[人工智能]【PyTorch】常用公开数据集构造 DataLoader 过程

图像数据集.

MNIST.

• 官方文档.

# In[Dataset]
transformation = transforms.Compose([transforms.ToTensor(),
                                     transforms.Normalize((0.1307,), (0.3081,))])

train_dataset = datasets.MNIST('data\\',
                               train = True,
                               transform = transformation,
                               download = True)

test_dataset = datasets.MNIST('data\\',
                               train = False,
                               transform = transformation,
                               download = True)

# In[DataLoader]
train_loader = DataLoader(train_dataset,
                          batch_size = 32,
                          shuffle = True,
                          num_workers = 0)

test_loader = DataLoader(test_dataset,
                         batch_size = 32,
                         shuffle = False,
                         num_workers = 0)

CIFAR10.

• 官方文档.

# In[DataLoader]
ds = datasets.CIFAR10(root = 'data',download = True,
                      transform = trans)

loader = torch.utils.data.DataLoader(ds,batch_size = 64,
                                     shuffle = True)

STL10.

• 官方文档.

IMAGE = STL10(root = 'data/',
              split = 'train',
              transform = transforms.ToTensor(),
              download = True)

VOC Segmentation.

• 官方文档.
• 下载后的目录结构如下所示：
  

文本数据集.

IMDB.

• 官方文档.

# In[Dataset]
TEXT = data.Field(lower = True,
                  fix_length = 200,
                  batch_first = True)

LABEL = data.Field(sequential = False)

train,test = datasets.IMDB.splits(TEXT,LABEL)

TEXT.build_vocab(train,
                 vectors = GloVe(name = '6B',
                                 dim = 300),
                 max_size = 10000,
                 min_freq = 10)
LABEL.build_vocab(train)

train_iter,test_iter = \
    data.BucketIterator.splits((train,test),
                               batch_size = 32,
                               device = torch.device('cpu'),
                               shuffle = True)

train_iter.repeat = False
test_iter.repeat = False

WikiText-2.

• 官方文档.

# In[Dataset]
TEXT = data.Field(batch_first = False,lower = True)

train,valid,test = datasets.WikiText2.splits(TEXT,root = 'data')

# In[Trunc]
batchsize = 4
for dataset in (train,valid,test):
    end = (len(dataset[0].text)//batchsize)*batchsize
    dataset[0].text = dataset[0].text[0:end]

train_iter,valid_iter,test_iter = \
    data.BPTTIterator.splits((train,valid,test),
                      batch_size = batchsize, 
                      bptt_len = 35)

# In[Vocab]
# TEXT.build_vocab(train,
#                  vectors = GloVe(name = '6B',
#                                  dim = 300),
#                  max_size = 10000,
#                  min_freq = 10)

TEXT.build_vocab(train)