All integrationsAI Models
Weights & Biases
Work with your Weights & Biases experiments: find and tag runs, read their metrics, record results, promote model versions with aliases, and list reports and sweeps.
- actions
- 18
- triggers
- 11
Triggers
Start a workflow when something happens in Weights & Biases.
- New runRuns when a run starts in a W&B project, such as a new training job. Runs already there when the workflow is turned on don't start it.
- Run finishedRuns when a run in a W&B project finishes successfully, so you can share its results or start the next step.
- Run failedRuns when a run in a W&B project fails (its script stopped with an error).
- Run crashedRuns when a run in a W&B project crashes: it stopped reporting in, for example because its machine went down.
- Run metric crossed a thresholdRuns when a run's latest value of a metric (its summary) goes above or below a number you choose, such as accuracy above 0.9. Each run starts it once. Runs already past it when the workflow is turned on don't.
- New artifact versionRuns when a new version of a W&B artifact (a model, dataset or other file set) finishes uploading.
- Alias added to an artifact versionRuns when an alias such as production or staging is put on a version of a W&B artifact, so a model's promotion can start a deployment or a review.
- New projectRuns when a project is created in a W&B team.
- New sweepRuns when a hyperparameter sweep is created in a W&B project.
- Sweep finishedRuns when a hyperparameter sweep in a W&B project finishes, with its best result so far.
- New reportRuns when someone saves a new report in a W&B project.
Actions
Steps a workflow can run in Weights & Biases.
- Get my accountGet the W&B account the API key belongs to: username, email, default team and every team you belong to.
- List projectsList the projects in a W&B team or your personal account.
- Get projectGet a W&B project, with how many runs it has.
- Create projectCreate a W&B project. If one with that name already exists it is returned as it is, unchanged.
- List runsList the runs in a W&B project, newest first, optionally only those in a state, with a tag, in a group or started after a time.
- Get runGet a W&B run with its config, latest metrics (summary) and system metrics.
- Update runRename a W&B run, change its notes or group, or add or remove tags. Fields left blank stay as they are.
- Set run summary metricsAdd or change values in a W&B run's summary, such as a review score or a test result. The run's other summary metrics stay.
- Get run metric historyGet the values a W&B run logged over time, such as its loss at each step, sampled down to the number of rows asked for.
- Delete runDelete a W&B run, and optionally the artifacts it logged. This cannot be undone.
- List artifactsList the artifacts (models, datasets and other file sets) in a W&B project.
- List artifact versionsList the versions of a W&B artifact, with their aliases.
- Get artifact versionGet one version of a W&B artifact by its version or alias, with its metadata.
- Add artifact aliasesPut aliases such as production on a version of a W&B artifact. An alias is on one version at a time, so it moves from any other.
- Remove artifact aliasesTake aliases off a version of a W&B artifact.
- List reportsList the reports in a W&B project, with a link to each.
- List sweepsList the hyperparameter sweeps in a W&B project, with each one's state and best result.
- Get sweepGet a W&B hyperparameter sweep: its state, settings, run count and best result.
No actions match that.
Automate Weights & Biases with Lodol.
Connect Weights & Biases, then hand Lodol the job in Slack, Claude or ChatGPT.
Free plan, no credit card