TensorBoard 2.2.0

The 2.2 minor series tracks TensorFlow 2.2.

Features

• Profile plugin now should be pip installed from tensorboard-plugin-profile.
  The new version works in Chrome 80 and Firefox, has better model insights and
  will be more actively maintained.
• Add S3_ENDPOINT variable (#3368) - thanks @thealphacod3r
• Confirm that the connection to tensorboard works or change to localhost
  (#2371) - thanks @miguelmorin
• Update --reload_multifile_inactive_secs default to 24 hours (#3243)
• New tensorboard dev update-metadata command allows for updating the name and
  description of experiments (#3277)
• Improved organization of artifacts downloaded during export from
  TensorBoard.dev (#3307)

Bug fixes

• Fix for #3282 where the tooltip would remain even after the mouse leaves the
  plot (#3347)
• Internal fix: HParams summary protos now properly include tensor values (#3386)
• Fixes to profiling tutorial (#3372 & #3381)

Breaking Changes

• Note: As of TensorBoard 2.1.1+, only Python 3 is supported. There will be no
  further releases for Python 2 as per
  https://groups.google.com/a/tensorflow.org/forum/#!topic/developers/ifEAGK3aPls

TensorBoard 2.1.1

Features

• Uploader: Added ability to upload and modify experiment name and description (#3277)

Breaking changes

• Only Python 3 is supported for this patch. There will be no further releases for Python 2 as per
  https://groups.google.com/a/tensorflow.org/forum/#!topic/developers/ifEAGK3aPls

TensorBoard 2.1.0

The 2.1 minor series tracks TensorFlow 2.1.

Features

• Debugger: added ability to display Tensors as images, with selectable color map and zooming (#2729, #2764)
• What-If Tool improvements:
  • Added ability to set custom distance function for counterfactuals (#2607)
  • Added ability to explore counterfactual examples for regression models (#2647)
  • Added ability to consume arbitrary prediction-time information (#2660)
  • Added ability to slice performance statistics by numeric features (in addition to categorical features) (#2678, #2704).
  • Added PR/ROC curves by class for multi-class classification models (#2755)
• Improvements for plugin developers:
  • Added support for communication between TensorBoard and plugins in iframes (#2309, #2703)
  • (Experimental) Added library for improved plugin integration (#2708)
  • Enabled dynamic plugins in TensorBoard within Colab (#2798)
• Security improvements, e.g. Content Security Policy configurations
• Reduced overhead of image, audio, and histogram summary writing API methods (#2899) - thanks @hongjunChoi

Bug fixes

• What-If Tool:
  • Fixed sometimes-stuck threshold sliders (#2682)
  • Fixed PD plots in notebook mode with py3 kernels (#2669)
  • Fixed info dialogs re. Fairness optimization (#2694)
• Scalars dashboard: fixed unreliable data loading over slow network connections (#2825)
• Fixed potential corruption when reading files from disk, when TensorFlow is not installed (#2791)
• Fixed writing of histogram summaries when using TPUs (#2883) - thanks @hongjunChoi

TensorBoard.dev updates

• The tensorboard dev list subcommand now provides detailed metadata about
  each experiment.

TensorBoard 2.0.2

Features

• Improvements to TensorBoard.dev support:
  • New tensorboard dev list subcommand lists all experiments uploaded to
    TensorBoard.dev (#2903)
  • In the event of a transient backend issue or permanent breaking change, the
    uploader can now gracefully degrade and print a diagnostic (#2879)

TensorBoard 2.0.1

Features

• Preview of TensorBoard.dev uploader! Check out https://tensorboard.dev/ for
  information and usage instructions.

TensorBoard 2.0.0

The 2.0 minor series tracks TensorFlow 2.0.

Breaking changes

• TensorBoard now serves on localhost only by default to avoid unintentional
  overexposure. To expose TensorBoard to the network, either use a proxy, bind
  to a specific hostname or IP address by using the --host flag, or explicitly
  enable the previous behavior of binding on all network interfaces by passing
  the flag --bind_all. See PR #2589.

• The --logdir flag no longer supports passing multiple comma-delimited paths,
  which means that it now supports paths containing literal comma and colon
  characters, like ./logs/m=10,n=20,lr=0.001 or ./logs/run_12:30:15. To
  mimic the old behavior, prefer using a tree of symlinks as it works with more
  plugins, but as a fallback the flag --logdir_spec exposes the old behavior.
  See PR #2664.

• Projector plugin visualize_embeddings() API now takes logdir as its first
  parameter rather than writer (which only supported TF 1.x summary writers).
  For backwards compatibility TF 1.x writers will still be accepted, but passing
  the logdir explicitly is preferred since it works without any dependency on
  TF 1.x or 2.x summary writing. See PR #2665.

• The namespace tensorboard.summary.* now aliases the summary API symbols in
  tensorboard.summary.v2.* rather than those in tensorboard.summary.v1.*.
  The old symbols can still be accessed under the .v1 names. Note that the
  new v2 API symbols are exposed in TF 2.0 as the new tf.summary.* API and
  this is normally how they should be used. See PR #2670.

Features

• Smarter log directory polling can be used by passing --reload_multifile=true
  to poll all "active" event files in a directory rather than only the last one.
  This avoids problems where data written to the non-last file never appears.
  See PR #1867 for details, including how to adjust the "active" threshold.

• What-If Tool now can sort PD plots by interestingness (#2461)

TensorBoard 1.15.0

The 1.15 minor series tracks TensorFlow 1.15.

Features

• Embeddings projector now shows sprite images in the nearest neighbors list
  (#2543) - thanks @beasteers
• When recording hyperparameters, the trial ID can now be customized, for easier
  integration with existing tuner systems (#2442)
• Improvements to Colab and Jupyter notebook integration:
  • The TENSORBOARD_BINARY environment variable can now be set to invoke a
    non-default tensorboard binary (#2386)
  • Error messages are now clearer when the TensorBoard binary fails to launch
    (#2395)
  • The %tensorboard magic no longer spams log messages when a different
    version of TensorBoard is already running on the same machine (#2470)
  • The %tensorboard magic can now be used in Jupyter notebooks running on
    hosts other than localhost (#2407)
• What-If Tool improvements:
  • Errors running inference are now surfaced in the What-If Tool UI (#2414)
  • Median error stats are now displayed in addition to mean error stats (#2434)
• Mesh plugin improvements:
  • Now compatible with TensorFlow 2.0 via a new summary_v2 module (#2443)
  • The number of vertices in the mesh can now be dynamic (#2373)
• Profile dashboard improvements:
  • Wasted time now appears in the node table, and can be used as a sort key
    (#2525)
  • Memory bandwidth utilization now appears in the dashboard header (#2525)
• Improvements for plugin developers:
  • Plugins can now be rendered in an iframe whose source is served from the
    plugin backend, eliminating the need to bundle a frontend with the
    TensorBoard binary
  • Plugins can now be discovered dynamically and loaded at runtime, by defining
    a tensorboard_plugins entry point
  • See our example dynamically loaded plugin for a plugin to
    use as a starting point, plus documentation
  • TensorBoard now uses Polymer 2.7 (#2392, et al.)

Bug fixes

• #2614 - “Toggle All Runs” button now behaves correctly on the first click when
  many runs are loaded (PR #2633)
• Scalar charts should no longer “become tiny” on certain kinds of rendering
  failures (PR #2605)
• #2028 - TensorBoard now logs less verbosely with Werkzeug 0.15.0 and up; it
  now behaves the same across Werkzeug versions (PR #2383)
• The What-If Tool can now properly compare two regression models in the initial
  Facets Dive view (PR #2414)
• Embedding projector metadata view now wraps long strings correctly (PR #2198)

TensorBoard 1.14.0

Features

• New hyperparameters dashboard: see tutorial and demo and
  summary APIs
• New dashboard for visualizing meshes and point clouds: see
  README
• Graph dashboard now shows Keras conceptual graph: see tutorial and
  demo
• Embedding projector now supports the UMAP dimensionality reduction
  algorithm (learn more about UMAP here) (#1901) -
  thanks @cannoneyed
• TensorBoard notebook support is stabilized: in a Colab or
  Jupyter notebook, run %load_ext tensorboard followed by %tensorboard --logdir ./path/to/logs
• Profile dashboard improvements:
  • New pod viewer tool to analyze TPU performance (#2111)
  • Now allows capturing profiles from TensorBoard (#1894)
• What-If Tool improvements:
  • Now available as a notebook widget for Jupyter and Colab: see
    demo
  • Now shows PR curves and F1 score (#2264)
  • Now supports Cloud AI Platform, including XGBoost models (#2194)
  • Now shows feature-level attributions for individual predictions, as
    applicable (#2252)
• Image dashboard now allows scrolling for large images (#2164) - thanks @lr1d
• Scalar chart smoothing now caps at 0.999 for convenience (#1974) - thanks
  @flostim
• Scalar chart scroll-to-zoom behavior now requires holding Alt (#2221)
• tensorboard now supports a --version command line argument (#2097) -
  thanks @shashvatshahi1998
• Python API now defines tensorboard.__version__ in addition to
  tensorboard.version.VERSION (#2026)

Bug fixes

• Projector metadata card now formats long words properly (PR #2016) - thanks
  @makseq
• #2010 - .tensorboard-info is now world-writable for multi-user *nix systems
  (PR #2131)
• #1989 - importlib.reload(tensorboard) now works properly (PR #2005)

TensorBoard 1.13.1

Bug fixes

• #1895 - Fix strftime-related launch error on Windows (PR #1900)
• #1794 - Fix What-If Tool loading examples without inference (PR #1898)
• #1914 - Disable the profile dashboard inside Colab, where it doesn’t work
• #1945 - Fix profile dashboard loading behavior

TensorBoard 1.13.0

The 1.13 minor series tracks TensorFlow 1.13.

Compatibility note: As of 1.13, TensorBoard has begun transitioning its own use
of some TensorFlow APIs to align with their naming in TF 2.0, and as a result
TensorBoard 1.13+ strictly requires TensorFlow 1.13+.

Features

• What-If tool notebook mode and general improvements
  • Now usable directly inside Jupyter and Colab notebooks (#1662, #1745, #1788)
  • Added comparison of multiple models (#1589, #1672)
  • Added CSV loading model (#1597)
  • Added global partial dependence plots (#1604)
  • Added custom prediction function support (#1842)
• (Alpha) TensorBoard can be embedded inside Juptyer and Colab notebooks via a
  %tensorboard magic, after loading the tb.notebook extension (#1813, #1822)
• Profile dashboard overview page now shows step time breakdown (PR #1683)
• Line chart "log" scale is now a true log scale (#1507)
• When no --port flag is specified, TensorBoard will now search for open ports
  near the default port (6006) if that port is already in use (#1851)

Performance improvements

• Faster event file loading by caching runtime check (PR #1686) - thanks @abiro

Bug fixes

• #786 (partial) - Avoid trying to smooth plots of constant y-value (PR #1698)
• #1515 - Fix image right-click accessiblity in non-Chromium browsers (PR #1561)
• #1541 - Fix --event_file flag when using --inspect
• #1566 - Fix error on trying to import "google.protobuf.pyext" (PR #1887)
• #1567 - Fix display bug on line chart after toggling series selection
• #1598 - Fix clipping in graph dashboard PNG download (PR #1600)
• #1601 - Fix chart SVG download option in Firefox
• #1623 - Fix --path_prefix interpretation
• #1838 - Fix run selector synchronization across already-loaded dashboards