Reference

• 1: Release Notes
• 1.1: 0.70.x
• 1.2: 0.69.x
• 1.3: 0.68.x
• 1.4: 0.67.x
• 1.5: 0.66.x
• 1.6: 0.65.x
• 1.7: 0.63.x
• 1.8: 0.61.0
• 1.9: 0.60.0
• 1.10: 0.58.1
• 1.11: 0.57.2
• 1.12: 0.56.0
• 1.13: Archived Releases
• 1.13.1: 0.54.0
• 1.13.2: 0.52.2
• 1.13.3: 0.51.0
• 1.13.4: 0.50.2
• 1.13.5: 0.49.0
• 1.13.6: 0.48.0
• 1.13.7: 0.47.3
• 1.13.8: 0.47.2
• 1.13.9: 0.46.0
• 1.13.10: 0.45.0
• 1.13.11: 0.44.1
• 1.13.12: 0.43.0
• 1.13.13: 0.42.0
• 1.13.14: 0.41.0
• 1.13.15: 0.40.0
• 1.13.16: 0.39.0
• 1.13.17: 0.38.0
• 1.13.18: 0.36.0
• 1.13.19: 0.35.0
• 1.13.20: 0.34.0
• 1.13.21: 0.33.0
• 1.14: Release policies and processes
• 2: Command Line Interface
• 2.1: wandb agent
• 2.2: wandb artifact
• 2.2.1: wandb artifact cache
• 2.2.1.1: wandb artifact cache cleanup
• 2.2.2: wandb artifact get
• 2.2.3: wandb artifact ls
• 2.2.4: wandb artifact put
• 2.3: wandb beta
• 2.3.1: wandb beta sync
• 2.4: wandb controller
• 2.5: wandb disabled
• 2.6: wandb docker
• 2.7: wandb docker-run
• 2.8: wandb enabled
• 2.9: wandb init
• 2.10: wandb job
• 2.10.1: wandb job create
• 2.10.2: wandb job describe
• 2.10.3: wandb job list
• 2.11: wandb launch
• 2.12: wandb launch-agent
• 2.13: wandb launch-sweep
• 2.14: wandb login
• 2.15: wandb offline
• 2.16: wandb online
• 2.17: wandb pull
• 2.18: wandb restore
• 2.19: wandb scheduler
• 2.20: wandb server
• 2.20.1: wandb server start
• 2.20.2: wandb server stop
• 2.21: wandb status
• 2.22: wandb sweep
• 2.23: wandb sync
• 2.24: wandb verify
• 3: JavaScript Library
• 4: Python Library v(0.21.0)
• 4.1: API Walkthrough
• 4.2: SDK
• 4.2.1: Functions
• 4.2.1.1: agent()
• 4.2.1.2: controller()
• 4.2.1.3: finish()
• 4.2.1.4: init()
• 4.2.1.5: login()
• 4.2.1.6: restore()
• 4.2.1.7: setup()
• 4.2.1.8: sweep()
• 4.2.1.9: teardown()
• 4.2.2: Classes
• 4.2.2.1: Artifact
• 4.2.2.2: Run
• 4.2.2.3: Settings
• 4.2.3: Data Types
• 4.2.3.1: Audio
• 4.2.3.2: box3d()
• 4.2.3.3: Html
• 4.2.3.4: Image
• 4.2.3.5: Molecule
• 4.2.3.6: Object3D
• 4.2.3.7: Plotly
• 4.2.3.8: Table
• 4.2.3.9: Video
• 4.2.4: Custom Charts
• 4.2.4.1: bar()
• 4.2.4.2: confusion_matrix()
• 4.2.4.3: histogram()
• 4.2.4.4: line_series()
• 4.2.4.5: line()
• 4.2.4.6: plot_table()
• 4.2.4.7: pr_curve()
• 4.2.4.8: roc_curve()
• 4.2.4.9: scatter()
• 4.3: Query API
• 4.3.1: api
• 4.3.2: artifacts
• 4.3.3: automations
• 4.3.4: files
• 4.3.5: history
• 4.3.6: integrations
• 4.3.7: projects
• 4.3.8: reports
• 4.3.9: runs
• 4.3.10: sweeps
• 4.3.11: teams
• 4.3.12: users
• 4.4: Automations
• 4.4.1: Automation
• 4.4.2: DoNothing
• 4.4.3: MetricChangeFilter
• 4.4.4: MetricThresholdFilter
• 4.4.5: NewAutomation
• 4.4.6: OnAddArtifactAlias
• 4.4.7: OnCreateArtifact
• 4.4.8: OnLinkArtifact
• 4.4.9: OnRunMetric
• 4.4.10: SendNotification
• 4.4.11: SendWebhook
• 4.5: wandb_workspaces
• 4.5.1: Reports
• 4.5.2: Workspaces
• 5: Query Expression Language
• 5.1: artifact
• 5.2: artifactType
• 5.3: artifactVersion
• 5.4: audio-file
• 5.5: bokeh-file
• 5.6: boolean
• 5.7: entity
• 5.8: file
• 5.9: float
• 5.10: html-file
• 5.11: image-file
• 5.12: int
• 5.13: joined-table
• 5.14: molecule-file
• 5.15: number
• 5.16: object3D-file
• 5.17: partitioned-table
• 5.18: project
• 5.19: pytorch-model-file
• 5.20: run
• 5.21: string
• 5.22: table
• 5.23: user
• 5.24: video-file

1 - Release Notes

This section includes release notes for supported W&B Server releases. For releases that are no longer supported, refer to Archived releases.

 RSS feed

1.1 - 0.70.x

W&B Server v0.70 includes features, enhancements, and performance improvements to help you gain and share insights more efficiently. For example:

• W&B Registry is now generally available, and organizations will be migrated from the legacy Model Registry in stages over several releases, with no action required.
• W&B organizations who use CoreWeave infrastructure can now observe CoreWeave infrastructure issues from within W&B, to streamline detection and resolution of problems. This release also brings CoreWeave AI Object Storage support for externally tracked files, as well as for BYOB storage for Dedicated Cloud and Self-Managed deployments.
• Rapidly prototype workspaces and visualizations to gain and share insights with workspace templates and additional bulk line plot and media panel settings at the workspace and section level.
• Monitor long-running running experiments before they finish with incremental table logging.

The latest patch is v0.70.1. Refer to Patches.

Support and end of life

• W&B Server v0.56 and below have reached end of life as of July 7, 2025.
• W&B Server v0.57 is scheduled to reach end of life on July 29, 2025.
• W&B Server v0.58 is scheduled to reach end of life on September 2, 2025.

A W&B Server release is supported for 12 months from its initial release date. As a reminder, customers using Self-managed are responsible to upgrade to a supported release in time to maintain support.

Refer to Release policies and processes. For assistance or questions, contact support.

Upcoming changes

• In an upcoming release, we will align multi-run and single-run workspace views, so when you drill into a single run, you’ll see the same layout and configuration of panels as you’ve configured for the multi-run view, including any custom charts, layouts, and settings. The goal is help you stay in context and avoid re-work around configuring your views. However, this will remove the ability to customize unique single-run views, isolated from the rest of the workspace. Reach out to your W&B team or support for any questions about the migration.
• Over the next several releases, we will migrate existing team-level Model Registry W&B Registry, For details and a timeline, see the W&B Registry GA announcement below.

Features

• W&B Registry is now generally available! Registry offers improved asset diversity, access management, and scalability, making it a true hub for collaborating on AI initiatives across large, diverse teams.

  • Registry can be turned on today with the ENABLE_REGISTRY_UI environment variable or by contacting your W&B support team.

  • We’re migrating existing team-level Model Registry assets to W&B Registry in stages over the next several releases. No action is needed on your end. W&B will automatically copy Model Registry assets for each team to a corresponding automatically-created team-level private registry.  Existing references to model collections in your old team-level Model Registry will still work.

    1. In Server v0.71, Registry will be on by default for all organizations.
    2. Server v0.72 (August) will contain an opt-out environment variable that will execute this migration automatically during the upgrade. Migration takes between 5 and 30 minutes for most organizations. During the migration, linking artifacts to either the old team’s Model Registry or the new Registry will fail with an error. **Artifact logging and retrieval will not be affected.
    3. In Server v0.73 (September), we will remove this flag and migration will happen during the upgrade.

    We’re optimizing for a seamless, minimally-inconveniencing upgrade to a powerful new experience. Reach out to your W&B team or support for any questions about the migration.

• CoreWeave infrastructure monitoring: During a W&B Run on CoreWeave infrastructure, CoreWeave Mission Control monitors your compute infrastructure, allowing for observation of infrastructure alerts such as GPU failures, thermal violations, and more. If an error occurs, CoreWeave sends that information to W&B. W&B populates infrastructure information onto your run’s plots in your project’s workspace. CoreWeave attempts to resolve some issues automatically, and W&B surfaces that information in the run’s page. W&B provides a link to the SLURM job’s Grafana dashboard for system-level details about the run. Learn more at Visualize CoreWeave infrastructure alerts.

• CoreWeave external file tracking: You can now track external files stored in CoreWeave AI Object Storage with reference artifacts. Learn more at Track external files.

• CoreWeave BYOB support: Dedicated Cloud and Self-Managed now support CoreWeave AI Object Storage for Instance and Team level BYOB. CoreWeave Team level BYOB on Multi-tenant Cloud is coming soon. Learn more at Bring your own bucket (BYOB).

• Bulk media settings: Just like line plots, you can now manage all your media panel settings at once—across an entire workspace or a specific section. Easily configure media panels to display by epoch or arrange them into customized grid patterns without adjusting each panel separately. Individual panel settings override the global settings.

• Use Workspace templates to quickly create workspaces using the same settings as an existing workspace. Currently, a workspace template can define custom line plot settings. Learn more at Workspace templates.

• With incremental table logging, you can log batches of rows to a table during a machine learning experiment. This is ideal for monitoring long-running jobs or when working with large tables that would be inefficient to log during the run for updates. Within the UI, the table is updated with new rows as they are logged, so you can view the latest data without having to wait for the entire run to finish. You can step through the increments to view the table at different points in time. Learn more at Log tables.

Fixes

• Added a banner at the top of the log viewer if only a subset of lines is shown.
• Fixed a bug in the running average calculation at data boundaries.
• Clarified x-axis unit labels for relative time plots.
• Fixed a bug in plots with smoothing where the original line would still display when Show Original was unchecked.
• Fixed a bug that could unexpectedly convert a grouped full-fidelity line plot with expressions into a bar chart.
• Hovering on histograms now shows the x-axis step in a tooltip.
• Fixed a bug where data exported from a panel could sort incorrectly. Exported data now sorts numerically.
• In the run selector, you can now filter and sort on updated at.
• Added right-handed system setting option to point cloud panels.
• Trailing slashes are now stripped from full screen panel URLs.
• Fixed a bug where a deleted saved view could reappear if you refreshed immediately after deleting it.
• Fixed a display bug that caused a flash when reloading a page with dark mode turned on.
• Fixed a bug that caused incorrect slider steps when using non-monotonic values.
• Added a convenience checkbox and notification when moving runs between groups.
• Fixed a bug where using the wildcard character * for the metric name in wandb.define_metric() would fail to match metrics with / in the name.
• W&B now no longer deletes source artifacts when collection portfolios are garbage collected.
• Fixed a bug that incorrectly allowed typing input in Allowed types and Registry visibility selection boxes.
• Collection cards now display the full artifact type name instead of truncating it.
• Fixed a bug where clicking Action History in a Registry collection would incorrectly load the Versions view.
• Registry now supports adding job type artifacts.
• Lineage tiles are now wider, displaying more text before truncating.
• Clarified text in project-level automation setup to refer to “artifacts” instead of the Registry terminology “collections”.
• The artifact browser now searches all artifacts when searching the artifact browser by name, rather than the previous limit of the first 500 artifacts.

Patches

0.70.1

July 9, 2025

• Fixed a bug where the status of crashed runs was not updated to Crashed after an upgrade.
• Fixed a bug where the State column was missing from the Sweeps Runs table.
• Fixed a Weave bug where the browser for a Registry collection could display files from an incorrect version because the cached path for an artifact manifest file depended only on the artifact’s version index.
• Fixed a Weave bug that could prevent JSON with valid syntax from being parsed correctly.

These bugs were introduced in v0.70.0.

1.2 - 0.69.x

W&B 0.69 focuses on making the workspace more intuitive, collaborative, and efficient. Clearer visualizations and faster artifact downloads streamline how you interact with your data, so you can gain and share insights more quickly. Updates to Weave improve team workflows and evaluation tracking. A range of quality-of-life fixes tidy up the overall user experience.

This release also marks the end of life for v0.54 and older, which are now officially unsupported.

The latest patch is v0.69.1.

Refer to Patches.

Support and end of life

• W&B Server v0.54 and below have reached end of life as of May 27, 2025.
• W&B Server v0.56 is scheduled to reach end of life in July 2025

A W&B Server release is supported for 12 months from its initial release date. As a reminder, customers using Self-managed are responsible to upgrade to a supported release in time to maintain support.

Refer to Release policies and processes. For assistance or questions, contact support.

Upgrading

To upgrade to W&B v0.69.x, you must use v0.31.4+ of the operator-wandb Helm chart. Otherwise, after the upgrade, the weave-cache-clear container can fail to start. Ensure that your deployment uses these values:

chart:
  url: https://charts.wandb.ai
  name: operator-wandb
  version: 0.31.4

If you have questions or are experiencing issues with an upgrade, contact support.

Features

• You can now set a custom display name for a run directly in the workspace. Customized run names show up in all plots and tables but only in your workspace, with no impact on your teammates’ views. This provides a clearer and cleaner view in your workspace, with no more labels like *...v6-final-restart...* in every legend and plot.
• When filtering or grouping runs, colors can sometimes overlap and become indistinct. The run selector’s new Randomize Colors option reassigns random colors from the default palette to your current run selection or groups, helping to make the colors more distinguishable.
• In line plots, you can now use Cmd+Click on a line to open a single-run view in a new tab.
• Video media panels now provide more playback controls to play, pause, seek, view full screen, and adjust playback speed.
• Settings for all types of media panels have been reorganized and improved.
• You can now customize the point and background colors for point cloud panels.
• Team-level and organization-level service accounts can now interact with Registry.
• Improved Exponentially-weighted Moving Average (EMA) smoothing provides more reliable smoothed lines when operating on complete, unbinned data. In most cases, smoothing is handled at the back end for improved performance. This feature was in private preview in v0.68.x.

Private preview

Private preview features are available by invitation only. To request enrollment in a private preview, contact support or your AISE.

• You can now color all of your runs based on a secondary metric, such as loss or custom efficiency metrics. This creates a clear gradient color scale across your runs in all plots, so you can spot patterns faster. Watch a video demo.
• Personal workspace templates allow you to save core line plot settings and automatically reapply them in new views. These settings include x-axis key, smoothing algorithm, smoothing factor, max number of lines, whether to use the run selector’s grouping, and which aggregation to apply.

Weave

• Saved views simplify team collaboration and allow you to persist filter and column settings.
• PDFs and generic files are now supported.
• The new EvaluationLogger API provides flexible imperative-style evaluation logging.
• You can now import human annotations into Weave datasets
• Playground now supports saved configurations and prompts.
• Decorators are now supported in TypeScript.
• Added support for tracing generator functions.
• The new dataset.add_rows helper improves the efficiency of appending to an existing dataset.
• To help you understand your usage, trace and object sizes are now shown through the UI.

Performance

• With wandb SDK v0.19.11, artifacts now download 3-5x faster on average. For example, an artifact that previously downloaded at around 100 MB/sec may now download at 450 MB/sec or faster. Actual download speeds vary based on factors such as your network and storage infrastructure.
• Improved caching on Project and User Settings pages.

Fixes

• Improved the startup process for the weave-cache-clear container to ensure compatibility with Python virtual environments.
• Added options for denser display of console logs.
• Workspace loading screens are now more informative.
• When adding a panel from a workspace to a report, the current project’s reports are now shown first in the destination report list.
• Fixed many cases where y-axes would over-round to a degree that caused duplicate values to display.
• Fixed confusing behavior when entering invalid smoothing parameters.
• Removed the Partial Media warning from media panels. This does not change the behavior of the media panels.
• When adding a run filter based on tags, the filter is now selected by default, as when filtering by other fields.
• Removed the green bell icon that could appear on active runs in the run selector.
• Removed the System page for individual runs.
• The project description field now respects new lines.
• Fixed URLs for legacy model registry collections.
• Fixed a bug where the Netron viewer did not expand to fill all available space on the page.
• When you click Delete on a project, the project name now displays in the confirmation modal.

Patches

0.69.1

June 10, 2025

• You can now set the initial run state when creating a run with Run.create() by setting the state parameter to pending or running.
• Fixed a bug where clicking Action History incorrectly loaded the Version view.
• Improved memory performance of the Parquet store service.

1.3 - 0.68.x

W&B Server v0.68 includes enhancements to various types of panels and visualizations, security improvements for Registry, Weave, and service accounts, performance improvements when forking and rewinding runs, and more.

The latest patch is v0.68.2.

Refer to Patches.

Features

• Release notes for W&B Server are now published in the W&B documentation in addition to on GitHub. Subscribe using RSS.
• Registry admins can define and assign protected aliases to represent key stages of your development pipeline. A protected alias can be assigned only by a registry admin. W&B blocks other users from adding or removing protected aliases from versions in a registry using the API or UI.
• You can now filter console logs based on a run’s x_label value. During distributed training, this optional parameter tracks the node that logged the run.
• You can now move runs between Groups, one by one or in bulk. Also, you can now create new Groups after the initial logging time.
• Line plots now support synchronized zooming mode, where zooming to a given range on one plot automatically zooms into the same range on all other line plots with a common x-axis. Turn this on in the workspace display settings for line plots.
• Line plots now support formatting custom metrics as timestamps. This is useful when synchronizing or uploading runs from a different system.
• You can now slide through media panels using non-_step fields such as epoch or train/global_step (or anything else).
• In Tables and plots in Query Panels that use runs or runs.history expressions, a step slider allows you to step through the progress on your metrics, text, or media through the course of your runs. The slider supports stepping through non-_step metrics.
• You can now customize bar chart labels using a font size control.

Private preview

Private preview features are available by invitation only. To request enrollment in a private preview, contact support or your AISE.

• Personal workspace templates allow you to save your workspace setup so it is automatically applied to your new projects. Initially, you can configure certain line plot settings such as the default X axis metric, smoothing algorithm, and smoothing factor.
• Improved Exponentially-weighted Moving Average (EMA) smoothing provides more reliable smoothed lines when operating on complete, unbinned data. In most cases, smoothing is handled at the back end for improved performance.

Weave

• Chat with fine-tuned models from within your W&B instance. Playground is now supported in Dedicated Cloud. Playground is a chat interface for comparing different LLMs on historical traces. Admins can add API keys to different model providers or hook up custom hosted LLM providers so your team can interact with them from within Weave.
• Open Telemetry Support. Now you can log traces via OpenTelemetry (OTel). Learn more.
• Weave tracing has new framework integrations: CrewAI, OpenAI’s Agent SDK, DSPy 2.x and Google’s genai Python SDK.
• Playground supports new OpenAI models: GPT‑4.1, GPT‑4.1 mini, and GPT‑4.1 nano.
• Build labeled datasets directly from traces, with your annotations automatically converted into dataset columns. Learn more.

Security

• Registry admins can now designate a service account in a registry as either a Registry Admin or a Member. Previously, the service account’s role was always Registry Admin. Learn more.

Performance

• Improved the performance of many workspace interactions, particularly in large workspaces. For example, expanding sections and using the run selector are significantly more responsive.

• Improved Fork and Rewind Performance.

  Forking a run creates a new run that uses the same configuration as an existing run. Changes to the forked run do not the parent run, and vice versa. A pointer is maintained between the forked run and the parent. Rewinding a run lets you log new data from that point in time without losing the existing data.

  In projects with many nested forks, forking new runs is now much more efficient due to improvements in caching.

Fixes

• Fixed a bug that could prevent an organization service account from being added to new teams.
• Fixed a bug that could cause hover marks to be missing for grouped lines.
• Fixed a bug that could include invalid project names in the Import dropdown of a Report panel.
• Fixed a display bug in the alignment of filters in the run selector.
• Fixed a page crash when adding a timestamp Within Last filter
• Fixed a bug that could prevent the X-axis from being set to Wall Time in global line plot settings.
• Fixed a bug that could prevent image captions from appearing when they are logged to a Table.
• Fixed a bug that could prevent sparse metrics from showing up in panels.
• In Run Overview pages, the Description field is now named Notes.

Patches

0.68.1

May 2, 2025

• Fixed a bug introduced in v0.68.0 that could prevent media from loading in media panels.

0.68.2

May 7, 2025

• Fixed a bug introduced in v0.68.0 that could cause background jobs to crash or run inconsistently. After upgrading to v0.68.2, affected background jobs will recover automatically. If you experience issues with background jobs after upgrading, contact Support.
• Fixed a long-standing UI bug where typing an invalid regular expression into the W&B App search field could crash the app. Now if you type an invalid regular expression, it is treated as a simple search string, and you can update the search field and try again.
• Fixed a bug where the SMTP port is set to 25 instead of the port specified in GORILLA_EMAIL_SINK.
• Fixed a bug where inviting a user to a team could fail with the misleading error You have no available seats.

1.4 - 0.67.x

Features

• In Reports, you can now give a run a custom display name per panel grid. This allows you to replace the run’s (often long and opaque) training-time name with one that is more meaningful to your audience. The report updates the name in all panel grids, helping you to explain your hard-won experimental insights to your colleagues in a concise and readable way. The original run name remain intact in the project, so doing this won’t disrupt your collaborators.
• When you expand a panel in the workspace, it now opens in full screen mode with more space. In this view, line plots now render with more granular detail, using up 10,000 bins. The run selector appear next to the panel, letting you easily toggle, group, or filter runs in context.
• From any panel, you can now copy a unique URL that links directly to that panel’s full screen view. This makes it even easier to share a link to dig into interesting or pathological patterns in your plots.
• Run Comparer is a powerful tool you can use to compare the configurations and key metrics of important runs alongside their loss curves. Run Comparer has been updated:
  • Faster to add a Run Comparer panel, as an expanded option in Add Panels.
  • By default, a Run Comparer panel takes up more space, so you can see the values right away.
  • Improved readability and legibility of a Run Comparer panel. You can use new controls to quickly change row and column sizes so you can read long or nested values.
  • You can copy any value in the panel to your clipboard with a single click.
  • You can search keys with regular expressions to quickly find exactly the subset of metrics you want to compare across. Your search history is saved to help you iterate efficiently between views.
  • Run Comparer is now more reliable at scale, and handles larger workspaces more efficiently, reducing the likelihood of poor performance or a crashed panel.

• Segmentation mask controls have been updated:
  • You can now toggle each mask type on or off in bulk, or toggle all masks or all images on or off.
  • You can now change each class’s assigned color, helping to avoid confusion if multiple classes use the same color.
• When you open a media panel in full screen mode, you can now use the left or right arrows on your keyboard to step through the images, without first clicking on the step slider.
• Media panels now color run names, matching the run selector. This makes it easier to associate a run’s media values with related metrics and plots.
• In the run selector, you can now filter by whether a run has certain media key or not.
• You can now move runs between groups in the W&B App UI, and you can create new groups after the run is logged.
• Automations can now be edited in the UI
• An automation can now notify a Slack channel for artifact events. When creating an automation, select “Slack notification” for the Action type.
• Registry now supports global search by default, allowing you to search across all registries by registry name, collection name, alias, or tag.
• In Tables and Query panels that use the runs expression, you can use the new Runs History step slider and drop-down controls to view a table of metrics at each step of a run.
• Playground in Weave supports new models: OpenAI’s gpt-4.5-preview and Deepseek’s deepseek-chat and deepseek-reasoner.
• Weave tracing has two new agent framework integrations: CrewAI and OpenAI’s Agent SDK.
• In the Weave UI, you can now build Datasets from traces. Learn more: https://weave-docs.wandb.ai/guides/core-types/datasets#create-edit-and-delete-a-dataset-in-the-ui
• The Weave Python SDK now provides a way to filter the inputs and outputs of your Weave data to ensure sensitive data does not leave your network perimeter. You can configure to redact sensitive data. Learn more: https://weave-docs.wandb.ai/guides/tracking/redact-pii/
• To streamline your experience, the System tab in the individual run workspace view will be removed in an upcoming release. View full information about system metrics in the System section of the workspace. For questions, contact support@wandb.com.

Security

• golang crypto has been upgraded to v0.36.0.
• golang oauth2 has been upgraded to v0.28.0.
• In Weave, pyarrow is now pinned to v17.0.0.

Performance

• Frontend updates significantly reduce workspace reload times by storing essential data in the browser cache across visits. The update optimizes loading of saved views, metric names, the run selector, run counts, W&B’s configuration details, and the recomputation of workspace views.
• Registry overview pages now load significantly faster.
• Improved the performance of selecting metrics for the X, Y, or Z values in a scatter plot in a workspace with thousands of runs or hundreds of metrics.
• Performance improvements to Weave evaluation logging.

Fixes

• Fixed a bug in Reports where following a link to a section in the report would not open to that section.
• Improved the behavior of how Gaussian smoothing handles index reflection, matching SciPy’s default “reflect” mode.
• A Report comment link sent via email now opens directly to the comment.
• Fixed a bug that could crash a workspace if a sweep takes longer than 2 billion compute seconds by changing the variable type for sweep compute seconds to int64 rather than int32.
• Fixed display bugs that could occur when a report included multiple run sets.
• Fixed a bug where panels Quick Added to an alphabetically sorted section were sorted incorrectly.
• Fixed a bug that generated malformed user invitation links.

1.5 - 0.66.x

Features

• In tables and query panels, columns you derive from other columns now persist, so you can use them for filtering or in query panel plots.

Security

• Limited the maximum depth for a GraphQL document to 20.
• Upgraded pyarrow to v17.0.0.

1.6 - 0.65.x

Features

• From a registry’s Settings, you can now update the owner to a different user with the Admin role. Select Owner from the user’s Role menu.
• You can now move a run to a different group in the same project. Hover over a run in the run list, click the three-vertical-dots menu, and choose Move to another group.
• You can now configure whether the Log Scale setting for line plots is enabled by default at the level of the workspace or section.
  • To configure the behavior for a workspace, click the action ... menu for the workspace, click Line plots, then toggle Log scale for the X or Y axis.
  • To configure the behavior for a section, click the gear icon for the section, then toggle Log scale for the X or Y axis.

1.7 - 0.63.x

Features

Weave is now generally available (GA) in Dedicated Cloud on AWS. Reach out to your W&B team if your teams are looking to build Generative AI apps with confidence and putting those in production.

The release includes the following additional updates:

• W&B Models now seamlessly integrates with Azure public cloud. You could now create a Dedicated Cloud instance in an Azure region directly from your Azure subscription and manage it as an Azure ISV resource. This integration is in private preview.
• Enable automations at the Registry level to monitor changes and events across all collections in the registry and trigger actions accordingly. This eliminates the need to configure separate webhooks and automations for individual collections.

• Ability to assign x_label, e.g. node-0, in run settings object to distinguish logs and metrics by label, e.g. node, in distributed runs. Enables grouping system metrics and console logs by label for visualization in the workspace.
• Coming soon with a patch release this week, you will be able to use organization-level service accounts to automate your W&B workloads across all teams in your instance. You would still be able to use existing team-level service accounts if you would like more control over the access scope of a service account.
  • Allow org-level service accounts to interact with Registry. Such service accounts can be invited to a registry using the invite modal and are displayed in the members table along with respective organization roles.

Fixes

• Fixed an issue where users creating custom roles including the Create Artifact permission were not able to log artifacts to a project.
• Fixed the issue with metadata logging for files in instances that have subpath support configured for BYOB.
• Block webhook deletion if used by organization registry automations.

1.8 - 0.61.0

Features

This is a mini-feature and patch release, delivered at a different schedule than the monthly W&B server major releases

• Organization admins can now configure Models seats and access control for both Models & Weave in a seamless manner from their organization dashboard. This change allows for a efficient user management when Weave is enabled for a Dedicated Cloud or Self-managed instance.
  • Weave pricing is consumption-based rather than based on number of seats used. Seat management only applies to the Models product.
• You can now configure access roles at the project level for team and restricted scoped projects. It allows assigning different access roles to a user within different projects in the same team, and thus adding another strong control to conform to enterprise governance needs.

Fixes

• Fixed an issue where underlying database schema changes as part of release upgrades could timeout during platform startup time.
• Added more performance improvements to the underlying parquet store service, to further improve the chart loading times for users. Parquet store service is only available on Dedicated Cloud, and Self-managed instances based on W&B kubernetes operator.
• Addressed the high CPU utilization issue for the underlying parquet store service, to make the efficient chart loading more reliable for users. Parquet store service is only available on Dedicated Cloud, and Self-managed instances based on W&B kubernetes operator.

1.9 - 0.60.0

Features

• Final updates for 1.1.1 Compliance of Level AA 2.2 for Web Content Accessibility Guidelines (WCAG) standards.
• W&B can now disable auto-version-upgrade for customer-managed instances using the W&B kubernetes operator. You can request this to your W&B team.
  • Note that W&B requires all instances to upgrade periodically to comply with the 6-month end-of-life period for each version. W&B does not support versions older than 6 months.

Fixes

• Fixed a bug to allow instance admins on Dedicated Cloud and Customer-managed instances to access workspaces in personal entities.
• SCIM Groups and Users GET endpoints now filter out service accounts from the responses. Only non service account users are now returned by those endpoints.
• Fixed a user management bug by removing the ability of team admins to simultaneously delete a user from the overall instance while deleting them from a team. Instance or Org admins are responsible to delete a user from the overall instance / organization.

Performance improvements

• Reduced the latency when adding a panel by up to 90% in workspaces with many metrics.
• Improved the reliability and performance of parquet exports to blob storage when runs are resumed often.
  • Runs export to blob storage in parquet format is available on Dedicated Cloud and on Customer-managed instances that are enabled using the W&B kubernetes operator.

1.10 - 0.58.1

Features

• W&B now supports sub-path for Secure storage connector i.e. Bring your own bucket capability. You can now provide a sub-path when configuring a bucket at the instance or team level. This is only available for new bucket configurations and not for existing configured buckets.
• W&B-managed storage on newer Dedicated Cloud instances in GCP & Azure will by default be encrypted with W&B managed cloud-native keys. This is already available on AWS instances. Each instance storage is encrypted with a key unique to the instance. Until now, all instances on GCP & Azure relied on default cloud provider-managed encryption keys.
• Makes the fields in the run config and summary copyable on click.
• If you’re using W&B kubernetes operator for a customer-managed instance, you can now optionally use a custom CA for the controller manager.
• We’ve modified the W&B kubernetes operator to run in a non-root context by default, aligning with OpenShift’s Security Context Constraints (SCCs). This change ensures smoother deployment of customer-managed instances on OpenShift by adhering to its security policies.

Fixes

• Fixed an issue where exporting panels from a workspace to a report now correctly respects the panel search regex.
• Fixed an issue where setting GORILLA_DISABLE_PERSONAL_ENTITY to true was not disabling users from creating projects and writing to existing projects in their personal entities.

Performance improvements

• We have significantly improved performance and stability for experiments with 100k+ logged points. If you’ve a customer-managed instance, this is available if the deployment is managed using the W&B kubernetes operator.
• Fixed issue where saving changes in large workspaces would be very slow or fail.
• Improved latency of opening workspace sections in large workspaces.

1.11 - 0.57.2

Features

You can now use JWTs (JSON Web Tokens) to access your W&B instance from the wandb SDK or CLI, using the identity federation capability. The feature is in preview. Refer to Identity federation and reach out to your W&B team for any questions.

The 0.57.2 release also includes these capabilities:

• New Add to reports drawer improvements for exporting Workspace panels into Reports.
• Artifacts metadata filtering in the artifact project browser.
• Pass in artifact metadata in webhook payload via ${artifact_metadata.KEY}.
• Added GPU memory usage panels to the RunSystemMetrics component, enhancing GPU metrics visualization for runs in the app frontend.
• Mobile users now enjoy a much smoother, more intuitive Workspace experience.
• If you’re using W&B Dedicated Cloud on GCP or Azure, you can now enable private connectivity for your instance, thus ensuring that all traffic from your AI workloads and optionally browser clients only transit the cloud provider private network. Refer to Private connectivity and reach out to your W&B team for any questions.
• Team-level service accounts are now shown separately in a new tab in the team settings view. The service accounts are not listed in the Members tab anymore. Also, the API key is now hidden and can only be copied by team admins.
• Dedicated Cloud is now available in GCP’s Seoul region.

Fixes

• Gaussian smoothing was extremely aggressive on many plots.
• Fixed issue where pressing the Ignore Outliers in Chart Scaling button currently has no effect in the UI workspace.
• Disallow inviting deactivated users to an organization.
• Fixed an issue where users added to an instance using SCIM API could not onbioard successfully.

Performance improvements

• Significantly improved performance when editing a panel’s settings and applying the changes.
• Improved the responsiveness of run visibility toggling in large workspaces.
• Improved chart hovering and brushing performance on plots in large workspaces.
• Reduced workspace memory usage and loading times in workspaces with many keys.

1.12 - 0.56.0

Features

The new Full Fidelity line plot in W&B Experiments enhances the visibility of training metrics by aggregating all data along the x-axis, displaying the minimum, maximum, and average values within each bucket, allowing users to easily spot outliers and zoom into high-fidelity details without downsampling loss. Learn more in our documentation.

The 0.56.0 release also includes these capabilities:

• You can now use cross-cloud storage buckets for team-level BYOB (secure storage connector) in Dedicated Cloud and Self-managed instances. For example, in a W&B instance on AWS, you can now configure Azure Blob Storage or Google Cloud Storage for team-level BYOB, and so on for each cross-cloud combination.
• In the same vein, you can now use S3-compatible storage buckets like MinIO for team-level BYOB (secure storage connector) in Dedicated Cloud and Self-managed instances. For example, in a W&B instance on GCP, you can configure a MinIO bucket hosted in cloud or on-prem for team-level BYOB.
• Admins can now automate full deletion of users in their Dedicated Cloud or Self-managed instances using the SCIM API’s DELETE User endpoint. The user deactivation operation has been reimplemented using the PATCH User endpoint, along with the introduction of user reactivation operation.
• If you use the SCIM API, you will also see a couple of minor improvements:
  • The API now has a more pertinent error message in case of authentication failures.
  • Relevant endpoints now return the full name of a user in the SCIM User object if it’s available.

Fixes

• The fix resolves an issue where deleting a search term from a runset in a report could delete the panel or cause the report to crash by ensuring proper handling of selected text during copy/paste operations.
• The fix addresses a problem with indenting bulleted items in reports, which was caused by an upgrade of slate and an additional check in the normalization process for elements.
• The fix resolves an issue where text could not be selected from a panel when the report was in edit mode.
• The fix addresses an issue where copy-pasting an entire panel grid in a Report using command-c was broken.
• The fix resolves an issue where report sharing with a magic link was broken when a team had the Hide this team from all non-members setting enabled.
• The fix introduces proper handling for restricted projects by allowing only explicitly invited users to access them, and implementing permissions based on project members and team roles.
• The fix allows instance admins to write to their own named workspaces, read other personal and shared workspaces, and write to shared views in private and public projects.
• The fix resolves an issue where the report would crash when trying to edit filters due to an out-of-bounds filter index caused by skipping non-individual filters while keeping the index count incremental.
• The fix addresses an issue where unselecting a runset caused media panels to crash in a report by ensuring only runs in enabled runsets are returned.
• The fix resolves an issue where the parameter importance panel crashes on initial load due to a violation of hooks error caused by a change in the order of hooks.
• The fix prevents chart data from being reloaded when scrolling down and then back up in small workspaces, enhancing performance and eliminating the feeling of slowness.

1.13 - Archived Releases

Archived releases have reached end of life and are no longer supported. A major release and its patches are supported for 12 months from the initial release date. Release notes for archived releases are provided for historical purposes. For supported releases, refer to Releases.

1.13.1 - 0.54.0

Features

• You can now configure Secure storage connector (BYOB) at team-level in Dedicated Cloud or Self-managed instances on Microsoft Azure.
  • Refer to this Terraform Module and reach out to your W&B team for any questions.
• Organization admins can now enforce privacy settings across all W&B teams by setting those at the organization level, from within the Settings tab in the Organization Dashboard.
  • W&B recommends to notify team admins and other users before making such enforcement changes.
• Enable direct lineage option for artifact lineage DAG
• It’s now possible to restrict Organization or Instance Admins from self-joining or adding themselves to a W&B team, thus ensuring that only Data & AI personas have access to the projects within the teams.
  • W&B advises to exercise caution and understand all implications before enabling this setting. Reach out to your W&B team for any questions.
• Dedicated Cloud on AWS is now also available in the Seoul (S. Korea) region.

Fixes

• Fix issue where Reports where failing to load on Mobile.
• Fix link to git diff file in run overview.
• Fixed the intermittently occurring issue related to loading of Organization Dashboard for certain users.

1.13.2 - 0.52.2

Features

• You can now enforce username and full name for users in your organization, by using OIDC claims from your SSO provider. Reach out to your W&B team or support if interested.
• You can now disable use of personal projects in your organization to ensure that all projects are created within W&B teams and governed using admin-enforced guidelines. Reach out to your W&B team or support if interested.
• Option to expand all versions in a cluster of runs or artifacts in the Artifacts Lineage DAG view.
• UI improvements to Artifacts Lineage DAG - the type will now be visible for each entry in a cluster.

Fixes

• Added pagination to image panels in media banks, displaying up to 32 images per page with enhanced grid aesthetics and improved pagination controls, while introducing a workaround for potential offset inconsistencies.
• Resolved an issue where tooltips on system charts were not displaying by enforcing the isHovered parameter, which is essential for the crosshair UI visibility.
• Unset the max-width property for images within media panels, addressing unintended style constraints previously applied to all images.
• Fixed broken config overrides in launch drawer.
• Fixed Launch drawer’s behavior when cloning from a run.

1.13.3 - 0.51.0

Features

You can now save multiple views of any workspace by clicking “Save as a new view” in the overflow menu of the workspace bar.

Learn more about how Saved views can enhance your team’s collaboration and project organization.

The release also includes these capabilities:

• You can now set a project’s visibility scope to Restricted if you want to collaborate on AI workflows related to sensitive or confidential data.
  • When you create a restricted project within a team, you can add specific members from the team. Unlike other project visibility scopes, all members of a team do not get implicit access to a restricted project.
• Enhanced Run Overview page performance: now 91% faster on load, with search functionality improved by 99.9%. Also enjoy RegEx search for Config and Summary data.
• New UX for Artifacts Lineage DAG introduces clustering for 5+ nodes at the same level, preview window to examine a node’s details, and a significant speedup in the graph’s loading time.
• The template variable values used for a run executed by launch, for example GPU type and quantity, are now shown on the queue’s list of runs. This makes it easier to see which runs are requesting which resources.
• Cloning a run with Launch now pre-selects the overrides, queue, and template variable values used by the cloned run.
• Instance admins will now see a Teams tab in the organization dashboard. It can be used to join a specific team when needed, whether it’s to monitor the team activity as per organizational guidelines or to help the team when team admins are not available.
• SCIM User API now returns the groups attribute as part of the GET endpoint, which includes the id of the groups / teams a user is part of.
• All Dedicated Cloud instances on GCP are now managed using the new W&B Kubernetes Operator. With that, the new Parquet Store service is also available.
  • Parquet store allows performant & cost efficient storage of run history data in parquet format in the blob storage. Dedicated Cloud instances on AWS & Azure are already managed using the operator and include the parquet store.
• Dedicated Cloud instances on AWS have been updated to use the latest version of the relational data storage, and the compute infrastructure has been upgraded to a newer generation with better performance.

Advanced Notice: We urge all customers who use Webhooks with Automations to add a valid A-record for their endpoints as we are going to disallow using IP address based Webhook URLs from the next release onwards. This is being done to protect against SSRF vulnerability and other related threat vectors.

Fixes

• Fixed issue where expressions tab was not rendering for line plots.
• Use display name for sweeps when grouped by sweeps in charts and runs table.
• Auto navigation to runs page when selecting job version.

1.13.4 - 0.50.2

Feature

• Add panel bank setting to auto-expand search results
• Better visibility for run queue item issues
• Dedicated Cloud customers on AWS can now use Privatelink to securely connect to their deployments.
  • The feature is in private preview and will be part of an advanced pricing tier at GA. Reach out to your W&B team if interested.
• You can now automate user role assignment for organization or team scopes using the SCIM role assignment API
• All Dedicated Cloud instances on AWS & Azure are now managed using the new W&B Kubernetes Operator. With that, the new Parquet Store service is also available. The service allows for performant & cost efficient storage of run history data in parquet format in the blob storage. That in turn leads to faster loading of relevant history data in charts & plots that are used to evaluate the runs.
• W&B Kubernetes Operator and along with that the Parquet Store service are now available for use in customer-managed instances. We encourage customers that already use Kubernetes to host W&B, to reach out to their W&B team on how they can use the operator. And we highly recommend others to migrate to Kubernetes in order to receive the latest performance improvements and new services in future via operator. We’re happy to assist with planning such a migration.

Fixes

• Properly pass template variables through sweep scheduler
• Scheduler polluting sweep yaml generator
• Display user roles correctly on team members page when search or sort is applied
• Org admins can again delete personal projects in their Dedicated Cloud or Self-managed server instance
• Add validation for SCIM GET groups API for pending users

1.13.5 - 0.49.0

Feature

• Set a default TTL (time-to-live or scheduled deletion) policy for a team in the team settings page.
  • Restrict setting or editing of a TTL policy to either of admin only or admin plus members.
• Test and debug a webhook during webhook creation or after in the team settings UI.
  • W&B will send a dummy payload and display the receiving server’s response.
• View Automation properties in the View Details slider.
  • This includes a summary of the triggering event and action, action configs, creation date, and a copy-able curl command to test webhook automations.
• Replace agent heartbeat with last successful run time in launch overview.
• Service accounts can now use the Report API to create reports.
• Use the new role management API to automate managing the custom roles.
• Enable Kubernetes Operator for Dedicated Cloud deployments on AWS.
• Configure a non-conflicting IP address range for managed cache used in Dedicated Cloud deployments on GCP.

Fixes

• Update the add runset button clickable area in reports
• Show proper truncate grouping message
• Prevent flashing of publish button in reports
• Horizonal Rule get collapsed in report section
• Add section button hidden in certain views
• Allow things like semantic versioning in the plot as string
• Remove requirements for quotes when using template variables in queue config definitions
• Improve Launch queue sorting order
• Don’t auto-open panel sections when searching large workspaces
• Change label text for grouped runs
• Open/close all sections while searching

1.13.6 - 0.48.0

Feature

• All required frontend changes for launch prioritization
  • Refer to this blog on how you can run more important jobs than others using Launch.
• Refer to below changes for access control and user attribution behavior of team service accounts:
  • When a team is configured in the training environment, a service account from that team can be used to log runs in either of private or public projects within that team, and additionally attribute those runs to a user only if WANDB_USERNAME or WANDB_USER_EMAIL variable is configured in the environment and the user is part of that team.
  • When a team is not configured in the training environment and a service account is used, the runs would be logged to the named project within the service account’s team, and those would be attributed to a user only if WANDB_USERNAME or WANDB_USER_EMAIL variable is configured in the environment and the user is part of that team.
  • A team service account can not log runs in a private project in another team, but it can log runs to public projects in other teams.

Fixes

• Reduce column widths for oversized runs selectors
• Fix a couple of bugs related to Custom Roles preview feature

1.13.7 - 0.47.3

Fixes

We’re releasing a couple of important fixes for the Custom Roles preview capability that was launched as part of v0.47.2. If you’re interested in using that feature to create fine-grained roles and better align with least privilege principle, please use this latest server release and reach out to your Weights & Biases team for an updated enterprise license.

1.13.8 - 0.47.2

Feature

Use custom roles with specific permissions to customize access control within a team

• Available in preview to enterprise customers. Please reach out to your Weights & Biases account team or support for any questions.

Also:

• Minor runs search improvements
• Auto-resize runs search for long texts
• View webhook details, including URL, secrets, date created directly from the automations table for webhook automations

Fixes

• Grouping of runs when group value is a string that looks like a number
• Janky report panel dragging behavior
• Update bar chart spec to match the one on public cloud
• Clean up panel padding and plot margins
• Restores workspace settings beta

1.13.9 - 0.46.0

Features

• Deployments on AWS can now use W&B Secrets with Webhooks and Automations
  • Secrets are stored securely in AWS Secret Manager - please use the terraform-aws-wandb module to provision one and
• Update webhooks table to display more information
• Better truncation of long strings to improve the usability of strings in the UI
• Reduce delay for scroll to report section
• Add white background to weave1 panels
• Allow deep link for weave1 panels in reports
• Allow weave1 panel resizing in reports
• Homepage banner will now show CLI login instructions
• User invites will now succeed if invite email can’t be sent for some reason
• Add list of associated queues to agent overview page

Fixes

• Copy function on panel overlay was dropping values
• CSS cleanup for import modal when creating report
• Fixes regression to hide legend when toggled off
• Report comment highlighting
• Remove all caching for view’s LoadMetadataList()
• Let run search stretch
• Associate launch agents with user id from X-WANDB-USERNAME header

1.13.10 - 0.45.0

Features

• Enable artifact garbage collection using environment variable GORILLA_ARTIFACT_GC_ENABLED=true and cloud object versioning or soft deletion.
• The terraform module terrraform-azurerm-wandb now supports Azure Key Vault as a secrets store.
  • Deployments on Azure can now use W&B Secrets with Webhooks and Automations. Secrets are stored securely in Azure Key Vault.

Fixes

• Remove invalid early exit preventing history deletion
• When moving/copying runs, don’t drop key-set info
• Update mutations to no longer use defunct storage plan or artifacts billing plan at all
• Respect skip flag in useRemoteServer

1.13.11 - 0.44.1

Features

Add OpenAI proxy UI to SaaS and Server

Also:

• New version v1.19.0 of our AWS Terraform module terraform-google-wandb is available
• Add support for AWS Secret Manager for Customer Secret Store, which can be enabled after the terraform module terrraform-aws-wandb is updated and released
• Add support for Azure Key Vault for Customer Secret Store, which can be enabled after the terraform module terrraform-azurerm-wandb is updated and released

Fixes

• Quality-of-life improvements in the model registry ui
• int values no longer ignored when determining if a run achieved a sweep’s optimization goal
• Cache runs data to improve workspace loading perf
• Improve TTL rendering in collection table
• Allow service accounts to be made workflow (registry) admins
• Add tooltip for truncated run tags in workspaces
• Fix report page scrolling
• Copy y data values for chart tooltip
• Query secrets for webhooks in local
• Fixing broken domain zoom in panel config
• Hide Customer Secret Store UI if GORILLA_CUSTOMER_SECRET_STORE_SOURCE env var not set

Chores

• Bump langchain to latest
• Adding WB Prompts to quickstart
• Update AWS MIs to use terraform-kubernetes-wandb v1.12.0
• Show correct Teams Plan tracked hours teams settings page and hide on usage page

1.13.12 - 0.43.0

Release 0.43.0 contains a number of minor bug fixes and performance improvements, including fixing the bottom of runs tables when there’s a scrollbar. Check out the other fixes below:

Fixes

• Dramatically improve workspace loading perf
• Fixing broken docs link in disabled add panel menu
• Render childPanel without editor in report
• Copying text from a panel grid when editing
• Run overview crashing with ’length’ key
• Padding for bottom of runs table when there’s a scrollbar
• Eliminate unnecessary history key cache read
• Error handling for Teams Checkout modal
• Memory leak, excess filestream sending, and orphaned processes in Weave Python autotracer

1.13.13 - 0.42.0

Features

W&B Artifacts now supports time-to-live (TTL) policies

Users can now gain more control over deleting and retention of Artifacts logged with W&B, with the ability to set retention and time-to-live (TTL) policies! Determine when you want specific Artifacts to be deleted, update policies on existing Artifacts, and set TTL policies on upstream or downstream Artifacts.

Here are the other new features include in this release:

• Use Launch drawer when creating Sweeps
• Delete run queue items
• Min/max aggregations nested dropdown
• Allow users to connect multiple S3-compatible buckets
• Add disk i/o system metrics
• Use the legacy way to set permissions
• Enable CustomerSecretStore
• Add Kubernetes as a backend for CustomerSecretStore

Fixes

• Disable storage and artifact invoices for ongoing storage calculations refractors
• Panel deletion bug
• Remove link-version event type from project automation slider
• Remove upper case styling for artifact type names
• Keep uncolored tags from changing color on render
• Stale defaults stuck in Launch drawer on reopen
• Trigger alias automations while creating artifact
• Edge case failure in infinite loading tag filters

1.13.14 - 0.41.0

Features

New Launch landing page

We’ve updated the Launch home page, so users looking to get started with Launch will have a much easier way to get setup quickly. Easily access detailed documentation, or simply follow the three Quickstart steps to create a Launch queue, agent, and start launching jobs immediately.

Here are the other new features included in this release:

• Add new reverse proxy to track OpenAI requests and responses
• Show agent version on agent overview page
• New model registry workflow removed from feature flag for all users

Fixes

• Empty projects causing infinite load on storage explorer
• Runs marked failed when run queue items are failed
• Use correct bucket for storing OpenAI proxy artifacts
• SEO tags not properly rendered by host
• Trigger export in background, on context deadline as well
• Transition runs in pending state to running when run is initialized
• Query so Launch queues show most recent completed and failed jobs

1.13.15 - 0.40.0

Features

Webhooks

Enable a seamless model CI/CD workflow using Webhook Automations to trigger specific actions within the CI/CD pipeline when certain events occur. Use webhooks to facilitate a clean hand-off point between ML engineering and devops. To see this in practice for Model Evaluation and Model Deployment, check out the linked demo videos. Learn more in our docs.

New user activity dashboard on for all customers

Fixes

• Removed limit on number of registered models an organization could have.
• Added search history to workspaces to make it easier to find commonly used plots.
• Changed reports “like” icon from hearts to stars.
• Users can now change the selected run in a workspace view with a group of runs.
• Fixed issue causing duplicate panel grids.
• Users can now pass in per-job resource config overrides for Sweeps on Launch
• Added redirect from /admin/users to new organization dashboard.
• Fixed issues with LDAP dropping connections.
• Improvements to run permadeletion.

1.13.16 - 0.39.0

Features

Revamped Organization Dashboard

We’ve made it easier to see who’s making the most W&B with our overhauled Organization Dashboard, accessible to W&B admins. You can now see details on who’s created runs and reports, who’s actively using W&B, who’s invites are pending–and you can export all this in CSV to share across your organization. Learn more in the docs.

For Dedicated Cloud customers, this feature has been turned on. For Customer-Managed W&B customers, contact W&B support and we’ll be happy to work with you to enable it.

Fixes

• Restrict service API keys to team admins
• Launch agent configuration is now shown on the Agents page
• Added navigation panel while viewing a single Launch job.
• Automations can now show configuration parameters for the associated job.
• Fixed issue with grouped runs not live updating
• Removed extra / in magic and normal link url
• Check base for incremental artifacts
• Inviting a user into multiple teams will no longer take up too many seats in the org

1.13.17 - 0.38.0

Features

Metric visualization enhancements

We’re continuing to enhance our core metric visualization experience. You can now define which metrics from regular expressions to render in your plots, up to 100 metrics at once. And to more accurately represent data at high scale, we’ve add a new time-weighted exponential moving average smoothing algorithm for plots (check out all of our supported algorithms).

Feedback surveys

W&B has always built our product based on customer feedback. Now, we’re happy to introduce a new way for you to shape the future of W&B: in-app feedback surveys in your Dedicated Cloud or Customer-Managed W&B install. Starting July 17th, W&B users will start periodically seeing simple 1 - 10 Net Promoter Score surveys in the application. All identifying information is anonymized. We appreciate all your feedback and look forward to making W&B even better, together.

Fixes

• Major improvement to artifact download speed: over a 6x speedup on our 1-million-file artifact benchmark. Please upgrade to SDK version 0.15.5+.
• (Launch) Optuna is now available as a sweeps scheduler with Sweeps on Launch, allowing more efficient exploration of hyperparameters.
• Run data permadeletion is now available (default off). This can be enabled with the GORILLA_DATA_RETENTION_PERIOD environment variable, specified in hours. Please take care before updating this variable and/or chat with W&B Support, since the deletion is permanent. Artifacts will not be deleted by this setting.
• Updated report sharing emails to include a preview.
• Relaxed HTML sanitation rules for reports in projects; this had been causing rare problems with report rendering.
• Expanded the maximum number of metrics that can be matched by a regex in chart configuration; previously this had been always 10, the maximum is now 100.
• Fixed issue with media panel step slider becoming unsynced with the media shown.
• Added time-weighted exponential moving average as an option for smoothing in plots.
• The “Search panels” textbox in workspaces now preserves the user’s last search.
• Applying a username filter when runs are grouped will no longer error.
• (Launch) The loading of the Launch tab should now be much faster, typically under two seconds.
• (Launch) There’s now an option to edit queue configs using YAML instead of JSON. It’s also now more clear how to edit queue configs.
• (Launch) Runs will now show error messages in the UI when they crash or fail.
• (Launch) If you don’t specify a project when creating a job, we’ll now use the value for WANDB_PROJECT from your wandb.init.
• (Launch) Updated support for custom accelerator images—these will run in noninteractive mode when building, which had been blocking some images.
• (Launch) Fixed issue where the run author for sweeps was the agent service account, rather than the real author
• (Launch) Clicking outside the Launch drawer will no longer close the drawer automatically.
• (Launch) Fixed issue where training jobs that had been enqueued by a sweep but not run yet were not correctly removed from the queue if you later stopped the sweep.
• (Launch) The Launch navigation link is now hidden for users who aren’t part of the team.
• (Launch) Fixed formatting and display issues on Agent logs.
• Fixed scrolling, resizing, and cloning issues in Automations panel.
• Fixed pagination on artifact action history.
• Added support for pre-signed URLs using a VPC endpoint URL if the AWS_S3_ENDPOINT_URL env var is set and passed in from the SDK side.
• Fixed enterprise dashboard link when organization name contains “&”
• Updated tag colors to be consistent

1.13.18 - 0.36.0

Features

Clone Runs with Launch

If you want to repeat a run but tweak a couple hyperparameters–say bump the batch size to take advantage of a larger machine–it’s now easy to clone a run using W&B Launch. Go to the run overview, click Clone, and you’ll be able to select new infrastructure to execute the job on, with new hyperparameters. Learn more in the Launch documentation.

Fixes

• Added report creation and update action to audit logs.
• Artifacts read through the SDK will now be captured in the audit logs.
• In report creation, added button to select all plots to add to the new report
• New view-only users signing up via a report link will now be fast tracked to the report, rather than going through the normal signup process.
• Team admins can now add protected aliases.
• Improved media panel handling of intermediate steps.
• Removed inactive ‘New Model’ button from Model Registry homepage for anonymous users
• Ability to copy data from plot legends has been rolled out to all users.
• Fixed incorrect progress indicator in Model Registry onboarding checklist.
• Fixed issue where the Automations page could crash when job name had slashes.
• Fixed issue where a user could update the wrong user profiles.
• Added option to permanently delete runs and their associated metrics after a duration specified in an environment variable.

1.13.19 - 0.35.0

Security

Fixed issue where API keys were logged for recently logged in users. Check for FetchAuthUserByAPIKey in the logs which you can find in gorilla.log from a debug bundle and rotate any keys that are found.

Features

Launch Agent Logs Now in the GUI

W&B Launch allows you to push machine learning jobs to a wide range of specialized compute environments. With this update, you can now use W&B to monitor and debug jobs running in these remote environments, without needing to log into your AWS or GCP console.

Fixes

• Logs tab is no longer trimmed to 1000 rows.
• Fixed scenario where artifact files pagination could get into an infinite loop
• Fixed bug where success toast messages were not appearing
• The Runs table will now correctly show the git commit value

1.13.20 - 0.34.0

Features

New Model Registry UI

We’re making it easier for users to manage a long list of models, and navigate seamlessly between entities in the model registry. With this new UI, users can:

• Look at all your registered models
• Filter to registered models within a specific team
• With the new list view, users can expand each panel to see the individual versions inside of it, including each version’s aliases, and metadata or run metrics. Clicking on a version from this quick view can take you to it’s version-view
• Look at an overview directly by clicking “View Details”
• See a preview of how many version, consumers, and automations are present for each registered model
• Create Automations directly
• See some metadata columns and details in preview
• Change Model Access Controls

Fixes

• Improved search functionality for better universal search ranking results.
• Added functionality to add/delete multiple tags at once in the model registry
• Enhanced the FileMarkdown feature to correctly scroll long content.
• Made the default team selection dropdown scrollable.
• Removed the UI access restriction for Tier 1/2/3 plans based on tracked hour usage.
• Added tooltips to for LLM trace viewer spans
• LLM trace timeline/detail now splits horizontally in fullscreen
• Added entity / team badges to Model Registry entries.
• Improved the navigation bar experience for logged out users
• Disabled storage/artifact banners to avoid issue where UI blocks for orgs with excess artifacts.
• Fixed issues where user avatars were not being displayed correctly.
• Fixed issue using Launch with Azure Git URLs
• Launch configuration boxes now work in airgapped environments
• In Launch queue creation, show teams as disabled (rather than hidden) for non-admins.
• Fixed issue with embedding projector rendering
• Fixes issue that prevented users from being able to reset their password in some cases involving mixed-case usernames.
• Files with special characters now show up in the media panel in Azure
• Added the ability to override the inline display format for timestamps.
• Reports with custom charts now load when not logged in.
• Wide GIFs no longer overflow fullscreen view
• Increase default automations limit from 20 to 200.
• Fixed bug allowing the appearance of deleting the version alias of a registered model (in fact, this could not be deleted on the backend).

1.13.21 - 0.33.0

Features

Prompts: Zoom and pan

Explore complex chains of LLM prompts more easily with new zoom and pan controls in our prompts tracer.

Model registry admin role

Control your model promotion process with a new role for model registry admins. These users can manage the list of protected aliases (for example, “challenger” or “prod”), as well as apply or remove protected aliases for model versions.

Viewer role

You can now share your W&B findings with a broader audience with the introduction of a Viewer role for W&B Server. Users with this role can view anything their team(s) make, but not create, edit, or delete anything. These seats are measured separately from traditional W&B Server seats, so reach out your W&B account team to request an updated license.

Improved sharing: optional magic link, and easier signup for viewers

Team admins can now disable magic link sharing for a team and its members. Disable public sharing on the team setting allows you increase team privacy controls. Meanwhile, it’s now easier for users who receive a report link to access the report in W&B after signing up.

Improved report composition

Reports help share your findings W&B further throughout an organization, including with people outside the ML team. We’ve made several investments to ensure it’s as simple and frictionless as possible to create and share them—including an improved report drafting experience with enhanced draft publication, editing, management, and sharing UX to improve how teams collaborate with Reports.

Updated navigation

As W&B has expanded the parts of the ML workflow we cover, we’ve heard your feedback that it can be hard to move around the application. So we’ve updated the navigation sidebar to include clearer labels on the product area, and added backlinks to certain detail screens. We’ve also renamed “Triggers” to “Automations” to better reflect the power of the feature.

Fixes

• When hovering over a plot in workspaces or a report, you can now use Cmd+C or Ctrl+C to copy run names and plot values shown in the hover control.
• Changes to default workspaces are now no longer auto-saved.
• Metrics in the Overview → Summary section now are formatted with commas.
• Added an install-level option to allow non-admin users to create teams (default off; contact W&B support to enable it).
• Weave plots now support log scales.
• The Launch panel can now be expanded horizontally to give more space for viewing parameters.
• The Launch panel now indicates whether a queue is active
• The Launch panel now allows you to choose a project for the run to be logged in.
• Launch queues can now only be created by team admins.
• Improved Markdown support in Launch panel.
• Improved error message on empty Launch queue configurations.
• Filters on the Sweeps parallel coordinates plot will now apply to all selected runsets.
• Sweeps now no longer require a metric.
• Added support for tracking reference artifact files saved outside W&B in Azure Blob Storage.
• Fixed bug in Markdown editing in Reports
• Fullscreen Weave panels can now share config changes with the original panel.
• Improved display of empty tables
• Fixed bug in which the first several characters of logs were cut off

1.14 - Release policies and processes

This page gives details about W&B Server releases and W&B’s release policies. This page relates to W&B Dedicated Cloud and Self-Managed deployments. To learn more about an individual W&B Server release, refer to W&B release notes.

W&B fully manages W&B Multi-tenant Cloud and the details in this page do not apply.

Release support and end of life policy

W&B supports a major W&B Server release for 12 months from its initial release date.

• Dedicated Cloud instances are automatically updated to maintain support.

• Customers with Self-managed instances are responsible for upgrading in time to maintain support. Avoid staying on an unsupported version.

  W&B strongly recommends customers with Self-managed instances to update their deployments with the latest release at minimum once per quarter to maintain support and receive the latest features, performance improvements, and fixes.

Release types and frequencies

• Major releases are produced monthly, and may include new features, enhancements, performance improvements, medium and low severity bug fixes, and deprecations. An example of a major release is 0.68.0.
• Patch releases within a major version are produced as needed, and include critical and high severity bug fixes. An example of a patch release is 0.67.1.

Release rollout

1. After testing and validation are complete, a release is first rolled out to all Dedicated Cloud instances to keep them fully updated.
2. After additional observation, the release is published, and Self-managed deployments can upgrade to it on their own schedule, and are responsible for upgrading in time to comply with the Release support and End of Life (EOL) policy. Learn more about upgrading W&B Server.

Downtime during upgrades

• When a Dedicated Cloud instance is upgraded, downtime is generally not expected, but may occur in certain situations:
  • If a new feature or enhancement requires changes to the underlying infrastructure, such as compute, storage or network.
  • To roll out a critical infrastructure change such as a security fix.
  • If the instance’s current version has reached its End of Life (EOL) and is upgraded by W&B to maintain support.
• For Self-managed deployments, the customer is responsible for implementing a rolling update process that meets their service level objectives (SLOs), such as by running W&B Server on Kubernetes.

Feature availability

After installing or upgrading, certain features may not be immediately available.

Enterprise features

An Enterprise license includes support for important security capabilities and other enterprise-friendly functionality. Some advanced features require an Enterprise license.

• Dedicated Cloud includes an Enterprise license and no action is required.
• On Self-managed deployments, features that require an Enterprise license are not available until it is set. To learn more or obtain an Enterprise license, refer to Obtain your W&B Server license.

Private preview and opt-in features

Most features are available immediately after installing or upgrading W&B Server. The W&B team must enable certain features before you can use them in your instance.

• Private preview: W&B invites design partners and early adopters to test these features and provide feedback. Private preview features are not recommended for production environments.

  The W&B team must enable a private preview feature for your instance before you can use it. Public documentation is not available; instructions are provided directly. Interfaces and APIs may change, and the feature may not be fully implemented.

• Public preview: Contact W&B to opt in to a public preview to try it out before it is generally available.

  The W&B team must enable a public preview feature before you can use it in your instance. Documentation may not be complete, interfaces and APIs may change, and the feature may not be fully implemented.

To learn more about an individual W&B Server release, including any limitations, refer to W&B Release notes.

2 - Command Line Interface

Usage

wandb [OPTIONS] COMMAND [ARGS]...

Options

Commands

2.1 - wandb agent

Usage

wandb agent [OPTIONS] SWEEP_ID

Summary

Run the W&B agent

Options

2.2 - wandb artifact

Usage

wandb artifact [OPTIONS] COMMAND [ARGS]...

Summary

Commands for interacting with artifacts

Options

Commands

2.2.1 - wandb artifact cache

Usage

wandb artifact cache [OPTIONS] COMMAND [ARGS]...

Summary

Commands for interacting with the artifact cache

Options

Commands

2.2.1.1 - wandb artifact cache cleanup

Usage

wandb artifact cache cleanup [OPTIONS] TARGET_SIZE

Summary

Clean up less frequently used files from the artifacts cache

Options

2.2.2 - wandb artifact get

Usage

wandb artifact get [OPTIONS] PATH

Summary

Download an artifact from wandb

Options

2.2.3 - wandb artifact ls

Usage

wandb artifact ls [OPTIONS] PATH

Summary

List all artifacts in a wandb project

Options

2.2.4 - wandb artifact put

Usage

wandb artifact put [OPTIONS] PATH

Summary

Upload an artifact to wandb

Options

2.3 - wandb beta

Usage

wandb beta [OPTIONS] COMMAND [ARGS]...

Summary

Beta versions of wandb CLI commands. Requires wandb-core.

Options

Commands

2.3.1 - wandb beta sync

Usage

wandb beta sync [OPTIONS] WANDB_DIR

Summary

Upload a training run to W&B

Options

2.4 - wandb controller

Usage

wandb controller [OPTIONS] SWEEP_ID

Summary

Run the W&B local sweep controller

Options

2.5 - wandb disabled

Usage

wandb disabled [OPTIONS]

Summary

Disable W&B.

Options

2.6 - wandb docker

Usage

wandb docker [OPTIONS] [DOCKER_RUN_ARGS]... [DOCKER_IMAGE]

Summary

Run your code in a docker container.

W&B docker lets you run your code in a docker image ensuring wandb is configured. It adds the WANDB_DOCKER and WANDB_API_KEY environment variables to your container and mounts the current directory in /app by default. You can pass additional args which will be added to docker run before the image name is declared, we’ll choose a default image for you if one isn’t passed:

images-public/tensorflow-1.12.0-notebook-cpu:v0.4.0 --jupyter wandb docker
wandb/deepo:keras-gpu --no-tty --cmd "python train.py --epochs=5" ```

By default, we override the entrypoint to check for the existence of wandb
and install it if not present.  If you pass the --jupyter flag we will
ensure jupyter is installed and start jupyter lab on port 8888.  If we
detect nvidia-docker on your system we will use the nvidia runtime.  If you
just want wandb to set environment variable to an existing docker run
command, see the wandb docker-run command.


**Options**

| **Option** | **Description** |
| :--- | :--- |
| `--nvidia / --no-nvidia` | Use the nvidia runtime, defaults to nvidia if   nvidia-docker is present |
| `--digest` | Output the image digest and exit |
| `--jupyter / --no-jupyter` | Run jupyter lab in the container |
| `--dir` | Which directory to mount the code in the container |
| `--no-dir` | Don't mount the current directory |
| `--shell` | The shell to start the container with |
| `--port` | The host port to bind jupyter on |
| `--cmd` | The command to run in the container |
| `--no-tty` | Run the command without a tty |

2.7 - wandb docker-run

Usage

wandb docker-run [OPTIONS] [DOCKER_RUN_ARGS]...

Summary

Wrap docker run and adds WANDB_API_KEY and WANDB_DOCKER environment variables.

This will also set the runtime to nvidia if the nvidia-docker executable is present on the system and –runtime wasn’t set.

See docker run --help for more details.

Options

2.8 - wandb enabled

Usage

wandb enabled [OPTIONS]

Summary

Enable W&B.

Options

2.9 - wandb init

Usage

wandb init [OPTIONS]

Summary

Configure a directory with Weights & Biases

Options

2.10 - wandb job

Usage

wandb job [OPTIONS] COMMAND [ARGS]...

Summary

Commands for managing and viewing W&B jobs

Options

Commands

2.10.1 - wandb job create

Usage

wandb job create [OPTIONS] {git|code|image} PATH

Summary

Create a job from a source, without a wandb run.

Jobs can be of three types, git, code, or image.

git: A git source, with an entrypoint either in the path or provided explicitly pointing to the main python executable. code: A code path, containing a requirements.txt file. image: A docker image.

Options

2.10.2 - wandb job describe

Usage

wandb job describe [OPTIONS] JOB

Summary

Describe a launch job. Provide the launch job in the form of: entity/project/job-name:alias-or-version

Options

2.10.3 - wandb job list

Usage

wandb job list [OPTIONS]

Summary

List jobs in a project

Options

2.11 - wandb launch

Usage

wandb launch [OPTIONS]

Summary

Launch or queue a W&B Job. See https://wandb.me/launch

Options

2.12 - wandb launch-agent

Usage

wandb launch-agent [OPTIONS]

Summary

Run a W&B launch agent.

Options

2.13 - wandb launch-sweep

Usage

wandb launch-sweep [OPTIONS] [CONFIG]

Summary

Run a W&B launch sweep (Experimental).

Options

2.14 - wandb login

Usage

wandb login [OPTIONS] [KEY]...

Summary

Login to Weights & Biases

Options

2.15 - wandb offline

Usage

wandb offline [OPTIONS]

Summary

Disable W&B sync

Options

2.16 - wandb online

Usage

wandb online [OPTIONS]

Summary

Enable W&B sync

Options

2.17 - wandb pull

Usage

wandb pull [OPTIONS] RUN

Summary

Pull files from Weights & Biases

Options

2.18 - wandb restore

Usage

wandb restore [OPTIONS] RUN

Summary

Restore code, config and docker state for a run

Options

2.19 - wandb scheduler

Usage

wandb scheduler [OPTIONS] SWEEP_ID

Summary

Run a W&B launch sweep scheduler (Experimental)

Options

2.20 - wandb server

Usage

wandb server [OPTIONS] COMMAND [ARGS]...

Summary

Commands for operating a local W&B server

Options

Commands

2.20.1 - wandb server start

Usage

wandb server start [OPTIONS]

Summary

Start a local W&B server

Options

2.20.2 - wandb server stop

Usage

wandb server stop [OPTIONS]

Summary

Stop a local W&B server

Options

2.21 - wandb status

Usage

wandb status [OPTIONS]

Summary

Show configuration settings

Options

2.22 - wandb sweep

Usage

wandb sweep [OPTIONS] CONFIG_YAML_OR_SWEEP_ID

Summary

Initialize a hyperparameter sweep. Search for hyperparameters that optimizes a cost function of a machine learning model by testing various combinations.

Options

2.23 - wandb sync

Usage

wandb sync [OPTIONS] [PATH]...

Summary

Upload an offline training directory to W&B

Options

2.24 - wandb verify

Usage

wandb verify [OPTIONS]

Summary

Verify your local instance

Options

3 - JavaScript Library

Similar to our Python library, we offer a client to track experiments in JavaScript/TypeScript.

• Log metrics from your Node server and display them in interactive plots on W&B
• Debug LLM applications with interactive traces
• Debug LangChain.js usage

This library is compatible with Node and modern JS run times.

You can find the source code for the JavaScript client in the Github repository.

Installation

npm install @wandb/sdk
# or ...
yarn add @wandb/sdk

Usage

TypeScript/ESM:

import wandb from '@wandb/sdk'

async function track() {
    await wandb.init({config: {test: 1}});
    wandb.log({acc: 0.9, loss: 0.1});
    wandb.log({acc: 0.91, loss: 0.09});
    await wandb.finish();
}

await track()

Node/CommonJS:

const wandb = require('@wandb/sdk').default;

We’re currently missing a lot of the functionality found in our Python SDK, but basic logging functionality is available. We’ll be adding additional features like Tables soon.

Authentication and Settings

In node environments we look for process.env.WANDB_API_KEY and prompt for it’s input if we have a TTY. In non-node environments we look for sessionStorage.getItem("WANDB_API_KEY"). Additional settings can be found here.

Integrations

Our Python integrations are widely used by our community, and we hope to build out more JavaScript integrations to help LLM app builders leverage whatever tool they want.

If you have any requests for additional integrations, we’d love you to open an issue with details about the request.

LangChain.js

This library integrates with the popular library for building LLM applications, LangChain.js version >= 0.0.75.

import {WandbTracer} from '@wandb/sdk/integrations/langchain';

const wbTracer = await WandbTracer.init({project: 'langchain-test'});
// run your langchain workloads...
chain.call({input: "My prompt"}, wbTracer)
await WandbTracer.finish();

See this test for a more detailed example.

4 - Python Library v(0.21.0)

4.1 - API Walkthrough

Learn when and how to use different W&B APIs to track, share, and manage model artifacts in your machine learning workflows. This page covers logging experiments, generating reports, and accessing logged data using the appropriate W&B API for each task.

W&B offers the following APIs:

• W&B Python SDK (wandb.sdk): Log and monitor experiments during training.
• W&B Public API (wandb.apis.public): Query and analyze logged experiment data.
• W&B Report and Workspace API (wandb.wandb-workspaces): Create reports to summarize findings.

Sign up and create an API key

To authenticate your machine with W&B, you must first generate an API key at wandb.ai/authorize. Copy the API key and store it securely.

Install and import packages

Install the W&B library and some other packages you will need for this walkthrough.

pip install wandb

Import W&B Python SDK:

import wandb

Specify the entity of your team in the following code block:

TEAM_ENTITY = "<Team_Entity>" # Replace with your team entity
PROJECT = "my-awesome-project"

Train model

The following code simulates a basic machine learning workflow: training a model, logging metrics, and saving the model as an artifact.

Use the W&B Python SDK (wandb.sdk) to interact with W&B during training. Log the loss using wandb.Run.log(), then save the trained model as an artifact using wandb.Artifact before finally adding the model file using Artifact.add_file.

import random # For simulating data

def model(training_data: int) -> int:
    """Model simulation for demonstration purposes."""
    return training_data * 2 + random.randint(-1, 1)  

# Simulate weights and noise
weights = random.random() # Initialize random weights
noise = random.random() / 5  # Small random noise to simulate noise

# Hyperparameters and configuration
config = {
    "epochs": 10,  # Number of epochs to train
    "learning_rate": 0.01,  # Learning rate for the optimizer
}

# Use context manager to initialize and close W&B runs
with wandb.init(project=PROJECT, entity=TEAM_ENTITY, config=config) as run:    
    # Simulate training loop
    for epoch in range(config["epochs"]):
        xb = weights + noise  # Simulated input training data
        yb = weights + noise * 2  # Simulated target output (double the input noise)
        
        y_pred = model(xb)  # Model prediction
        loss = (yb - y_pred) ** 2  # Mean Squared Error loss

        print(f"epoch={epoch}, loss={y_pred}")
        # Log epoch and loss to W&B
        run.log({
            "epoch": epoch,
            "loss": loss,
        })

    # Unique name for the model artifact,
    model_artifact_name = f"model-demo"  

    # Local path to save the simulated model file
    PATH = "model.txt" 

    # Save model locally
    with open(PATH, "w") as f:
        f.write(str(weights)) # Saving model weights to a file

    # Create an artifact object
    # Add locally saved model to artifact object
    artifact = wandb.Artifact(name=model_artifact_name, type="model", description="My trained model")
    artifact.add_file(local_path=PATH)
    artifact.save()

The key takeaways from the previous code block are:

• Use wandb.Run.log() to log metrics during training.
• Use wandb.Artifact to save models (datasets, and so forth) as an artifact to your W&B project.

Now that you have trained a model and saved it as an artifact, you can publish it to a registry in W&B. Use wandb.Run.use_artifact() to retrieve the artifact from your project and prepare it for publication in the Model registry. wandb.Run.use_artifact() serves two key purposes:

• Retrieves the artifact object from your project.
• Marks the artifact as an input to the run, ensuring reproducibility and traceability. See Create and view lineage map for details.

Publish the model to the Model registry

To share the model with others in your organization, publish it to a collection using wandb.Run.link_artifact(). The following code links the artifact to the core Model registry, making it accessible to your team.

# Artifact name specifies the specific artifact version within our team's project
artifact_name = f'{TEAM_ENTITY}/{PROJECT}/{model_artifact_name}:v0'
print("Artifact name: ", artifact_name)

REGISTRY_NAME = "Model" # Name of the registry in W&B
COLLECTION_NAME = "DemoModels"  # Name of the collection in the registry

# Create a target path for our artifact in the registry
target_path = f"wandb-registry-{REGISTRY_NAME}/{COLLECTION_NAME}"
print("Target path: ", target_path)

run = wandb.init(entity=TEAM_ENTITY, project=PROJECT)
model_artifact = run.use_artifact(artifact_or_name=artifact_name, type="model")
run.link_artifact(artifact=model_artifact, target_path=target_path)
run.finish()

After running wandb.Run.link_artifact(), the model artifact will be in the DemoModels collection in your registry. From there, you can view details such as the version history, lineage map, and other metadata.

For additional information on how to link artifacts to a registry, see Link artifacts to a registry.

Retrieve model artifact from registry for inference

To use a model for inference, use wandb.Run.use_artifact() to retrieve the published artifact from the registry. This returns an artifact object that you can then use wandb.Artifact.download() to download the artifact to a local file.

REGISTRY_NAME = "Model"  # Name of the registry in W&B
COLLECTION_NAME = "DemoModels"  # Name of the collection in the registry
VERSION = 0 # Version of the artifact to retrieve

model_artifact_name = f"wandb-registry-{REGISTRY_NAME}/{COLLECTION_NAME}:v{VERSION}"
print(f"Model artifact name: {model_artifact_name}")

run = wandb.init(entity=TEAM_ENTITY, project=PROJECT)
registry_model = run.use_artifact(artifact_or_name=model_artifact_name)
local_model_path = registry_model.download()

For more information on how to retrieve artifacts from a registry, see Download an artifact from a registry.

Depending on your machine learning framework, you may need to recreate the model architecture before loading the weights. This is left as an exercise for the reader, as it depends on the specific framework and model you are using.

Share your finds with a report

Create and share a report to summarize your work. To create a report programmatically, use the W&B Report and Workspace API.

First, install the W&B Reports API:

pip install wandb wandb-workspaces -qqq

The following code block creates a report with multiple blocks, including markdown, panel grids, and more. You can customize the report by adding more blocks or changing the content of existing blocks.

The output of the code block prints a link to the URL report created. You can open this link in your browser to view the report.

import wandb_workspaces.reports.v2 as wr

experiment_summary = """This is a summary of the experiment conducted to train a simple model using W&B."""
dataset_info = """The dataset used for training consists of synthetic data generated by a simple model."""
model_info = """The model is a simple linear regression model that predicts output based on input data with some noise."""

report = wr.Report(
    project=PROJECT,
    entity=TEAM_ENTITY,
    title="My Awesome Model Training Report",
    description=experiment_summary,
    blocks= [
        wr.TableOfContents(),
        wr.H2("Experiment Summary"),
        wr.MarkdownBlock(text=experiment_summary),
        wr.H2("Dataset Information"),
        wr.MarkdownBlock(text=dataset_info),
        wr.H2("Model Information"),
        wr.MarkdownBlock(text = model_info),
        wr.PanelGrid(
            panels=[
                wr.LinePlot(title="Train Loss", x="Step", y=["loss"], title_x="Step", title_y="Loss")
                ],
            ),  
    ]

)

# Save the report to W&B
report.save()

For more information on how to create a report programmatically or how to create a report interactively with the W&B App, see Create a report in the W&B Docs Developer guide.

Query the registry

Use the W&B Public APIs to query, analyze, and manage historical data from W&B. This can be useful for tracking the lineage of artifacts, comparing different versions, and analyzing the performance of models over time.

The following code block demonstrates how to query the Model registry for all artifacts in a specific collection. It retrieves the collection and iterates through its versions, printing out the name and version of each artifact.

import wandb

# Initialize wandb API
api = wandb.Api()

# Find all artifact versions that contains the string `model` and 
# has either the tag `text-classification` or an `latest` alias
registry_filters = {
    "name": {"$regex": "model"}
}

# Use logical $or operator to filter artifact versions
version_filters = {
    "$or": [
        {"tag": "text-classification"},
        {"alias": "latest"}
    ]
}

# Returns an iterable of all artifact versions that match the filters
artifacts = api.registries(filter=registry_filters).collections().versions(filter=version_filters)

# Print out the name, collection, aliases, tags, and created_at date of each artifact found
for art in artifacts:
    print(f"artifact name: {art.name}")
    print(f"collection artifact belongs to: { art.collection.name}")
    print(f"artifact aliases: {art.aliases}")
    print(f"tags attached to artifact: {art.tags}")
    print(f"artifact created at: {art.created_at}\n")

For more information on querying the registry, see the Query registry items with MongoDB-style queries.

4.2 - SDK

Train and fine-tune models, manage models from experimentation to production. For guides and examples, see https://docs.wandb.ai.

4.2.1 - Functions

4.2.1.1 - agent()

function agent

agent(
    sweep_id: str,
    function: Optional[Callable] = None,
    entity: Optional[str] = None,
    project: Optional[str] = None,
    count: Optional[int] = None
) → None

Start one or more sweep agents.

The sweep agent uses the sweep_id to know which sweep it is a part of, what function to execute, and (optionally) how many agents to run.

Args:

• sweep_id: The unique identifier for a sweep. A sweep ID is generated by W&B CLI or Python SDK.
• function: A function to call instead of the “program” specified in the sweep config.
• entity: The username or team name where you want to send W&B runs created by the sweep to. Ensure that the entity you specify already exists. If you don’t specify an entity, the run will be sent to your default entity, which is usually your username.
• project: The name of the project where W&B runs created from the sweep are sent to. If the project is not specified, the run is sent to a project labeled “Uncategorized”.
• count: The number of sweep config trials to try.

4.2.1.2 - controller()

function controller

controller(
    sweep_id_or_config: Optional[str, Dict] = None,
    entity: Optional[str] = None,
    project: Optional[str] = None
) → _WandbController

Public sweep controller constructor.

Examples:

import wandb

tuner = wandb.controller(...)
print(tuner.sweep_config)
print(tuner.sweep_id)
tuner.configure_search(...)
tuner.configure_stopping(...)

4.2.1.3 - finish()

function finish

finish(exit_code: 'int | None' = None, quiet: 'bool | None' = None) → None

Finish a run and upload any remaining data.

Marks the completion of a W&B run and ensures all data is synced to the server. The run’s final state is determined by its exit conditions and sync status.

Run States:

• Running: Active run that is logging data and/or sending heartbeats.
• Crashed: Run that stopped sending heartbeats unexpectedly.
• Finished: Run completed successfully (exit_code=0) with all data synced.
• Failed: Run completed with errors (exit_code!=0).

Args:

• exit_code: Integer indicating the run’s exit status. Use 0 for success, any other value marks the run as failed.
• quiet: Deprecated. Configure logging verbosity using wandb.Settings(quiet=...).

4.2.1.4 - init()

function init

init(
    entity: 'str | None' = None,
    project: 'str | None' = None,
    dir: 'StrPath | None' = None,
    id: 'str | None' = None,
    name: 'str | None' = None,
    notes: 'str | None' = None,
    tags: 'Sequence[str] | None' = None,
    config: 'dict[str, Any] | str | None' = None,
    config_exclude_keys: 'list[str] | None' = None,
    config_include_keys: 'list[str] | None' = None,
    allow_val_change: 'bool | None' = None,
    group: 'str | None' = None,
    job_type: 'str | None' = None,
    mode: "Literal['online', 'offline', 'disabled'] | None" = None,
    force: 'bool | None' = None,
    anonymous: "Literal['never', 'allow', 'must'] | None" = None,
    reinit: "bool | Literal[None, 'default', 'return_previous', 'finish_previous', 'create_new']" = None,
    resume: "bool | Literal['allow', 'never', 'must', 'auto'] | None" = None,
    resume_from: 'str | None' = None,
    fork_from: 'str | None' = None,
    save_code: 'bool | None' = None,
    tensorboard: 'bool | None' = None,
    sync_tensorboard: 'bool | None' = None,
    monitor_gym: 'bool | None' = None,
    settings: 'Settings | dict[str, Any] | None' = None
) → Run

Start a new run to track and log to W&B.

In an ML training pipeline, you could add wandb.init() to the beginning of your training script as well as your evaluation script, and each piece would be tracked as a run in W&B.

wandb.init() spawns a new background process to log data to a run, and it also syncs data to https://wandb.ai by default, so you can see your results in real-time. When you’re done logging data, call wandb.finish() to end the run. If you don’t call run.finish(), the run will end when your script exits.

Run IDs must not contain any of the following special characters / \ # ? % :

Args:

• entity: The username or team name the runs are logged to. The entity must already exist, so ensure you create your account or team in the UI before starting to log runs. If not specified, the run will default your default entity. To change the default entity, go to your settings and update the “Default location to create new projects” under “Default team”.
• project: The name of the project under which this run will be logged. If not specified, we use a heuristic to infer the project name based on the system, such as checking the git root or the current program file. If we can’t infer the project name, the project will default to "uncategorized".
• dir: The absolute path to the directory where experiment logs and metadata files are stored. If not specified, this defaults to the ./wandb directory. Note that this does not affect the location where artifacts are stored when calling download().
• id: A unique identifier for this run, used for resuming. It must be unique within the project and cannot be reused once a run is deleted. For a short descriptive name, use the name field, or for saving hyperparameters to compare across runs, use config.
• name: A short display name for this run, which appears in the UI to help you identify it. By default, we generate a random two-word name allowing easy cross-reference runs from table to charts. Keeping these run names brief enhances readability in chart legends and tables. For saving hyperparameters, we recommend using the config field.
• notes: A detailed description of the run, similar to a commit message in Git. Use this argument to capture any context or details that may help you recall the purpose or setup of this run in the future.
• tags: A list of tags to label this run in the UI. Tags are helpful for organizing runs or adding temporary identifiers like “baseline” or “production.” You can easily add, remove tags, or filter by tags in the UI. If resuming a run, the tags provided here will replace any existing tags. To add tags to a resumed run without overwriting the current tags, use run.tags += ("new_tag",) after calling run = wandb.init().
• config: Sets wandb.config, a dictionary-like object for storing input parameters to your run, such as model hyperparameters or data preprocessing settings. The config appears in the UI in an overview page, allowing you to group, filter, and sort runs based on these parameters. Keys should not contain periods (.), and values should be smaller than 10 MB. If a dictionary, argparse.Namespace, or absl.flags.FLAGS is provided, the key-value pairs will be loaded directly into wandb.config. If a string is provided, it is interpreted as a path to a YAML file, from which configuration values will be loaded into wandb.config.
• config_exclude_keys: A list of specific keys to exclude from wandb.config.
• config_include_keys: A list of specific keys to include in wandb.config.
• allow_val_change: Controls whether config values can be modified after their initial set. By default, an exception is raised if a config value is overwritten. For tracking variables that change during training, such as a learning rate, consider using wandb.log() instead. By default, this is False in scripts and True in Notebook environments.
• group: Specify a group name to organize individual runs as part of a larger experiment. This is useful for cases like cross-validation or running multiple jobs that train and evaluate a model on different test sets. Grouping allows you to manage related runs collectively in the UI, making it easy to toggle and review results as a unified experiment.
• job_type: Specify the type of run, especially helpful when organizing runs within a group as part of a larger experiment. For example, in a group, you might label runs with job types such as “train” and “eval”. Defining job types enables you to easily filter and group similar runs in the UI, facilitating direct comparisons.
• mode: Specifies how run data is managed, with the following options:
  • "online" (default): Enables live syncing with W&B when a network connection is available, with real-time updates to visualizations.
  • "offline": Suitable for air-gapped or offline environments; data is saved locally and can be synced later. Ensure the run folder is preserved to enable future syncing.
  • "disabled": Disables all W&B functionality, making the run’s methods no-ops. Typically used in testing to bypass W&B operations.
• force: Determines if a W&B login is required to run the script. If True, the user must be logged in to W&B; otherwise, the script will not proceed. If False (default), the script can proceed without a login, switching to offline mode if the user is not logged in.
• anonymous: Specifies the level of control over anonymous data logging. Available options are:
  • "never" (default): Requires you to link your W&B account before tracking the run. This prevents unintentional creation of anonymous runs by ensuring each run is associated with an account.
  • "allow": Enables a logged-in user to track runs with their account, but also allows someone running the script without a W&B account to view the charts and data in the UI.
  • "must": Forces the run to be logged to an anonymous account, even if the user is logged in.
• reinit: Shorthand for the “reinit” setting. Determines the behavior of wandb.init() when a run is active.
• resume: Controls the behavior when resuming a run with the specified id. Available options are:
  • "allow": If a run with the specified id exists, it will resume from the last step; otherwise, a new run will be created.
  • "never": If a run with the specified id exists, an error will be raised. If no such run is found, a new run will be created.
  • "must": If a run with the specified id exists, it will resume from the last step. If no run is found, an error will be raised.
  • "auto": Automatically resumes the previous run if it crashed on this machine; otherwise, starts a new run.
  • True: Deprecated. Use "auto" instead.
  • False: Deprecated. Use the default behavior (leaving resume unset) to always start a new run. If resume is set, fork_from and resume_from cannot be used. When resume is unset, the system will always start a new run.
• resume_from: Specifies a moment in a previous run to resume a run from, using the format {run_id}?_step={step}. This allows users to truncate the history logged to a run at an intermediate step and resume logging from that step. The target run must be in the same project. If an id argument is also provided, the resume_from argument will take precedence. resume, resume_from and fork_from cannot be used together, only one of them can be used at a time. Note that this feature is in beta and may change in the future.
• fork_from: Specifies a point in a previous run from which to fork a new run, using the format {id}?_step={step}. This creates a new run that resumes logging from the specified step in the target run’s history. The target run must be part of the current project. If an id argument is also provided, it must be different from the fork_from argument, an error will be raised if they are the same. resume, resume_from and fork_from cannot be used together, only one of them can be used at a time. Note that this feature is in beta and may change in the future.
• save_code: Enables saving the main script or notebook to W&B, aiding in experiment reproducibility and allowing code comparisons across runs in the UI. By default, this is disabled, but you can change the default to enable on your settings page.
• tensorboard: Deprecated. Use sync_tensorboard instead.
• sync_tensorboard: Enables automatic syncing of W&B logs from TensorBoard or TensorBoardX, saving relevant event files for viewing in the W&B UI.
• saving relevant event files for viewing in the W&B UI. (Default: False)
• monitor_gym: Enables automatic logging of videos of the environment when using OpenAI Gym.
• settings: Specifies a dictionary or wandb.Settings object with advanced settings for the run.

Raises:

• Error: If some unknown or internal error happened during the run initialization.
• AuthenticationError: If the user failed to provide valid credentials.
• CommError: If there was a problem communicating with the WandB server.
• UsageError: If the user provided invalid arguments.
• KeyboardInterrupt: If user interrupts the run.

Returns: A Run object.

Examples: wandb.init() returns a Run object. Use the run object to log data, save artifacts, and manage the run lifecycle.

import wandb

config = {"lr": 0.01, "batch_size": 32}
with wandb.init(config=config) as run:
    # Log accuracy and loss to the run
    acc = 0.95  # Example accuracy
    loss = 0.05  # Example loss
    run.log({"accuracy": acc, "loss": loss})

4.2.1.5 - login()

function login

login(
    anonymous: Optional[Literal['allow', 'must', 'never']] = None,
    key: Optional[str] = None,
    relogin: Optional[bool] = None,
    host: Optional[str] = None,
    force: Optional[bool] = None,
    timeout: Optional[int] = None,
    verify: bool = False,
    referrer: Optional[str] = None
) → bool

Set up W&B login credentials.

By default, this will only store credentials locally without verifying them with the W&B server. To verify credentials, pass verify=True.

Args:

• anonymous: Set to “must”, “allow”, or “never”. If set to “must”, always log a user in anonymously. If set to “allow”, only create an anonymous user if the user isn’t already logged in. If set to “never”, never log a user anonymously. Default set to “never”.
• key: The API key to use.
• relogin: If true, will re-prompt for API key.
• host: The host to connect to.
• force: If true, will force a relogin.
• timeout: Number of seconds to wait for user input.
• verify: Verify the credentials with the W&B server.
• referrer: The referrer to use in the URL login request.

Returns:

• bool: If key is configured.

Raises:

• AuthenticationError: If api_key fails verification with the server.
• UsageError: If api_key cannot be configured and no tty.

4.2.1.6 - restore()

function restore

restore(
    name: 'str',
    run_path: 'str | None' = None,
    replace: 'bool' = False,
    root: 'str | None' = None
) → None | TextIO

Download the specified file from cloud storage.

File is placed into the current directory or run directory. By default, will only download the file if it doesn’t already exist.

Args:

• name: The name of the file.
• run_path: Optional path to a run to pull files from, i.e. username/project_name/run_id if wandb.init has not been called, this is required.
• replace: Whether to download the file even if it already exists locally
• root: The directory to download the file to. Defaults to the current directory or the run directory if wandb.init was called.

Returns: None if it can’t find the file, otherwise a file object open for reading.

Raises:

• CommError: If W&B can’t connect to the W&B backend.
• ValueError: If the file is not found or can’t find run_path.

4.2.1.7 - setup()

function setup

setup(settings: 'Settings | None' = None) → _WandbSetup

Prepares W&B for use in the current process and its children.

You can usually ignore this as it is implicitly called by wandb.init().

When using wandb in multiple processes, calling wandb.setup() in the parent process before starting child processes may improve performance and resource utilization.

Note that wandb.setup() modifies os.environ, and it is important that child processes inherit the modified environment variables.

See also wandb.teardown().

Args:

• settings: Configuration settings to apply globally. These can be overridden by subsequent wandb.init() calls.

Example:

import multiprocessing

import wandb


def run_experiment(params):
   with wandb.init(config=params):
        # Run experiment
        pass


if __name__ == "__main__":
   # Start backend and set global config
   wandb.setup(settings={"project": "my_project"})

   # Define experiment parameters
   experiment_params = [
        {"learning_rate": 0.01, "epochs": 10},
        {"learning_rate": 0.001, "epochs": 20},
   ]

   # Start multiple processes, each running a separate experiment
   processes = []
   for params in experiment_params:
        p = multiprocessing.Process(target=run_experiment, args=(params,))
        p.start()
        processes.append(p)

   # Wait for all processes to complete
   for p in processes:
        p.join()

   # Optional: Explicitly shut down the backend
   wandb.teardown()

4.2.1.8 - sweep()

function sweep

sweep(
    sweep: Union[dict, Callable],
    entity: Optional[str] = None,
    project: Optional[str] = None,
    prior_runs: Optional[List[str]] = None
) → str

Initialize a hyperparameter sweep.

Search for hyperparameters that optimizes a cost function of a machine learning model by testing various combinations.

Make note the unique identifier, sweep_id, that is returned. At a later step provide the sweep_id to a sweep agent.

See Sweep configuration structure for information on how to define your sweep.

Args:

• sweep: The configuration of a hyperparameter search. (or configuration generator). If you provide a callable, ensure that the callable does not take arguments and that it returns a dictionary that conforms to the W&B sweep config spec.
• entity: The username or team name where you want to send W&B runs created by the sweep to. Ensure that the entity you specify already exists. If you don’t specify an entity, the run will be sent to your default entity, which is usually your username.
• project: The name of the project where W&B runs created from the sweep are sent to. If the project is not specified, the run is sent to a project labeled ‘Uncategorized’.
• prior_runs: The run IDs of existing runs to add to this sweep.

Returns:

• sweep_id: (str) A unique identifier for the sweep.

4.2.1.9 - teardown()

function teardown

teardown(exit_code: 'int | None' = None) → None

Waits for W&B to finish and frees resources.

Completes any runs that were not explicitly finished using run.finish() and waits for all data to be uploaded.

It is recommended to call this at the end of a session that used wandb.setup(). It is invoked automatically in an atexit hook, but this is not reliable in certain setups such as when using Python’s multiprocessing module.

4.2.2 - Classes

4.2.2.1 - Artifact

class Artifact

Flexible and lightweight building block for dataset and model versioning.

Construct an empty W&B Artifact. Populate an artifacts contents with methods that begin with add. Once the artifact has all the desired files, you can call run.log_artifact() to log it.

Args:

• name (str): A human-readable name for the artifact. Use the name to identify a specific artifact in the W&B App UI or programmatically. You can interactively reference an artifact with the use_artifact Public API. A name can contain letters, numbers, underscores, hyphens, and dots. The name must be unique across a project.
• type (str): The artifact’s type. Use the type of an artifact to both organize and differentiate artifacts. You can use any string that contains letters, numbers, underscores, hyphens, and dots. Common types include dataset or model. Include model within your type string if you want to link the artifact to the W&B Model Registry. Note that some types reserved for internal use and cannot be set by users. Such types include job and types that start with wandb-.
• description (str | None) = None: A description of the artifact. For Model or Dataset Artifacts, add documentation for your standardized team model or dataset card. View an artifact’s description programmatically with the Artifact.description attribute or programmatically with the W&B App UI. W&B renders the description as markdown in the W&B App.
• metadata (dict[str, Any] | None) = None: Additional information about an artifact. Specify metadata as a dictionary of key-value pairs. You can specify no more than 100 total keys.
• incremental: Use Artifact.new_draft() method instead to modify an existing artifact.
• use_as: Deprecated.
• is_link: Boolean indication of if the artifact is a linked artifact(True) or source artifact(False).

Returns: An Artifact object.

method Artifact.__init__

__init__(
    name: 'str',
    type: 'str',
    description: 'str | None' = None,
    metadata: 'dict[str, Any] | None' = None,
    incremental: 'bool' = False,
    use_as: 'str | None' = None
) → None

property Artifact.aliases

List of one or more semantically-friendly references or

identifying “nicknames” assigned to an artifact version.

Aliases are mutable references that you can programmatically reference. Change an artifact’s alias with the W&B App UI or programmatically. See Create new artifact versions for more information.

property Artifact.collection

The collection this artifact was retrieved from.

A collection is an ordered group of artifact versions. If this artifact was retrieved from a portfolio / linked collection, that collection will be returned rather than the collection that an artifact version originated from. The collection that an artifact originates from is known as the source sequence.

property Artifact.commit_hash

The hash returned when this artifact was committed.

property Artifact.created_at

Timestamp when the artifact was created.

property Artifact.description

A description of the artifact.

property Artifact.digest

The logical digest of the artifact.

The digest is the checksum of the artifact’s contents. If an artifact has the same digest as the current latest version, then log_artifact is a no-op.

property Artifact.entity

The name of the entity that the artifact collection belongs to.

If the artifact is a link, the entity will be the entity of the linked artifact.

property Artifact.file_count

The number of files (including references).

property Artifact.history_step

The nearest step at which history metrics were logged for the source run of the artifact.

Examples:

run = artifact.logged_by()
if run and (artifact.history_step is not None):
    history = run.sample_history(
        min_step=artifact.history_step,
        max_step=artifact.history_step + 1,
        keys=["my_metric"],
    )

property Artifact.id

The artifact’s ID.

property Artifact.is_link

Boolean flag indicating if the artifact is a link artifact.

True: The artifact is a link artifact to a source artifact. False: The artifact is a source artifact.

property Artifact.linked_artifacts

Returns a list of all the linked artifacts of a source artifact.

If the artifact is a link artifact (artifact.is_link == True), it will return an empty list. Limited to 500 results.

property Artifact.manifest

The artifact’s manifest.

The manifest lists all of its contents, and can’t be changed once the artifact has been logged.

property Artifact.metadata

User-defined artifact metadata.

Structured data associated with the artifact.

property Artifact.name

The artifact name and version of the artifact.

A string with the format {collection}:{alias}. If fetched before an artifact is logged/saved, the name won’t contain the alias. If the artifact is a link, the name will be the name of the linked artifact.

property Artifact.project

The name of the project that the artifact collection belongs to.

If the artifact is a link, the project will be the project of the linked artifact.

property Artifact.qualified_name

The entity/project/name of the artifact.

If the artifact is a link, the qualified name will be the qualified name of the linked artifact path.

property Artifact.size

The total size of the artifact in bytes.

Includes any references tracked by this artifact.

property Artifact.source_artifact

Returns the source artifact. The source artifact is the original logged artifact.

If the artifact itself is a source artifact (artifact.is_link == False), it will return itself.

property Artifact.source_collection

The artifact’s source collection.

The source collection is the collection that the artifact was logged from.

property Artifact.source_entity

The name of the entity of the source artifact.

property Artifact.source_name

The artifact name and version of the source artifact.

A string with the format {source_collection}:{alias}. Before the artifact is saved, contains only the name since the version is not yet known.

property Artifact.source_project

The name of the project of the source artifact.

property Artifact.source_qualified_name

The source_entity/source_project/source_name of the source artifact.

property Artifact.source_version

The source artifact’s version.

A string with the format v{number}.

property Artifact.state

The status of the artifact. One of: “PENDING”, “COMMITTED”, or “DELETED”.

property Artifact.tags

List of one or more tags assigned to this artifact version.

property Artifact.ttl

The time-to-live (TTL) policy of an artifact.

Artifacts are deleted shortly after a TTL policy’s duration passes. If set to None, the artifact deactivates TTL policies and will be not scheduled for deletion, even if there is a team default TTL. An artifact inherits a TTL policy from the team default if the team administrator defines a default TTL and there is no custom policy set on an artifact.

Raises:

• ArtifactNotLoggedError: Unable to fetch inherited TTL if the artifact has not been logged or saved.

property Artifact.type

The artifact’s type. Common types include dataset or model.

property Artifact.updated_at

The time when the artifact was last updated.

property Artifact.url

Constructs the URL of the artifact.

Returns:

• str: The URL of the artifact.

property Artifact.use_as

Deprecated.

property Artifact.version

The artifact’s version.

A string with the format v{number}. If the artifact is a link artifact, the version will be from the linked collection.

method Artifact.add

add(
    obj: 'WBValue',
    name: 'StrPath',
    overwrite: 'bool' = False
) → ArtifactManifestEntry

Add wandb.WBValue obj to the artifact.

Args:

• obj: The object to add. Currently support one of Bokeh, JoinedTable, PartitionedTable, Table, Classes, ImageMask, BoundingBoxes2D, Audio, Image, Video, Html, Object3D
• name: The path within the artifact to add the object.
• overwrite: If True, overwrite existing objects with the same file path if applicable.

Returns: The added manifest entry

Raises:

• ArtifactFinalizedError: You cannot make changes to the current artifact version because it is finalized. Log a new artifact version instead.

method Artifact.add_dir

add_dir(
    local_path: 'str',
    name: 'str | None' = None,
    skip_cache: 'bool | None' = False,
    policy: "Literal['mutable', 'immutable'] | None" = 'mutable',
    merge: 'bool' = False
) → None

Add a local directory to the artifact.

Args:

• local_path: The path of the local directory.
• name: The subdirectory name within an artifact. The name you specify appears in the W&B App UI nested by artifact’s type. Defaults to the root of the artifact.
• skip_cache: If set to True, W&B will not copy/move files to the cache while uploading
• policy: By default, “mutable”.
  • mutable: Create a temporary copy of the file to prevent corruption during upload.
  • immutable: Disable protection, rely on the user not to delete or change the file.
• merge: If False (default), throws ValueError if a file was already added in a previous add_dir call and its content has changed. If True, overwrites existing files with changed content. Always adds new files and never removes files. To replace an entire directory, pass a name when adding the directory using add_dir(local_path, name=my_prefix) and call remove(my_prefix) to remove the directory, then add it again.

Raises:

• ArtifactFinalizedError: You cannot make changes to the current artifact version because it is finalized. Log a new artifact version instead.
• ValueError: Policy must be “mutable” or “immutable”

method Artifact.add_file

add_file(
    local_path: 'str',
    name: 'str | None' = None,
    is_tmp: 'bool | None' = False,
    skip_cache: 'bool | None' = False,
    policy: "Literal['mutable', 'immutable'] | None" = 'mutable',
    overwrite: 'bool' = False
) → ArtifactManifestEntry

Add a local file to the artifact.

Args:

• local_path: The path to the file being added.
• name: The path within the artifact to use for the file being added. Defaults to the basename of the file.
• is_tmp: If true, then the file is renamed deterministically to avoid collisions.
• skip_cache: If True, do not copy files to the cache after uploading.
• policy: By default, set to “mutable”. If set to “mutable”, create a temporary copy of the file to prevent corruption during upload. If set to “immutable”, disable protection and rely on the user not to delete or change the file.
• overwrite: If True, overwrite the file if it already exists.

Returns: The added manifest entry.

Raises:

• ArtifactFinalizedError: You cannot make changes to the current artifact version because it is finalized. Log a new artifact version instead.
• ValueError: Policy must be “mutable” or “immutable”

method Artifact.add_reference

add_reference(
    uri: 'ArtifactManifestEntry | str',
    name: 'StrPath | None' = None,
    checksum: 'bool' = True,
    max_objects: 'int | None' = None
) → Sequence[ArtifactManifestEntry]

Add a reference denoted by a URI to the artifact.

Unlike files or directories that you add to an artifact, references are not uploaded to W&B. For more information, see Track external files.

By default, the following schemes are supported:

• http(s): The size and digest of the file will be inferred by the Content-Length and the ETag response headers returned by the server.
• s3: The checksum and size are pulled from the object metadata. If bucket versioning is enabled, then the version ID is also tracked.
• gs: The checksum and size are pulled from the object metadata. If bucket versioning is enabled, then the version ID is also tracked.
• https, domain matching *.blob.core.windows.net
• Azure: The checksum and size are be pulled from the blob metadata. If storage account versioning is enabled, then the version ID is also tracked.
• file: The checksum and size are pulled from the file system. This scheme is useful if you have an NFS share or other externally mounted volume containing files you wish to track but not necessarily upload.

For any other scheme, the digest is just a hash of the URI and the size is left blank.

Args:

• uri: The URI path of the reference to add. The URI path can be an object returned from Artifact.get_entry to store a reference to another artifact’s entry.
• name: The path within the artifact to place the contents of this reference.
• checksum: Whether or not to checksum the resource(s) located at the reference URI. Checksumming is strongly recommended as it enables automatic integrity validation. Disabling checksumming will speed up artifact creation but reference directories will not iterated through so the objects in the directory will not be saved to the artifact. We recommend setting checksum=False when adding reference objects, in which case a new version will only be created if the reference URI changes.
• max_objects: The maximum number of objects to consider when adding a reference that points to directory or bucket store prefix. By default, the maximum number of objects allowed for Amazon S3, GCS, Azure, and local files is 10,000,000. Other URI schemas do not have a maximum.

Returns: The added manifest entries.

Raises:

• ArtifactFinalizedError: You cannot make changes to the current artifact version because it is finalized. Log a new artifact version instead.

method Artifact.checkout

checkout(root: 'str | None' = None) → str

Replace the specified root directory with the contents of the artifact.

WARNING: This will delete all files in root that are not included in the artifact.

Args:

• root: The directory to replace with this artifact’s files.

Returns: The path of the checked out contents.

Raises:

• ArtifactNotLoggedError: If the artifact is not logged.

method Artifact.delete

delete(delete_aliases: 'bool' = False) → None

Delete an artifact and its files.

If called on a linked artifact, only the link is deleted, and the source artifact is unaffected.

Use artifact.unlink() instead of artifact.delete() to remove a link between a source artifact and a linked artifact.

Args:

• delete_aliases: If set to True, deletes all aliases associated with the artifact. Otherwise, this raises an exception if the artifact has existing aliases. This parameter is ignored if the artifact is linked (a member of a portfolio collection).

Raises:

• ArtifactNotLoggedError: If the artifact is not logged.

method Artifact.download

download(
    root: 'StrPath | None' = None,
    allow_missing_references: 'bool' = False,
    skip_cache: 'bool | None' = None,
    path_prefix: 'StrPath | None' = None,
    multipart: 'bool | None' = None
) → FilePathStr

Download the contents of the artifact to the specified root directory.

Existing files located within root are not modified. Explicitly delete root before you call download if you want the contents of root to exactly match the artifact.

Args:

• root: The directory W&B stores the artifact’s files.
• allow_missing_references: If set to True, any invalid reference paths will be ignored while downloading referenced files.
• skip_cache: If set to True, the artifact cache will be skipped when downloading and W&B will download each file into the default root or specified download directory.
• path_prefix: If specified, only files with a path that starts with the given prefix will be downloaded. Uses unix format (forward slashes).
• multipart: If set to None (default), the artifact will be downloaded in parallel using multipart download if individual file size is greater than 2GB. If set to True or False, the artifact will be downloaded in parallel or serially regardless of the file size.

Returns: The path to the downloaded contents.

Raises:

• ArtifactNotLoggedError: If the artifact is not logged.

method Artifact.file

file(root: 'str | None' = None) → StrPath

Download a single file artifact to the directory you specify with root.

Args:

• root: The root directory to store the file. Defaults to ./artifacts/self.name/.

Returns: The full path of the downloaded file.

Raises:

• ArtifactNotLoggedError: If the artifact is not logged.
• ValueError: If the artifact contains more than one file.

method Artifact.files

files(names: 'list[str] | None' = None, per_page: 'int' = 50) → ArtifactFiles

Iterate over all files stored in this artifact.

Args:

• names: The filename paths relative to the root of the artifact you wish to list.
• per_page: The number of files to return per request.

Returns: An iterator containing File objects.

Raises:

• ArtifactNotLoggedError: If the artifact is not logged.

method Artifact.finalize

finalize() → None

Finalize the artifact version.

You cannot modify an artifact version once it is finalized because the artifact is logged as a specific artifact version. Create a new artifact version to log more data to an artifact. An artifact is automatically finalized when you log the artifact with log_artifact.

method Artifact.get

get(name: 'str') → WBValue | None

Get the WBValue object located at the artifact relative name.

Args:

• name: The artifact relative name to retrieve.

Returns: W&B object that can be logged with run.log() and visualized in the W&B UI.

Raises:

• ArtifactNotLoggedError: if the artifact isn’t logged or the run is offline.

method Artifact.get_added_local_path_name

get_added_local_path_name(local_path: 'str') → str | None

Get the artifact relative name of a file added by a local filesystem path.

Args:

• local_path: The local path to resolve into an artifact relative name.

Returns: The artifact relative name.

method Artifact.get_entry

get_entry(name: 'StrPath') → ArtifactManifestEntry

Get the entry with the given name.

Args:

• name: The artifact relative name to get

Returns: A W&B object.

Raises:

• ArtifactNotLoggedError: if the artifact isn’t logged or the run is offline.
• KeyError: if the artifact doesn’t contain an entry with the given name.

method Artifact.get_path

get_path(name: 'StrPath') → ArtifactManifestEntry

Deprecated. Use get_entry(name).

method Artifact.is_draft

is_draft() → bool

Check if artifact is not saved.

Returns: Boolean. False if artifact is saved. True if artifact is not saved.

method Artifact.json_encode

json_encode() → dict[str, Any]

Returns the artifact encoded to the JSON format.

Returns: A dict with string keys representing attributes of the artifact.

method Artifact.link

link(target_path: 'str', aliases: 'list[str] | None' = None) → Artifact | None

Link this artifact to a portfolio (a promoted collection of artifacts).

Args:

• target_path: The path to the portfolio inside a project. The target path must adhere to one of the following schemas {portfolio}, {project}/{portfolio} or {entity}/{project}/{portfolio}. To link the artifact to the Model Registry, rather than to a generic portfolio inside a project, set target_path to the following schema {"model-registry"}/{Registered Model Name} or {entity}/{"model-registry"}/{Registered Model Name}.
• aliases: A list of strings that uniquely identifies the artifact inside the specified portfolio.

Raises:

• ArtifactNotLoggedError: If the artifact is not logged.

Returns: The linked artifact if linking was successful, otherwise None.

method Artifact.logged_by

logged_by() → Run | None

Get the W&B run that originally logged the artifact.

Returns: The name of the W&B run that originally logged the artifact.

Raises:

• ArtifactNotLoggedError: If the artifact is not logged.

method Artifact.new_draft

new_draft() → Artifact

Create a new draft artifact with the same content as this committed artifact.

Modifying an existing artifact creates a new artifact version known as an “incremental artifact”. The artifact returned can be extended or modified and logged as a new version.

Returns: An Artifact object.

Raises:

• ArtifactNotLoggedError: If the artifact is not logged.

method Artifact.new_file

new_file(
    name: 'str',
    mode: 'str' = 'x',
    encoding: 'str | None' = None
) → Iterator[IO]

Open a new temporary file and add it to the artifact.

Args:

• name: The name of the new file to add to the artifact.
• mode: The file access mode to use to open the new file.
• encoding: The encoding used to open the new file.

Returns: A new file object that can be written to. Upon closing, the file is automatically added to the artifact.

Raises:

• ArtifactFinalizedError: You cannot make changes to the current artifact version because it is finalized. Log a new artifact version instead.

method Artifact.remove

remove(item: 'StrPath | ArtifactManifestEntry') → None

Remove an item from the artifact.

Args:

• item: The item to remove. Can be a specific manifest entry or the name of an artifact-relative path. If the item matches a directory all items in that directory will be removed.

Raises:

• ArtifactFinalizedError: You cannot make changes to the current artifact version because it is finalized. Log a new artifact version instead.
• FileNotFoundError: If the item isn’t found in the artifact.

method Artifact.save

save(
    project: 'str | None' = None,
    settings: 'wandb.Settings | None' = None
) → None

Persist any changes made to the artifact.

If currently in a run, that run will log this artifact. If not currently in a run, a run of type “auto” is created to track this artifact.

Args:

• project: A project to use for the artifact in the case that a run is not already in context.
• settings: A settings object to use when initializing an automatic run. Most commonly used in testing harness.

method Artifact.unlink

unlink() → None

Unlink this artifact if it is currently a member of a promoted collection of artifacts.

Raises:

• ArtifactNotLoggedError: If the artifact is not logged.
• ValueError: If the artifact is not linked, in other words, it is not a member of a portfolio collection.

method Artifact.used_by

used_by() → list[Run]

Get a list of the runs that have used this artifact and its linked artifacts.

Returns: A list of Run objects.

Raises:

• ArtifactNotLoggedError: If the artifact is not logged.

method Artifact.verify

verify(root: 'str | None' = None) → None

Verify that the contents of an artifact match the manifest.

All files in the directory are checksummed and the checksums are then cross-referenced against the artifact’s manifest. References are not verified.

Args:

• root: The directory to verify. If None artifact will be downloaded to ‘./artifacts/self.name/’.

Raises:

• ArtifactNotLoggedError: If the artifact is not logged.
• ValueError: If the verification fails.

method Artifact.wait

wait(timeout: 'int | None' = None) → Artifact

If needed, wait for this artifact to finish logging.

Args:

• timeout: The time, in seconds, to wait.

Returns: An Artifact object.

4.2.2.2 - Run

class Run

A unit of computation logged by W&B. Typically, this is an ML experiment.

Call wandb.init() to create a new run. wandb.init() starts a new run and returns a wandb.Run object. Each run is associated with a unique ID (run ID). W&B recommends using a context (with statement) manager to automatically finish the run.

For distributed training experiments, you can either track each process separately using one run per process or track all processes to a single run. See Log distributed training experiments for more information.

You can log data to a run with wandb.Run.log(). Anything you log using wandb.Run.log() is sent to that run. See Create an experiment or wandb.init API reference page or more information.

There is a another Run object in the wandb.apis.public namespace. Use this object is to interact with runs that have already been created.

Attributes:

• summary: (Summary) A summary of the run, which is a dictionary-like object. For more information, see
• [Log summary metrics](https: //docs.wandb.ai/guides/track/log/log-summary/).

Examples: Create a run with wandb.init():

import wandb

# Start a new run and log some data
# Use context manager (`with` statement) to automatically finish the run
with wandb.init(entity="entity", project="project") as run:
    run.log({"accuracy": acc, "loss": loss})

property Run.config

Config object associated with this run.

property Run.config_static

Static config object associated with this run.

property Run.dir

The directory where files associated with the run are saved.

property Run.disabled

True if the run is disabled, False otherwise.

property Run.entity

The name of the W&B entity associated with the run.

Entity can be a username or the name of a team or organization.

property Run.group

Returns the name of the group associated with this run.

Grouping runs together allows related experiments to be organized and visualized collectively in the W&B UI. This is especially useful for scenarios such as distributed training or cross-validation, where multiple runs should be viewed and managed as a unified experiment.

In shared mode, where all processes share the same run object, setting a group is usually unnecessary, since there is only one run and no grouping is required.

property Run.id

Identifier for this run.

property Run.job_type

Name of the job type associated with the run.

View a run’s job type in the run’s Overview page in the W&B App.

You can use this to categorize runs by their job type, such as “training”, “evaluation”, or “inference”. This is useful for organizing and filtering runs in the W&B UI, especially when you have multiple runs with different job types in the same project. For more information, see Organize runs.

property Run.name

Display name of the run.

Display names are not guaranteed to be unique and may be descriptive. By default, they are randomly generated.

property Run.notes

Notes associated with the run, if there are any.

Notes can be a multiline string and can also use markdown and latex equations inside $$, like $x + 3$.

property Run.offline

True if the run is offline, False otherwise.

property Run.path

Path to the run.

Run paths include entity, project, and run ID, in the format entity/project/run_id.

property Run.project

Name of the W&B project associated with the run.

property Run.project_url

URL of the W&B project associated with the run, if there is one.

Offline runs do not have a project URL.

property Run.resumed

True if the run was resumed, False otherwise.

property Run.settings

A frozen copy of run’s Settings object.

property Run.start_time

Unix timestamp (in seconds) of when the run started.

property Run.sweep_id

Identifier for the sweep associated with the run, if there is one.

property Run.sweep_url

URL of the sweep associated with the run, if there is one.

Offline runs do not have a sweep URL.

property Run.tags

Tags associated with the run, if there are any.

property Run.url

The url for the W&B run, if there is one.

Offline runs will not have a url.

method Run.alert

alert(
    title: 'str',
    text: 'str',
    level: 'str | AlertLevel | None' = None,
    wait_duration: 'int | float | timedelta | None' = None
) → None

Create an alert with the given title and text.

Args:

• title: The title of the alert, must be less than 64 characters long.
• text: The text body of the alert.
• level: The alert level to use, either: INFO, WARN, or ERROR.
• wait_duration: The time to wait (in seconds) before sending another alert with this title.

method Run.define_metric

define_metric(
    name: 'str',
    step_metric: 'str | wandb_metric.Metric | None' = None,
    step_sync: 'bool | None' = None,
    hidden: 'bool | None' = None,
    summary: 'str | None' = None,
    goal: 'str | None' = None,
    overwrite: 'bool | None' = None
) → wandb_metric.Metric

Customize metrics logged with wandb.Run.log().

Args:

• name: The name of the metric to customize.
• step_metric: The name of another metric to serve as the X-axis for this metric in automatically generated charts.
• step_sync: Automatically insert the last value of step_metric into wandb.Run.log() if it is not provided explicitly. Defaults to True if step_metric is specified.
• hidden: Hide this metric from automatic plots.
• summary: Specify aggregate metrics added to summary. Supported aggregations include “min”, “max”, “mean”, “last”, “best”, “copy” and “none”. “best” is used together with the goal parameter. “none” prevents a summary from being generated. “copy” is deprecated and should not be used.
• goal: Specify how to interpret the “best” summary type. Supported options are “minimize” and “maximize”.
• overwrite: If false, then this call is merged with previous define_metric calls for the same metric by using their values for any unspecified parameters. If true, then unspecified parameters overwrite values specified by previous calls.

Returns: An object that represents this call but can otherwise be discarded.

method Run.display

display(height: 'int' = 420, hidden: 'bool' = False) → bool

Display this run in Jupyter.

method Run.finish

finish(exit_code: 'int | None' = None, quiet: 'bool | None' = None) → None

Finish a run and upload any remaining data.

Marks the completion of a W&B run and ensures all data is synced to the server. The run’s final state is determined by its exit conditions and sync status.

Run States:

• Running: Active run that is logging data and/or sending heartbeats.
• Crashed: Run that stopped sending heartbeats unexpectedly.
• Finished: Run completed successfully (exit_code=0) with all data synced.
• Failed: Run completed with errors (exit_code!=0).
• Killed: Run was forcibly stopped before it could finish.

Args:

• exit_code: Integer indicating the run’s exit status. Use 0 for success, any other value marks the run as failed.
• quiet: Deprecated. Configure logging verbosity using wandb.Settings(quiet=...).

method Run.finish_artifact

finish_artifact(
    artifact_or_path: 'Artifact | str',
    name: 'str | None' = None,
    type: 'str | None' = None,
    aliases: 'list[str] | None' = None,
    distributed_id: 'str | None' = None
) → Artifact

Finishes a non-finalized artifact as output of a run.

Subsequent “upserts” with the same distributed ID will result in a new version.

Args:

• artifact_or_path: A path to the contents of this artifact, can be in the following forms: - /local/directory - /local/directory/file.txt - s3://bucket/path You can also pass an Artifact object created by calling wandb.Artifact.
• name: An artifact name. May be prefixed with entity/project. Valid names can be in the following forms: - name:version - name:alias - digest This will default to the basename of the path prepended with the current run id if not specified.
• type: The type of artifact to log, examples include dataset, model
• aliases: Aliases to apply to this artifact, defaults to ["latest"]
• distributed_id: Unique string that all distributed jobs share. If None, defaults to the run’s group name.

Returns: An Artifact object.

method Run.link_artifact

link_artifact(
    artifact: 'Artifact',
    target_path: 'str',
    aliases: 'list[str] | None' = None
) → Artifact | None

Link the given artifact to a portfolio (a promoted collection of artifacts).

Linked artifacts are visible in the UI for the specified portfolio.

Args:

• artifact: the (public or local) artifact which will be linked
• target_path: str - takes the following forms: {portfolio}, {project}/{portfolio}, or {entity}/{project}/{portfolio}
• aliases: List[str] - optional alias(es) that will only be applied on this linked artifact inside the portfolio. The alias “latest” will always be applied to the latest version of an artifact that is linked.

Returns: The linked artifact if linking was successful, otherwise None.

method Run.link_model

link_model(
    path: 'StrPath',
    registered_model_name: 'str',
    name: 'str | None' = None,
    aliases: 'list[str] | None' = None
) → Artifact | None

Log a model artifact version and link it to a registered model in the model registry.

Linked model versions are visible in the UI for the specified registered model.

This method will:

• Check if ’name’ model artifact has been logged. If so, use the artifact version that matches the files located at ‘path’ or log a new version. Otherwise log files under ‘path’ as a new model artifact, ’name’ of type ‘model’.
• Check if registered model with name ‘registered_model_name’ exists in the ‘model-registry’ project. If not, create a new registered model with name ‘registered_model_name’.
• Link version of model artifact ’name’ to registered model, ‘registered_model_name’.
• Attach aliases from ‘aliases’ list to the newly linked model artifact version.

Args:

• path: (str) A path to the contents of this model, can be in the following forms:
  • /local/directory
  • /local/directory/file.txt
  • s3://bucket/path
• registered_model_name: The name of the registered model that the model is to be linked to. A registered model is a collection of model versions linked to the model registry, typically representing a team’s specific ML Task. The entity that this registered model belongs to will be derived from the run.
• name: The name of the model artifact that files in ‘path’ will be logged to. This will default to the basename of the path prepended with the current run id if not specified.
• aliases: Aliases that will only be applied on this linked artifact inside the registered model. The alias “latest” will always be applied to the latest version of an artifact that is linked.

Raises:

• AssertionError: If registered_model_name is a path or if model artifact ’name’ is of a type that does not contain the substring ‘model’.
• ValueError: If name has invalid special characters.

Returns: The linked artifact if linking was successful, otherwise None.

method Run.log

log(
    data: 'dict[str, Any]',
    step: 'int | None' = None,
    commit: 'bool | None' = None
) → None

Upload run data.

Use log to log data from runs, such as scalars, images, video, histograms, plots, and tables. See Log objects and media for code snippets, best practices, and more.

Basic usage:

import wandb

with wandb.init() as run:
     run.log({"train-loss": 0.5, "accuracy": 0.9})

The previous code snippet saves the loss and accuracy to the run’s history and updates the summary values for these metrics.

Visualize logged data in a workspace at wandb.ai, or locally on a self-hosted instance of the W&B app, or export data to visualize and explore locally, such as in a Jupyter notebook, with the Public API.

Logged values don’t have to be scalars. You can log any W&B supported Data Type such as images, audio, video, and more. For example, you can use wandb.Table to log structured data. See Log tables, visualize and query data tutorial for more details.

W&B organizes metrics with a forward slash (/) in their name into sections named using the text before the final slash. For example, the following results in two sections named “train” and “validate”:

with wandb.init() as run:
     # Log metrics in the "train" section.
     run.log(
         {
             "train/accuracy": 0.9,
             "train/loss": 30,
             "validate/accuracy": 0.8,
             "validate/loss": 20,
         }
     )

Only one level of nesting is supported; run.log({"a/b/c": 1}) produces a section named “a/b”.

run.log() is not intended to be called more than a few times per second. For optimal performance, limit your logging to once every N iterations, or collect data over multiple iterations and log it in a single step.

By default, each call to log creates a new “step”. The step must always increase, and it is not possible to log to a previous step. You can use any metric as the X axis in charts. See Custom log axes for more details.

In many cases, it is better to treat the W&B step like you’d treat a timestamp rather than a training step.

with wandb.init() as run:
     # Example: log an "epoch" metric for use as an X axis.
     run.log({"epoch": 40, "train-loss": 0.5})

It is possible to use multiple wandb.Run.log() invocations to log to the same step with the step and commit parameters. The following are all equivalent:

with wandb.init() as run:
     # Normal usage:
     run.log({"train-loss": 0.5, "accuracy": 0.8})
     run.log({"train-loss": 0.4, "accuracy": 0.9})

     # Implicit step without auto-incrementing:
     run.log({"train-loss": 0.5}, commit=False)
     run.log({"accuracy": 0.8})
     run.log({"train-loss": 0.4}, commit=False)
     run.log({"accuracy": 0.9})

     # Explicit step:
     run.log({"train-loss": 0.5}, step=current_step)
     run.log({"accuracy": 0.8}, step=current_step)
     current_step += 1
     run.log({"train-loss": 0.4}, step=current_step)
     run.log({"accuracy": 0.9}, step=current_step)

Args:

• data: A dict with str keys and values that are serializable
• Python objects including: int, float and string; any of the wandb.data_types; lists, tuples and NumPy arrays of serializable Python objects; other dicts of this structure.
• step: The step number to log. If None, then an implicit auto-incrementing step is used. See the notes in the description.
• commit: If true, finalize and upload the step. If false, then accumulate data for the step. See the notes in the description. If step is None, then the default is commit=True; otherwise, the default is commit=False.

Examples: For more and more detailed examples, see our guides to logging.

Basic usage

import wandb

with wandb.init() as run:
    run.log({"train-loss": 0.5, "accuracy": 0.9

Incremental logging

import wandb

with wandb.init() as run:
    run.log({"loss": 0.2}, commit=False)
    # Somewhere else when I'm ready to report this step:
    run.log({"accuracy": 0.8})

Histogram

import numpy as np
import wandb

# sample gradients at random from normal distribution
gradients = np.random.randn(100, 100)
with wandb.init() as run:
    run.log({"gradients": wandb.Histogram(gradients)})

Image from NumPy

import numpy as np
import wandb

with wandb.init() as run:
    examples = []
    for i in range(3):
         pixels = np.random.randint(low=0, high=256, size=(100, 100, 3))
         image = wandb.Image(pixels, caption=f"random field {i}")
         examples.append(image)
    run.log({"examples": examples})

Image from PIL

import numpy as np
from PIL import Image as PILImage
import wandb

with wandb.init() as run:
    examples = []
    for i in range(3):
         pixels = np.random.randint(
             low=0,
             high=256,
             size=(100, 100, 3),
             dtype=np.uint8,
         )
         pil_image = PILImage.fromarray(pixels, mode="RGB")
         image = wandb.Image(pil_image, caption=f"random field {i}")
         examples.append(image)
    run.log({"examples": examples})

Video from NumPy

import numpy as np
import wandb

with wandb.init() as run:
    # axes are (time, channel, height, width)
    frames = np.random.randint(
         low=0,
         high=256,
         size=(10, 3, 100, 100),
         dtype=np.uint8,
    )
    run.log({"video": wandb.Video(frames, fps=4)})

Matplotlib plot

from matplotlib import pyplot as plt
import numpy as np
import wandb

with wandb.init() as run:
    fig, ax = plt.subplots()
    x = np.linspace(0, 10)
    y = x * x
    ax.plot(x, y)  # plot y = x^2
    run.log({"chart": fig})

PR Curve

import wandb

with wandb.init() as run:
    run.log({"pr": wandb.plot.pr_curve(y_test, y_probas, labels)})

3D Object

import wandb

with wandb.init() as run:
    run.log(
         {
             "generated_samples": [
                 wandb.Object3D(open("sample.obj")),
                 wandb.Object3D(open("sample.gltf")),
                 wandb.Object3D(open("sample.glb")),
             ]
         }
    )

Raises:

• wandb.Error: If called before wandb.init().
• ValueError: If invalid data is passed.

method Run.log_artifact

log_artifact(
    artifact_or_path: 'Artifact | StrPath',
    name: 'str | None' = None,
    type: 'str | None' = None,
    aliases: 'list[str] | None' = None,
    tags: 'list[str] | None' = None
) → Artifact

Declare an artifact as an output of a run.

Args:

• artifact_or_path: (str or Artifact) A path to the contents of this artifact, can be in the following forms: - /local/directory - /local/directory/file.txt - s3://bucket/path You can also pass an Artifact object created by calling wandb.Artifact.
• name: (str, optional) An artifact name. Valid names can be in the following forms: - name:version - name:alias - digest This will default to the basename of the path prepended with the current run id if not specified.
• type: (str) The type of artifact to log, examples include dataset, model
• aliases: (list, optional) Aliases to apply to this artifact, defaults to ["latest"]
• tags: (list, optional) Tags to apply to this artifact, if any.

Returns: An Artifact object.

method Run.log_code

log_code(
    root: 'str | None' = '.',
    name: 'str | None' = None,
    include_fn: 'Callable[[str, str], bool] | Callable[[str], bool]' = <function _is_py_requirements_or_dockerfile at 0x102da5f30>,
    exclude_fn: 'Callable[[str, str], bool] | Callable[[str], bool]' = <function exclude_wandb_fn at 0x103b4c5e0>
) → Artifact | None

Save the current state of your code to a W&B Artifact.

By default, it walks the current directory and logs all files that end with .py.

Args:

• root: The relative (to os.getcwd()) or absolute path to recursively find code from.
• name: (str, optional) The name of our code artifact. By default, we’ll name the artifact source-$PROJECT_ID-$ENTRYPOINT_RELPATH. There may be scenarios where you want many runs to share the same artifact. Specifying name allows you to achieve that.
• include_fn: A callable that accepts a file path and (optionally) root path and returns True when it should be included and False otherwise. This
• defaults to lambda path, root: path.endswith(".py").
• exclude_fn: A callable that accepts a file path and (optionally) root path and returns True when it should be excluded and False otherwise. This defaults to a function that excludes all files within <root>/.wandb/ and <root>/wandb/ directories.

Examples: Basic usage

import wandb

with wandb.init() as run:
    run.log_code()

Advanced usage

import wandb

with wandb.init() as run:
    run.log_code(
         root="../",
         include_fn=lambda path: path.endswith(".py") or path.endswith(".ipynb"),
         exclude_fn=lambda path, root: os.path.relpath(path, root).startswith(
             "cache/"
         ),
    )

Returns: An Artifact object if code was logged

method Run.log_model

log_model(
    path: 'StrPath',
    name: 'str | None' = None,
    aliases: 'list[str] | None' = None
) → None

Logs a model artifact containing the contents inside the ‘path’ to a run and marks it as an output to this run.

The name of model artifact can only contain alphanumeric characters, underscores, and hyphens.

Args:

• path: (str) A path to the contents of this model, can be in the following forms: - /local/directory - /local/directory/file.txt - s3://bucket/path
• name: A name to assign to the model artifact that the file contents will be added to. This will default to the basename of the path prepended with the current run id if not specified.
• aliases: Aliases to apply to the created model artifact, defaults to ["latest"]

Raises:

• ValueError: If name has invalid special characters.

Returns: None

method Run.mark_preempting

mark_preempting() → None

Mark this run as preempting.

Also tells the internal process to immediately report this to server.

method Run.restore

restore(
    name: 'str',
    run_path: 'str | None' = None,
    replace: 'bool' = False,
    root: 'str | None' = None
) → None | TextIO

Download the specified file from cloud storage.

File is placed into the current directory or run directory. By default, will only download the file if it doesn’t already exist.

Args:

• name: The name of the file.
• run_path: Optional path to a run to pull files from, i.e. username/project_name/run_id if wandb.init has not been called, this is required.
• replace: Whether to download the file even if it already exists locally
• root: The directory to download the file to. Defaults to the current directory or the run directory if wandb.init was called.

Returns: None if it can’t find the file, otherwise a file object open for reading.

Raises:

• CommError: If W&B can’t connect to the W&B backend.
• ValueError: If the file is not found or can’t find run_path.

method Run.save

save(
    glob_str: 'str | os.PathLike',
    base_path: 'str | os.PathLike | None' = None,
    policy: 'PolicyName' = 'live'
) → bool | list[str]

Sync one or more files to W&B.

Relative paths are relative to the current working directory.

A Unix glob, such as “myfiles/*”, is expanded at the time save is called regardless of the policy. In particular, new files are not picked up automatically.

A base_path may be provided to control the directory structure of uploaded files. It should be a prefix of glob_str, and the directory structure beneath it is preserved.

When given an absolute path or glob and no base_path, one directory level is preserved as in the example above.

Args:

• glob_str: A relative or absolute path or Unix glob.
• base_path: A path to use to infer a directory structure; see examples.
• policy: One of live, now, or end.
  • live: upload the file as it changes, overwriting the previous version
  • now: upload the file once now
  • end: upload file when the run ends

Returns: Paths to the symlinks created for the matched files.

For historical reasons, this may return a boolean in legacy code.

import wandb

run = wandb.init()

run.save("these/are/myfiles/*")
# => Saves files in a "these/are/myfiles/" folder in the run.

run.save("these/are/myfiles/*", base_path="these")
# => Saves files in an "are/myfiles/" folder in the run.

run.save("/User/username/Documents/run123/*.txt")
# => Saves files in a "run123/" folder in the run. See note below.

run.save("/User/username/Documents/run123/*.txt", base_path="/User")
# => Saves files in a "username/Documents/run123/" folder in the run.

run.save("files/*/saveme.txt")
# => Saves each "saveme.txt" file in an appropriate subdirectory
#    of "files/".

# Explicitly finish the run since a context manager is not used.
run.finish()

method Run.status

status() → RunStatus

Get sync info from the internal backend, about the current run’s sync status.

method Run.unwatch

unwatch(
    models: 'torch.nn.Module | Sequence[torch.nn.Module] | None' = None
) → None

Remove pytorch model topology, gradient and parameter hooks.

Args:

• models: Optional list of pytorch models that have had watch called on them.

method Run.upsert_artifact

upsert_artifact(
    artifact_or_path: 'Artifact | str',
    name: 'str | None' = None,
    type: 'str | None' = None,
    aliases: 'list[str] | None' = None,
    distributed_id: 'str | None' = None
) → Artifact

Declare (or append to) a non-finalized artifact as output of a run.

Note that you must call run.finish_artifact() to finalize the artifact. This is useful when distributed jobs need to all contribute to the same artifact.

Args:

• artifact_or_path: A path to the contents of this artifact, can be in the following forms:
  • /local/directory
  • /local/directory/file.txt
  • s3://bucket/path
• name: An artifact name. May be prefixed with “entity/project”. Defaults to the basename of the path prepended with the current run ID if not specified. Valid names can be in the following forms:
  • name:version
  • name:alias
  • digest
• type: The type of artifact to log. Common examples include dataset, model.
• aliases: Aliases to apply to this artifact, defaults to ["latest"].
• distributed_id: Unique string that all distributed jobs share. If None, defaults to the run’s group name.

Returns: An Artifact object.

method Run.use_artifact

use_artifact(
    artifact_or_name: 'str | Artifact',
    type: 'str | None' = None,
    aliases: 'list[str] | None' = None,
    use_as: 'str | None' = None
) → Artifact

Declare an artifact as an input to a run.

Call download or file on the returned object to get the contents locally.

Args:

• artifact_or_name: The name of the artifact to use. May be prefixed with the name of the project the artifact was logged to ("" or “/”). If no entity is specified in the name, the Run or API setting’s entity is used. Valid names can be in the following forms
  • name:version
  • name:alias
• type: The type of artifact to use.
• aliases: Aliases to apply to this artifact
• use_as: This argument is deprecated and does nothing.