Closed manongXT closed 2 years ago
使用“CUDA_VISIBLE_DEVICES=3 python3 train_ssd.py --save_interval 12000”训练时,只训练了一轮就停止,也没报错,请问是啥原因?结果显示如下: @yunxuan01:~/CrossRectify-SSOD/SSD300$ CUDA_VISIBLE_DEVICES=0,3 python3 train_ssd.py Loading base network... Initializing weights... random seed is set as 123 Loading the dataset... Training SSD on: VOC300 Using the specified args: Namespace(basenet='vgg16_reducedfc.pth', batch_size=4, cuda=True, dataset='VOC300', dataset_root='/CrossRectify-SSOD/Data/voc/VOCdevkit', gamma=0.1, lr=0.001, momentum=0.9, num_workers=4, resume=None, save_folder='weights/', save_interval=12000, seed=123, sup_aug_type='default', weight_decay=0.0005) iter 0 || loss: 30.9676, loss_c: 27.1707, loss_l: 3.7969, super_len: 1
SSD的问题,SSD训练不稳定,多重启几次试试
您好,我这边又遇到一个问题,在执行train_pseudo39.py脚本时,336行的priors_single = torch.load("priors.pt")无法执行,这个“priors.pt"该怎样加载?
它在这里
还是无法运行,直接结束,debug的话就是卡在这一行一直不动。
使用“CUDA_VISIBLE_DEVICES=3 python3 train_ssd.py --save_interval 12000”训练时,只训练了一轮就停止,也没报错,请问是啥原因?结果显示如下: @yunxuan01:~/CrossRectify-SSOD/SSD300$ CUDA_VISIBLE_DEVICES=0,3 python3 train_ssd.py Loading base network... Initializing weights... random seed is set as 123 Loading the dataset... Training SSD on: VOC300 Using the specified args: Namespace(basenet='vgg16_reducedfc.pth', batch_size=4, cuda=True, dataset='VOC300', dataset_root='/CrossRectify-SSOD/Data/voc/VOCdevkit', gamma=0.1, lr=0.001, momentum=0.9, num_workers=4, resume=None, save_folder='weights/', save_interval=12000, seed=123, sup_aug_type='default', weight_decay=0.0005) iter 0 || loss: 30.9676, loss_c: 27.1707, loss_l: 3.7969, super_len: 1