将 .txt 文件的特定行（包含大量行）写入新的 txt 文件

Question

我有一个张量格式的文本 (.txt) 文件，如下所示：

[tensor([[5.7744e+02, 1.7730e+02, 6.2396e+02, 2.1678e+02, 9.9988e-01],
    [5.8410e+02, 1.7532e+02, 6.2769e+02, 2.1083e+02, 9.9969e-01],
    [6.4279e+02, 1.7254e+02, 6.7784e+02, 2.0478e+02, 9.9957e-01],
    [6.9179e+02, 1.6719e+02, 8.2207e+02, 2.6694e+02, 9.9954e-01],
   [1.0050e+03, 2.9167e+02, 1.0792e+03, 3.7500e+02, 5.7740e-05]],
   device='cuda:0')]
[tensor([[7.6615e+02, 1.7577e+02, 8.2579e+02, 2.0013e+02, 9.9981e-01],
    [7.7214e+02, 1.7320e+02, 8.3133e+02, 1.9800e+02, 9.9901e-01],
    [6.3909e+02, 1.7303e+02, 6.6037e+02, 1.8797e+02, 9.9790e-01],
    [7.6184e+02, 1.7468e+02, 8.3014e+02, 2.0274e+02, 9.9395e-01],
   [6.9179e+02, 1.6719e+02, 8.2207e+02, 2.6694e+02, 9.9954e-01],
   [1.0050e+03, 2.9167e+02, 1.0792e+03, 3.7500e+02, 5.7740e-05]],
   device='cuda:0')]

如何将每个张量写入新的 (.txt) 文件。 例如我想要：

[tensor([[5.7744e+02, 1.7730e+02, 6.2396e+02, 2.1678e+02, 9.9988e-01],
    [5.8410e+02, 1.7532e+02, 6.2769e+02, 2.1083e+02, 9.9969e-01],
    [6.4279e+02, 1.7254e+02, 6.7784e+02, 2.0478e+02, 9.9957e-01],
    [6.9179e+02, 1.6719e+02, 8.2207e+02, 2.6694e+02, 9.9954e-01],
   [1.0050e+03, 2.9167e+02, 1.0792e+03, 3.7500e+02, 5.7740e-05]],
   device='cuda:0')]

这个张量在文件 one.txt 和另一个张量到 two.txt 等等。

Answer 1

使用 Matrkeenerh 的建议，如果我们把你的文件作为input.txt那么你可以做一些非常简单的事情。

file_in = 'input.txt'

counter = 0 # counter used ofr the output files

with open(file_in, 'r') as fin:    
    line = fin.readline() # read file in line by line
    while line: # so long as line is not empty,
                # if you do have blank lines in your file
                # you'd need to do something else
        if 'tensor' in line:
            # the very first time there won't be a fout so use try
            # there are other options, like "if count > 0", etc.
            try:
                fout.close() # close the old file
            except:
                pass
            # augment the counter and open a new file
            counter += 1 
            fout = open('output_%03d.txt' % counter, 'w')
        fout.write(line) # write the line to the output file
        line = fin.readline() # read the next line of the input file
fout.close() # close the last output file

这将使用您要查找的输出创建output_001.txt 、 output_002.txt等文件。

请注意，我在上面选择了逐行阅读文件。 如果文件不大，你可以用类似的东西一次性读完

with open(file_in, 'r'):
      lines = fid.readlines()
for line in lines:
      if 'tensor' in line:
            ...
            ...
      fout.write(line)
fout.close()

有可能更简单的方法来做到这一点。 例如，如果您的文件是一致的，并且每个tensor条目都在输入文件的 6 行上，那么您可以每六行启动一个新文件，例如。 或者您可以将所有内容都保存在一个巨大的文件中，知道要在文件中查看的位置，因为条目n将从(n-1) * 6 + 1 。