Pytorch使用DistributedDataParallel单机多卡训练时遇到
ValueError: sampler option is mutually exclusive with shuffle的报错
是因为sampler选项与shuffle互斥
在数据读取这一块,需要将shuffle改成 False
train_dataset = datasets.ImageFolder(root='data/train/', transform=data_transform)
train_sampler = torch.utils.data.distributed.DistributedSampler(train_dataset)
train_loader = torch.utils.data.DataLoader(train_dataset,
batch_size=batch_size,
shuffle=False, # 得改成False
num_workers=num_workers,
sampler=train_sampler
)