模型训练结束后自动整理记录各项参数

miracle

模型训练完成后，要注意及时记录保存各种参数，网络结构，分类存档以供后续对比出各种结论，但问题是填写一把这个表格太慢了而且太难受了。。

废话不多说，上脚本：

def write_summary(net, opt, summary_info):
    current_time = datetime.datetime.now().strftime("%Y-%m-%d_%H:%M:%S")
    prefix   = './source/summaries/'+net.model_name
    if not os.path.exists(prefix): os.mkdir(prefix)
    sum_path = prefix + '/MiracleYoo_'+current_time+'_'+net.model_name+'_Model_Testing_Record_Form.md'
    with codecs.open('./config.py', 'r', encoding='utf-8') as f:
        raw_data = f.readlines()
        configs  = ''
        for line in raw_data:
            if line.strip().startswith('self.'):
                configs += line.strip().strip('self.')+'\n'

    content = '''
# Model Testing Record Form
| Item Name        | Information |
| ---------        | ----------- |
| Model Name       | %s          |
| Tester's Name    | Miracle Yoo |
| Author's Nmae    | Miracle Yoo |
| Test Time        | %s          |
| Test Position    | %s          |
| Training Epoch   | %d          |
| Highest Test Acc | %.4f        |
| Loss of highest Test Acc| %.4f |
| Last epoch test acc   | %.4f   |
| Last epoch test loss  | %.4f   |
| Last epoch train acc  | %.4f   |
| Last epoch train loss | %.4f   |
| Train Dataset Path    | %s     |
| Test Dataset Path     | %s     |
| Class Number     | %d          |
| Framwork         | Pytorch     |
| Basic Method     | Classify    |
| Input Type       | Char        |
| Criterion        | CrossEntropy|
| Optimizer        | %s          |
| Learning Rate    | %.4f        |
| Embedding dimension   | %d     |
| Data Homogenization   | True   |
| Pretreatment|Remove punctuation|
| Other Major Param |            |
| Other Operation   |            |


## Configs
\```
%s
\```

## Net Structure
\```
%s
\```
    '''%(
        net.model_name,
        current_time,
        opt.TEST_POSITION,
        summary_info['total_epoch'],
        summary_info['best_acc'],
        summary_info['best_acc_loss'],
        summary_info['ave_test_acc'],
        summary_info['ave_test_loss'],
        summary_info['ave_train_acc'],
        summary_info['ave_train_loss'],
        os.path.basename(opt.TRAIN_DATASET_PATH),
        os.path.basename(opt.TEST_DATASET_PATH),
        opt.NUM_CLASSES,
        opt.OPTIMIZER,
        opt.LEARNING_RATE,
        opt.EMBEDDING_DIM,

        configs.strip('\n'),
        str(net)
    )
    with codecs.open(sum_path, 'w+', encoding='utf-8') as f:
        f.writelines(content)

记着把上面```前面的\去掉食用~
这个表不全，后面会有补充，内容也可以根据你自己模型和项目的具体情况修改。

演示效果如下：

Model Testing Record Form

Item Name	Information
Model Name	TextCNNInc
Tester's Name	Miracle Yoo
Author's Nmae	Miracle Yoo
Test Time	2018-05-13_15:24:43
Test Position	Gangge Server
Training Epoch	1
Highest Test Acc	0.0006
Loss of highest Test Acc	2.1721
Last epoch test acc	0.0006
Last epoch test loss	2.1721
Last epoch train acc	0.0004
Last epoch train loss	1.2189
Train Dataset Path	knowledge&log_data.txt
Test Dataset Path	yibot_two_year_test.txt
Class Number	2411
Framwork	Pytorch
Basic Method	Classify
Input Type	Char
Criterion	CrossEntropy
Optimizer	Adam
Learning Rate	0.0010
Embedding dimension	512
Data Homogenization	True
Pretreatment	Remove punctuation
Other Major Param
Other Operation

Configs

USE_CUDA           = torch.cuda.is_available()
RUNNING_ON_SERVER  = Fa
NET_SAVE_PATH      = "./source/trained_net/"
TRAIN_DATASET_PATH = "../database/test_train/knowledge&log_data.txt"
TEST_DATASET_PATH  = "../database/test_train/yibot_two_year_test.txt"
NUM_EPOCHS         = 1
BATCH_SIZE         = 8
TOP_NUM            = 4
NUM_WORKERS        = 1
IS_TRAINING        = Tru
ENSEMBLE_TEST      = Fa
LEARNING_RATE      = 0.001
RE_TRAIN           = Fa
TEST_POSITION      = 'Gangge Server'
OPTIMIZER          = 'Adam'
USE_CHAR           = Tru
USE_WORD2VEC       = Tru
NUM_CLASSES        = 1890#len(get_labels2idx()[0])
EMBEDDING_DIM      = 512
VOCAB_SIZE         = 20029
CHAR_SIZE          = 3403
LSTM_HID_SIZE      = 512
LSTM_LAYER_NUM     = 2
TITLE_DIM          = 200
SENT_LEN           = 20
LINER_HID_SIZE     = 2000
KERNEL_SIZE        = [1,2,3,4,5]
DILA_TITLE_DIM     = 20

Net Structure

TextCNNInc(
  (encoder): Embedding(3394, 512)
  (question_convs): ModuleList(
    (0): Sequential(
      (0): Conv1d(512, 200, kernel_size=(1,), stride=(1,))
      (1): BatchNorm1d(200, eps=1e-05, momentum=0.1, affine=True)
      (2): ReLU(inplace)
      (3): MaxPool1d(kernel_size=20, stride=20, padding=0, dilation=1, ceil_mode=False)
    )
    (1): Sequential(
      (0): Conv1d(512, 200, kernel_size=(3,), stride=(1,))
      (1): BatchNorm1d(200, eps=1e-05, momentum=0.1, affine=True)
      (2): ReLU(inplace)
      (3): MaxPool1d(kernel_size=18, stride=18, padding=0, dilation=1, ceil_mode=False)
    )
    (2): Sequential(
      (0): Conv1d(512, 200, kernel_size=(1,), stride=(1,))
      (1): BatchNorm1d(200, eps=1e-05, momentum=0.1, affine=True)
      (2): ReLU(inplace)
      (3): Conv1d(200, 200, kernel_size=(3,), stride=(1,))
      (4): BatchNorm1d(200, eps=1e-05, momentum=0.1, affine=True)
      (5): ReLU(inplace)
      (6): MaxPool1d(kernel_size=18, stride=18, padding=0, dilation=1, ceil_mode=False)
    )
    (3): Sequential(
      (0): Conv1d(512, 200, kernel_size=(3,), stride=(1,))
      (1): BatchNorm1d(200, eps=1e-05, momentum=0.1, affine=True)
      (2): ReLU(inplace)
      (3): Conv1d(200, 200, kernel_size=(5,), stride=(1,))
      (4): BatchNorm1d(200, eps=1e-05, momentum=0.1, affine=True)
      (5): ReLU(inplace)
      (6): MaxPool1d(kernel_size=14, stride=14, padding=0, dilation=1, ceil_mode=False)
    )
  )
  (fc): Sequential(
    (0): Linear(in_features=800, out_features=2000, bias=True)
    (1): BatchNorm1d(2000, eps=1e-05, momentum=0.1, affine=True)
    (2): ReLU(inplace)
    (3): Dropout(p=0.5)
    (4): Linear(in_features=2000, out_features=2411, bias=True)
  )
)

喵喵喵~
0_1526197423071_SAO_刀剑神域_57.jpg