记录参数、指标和制品
为每次运行记录超参数、分数和文件。
记录参数、指标和制品 是 CoddyKit 上的免费 MLOps Academy 课时。 这是第 2 节课，共 4 节。 你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。 这是 MLOps Academy 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 MLOps Academy 课程共包含 4 节课。
本课时的部分内容尚未翻译，以英文显示。
Three Things to Log
Inside a run you capture three kinds of data: the parameters you chose, the metrics you measured, and the files you produced.
Log a Parameter
A parameter is an input you picked, like learning rate or tree depth. Log it once so you always know how a run was configured.
mlflow.log_param("learning_rate", 0.01)Log Many Params at Once
Got a whole config dict? Use log_params to record every key and value in one call instead of writing them line by line.
mlflow.log_params({"max_depth": 6, "n_estimators": 200})Log a Metric
A metric is a number your model earned, like accuracy or loss. Log it so runs can be ranked and compared later.
mlflow.log_metric("accuracy", 0.92)Metrics Can Have Steps
Metrics change over training. Pass a step to log loss at each epoch, and MLflow draws a curve you can inspect in the UI.
for epoch in range(10):
mlflow.log_metric("loss", loss, step=epoch)What an Artifact Is
An artifact is any file a run produces: a saved model, a plot, a confusion matrix image, or a CSV of predictions.
Log an Artifact
Use log_artifact to attach a local file to the run. MLflow copies it to the artifact store so it stays tied to that experiment.
mlflow.log_artifact("confusion_matrix.png")Params vs Metrics
The rule of thumb: a param is an input you control, a metric is an output you measure. Logging both makes runs fully comparable.
Tag Your Runs
Add a tag to label a run with free-form notes, like the data version or who triggered it. Tags make later searching far easier.
mlflow.set_tag("data_version", "v3")It All Lives in the Run
Every param, metric, and artifact attaches to the active run. Close the run and that snapshot is frozen forever for you to revisit.
Log Early, Log Often
Cheap to log, painful to lose. Capture everything that might matter, because you can never reconstruct a run you failed to record.
Quick Check
Let us check that you can tell params from metrics.
Recap
You logged params, metrics with steps, artifacts, and tags. A run now holds its full story, ready to compare against the rest. ✅
常见问题解答
「记录参数、指标和制品」课时是免费的吗？
是的 — 「记录参数、指标和制品」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 MLOps Academy 课程的其余内容，请升级到 CoddyKit PRO。 MLOps Academy 课程共包含 4 节课。
「记录参数、指标和制品」这节课中我会学到什么？
为每次运行记录超参数、分数和文件。 你通过在浏览器中直接运行的动手代码来练习 MLOps Academy，全天候 AI 导师会在你学习这节课的过程中回答你的问题。
学习 MLOps Academy 需要有经验吗？
无需任何先前经验。CoddyKit 上的 MLOps Academy 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。 这是第 2 节课，共 4 节。
「记录参数、指标和制品」课时需要多长时间？
大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。
我能在这节 MLOps Academy 课中编写并运行代码吗？
能。每节 MLOps Academy 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。