Question

我有一个类似这样的文本文件。我想打破界限，这样我就可以得到单独的列，以便能够绘制图形。

node name | requested bytes | total execution time | accelerator execution time | cpu execution time
prefix/up23/conv2d_transpose     37.75MB (100.00%, 15.34%),      150.71ms (100.00%, 4.83%),             0us (0.00%, 0.00%),      150.71ms (100.00%, 4.83%)
prefix/up20/conv2d_transpose       18.87MB (84.66%, 7.67%),       115.01ms (95.17%, 3.68%),             0us (0.00%, 0.00%),       115.01ms (95.17%, 3.68%)
prefix/up17/conv2d_transpose       18.87MB (76.99%, 7.67%),        91.43ms (91.49%, 2.93%),             0us (0.00%, 0.00%),        91.43ms (91.49%, 2.93%)
prefix/fres19/conv_b_1x3/Conv2D        2.10MB (69.33%, 0.85%),        46.41ms (88.56%, 1.49%),             0us (0.00%, 0.00%),        46.41ms (88.56%, 1.49%)
prefix/fres5/conv_b_3x1/Conv2D        2.10MB (68.47%, 0.85%),        44.63ms (87.07%, 1.43%),             0us (0.00%, 0.00%),        44.63ms (87.07%, 1.43%)
prefix/fres6/conv_a_3x1/Conv2D        2.10MB (67.62%, 0.85%),        40.19ms (85.64%, 1.29%),             0us (0.00%, 0.00%),        40.19ms (85.64%, 1.29%)
prefix/fres22/conv_a_3x1/Conv2D        2.10MB (66.77%, 0.85%),        39.97ms (84.36%, 1.28%),             0us (0.00%, 0.00%),        39.97ms (84.36%, 1.28%)
prefix/fres21/conv_a_3x1/Conv2D        2.10MB (65.92%, 0.85%),        38.85ms (83.08%, 1.24%),             0us (0.00%, 0.00%),        38.85ms (83.08%, 1.24%)
pref

我尝试了以下内容，但它给出的输出如下：

with open('file.txt','r') as inp:
    arr = []
    for f in inp:
        arr.append(f)
b = arr[514]
c = b.split(' ')

输出：

['prefix/up23/conv2d_transpose', '', '', '', '', '37.75MB', '(100.00%,', '15.34%),', '', '', '', '', '', '150.71ms', '(100.00%,', '4.83%),', '', '', '', '', '', '', '', '', '', '', '', '', '0us', '(0.00%,', '0.00%),', '', '', '', '', '', '150.71ms', '(100.00%,', '4.83%)\n']

请提出建议，以使数据进入csv中的不同列。

Answer 1

您可以这样做：

with open('test.txt','r') as inp:
    for f in inp.readlines():
        print(f.split())

哪些印刷品：

['node', 'name', '|', 'requested', 'bytes', '|', 'total', 'execution', 'time', '|', 'accelerator', 'execution', 'time', '|', 'cpu', 'execution', 'time']
['prefix/up23/conv2d_transpose', '37.75MB', '(100.00%,', '15.34%),', '150.71ms', '(100.00%,', '4.83%),', '0us', '(0.00%,', '0.00%),', '150.71ms', '(100.00%,', '4.83%)']
['prefix/up20/conv2d_transpose', '18.87MB', '(84.66%,', '7.67%),', '115.01ms', '(95.17%,', '3.68%),', '0us', '(0.00%,', '0.00%),', '115.01ms', '(95.17%,', '3.68%)']
['prefix/up17/conv2d_transpose', '18.87MB', '(76.99%,', '7.67%),', '91.43ms', '(91.49%,', '2.93%),', '0us', '(0.00%,', '0.00%),', '91.43ms', '(91.49%,', '2.93%)']
['prefix/fres19/conv_b_1x3/Conv2D', '2.10MB', '(69.33%,', '0.85%),', '46.41ms', '(88.56%,', '1.49%),', '0us', '(0.00%,', '0.00%),', '46.41ms', '(88.56%,', '1.49%)']
['prefix/fres5/conv_b_3x1/Conv2D', '2.10MB', '(68.47%,', '0.85%),', '44.63ms', '(87.07%,', '1.43%),', '0us', '(0.00%,', '0.00%),', '44.63ms', '(87.07%,', '1.43%)']
['prefix/fres6/conv_a_3x1/Conv2D', '2.10MB', '(67.62%,', '0.85%),', '40.19ms', '(85.64%,', '1.29%),', '0us', '(0.00%,', '0.00%),', '40.19ms', '(85.64%,', '1.29%)']
['prefix/fres22/conv_a_3x1/Conv2D', '2.10MB', '(66.77%,', '0.85%),', '39.97ms', '(84.36%,', '1.28%),', '0us', '(0.00%,', '0.00%),', '39.97ms', '(84.36%,', '1.28%)']
['prefix/fres21/conv_a_3x1/Conv2D', '2.10MB', '(65.92%,', '0.85%),', '38.85ms', '(83.08%,', '1.24%),', '0us', '(0.00%,', '0.00%),', '38.85ms', '(83.08%,', '1.24%)']

这就是你想要的吗？

Answer 2

您可以使用：

with open('file.txt','r') as inp: 
        arr = []
        for f in inp: 
           arr.append(f)
b = arr[514]
c = b.split()

如果您想坚持自己的代码，则可以从列表中删除空元素使用

Output = [x for x in c if x]

Answer 3

这不是有史以来最漂亮的代码，但我相信它可以解决您的问题。我曾考虑过使用一些正则表达式来避免分割百分比，但认为数据始终遵循相同的模式，因此应该可以。

def remove_dangling_comma(content):
    if content[-1] == ',':
        return content[:-1]
    return content

data_columns = []
with open("words.txt", 'r') as f:
    for i, line in enumerate(f):
        if i == 0:
            continue  # skip header
        parts = line.split()
        node_name = parts[0]
        # concatenate broken parts of the same data and remove dangling commas, if any
        requested_bytes = remove_dangling_comma(' '.join([parts[1], parts[2], parts[3]]))
        total_time = remove_dangling_comma(' '.join([parts[4], parts[5], parts[6]]))
        accelerator_time = remove_dangling_comma(' '.join([parts[7], parts[8], parts[9]]))
        cpu_time = remove_dangling_comma(' '.join([parts[10], parts[11], parts[12]]))

        # append the processed data to the list
        data_columns.append([node_name, requested_bytes, total_time, accelerator_time, cpu_time])       

    print(data_columns)

输出：

[
    ['prefix/up20/conv2d_transpose', '18.87MB (84.66%, 7.67%)', '115.01ms (95.17%, 3.68%)', '0us (0.00%, 0.00%)', '115.01ms (95.17%, 3.68%)'],
    ['prefix/up17/conv2d_transpose', '18.87MB (76.99%, 7.67%)', '91.43ms (91.49%, 2.93%)', '0us (0.00%, 0.00%)', '91.43ms (91.49%, 2.93%)'], 
    ['prefix/fres19/conv_b_1x3/Conv2D', '2.10MB (69.33%, 0.85%)', '46.41ms (88.56%, 1.49%)', '0us (0.00%, 0.00%)', '46.41ms (88.56%, 1.49%)'], 
    ['prefix/fres5/conv_b_3x1/Conv2D', '2.10MB (68.47%, 0.85%)', '44.63ms (87.07%, 1.43%)', '0us (0.00%, 0.00%)', '44.63ms (87.07%, 1.43%)'],
    ['prefix/fres6/conv_a_3x1/Conv2D', '2.10MB (67.62%, 0.85%)', '40.19ms (85.64%, 1.29%)', '0us (0.00%, 0.00%)', '40.19ms (85.64%, 1.29%)'], 
    ['prefix/fres22/conv_a_3x1/Conv2D', '2.10MB (66.77%, 0.85%)', '39.97ms (84.36%, 1.28%)', '0us (0.00%, 0.00%)', '39.97ms (84.36%, 1.28%)'], 
    ['prefix/fres21/conv_a_3x1/Conv2D', '2.10MB (65.92%, 0.85%)', '38.85ms (83.08%, 1.24%)', '0us (0.00%, 0.00%)', '38.85ms (83.08%, 1.24%)']
]

将文本文件中的行分成几列，用于CSV

3 个答案: