模型训练结束后自动整理记录各项参数
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:
1 | def write_summary(net, opt, summary_info): |
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 | 100 |
Highest Test Acc | 0.7102 |
Loss of highest Test Acc | 0.1721 |
Last epoch test acc | 0.6706 |
Last epoch test loss | 0.1721 |
Last epoch train acc | 0.8904 |
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
1 | USE_CUDA = torch.cuda.is_available() |
Net Structure
1 | TextCNNInc( |
喵喵喵~
Related Posts