Add missing algorithms, add stylistic improvements, update with spec conventions #446

zolkis · 2023-08-10T12:37:15Z

A series of changes including:

add collapsible stylistic boxes for algorithms and other stylistic improvements
add missing algorithms
some white space alignments
consistent args and variable names across the spec.

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Squashed from the following commits: Replace MLOperand.[[descriptor]] with type and dimensions Clarify the algorithm for only setting up the op Improve the clamp() algorithm, use the prose assuming the create steps for MLOperand and MLActivation Rework clamp with polymorphic behavior. Update for changes in MLOperand. Rework clamp() like constant(), polymorphic forms in separate sections, argument and return descriptions as notes. Fix platform related steps and reference to internal slots Address review, remove note Remove back quotes from title Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Squashed from the following commits: Add reference to the platform operand object Address review comment, change note. Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

…ithm Add the batch normalization algorithm

Add algorithm to concat()

Clamp algorithm

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

…ing algorithms Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

…integration Add stylistic changes

…to zk-conventions-integration

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Squashed from the following commits: Add reference to the platform operand object Address review comment, change note. Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Squashed from the following commits: Replace MLOperand.[[descriptor]] with type and dimensions Clarify the algorithm for only setting up the op Improve the clamp() algorithm, use the prose assuming the create steps for MLOperand and MLActivation Rework clamp with polymorphic behavior. Update for changes in MLOperand. Rework clamp() like constant(), polymorphic forms in separate sections, argument and return descriptions as notes. Fix platform related steps and reference to internal slots Address review, remove note Remove back quotes from title Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

…ing algorithms Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

…s/webnn into zk-conventions-integration

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

…, clamp, concat, batch norm Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

index.bs

lisa0314

@zolkis Leave some comments for remaining issue. PTAL. Thanks!

index.bs

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

miaobin

Just checked again and found an issue, PTAL.

index.bs

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

index.bs

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

lisa0314

@zolkis LGTM with one nit. PTAL. Thanks!

lisa0314 · 2023-08-25T08:47:42Z

index.bs

+  <div class=algorithm-steps>
+    1. [=Assert=]: |op| is one of "reduceL1", "reduceL2", "reduceLogSum", "reduceLogSumExp", "reduceMax", "reduceMean", "reduceMin", "reduceProduct", "reduceSum", "reduceSumSquare".
+    1. [=Assert=]: the type of |input| is {{MLOperand}}.
+    1. If |options|.{{MLReduceOptions/axes}} [=map/exists=], if any of its elements is not in [=the range=] 0 to the [=rank=] of |input|, exclusive, then [=exception/throw=] a "{{DataError}}" {{DOMException}}.


in [=the range=] 0 to N-1 where N is the [=rank=] of the input tensor.

The "exclusive" word does that. :)
The full expression is

in [=the range=] 0 to the [=rank=] of |input|, exclusive

huningxin

After the final check with other reviewers, all the comments are either addressed or tracked by issues. With that, this PR is ready for merging.

Thanks all for the productive review! Special thanks to @zolkis 's great work!

@wchao1115 , you may want to take another look before merging. Thanks!

anssiko · 2023-08-28T14:52:02Z

Thank you everyone for taking this major effort to completion with special thanks to @zolkis!

I asked @zolkis to acknowledge the contributors in 18442ad -- if any names are missing (or misspelled), please let us know and we'll fix.

@wchao1115, feel free to proceed at your earliest convenience.

anssiko · 2023-08-29T07:27:05Z

@wchao1115 @huningxin thank you for your approval. As the editors, I will let you discuss among yourself whether you would like to squash this big PR or retain the full commit history. You're free to merge when you are decided on the best path forward.

I have a slight preference for retaining the commits to help future spec archeologists. There's this idiom "how the sausage is made". :)

huningxin · 2023-08-29T07:38:56Z

I suppose even if we squash this PR and merge it as one commit, the full commit history in this PR would still be preserved, correct?

anssiko · 2023-08-29T08:03:11Z

@huningxin, if that works it would be best of both worlds: keep the main commit history clean while retain the detailed commit history in the PR.

I don't remember whether we need to keep the integration branch around (not delete it) for this to work. Perhaps you want to test this in a test repo to find out. Sometimes we do a clean up of stale branches and this branch might accidentally get deleted in the future.

zolkis · 2023-08-29T15:23:26Z

I suppose even if we squash this PR and merge it as one commit, the full commit history in this PR would still be preserved, correct?

I'd keep the separate commits because 1. there are different authors, and 2. one such big commit would be impractical for its sheer size. Up to the editors, though.

wchao1115 · 2023-08-29T18:37:48Z

@wchao1115 @huningxin thank you for your approval. As the editors, I will let you discuss among yourself whether you would like to squash this big PR or retain the full commit history. You're free to merge when you are decided on the best path forward.

I have a slight preference for retaining the commits to help future spec archeologists. There's this idiom "how the sausage is made". :)

I do not have a strong preference one way or the other. Either works for me.

huningxin · 2023-08-30T07:09:53Z

Thanks for the feedbacks. Let's merge this pull request with all commits history.

SHA: 34a4d80 Reason: push, by huningxin Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

zolkis and others added 30 commits May 16, 2023 18:28

Add the batch normalization algorithm

0eea13e

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Add the 'copy MLOperand' and 'create MLActivation' steps

d68bd87

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Add the concat algorithm

7082ee7

Squashed from the following commits: Add reference to the platform operand object Address review comment, change note. Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Add steps for 'rank', 'validate MLOperand'. Fix the concat() steps

1651b73

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

clamp(): improve platform related steps

dc4d3d6

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

concat: remove style from title

bce29de

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

concat: remove note styling from argument description

2845074

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Merge pull request webmachinelearning#397 from zolkis/batchnorm-algor…

093e38b

…ithm Add the batch normalization algorithm

Merge pull request webmachinelearning#398 from zolkis/concat-algorithm

828f044

Add algorithm to concat()

Merge pull request webmachinelearning#399 from zolkis/clamp-algorithm

9ad554d

Clamp algorithm

Add stylistic definitions for hiding algorithms, stylistic boxes

20499ba

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Adapt the existing main version to new style, without indenting exist…

257bcad

…ing algorithms Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Indent algorithms and fix make errors

e260d9f

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Merge pull request webmachinelearning#401 from zolkis/zk-conventions-…

329f41c

…integration Add stylistic changes

Merge branch 'main' of https://github.com/webmachinelearning/webnn in…

9b52017

…to zk-conventions-integration

Add the batch normalization algorithm

a6178e9

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Add the concat algorithm

013250e

Squashed from the following commits: Add reference to the platform operand object Address review comment, change note. Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Add steps for 'rank', 'validate MLOperand'. Fix the concat() steps

ef88d7c

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

concat: remove style from title

236bec6

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

concat: remove note styling from argument description

780d1a9

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Add the 'copy MLOperand' and 'create MLActivation' steps

02e977b

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

clamp(): improve platform related steps

5faf170

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Add stylistic definitions for hiding algorithms, stylistic boxes

1c89402

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Adapt the existing main version to new style, without indenting exist…

4a26973

…ing algorithms Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Indent algorithms and fix make errors

75925de

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Merge branch 'zk-conventions-integration' of https://github.com/zolki…

d5dd4f9

…s/webnn into zk-conventions-integration

Align the algorithms for input, constant, clamp, concat, batch norm etc

8fac5ca

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Adress review comments on aligning the algorithms for input, constant…

30ba12e

…, clamp, concat, batch norm Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

zolkis requested a review from huningxin August 24, 2023 09:11

zolkis added 2 commits August 24, 2023 14:33

Fix resample2d() steps

f937b88

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Add acks for webmachinelearning#446 and related work

18442ad

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

huningxin reviewed Aug 24, 2023

View reviewed changes

index.bs Show resolved Hide resolved

lisa0314 reviewed Aug 25, 2023

View reviewed changes

index.bs Outdated Show resolved Hide resolved

index.bs Outdated Show resolved Hide resolved

mingmingtasd reviewed Aug 25, 2023

View reviewed changes

index.bs Show resolved Hide resolved

mingmingtasd mentioned this pull request Aug 25, 2023

[Split] Validate the output shape of each output #458

Closed

zolkis added 2 commits August 25, 2023 09:33

Fix steps in resample2d(). Fix typo in conv2d().

3f9f3b3

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Fix webmachinelearning#459: correct the concat() steps

6818a1c

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

zolkis force-pushed the zk-conventions-integration branch from d04f364 to 6818a1c Compare August 25, 2023 07:47

miaobin reviewed Aug 25, 2023

View reviewed changes

index.bs Outdated Show resolved Hide resolved

zolkis added 2 commits August 25, 2023 11:02

Fix webmachinelearning#451: check options.axes in reduce operations

ffce2d6

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

Remove spurious step from matmul()

b5b95d4

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

mingmingtasd reviewed Aug 25, 2023

View reviewed changes

index.bs Outdated Show resolved Hide resolved

Improve concat() steps

c067b22

Signed-off-by: Zoltan Kis <zoltan.kis@intel.com>

lisa0314 reviewed Aug 25, 2023

View reviewed changes

huningxin approved these changes Aug 26, 2023

View reviewed changes

wchao1115 approved these changes Aug 29, 2023

View reviewed changes

huningxin merged commit 34a4d80 into webmachinelearning:main Aug 30, 2023
1 check passed

github-actions bot added a commit that referenced this pull request Aug 30, 2023

Merge pull request #446 from zolkis/zk-conventions-integration

2ad3f5c

SHA: 34a4d80 Reason: push, by huningxin Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

anssiko mentioned this pull request Aug 31, 2023

Staging area: discuss issues for the zk-conventions-integration branch #404

Closed

zolkis mentioned this pull request Oct 20, 2023

Empty tensors are valid for new shape #472

Closed

anssiko mentioned this pull request Dec 7, 2023

Add support for operations needed for well-known transformers e.g. Segment Anything, Stable Diffusion, etc. #478

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add missing algorithms, add stylistic improvements, update with spec conventions #446

Add missing algorithms, add stylistic improvements, update with spec conventions #446

zolkis commented Aug 10, 2023 •

edited by pr-preview bot

Loading

lisa0314 left a comment

miaobin left a comment

lisa0314 left a comment

lisa0314 Aug 25, 2023

zolkis Aug 25, 2023 •

edited

Loading

huningxin left a comment

anssiko commented Aug 28, 2023

anssiko commented Aug 29, 2023

huningxin commented Aug 29, 2023

anssiko commented Aug 29, 2023

zolkis commented Aug 29, 2023

wchao1115 commented Aug 29, 2023

huningxin commented Aug 30, 2023

Add missing algorithms, add stylistic improvements, update with spec conventions #446

Add missing algorithms, add stylistic improvements, update with spec conventions #446

Conversation

zolkis commented Aug 10, 2023 • edited by pr-preview bot Loading

lisa0314 left a comment

Choose a reason for hiding this comment

miaobin left a comment

Choose a reason for hiding this comment

lisa0314 left a comment

Choose a reason for hiding this comment

lisa0314 Aug 25, 2023

Choose a reason for hiding this comment

zolkis Aug 25, 2023 • edited Loading

Choose a reason for hiding this comment

huningxin left a comment

Choose a reason for hiding this comment

anssiko commented Aug 28, 2023

anssiko commented Aug 29, 2023

huningxin commented Aug 29, 2023

anssiko commented Aug 29, 2023

zolkis commented Aug 29, 2023

wchao1115 commented Aug 29, 2023

huningxin commented Aug 30, 2023

zolkis commented Aug 10, 2023 •

edited by pr-preview bot

Loading

zolkis Aug 25, 2023 •

edited

Loading