yfguo91
commented
2 months ago Hi yfguo, I'd like to run the project on the cifar10 dataset. Unfortunately, I've encountered some unusual results as follows. Could you please help explain why the code isn't working? Thank you very much.
Time elapsed: 2.5160064697265625 tensor(2.3065, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0957, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 5.4820191860198975 tensor(2.3005, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0937, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 8.224401235580444 tensor(2.3013, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0950, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 10.806175947189331 tensor(2.3032, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0950, device='cuda:0', grad_fn=<DivBackward0>) Test Accuracy of the model on the 10000 test images: 10.0 best_acc is: tensor([10.], device='cuda:0') best_iter: 0 Iters: 19 Time elapsed: 2.621997594833374 tensor(2.3065, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0950, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 5.071321487426758 tensor(2.3005, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0954, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 7.414266109466553 tensor(2.3013, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0955, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 9.854078531265259 tensor(2.3032, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0948, device='cuda:0', grad_fn=<DivBackward0>) Test Accuracy of the model on the 10000 test images: 10.0 best_acc is: tensor([10.], device='cuda:0') best_iter: 0 Iters: 20 Time elapsed: 2.3380064964294434 tensor(2.3064, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0969, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 4.684789180755615 tensor(2.3006, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0970, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 7.105907440185547 tensor(2.3012, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0950, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 9.586540222167969 tensor(2.3031, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0954, device='cuda:0', grad_fn=<DivBackward0>) Test Accuracy of the model on the 10000 test images: 10.0 best_acc is: tensor([10.], device='cuda:0') best_iter: 0 Iters: 21 Time elapsed: 2.4517087936401367 tensor(2.3063, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0980, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 4.906440019607544 tensor(2.3007, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0973, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 7.477343797683716 tensor(2.3012, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0970, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 10.011773109436035 tensor(2.3031, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0984, device='cuda:0', grad_fn=<DivBackward0>) Test Accuracy of the model on the 10000 test images: 10.0 best_acc is: tensor([10.], device='cuda:0') best_iter: 0 Iters: 22 Time elapsed: 2.432337999343872 tensor(2.3063, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0990, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 4.947934150695801 tensor(2.3008, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1017, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 7.39822793006897 tensor(2.3012, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0984, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 9.842048406600952 tensor(2.3030, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.0987, device='cuda:0', grad_fn=<DivBackward0>) Test Accuracy of the model on the 10000 test images: 10.0 best_acc is: tensor([10.], device='cuda:0') best_iter: 0 Iters: 23 Time elapsed: 2.4248645305633545 tensor(2.3064, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1363, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 4.903534412384033 tensor(2.3008, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1039, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 7.391828298568726 tensor(2.3012, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1031, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 9.873691082000732 tensor(2.3030, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1031, device='cuda:0', grad_fn=<DivBackward0>) Test Accuracy of the model on the 10000 test images: 10.0 best_acc is: tensor([10.], device='cuda:0') best_iter: 0 Iters: 24 Time elapsed: 2.4539432525634766 tensor(2.3063, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1411, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 4.891488790512085 tensor(2.3009, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1096, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 7.318920612335205 tensor(2.3012, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1058, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 9.833449125289917 tensor(2.3030, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1110, device='cuda:0', grad_fn=<DivBackward0>) Test Accuracy of the model on the 10000 test images: 10.0 best_acc is: tensor([10.], device='cuda:0') best_iter: 0 Iters: 25 Time elapsed: 2.467111349105835 tensor(2.3061, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1167, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 4.8138267993927 tensor(2.3010, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1161, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 7.239405870437622 tensor(2.3012, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1137, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 9.60375189781189 tensor(2.3030, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1148, device='cuda:0', grad_fn=<DivBackward0>) Test Accuracy of the model on the 10000 test images: 10.0 best_acc is: tensor([10.], device='cuda:0') best_iter: 0 Iters: 26 Time elapsed: 2.345698833465576 tensor(2.3062, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1417, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 4.72053599357605 tensor(2.3011, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1183, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 7.054431676864624 tensor(2.3012, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1157, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 9.43491816520691 tensor(2.3029, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1158, device='cuda:0', grad_fn=<DivBackward0>) Test Accuracy of the model on the 10000 test images: 10.0 best_acc is: tensor([10.], device='cuda:0') best_iter: 0 Iters: 27 Time elapsed: 2.4523515701293945 tensor(2.3060, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1175, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 4.904651880264282 tensor(2.3012, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1183, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 7.314801216125488 tensor(2.3012, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1172, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 9.779573917388916 tensor(2.3029, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1164, device='cuda:0', grad_fn=<DivBackward0>) Test Accuracy of the model on the 10000 test images: 10.0 best_acc is: tensor([10.], device='cuda:0') best_iter: 0 Iters: 28 Time elapsed: 2.426724433898926 tensor(2.3059, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1185, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 4.901515483856201 tensor(2.3013, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1180, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 7.466801404953003 tensor(2.3012, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1173, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 10.177380561828613 tensor(2.3028, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1164, device='cuda:0', grad_fn=<DivBackward0>) Test Accuracy of the model on the 10000 test images: 10.0 best_acc is: tensor([10.], device='cuda:0') best_iter: 0 Iters: 29 Time elapsed: 2.430713176727295 tensor(2.3058, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1175, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 4.775119066238403 tensor(2.3014, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1214, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 7.120392084121704 tensor(2.3011, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1182, device='cuda:0', grad_fn=<DivBackward0>) Time elapsed: 9.459439039230347 tensor(2.3027, device='cuda:0', grad_fn=<AddBackward0>) tensor(0.1171, device='cuda:0', grad_fn=<DivBackward0>) Test Accuracy of the model on the 10000 test images: 10.0 best_acc is: tensor([10.], device='cuda:0') best_iter: 0 Iters: 30
看上去是没训起来，权重初始不够好，或者没有加正则化
Hi yfguo, I'd like to run the project on the cifar10 dataset. Unfortunately, I've encountered some unusual results as follows. Could you please help explain why the code isn't working? Thank you very much.