如何将字符串对象类型转换回张量对象

Question

我正在处理 csv。我嵌入了一个列并转换为张量。就像，

tensor([-1.7110e-01,  1.3811e-01, -2.5881e-01, -1.8281e-01, -3.3073e-01,
    -1.1071e-01])

将这些张量保存为新列并将其保存到 csv 中。现在，当我再次加载该 csv 并查看嵌入列的值时，它看起来像这样

'tensor([-1.7110e-01,  1.3811e-01, -2.5881e-01, -1.8281e-01, -3.3073e-01,\n        -1.1071e-01])'

让我知道如何将其转换回来并使用？

Answer 1

显然，您写入 CSV 文件的内容是张量对象的

repr

，这就是您打印对象时看到的内容。要正确处理 CSV 文件中的张量，您应该将张量对象序列化为正确的数据格式以供导出，并在加载时反序列化。

以下是如何实现此目的的示例：

import csv
import ast

# Sample tensor
sample_tensor = torch.tensor([-1.7110e-01, 1.3811e-01, -2.5881e-01, -1.8281e-01, -3.3073e-01, -1.1071e-01])

# Serialize the tensor as a list and write to CSV
csv_file = 'tensor_data.csv'
with open(csv_file, 'w') as file:
    writer = csv.writer(file)
    writer.writerow([sample_tensor.tolist()])

# Read the CSV file and reconstruct the tensor
with open(csv_file, 'r') as file:
    reader = csv.reader(file)
    serialized_tensor = next(reader)[0]
reconstructed_tensor = torch.tensor(ast.literal_eval(serialized_tensor))

# Print the reconstructed tensor
print(reconstructed_tensor)