerr << "Difference at pos " << i << std::endl; 29 | //the + is magic to convert char to int without messing 30 | //with other types 31 | std::cerr << "Reference: " << std::setprecision(17) << +ref[i] << 32 | "\nGPU : " << +gpu[i] << std::endl; 33 | exit(1); 34 | } 35 | } 36 | } 37 | 38 | template 39 | void checkResultsEps(const T* const ref, const T* const gpu, size_t numElem, double eps1, double eps2) { 40 | assert(eps1 >= 0 && eps2 >= 0); 41 | unsigned long long totalDiff = 0; 42 | unsigned numSmallDifferences = 0; 43 | for (size_t i = 0; i < numElem; ++i) { 44 | //subtract smaller from larger in case of unsigned types 45 | T smaller = std::min(ref[i], gpu[i]); 46 | T larger = std::max(ref[i], gpu[i]); 47 | T diff = larger - smaller; 48 | if (diff > 0 && diff <= eps1) { 49 | numSmallDifferences++; 50 | } 51 | else if (diff > eps1) { 52 | std::cerr << "Difference at pos " << +i << " exceeds tolerance of " << eps1 << std::endl; 53 | std::cerr << "Reference: " << std::setprecision(17) << +ref[i] << 54 | "\nGPU : " << +gpu[i] << std::endl; 55 | exit(1); 56 | } 57 | totalDiff += diff * diff; 58 | } 59 | double percentSmallDifferences = (double)numSmallDifferences / (double)numElem; 60 | if (percentSmallDifferences > eps2) { 61 | std::cerr << "Total percentage of non-zero pixel difference between the two images exceeds " << 100.0 * eps2 << "%" << std::endl; 62 | std::cerr << "Percentage of non-zero pixel differences: " << 100.0 * percentSmallDifferences << "%" << std::endl; 63 | exit(1); 64 | } 65 | } 66 | 67 | //Uses the autodesk method of image comparison 68 | //Note the the tolerance here is in PIXELS not a percentage of input pixels 69 | template 70 | void checkResultsAutodesk(const T* const ref, const T* const gpu, size_t numElem, double variance, size_t tolerance) 71 | { 72 | 73 | size_t numBadPixels = 0; 74 | for (size_t i = 0; i < numElem; ++i) { 75 | T smaller = std::min(ref[i], gpu[i]); 76 | T larger = std::max(ref[i], gpu[i]); 77 | T diff = larger - smaller; 78 | if (diff > variance) 79 | ++numBadPixels; 80 | } 81 | 82 | if (numBadPixels > tolerance) { 83 | std::cerr << "Too many bad pixels in the image." << numBadPixels << "/" << tolerance << std::endl; 84 | exit(1); 85 | } 86 | } 87 | 88 | #endif 89 | -------------------------------------------------------------------------------- /Models/PCSA/build/lib.linux-x86_64-3.6/self_cuda_backend.cpython-36m-x86_64-linux-gnu.so: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/PCSA/build/lib.linux-x86_64-3.6/self_cuda_backend.cpython-36m-x86_64-linux-gnu.so -------------------------------------------------------------------------------- /Models/PCSA/build/temp.linux-x86_64-3.6/SelfAttention_Module/reference.o: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/PCSA/build/temp.linux-x86_64-3.6/SelfAttention_Module/reference.o -------------------------------------------------------------------------------- /Models/PCSA/build/temp.linux-x86_64-3.6/SelfAttention_Module/sa.o: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/PCSA/build/temp.linux-x86_64-3.6/SelfAttention_Module/sa.o -------------------------------------------------------------------------------- /Models/PCSA/build/temp.linux-x86_64-3.6/SelfAttention_Module/sa_ext.o: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/PCSA/build/temp.linux-x86_64-3.6/SelfAttention_Module/sa_ext.o -------------------------------------------------------------------------------- /Models/PCSA/dist/self_cuda-0.0.0-py3.6-linux-x86_64.egg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/PCSA/dist/self_cuda-0.0.0-py3.6-linux-x86_64.egg -------------------------------------------------------------------------------- /Models/PCSA/self_cuda.egg-info/PKG-INFO: -------------------------------------------------------------------------------- 1 | Metadata-Version: 1.0 2 | Name: self-cuda 3 | Version: 0.0.0 4 | Summary: UNKNOWN 5 | Home-page: UNKNOWN 6 | Author: UNKNOWN 7 | Author-email: UNKNOWN 8 | License: UNKNOWN 9 | Description-Content-Type: UNKNOWN 10 | Description: UNKNOWN 11 | Platform: UNKNOWN 12 | -------------------------------------------------------------------------------- /Models/PCSA/self_cuda.egg-info/SOURCES.txt: -------------------------------------------------------------------------------- 1 | setup.py 2 | SelfAttention_Module/reference.cpp 3 | SelfAttention_Module/sa.cu 4 | SelfAttention_Module/sa_ext.cpp 5 | self_cuda.egg-info/PKG-INFO 6 | self_cuda.egg-info/SOURCES.txt 7 | self_cuda.egg-info/dependency_links.txt 8 | self_cuda.egg-info/top_level.txt -------------------------------------------------------------------------------- /Models/PCSA/self_cuda.egg-info/dependency_links.txt: -------------------------------------------------------------------------------- 1 | 2 | -------------------------------------------------------------------------------- /Models/PCSA/self_cuda.egg-info/top_level.txt: -------------------------------------------------------------------------------- 1 | self_cuda_backend 2 | -------------------------------------------------------------------------------- /Models/PCSA/setup.py: -------------------------------------------------------------------------------- 1 | from setuptools import setup 2 | from torch.utils.cpp_extension import BuildExtension, CUDAExtension 3 | from os.path import join 4 | 5 | project_root = 'PCSA_Module' 6 | sources = [join(project_root, file) for file in ['sa_ext.cpp', 7 | 'sa.cu','reference.cpp']] 8 | ''' 9 | with open("README.md", "r") as fh: 10 | long_description = fh.read() 11 | ''' 12 | 13 | nvcc_args = [ 14 | '-gencode', 'arch=compute_61,code=sm_61', 15 | '-gencode', 'arch=compute_70,code=sm_70', 16 | '-gencode', 'arch=compute_70,code=compute_70' 17 | ] 18 | cxx_args = ['-std=c++11'] 19 | 20 | setup( 21 | name='self_cuda', 22 | #version="0.1.0", 23 | #author="Clément Pinard", 24 | #author_email="clement.pinard@ensta-paristech.fr", 25 | #description="Correlation module for pytorch", 26 | #long_description=long_description, 27 | #long_description_content_type="text/markdown", 28 | #url="https://github.com/ClementPinard/Pytorch-Correlation-extension", 29 | #install_requires=['torch>=1.0.1','numpy'], 30 | ext_modules=[ 31 | CUDAExtension('self_cuda_backend', 32 | sources, extra_compile_args={'cxx': cxx_args,'nvcc': nvcc_args}) 33 | ], 34 | #package_dir={'': project_root}, 35 | #packages=['spatial_correlation_sampler'], 36 | cmdclass={ 37 | 'build_ext': BuildExtension 38 | }) 39 | -------------------------------------------------------------------------------- /Models/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/__init__.py -------------------------------------------------------------------------------- /Models/lightrfb.py: -------------------------------------------------------------------------------- 1 | import torch 2 | import torch.nn as nn 3 | 4 | class h_sigmoid(nn.Module): 5 | def __init__(self, inplace=True): 6 | super(h_sigmoid, self).__init__() 7 | self.relu = nn.ReLU6(inplace=inplace) 8 | 9 | def forward(self, x): 10 | return self.relu(x + 3) / 6 11 | 12 | class SELayer(nn.Module): 13 | def __init__(self, channel, reduction=4): 14 | super(SELayer, self).__init__() 15 | self.avg_pool = nn.AdaptiveAvgPool2d(1) 16 | self.fc = nn.Sequential( 17 | nn.Linear(channel, channel // reduction), 18 | nn.ReLU(inplace=True), 19 | nn.Linear(channel // reduction, channel), 20 | h_sigmoid() 21 | ) 22 | 23 | def forward(self, x): 24 | b, c, _, _ = x.size() 25 | y = self.avg_pool(x).view(b, c) 26 | y = self.fc(y).view(b, c, 1, 1) 27 | return x * y 28 | 29 | class BasicConv(nn.Module): 30 | 31 | def __init__(self, in_planes, out_planes, kernel_size, stride=1, padding=0, dilation=1, groups=1, relu=True, bn=True, bias=False): 32 | super(BasicConv, self).__init__() 33 | self.out_channels = out_planes 34 | self.conv = nn.Conv2d(in_planes, out_planes, kernel_size=kernel_size, stride=stride, padding=padding, dilation=dilation, groups=groups, bias=bias) 35 | self.bn = nn.BatchNorm2d(out_planes,eps=1e-5, momentum=0.01, affine=True) if bn else None 36 | self.relu = nn.PReLU() if relu else None 37 | #self.relu = h_sigmoid() if relu else None 38 | 39 | def forward(self, x): 40 | x = self.conv(x) 41 | if self.bn is not None: 42 | x = self.bn(x) 43 | if self.relu is not None: 44 | x = self.relu(x) 45 | return x 46 | 47 | class LightRFB(nn.Module): 48 | def __init__(self, channels_in=160,channels_mid=128,channels_out=32): 49 | super(LightRFB, self).__init__() 50 | self.global_se=SELayer(channels_in) 51 | self.reduce=nn.Sequential(nn.Conv2d(channels_in,channels_mid,kernel_size=1,bias=False), 52 | nn.BatchNorm2d(channels_mid), 53 | nn.PReLU(channels_mid)) 54 | self.br0 = nn.Sequential( 55 | BasicConv(channels_mid, channels_mid, kernel_size=1,bias=False, 56 | bn=True, relu=True), 57 | BasicConv(channels_mid, channels_mid, kernel_size=3, dilation=1, padding=1, groups=channels_mid, bias=False, 58 | relu=False), 59 | ) 60 | self.br1 = nn.Sequential( 61 | BasicConv(channels_mid, channels_mid, kernel_size=3, dilation=1, padding=1, groups=channels_mid, bias=False,bn=True,relu=False), 62 | BasicConv(channels_mid, channels_mid, kernel_size=1, dilation=1, bias=False,bn=True,relu=True), 63 | 64 | BasicConv(channels_mid, channels_mid, kernel_size=3, dilation=3, padding=3, groups=channels_mid, bias=False, 65 | relu=False), 66 | ) 67 | self.br2 = nn.Sequential( 68 | BasicConv(channels_mid, channels_mid, kernel_size=5, dilation=1, padding=2, groups=channels_mid, bias=False, 69 | bn=True, relu=False), 70 | BasicConv(channels_mid, channels_mid, kernel_size=1, dilation=1, bias=False, bn=True, relu=True), 71 | 72 | BasicConv(channels_mid, channels_mid, kernel_size=3, dilation=5, padding=5, groups=channels_mid, bias=False, 73 | relu=False), 74 | ) 75 | self.br3 = nn.Sequential( 76 | BasicConv(channels_mid, channels_mid, kernel_size=7, dilation=1, padding=3, groups=channels_mid, bias=False, 77 | bn=True, relu=False), 78 | BasicConv(channels_mid, channels_mid, kernel_size=1, dilation=1, bias=False, bn=True, relu=True), 79 | 80 | BasicConv(channels_mid, channels_mid, kernel_size=3, dilation=7, padding=7, groups=channels_mid, bias=False, 81 | relu=False), 82 | ) 83 | self.point_global=BasicConv(channels_mid*4+channels_in,channels_out,kernel_size=1,bias=False,bn=True, relu=True) 84 | 85 | def forward(self, x): 86 | x_reduce=self.reduce(self.global_se(x)) 87 | x0=self.br0(x_reduce) 88 | x1=self.br1(x_reduce) 89 | x2=self.br2(x_reduce) 90 | x3=self.br3(x_reduce) 91 | out=self.point_global(torch.cat([x,x0,x1,x2,x3],dim=1)) 92 | return out 93 | 94 | 95 | 96 | 97 | if __name__=="__main__": 98 | m=GycModule(196,128,32) 99 | t=torch.zeros(1,196,14,14) 100 | print(m(t).shape) 101 | -------------------------------------------------------------------------------- /Models/mobilenetv3_pretrain.py: -------------------------------------------------------------------------------- 1 | import torch.nn as nn 2 | import torch 3 | import numpy as np 4 | import math 5 | import torch.nn.functional as F 6 | from config import config 7 | from Models.lightrfb import LightRFB 8 | __all__ = ['mobilenetv3_large'] 9 | 10 | 11 | def _make_divisible(v, divisor, min_value=None): 12 | """ 13 | This function is taken from the original tf repo. 14 | It ensures that all layers have a channel number that is divisible by 8 15 | It can be seen here: 16 | https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet/mobilenet.py 17 | :param v: 18 | :param divisor: 19 | :param min_value: 20 | :return: 21 | """ 22 | if min_value is None: 23 | min_value = divisor 24 | new_v = max(min_value, int(v + divisor / 2) // divisor * divisor) 25 | # Make sure that round down does not go down by more than 10%. 26 | if new_v < 0.9 * v: 27 | new_v += divisor 28 | return new_v 29 | 30 | 31 | class h_sigmoid(nn.Module): 32 | def __init__(self, inplace=True): 33 | super(h_sigmoid, self).__init__() 34 | self.relu = nn.ReLU6(inplace=inplace) 35 | 36 | def forward(self, x): 37 | return self.relu(x + 3) / 6 38 | 39 | 40 | class h_swish(nn.Module): 41 | def __init__(self, inplace=True): 42 | super(h_swish, self).__init__() 43 | self.sigmoid = h_sigmoid(inplace=inplace) 44 | 45 | def forward(self, x): 46 | return x * self.sigmoid(x) 47 | 48 | 49 | class SELayer(nn.Module): 50 | def __init__(self, channel, reduction=4): 51 | super(SELayer, self).__init__() 52 | self.avg_pool = nn.AdaptiveAvgPool2d(1) 53 | self.fc = nn.Sequential( 54 | nn.Linear(channel, channel // reduction), 55 | nn.ReLU(inplace=True), 56 | nn.Linear(channel // reduction, channel), 57 | h_sigmoid() 58 | ) 59 | 60 | def forward(self, x): 61 | b, c, _, _ = x.size() 62 | y = self.avg_pool(x).view(b, c) 63 | y = self.fc(y).view(b, c, 1, 1) 64 | return x * y 65 | 66 | 67 | def conv_3x3_bn(inp, oup, stride): 68 | return nn.Sequential( 69 | nn.Conv2d(inp, oup, 3, stride, 1, bias=False), 70 | nn.BatchNorm2d(oup), 71 | h_swish() 72 | ) 73 | 74 | 75 | def conv_1x1_bn(inp, oup): 76 | return nn.Sequential( 77 | nn.Conv2d(inp, oup, 1, 1, 0, bias=False), 78 | nn.BatchNorm2d(oup), 79 | h_swish() 80 | ) 81 | 82 | 83 | class InvertedResidual(nn.Module): 84 | def __init__(self, inp, hidden_dim, oup, kernel_size, stride, use_se, use_hs): 85 | super(InvertedResidual, self).__init__() 86 | assert stride in [1, 2] 87 | 88 | self.identity = stride == 1 and inp == oup 89 | 90 | if inp == hidden_dim: 91 | self.conv = nn.Sequential( 92 | # dw 93 | nn.Conv2d(hidden_dim, hidden_dim, kernel_size, stride, (kernel_size - 1) // 2, groups=hidden_dim, bias=False), 94 | nn.BatchNorm2d(hidden_dim), 95 | h_swish() if use_hs else nn.ReLU(inplace=True), 96 | # Squeeze-and-Excite 97 | SELayer(hidden_dim) if use_se else nn.Sequential(), 98 | # pw-linear 99 | nn.Conv2d(hidden_dim, oup, 1, 1, 0, bias=False), 100 | nn.BatchNorm2d(oup), 101 | ) 102 | else: 103 | self.conv = nn.Sequential( 104 | # pw 105 | nn.Conv2d(inp, hidden_dim, 1, 1, 0, bias=False), 106 | nn.BatchNorm2d(hidden_dim), 107 | h_swish() if use_hs else nn.ReLU(inplace=True), 108 | # dw 109 | nn.Conv2d(hidden_dim, hidden_dim, kernel_size, stride, (kernel_size - 1) // 2, groups=hidden_dim, bias=False), 110 | nn.BatchNorm2d(hidden_dim), 111 | # Squeeze-and-Excite 112 | SELayer(hidden_dim) if use_se else nn.Sequential(), 113 | h_swish() if use_hs else nn.ReLU(inplace=True), 114 | # pw-linear 115 | nn.Conv2d(hidden_dim, oup, 1, 1, 0, bias=False), 116 | nn.BatchNorm2d(oup), 117 | ) 118 | 119 | def forward(self, x): 120 | if self.identity: 121 | return x + self.conv(x) 122 | else: 123 | return self.conv(x) 124 | 125 | 126 | class MobileNetV3(nn.Module): 127 | def __init__(self, cfgs, mode,width_mult=1.): 128 | super(MobileNetV3, self).__init__() 129 | # setting of inverted residual blocks 130 | self.cfgs = cfgs 131 | assert mode in ['large', 'small'] 132 | 133 | # building first layer 134 | input_channel = _make_divisible(16 * width_mult, 8) 135 | layers = [conv_3x3_bn(3, input_channel, 2)] 136 | # building inverted residual blocks 137 | block = InvertedResidual 138 | for k, exp_size, c, use_se, use_hs, s in self.cfgs: 139 | output_channel = _make_divisible(c * width_mult, 8) 140 | layers.append(block(input_channel, exp_size, output_channel, k, s, use_se, use_hs)) 141 | input_channel = output_channel 142 | self.features = nn.Sequential(*layers) 143 | self.load_backbone( 144 | torch.load(config.backbone_imagenet_pretrain,map_location=torch.device('cpu'))) 145 | 146 | def forward(self, x): 147 | x = self.features[:4](x) 148 | low_level=x 149 | high_level = self.features[4:](x) 150 | return low_level,high_level 151 | 152 | def load_backbone(self,pretrained_dict): 153 | # step2: get model state_dict 154 | model_dict = self.state_dict() 155 | # step3: remove pretrained_dict params which is not in model_dict 156 | pretrained_dict = {k: v for k, v in pretrained_dict.items() if (k in model_dict)} 157 | # step4: update model_dict using pretrained_dict 158 | model_dict.update(pretrained_dict) 159 | # step5: update model using model_dict 160 | self.load_state_dict(model_dict) 161 | 162 | def _initialize_weights(self): 163 | for m in self.modules(): 164 | if isinstance(m, nn.Conv2d): 165 | n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels 166 | m.weight.data.normal_(0, math.sqrt(2. / n)) 167 | if m.bias is not None: 168 | m.bias.data.zero_() 169 | elif isinstance(m, nn.BatchNorm2d): 170 | m.weight.data.fill_(1) 171 | m.bias.data.zero_() 172 | elif isinstance(m, nn.Linear): 173 | n = m.weight.size(1) 174 | m.weight.data.normal_(0, 0.01) 175 | m.bias.data.zero_() 176 | 177 | class conbine_feature(nn.Module): 178 | def __init__(self): 179 | super(conbine_feature, self).__init__() 180 | self.up2_high = DilatedParallelConvBlockD2(32, 16) 181 | self.up2_low = nn.Conv2d(24, 16, 1, stride=1, padding=0,bias=False) 182 | self.up2_bn2 = nn.BatchNorm2d(16) 183 | self.up2_act = nn.PReLU(16) 184 | self.refine=nn.Sequential(nn.Conv2d(16,16,3,padding=1,bias=False),nn.BatchNorm2d(16),nn.PReLU()) 185 | 186 | def forward(self, low_fea,high_fea): 187 | high_fea = self.up2_high(high_fea) 188 | low_fea = self.up2_bn2(self.up2_low(low_fea)) 189 | refine_feature = self.refine(self.up2_act(high_fea+low_fea)) 190 | return refine_feature 191 | 192 | 193 | class DilatedParallelConvBlockD2(nn.Module): 194 | def __init__(self, nIn, nOut, add=False): 195 | super(DilatedParallelConvBlockD2, self).__init__() 196 | n = int(np.ceil(nOut / 2.)) 197 | n2 = nOut - n 198 | 199 | self.conv0 = nn.Conv2d(nIn, nOut, 1, stride=1, padding=0, dilation=1, bias=False) 200 | self.conv1 = nn.Conv2d(n, n, 3, stride=1, padding=1, dilation=1, bias=False) 201 | self.conv2 = nn.Conv2d(n2, n2, 3, stride=1, padding=2, dilation=2, bias=False) 202 | 203 | self.bn = nn.BatchNorm2d(nOut) 204 | #self.act = nn.PReLU(nOut) 205 | self.add = add 206 | 207 | def forward(self, input): 208 | in0 = self.conv0(input) 209 | in1, in2 = torch.chunk(in0, 2, dim=1) 210 | b1 = self.conv1(in1) 211 | b2 = self.conv2(in2) 212 | output = torch.cat([b1, b2], dim=1) 213 | 214 | if self.add: 215 | output = input + output 216 | output = self.bn(output) 217 | #output = self.act(output) 218 | 219 | return output 220 | 221 | class Fastnet(nn.Module): 222 | def __init__(self): 223 | super(Fastnet, self).__init__() 224 | self.feature_extractor=mobilenetv3_large() 225 | self.dilation_conv_group=LightRFB() 226 | self.combine=conbine_feature() 227 | self.SegNIN = nn.Sequential(nn.Dropout2d(0.1),nn.Conv2d(16, 1, kernel_size=1,bias=False)) 228 | 229 | def forward(self, x): 230 | if len(x.shape)==4: 231 | origin_shape=x.shape 232 | low_feature,high_feature=self.feature_extractor(x) 233 | high_feature=self.dilation_conv_group(high_feature) 234 | high_feature = F.interpolate(high_feature, size=(low_feature.shape[-2], low_feature.shape[-1]), 235 | mode="bilinear", 236 | align_corners=False) 237 | out = self.combine(low_feature, high_feature) 238 | out = torch.sigmoid(F.interpolate(self.SegNIN(out), size=(origin_shape[-2], origin_shape[-1]), mode="bilinear", 239 | align_corners=False)) 240 | else: 241 | origin_shape=x.shape 242 | x=x.view(-1,*origin_shape[2:]) 243 | low_feature,high_feature=self.feature_extractor(x) 244 | high_feature=self.dilation_conv_group(high_feature) 245 | high_feature = F.interpolate(high_feature, size=(low_feature.shape[-2], low_feature.shape[-1]), 246 | mode="bilinear", 247 | align_corners=False) 248 | out = self.combine(low_feature, high_feature) 249 | out = torch.sigmoid( 250 | F.interpolate(self.SegNIN(out), size=(origin_shape[-2], origin_shape[-1]), mode="bilinear", 251 | align_corners=False)) 252 | return out 253 | 254 | 255 | def mobilenetv3_large(**kwargs): 256 | """ 257 | Constructs a MobileNetV3-Large model 258 | """ 259 | cfgs = [ 260 | # k, t, c, SE, NL, s 261 | [3, 16, 16, 0, 0, 1], 262 | [3, 64, 24, 0, 0, 2], 263 | [3, 72, 24, 0, 0, 1], 264 | [5, 72, 40, 1, 0, 1], 265 | [5, 120, 40, 1, 0, 1], 266 | [5, 120, 40, 1, 0, 1], 267 | [3, 240, 80, 0, 1, 2], 268 | [3, 200, 80, 0, 1, 1], 269 | [3, 184, 80, 0, 1, 1], 270 | [3, 184, 80, 0, 1, 1], 271 | [3, 480, 112, 1, 1, 1], 272 | [3, 672, 112, 1, 1, 1], 273 | [5, 672, 160, 1, 1, 1], 274 | [5, 672, 160, 1, 1, 1], 275 | [5, 960, 160, 1, 1, 1] 276 | ] 277 | return MobileNetV3(cfgs, mode='large', **kwargs) 278 | 279 | if __name__=="__main__": 280 | from torchstat import stat 281 | a=torch.zeros(1,3,224,336) 282 | mobile=Fastnet() 283 | mobile(a) 284 | dsa 285 | 286 | 287 | print(mobile(a)[0].shape) 288 | total_paramters = sum([np.prod(p.size()) for p in mobile.parameters()]) 289 | print('Total network parameters: ' + str(total_paramters)) -------------------------------------------------------------------------------- /Models/mobilenetv3temporal_PCSA.py: -------------------------------------------------------------------------------- 1 | import torch.nn as nn 2 | import torch 3 | import numpy as np 4 | import math 5 | import torch.nn.functional as F 6 | from config import config 7 | from Models.lightrfb import LightRFB 8 | from Models.PCSA import T_Moduel 9 | __all__ = ['mobilenetv3_large'] 10 | 11 | 12 | def _make_divisible(v, divisor, min_value=None): 13 | """ 14 | This function is taken from the original tf repo. 15 | It ensures that all layers have a channel number that is divisible by 8 16 | It can be seen here: 17 | https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet/mobilenet.py 18 | :param v: 19 | :param divisor: 20 | :param min_value: 21 | :return: 22 | """ 23 | if min_value is None: 24 | min_value = divisor 25 | new_v = max(min_value, int(v + divisor / 2) // divisor * divisor) 26 | # Make sure that round down does not go down by more than 10%. 27 | if new_v < 0.9 * v: 28 | new_v += divisor 29 | return new_v 30 | 31 | 32 | class h_sigmoid(nn.Module): 33 | def __init__(self, inplace=True): 34 | super(h_sigmoid, self).__init__() 35 | self.relu = nn.ReLU6(inplace=inplace) 36 | 37 | def forward(self, x): 38 | return self.relu(x + 3) / 6 39 | 40 | 41 | class h_swish(nn.Module): 42 | def __init__(self, inplace=True): 43 | super(h_swish, self).__init__() 44 | self.sigmoid = h_sigmoid(inplace=inplace) 45 | 46 | def forward(self, x): 47 | return x * self.sigmoid(x) 48 | 49 | 50 | class SELayer(nn.Module): 51 | def __init__(self, channel, reduction=4): 52 | super(SELayer, self).__init__() 53 | self.avg_pool = nn.AdaptiveAvgPool2d(1) 54 | self.fc = nn.Sequential( 55 | nn.Linear(channel, channel // reduction), 56 | nn.ReLU(inplace=True), 57 | nn.Linear(channel // reduction, channel), 58 | h_sigmoid() 59 | ) 60 | 61 | def forward(self, x): 62 | b, c, _, _ = x.size() 63 | y = self.avg_pool(x).view(b, c) 64 | y = self.fc(y).view(b, c, 1, 1) 65 | return x * y 66 | 67 | 68 | def conv_3x3_bn(inp, oup, stride): 69 | return nn.Sequential( 70 | nn.Conv2d(inp, oup, 3, stride, 1, bias=False), 71 | nn.BatchNorm2d(oup), 72 | h_swish() 73 | ) 74 | 75 | 76 | def conv_1x1_bn(inp, oup): 77 | return nn.Sequential( 78 | nn.Conv2d(inp, oup, 1, 1, 0, bias=False), 79 | nn.BatchNorm2d(oup), 80 | h_swish() 81 | ) 82 | 83 | 84 | class InvertedResidual(nn.Module): 85 | def __init__(self, inp, hidden_dim, oup, kernel_size, stride, use_se, use_hs): 86 | super(InvertedResidual, self).__init__() 87 | assert stride in [1, 2] 88 | 89 | self.identity = stride == 1 and inp == oup 90 | 91 | if inp == hidden_dim: 92 | self.conv = nn.Sequential( 93 | # dw 94 | nn.Conv2d(hidden_dim, hidden_dim, kernel_size, stride, (kernel_size - 1) // 2, groups=hidden_dim, bias=False), 95 | nn.BatchNorm2d(hidden_dim), 96 | h_swish() if use_hs else nn.ReLU(inplace=True), 97 | # Squeeze-and-Excite 98 | SELayer(hidden_dim) if use_se else nn.Sequential(), 99 | # pw-linear 100 | nn.Conv2d(hidden_dim, oup, 1, 1, 0, bias=False), 101 | nn.BatchNorm2d(oup), 102 | ) 103 | else: 104 | self.conv = nn.Sequential( 105 | # pw 106 | nn.Conv2d(inp, hidden_dim, 1, 1, 0, bias=False), 107 | nn.BatchNorm2d(hidden_dim), 108 | h_swish() if use_hs else nn.ReLU(inplace=True), 109 | # dw 110 | nn.Conv2d(hidden_dim, hidden_dim, kernel_size, stride, (kernel_size - 1) // 2, groups=hidden_dim, bias=False), 111 | nn.BatchNorm2d(hidden_dim), 112 | # Squeeze-and-Excite 113 | SELayer(hidden_dim) if use_se else nn.Sequential(), 114 | h_swish() if use_hs else nn.ReLU(inplace=True), 115 | # pw-linear 116 | nn.Conv2d(hidden_dim, oup, 1, 1, 0, bias=False), 117 | nn.BatchNorm2d(oup), 118 | ) 119 | 120 | def forward(self, x): 121 | if self.identity: 122 | return x + self.conv(x) 123 | else: 124 | return self.conv(x) 125 | 126 | 127 | class MobileNetV3(nn.Module): 128 | def __init__(self, cfgs, mode,width_mult=1.): 129 | super(MobileNetV3, self).__init__() 130 | # setting of inverted residual blocks 131 | self.cfgs = cfgs 132 | assert mode in ['large', 'small'] 133 | 134 | # building first layer 135 | input_channel = _make_divisible(16 * width_mult, 8) 136 | layers = [conv_3x3_bn(3, input_channel, 2)] 137 | # building inverted residual blocks 138 | block = InvertedResidual 139 | for k, exp_size, c, use_se, use_hs, s in self.cfgs: 140 | output_channel = _make_divisible(c * width_mult, 8) 141 | layers.append(block(input_channel, exp_size, output_channel, k, s, use_se, use_hs)) 142 | input_channel = output_channel 143 | self.features = nn.Sequential(*layers) 144 | self.load_backbone( 145 | torch.load(config.backbone_imagenet_pretrain,map_location=torch.device('cpu'))) 146 | 147 | def forward(self, x): 148 | x = self.features[:4](x) 149 | low_level = x 150 | high_level = self.features[4:](x) 151 | return low_level, high_level 152 | 153 | def load_backbone(self,pretrained_dict): 154 | # step2: get model state_dict 155 | model_dict = self.state_dict() 156 | # step3: remove pretrained_dict params which is not in model_dict 157 | pretrained_dict = {k: v for k, v in pretrained_dict.items() if (k in model_dict)} 158 | 159 | # step4: update model_dict using pretrained_dict 160 | model_dict.update(pretrained_dict) 161 | # step5: update model using model_dict 162 | self.load_state_dict(model_dict) 163 | 164 | def _initialize_weights(self): 165 | for m in self.modules(): 166 | if isinstance(m, nn.Conv2d): 167 | n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels 168 | m.weight.data.normal_(0, math.sqrt(2. / n)) 169 | if m.bias is not None: 170 | m.bias.data.zero_() 171 | elif isinstance(m, nn.BatchNorm2d): 172 | m.weight.data.fill_(1) 173 | m.bias.data.zero_() 174 | elif isinstance(m, nn.Linear): 175 | n = m.weight.size(1) 176 | m.weight.data.normal_(0, 0.01) 177 | m.bias.data.zero_() 178 | 179 | class conbine_feature(nn.Module): 180 | def __init__(self): 181 | super(conbine_feature, self).__init__() 182 | self.up2_high = DilatedParallelConvBlockD2(32, 16) 183 | self.up2_low = nn.Conv2d(24, 16, 1, stride=1, padding=0,bias=False) 184 | self.up2_bn2 = nn.BatchNorm2d(16) 185 | self.up2_act = nn.PReLU(16) 186 | self.refine=nn.Sequential(nn.Conv2d(16,16,3,padding=1,bias=False),nn.BatchNorm2d(16),nn.PReLU()) 187 | 188 | def forward(self, low_fea,high_fea): 189 | high_fea = self.up2_high(high_fea) 190 | low_fea = self.up2_bn2(self.up2_low(low_fea)) 191 | refine_feature = self.refine(self.up2_act(high_fea+low_fea)) 192 | return refine_feature 193 | 194 | 195 | class DilatedParallelConvBlockD2(nn.Module): 196 | def __init__(self, nIn, nOut, add=False): 197 | super(DilatedParallelConvBlockD2, self).__init__() 198 | n = int(np.ceil(nOut / 2.)) 199 | n2 = nOut - n 200 | 201 | self.conv0 = nn.Conv2d(nIn, nOut, 1, stride=1, padding=0, dilation=1, bias=False) 202 | self.conv1 = nn.Conv2d(n, n, 3, stride=1, padding=1, dilation=1, bias=False) 203 | self.conv2 = nn.Conv2d(n2, n2, 3, stride=1, padding=2, dilation=2, bias=False) 204 | 205 | self.bn = nn.BatchNorm2d(nOut) 206 | #self.act = nn.PReLU(nOut) 207 | self.add = add 208 | 209 | def forward(self, input): 210 | in0 = self.conv0(input) 211 | in1, in2 = torch.chunk(in0, 2, dim=1) 212 | b1 = self.conv1(in1) 213 | b2 = self.conv2(in2) 214 | output = torch.cat([b1, b2], dim=1) 215 | 216 | if self.add: 217 | output = input + output 218 | output = self.bn(output) 219 | #output = self.act(output) 220 | 221 | return output 222 | 223 | class Fastnet(nn.Module): 224 | def __init__(self): 225 | super(Fastnet, self).__init__() 226 | self.feature_extractor=mobilenetv3_large() 227 | self.dilation_conv_group=LightRFB() 228 | self.combine=conbine_feature() 229 | self.SegNIN = nn.Sequential(nn.Dropout2d(0.1),nn.Conv2d(16, 1, kernel_size=1,bias=False)) 230 | self.temporal_high=T_Moduel(32) 231 | 232 | def load_backbone(self,pretrained_dict,logger): 233 | # step2: get model state_dict 234 | model_dict = self.state_dict() 235 | # step3: remove pretrained_dict params which is not in model_dict 236 | logger.info("load_state_dict!!!") 237 | for k, v in pretrained_dict.items(): 238 | if (k in model_dict): 239 | logger.info("load:%s"%k) 240 | else: 241 | logger.info("jump over:%s"%k) 242 | 243 | pretrained_dict = {k: v for k, v in pretrained_dict.items() if (k in model_dict)} 244 | 245 | # step4: update model_dict using pretrained_dict 246 | model_dict.update(pretrained_dict) 247 | # step5: update model using model_dict 248 | self.load_state_dict(model_dict) 249 | 250 | def freezen_feature_extractor(self,logger): 251 | logger.info("freeze feature extractor!!!!!!") 252 | for params in self.feature_extractor.parameters(): 253 | params.requires_grad=False 254 | 255 | def forward(self, x): 256 | if len(x.shape)==4: 257 | origin_shape=x.shape 258 | low_feature, high_feature = self.feature_extractor(x) 259 | high_feature=self.dilation_conv_group(high_feature) 260 | high_feature = F.interpolate(high_feature, size=(low_feature.shape[-2], low_feature.shape[-1]), 261 | mode="bilinear", 262 | align_corners=False) 263 | out = self.combine(low_feature, high_feature) 264 | out = torch.sigmoid(F.interpolate(self.SegNIN(out), size=(origin_shape[-2], origin_shape[-1]), mode="bilinear", 265 | align_corners=False)) 266 | else: 267 | origin_shape = x.shape 268 | x = x.view(-1, *origin_shape[2:]) 269 | low_feature, high_feature = self.feature_extractor(x) 270 | high_feature = self.dilation_conv_group(high_feature) 271 | high_feature=high_feature.view(*origin_shape[:2],*high_feature.shape[1:]) 272 | high_feature=self.temporal_high(high_feature) 273 | high_feature = high_feature.view(-1, *high_feature.shape[2:]) 274 | 275 | high_feature = F.interpolate(high_feature, size=(low_feature.shape[-2], low_feature.shape[-1]), 276 | mode="bilinear", 277 | align_corners=False) 278 | 279 | out = self.combine(low_feature, high_feature) 280 | out = torch.sigmoid( 281 | F.interpolate(self.SegNIN(out), size=(origin_shape[-2], origin_shape[-1]), mode="bilinear", 282 | align_corners=False)) 283 | return out 284 | 285 | 286 | def mobilenetv3_large(**kwargs): 287 | """ 288 | Constructs a MobileNetV3-Large model 289 | """ 290 | cfgs = [ 291 | # k, t, c, SE, NL, s 292 | [3, 16, 16, 0, 0, 1], 293 | [3, 64, 24, 0, 0, 2], 294 | [3, 72, 24, 0, 0, 1], 295 | [5, 72, 40, 1, 0, 1], 296 | [5, 120, 40, 1, 0, 1], 297 | [5, 120, 40, 1, 0, 1], 298 | [3, 240, 80, 0, 1, 2], 299 | [3, 200, 80, 0, 1, 1], 300 | [3, 184, 80, 0, 1, 1], 301 | [3, 184, 80, 0, 1, 1], 302 | [3, 480, 112, 1, 1, 1], 303 | [3, 672, 112, 1, 1, 1], 304 | [5, 672, 160, 1, 1, 1], 305 | [5, 672, 160, 1, 1, 1], 306 | [5, 960, 160, 1, 1, 1] 307 | ] 308 | return MobileNetV3(cfgs, mode='large', **kwargs) 309 | 310 | if __name__=="__main__": 311 | a=torch.zeros(1,5,3,224,336).cuda() 312 | mobile=Fastnet().cuda() 313 | print(mobile(a)[0].shape) 314 | -------------------------------------------------------------------------------- /Models/statedict/mobilenetv3-large.pth: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/guyuchao/PyramidCSA/45025dbfb9e95b832be8a82de281eadf9a2c2e5c/Models/statedict/mobilenetv3-large.pth -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # PyramidCSA 2 | 3 | Code for "Pyramid Constrained Self-Attention Network for Fast Video Salient Object Detection" (AAAI 2020) 4 | 5 | ## Build 6 | 7 | ```bash 8 | conda create -n PCSA python=3.6 9 | conda activate PCSA 10 | conda install pytorch=1.1.0 torchvision -c pytorch 11 | pip install tensorboardX tqdm Pillow==6.2.2 12 | pip install git+https://github.com/pytorch/tnt.git@master 13 | cd Models/PCSA 14 | python setup.py build develop 15 | ``` 16 | 17 | ## Training 18 | 19 | ### pretrain phase 20 | ```bash 21 | bash pretrain.sh 22 | ``` 23 | ### finetune phase 24 | ```bash 25 | bash finetune.sh 26 | ``` 27 | 28 | ## Results 29 | The result saliency map and model can be downloaded [baidu pan](https://pan.baidu.com/s/1bktiBwBUprIpfstK9fDehg) (password t781), or [google drive](https://drive.google.com/drive/folders/1Xe2Eob173M6e6dcapThrOD0rmnuDA2MN?usp=sharing). 30 | 31 | ## Evaluation 32 | For VSOD, we use the evaluation code provided by [DAVSOD](https://github.com/DengPingFan/DAVSOD). 33 | 34 | For UVOS, we use the evaluation code provided by [Davis16](https://github.com/fperazzi/davis). 35 | 36 | ## Speed Evaluation 37 | ```python3 38 | python speed.py 39 | ``` 40 | 41 | ## Cite 42 | If you think this work is helpful, please cite 43 | ```latex 44 | @inproceedings{gu2020PCSA, 45 | title={Pyramid Constrained Self-Attention Network for Fast Video Salient Object Detection}, 46 | author={Gu, Yuchao and Wang, Lijuan and Wang, Ziqin and Liu, Yun and Cheng, Ming-Ming and Lu, Shao-Ping}, 47 | booktitle={Proceedings of the AAAI Conference on Artificial Intelligence}, 48 | year={2020}, 49 | } 50 | ``` 51 | 52 | ## License 53 | This project is licensed under the [Creative Commons NonCommercial (CC BY-NC 3.0)](https://creativecommons.org/licenses/by-nc/3.0/) license where only 54 | non-commercial usage is allowed. For commercial usage, please contact us. 55 | 56 | ## Related Project 57 | The feature extraction backbone is borrowed from [d-li14/mobilenetv3.pytorch](https://github.com/d-li14/mobilenetv3.pytorch) 58 | 59 | ## Contact 60 | Any questions and suggestions, please email [ycgu@mail.nankai.edu.cn](ycgu@mail.nankai.edu.cn). 61 | -------------------------------------------------------------------------------- /build.sh: -------------------------------------------------------------------------------- 1 | conda create -n PCSA python=3.6 2 | conda activate PCSA 3 | conda install pytorch=1.1.0 torchvision -c pytorch 4 | pip install tensorboardX tqdm Pillow==6.2.2 5 | pip install git+https://github.com/pytorch/tnt.git@master 6 | cd Models/PCSA 7 | python setup.py build develop 8 | -------------------------------------------------------------------------------- /config.py: -------------------------------------------------------------------------------- 1 | import argparse 2 | parser = argparse.ArgumentParser() 3 | #optimizer 4 | parser.add_argument('--lr_mode', type=str, default="poly") 5 | parser.add_argument('--base_lr', type=float, default=1e-4) 6 | parser.add_argument('--finetune_lr', type=float, default=1e-6) 7 | 8 | #train schedule 9 | parser.add_argument('--pretrain_epoches', type=int, default=15) 10 | parser.add_argument('--finetune_epoches', type=int, default=15) 11 | parser.add_argument('--log_inteval', type=int, default=50) 12 | 13 | 14 | ##data 15 | parser.add_argument('--data_statistics', type=str, default="Data/['DUTS-TR']_statistics.pth") 16 | parser.add_argument('--img_dataset_list', type=str, default=["DUTS-TR"]) 17 | parser.add_argument('--video_dataset_list', type=str, default=["DAVIS","DAVSOD"]) 18 | parser.add_argument('--img_dataset_root', type=str,default="/media/data/guyuchao/dataset/fastsaliency") 19 | parser.add_argument('--video_dataset_root', type=str,default="/media/data/guyuchao/dataset/saliency/trainDataset") 20 | parser.add_argument('--size', type=tuple,default=(256,448)) 21 | parser.add_argument('--pretrain_batchsize', type=int, default=24) 22 | parser.add_argument('--video_batchsize', type=int, default=12) 23 | parser.add_argument('--video_time_clips', type=int, default=5) 24 | parser.add_argument('--video_testset_root', type=str,default="/media/data/guyuchao/dataset/saliency/testDataset") 25 | parser.add_argument('--parallel', type=bool, default=True) 26 | parser.add_argument('--device_idxs', type=list, default=[0,1,2,3]) 27 | parser.add_argument('--local_rank', type=int,default=0) 28 | 29 | #pretrain 30 | parser.add_argument('--pretrain_state_dict', type=str, default="/media/data/guyuchao/project/released/videofastsal/checkpoints/tensorboard/pretrain_baseline/epoch_15_batch_0/checkpoint.pth") 31 | parser.add_argument('--backbone_imagenet_pretrain', type=str, default="Models/statedict/mobilenetv3-large.pth") 32 | 33 | config = parser.parse_args() 34 | -------------------------------------------------------------------------------- /finetune.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | source /home/ycgu/anaconda3/bin/activate pt10py36 4 | 5 | GPU_IDS=0,1,2,3 6 | NUM_THREADS=2 7 | CUDA_VISIBLE_DEVICES=${GPU_IDS} OMP_NUM_THREADS=${NUM_THREADS} python -m torch.distributed.launch --nproc_per_node=4 --master_port=1234 finetune_temporal_distribute.py 8 | -------------------------------------------------------------------------------- /finetune_temporal_distribute.py: -------------------------------------------------------------------------------- 1 | from torch.optim import SGD,Adam 2 | import torch 3 | from Data.dataloader import get_video_dataset 4 | from analysis.evaluate import AutoEvaluate 5 | from analysis.onlinetest import AutoTest 6 | import torch.nn.functional as F 7 | from utils.utils import get_Logger_and_SummaryWriter 8 | import os 9 | from utils.Distribute.engine import Engine 10 | from config import config 11 | from Models.mobilenetv3temporal_PCSA import Fastnet 12 | import torch.nn as nn 13 | from utils.SalEval import SalEval 14 | from torchnet.meter import AverageValueMeter 15 | import numpy as np 16 | class CrossEntropyLoss(nn.Module): 17 | def __init__(self): 18 | super(CrossEntropyLoss, self).__init__() 19 | 20 | def forward(self, *inputs): 21 | pred, target = tuple(inputs) 22 | total_loss = F.binary_cross_entropy(pred, target.float()) 23 | return total_loss 24 | 25 | 26 | class TrainSchedule(object): 27 | def __init__(self, batches_per_epoch): 28 | self.cur_epoch = 0 29 | self.total_epoch = config.finetune_epoches 30 | self.cur_batches = 0 31 | self.batches_per_epoch = batches_per_epoch 32 | 33 | def update(self): 34 | self.cur_batches += 1 35 | if not self.cur_batches < self.batches_per_epoch: 36 | self.cur_batches = 0 37 | self.cur_epoch += 1 38 | 39 | def state_dict(self): 40 | state_dict = {"cur_batches": self.cur_batches, 41 | "cur_epoch": self.cur_epoch, 42 | "total_epoch": self.total_epoch, 43 | "batches_per_epoch": self.batches_per_epoch} 44 | return state_dict 45 | 46 | def load_state_dict(self, state_dict): 47 | self.cur_batches = state_dict["cur_batches"] 48 | self.cur_epoch = state_dict["cur_epoch"] 49 | self.total_epoch = state_dict["total_epoch"] 50 | self.batches_per_epoch = state_dict["batches_per_epoch"] 51 | 52 | 53 | class Train(object): 54 | 55 | def __init__(self): 56 | self.logger, self.writer, self.tag_dir = get_Logger_and_SummaryWriter() 57 | self.engine=Engine(self.logger) 58 | self.device = torch.device("cuda") 59 | 60 | self.network=Fastnet() 61 | self.load_backbone(self.logger) 62 | self.network=self.network.cuda() 63 | self.network = self.engine.data_parallel(self.network) 64 | 65 | self.criterion = CrossEntropyLoss().to(self.device) 66 | 67 | base_params = [params for name, params in self.network.named_parameters() if ("temporal_high" in name)] 68 | finetune_params = [params for name, params in self.network.named_parameters() if ("temporal_high" not in name)] 69 | 70 | self.optim = Adam([ 71 | {'params': base_params, 'lr': config.base_lr,'weight_decay':1e-4, 'name': "base_params"}, 72 | {'params': finetune_params, 'lr': config.finetune_lr,'weight_decay':1e-4, 'name': 'finetune_params'}]) 73 | 74 | self.train_dataset, self.train_multiscale_dataset, statistics = get_video_dataset() 75 | 76 | self.train_multiscale_loader = [] 77 | self.train_multiscale_smapler = [] 78 | for dst in self.train_multiscale_dataset: 79 | ld, sp = self.engine.get_train_loader(dst,config.video_batchsize) 80 | self.train_multiscale_loader.append(ld) 81 | self.train_multiscale_smapler.append(sp) 82 | self.train_loader,self.train_sampler=self.engine.get_train_loader(self.train_dataset,config.video_batchsize) 83 | 84 | batches_per_epoch = 0 85 | batches_per_epoch += len(self.train_loader) 86 | for loader in self.train_multiscale_loader: 87 | batches_per_epoch += len(loader) 88 | 89 | self.sche = TrainSchedule(batches_per_epoch) 90 | if self.engine.local_rank==0: 91 | self.logger.info(config) 92 | self.logger.info(self.network) 93 | total_paramters = sum([np.prod(p.size()) for p in self.network.parameters()]) 94 | self.logger.info('Total network parameters: ' + str(total_paramters)) 95 | 96 | def save_checkpoint(self): 97 | os.makedirs(os.path.join(self.tag_dir, "epoch_%d_batch_%d" % ( 98 | self.sche.cur_epoch, self.sche.cur_batches)), exist_ok=True) 99 | save_root = os.path.join(self.tag_dir, "epoch_%d_batch_%d" % ( 100 | self.sche.cur_epoch, self.sche.cur_batches)) 101 | torch.save(self.state_dict(), os.path.join(save_root, "checkpoint.pth")) 102 | 103 | def adjust_learning_rate(self): 104 | if config.lr_mode == 'poly': 105 | cur_iter = self.sche.batches_per_epoch * self.sche.cur_epoch + self.sche.cur_batches 106 | max_iter = self.sche.batches_per_epoch * self.sche.total_epoch 107 | base_lr = config.base_lr * (1 - cur_iter * 1.0 / max_iter) ** 0.9 108 | finetune_lr = config.finetune_lr * (1 - cur_iter * 1.0 / max_iter) ** 0.9 109 | 110 | for param_group in self.optim.param_groups: 111 | if param_group["name"] == "base_params": 112 | param_group['lr'] = base_lr 113 | if param_group["name"] == "finetune_params": 114 | param_group['lr'] = finetune_lr 115 | 116 | return base_lr, finetune_lr 117 | 118 | def train_per_loader(self, trainloader): 119 | self.network.train() 120 | loss_meter = AverageValueMeter() 121 | for idx, (img, label) in enumerate(trainloader): 122 | baselr,finetunelr = self.adjust_learning_rate() 123 | img = img.to(self.device) 124 | label = label.to(self.device) 125 | if len(label.shape) == 5: 126 | label = label.view(-1, *(label.shape[2:])) 127 | output = self.network(img) 128 | loss = self.criterion(output, label) 129 | self.optim.zero_grad() 130 | loss.backward() 131 | self.optim.step() 132 | loss_meter.add(float(loss)) 133 | 134 | if self.engine.local_rank == 0: 135 | if self.sche.cur_batches % config.log_inteval == 0: 136 | self.logger.info("%s-epoch:%d/%d batch:%d/%d loss:%.4f base_lr:%e finetune_lr:%e" % ( 137 | self.tag_dir.split("/")[-1], self.sche.cur_epoch, self.sche.total_epoch, self.sche.cur_batches, 138 | self.sche.batches_per_epoch, loss_meter.value()[0], baselr,finetunelr)) 139 | self.sche.update() 140 | 141 | return loss_meter.value()[0] 142 | 143 | def train_per_epoch(self): 144 | for idx,loader in enumerate(self.train_multiscale_loader): 145 | self.train_per_loader(loader) 146 | loss_train= self.train_per_loader(self.train_loader) 147 | if self.engine.local_rank == 0: 148 | self.logger.info("train_img_loss:%.4f" % (loss_train)) 149 | 150 | def train(self): 151 | while self.sche.cur_epoch < self.sche.total_epoch: 152 | self.train_sampler.set_epoch(self.sche.cur_epoch) 153 | for sp in self.train_multiscale_smapler: 154 | sp.set_epoch(self.sche.cur_epoch) 155 | self.train_per_epoch() 156 | if self.engine.local_rank == 0: 157 | self.save_checkpoint() 158 | 159 | def state_dict(self): 160 | if config.parallel is True: 161 | state_dict = {"net": self.network.module.state_dict(), 162 | 'optimizer': self.optim.state_dict(), 163 | 'sche': self.sche.state_dict()} 164 | else: 165 | state_dict = {"net": self.network.state_dict(), 166 | 'optimizer': self.optim.state_dict(), 167 | 'sche': self.sche.state_dict()} 168 | return state_dict 169 | 170 | def load_state_dict(self, state_dict): 171 | if config.parallel is True: 172 | #self.sche.load_state_dict(state_dict["sche"]) 173 | #self.optim.load_state_dict(state_dict["optimizer"]) 174 | self.network.module.load_state_dict(state_dict["net"]) 175 | else: 176 | self.sche.load_state_dict(state_dict["sche"]) 177 | self.optim.load_state_dict(state_dict["optimizer"]) 178 | self.network.load_state_dict(state_dict["net"]) 179 | 180 | def load_backbone(self,logger): 181 | assert config.pretrain_state_dict is not None,"error" 182 | self.network.load_backbone(torch.load(config.pretrain_state_dict,map_location=torch.device('cpu'))["net"],logger) 183 | 184 | if __name__ == "__main__": 185 | torch.backends.cudnn.benchmark = True 186 | trainer = Train() 187 | trainer.train() 188 | 189 | -------------------------------------------------------------------------------- /pretrain.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | 3 | source /home/ycgu/anaconda3/bin/activate pt10py36 4 | 5 | GPU_IDS=0,1,2,3 6 | NUM_THREADS=2 7 | CUDA_VISIBLE_DEVICES=${GPU_IDS} OMP_NUM_THREADS=${NUM_THREADS} python -m torch.distributed.launch --nproc_per_node=4 --master_port=1235 pretrain_distribute.py 8 | -------------------------------------------------------------------------------- /pretrain_distribute.py: -------------------------------------------------------------------------------- 1 | from torch.optim import Adam 2 | import torch 3 | from utils.Distribute.engine import Engine 4 | from Data.dataloader import get_pretrain_dataset 5 | import torch.nn.functional as F 6 | from utils.utils import get_Logger_and_SummaryWriter 7 | import os 8 | from config import config 9 | from Models.mobilenetv3_pretrain import Fastnet 10 | import torch.nn as nn 11 | from utils.SalEval import SalEval 12 | from torchnet.meter import AverageValueMeter 13 | import numpy as np 14 | import random 15 | 16 | def setup_seed(seed): 17 | torch.manual_seed(seed) 18 | torch.cuda.manual_seed_all(seed) 19 | np.random.seed(seed) 20 | random.seed(seed) 21 | torch.backends.cudnn.deterministic = True 22 | 23 | class CrossEntropyLoss(nn.Module): 24 | def __init__(self): 25 | super(CrossEntropyLoss, self).__init__() 26 | 27 | def forward(self, *inputs): 28 | pred,target = tuple(inputs) 29 | total_loss = F.binary_cross_entropy(pred, target.float()) 30 | return total_loss 31 | 32 | class TrainSchedule(object): 33 | def __init__(self,batches_per_epoch): 34 | self.cur_epoch = 0 35 | self.total_epoch=config.pretrain_epoches 36 | self.cur_batches = 0 37 | self.batches_per_epoch=batches_per_epoch 38 | 39 | def update(self): 40 | self.cur_batches+=1 41 | if not self.cur_batches 10: 22 | time_spent.append(time.time() - start_time) 23 | print('Avg execution time (ms): %.4f, FPS:%d'%(np.mean(time_spent),1*1//np.mean(time_spent))) 24 | return 1*1//np.mean(time_spent) 25 | 26 | if __name__=="__main__": 27 | 28 | torch.backends.cudnn.benchmark = True 29 | 30 | from Models import mobilenetv3temporal_PCSA as net 31 | model = net.Fastnet() 32 | 33 | computeTime(model) 34 | -------------------------------------------------------------------------------- /utils/Distribute/engine.py: -------------------------------------------------------------------------------- 1 | import os 2 | from config import config 3 | import torch 4 | import torch.distributed as dist 5 | from torch.utils.data import DistributedSampler,DataLoader 6 | import torch.nn as nn 7 | 8 | def all_reduce_tensor(tensor, op=dist.ReduceOp.SUM, world_size=1, norm=True): 9 | tensor = tensor.clone() 10 | dist.all_reduce(tensor, op) 11 | if norm: 12 | tensor.div_(world_size) 13 | return tensor 14 | 15 | class Engine(object): 16 | 17 | ## distribute init 18 | def __init__(self, logger=None): 19 | self.distributed = False 20 | self.logger=logger 21 | if 'WORLD_SIZE' in os.environ: 22 | self.distributed = int(os.environ['WORLD_SIZE']) > 1 23 | else: 24 | raise NotImplementedError 25 | 26 | if self.distributed: 27 | self.local_rank = config.local_rank 28 | self.world_size = int(os.environ['WORLD_SIZE']) 29 | torch.cuda.set_device(self.local_rank) 30 | dist.init_process_group(backend="nccl", init_method='env://') 31 | else: 32 | raise NotImplementedError 33 | 34 | ## convert model 35 | 36 | def data_parallel(self, model): 37 | if self.distributed: 38 | #model = nn.SyncBatchNorm.convert_sync_batchnorm(model) 39 | model = nn.parallel.DistributedDataParallel(model,device_ids=[self.local_rank],output_device=self.local_rank) 40 | else: 41 | raise NotImplementedError 42 | return model 43 | 44 | def get_train_loader(self, train_dataset,batchsize): 45 | if self.distributed: 46 | train_sampler = DistributedSampler( 47 | train_dataset) 48 | local_bs = batchsize // self.world_size 49 | is_shuffle = False 50 | train_loader = DataLoader(train_dataset, 51 | batch_size=local_bs, 52 | num_workers=2, 53 | drop_last=False, 54 | shuffle=is_shuffle, 55 | pin_memory=False, 56 | sampler=train_sampler) 57 | 58 | else: 59 | raise NotImplementedError 60 | 61 | return train_loader, train_sampler 62 | 63 | def all_reduce_tensor(self, tensor, norm=True): 64 | if self.distributed: 65 | return all_reduce_tensor(tensor, world_size=self.world_size, norm=norm) 66 | else: 67 | raise NotImplementedError 68 | 69 | 70 | def __enter__(self): 71 | return self 72 | 73 | def __exit__(self, type, value, tb): 74 | torch.cuda.empty_cache() 75 | if type is not None: 76 | self.logger.warning( 77 | "A exception occurred during Engine initialization, " 78 | "give up running process") 79 | return False 80 | -------------------------------------------------------------------------------- /utils/utils.py: -------------------------------------------------------------------------------- 1 | from tensorboardX import SummaryWriter 2 | import logging 3 | import os 4 | import torch 5 | from PIL import Image 6 | from math import ceil 7 | 8 | def get_Logger_and_SummaryWriter(): 9 | for i in range(1000): 10 | tag_dir = 'checkpoints/tensorboard/try_{}'.format(i) 11 | if not os.path.exists(tag_dir): 12 | os.makedirs(tag_dir, exist_ok=True) 13 | logger = logging.getLogger("PGGAN") 14 | file_handler = logging.FileHandler(os.path.join(tag_dir, 'log.txt'), "w") 15 | stdout_handler = logging.StreamHandler() 16 | logger.addHandler(file_handler) 17 | logger.addHandler(stdout_handler) 18 | stdout_handler.setFormatter(logging.Formatter('%(asctime)s %(levelname)s %(message)s')) 19 | file_handler.setFormatter(logging.Formatter('%(asctime)s %(levelname)s %(message)s')) 20 | logger.setLevel(logging.INFO) 21 | return logger, SummaryWriter(tag_dir),tag_dir 22 | 23 | def safe_save(img,save_path): 24 | os.makedirs(save_path.replace(save_path.split('/')[-1],""),exist_ok=True) 25 | img.save(save_path) 26 | --------------------------------------------------------------------------------