2023-01-04 18:06:16 +08:00
|
|
|
# 分析工具
|
|
|
|
|
|
|
|
|
|
<!-- TOC -->
|
|
|
|
|
|
|
|
|
|
- [分析工具](#分析工具)
|
|
|
|
|
- [统计参数量](#统计参数量)
|
|
|
|
|
- [发布模型](#发布模型)
|
|
|
|
|
- [结果复现](#结果复现)
|
|
|
|
|
- [日志分析](#日志分析)
|
|
|
|
|
|
|
|
|
|
## 统计参数量
|
|
|
|
|
|
|
|
|
|
```shell
|
|
|
|
|
python tools/analysis_tools/count_parameters.py ${CONFIG_FILE}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
一个例子如下：
|
|
|
|
|
|
|
|
|
|
```shell
|
|
|
|
|
python tools/analysis_tools/count_parameters.py configs/selfsup/mocov2/mocov2_resnet50_8xb32-coslr-200e_in1k.py
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
## 发布模型
|
|
|
|
|
|
|
|
|
|
发布模型之前，你可能是想：
|
|
|
|
|
|
|
|
|
|
- 把模型权重转换为 CPU 张量。
|
|
|
|
|
- 删除优化器相关状态。
|
|
|
|
|
- 计算检查点文件的哈希值并把哈希 ID 加到文件名上。
|
|
|
|
|
|
|
|
|
|
```shell
|
|
|
|
|
python tools/model_converters/publish_model.py ${INPUT_FILENAME} ${OUTPUT_FILENAME}
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
例子如下：
|
|
|
|
|
|
|
|
|
|
```shell
|
|
|
|
|
python tools/model_converters/publish_model.py YOUR/PATH/epoch_100.pth YOUR/PATH/epoch_100_output.pth
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
## 结果复现
|
|
|
|
|
|
|
|
|
|
想让你的结果完全可以复现的话，训练最终模型时请设置 `--cfg-options randomness.deterministic=True` 。值得一提的是，这会关掉 `torch.backends.cudnn.benchmark` 并降低训练速度。
|
|
|
|
|
|
|
|
|
|
## 日志分析
|
|
|
|
|
|
|
|
|
|
`tools/analysis_tools/analyze_logs.py` 用训练日志文件画损失/学习率曲线。首先 `pip install seaborn` 安装依赖库。
|
|
|
|
|
|
|
|
|
|
```shell
|
|
|
|
|
python tools/analysis_tools/analyze_logs.py plot_curve [--keys ${KEYS}] [--title ${TITLE}] [--legend ${LEGEND}] [--backend ${BACKEND}] [--style ${STYLE}] [--out ${OUT_FILE}]
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
<div align="center">
|
|
|
|
|
<img src="https://raw.githubusercontent.com/open-mmlab/mmdetection/master/resources/loss_curve.png" width="400" />
|
|
|
|
|
</div>
|
|
|
|
|
|
|
|
|
|
例子如下:
|
|
|
|
|
|
|
|
|
|
- 画部分运行过程中分类的损失函数图像。
|
|
|
|
|
|
|
|
|
|
```shell
|
|
|
|
|
python tools/analysis_tools/analyze_logs.py plot_curve log.json --keys loss_dense --legend loss_dense
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
- 画部分运行过程中分类和倒退的损失函数图像并存到 pdf 文件里。
|
|
|
|
|
|
|
|
|
|
```shell
|
|
|
|
|
python tools/analysis_tools/analyze_logs.py plot_curve log.json --keys loss_dense loss_single --out losses.pdf
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
- 在同一张图内，比较两次训练的损失。
|
|
|
|
|
|
|
|
|
|
```shell
|
|
|
|
|
python tools/analysis_tools/analyze_logs.py plot_curve log1.json log2.json --keys loss --legend run1 run2
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
- 计算平均训练速度。
|
|
|
|
|
|
|
|
|
|
```shell
|
|
|
|
|
python tools/analysis_tools/analyze_logs.py cal_train_time log.json [--include-outliers]
|
|
|
|
|
```
|
|
|
|
|
|
|
|
|
|
输出应该像下面这样：
|
|
|
|
|
|
|
|
|
|
```text
|
|
|
|
|
-----Analyze train time of work_dirs/some_exp/20190611_192040.log.json-----
|
|
|
|
|
slowest epoch 11, average time is 1.2024
|
|
|
|
|
fastest epoch 1, average time is 1.1909
|
|
|
|
|
time std over epochs is 0.0028
|
|
|
|
|
average iter time: 1.1959 s/iter
|
|
|
|
|
```
|
