profile
viewpoint
Shanqing Cai caisq Google https://caisq.github.io/ Software Engineer @ Google Research

caisq/tensorflow 2

Computation using data flow graphs for scalable machine learning

caisq/NeuroEvolution-Vehicles 1

Raw code from Live Stream May 13 2019

caisq/tfjs-converter-1 1

Convert TensorFlow SavedModel and Keras models to TensorFlow.js

caisq/tfjs-layers-1 1

TensorFlow.js high-level layers API

caisq/tfjs-models 1

Pretrained models for TensorFlow.js

caisq/audapter_matlab 0

MATLAB code for Audapter

caisq/caisq.github.io 0

Shanqing Cai's Homepage (GitHub-hosted)

caisq/chrome-extension-typescript-starter 0

Chrome Extension TypeScript Starter

pull request commenttensorflow/tfjs

Subject: Add swish activation function support (#1804)

@rthadur @pyu10055 feel free to merge the PR when you're comfortable.

Triforcey

comment created time in 24 days

PullRequestReviewEvent
PullRequestReviewEvent

pull request commenttensorflow/tfjs

Support for ConvLSTM2D

@dikatok I think it is acceptable. I'd suggest doing it as a separate PR from this one, though.

dikatok

comment created time in 25 days

pull request commenttensorflow/tfjs

Support for ConvLSTM2D

@pyu10055 PTAL? I think PR is ready to merge.

dikatok

comment created time in a month

pull request commenttensorflow/tfjs

Support for ConvLSTM2D

Actually, running yarn test-snippet in the tfjs-layers folder should be the way. See: https://github.com/tensorflow/tfjs/blob/master/tfjs-layers/package.json#L65

dikatok

comment created time in a month

pull request commenttensorflow/tfjs

Support for ConvLSTM2D

@dikatok The snippet-testing code is here: https://github.com/tensorflow/tfjs/blob/master/tfjs-layers/scripts/test_snippets.ts

I'm not sure about the correct command to invoke it. But you can try figuring it out or maybe @pyu10055 can tell you more. It runs on CI for sure.

dikatok

comment created time in a month

pull request commenttensorflow/tfjs

Support for ConvLSTM2D

@dikatok The code snippet in the comments should use tf.* instead of tfc.*. This is why the CI is failing.

const filters = 3;
const kernelSize = 3;

const sequenceLength = 1;
const size = 5;
const channels = 3;

const inputShape = [sequenceLength, size, size, channels];
const input = tfc.ones(inputShape);

const cell = tf.layers.convLstm2dCell({filters, kernelSize});

cell.build(input.shape);

const outputSize = size - kernelSize + 1;
const outShape = [sequenceLength, outputSize, outputSize, filters];

const initialH = tfc.zeros(outShape);
const initialC = tfc.zeros(outShape);

const [o, h, c] = cell.call([x, initialH, initialC], {});
dikatok

comment created time in a month

pull request commenttensorflow/tfjs

Support for ConvLSTM2D

Looks like there are lint errors.

ERROR: (no-unnecessary-type-assertion) /workspace/tfjs-layers/src/layers/convolutional_recurrent.ts[160, 35]: This assertion is unnecessary since it does not change the type of the expression.

@dikatok Can you run yarn lint locally and fix all the errors?

dikatok

comment created time in a month

PullRequestReviewEvent
PullRequestReviewEvent
PullRequestReviewEvent

pull request commenttensorflow/tfjs

Support for ConvLSTM2D

@dikatok Sorry for the delay. I'm not reviewing this PR and will get back to you with a review by end of day today.

dikatok

comment created time in a month

push eventtensorflow/tfjs-models

Shanqing Cai

commit sha 33c84070b80c8b48826a77f8490a5fb7a82fc983

[SpeechCommands] Clean up training/browser-fft directory (#513) - Removes obsolete Python files, now that we have notebooks showing the complete workflow - Removes supporting TypeScript scripts and package.json file. - Updates README.md. - Also adds the step to write meta

view details

push time in a month

PR merged tensorflow/tfjs-models

Reviewers
[SpeechCommands] Clean up training/browser-fft directory cla: yes
  • Removes obsolete Python files, now that we have notebooks showing the complete workflow
  • Removes supporting TypeScript scripts and package.json file.
  • Updates README.md.
  • Also adds the step to write metadata.json to traning_custom_audio_model_in_python.ipynb

<!-- Reviewable:start --> This change is <img src="https://reviewable.io/review_button.svg" height="34" align="absmiddle" alt="Reviewable"/> <!-- Reviewable:end -->

+55 -1562

1 comment

11 changed files

caisq

pr closed time in a month

push eventcaisq/tfjs-models

Shanqing Cai

commit sha cd17028cf2d1b970da99b508f74ee6eaa8fa10be

[SpeechCommands] Update tflite_conversion.ipynb to TF 2.3.0 (#505) Co-authored-by: Ping Yu <4018+pyu10055@users.noreply.github.com>

view details

Shanqing Cai

commit sha 472011228bf4fcf57a417dd5af95036821d8747e

Merge branch 'master' into sc-train-cleanup

view details

push time in a month

PullRequestReviewEvent

push eventtensorflow/tfjs-models

Shanqing Cai

commit sha cd17028cf2d1b970da99b508f74ee6eaa8fa10be

[SpeechCommands] Update tflite_conversion.ipynb to TF 2.3.0 (#505) Co-authored-by: Ping Yu <4018+pyu10055@users.noreply.github.com>

view details

push time in a month

push eventcaisq/tfjs-models

Ann Yuan

commit sha 0966d7cd9bb2e2b7ef4d3d6cfa584010a2951fa7

fix (#511)

view details

Shanqing Cai

commit sha 0bab4198da2ec63077e7ccdeb6d095af0fc30c3a

Merge branch 'master' into sc-train-cleanup

view details

push time in a month

startedpolyzer/rllib.js

started time in a month

issue commenttensorflow/tensorboard

Using TensorBoard Debugger

+1 to what @bileschi said. Please describe what you did that led to the UnimlpementedError, and what version of TensorFlow and TensorBoard you are using.

anandh51295

comment created time in a month

push eventcaisq/tfjs-models

Shanqing Cai

commit sha aed8756ec329cef48cb2f003347844be9597953a

Remove obsolete Python test step

view details

push time in a month

push eventcaisq/tfjs-models

Shanqing Cai

commit sha c3560e3a7763bc269504d98520a194c626a4a2cf

Clarify deployment steps more

view details

push time in a month

push eventcaisq/tfjs-models

Shanqing Cai

commit sha 4fa44e759b99732a5acab955a5b3b9d05bdda7a1

Remove more files

view details

push time in a month

push eventcaisq/tfjs-models

Shanqing Cai

commit sha d1534b86f129bbd0b7964734a8c047911d5dc6d3

Add JavaScript code example to training_custom_audio_model_in_python.ipynb

view details

push time in a month

create barnchcaisq/tfjs-models

branch : sc-train-cleanup

created branch time in a month

push eventcaisq/tfjs-models

Ann Yuan

commit sha 0855c08bf4c9463438f82da74bddb5f55c365788

[handpose] Add WASM support. (#501)

view details

Ann Yuan

commit sha 27eea88a3b5a2edcc992c6a9522338b5dc402f61

[facemesh] Support rotated faces, improve docs. (#486)

view details

Shanqing Cai

commit sha e288c6ab88abadc3b1983730ad79bc43a95f030b

[Speech Commands] Add notebook for training custom models in Python (#506) * [Speech Commands] Add notebook to show how to train custom audio model in Python, using a set of .wav files.

view details

push time in a month

issue commenttensorflow/tfjs

Feature Request: Add Colab Notebook to show how to train an audio model from scratch in Python

Fixed by https://github.com/tensorflow/tfjs-models/pull/506

caisq

comment created time in a month

push eventtensorflow/tfjs-models

Shanqing Cai

commit sha e288c6ab88abadc3b1983730ad79bc43a95f030b

[Speech Commands] Add notebook for training custom models in Python (#506) * [Speech Commands] Add notebook to show how to train custom audio model in Python, using a set of .wav files.

view details

push time in a month

PR merged tensorflow/tfjs-models

[Speech Commands] Add notebook for training custom models in Python cla: yes

in Python, using a set of .wav files.

Fixes https://github.com/tensorflow/tfjs/issues/3719

<!-- Reviewable:start -->

This change is <img src="https://reviewable.io/review_button.svg" height="34" align="absmiddle" alt="Reviewable"/> <!-- Reviewable:end -->

+835 -0

5 comments

1 changed file

caisq

pr closed time in a month

issue closedtensorflow/tfjs

Feature Request: Add Colab Notebook to show how to train an audio model from scratch in Python

The audio model should use the WebAudio preprocessing layer like https://github.com/tensorflow/tfjs-models/blob/master/speech-commands/training/browser-fft/tflite_conversion.ipynb to support easy conversion to TF.js and TFLite.

Public datasets we can use for this:

  • https://www.tensorflow.org/datasets/catalog/speech_commands

closed time in a month

caisq

pull request commenttensorflow/tfjs-models

[Speech Commands] Add notebook for training custom models in Python

Thanks for the reviews, @pyu10055 and @lintian06. Merging PR.

caisq

comment created time in a month

pull request commenttensorflow/tfjs-models

[Speech Commands] Add notebook for training custom models in Python

Ping @pyu10055 if you can take another look at the PR, it would be great. Thanks.

caisq

comment created time in a month

push eventcaisq/tfjs-models

Shanqing Cai

commit sha d4bca85f506e6d72acf5c7fe5e6b6197437130ea

Add a sentence for clarity

view details

push time in a month

push eventcaisq/tfjs-models

Shanqing Cai

commit sha 0d0888b12a1ff2d10b61c57523f7d9b87b077368

Add paragraph about data format; Apply nbfmt.py

view details

push time in a month

Pull request review commenttensorflow/tfjs-models

[Speech Commands] Add notebook for training custom models in Python

+{

<p>Yes. We can safely assume that these libraries are all pre-installed for the following reasons:</p><p> </p><ol><li>They are installed by default on Google Colab.</li><li>These three are such widely-used libraries that virtually all data scientists and ML practitioners know how to install them. </li></ol>

caisq

comment created time in a month

PullRequestReviewEvent

Pull request review commenttensorflow/tfjs-models

[Speech Commands] Add notebook for training custom models in Python

+{

<p>Good question. I added a paragraph in this header section to answer that question.</p>

caisq

comment created time in a month

PullRequestReviewEvent

push eventtensorflow/tensorboard

Shanqing Cai

commit sha 55ec4ba857e9740f5eed5968c968e56f4622e9d6

[DebuggerV1] Remove code of deprecated plugin (#4007) * Motivation for features / changes * This PR removes the Python backend and web frontend code of the Debugger (V1) plugin. This plugin has been marked as deprecated and for removal in the 2.3.0 release notes. * Technical description of changes * #4033 has removed the reference to this plugin from other parts of TensorBoard. * This PR in turn removes code in tensorboard/plugins/debugger. * Detailed steps to verify changes work correctly (as executed by you) * Verified that Debugger V2 still functions properly * Internal CL for corresponding code removal is tested globally

view details

push time in a month

PR merged tensorflow/tensorboard

[DebuggerV1] Remove code of deprecated plugin cla: yes
  • Motivation for features / changes
    • This PR removes the Python backend and web frontend code of the Debugger (V1) plugin. This plugin has been marked as deprecated and for removal in the 2.3.0 release notes.
  • Technical description of changes
    • #4033 has removed the reference to this plugin from other parts of TensorBoard.
    • This PR in turn removes code in tensorboard/plugins/debugger.
  • Detailed steps to verify changes work correctly (as executed by you)
    • Verified that Debugger V2 still functions properly
    • Internal CL for corresponding code removal is tested globally
+0 -20673

2 comments

74 changed files

caisq

pr closed time in a month

push eventcaisq/tensorboard

Stephan Lee

commit sha 048695c9ef46f3534082198096e28335932ca4ae

add 'data:' to font-src CSP (#4027) TensorBoard depends on vaadin package and it uses data uri to bundle the font icons. This change also removes the global CSP whitelist validation since we already have validation in the security validator module.

view details

Stephan Lee

commit sha 95aed7141ae095a773145154b274b3963d23abee

polymer3: add web-animations-js shim to Polymer 3 (#4029) In this change, we use genrule to inline the content since the module cannot be safely imported from TypeScript internally. In order to have congruent build system, we just use genrule to concatenate already compiled sources.

view details

Shanqing Cai

commit sha 0e63b90fa999ffeed558654f68a61e91bd41dbab

[Debuggger V1] Remove external reference to the deprecated plugin (#4033) * Motivation for features / changes * This PR removes the references to the deprecated debugger (v1) plugin. This plugin has been marked as deprecated and for removal in the 2.3.0 release notes. The code itself will be removed in a follow-up change. * Technical description of changes * Removes external references to tensorboard/plugins/debugger, from BUILD files, Python modules and Polymer .html files that are not a part of plugins/debugger. * During testing, it was discovered that the BUILD files in plugins/audio and plugins/images needed to be fixed (i.e., dependency on the ":metadata" BUILD target added in each respective BUILD file) after plugins/debugger is removed. This probably was due to an unmasking of transitive dependencies of some sort. * Detailed steps to verify changes work correctly (as executed by you) * Verified that Debugger V2 still functions properly

view details

William Chargin

commit sha 67f8ab9c1583a9a14e6c52b2b30fb62234c4afe0

polymer: lock down component visibility (#4034) Summary: We’ve migrated all internal dependencies off of these targets, so we can now lock them down to prevent backsliding. Generated via Buildozer with: ``` buildozer 'remove visibility //visibility:public' \ //tensorboard/components{,_polymer3}/...:\* \ //tensorboard/plugins/...:%tf_web_library \ ; ``` …which imposes `//tensorboard:internal` or stricter due to #3566. Test Plan: A test sync (<http://cl/326725258>) passes presubmits, and a build graph query shows that these targets don’t have direct reverse dependencies from outside the internal package group. wchargin-branch: polymer-lockdown

view details

Stephan Lee

commit sha 75b58b74d4cb194b19653fc6719639c5c9ccc0bc

polymer3: make the polymer 3 entry point the default (#4035) This change changes the default entry point (/index.html) to the TensorBoard to Polymer 3. We rename/clean-up polymer3 names in files in a follow up.

view details

Shanqing Cai

commit sha 438bb408cdbed6e9d6297890913f67effe446bb6

Merge branch 'master' into debugger-v1-removal

view details

push time in a month

push eventcaisq/tensorboard

Stephan Lee

commit sha 95aed7141ae095a773145154b274b3963d23abee

polymer3: add web-animations-js shim to Polymer 3 (#4029) In this change, we use genrule to inline the content since the module cannot be safely imported from TypeScript internally. In order to have congruent build system, we just use genrule to concatenate already compiled sources.

view details

Shanqing Cai

commit sha 0e63b90fa999ffeed558654f68a61e91bd41dbab

[Debuggger V1] Remove external reference to the deprecated plugin (#4033) * Motivation for features / changes * This PR removes the references to the deprecated debugger (v1) plugin. This plugin has been marked as deprecated and for removal in the 2.3.0 release notes. The code itself will be removed in a follow-up change. * Technical description of changes * Removes external references to tensorboard/plugins/debugger, from BUILD files, Python modules and Polymer .html files that are not a part of plugins/debugger. * During testing, it was discovered that the BUILD files in plugins/audio and plugins/images needed to be fixed (i.e., dependency on the ":metadata" BUILD target added in each respective BUILD file) after plugins/debugger is removed. This probably was due to an unmasking of transitive dependencies of some sort. * Detailed steps to verify changes work correctly (as executed by you) * Verified that Debugger V2 still functions properly

view details

William Chargin

commit sha 67f8ab9c1583a9a14e6c52b2b30fb62234c4afe0

polymer: lock down component visibility (#4034) Summary: We’ve migrated all internal dependencies off of these targets, so we can now lock them down to prevent backsliding. Generated via Buildozer with: ``` buildozer 'remove visibility //visibility:public' \ //tensorboard/components{,_polymer3}/...:\* \ //tensorboard/plugins/...:%tf_web_library \ ; ``` …which imposes `//tensorboard:internal` or stricter due to #3566. Test Plan: A test sync (<http://cl/326725258>) passes presubmits, and a build graph query shows that these targets don’t have direct reverse dependencies from outside the internal package group. wchargin-branch: polymer-lockdown

view details

Stephan Lee

commit sha 75b58b74d4cb194b19653fc6719639c5c9ccc0bc

polymer3: make the polymer 3 entry point the default (#4035) This change changes the default entry point (/index.html) to the TensorBoard to Polymer 3. We rename/clean-up polymer3 names in files in a follow up.

view details

push time in a month

push eventtensorflow/tensorboard

Shanqing Cai

commit sha 0e63b90fa999ffeed558654f68a61e91bd41dbab

[Debuggger V1] Remove external reference to the deprecated plugin (#4033) * Motivation for features / changes * This PR removes the references to the deprecated debugger (v1) plugin. This plugin has been marked as deprecated and for removal in the 2.3.0 release notes. The code itself will be removed in a follow-up change. * Technical description of changes * Removes external references to tensorboard/plugins/debugger, from BUILD files, Python modules and Polymer .html files that are not a part of plugins/debugger. * During testing, it was discovered that the BUILD files in plugins/audio and plugins/images needed to be fixed (i.e., dependency on the ":metadata" BUILD target added in each respective BUILD file) after plugins/debugger is removed. This probably was due to an unmasking of transitive dependencies of some sort. * Detailed steps to verify changes work correctly (as executed by you) * Verified that Debugger V2 still functions properly

view details

push time in a month

PR merged tensorflow/tensorboard

[Debuggger V1] Remove external reference to the deprecated plugin cla: yes plugin:debugger
  • Motivation for features / changes
    • This PR removes the references to the deprecated debugger (v1) plugin. This plugin has been marked as deprecated and for removal in the 2.3.0 release notes. The code itself will be removed in a follow-up change.
  • Technical description of changes
    • Removes external references to tensorboard/plugins/debugger, from BUILD files, Python modules and Polymer .html files that are not a part of plugins/debugger.
  • During testing, it was discovered that the BUILD files in plugins/audio and plugins/images needed to be fixed (i.e., dependency on the ":metadata" BUILD target added in each respective BUILD file) after plugins/debugger is removed. This probably was due to an unmasking of transitive dependencies of some sort.
  • Detailed steps to verify changes work correctly (as executed by you)
    • Verified that Debugger V2 still functions properly
+2 -7

1 comment

8 changed files

caisq

pr closed time in a month

pull request commenttensorflow/tensorboard

[Debuggger V1] Remove external reference to the deprecated plugin

@nickfelt Done.

caisq

comment created time in a month

push eventcaisq/tensorboard

Shanqing Cai

commit sha 9b4ff006d13026ced8697f3e2e6ef9e1ec24a42f

Add :metadata dependencies to audio and image plugins

view details

push time in a month

PR opened tensorflow/tensorboard

[Debuggger V1] Remove external reference to the deprecated plugin
  • Motivation for features / changes
    • This PR removes the Python backend and web frontend code of the Debugger (V1) plugin. This plugin has been marked as deprecated and for removal in the 2.3.0 release notes.
  • Technical description of changes
    • Removes external references to tensorboard/plugins/debugger, from BUILD files, Python modules and Polymer .html files that are not a part of plugins/debugger.
  • Detailed steps to verify changes work correctly (as executed by you)
    • Verified that Debugger V2 still functions properly
+0 -7

0 comment

6 changed files

pr created time in a month

create barnchcaisq/tensorboard

branch : debugger-v1-removal-2

created branch time in a month

push eventcaisq/tensorboard

Stephan Lee

commit sha 82cbcda4ddfe061bce3ae36f26952df8df763f07

polymer3: fix data location rendering (#4026) Because we were using readOnly and was setting the value in the constructor, the delimitter pattern was never set.

view details

Nick Felt

commit sha 7e06535c7eae341ffdfcb7d276cc2e75628ab38a

migrate: fix regressions in hparams table view (#4028) * fix slightly too-narrow width for Show Metrics column * guard against empty sessionGroup in cell value rendering * guard against empty sessionGroup in details pane rendering

view details

Stephan Lee

commit sha 048695c9ef46f3534082198096e28335932ca4ae

add 'data:' to font-src CSP (#4027) TensorBoard depends on vaadin package and it uses data uri to bundle the font icons. This change also removes the global CSP whitelist validation since we already have validation in the security validator module.

view details

push time in a month

push eventcaisq/tensorboard

Stephan Lee

commit sha 82cbcda4ddfe061bce3ae36f26952df8df763f07

polymer3: fix data location rendering (#4026) Because we were using readOnly and was setting the value in the constructor, the delimitter pattern was never set.

view details

Nick Felt

commit sha 7e06535c7eae341ffdfcb7d276cc2e75628ab38a

migrate: fix regressions in hparams table view (#4028) * fix slightly too-narrow width for Show Metrics column * guard against empty sessionGroup in cell value rendering * guard against empty sessionGroup in details pane rendering

view details

Shanqing Cai

commit sha a6ee436d1584e5d8748dcc3c56749b742c1cb633

Merge branch 'master' of github.com:tensorflow/tensorboard into debugger-v1-removal

view details

push time in a month

push eventcaisq/tensorboard

Stephan Lee

commit sha e62b483232c8d982a9e900c74b01e4538bf2723a

chore: remove an ability to customize tsconfig (#3991) Two things: 1. internally, we cannot have a custom tsconfig so allowing custom one will only make sync harder 2. passing tsconfig-testing for test_only targets were quite tedious. Automatng that.

view details

Stephan Lee

commit sha 17909dca14e435ee862f79fcc510db77b46cf8fc

migrate: mesh plugin (#3940) Do note that for ease of migration, we have copied two modules from the tensorflow/graphics repository[1]. Do note that we made some manual changes to better type (to best of our knowledge based on the closure type annotations). Because version of THREE has been upgraded and changed, we may be using deprecated & removed API. We will be able to better tell whether we are doing the right thing only after when we do the functional testing. [1]: https://github.com/tensorflow/graphics/blob/master/tensorflow_graphics/tensorboard/mesh_visualizer/tf_mesh_dashboard/

view details

E

commit sha 736e5f68787f412617acfb8d2852c27bbc6f528f

Remove unnecessary py_binary loader (#4009) Internally, py_binary exists in the ambiently, and will cause errors if loaded manually. This removes the unnecessary load for py_binary. Tested internally with: cl/325902456

view details

William Chargin

commit sha 7b761f36f1a0ce3916b10bda9c17b8dbc10ca0f3

migrate: tf_custom_scalar_dashboard (#4002) Summary: Notable changes: - Fields of the `DataSeries` TypeScript class are no longer `private`, since the view code (previously in HTML and thus not typechecked) poked into them. - Categories now set `type: PREFIX_GROUP`, which is necessary to type-check even though it doesn’t end up making a functional difference due to implementation details of the category view. Test Plan: With `TB_POLYMER3=1 bazel run //tensorboard`: - custom scalars render, both margin charts and multi-line charts; - custom scalar charts are not paginated (by design); - smoothing works; - the axis selector works; - the data download links work; - the outliers toggle works. Googlers, see test sync: <Summary: Notable changes: - Fields of the `DataSeries` TypeScript class are no longer `private`, since the view code (previously in HTML and thus not typechecked) poked into them. - Categories now set `type: PREFIX_GROUP`, which is necessary to type-check even though it doesn’t end up making a functional difference due to implementation details of the category view. Test Plan: With `TB_POLYMER3=1 bazel run //tensorboard`: - custom scalars render, both margin charts and multi-line charts; - custom scalar charts are not paginated (by design); - smoothing works; - the axis selector works; - the data download links work; - the outliers toggle works. Googlers, see test sync: <http://cl/325552918>.

view details

E

commit sha 18c8cfb2599b98acdf55e7067b7aac63710596a2

migrate: prepare graph for conversion part 1 (#3921) Followup: https://github.com/tensorflow/tensorboard/pull/3932 This change adds npm dependencies for the graph dashboard: - dagre - @polymer/paper-radio-button - @polymer/paper-progress - @polymer/iron-list Googlers: This PR should be merged after cl/325528860 After this PR is synced internally, cl/325559670 can be submitted to unblock the final change that enables the Graph dashboard: https://github.com/tensorflow/tensorboard/pull/4005

view details

E

commit sha 16a2e7f189471d360d45538606fac51cfdb203b7

migrate: convert graph externally part 2 (#3932) Diffbase: https://github.com/tensorflow/tensorboard/pull/3921 Followup: https://github.com/tensorflow/tensorboard/pull/4004 Part 2 of the Polymer 3 conversion of the graph dashboard. This adds tags = manual to graph targets, but ensures they build successfully externally, including tf-graph-app's binary.

view details

E

commit sha 151373805447f46fc8ebb87837de0f069fea0609

migrate: graph part 2 (#4004) Diffbase: https://github.com/tensorflow/tensorboard/pull/3932 Followup: https://github.com/tensorflow/tensorboard/pull/4005 Migrating the graph dashboard requires some extra changes to make it pass more strict conformance rules internally. Most of the changes involve making type annotations looser. Notably: Fixed circular dependencies between JS modules. `tf_graph_common` contained lots and lots of cycles, including - scene.ts --> edge.ts --> scene.ts - node.ts --> annotation.ts --> node.ts These were resolved by either moving methods into `common.ts`, or in the case of `annotation.ts`, merging it into `node.ts`. The `tf_graph_common` modules now follow a proper layering scheme (higher layers can depend on lower layers, but never the other way around). layer 0: common, proto layer 1: util layer 2: parser, graph layer 3: op, template layer 4: hierarchy layer 5: render, loader layer 6: tf-graph-scene layer 7: contextmenu, edge, layout layer 8: scene layer 9: annotation/node

view details

E

commit sha 2a1a7234bd96eeaa06d0fb913b1cd5202ef7fae3

migrate: enable the converted graph dashboard part 4 (#4005) Diffbase: https://github.com/tensorflow/tensorboard/pull/4004 This is the last in a series of changes to convert plugins/graph to Polymer 3. This change - removes `tags=["manual"]` - enables the dashboard Tested externally with `TB_POLYMER3=1 bazel run tensorboard` and internally. Googlers, see cl/325351322.

view details

Stanley Bileschi

commit sha 465d0fb27761a56de27d0ac020e51abca22c3e2d

Update Getting Started notebook to use `--one_shot` uploader.

view details

Stanley Bileschi

commit sha 780ea6f0a76f34e4ffeb81611cde20d8d028c55c

spelling & newline

view details

Stephan Lee

commit sha b8d1d4cb3f9fa0279b1c897292ff9b0c90f6eac4

text_v2: actions and effects (#4012) Note that we still are not implementing reducers. Details: - sample data flow: 1. load the text plugin, it fires pluginChanged action. That fetches runTag map 2. based on the run tag map, we draw few tag groups. Few tag groups will be opened by default and some won't. The opened ones fire action for their visiblity changes with <run, tag> tuple of current page 3. that will trigger data fetches 4. on reload, we will reload what is currently on screen

view details

Stephan Lee

commit sha d5e5a4d1b5fb01b30560622b12e8f2394983bb91

polymer3: integrate mesh plugin to the migrated code (#4011) This change enables mesh plugin in the Polymer 3 world.

view details

Stanley Bileschi

commit sha 7609dc568711276a5ac8daf66f6da7e4c6932b8a

Fix missing metadata in tbdev getting started notebook.

view details

Stanley Bileschi

commit sha dba7b9ed8cea082944096a75834561fe9b0f1b77

Clarifies output during `--one_shot` upload. (#4015) Clarifies output during `--one_shot` upload. When `--one_shot` flag is provided, the user expects the uploader to pass through the logdir exactly once and then quit. This change replaces the log line "Listening for new data in logdir" with a more accurate line "Done passing through logdir", but only when `--one_shot` is provided.

view details

Stephan Lee

commit sha 8e90e89cff90cf2a941021806953a463107f8546

Migrate plugins/hparams (#4010) Migrates the hparams plugin. Do note that removed the vaadin-split-layout by simply using flex (824f9f5). Relates to #3955 Co-authored-by: Erik Luoe <luoe@chromium.org>

view details

Stephan Lee

commit sha 76c452c4ec78b51b03c755dbdcbced46da2c418a

Revert "Migrate plugins/hparams (#4010)" (#4016) This reverts commit 8e90e89cff90cf2a941021806953a463107f8546.

view details

Nick Felt

commit sha 8d73b81e249f8c45890b9951615ec27d3d8b1e95

migrate: plugins/pr_curve/polymer3 (#4013) * rename html to ts for better diff * tbmigrate output for better diff * yarn fix-lint * migrate: bulk-reformat @property declarations Done via script: ``` $ for file in tensorboard/plugins/pr_curve/polymer3/tf_pr_curve_dashboard/*.ts; do sed -i -r -z -e 's/([^\n])\n( *)@property/\1\n\n\2@property/g' -e 's/@property\(\{\n *(type: [^\n]*),\n *\}\)/@property({\1})/g' "$file"; done ``` * link in polymer3 pr_curve dashboard * migrate: manual changes to get PR curves working * migrate: make internal tslint happier

view details

Stephan Lee

commit sha 10b678badefef98f6fca1a8446dff430aa4c6a12

polymer3: add hparams dependencies (#4018)

view details

Alex Bäuerle

commit sha 8fee03124fb2c276ba052a2b7ce21982e9c7d00d

State for nPMI Plugin (#3950) Added state to the nPMI plugin This will enable the development of all the frontend code.

view details

Stephan Lee

commit sha 5e54c6853121a20e6a70dd51974b759152ced493

polymer3: add indirection for polymer/lib/legacy/class (#4017)

view details

push time in a month

pull request commenttensorflow/tensorboard

[DebuggerV1] Remove deprecated plugin

@psybuzz Thanks for the reminder. I was just blocked by a test failure in this PR. I've now pushed a fix. If tests passes, I'll send it for review.

caisq

comment created time in a month

push eventcaisq/tensorboard

Shanqing Cai

commit sha 4d94f1ab366ba2836bfa7b387be9ad5aacd492c0

Add missing :metadata dependencies to audio and image plugins

view details

push time in a month

push eventcaisq/tensorboard

Stephan Lee

commit sha e62b483232c8d982a9e900c74b01e4538bf2723a

chore: remove an ability to customize tsconfig (#3991) Two things: 1. internally, we cannot have a custom tsconfig so allowing custom one will only make sync harder 2. passing tsconfig-testing for test_only targets were quite tedious. Automatng that.

view details

Stephan Lee

commit sha 17909dca14e435ee862f79fcc510db77b46cf8fc

migrate: mesh plugin (#3940) Do note that for ease of migration, we have copied two modules from the tensorflow/graphics repository[1]. Do note that we made some manual changes to better type (to best of our knowledge based on the closure type annotations). Because version of THREE has been upgraded and changed, we may be using deprecated & removed API. We will be able to better tell whether we are doing the right thing only after when we do the functional testing. [1]: https://github.com/tensorflow/graphics/blob/master/tensorflow_graphics/tensorboard/mesh_visualizer/tf_mesh_dashboard/

view details

E

commit sha 736e5f68787f412617acfb8d2852c27bbc6f528f

Remove unnecessary py_binary loader (#4009) Internally, py_binary exists in the ambiently, and will cause errors if loaded manually. This removes the unnecessary load for py_binary. Tested internally with: cl/325902456

view details

William Chargin

commit sha 7b761f36f1a0ce3916b10bda9c17b8dbc10ca0f3

migrate: tf_custom_scalar_dashboard (#4002) Summary: Notable changes: - Fields of the `DataSeries` TypeScript class are no longer `private`, since the view code (previously in HTML and thus not typechecked) poked into them. - Categories now set `type: PREFIX_GROUP`, which is necessary to type-check even though it doesn’t end up making a functional difference due to implementation details of the category view. Test Plan: With `TB_POLYMER3=1 bazel run //tensorboard`: - custom scalars render, both margin charts and multi-line charts; - custom scalar charts are not paginated (by design); - smoothing works; - the axis selector works; - the data download links work; - the outliers toggle works. Googlers, see test sync: <Summary: Notable changes: - Fields of the `DataSeries` TypeScript class are no longer `private`, since the view code (previously in HTML and thus not typechecked) poked into them. - Categories now set `type: PREFIX_GROUP`, which is necessary to type-check even though it doesn’t end up making a functional difference due to implementation details of the category view. Test Plan: With `TB_POLYMER3=1 bazel run //tensorboard`: - custom scalars render, both margin charts and multi-line charts; - custom scalar charts are not paginated (by design); - smoothing works; - the axis selector works; - the data download links work; - the outliers toggle works. Googlers, see test sync: <http://cl/325552918>.

view details

E

commit sha 18c8cfb2599b98acdf55e7067b7aac63710596a2

migrate: prepare graph for conversion part 1 (#3921) Followup: https://github.com/tensorflow/tensorboard/pull/3932 This change adds npm dependencies for the graph dashboard: - dagre - @polymer/paper-radio-button - @polymer/paper-progress - @polymer/iron-list Googlers: This PR should be merged after cl/325528860 After this PR is synced internally, cl/325559670 can be submitted to unblock the final change that enables the Graph dashboard: https://github.com/tensorflow/tensorboard/pull/4005

view details

E

commit sha 16a2e7f189471d360d45538606fac51cfdb203b7

migrate: convert graph externally part 2 (#3932) Diffbase: https://github.com/tensorflow/tensorboard/pull/3921 Followup: https://github.com/tensorflow/tensorboard/pull/4004 Part 2 of the Polymer 3 conversion of the graph dashboard. This adds tags = manual to graph targets, but ensures they build successfully externally, including tf-graph-app's binary.

view details

E

commit sha 151373805447f46fc8ebb87837de0f069fea0609

migrate: graph part 2 (#4004) Diffbase: https://github.com/tensorflow/tensorboard/pull/3932 Followup: https://github.com/tensorflow/tensorboard/pull/4005 Migrating the graph dashboard requires some extra changes to make it pass more strict conformance rules internally. Most of the changes involve making type annotations looser. Notably: Fixed circular dependencies between JS modules. `tf_graph_common` contained lots and lots of cycles, including - scene.ts --> edge.ts --> scene.ts - node.ts --> annotation.ts --> node.ts These were resolved by either moving methods into `common.ts`, or in the case of `annotation.ts`, merging it into `node.ts`. The `tf_graph_common` modules now follow a proper layering scheme (higher layers can depend on lower layers, but never the other way around). layer 0: common, proto layer 1: util layer 2: parser, graph layer 3: op, template layer 4: hierarchy layer 5: render, loader layer 6: tf-graph-scene layer 7: contextmenu, edge, layout layer 8: scene layer 9: annotation/node

view details

E

commit sha 2a1a7234bd96eeaa06d0fb913b1cd5202ef7fae3

migrate: enable the converted graph dashboard part 4 (#4005) Diffbase: https://github.com/tensorflow/tensorboard/pull/4004 This is the last in a series of changes to convert plugins/graph to Polymer 3. This change - removes `tags=["manual"]` - enables the dashboard Tested externally with `TB_POLYMER3=1 bazel run tensorboard` and internally. Googlers, see cl/325351322.

view details

Stanley Bileschi

commit sha 465d0fb27761a56de27d0ac020e51abca22c3e2d

Update Getting Started notebook to use `--one_shot` uploader.

view details

Stanley Bileschi

commit sha 780ea6f0a76f34e4ffeb81611cde20d8d028c55c

spelling & newline

view details

Stephan Lee

commit sha b8d1d4cb3f9fa0279b1c897292ff9b0c90f6eac4

text_v2: actions and effects (#4012) Note that we still are not implementing reducers. Details: - sample data flow: 1. load the text plugin, it fires pluginChanged action. That fetches runTag map 2. based on the run tag map, we draw few tag groups. Few tag groups will be opened by default and some won't. The opened ones fire action for their visiblity changes with <run, tag> tuple of current page 3. that will trigger data fetches 4. on reload, we will reload what is currently on screen

view details

Stephan Lee

commit sha d5e5a4d1b5fb01b30560622b12e8f2394983bb91

polymer3: integrate mesh plugin to the migrated code (#4011) This change enables mesh plugin in the Polymer 3 world.

view details

Stanley Bileschi

commit sha 7609dc568711276a5ac8daf66f6da7e4c6932b8a

Fix missing metadata in tbdev getting started notebook.

view details

Stanley Bileschi

commit sha dba7b9ed8cea082944096a75834561fe9b0f1b77

Clarifies output during `--one_shot` upload. (#4015) Clarifies output during `--one_shot` upload. When `--one_shot` flag is provided, the user expects the uploader to pass through the logdir exactly once and then quit. This change replaces the log line "Listening for new data in logdir" with a more accurate line "Done passing through logdir", but only when `--one_shot` is provided.

view details

Stephan Lee

commit sha 8e90e89cff90cf2a941021806953a463107f8546

Migrate plugins/hparams (#4010) Migrates the hparams plugin. Do note that removed the vaadin-split-layout by simply using flex (824f9f5). Relates to #3955 Co-authored-by: Erik Luoe <luoe@chromium.org>

view details

Stephan Lee

commit sha 76c452c4ec78b51b03c755dbdcbced46da2c418a

Revert "Migrate plugins/hparams (#4010)" (#4016) This reverts commit 8e90e89cff90cf2a941021806953a463107f8546.

view details

Nick Felt

commit sha 8d73b81e249f8c45890b9951615ec27d3d8b1e95

migrate: plugins/pr_curve/polymer3 (#4013) * rename html to ts for better diff * tbmigrate output for better diff * yarn fix-lint * migrate: bulk-reformat @property declarations Done via script: ``` $ for file in tensorboard/plugins/pr_curve/polymer3/tf_pr_curve_dashboard/*.ts; do sed -i -r -z -e 's/([^\n])\n( *)@property/\1\n\n\2@property/g' -e 's/@property\(\{\n *(type: [^\n]*),\n *\}\)/@property({\1})/g' "$file"; done ``` * link in polymer3 pr_curve dashboard * migrate: manual changes to get PR curves working * migrate: make internal tslint happier

view details

Stephan Lee

commit sha 10b678badefef98f6fca1a8446dff430aa4c6a12

polymer3: add hparams dependencies (#4018)

view details

Alex Bäuerle

commit sha 8fee03124fb2c276ba052a2b7ce21982e9c7d00d

State for nPMI Plugin (#3950) Added state to the nPMI plugin This will enable the development of all the frontend code.

view details

Stephan Lee

commit sha 5e54c6853121a20e6a70dd51974b759152ced493

polymer3: add indirection for polymer/lib/legacy/class (#4017)

view details

push time in a month

pull request commenttensorflow/tfjs-models

[Speech Commands] Add notebook for training custom models in Python

Gentle ping @pyu10055. Do you have time to review the PR? Thanks.

caisq

comment created time in 2 months

push eventcaisq/tfjs-models

Shanqing Cai

commit sha 9b7863f6917f56ba4e78b9032e20cc5093886574

Remove unused python lines

view details

Shanqing Cai

commit sha 0f4d99da81177a545d1cfb54c37b274c8c588f26

Merge branch 'sc-train-custom-audio-model-notebook' of github.com:caisq/tfjs-models into sc-train-custom-audio-model-notebook

view details

push time in 2 months

Pull request review commenttensorflow/tensorboard

Clarifies output during `--one_shot` upload.

 def send_tracker(self):             yield         finally:             self._update_cumulative_status()+            if self._one_shot:+                scan_complete_msg = "Done scanning through logdir"

I don't see a change to update_tracker_test.py. Did you forget to commit and push the change?

bileschi

comment created time in 2 months

push eventcaisq/punt37

Dev@TG

commit sha f878d99730f8292c6a5ffe062a241cc32c0c2493

Add .gitignore to project solution

view details

Dev@TG

commit sha 90821a4a079f345db943508579837bff81b15000

Add support for PUT PUNT37

view details

Dev@TG

commit sha f5a1cddbf34920be9ef92030d24e063b8562bad7

Merge remote-tracking branch 'origin/master'

view details

Shanqing Cai

commit sha 9170af1cde97ddff08dd212e8c7c3be28ee28c05

Switche from http to https

view details

push time in 2 months

push eventcaisq/tfjs-models

Ann Yuan

commit sha 0855c08bf4c9463438f82da74bddb5f55c365788

[handpose] Add WASM support. (#501)

view details

Ann Yuan

commit sha 27eea88a3b5a2edcc992c6a9522338b5dc402f61

[facemesh] Support rotated faces, improve docs. (#486)

view details

Shanqing Cai

commit sha e5bc6f57aea0a32e6fab4c73b7fac56abf4349b8

Merge branch 'master' into sc-train-custom-audio-model-notebook

view details

push time in 2 months

push eventcaisq/tfjs-models

Shanqing Cai

commit sha f3419eb404ba81cd3edbc698138d201ac2e6fd0a

Adjust some wording; Remove unused import

view details

push time in 2 months

pull request commenttensorflow/tfjs-models

[Speech Commands] Add notebook to show how to train custom audio model

cc @khanhlvg @lintian06

caisq

comment created time in 2 months

push eventcaisq/tfjs-models

Shanqing Cai

commit sha 617474390bf5a2af1c336feaa88f98f453cb87d2

Improve text and doc strings. Add tfjs and tflite conversion steps

view details

push time in 2 months

PR opened tensorflow/tfjs-models

[Speech Commands] Add notebook to show how to train custom audio model

in Python, using a set of .wav files.

+688 -0

0 comment

1 changed file

pr created time in 2 months

create barnchcaisq/tfjs-models

branch : sc-train-custom-audio-model-notebook

created branch time in 2 months

push eventcaisq/tfjs-models

Shanqing Cai

commit sha e8f38994bd742678d93b0583a3fa745395b13f9c

[SpeechCommands] Add notebook to show how to convert model to tflite (#464) * Fixes https://github.com/googlecreativelab/teachablemachine-community/issues/36 The notebook shows * how to convert a speech-commands model from the TF.js format to the Python (tf.keras) and TFLite formats * how to run the Python (tf.keras) model for inference.

view details

Rajeshwar Reddy T

commit sha 613e163bc9b21fc9691af578c084aee615fc95fa

Update bodypix README.md (#469)

view details

Johan Euphrosine

commit sha 9c54814fac792c5d91bf6ac3c04bca979e417b28

speech-commands/training: fix model links (#489)

view details

Justin Greene

commit sha f4a22ea394e457ccac98d0a26f9c68ac11609d09

iOS safari shader compilation issue (#471) Co-authored-by: Ping Yu <4018+pyu10055@users.noreply.github.com>

view details

Ping Yu

commit sha b02310745ceac6b8e4a475719c343da53e3cade2

USE QnA API (#490) * add use-qna code * add test for use qna model * added qna demo and update the model input for qna * updated the model source to tfhub * updated README and versions in package.json * remove yalc link * fix tests * address comments * fix lint * address comments * updated README

view details

Ping Yu

commit sha d57f29cdda506b44779b16598d4974aede940f64

fix npm loading for use (#499)

view details

Dale Markowitz

commit sha afeef9ca89b6e91eeb962e983cf943b9fad5b355

Update README.md (#502) Require tfjs in USE Node code sample so that users can run it as-is.

view details

Ping Yu

commit sha be69bace1e34fe9f77d17c9d6e606a86e3c49d10

use the gcp for image storage (#503)

view details

Ping Yu

commit sha 4fb931ab56b1171d438527254b72804eb3fd4448

update the version for the README change (#504)

view details

Yannick Assogba

commit sha 12f4c492c60066ecc7461e641a2ce32f5fe0123b

revert default tsconfig settings to output commonjs+es5 (#500) Also update cocoSsd build config and package.json to support esm bundles.

view details

push time in 2 months

PR opened tensorflow/tensorboard

[DebuggerV1] Remove deprecated plugin
  • Motivation for features / changes
    • This PR removes the Python backend and web frontend code of the Debugger (V1) plugin. This plugin has been marked as deprecated and for removal in the 2.3.0 release notes.
  • Technical description of changes
    • Removes tensorboard/plugins/debugger, along with references to it from BUILD files, Python modules and Polymer .html files
  • Detailed steps to verify changes work correctly (as executed by you)
    • Verified that Debugger V2 still functions properly
    • Internal CL for corresponding code removal is tested globally
+0 -20680

0 comment

80 changed files

pr created time in 2 months

create barnchcaisq/tensorboard

branch : debugger-v1-removal

created branch time in 2 months

push eventcaisq/tensorboard

Stephan Lee

commit sha cec9ff2520cdb428a666d47f5b947db88f49f134

sync: use tb re-exported legacy element mixin (#3939)

view details

Stephan Lee

commit sha b965e759047cd011973de9191163444a3d538d51

migrate: projector (#3933) The plugin is fully function: `bazel run tensorboard/plugins/polymer3/vz_projector:standalone` works end to end.

view details

Preston Fossee

commit sha 5fccd86684993014377b592ad9cd62ffe6d0679d

Text demo v2 (#3864) Update the demo file to better test the text plugin's frontend UI. This will help in creating/testing the new Angular frontend UI for the text plugin. Also ports the demo to TF V2 to make it easier to use (no longer need to switch TensorFlow versions to use it) Adds multiple runs and a function to test the pagination view in the frontend UI.

view details

Stephan Lee

commit sha a5b6ce6aca8fdca30ec0fd2f70c262b6e77b10d7

bug: remove extra unused import in tf-multi-checkbox (#3945)

view details

Stephan Lee

commit sha fa8556f4b215825ec76e1ef7ee0813d0cd9e08ba

fix: run filter in the runs selector (#3943) Our run filter was broken because the callback was keep returning a function instead of doing anything useful.

view details

Stephan Lee

commit sha 26ba8062fbe0dca671a602f5b7452ee3fe3eae9f

bug: fixed wrong ref to urlDict (#3946) We referred to the module level variable as `this.urlDict`. Co-authored-by: William Chargin <wchargin@gmail.com>

view details

Stephan Lee

commit sha 319c92be3cb0948da05049c27cc26d62e6b2f01c

fix: import iron-flex by depending on appropriate module (#3942) When using `iron-flex` style dom-module, we need to have hard dependency on the `iron-flex-layout-classes` for the dom-module definition to be pulled in.

view details

Stephan Lee

commit sha 045800356caa4ac361f908ab26b96c91835546af

fix register_style_dom_module for styleIncludes (#3941) Two issues: - When adding children to template, we need to set it to `template.content`. Added the style include in the appropriate part of the DOM - added explicit dependency on @polymer/polymer

view details

William Chargin

commit sha 80526dc6703ac0e286d155655996cc3d6e333a0d

migrate: tf_audio_dashboard (#3936)

view details

Stephan Lee

commit sha 85738083f83deeee9501fd5c345fb8bd94ad44f7

fix: call *Initializer with reference to (#3948) Our Polymer based storage utility required Polymer class instance for the initializer (so we can set the component prop value when hash changes). This broke in the class syntax so we reinstated it by explicitly passing `this`. Co-authored-by: William Chargin <wchargin@gmail.com>

view details

Stephan Lee

commit sha 40fe4109970b95058476822b006d1ae9a4944093

bug: fixed the text dashboard (#3944) 1. ready without super call 2. wrong imports

view details

Stephan Lee

commit sha c3829199dd0674f244d498b262a7ffd0957e3bcd

add eplicit global HTML style and add roboto dependency (#3947) Do note that after the library upgrade, we need to explicitly tell Polymer not to fetch Roboto from CDN. Manually ran the application to check for no regression. Note that the roboto genrule file change does not impact Polymer 2.

view details

Nick Felt

commit sha d412541810115542d3ac3c59e20a9ab74dc637e2

migrate: fix data-loader-behavior.ts observer spec (#3951)

view details

Stephan Lee

commit sha e63114831b42f8a6f0e31c5f3dbf8bcb504263f3

migrate: misc fixes (#3949) - we add redundant import to cause side-effect of importing Polymer module. Previously, we were importing specific symbols for type checking and the Polymer modules were tree shaken - Removed readOnly. Even in Polymer 3, readOnly creates an implicit method (if prop name was foo, Polymer creates _setFoo) which is impossible to type. We just removed it for now. - Removed the default value on _contentActive so the consumer can override it with a getter. Co-authored-by: William Chargin <wchargin@gmail.com>

view details

Nick Felt

commit sha 6ca6f0bd3533b4c9ca412207575b11438241898f

migrate: add plottable-style DOM module for vz-line-chart2 (#3952)

view details

Stephan Lee

commit sha 2054318cfd1d3ed6235b9ad2cef638142ecdbea6

Allow tensorboard_html_binary to serve JS file in the devserver (#3937) With this change, when you `bazel run` a target, loading the HTML file in the browser will work since extracted JavaScript would be served by the server. Previously above flow failed since the server never served the JavaScript file.

view details

Stephan Lee

commit sha 8ad70a830f9499edc3cf94a464222e1ebf81a882

sync: fixed projector for sync (#3953) Few issues: 1. depend on `numeric`. `numericjs` is not the popular package. 2. remove export if using `LegacyElementMixin`. Fixed component class imports accordingly 3. depend on internal legacy element mixin 4. remove innerHTML and change it to textContent since we are all setting textContent 5. replaced custom word break with tf-wbr-string 6. side-stepped issue with tsetse regarding hyperlinks

view details

Alex Bäuerle

commit sha 0cbc74ea0b07993ca32123f8ff71a7b9e369d4b2

Main and Inactive View (#3931) * Added nPMI Plugin Backend Files Added files for the nPMI plugin backend. This is not yet completely working. * Added Comment to Active Function Commented the isActive function of the plugin. * Fix Plugin Test File The test file did not work correctly as of some indentation issue. THis has been fixed now. * Summary Tests Added tests for the summary writer. * Corrected Indentation for Main File Corrected the indentation for the main plugin backend file. * Added Missing License Added a missing license disclaimer to the summary test file. * More Indentation Correction Corrected more of the indentation issues in the python files. * Further Linter Issues Fixed further linter issues. * Linting Correction Further Linting Correction * Further Linting More linting changes. * Fixed BUILD File Fixed linter error for BUILD file. * Fixed Test Issue on TF 1.15 Fixed failing tests because of wrong imports on TF 1.15. * Rename NPMI to Npmi in Classes For consistency, this renaming has been done. * Missed Rename Missed renaming in one place. * More Explanation for the Plugin Added more explanation for the plugin with a readme and one more comment. * Removed Keys and added Comments Removed some keys that are not really needed for the plugin and added more comments to clarify routes. * Build Deps Cleaned Cleaned up a little in the build deps. * Removed Unused Imports Had several imports that were not used. Removed them now. * Added Dependency Added a missing dependency for the plugin. * Changed dependencies to actual Import Changed dependecies of the summary writer to the actual imports of this file. * Restructured Summary Writer Summary writer now with three different functions for the routes. * Fixed Linter Issues After this restructure, I introduced some linter issues, which are now fixed. * Added Safe JSON Encoder Added a encoder that safely handles numpy nan, inf, and -inf values. * Linting Fixes After adding the SafeEncoder, needed some linter fixes. * Final Linter Change One more Linter change for the safe encoder. * Added Proto for Metadata Added aproto for the metadata field. * Metadata is Own File Now, the metadata is its own file. This had to be done for the metadata content is a proto now and thus metadata is accessed at different places. * Begin data_provider Transition Beginning the transition from multiplexer to data_provider. * Extra Whitespace Bug Removed an extra whitespace where there should not be one. * Plugin Name from Metadata Now getting the plugin name from metadata to avoid duplication. * Fix BUILD File Some fixes in the BUILD file for the CL. * Added Version Field Added a version field to the metadata. * Switch from Multiplexer to Data Provider Switched from multiplexter to using data provider for serving data. * Linting Corrected linting issues. * More Linting More linting corrections. * Removed Safe Encoder for Parse Function Now parsing the values that we get from a tensor and converting nan to None. * Fixed Linting in Convert Function The convert function now is correctly formatted as well. * Minor Comment Changes Changed some comments for better explanation. * Removed Table Field Removed the table field from all the data. * Linting Fix Fixed a minor linter error. * Moved To Correct Directory The backend has been moved to the tensorboard/plugins directory. * Build Lint Fix * Fixed Wrong Import The default.py was still using the old path. * Added The bare-bones of the nPMI Frontend Added just the main comopnent without any data fetching whatsoever. * Remove Duplicate Plugin Class Removed a duplicate declaration of the npmi plugin class in plugin.py. * NPMI -> Npmi To match the backend way of writing this. * Started Test Implementation Started implementing tests for the frontend. * Added Test to Webapp * Added Tests for NPMI Added tests for the npmi component. * BUILD Lint * BUILD Lint Fix * Prettier Formatting * Fixed Issues and Adressed Stephans Comments Fixed an issue with the a failing test and adressed Stephans PR comments. * Require Runs Runs now required to be passed. * Corrected Usage of AppState Corrected usage of State to match the model for tb plugins. * Missing Dependency Added a missing dependency which is required for the run selector. * BUILD Lint * Add Inactive and Main View Added a view for when the plugin is inactive, and one as a main starting view of the plugin. * License Header * HTML and SCSS Formatting * Added Tests for new Modules * Removed Unused CSS * Addressed more PR comments * Fixed Selector in Test Fixed an incorrect selector name in a test. * Renamed Selector for Inactive Component

view details

Stephan Lee

commit sha ebf353e80c96f6a06896128d6cd61c170b8e533a

bug: fix broken contentActive in tf-dom-repeat (#3957) tf-dom-repeat and tf-category-paginated-view were weird because the sub class was implementing `_contentActive` protected Polymer property via getter (because it was a computed property). It looks like it is impossible for subclass to override a property using a getter. For instance, in pure JavaScript, ```js (function() { class Foo { foo = 'foo'; } class Bar extends Foo { foo = 'bar'; } class Baz extends Foo { get foo() { return 'baz'; } } console.log(new Bar().foo); // 'bar' console.log(new Baz().foo); // 'foo' })(); ``` This change no longer implements `_contentActive` setting via getter.

view details

Stephan Lee

commit sha 05213460690fe800780edee72cc6bfafb8a926aa

polymer3: use hourglass import for paper & irons (#3958) Polymer's iron and paper components are hard to depend in a way that is compatible both in OSS and google3. With this change, by creating an hourglass dependency, we can resolve the discrepancy easily by making the changes in only the new file.

view details

push time in 2 months

MemberEvent

Pull request review commenttensorflow/tensorboard

Clarifies output during `--one_shot` upload.

 def send_tracker(self):             yield         finally:             self._update_cumulative_status()+            if self._one_shot:+                scan_complete_msg = "Done scanning through logdir"+            else:

L345 - 350 can be simplified as

self._update_uploading_status(
    "Done scanning logdir" if self._one_shot else "Listening for new data in logdir",
    color_code=_STYLE_YELLOW)

Notice that I suggested simplifying "Done scanning through..." to "Done scanning..." too, for conciseness.

bileschi

comment created time in 2 months

Pull request review commenttensorflow/tensorboard

Clarifies output during `--one_shot` upload.

 def send_tracker(self):             yield         finally:             self._update_cumulative_status()+            if self._one_shot:+                scan_complete_msg = "Done scanning through logdir"

Can you add a new test in this file in the vicinity of this line?

https://github.com/tensorflow/tensorboard/blob/master/tensorboard/uploader/upload_tracker_test.py#L253

bileschi

comment created time in 2 months

create barnchtensorflow/tfjs-models

branch : sc-notebook-tf2.3.0

created branch time in 2 months

push eventtensorflow/tfjs

Shanqing Cai

commit sha 9b8c96363e0fbd7c19961828f56e0e7d32c82f6b

[tfjs-node] Fix bug in node.tensorBoard() callback re initialEpoch (#3714) Fixes https://github.com/tensorflow/tfjs/issues/3705 Previously, the `tensorBoard` callback in tfjs-node does not honor the `initialEpoch` arg passed to the `fit()` call that uses the callback. It always incorrectly starts from 0. This CL fixes this bug by using the `epoch` arg passed to `onEpochEnd()` instead of an `epochsSeen` counter maintained by the callback object itself.

view details

push time in 2 months

PR merged tensorflow/tfjs

[tfjs-node] Fix bug in node.tensorBoard() callback re initialEpoch cla: yes

Fixes https://github.com/tensorflow/tfjs/issues/3705

Previously, the tensorBoard callback in tfjs-node does not honor the initialEpoch arg passed to the fit() call that uses the callback. It always incorrectly starts from 0. This CL fixes this bug by using the epoch arg passed to onEpochEnd() instead of an epochsSeen counter maintained by the callback object itself.

<!-- Reviewable:start -->

This change is <img src="https://reviewable.io/review_button.svg" height="34" align="absmiddle" alt="Reviewable"/> <!-- Reviewable:end -->

+57 -4

3 comments

2 changed files

caisq

pr closed time in 2 months

issue closedtensorflow/tfjs

Tensorboard logging does not honor `initialEpoch`

TensorFlow.js version

2.0.1

Describe the problem or feature request

When training a model, using callback: tf.node.tensorBoard() and initialEpoch: x, the series graph from step=0, not step=initialEpoch

Code to reproduce the bug / link to feature request

const history = await model.fitDataset(trainData, {
	validationData,
	epochs:100,
	initialEpoch: 23,
	callbacks: tf.node.tensorBoard(tbPath),
})

Probably, this is the source of the issue: https://github.com/tensorflow/tfjs/commit/303345fb161e5532e5b6d6e49cd035c4a3d3b11c#r41091644

closed time in 2 months

spazmodius

push eventcaisq/tfjs-1

Linchenn

commit sha 0f629f8f0be7829e58ecf855188283965dc62887

Decouple benchmark parameters from the unit test (#3701) FEATURE

view details

Shanqing Cai

commit sha cb49707dd8f095962521ea1d33d995f284ab9f86

Merge branch 'master' into node-tensorboard-callback-initial-epoch-fix

view details

push time in 2 months

push eventcaisq/tfjs-1

Shanqing Cai

commit sha edd03c154c129b04e1ce8a5b89e19bde35a1d259

Disable lint for any

view details

Shanqing Cai

commit sha 501d65e0a22f71fc7fed7420207d7a394a9e8b28

Merge branch 'node-tensorboard-callback-initial-epoch-fix' of github.com:caisq/tfjs-1 into node-tensorboard-callback-initial-epoch-fix

view details

push time in 2 months

pull request commenttensorflow/tfjs

[tfjs-node] Fix bug in node.tensorBoard() callback re initialEpoch

@lina128 @pyu10055 Ah - good find. Sorry I didn't notice that.

caisq

comment created time in 2 months

pull request commenttensorflow/tfjs

[tfjs-node] Fix bug in node.tensorBoard() callback re initialEpoch

@pyu10055 @lina128 Any idea why the CI of tfjs-node is failing? It seems to be unrelated to this PR.

image

caisq

comment created time in 2 months

push eventcaisq/tfjs-1

Linchenn

commit sha b9a8b0eae765abc50c8a7080322982fec3f87098

Add UI for tuning parameters for the BrowserStack-benchmark tool (#3704) FEATURE

view details

Na Li

commit sha 7769bdc34a83b3fef2e74e04e6d300537b2776b4

Use custom docker for tfjs-wasm. (#3720)

view details

Linchenn

commit sha 9a1c5670922c7acb09ca05047867e6bca994d1f4

Decouple logging from profiler.profileKernel (#3606) FEATURE

view details

Linchenn

commit sha 5ca0e8d410ddb9cc1948337e615a2706ad283248

Add flag feature for the benchmark webpage tool (#3625) FEATURE

view details

Shanqing Cai

commit sha 703797b5dfc1c0fa5d04b9a4109c49a2ac98598c

Merge branch 'master' into node-tensorboard-callback-initial-epoch-fix

view details

push time in 2 months

issue openedtensorflow/tfjs

Feature Request: Add Colab Notebook to show how to train an audio model from scratch in Python

The audio model should use the WebAudio preprocessing layer like https://github.com/tensorflow/tfjs-models/blob/master/speech-commands/training/browser-fft/tflite_conversion.ipynb to support easy conversion to TF.js and TFLite.

Public datasets we can use for this:

  • https://www.tensorflow.org/datasets/catalog/speech_commands

created time in 2 months

push eventcaisq/tfjs-1

Na Li

commit sha cc8399a62dfd00e922ef75c330788b8a1a29ab4f

Use custom docker for tfjs-wasm. (#3710)

view details

Shanqing Cai

commit sha f282f1249b2bd28a80b5dca03ae6b2c7382d4d8e

Merge branch 'master' into node-tensorboard-callback-initial-epoch-fix

view details

push time in 2 months

PR opened tensorflow/tfjs

[tfjs-node] Fix bug in node.tensorBoard() callback re initialEpoch

Fixes https://github.com/tensorflow/tfjs/issues/3705

Previously, the tensorBoard callback in tfjs-node does not honor the initialEpoch arg passed to the fit() call that uses the callback. It always incorrectly starts from 0. This CL fixes this bug by using the epoch arg passed to onEpochEnd() instead of an epochsSeen counter maintained by the callback object itself.

+53 -4

0 comment

2 changed files

pr created time in 2 months

create barnchcaisq/tfjs-1

branch : node-tensorboard-callback-initial-epoch-fix

created branch time in 2 months

issue commenttensorflow/tfjs

Uncaught (in promise) Error: Argument tensors passed to stack must be a Tensor[] or TensorLike[]

More info: convertToTensorArray(), which is a part of the gradient tape mechanism, expects its argument to be an array of Tensor or TensorLike. But for some reason, when input shape of the LSTM is [1, 1], i.e., only one temporal step, convertToTensorArray() receives a single Tensor as argument. If the argument is an Array<Tensor> of length 1, then everything works.

chittalpatel

comment created time in 2 months

issue commenttensorflow/tfjs

Uncaught (in promise) Error: Argument tensors passed to stack must be a Tensor[] or TensorLike[]

@tafsiri Thanks for letting me know. I can reproduce it too. An additional interesting observation is that the model can always run predict() correctly even though its fit() method throws the error. This seems to indicate that the problem is somewhere in the gradient tape code related to the convertToTensorArray method in tfjs-core. This is consistent with the stack trace of the error.

In addition, I observed that if the input shape is [2, 1], instead of [1, 1], then the model trains without error. This again is a hint that the bug is an edge case in the gradient tape code.

cc @dsmilkov

chittalpatel

comment created time in 2 months

push eventcaisq/tfjs-1

Yannick Assogba

commit sha 4c9f4f632312f669d5ad1166d0780f9d0875bd99

[tfjs-core] Modularize squaredDiff INTERNAL

view details

Paul Van Eck

commit sha f0580750bc8d4e106b1801aa65b3ba8c9cb2f499

Add some fixes to converter wizard (#2799) BUG * Fix dryrun arg not working * Enable tfjs-layers to tfjs-layers conversion path * Change conditions for split_weights_by_layer argument Co-authored-by: Ping Yu <4018+pyu10055@users.noreply.github.com>

view details

Paul Van Eck

commit sha af46b87f2c110fcce028d609b6328857171df0cb

[tfjs-converter] Allow custom shard size for more conversion pairs (#2774) FEATURE * Allow custom shard size for more conversion pairs * Adjust arg size arg check * Add tests for sharded argument * Update readme with new arg * Add shard size support to tensorflowjs_wizard Co-authored-by: Ping Yu <4018+pyu10055@users.noreply.github.com>

view details

Na Li

commit sha 0162a295d16795b47e47136c21f91406ec09ba6c

Add affected packages analysis. (#2793)

view details

Etienne Dupont

commit sha 1a942195b713886998123568b9a35f34aa2077de

Update package.json (#2762) * Update package.json Solves #2761 * Line up tfjs-node-gpu's package.json

view details

Yannick Assogba

commit sha 5d5d091082f339e237bbdc9e7836d46a36b335ba

[tfjs-react-native] add expo-camera to deps list DOC Fixes https://github.com/tensorflow/tfjs/issues/2795

view details

Ping Yu

commit sha 8bd65f5229f8c7bf72565649325c06fb9eaedfc6

fix g3 sync errors (#2810) INTERNAL

view details

Ping Yu

commit sha 5b2b48722708929b7220160ddf20e1ccfbdedb88

more g3 sync fixes (#2811) INTERNAL * more g3 sync fixes * more fixes

view details

Ping Yu

commit sha a29d40cb4e939717a2629487071c50a26f41a8eb

Update tfjs-core to 1.5.3. (#2814) INTERNAL

view details

Nikhil Thorat

commit sha 5cf6e9ccaf877dcdbaf9ee54d1622d109a26eac9

Change test-integration to look for 'run-ci' instead of 'diff'. (#2823) BUG

view details

Na Li

commit sha e19a917ae6de3647633e9fb1be66bbcfb46d3ff4

Update tfjs-core to 1.6.0. (#2824)

view details

Ping Yu

commit sha 5eb777a03b07f78f43fb1f46057864fe35642558

be more specific on keras packages (#2813) INTERNAL

view details

Ann Yuan

commit sha 28e56fcb6e6bb6954a1a32101fda857666fbb10c

[wasm] Export version. (#2819) FEATURE

view details

Nikhil Thorat

commit sha d658f3e13fc00ff44d9b10753cf66fcb00b55613

Update tfjs-layers, tfjs-converter to 1.6.0, 1.6.0. (#2825) INTERNAL

view details

Ping Yu

commit sha 9594ad9cb38671c085ce30633c5061025a67d656

update the README file to introduce the converter wizard CLI (#2791) DOC * update the README file to introduce the converter wizard CLI * addressed comments * use single quote for inline code

view details

Na Li

commit sha ebde93180c2a11eb93114478f0cbc6c7dd38d7ea

Update tfjs-data to 1.6.0. (#2826)

view details

Ping Yu

commit sha adfb9bf30de8a96b578c72d2649e0ee68bbc310f

Update tfjs to 1.6.0. (#2827) INTERNAL

view details

Ping Yu

commit sha f8b87860ef47f786b975ca144868972cd25c5986

Update tfjs-node, tfjs-node-gpu to 1.6.0, 1.6.0. (#2829) INTERNAL * Update tfjs-node, tfjs-node-gpu to 1.6.0, 1.6.0. * udpate tfjs-core deps * update gpu core version

view details

Na Li

commit sha e71c512ee3e54ec2b2a8e0356bd5d7a0b3fcf69a

[layers]Test against head. (#2831)

view details

Na Li

commit sha d989856c78211ec0d0ef5472254334797012afcb

[data]Test against head. (#2832)

view details

push time in 2 months

push eventcaisq/tensorboard

Stephan Lee

commit sha 97d2683eaac83a9b0b6735cc8425c9226fba7c52

cleanup: remove unused iron_ajax (#3848) Unused dependency.

view details

William Chargin

commit sha 6ed7c6dcf69124b2d0aa13b8e5e7567cbf3231a7

lib: make `errors` submodule lazy (#3851) Summary: All the other submodules are lazy. We originally made `errors` eager because it seemed like a small surface (#2631), but some Google-internal Python 2 binaries unnecessarily transitively depend on `errors` because of this edge. Moving it to a lazy dependency makes it easier to make `errors` use py3-only features. Test Plan: Existing smoke test in `test_pip_package.sh` suffices. wchargin-branch: lib-lazy-errors

view details

William Chargin

commit sha 62f586650b20a4eb339005c6af63e1bb463249a7

errors: add `UnauthenticatedError` (#3852) Summary: We add a new type to distinguish authentication errors (“we can’t confirm who you are”) from authorization errors (“we know who you are, but you’re not allowed to perform this operation”). This was omitted from the initial draft of the `errors` module because the natural status code is HTTP 401, but HTTP 401 responses are required by RFC 7235, §3.1 to include a `WWW-Authenticate` challenge header. In this patch, we add the necessary error handling machinery to support that header. Test Plan: Unit tests suffice for the core implementation. As a manual test, changing a `NotFoundError` to an `UnauthorizedError` with an arbitrary challenge in `scalars_plugin.py` properly sends a 401 with the intended response and challenge header. Since this patch changes `:errors` to include Python 3 syntax, Google-internal <http://cl/321666883> witnesses that there are no more internal Python 2 dependents after #3851. wchargin-branch: errors-unauthenticated

view details

William Chargin

commit sha cb87cd0ed4019335e51ffe0f7db67ff8d6ca08b6

ci: unpin 2020-07-12 nightlies (#3849) Summary: Fixed upstream: <https://github.com/tensorflow/tensorflow/issues/41388> This reverts commit d24e136ede5bab292afed0e62b813a2740c0a648. Test Plan: CI suffices. wchargin-branch: unpin-20200712-nightly

view details

Stephan Lee

commit sha 116e5eb9a22953830e2b5a96d4ebea7286a22f34

Added the Angular wrapper for tf-runs-selector (#3825) For the ease of our development in Angular, especially for porting a fixing dashboard without any functional differences, we decided to just wrap a Polymer based runs selector. This change introduced that. However, do note that the selection state is in the ngrx store so you must use the regular Angular pattern to render and access the data related to runs.

view details

Shanqing Cai

commit sha 72a104edb0f8d83ef8889ebee7dd39be684461c1

[DebugerV2] Fix unexpected markdown section header caused by hash (#3853) * Motivation for features / changes * Fix a bad formatting in the tutorial page at https://www.tensorflow.org/tensorboard/debugger_v2 caused by a `#` at the beginning of a source line of the markdown file. See symptom in screenshot: ![image](https://user-images.githubusercontent.com/16824702/87725255-f222c280-c78a-11ea-8064-9057af91b24f.png) * Technical description of changes * Remove the hash `#`, which is not necessary for the clarity of the text anyway.

view details

Alex Bäuerle

commit sha 1e7a722c49829351235308d8fdc31e28d4a04383

nPMI Plugin Backend (experimental) (#3799) * Added nPMI Plugin Backend Files Added files for the nPMI plugin backend. This is not yet completely working. * Added Comment to Active Function Commented the isActive function of the plugin. * Fix Plugin Test File The test file did not work correctly as of some indentation issue. THis has been fixed now. * Summary Tests Added tests for the summary writer. * Corrected Indentation for Main File Corrected the indentation for the main plugin backend file. * Added Missing License Added a missing license disclaimer to the summary test file. * More Indentation Correction Corrected more of the indentation issues in the python files. * Further Linter Issues Fixed further linter issues. * Linting Correction Further Linting Correction * Further Linting More linting changes. * Fixed BUILD File Fixed linter error for BUILD file. * Fixed Test Issue on TF 1.15 Fixed failing tests because of wrong imports on TF 1.15. * Rename NPMI to Npmi in Classes For consistency, this renaming has been done. * Missed Rename Missed renaming in one place. * More Explanation for the Plugin Added more explanation for the plugin with a readme and one more comment. * Removed Keys and added Comments Removed some keys that are not really needed for the plugin and added more comments to clarify routes. * Build Deps Cleaned Cleaned up a little in the build deps. * Removed Unused Imports Had several imports that were not used. Removed them now. * Added Dependency Added a missing dependency for the plugin. * Changed dependencies to actual Import Changed dependecies of the summary writer to the actual imports of this file. * Restructured Summary Writer Summary writer now with three different functions for the routes. * Fixed Linter Issues After this restructure, I introduced some linter issues, which are now fixed. * Added Safe JSON Encoder Added a encoder that safely handles numpy nan, inf, and -inf values. * Linting Fixes After adding the SafeEncoder, needed some linter fixes. * Final Linter Change One more Linter change for the safe encoder. * Added Proto for Metadata Added aproto for the metadata field. * Metadata is Own File Now, the metadata is its own file. This had to be done for the metadata content is a proto now and thus metadata is accessed at different places. * Begin data_provider Transition Beginning the transition from multiplexer to data_provider. * Extra Whitespace Bug Removed an extra whitespace where there should not be one. * Plugin Name from Metadata Now getting the plugin name from metadata to avoid duplication. * Fix BUILD File Some fixes in the BUILD file for the CL. * Added Version Field Added a version field to the metadata. * Switch from Multiplexer to Data Provider Switched from multiplexter to using data provider for serving data. * Linting Corrected linting issues. * More Linting More linting corrections. * Removed Safe Encoder for Parse Function Now parsing the values that we get from a tensor and converting nan to None. * Fixed Linting in Convert Function The convert function now is correctly formatted as well. * Minor Comment Changes Changed some comments for better explanation. * Removed Table Field Removed the table field from all the data. * Linting Fix Fixed a minor linter error. * Moved To Correct Directory The backend has been moved to the tensorboard/plugins directory. * Build Lint Fix * Fixed Wrong Import The default.py was still using the old path. Co-authored-by: Alex Bäuerle <abauerle@google.com>

view details

William Chargin

commit sha 1b9e1f58cd1e90a2999857e3732f3fe343293d67

application: add injection point for middleware (#3855) Summary: Clients of `TensorBoardWSGIApp` may now inject middleware around the core TensorBoard app routing but still within the experiment ID, path prefix, request context, etc. middlewares. This is useful for handlers for extra routes that need to access the request context. Test Plan: Unit tests included. wchargin-branch: application-middleware-injection

view details

Brian Dubois

commit sha 0c88c1f8c98e68898bd5eabfed47e90ae1267a44

Add utm_source to outbound TensorBoard.dev links (#3822) The "Upload to TensorBoard.dev" dialog provides a couple anchor links to http://TensorBoard.dev. TensorBoard.dev would still like to know the source of the traffic. Append a utm_source query parameter to the URL, which will be used by TensorBoard.dev's Google Analytics library to identify the traffic as coming from "tensorboard". For the moment we intentionally exclude other GA utm parameters since we're not interested in other dimensions and they pollute the URL further.

view details

William Chargin

commit sha 03b7edf797bca656308322e7d628edf6a2cc966c

data: add dispatching data provider (#3847) Summary: A new `DispatchingDataProvider` composes multiple sub-providers under different experiment ID prefixes. Thus, if `FooDataProvider` expects experiment IDs like `123` and `BarDataProvider` expects IDs like `abc`, you could use a `DispatchingDataProvider` to simultaneously expose experiment IDs like `foo:123` and `bar:abc`. Most method implementations of `DispatchingDataProvider` simply involve determining the appropriate sub-provider and forwarding all arguments. We implement those with a tiny bit of metaprogramming (since there are nine such methods), stopping short of [more drastic measures][ddd]. [ddd]: https://wchargin.github.io/posts/dubious-dynamic-delegation-in-python/ Test Plan: Unit tests included. For Google-internal use, <http://cl/321472166> shows an application of this data provider. wchargin-branch: data-dispatching-provider

view details

William Chargin

commit sha 869e2b4931c560505847e8451202a69f1953d52f

data: simplify dispatching provider tests (#3858) Summary: Per comment of @nfelt on #3847, we can still test for delegate equality: <https://github.com/tensorflow/tensorboard/pull/3847#discussion_r457698571> wchargin-branch: data-dispatch-test-bare

view details

E

commit sha 7182735ef564b41bb3b63d85367a1999c8a819fd

fix data_ingester_test on Python 3.8 (#3861) Python 3.8 includes breaking changes to the `os.path.expanduser` method, which no longer reads from the `HOME` environment variable on Windows, but rather uses `USERPROFILE`. This broke the `data_ingester_test.py`, which tries to emulate platform specific expansion by setting `os.environ["HOME"]`. To allow this test to run on both Python 3.8 and before, this change makes the test simulate home directory expansion by setting both environment variables. Manually tested that the test fails before and passes after this change, with `bazel run //tensorboard/backend/event_processing:data_ingester_test`. Note that our Travis config does not use Python 3.8, so our CI does not validate whether this change works. See https://bugs.python.org/issue36264

view details

E

commit sha a64659b4a17001c6ac860153139be14d760ddb4a

Update protos for TF v2.3.0-rc2 (#3862) This updates TensorBoard protos to match those of TensorFlow v2.3.0-rc2 to make the compat/proto_test pass. This change was generated by expanding `update.sh`'s replacement pattern, and running `git checkout v2.3.0-rc2` in a TensorFlow local repo, and running `./tensorboard/compat/proto/update.sh PATH_TO_TENSORFLOW_REPO`.

view details

E

commit sha ed14106def5eb06b66f96543580aee03db4e50ac

Bump version at HEAD to 2.4.0a0

view details

E

commit sha d9dca6df3b05fa9eb6ac553daeae86ecd6341d16

Add 2.3.0 relnotes to README.md

view details

dependabot[bot]

commit sha 7ec24d552ba10af2210ecf3c64edd9b94e124987

build(deps): bump lodash from 4.17.15 to 4.17.19 (#3854) Bumps [lodash](https://github.com/lodash/lodash) from 4.17.15 to 4.17.19. - [Release notes](https://github.com/lodash/lodash/releases) - [Commits](https://github.com/lodash/lodash/compare/4.17.15...4.17.19) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

view details

Stephan Lee

commit sha 69c694619d44d350e9937fda65846b0ab14c7cd4

run selector interop: populate initial state (#3863) Bug: currently, the initial state of the legacy runs selector is not reflected in the store. Fix: read the state from the Polymer component and fire an action.

view details

Mani Varadarajan

commit sha 2a14101a04f339cadd536bd4c9b239664c51adc7

Scalars plugin: Return 400 if run or tag is not specified (#3874) * Motivation for features / changes If the "/data/plugin/scalars/scalars" route is hit without specifying the run or tag query parameters, TensorBoard will return 500. Corresponding errors were noticed in tensorboard.dev server logs. Internal Google bug reference: b/161680845 * Technical description of changes Scalars plugin now checks that tag and run are not None before proceeding with the request. If either is None, it raises InvalidArgumentError. * Screenshots of UI changes None relevant. * Detailed steps to verify changes work correctly (as executed by you) 1) Run local TensorBoard with changes 2) Visit "http://localhost:6006/data/plugin/scalars/scalars?run=foo", "http://localhost:6006/data/plugin/scalars/scalars?tag=foo" and "http://localhost:6006/data/plugin/scalars/scalars" 3) Ensure that "Invalid argument: Both run and tag must be specified: tag=X, run=Y" are returned as appropriate * Alternate designs / implementations considered None

view details

Mani Varadarajan

commit sha 5a5d123f47f8d03a1b33b7c70c5536728b5b054e

Scalars plugin: Clean up unused Python imports (#3875) Clean up unused imports and update BUILD file.

view details

Stephan Lee

commit sha 7e0611bf4ba309a3284cd0d1f5c848fd600c1ede

Change our Plottable monkey patch to TypeScript (#3876) This change is made so that we can more easily migrate to tf_ts_library. Background: we have plans to migrate our build system from tf_web_library to tf_ts_library. In doing so, we will, without this change, make ts_library depend on js_library (this file) but that is impossible since there is no such thing as js_library. Since the module largely monkey patches and accesses private methods we treated many violations with casting to any for now.

view details

push time in 2 months

push eventtensorflow/tfjs-models

Johan Euphrosine

commit sha 9c54814fac792c5d91bf6ac3c04bca979e417b28

speech-commands/training: fix model links (#489)

view details

push time in 2 months

PR merged tensorflow/tfjs-models

speech-commands/training: fix model links cla: yes

/cc @khanhlvg

<!-- Reviewable:start --> This change is <img src="https://reviewable.io/review_button.svg" height="34" align="absmiddle" alt="Reviewable"/> <!-- Reviewable:end -->

+4 -4

0 comment

2 changed files

proppy

pr closed time in 2 months

issue commenttensorflow/tensorflow

tf.debugging.enable_check_numerics() doesn't work on TPUs

Please note that in order to use tf.debugging.enable_check_numerics() on TPUs, currently you need to call

tf.config.set_soft_device_placement(True)

before calling tf.debugging.enable_check_numerics()

as this API uses automatic outside compilation on TPUs.

This is in the documentation in the source code at https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/debug/lib/check_numerics_callback.py#L415, but isn't reflected at the official documentation at https://www.tensorflow.org/api_docs/python/tf/debugging/enable_check_numerics yet because it's pending the next version update (2.3.0). Sorry for the confusion.

bjourne

comment created time in 2 months

issue commenttensorflow/tensorflow

Debugging predictions step with tf.debugging.experimental.enable_dump_debug_info throws an error.

Yep. The new error "Variable to save is not a Variable" is related to a known issue that the enable_dump_debug_info() API is incompatible with checkpoint saving. Can you disable saving in your program during debugging and try again?

mshavlovsky

comment created time in 2 months

issue commenttensorflow/tensorflow

Debugging predictions step with tf.debugging.experimental.enable_dump_debug_info throws an error.

@mshavlovsky Can you try the following

tf.debugging.experimental.enable_dump_debug_info("/tmp/debug", tensor_debug_mode="FULL_HEALTH", circular_buffer_size=-1)

The tensor_debug_mode specifies a mode that gives more information compared to the default "NO_TENSOR" mode; it'll also probably bypass the problem you are experiencing. The circular_buffer_size=-1 lets the debugger dump all tensor info, instead of only the most recent 1000.

See https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/debug/lib/dumping_callback.py#L747 for more info about the two arguments.

Let me know if this resolves your issue.

mshavlovsky

comment created time in 2 months

issue closedtensorflow/tensorflow

TFDBG doesn't display any tensor

<em>Please make sure that this is a bug. As per our GitHub Policy, we only address code/doc bugs, performance issues, feature requests and build/installation issues on GitHub. tag:bug_template</em>

System information

  • Have I written custom code (as opposed to using a stock example script provided in TensorFlow): No
  • OS Platform and Distribution (e.g., Linux Ubuntu 16.04): ubuntu18.04
  • Mobile device (e.g. iPhone 8, Pixel 2, Samsung Galaxy) if the issue happens on mobile device:
  • TensorFlow installed from (source or binary): binary
  • TensorFlow version (use command below): 2.3.0rc2
  • Python version: 3.6.8
  • Bazel version (if compiling from source):
  • GCC/Compiler version (if compiling from source):
  • CUDA/cuDNN version:
  • GPU model and memory:

You can collect some of this information using our environment capture script You can also obtain the TensorFlow version with:

  1. TF 1.0: python -c "import tensorflow as tf; print(tf.GIT_VERSION, tf.VERSION)"
  2. TF 2.0: python -c "import tensorflow as tf; print(tf.version.GIT_VERSION, tf.version.VERSION)"

Describe the current behavior Screenshot from 2020-07-21 10-49-27

Describe the expected behavior Display record tensors which like tf1.x Standalone code to reproduce the issue Provide a reproducible test case that is the bare minimum necessary to generate the problem. If possible, please share a link to Colab/Jupyter/any notebook.

python -m tensorflow.python.debug.examples.v2.debug_mnist_v2     --dump_dir /tmp/tfdbg2_logdir --dump_tensor_debug_mode FULL_HEALTH
python -m tensorflow.python.debug.cli.offline_analyzer --dump_dir="/tmp/tfdbg2_logdir"

Other info / logs Include any logs or source code that would be helpful to diagnose the problem. If including tracebacks, please include the full traceback. Large logs and files should be attached.

closed time in 2 months

fsx950223

issue commenttensorflow/tensorflow

TFDBG doesn't display any tensor

tensorflow.python.debug.cli.offline_analyzer is not compatible with the debugger v2 API in TF 2.x (i.e., the tf.debugging.experimental.enable_dump_debug_info() as used in the debug_mnist_v2 example). Instead, please use tensorboard (2.3+, to be released soon) to look at the dumped debug data.

Please take a look at https://www.tensorflow.org/tensorboard/debugger_v2 for more details.

fsx950223

comment created time in 2 months

more