使用 DVC 进行超参数调优

面向机器学习的 CI/CD

Ravi Bhadauria

Machine Learning Engineer

超参数调优工作流

超参数调优
- 输入：参数搜索范围
- 输出：最佳参数
训练
- 输入：最佳参数
- 输出：指标与图表（已介绍）
松耦合，便于独立训练
- 超参调优是充分但非必要条件
两个作业都依赖数据集

# Contents of hp configuration
{
    "n_estimators": [2, 4, 5],
    "max_depth": [10, 20, 50],
    "random_state": [1993]
}

# Contents of best parameters
{
    "n_estimators": 5,
    "max_depth": 20,
    "random_state": 1993
}

训练代码变更

Python 训练代码的变更

# Load hyperparameters from the JSON file
with open("rfc_best_params.json", "r") as params_file:
  rfc_params = json.load(params_file)


# Define and train model
model = RandomForestClassifier(**rfc_params)
model.fit(X_train, y_train)

使用 GridSearch 的超参调优

# Define the model and hyperparameter search space
model = RandomForestClassifier()
param_grid = json.load(open("hp_config.json", "r"))


# Perform GridSearch with five fold CV
grid_search = GridSearchCV(model, param_grid, cv=5)
grid_search.fit(X_train, y_train)


# Get the best hyperparameters
best_params = grid_search.best_params_
with open("rfc_best_params.json", "w") as outfile:
        json.dump(best_params, outfile)

DVC YAML 变更

超参数调优

stages:
  preprocess: ...
  train: ...
  hp_tune:
    cmd: python hp_tuning.py
    deps:
    - processed_dataset/weather.csv
    - hp_config.json
    - hp_tuning.py
    outs: # 不跟踪最佳参数
      - hp_tuning_results.md:
          cache: false

训练

stages:
  preprocess: ...
  hp_tune: ...
  train:
    cmd: python train.py
    deps:
    - processed_dataset/weather.csv
    - rfc_best_params.json # 最佳参数
    - train.py
    metrics:
      - metrics.json:
          cache: false

触发单个阶段

可独立触发阶段 dvc repro <stage_name>
强制运行超参调优阶段 dvc repro -f hp_tune
- 确保最佳参数文件会更新
使用 dvc repro train 运行训练
两个阶段都会触发依赖的预处理步骤

超参运行输出

mean_test_score	std_test_score	max_depth	n_estimators	random_state
0.999733	0.000413118	20	5	1993
0.999307	0.000574418	50	5	1993
0.99888	0.000617378	10	5	1993
0.997813	0.00117333	10	4	1993

Python 超参数调优脚本的变更

# Save the results of hyperparameter tuning
cv_results = pd.DataFrame(grid_search.cv_results_)
markdown_table = cv_results.to_markdown(index=False)
with open("hp_tuning_results.md", "w") as markdown_file:
  markdown_file.write(markdown_table)

小结

超参数调优路径
- 分支名 hp_tune/<some-string>
- 修改搜索配置
- 手动打开 PR
  - 强制运行 DVC 流水线 dvc repro -f hp_tune
  - 使用 cml pr create 基于最佳参数创建新的训练 PR
- 强制推送一次提交到训练 PR，触发训练作业
手动路径
- 分支名 train/<some-string>
- 编辑最佳参数文件并提交
- 手动打开 PR 触发训练作业

¡Vamos a practicar!

面向机器学习的 CI/CD