Release 1.6.0 #959

nfelt · 2018-02-08T00:32:45Z

This PR consists of cherry-picks for bug fixes and documentation updates for 1.6.0, plus the 1.6.0 release notes.

PTAL at the wording of the proposed 1.6.0 release notes. I'd like to avoid making any further edits to them when merging them back into master's RELEASE.md, so please speak up now if you see anything that needs fixing.

cc @qiuminxu @ckluk @caisq

* Start on docs for the debugger dashboard

That height caused blank space to be under the tensor data table. Furthermore, we make the tensor values data vertically scrollable so that rows of tensor values do not become inaccessible.

The custom scalars plugin lets the user overlap custom run-tag pairs within the same plot as well as visualize margins. This change modifies the main TensorBoard README to note those new features.

Cherrypick to add 1.5.1 relnotes back to master.

qiuminxu · 2018-02-08T01:19:56Z

LGTM! Thanks Nick.

jart · 2018-02-08T22:11:12Z

RELEASE.md

@@ -1,6 +1,57 @@
 # Release 1.6.0

-The 1.6 minor series tracks TensorFlow 1.6. Release notes upcoming.
+NOTICE: TensorBoard 1.6.0+ has moved to the `tensorboard` package name on PyPI:


Optional nit: in spirit of go/java-practices/javadoc IIRC? consider **Note:**

Ack. I already used NOTICE for 1.5.1 relnotes, which were approved twice already, and the PyPI description for 1.5.1, and I think it's better to be consistent.

jart · 2018-02-08T22:11:40Z

RELEASE.md

+
+## Highlights
+
+- (beta release) New Profile dashboard, which provides a suite of tools for


Optional Consider (Beta) New ...

jart · 2018-02-08T22:36:19Z

RELEASE.md

+
+- (beta release) New Profile dashboard, which provides a suite of tools for
+  inspecting TPU performance.  See for details:
+  https://github.com/tensorflow/tensorboard/tree/1.6/tensorboard/plugins/profile


Optional Hint: I like to sometimes put <http://blah> notation around markdown links, which is more proper and compatible IIRC.

Ack. I'd prefer not to, since it's unnecessary in most modern markdown renderers and in some it's definitely less compatible in that it might get interpreted as an unknown HTML tag and stripped rather than just rendered in plain text, which is a worse failure mode.

https://meta.stackexchange.com/questions/266263/links-in-angle-brackets-dont-show-up
jgm/peg-markdown#33

jart · 2018-02-08T22:38:29Z

RELEASE.md

-The 1.6 minor series tracks TensorFlow 1.6. Release notes upcoming.
+NOTICE: TensorBoard 1.6.0+ has moved to the `tensorboard` package name on PyPI:
+https://pypi.python.org/pypi/tensorboard. Only bugfix updates on 1.5.x will be
+applied to the old package name (`tensorflow-tensorboard`). To upgrade to


Optional alternative to consider:

... applied to the old package name (`tensorflow-tensorboard`). To upgrade: pip uninstall tensorflow-tensorboard pip install tensorboard See also "Known Issues" below, if you run into problems.

Ack. I already used the inline style for 1.5.1 relnotes, which were approved twice already, and the PyPI description for 1.5.1, and I think it's better to be consistent.

jart · 2018-02-08T22:40:48Z

tensorboard/plugins/debugger/README.md

@@ -0,0 +1,175 @@
+# The Debugger Dashboard
+
+The debugger dashboard offers a graphical user interface for the [TensorFlow debugger](https://www.tensorflow.org/programmers_guide/debugger). For instance, this dashboard enables users to


Optional: s/The debugger dashboard/This dashboard/

This is a cherry-pick of a README written by Chi in #942. If you think these edits are release-blocking, please send a PR with changes to Chi ASAP and I will cherry-pick it into the release PR once it's merged into master - I'm not editing this in the release PR. Ditto for all the other comments on this file.

jart · 2018-02-08T22:41:48Z

tensorboard/plugins/debugger/README.md

@@ -0,0 +1,175 @@
+# The Debugger Dashboard
+
+The debugger dashboard offers a graphical user interface for the [TensorFlow debugger](https://www.tensorflow.org/programmers_guide/debugger). For instance, this dashboard enables users to


Optional ProTip: Did you know it's possible to do hyperlinks in markdown where you punt the long ugly URLs to after the section, or to the very bottom entirely?

jart · 2018-02-08T22:42:17Z

tensorboard/plugins/debugger/README.md

+
+This command demonstrates how to set the debugger port to 6064. 
+
+```


maybe put sh or bash after backticks, just for more semantic web sake?

jart · 2018-02-08T22:42:25Z

tensorboard/plugins/debugger/README.md

+
+## Start TensorBoard with the `debugger_port` flag.
+
+To enable the debugger dashboard, pass the `debugger_port` flag to TensorBoard. TensorBoard will then both receive gRPC messages from and issue gRPC messages to model logic via this port.


Fill this paragraph please. 72 chars is good. 80 is ok. Note we have a markdown auto formatter internally you can probably run.

jart · 2018-02-08T22:44:23Z

tensorboard/plugins/profile/tf_profile_dashboard/tf-profile-dashboard.html

+    <a href="https://github.com/tensorflow/tensorboard/blob/1.6/tensorboard/plugins/profile/README.md">README</a>
+    and perhaps the
+    <a
+     href="https://cloud.google.com/tpu/docs/cloud-tpu-tools"


Nit: HTML style is 100 column limit. say things like this:

<doodle foo="bar" omg="lol>a</doodle>

Per go/htmlstyle even though we currently aren't compliant in many places.

Note: We sometimes need to be ugly because, unlike Soy, Polymer doesn't let us control whitespace.

This is a cherry-pick of docs written by Qiumin in #957. If you think these edits are release-blocking, please send a PR with changes to me or Qiumin ASAP and I will cherry-pick it into the release PR once it's merged into master - I'm not editing this in the release PR. Ditto for all the other comments on this file.

jart · 2018-02-08T22:46:00Z

tensorboard/plugins/profile/tf_profile_dashboard/tf-profile-dashboard.html

+    <a href="https://pypi.python.org/pypi/cloud-tpu-profiler">cloud-tpu-profiler</a>.
+    </p>
+    <p>
+    If you’re new to TPUs, and want to find out how


Optional nit: I've been trying to train myself personally to avoid using "you" in documentation. I also try to go so far as to figure out prose that avoids saying "one" in its place. Not worth worrying about now. Something to keep in mind.

chihuahua and others added 6 commits February 6, 2018 18:23

Add docs and screenshots for the debugger plugin (tensorflow#942)

8fc80a0

* Start on docs for the debugger dashboard

Remove height 29% from tensor data table (tensorflow#943)

3ba74e9

That height caused blank space to be under the tensor data table. Furthermore, we make the tensor values data vertically scrollable so that rows of tensor values do not become inaccessible.

Minor tweaks to debugger's README.md (tensorflow#950)

fd28791

Update README to consider custom scalars plugin (tensorflow#951)

e70b8df

The custom scalars plugin lets the user overlap custom run-tag pairs within the same plot as well as visualize margins. This change modifies the main TensorBoard README to note those new features.

Added error message when no profile data was found. (tensorflow#957)

bdabc17

Add 1.5.1 relnotes to RELEASE.md (tensorflow#956)

241a762

Cherrypick to add 1.5.1 relnotes back to master.

nfelt requested review from jart and chihuahua February 8, 2018 00:32

chihuahua approved these changes Feb 8, 2018

View reviewed changes

jart approved these changes Feb 8, 2018

View reviewed changes

nfelt added 2 commits February 9, 2018 10:55

Add 1.6.0 relnotes to RELEASE.md

eb8b676

TensorBoard 1.6.0

7d9179f

nfelt force-pushed the 1.6 branch from 989d593 to 7d9179f Compare February 9, 2018 19:19

nfelt merged commit 83e3fad into tensorflow:1.6 Feb 9, 2018


		## Highlights

		- (beta release) New Profile dashboard, which provides a suite of tools for

		@@ -0,0 +1,175 @@
		# The Debugger Dashboard

		The debugger dashboard offers a graphical user interface for the [TensorFlow debugger](https://www.tensorflow.org/programmers_guide/debugger). For instance, this dashboard enables users to


		This command demonstrates how to set the debugger port to 6064.

		```


		## Start TensorBoard with the `debugger_port` flag.

		To enable the debugger dashboard, pass the `debugger_port` flag to TensorBoard. TensorBoard will then both receive gRPC messages from and issue gRPC messages to model logic via this port.

Release 1.6.0 #959

Release 1.6.0 #959

Uh oh!

Conversation

nfelt commented Feb 8, 2018

Uh oh!

qiuminxu commented Feb 8, 2018

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!