Web Machine Learning Working Group - Activity in 2023
No matches.
TR publications
-
3 documents published.
-
Web Neural Network API published 3 times as a Candidate Recommendation Draft. History of Web Neural Network API
-
Past meetings
-
1 meeting.
Participants
-
3 participants joined.
-
Tianqi Chen (W3C Invited Experts) joined.
-
Bryan Bernhart (Intel Corporation) joined.
-
Christos Bacharakis (eyeo GmbH) joined.
-
GitHub
webmachinelearning/webnn
-
11 issues opened.
-
Bitwise operators and logical operators naming (rename not to logicalNot) webmachinelearning/webnn#496, by @fdwr.
-
The input of `triangular` should be at least 2D tensor webmachinelearning/webnn#494, by @BruceDai.
-
Clarify the expected behavior for `axes` being empty `[]` webmachinelearning/webnn#493, by @BruceDai.
-
Constant sequential filling operation needs output shape parameter webmachinelearning/webnn#492, by @fdwr.
-
Clarify `instanceNormalization`'s support for 1d, 2d, 3d use cases webmachinelearning/webnn#490, by @huningxin.
-
Clarify the casting behavior from floating-point / signed integers <-> unsigned integers webmachinelearning/webnn#489, by @huningxin.
-
Please correct the emulation for where op webmachinelearning/webnn#488, by @BruceDai.
-
Should `axes` be required parameter for layerNormalization build method? webmachinelearning/webnn#487, by @huningxin.
-
Add "implementation consideration" about how out-of-bound indices of Gather/Scatter should be handled webmachinelearning/webnn#486, by @huningxin.
-
Implement webnn-baseline and WPT for operations needed for well-known transformers webmachinelearning/webnn#485, by @huningxin.
-
Should Gather's indices data type be integers and support negative value? webmachinelearning/webnn#484, by @huningxin.
-
-
3 issues closed.
-
(bikeshed naming) MLOperandDataType webmachinelearning/webnn#454.
-
Add "Die On: warning" to spec metadata webmachinelearning/webnn#468.
-
Please correct the emulation for where op webmachinelearning/webnn#488.
-
-
15 issues received 29 new comments.
-
Support for device-based tensor storage objects webmachinelearning/webnn#482 (9 comments by @bbernhar, @huningxin, @inexorabletash, @wchao1115, @zolkis)
-
Please correct the emulation for where op webmachinelearning/webnn#488 (3 comments by @anssiko, @BruceDai, @fdwr)
-
Add traversal algorithm to buildSync() webmachinelearning/webnn#448 (2 comments by @inexorabletash, @reillyeon)
-
The input of `triangular` should be at least 2D tensor webmachinelearning/webnn#494 (1 comment by @fdwr)
-
Clarify the expected behavior for `axes` being empty `[]` webmachinelearning/webnn#493 (1 comment by @fdwr)
-
Define the algorithm of calculating the effective padding for "same-upper" and "same-lower" option webmachinelearning/webnn#326 (1 comment by @Honry)
-
Clarify `instanceNormalization`'s support for 1d, 2d, 3d use cases webmachinelearning/webnn#490 (1 comment by @fdwr)
-
Implement webnn-baseline and WPT for operations needed for well-known transformers webmachinelearning/webnn#485 (1 comment by @BruceDai)
-
Clarify the casting behavior from floating-point / signed integers <-> unsigned integers webmachinelearning/webnn#489 (1 comment by @zolkis)
-
Should `axes` be required parameter for layerNormalization build method? webmachinelearning/webnn#487 (1 comment by @fdwr)
-
Should `scale` and `bias` be required inputs for `batchNormalization` op? webmachinelearning/webnn#481 (1 comment by @huningxin)
-
-
2 pull requests opened.
-
Adopt Bikeshed best practices and Infra spec conventions webmachinelearning/webnn#495, by @inexorabletash.
-
Fix a handful of markdown glitches, probably c/o a find/replace. webmachinelearning/webnn#491, by @inexorabletash.
-
-
3 pull requests merged.
-
Define the Bikeshed build to fail on warnings webmachinelearning/webnn#469.
-
Fix a handful of markdown glitches, probably c/o a find/replace. webmachinelearning/webnn#491.
-
Add support for operations needed for well-known transformers e.g. Segment Anything, Stable Diffusion, etc. webmachinelearning/webnn#478.
-
-
4 pull requests received 84 new comments.
-
Add support for operations needed for well-known transformers e.g. Segment Anything, Stable Diffusion, etc. webmachinelearning/webnn#478 (79 comments by @anssiko, @BruceDai, @fdwr, @huningxin, @miaobin, @wchao1115, @zolkis)
-
Adopt Bikeshed best practices and Infra spec conventions webmachinelearning/webnn#495 (2 comments by @zolkis)
-
Fix a handful of markdown glitches, probably c/o a find/replace. webmachinelearning/webnn#491 (2 comments by @huningxin, @inexorabletash)
-
Define the Bikeshed build to fail on warnings webmachinelearning/webnn#469 (1 comment by @inexorabletash)
-
Past meetings
Participants
-
2 participants joined.
-
Laszlo Gombos (Samsung Electronics Co., Ltd.) joined.
-
Chus Garcia (Fundacion CTIC) joined.
-
-
1 organization joined.
-
Samsung Electronics Co., Ltd. joined.
-
GitHub
webmachinelearning/webnn
-
6 issues opened.
-
Consistency for dictionary member definitions webmachinelearning/webnn#483, by @inexorabletash.
-
Support for device-based tensor storage objects webmachinelearning/webnn#482, by @bbernhar.
-
Should `scale` and `bias` be required inputs for `batchNormalization` op? webmachinelearning/webnn#481, by @huningxin.
-
Proposed use cases from Web LLM project webmachinelearning/webnn#480, by @anssiko.
-
API lacks handling for async ML device errors on the context webmachinelearning/webnn#477, by @bbernhar.
-
Remove `builder.constant(value, type)` variant webmachinelearning/webnn#475, by @huningxin.
-
-
11 issues received 33 new comments.
-
Support for transformers webmachinelearning/webnn#375 (9 comments by @anssiko, @fdwr, @philloooo, @tqchen)
-
Allow 0 size dimensions (dimensions containing a 0 in the list of sizes, not a rank of 0 which is valid) webmachinelearning/webnn#391 (5 comments by @fdwr, @huningxin)
-
Specify the operand data type constraints of operation webmachinelearning/webnn#283 (3 comments by @fdwr, @huningxin)
-
Should `scale` and `bias` be required inputs for `batchNormalization` op? webmachinelearning/webnn#481 (2 comments by @fdwr, @wchao1115)
-
A conformance suite with disallowed intra-op examples would be helpful for hardening. webmachinelearning/webnn#244 (2 comments by @fdwr, @huningxin)
-
Define the algorithm of calculating the effective padding for "same-upper" and "same-lower" option webmachinelearning/webnn#326 (2 comments by @fdwr, @huningxin)
-
Remove `builder.constant(value, type)` variant webmachinelearning/webnn#475 (2 comments by @fdwr, @wacky6)
-
Clarify the restriction for `minValue` and `maxValue` of `MLClampOptions` webmachinelearning/webnn#396 (1 comment by @wchao1115)
-
-
3 pull requests opened.
-
Just delete this webmachinelearning/webnn#479, by @ai-must-die.
-
Add support for operations needed for well-known transformers e.g. Segment Anything, Stable Diffusion, etc. webmachinelearning/webnn#478, by @wchao1115.
-
Fix dimensions for 0D scalars webmachinelearning/webnn#476, by @fdwr.
-
-
1 pull request closed.
-
Just delete this webmachinelearning/webnn#479.
-
-
2 pull requests received 125 new comments.
-
Add support for operations needed for well-known transformers e.g. Segment Anything, Stable Diffusion, etc. webmachinelearning/webnn#478 (110 comments by @anssiko, @BruceDai, @fdwr, @huningxin, @inexorabletash, @shiyi9801, @wacky6, @wchao1115)
-
Fix dimensions for 0D scalars webmachinelearning/webnn#476 (15 comments by @fdwr, @huningxin, @inexorabletash, @zolkis)
-
TR publications
-
1 document published.
-
Web Neural Network API published as a Candidate Recommendation Draft. History of Web Neural Network API
-
Past meetings
Participants
-
5 participants joined.
-
Austin Sullivan (Google LLC) joined.
-
Jason Stephen (Microsoft Corporation) joined.
-
Christian Weyer (Thinktecture) joined.
-
Reilly Grant (Google LLC) joined.
-
Phillis Tang (Google LLC) joined.
-
-
1 organization joined.
-
Thinktecture joined.
-
GitHub
webmachinelearning/webnn
-
7 issues opened.
-
Simplify `resample2d` op webmachinelearning/webnn#474, by @huningxin.
-
Facial Landmark Detection for spatial computing webmachinelearning/webnn#473, by @munrocket.
-
Empty tensors are valid for new shape webmachinelearning/webnn#472, by @zolkis.
-
Should softmax be a fused activation ? webmachinelearning/webnn#471, by @mingmingtasd.
-
Simplify `matmul` op webmachinelearning/webnn#470, by @huningxin.
-
Add "Die On: warning" to spec metadata webmachinelearning/webnn#468, by @inexorabletash.
-
Where is scatter and gather op? webmachinelearning/webnn#467, by @muazhuda.
-
-
1 issue closed.
-
Facial Landmark Detection for spatial computing webmachinelearning/webnn#473.
-
-
11 issues received 23 new comments.
-
Specify the operand data type constraints of operation webmachinelearning/webnn#283 (4 comments by @fdwr, @huningxin, @wacky6)
-
Add "Die On: warning" to spec metadata webmachinelearning/webnn#468 (4 comments by @anssiko, @dontcallmedom, @inexorabletash)
-
Support for transformers webmachinelearning/webnn#375 (3 comments by @Honry, @sushraja-msft, @wacky6)
-
Facial Landmark Detection for spatial computing webmachinelearning/webnn#473 (1 comment by @anssiko)
-
Unsqueeze operator is missing (so add unsqueeze, or better yet, delete squeeze) webmachinelearning/webnn#296 (1 comment by @fdwr)
-
Type of some parameters should match the input data type webmachinelearning/webnn#442 (1 comment by @inexorabletash)
-
1 pull request opened.
-
Define the Bikeshed build to fail on warnings webmachinelearning/webnn#469, by @anssiko.
-
-
1 pull request merged.
-
MLOperandType -> MLOperandDataType webmachinelearning/webnn#464.
-
-
3 pull requests received 10 new comments.
-
MLOperandType -> MLOperandDataType webmachinelearning/webnn#464 (8 comments by @anssiko, @fdwr, @Honry, @inexorabletash)
-
Add missing algorithms, add stylistic improvements, update with spec conventions webmachinelearning/webnn#446 (1 comment by @fdwr)
-
Past meetings
Participants
-
3 participants joined.
-
Deepti Gandluri (Google LLC) joined.
-
Etienne Noël (Google LLC) joined.
-
Tzviya Siegman (Wiley) joined.
-
-
1 organization joined.
-
Wiley joined.
-
GitHub
webmachinelearning/webnn
-
4 issues opened.
-
Softmax axis absent webmachinelearning/webnn#466, by @fdwr.
-
`keepDimensions` of reduction ops webmachinelearning/webnn#465, by @lisa0314.
-
Allow checking whether operators/types are supported for a backend before creating a graph webmachinelearning/webnn#463, by @huningxin.
-
Improve "underlying platform" references in spec webmachinelearning/webnn#462, by @inexorabletash.
-
-
9 issues received 24 new comments.
-
Allow to hint number of threads for CPU MLContext webmachinelearning/webnn#436 (9 comments by @dontcallmedom, @fdwr, @huningxin, @wacky6)
-
Allow checking whether operators/types are supported for a backend before creating a graph webmachinelearning/webnn#463 (4 comments by @huningxin, @wacky6)
-
Improve "underlying platform" references in spec webmachinelearning/webnn#462 (3 comments by @huningxin, @inexorabletash, @zolkis)
-
Allow 0 size dimensions (dimensions containing a 0 in the list of sizes, not a rank of 0 which is valid) webmachinelearning/webnn#391 (1 comment by @fdwr)
-
-
1 pull request opened.
-
MLOperandType -> MLOperandDataType webmachinelearning/webnn#464, by @fdwr.
-
-
1 pull request received 12 new comments.
-
MLOperandType -> MLOperandDataType webmachinelearning/webnn#464 (12 comments by @BruceDai, @fdwr, @Honry, @huningxin, @wchao1115)
-
TR publications
-
3 documents published.
-
Web Neural Network API published twice as a Candidate Recommendation Draft. History of Web Neural Network API
-
Ethical Principles for Web Machine Learning published as a Draft Note. History of Ethical Principles for Web Machine Learning
-
Past meetings
Participants
-
1 participant joined.
-
Joshua Bell (Google LLC) joined.
-
GitHub
webmachinelearning/webnn
-
14 issues opened.
-
[Concat] Fix error of validating each input operand type webmachinelearning/webnn#459, by @mingmingtasd.
-
[Split] Validate the output shape of each output webmachinelearning/webnn#458, by @mingmingtasd.
-
Complete the build algorithms webmachinelearning/webnn#457, by @zolkis.
-
Define the maximum number of operand dimensions (maximum rank) webmachinelearning/webnn#456, by @huningxin.
-
Reserve asserts only for non-obvious things webmachinelearning/webnn#455, by @zolkis.
-
(bikeshed naming) MLOperandDataType webmachinelearning/webnn#454, by @fdwr.
-
Google Chrome Feedback on WebNN: aiming for broad device coverage and maintainability webmachinelearning/webnn#453, by @vsekhar.
-
Reshape: what if |newShape| is a scalar number webmachinelearning/webnn#452, by @zolkis.
-
The default values of |options|.{{MLReduceOptions/axes}} should be {0, ...., N-1} rather than null. webmachinelearning/webnn#451, by @zolkis.
-
Use web spec best practices webmachinelearning/webnn#450, by @zolkis.
-
Fix description and formula in pooling operations webmachinelearning/webnn#449, by @zolkis.
-
Add traversal algorithm to buildSync() webmachinelearning/webnn#448, by @zolkis.
-
Improve algorithmic steps definitions webmachinelearning/webnn#447, by @zolkis.
-
Zero dilations needs spec clarification webmachinelearning/webnn#444, by @fdwr.
-
-
7 issues closed.
-
Staging area: discuss issues for the zk-conventions-integration branch webmachinelearning/webnn#404.
-
The default values of |options|.{{MLReduceOptions/axes}} should be {0, ...., N-1} rather than null. webmachinelearning/webnn#451.
-
[Concat] Fix error of validating each input operand type webmachinelearning/webnn#459.
-
[Split] Validate the output shape of each output webmachinelearning/webnn#458.
-
Improve algorithmic steps definitions webmachinelearning/webnn#447.
-
Add missing conv2d validation steps webmachinelearning/webnn#439.
-
Review sync vs async compute differences webmachinelearning/webnn#316.
-
-
1 issue reopened.
-
Use web spec best practices webmachinelearning/webnn#450, by @anssiko.
-
-
20 issues received 84 new comments.
-
Use web spec best practices webmachinelearning/webnn#450 (43 comments by @anssiko, @inexorabletash, @zolkis)
-
Add security consideration for computation control-flow attack based on weights / constants change webmachinelearning/webnn#443 (5 comments by @fdwr, @quidity, @wacky6)
-
Use modern WebIDL and Infra standard conventions webmachinelearning/webnn#210 (4 comments by @anssiko, @wchao1115, @zolkis)
-
Define the maximum number of operand dimensions (maximum rank) webmachinelearning/webnn#456 (4 comments by @fdwr, @Maratyszcza, @wacky6)
-
Simplify the operand layout support of conv2d and pooling 2d operations webmachinelearning/webnn#324 (4 comments by @huningxin, @wacky6, @wchao1115, @zolkis)
-
[Split] Validate the output shape of each output webmachinelearning/webnn#458 (3 comments by @mingmingtasd, @zolkis)
-
(bikeshed naming) MLOperandDataType webmachinelearning/webnn#454 (3 comments by @anssiko, @fdwr, @zolkis)
-
The default values of |options|.{{MLReduceOptions/axes}} should be {0, ...., N-1} rather than null. webmachinelearning/webnn#451 (2 comments by @lisa0314, @zolkis)
-
Improve algorithmic steps definitions webmachinelearning/webnn#447 (2 comments by @inexorabletash, @zolkis)
-
Staging area: discuss issues for the zk-conventions-integration branch webmachinelearning/webnn#404 (1 comment by @anssiko)
-
[Concat] Fix error of validating each input operand type webmachinelearning/webnn#459 (1 comment by @mingmingtasd)
-
Fix description and formula in pooling operations webmachinelearning/webnn#449 (1 comment by @zolkis)
-
Google Chrome Feedback on WebNN: aiming for broad device coverage and maintainability webmachinelearning/webnn#453 (1 comment by @anssiko)
-
Define the data type of the padding, strides and dilations of MLConv2dOptions as sequence<unsigned long> webmachinelearning/webnn#269 (1 comment by @fdwr)
-
-
3 pull requests opened.
-
Do not include metadata collapsible into expand/collapse all button webmachinelearning/webnn#460, by @dontcallmedom.
-
Add missing algorithms, add stylistic improvements, update with spec conventions webmachinelearning/webnn#446, by @zolkis.
-
Fix missing <details open> tags webmachinelearning/webnn#445, by @wchao1115.
-
-
6 pull requests merged.
-
Do not include metadata collapsible into expand/collapse all button webmachinelearning/webnn#460.
-
Add missing algorithms, add stylistic improvements, update with spec conventions webmachinelearning/webnn#446.
-
Fix #439: add missing conv2d() and convTranspose2d() validation steps webmachinelearning/webnn#441.
-
Add the command finish steps webmachinelearning/webnn#440.
-
Rework the sync async algorithms based on PR 323 webmachinelearning/webnn#329.
-
Fix missing <details open> tags webmachinelearning/webnn#445.
-
-
6 pull requests received 239 new comments.
-
Add missing algorithms, add stylistic improvements, update with spec conventions webmachinelearning/webnn#446 (230 comments by @anssiko, @BruceDai, @fujunwei, @huningxin, @inexorabletash, @lisa0314, @miaobin, @mingmingtasd, @shiyi9801, @wchao1115, @zolkis)
-
Rework the sync async algorithms based on PR 323 webmachinelearning/webnn#329 (3 comments by @anssiko, @wchao1115, @zolkis)
-
Do not include metadata collapsible into expand/collapse all button webmachinelearning/webnn#460 (1 comment by @anssiko)
-
Fix #439: add missing conv2d() and convTranspose2d() validation steps webmachinelearning/webnn#441 (1 comment by @anssiko)
-
webmachinelearning/webmachinelearning-ethics
-
3 pull requests opened.
-
Update references to Process Document webmachinelearning/webmachinelearning-ethics#31, by @dontcallmedom.
-
Update github workflow webmachinelearning/webmachinelearning-ethics#30, by @dontcallmedom.
-
Update James' status to Former Editor webmachinelearning/webmachinelearning-ethics#29, by @anssiko.
-
-
3 pull requests merged.
-
Update references to Process Document webmachinelearning/webmachinelearning-ethics#31.
-
Update github workflow webmachinelearning/webmachinelearning-ethics#30.
-
Update James' status to Former Editor webmachinelearning/webmachinelearning-ethics#29.
-
-
1 pull request received 6 new comments.
-
Update James' status to Former Editor webmachinelearning/webmachinelearning-ethics#29 (6 comments by @anssiko, @dontcallmedom)
-
Participants
-
1 participant left.
-
James Fletcher (British Broadcasting Corporation) left.
-
GitHub
webmachinelearning/webnn
-
3 issues opened.
-
Add security consideration for computation control-flow attack based on weights / constants change webmachinelearning/webnn#443, by @huningxin.
-
Type of some parameters should match the input data type webmachinelearning/webnn#442, by @Honry.
-
Add missing conv2d validation steps webmachinelearning/webnn#439, by @zolkis.
-
-
4 issues received 13 new comments.
-
Should we add some model-based tests into WPT webmachinelearning/webnn#402 (9 comments by @BruceDai, @fdwr, @huningxin, @wacky6, @wchao1115)
-
Allow 0 size dimensions (dimensions containing a 0 in the list of sizes, not a rank of 0 which is valid) webmachinelearning/webnn#391 (2 comments by @fdwr, @miaobin)
-
Simplify the operand layout support of conv2d and pooling 2d operations webmachinelearning/webnn#324 (1 comment by @wacky6)
-
Type of some parameters should match the input data type webmachinelearning/webnn#442 (1 comment by @fdwr)
-
-
2 pull requests opened.
-
Fix #439: add missing conv2d() and convTranspose2d() validation steps webmachinelearning/webnn#441, by @zolkis.
-
Add the command finish steps webmachinelearning/webnn#440, by @zolkis.
-
-
2 pull requests received 2 new comments.
TR publications
-
2 documents published.
-
Web Neural Network API published twice as a Candidate Recommendation Draft. History of Web Neural Network API
-
Past meetings
Participants
-
2 participants joined.
-
Joshua Lochner (Hugging Face) joined.
-
Margery Clacy (Centre for Inclusive Design) joined.
-
-
1 organization joined.
-
Centre for Inclusive Design joined.
-
GitHub
webmachinelearning/webnn
-
7 issues opened.
-
Support dedicated square root operator webmachinelearning/webnn#438, by @huningxin.
-
Allow to hint number of threads for CPU MLContext webmachinelearning/webnn#436, by @huningxin.
-
Staging area: discuss issues for the zk-conventions-integration branch webmachinelearning/webnn#404, by @zolkis.
-
Should we add some model-based tests into WPT webmachinelearning/webnn#402, by @BruceDai.
-
Clarify the restriction for `minValue` and `maxValue` of `MLClampOptions` webmachinelearning/webnn#396, by @huningxin.
-
Use deep copy for setting internal slots? webmachinelearning/webnn#395, by @zolkis.
-
The `MLGraphBuilder.input()` algorithm passes one more argument to `create MLOperand` algorithm webmachinelearning/webnn#393, by @huningxin.
-
-
3 issues closed.
-
Add internal slots to MLOperand, MLActivation and basic algorithms webmachinelearning/webnn#336.
-
The `MLGraphBuilder.input()` algorithm passes one more argument to `create MLOperand` algorithm webmachinelearning/webnn#393.
-
heads up re: proposal to add Float16Array to JavaScript webmachinelearning/webnn#373.
-
-
10 issues received 26 new comments.
-
API simplification: context owns builder, graph becomes internal slot webmachinelearning/webnn#303 (6 comments by @anssiko, @wacky6, @wchao1115, @zolkis)
-
Use deep copy for setting internal slots? webmachinelearning/webnn#395 (5 comments by @anssiko, @zolkis)
-
Use modern WebIDL and Infra standard conventions webmachinelearning/webnn#210 (4 comments by @anssiko, @wchao1115, @zolkis)
-
Add internal slots to MLOperand, MLActivation and basic algorithms webmachinelearning/webnn#336 (2 comments by @anssiko, @zolkis)
-
heads up re: proposal to add Float16Array to JavaScript webmachinelearning/webnn#373 (2 comments by @anssiko, @bakkot)
-
Allow 0 size dimensions (dimensions containing a 0 in the list of sizes, not a rank of 0 which is valid) webmachinelearning/webnn#391 (1 comment by @huningxin)
-
Clarify the restriction for `minValue` and `maxValue` of `MLClampOptions` webmachinelearning/webnn#396 (1 comment by @fdwr)
-
-
39 pull requests opened.
-
Fix merge errors webmachinelearning/webnn#437, by @zolkis.
-
Add the transpose algorithm webmachinelearning/webnn#435, by @zolkis.
-
Add the tanh algorithm webmachinelearning/webnn#434, by @zolkis.
-
Add the squeeze() algorithm webmachinelearning/webnn#433, by @zolkis.
-
Add the split algorithm webmachinelearning/webnn#432, by @zolkis.
-
Add the softsign method webmachinelearning/webnn#431, by @zolkis.
-
Add the softplus algorithm webmachinelearning/webnn#430, by @zolkis.
-
Add the softmax algorithm webmachinelearning/webnn#429, by @zolkis.
-
Add the slice algorithm webmachinelearning/webnn#428, by @zolkis.
-
Add the sigmoid algorithm webmachinelearning/webnn#427, by @zolkis.
-
Add the reshape algorithm webmachinelearning/webnn#426, by @zolkis.
-
Add the resample2d() algorithm webmachinelearning/webnn#425, by @zolkis.
-
Add the relu() algorithm webmachinelearning/webnn#424, by @zolkis.
-
Add the reduction algorithms webmachinelearning/webnn#423, by @zolkis.
-
Add the prelu() algorithm webmachinelearning/webnn#422, by @zolkis.
-
Add the pooling algorithms webmachinelearning/webnn#421, by @zolkis.
-
Add the pad() algorithm webmachinelearning/webnn#420, by @zolkis.
-
Add the matmul() algorithm webmachinelearning/webnn#419, by @zolkis.
-
Add the LSTM and LSTM Cell algorithms webmachinelearning/webnn#418, by @zolkis.
-
Add the linear function algorithm webmachinelearning/webnn#417, by @zolkis.
-
Add the Leaky RELU algorithm webmachinelearning/webnn#416, by @zolkis.
-
Add the instance normalization algorithm webmachinelearning/webnn#415, by @zolkis.
-
Add the hard-swish algorithm webmachinelearning/webnn#414, by @zolkis.
-
Add the hard-sigmoid algorithms webmachinelearning/webnn#413, by @zolkis.
-
Add the gru() and gruCell() algorithms webmachinelearning/webnn#412, by @zolkis.
-
Add the gemm() algorithm webmachinelearning/webnn#411, by @zolkis.
-
Add the elu() algorithm webmachinelearning/webnn#410, by @zolkis.
-
Add element-wise unary algorithms webmachinelearning/webnn#409, by @zolkis.
-
Add element-wise binary operation algorithms webmachinelearning/webnn#408, by @zolkis.
-
Add the conv2d and convTranspose2d algorithms webmachinelearning/webnn#407, by @zolkis.
-
Add the build() and buildSync() algorithms webmachinelearning/webnn#406, by @zolkis.
-
Align algorithms defined so far webmachinelearning/webnn#405, by @zolkis.
-
pull from main webmachinelearning/webnn#403, by @zolkis.
-
Add stylistic changes webmachinelearning/webnn#401, by @zolkis.
-
Operand activation improvements webmachinelearning/webnn#400, by @zolkis.
-
Clamp algorithm webmachinelearning/webnn#399, by @zolkis.
-
Add algorithm to concat() webmachinelearning/webnn#398, by @zolkis.
-
Add the batch normalization algorithm webmachinelearning/webnn#397, by @zolkis.
-
Fix #393 for input() webmachinelearning/webnn#394, by @zolkis.
-
-
39 pull requests merged.
-
Fix merge errors webmachinelearning/webnn#437.
-
Add the tanh algorithm webmachinelearning/webnn#434.
-
Add the softsign method webmachinelearning/webnn#431.
-
Add the softmax algorithm webmachinelearning/webnn#429.
-
Add the resample2d() algorithm webmachinelearning/webnn#425.
-
Add the prelu() algorithm webmachinelearning/webnn#422.
-
Add the transpose algorithm webmachinelearning/webnn#435.
-
Add the squeeze() algorithm webmachinelearning/webnn#433.
-
Add the split algorithm webmachinelearning/webnn#432.
-
Add the sigmoid algorithm webmachinelearning/webnn#427.
-
Add the softplus algorithm webmachinelearning/webnn#430.
-
Add the slice algorithm webmachinelearning/webnn#428.
-
Add the reshape algorithm webmachinelearning/webnn#426.
-
Add the relu() algorithm webmachinelearning/webnn#424.
-
Add the reduction algorithms webmachinelearning/webnn#423.
-
Add the pooling algorithms webmachinelearning/webnn#421.
-
Add the pad() algorithm webmachinelearning/webnn#420.
-
Add the matmul() algorithm webmachinelearning/webnn#419.
-
Add the LSTM and LSTM Cell algorithms webmachinelearning/webnn#418.
-
Add the linear function algorithm webmachinelearning/webnn#417.
-
Add the Leaky RELU algorithm webmachinelearning/webnn#416.
-
Add the instance normalization algorithm webmachinelearning/webnn#415.
-
Add the hard-swish algorithm webmachinelearning/webnn#414.
-
Add the hard-sigmoid algorithms webmachinelearning/webnn#413.
-
Add the gru() and gruCell() algorithms webmachinelearning/webnn#412.
-
Add the gemm() algorithm webmachinelearning/webnn#411.
-
Add the elu() algorithm webmachinelearning/webnn#410.
-
Add element-wise unary algorithms webmachinelearning/webnn#409.
-
Add element-wise binary operation algorithms webmachinelearning/webnn#408.
-
Add the conv2d and convTranspose2d algorithms webmachinelearning/webnn#407.
-
Add the build() and buildSync() algorithms webmachinelearning/webnn#406.
-
Align algorithms defined so far webmachinelearning/webnn#405.
-
Add stylistic changes webmachinelearning/webnn#401.
-
Fix #393 for input() webmachinelearning/webnn#394.
-
Fix typo error for lstm operations webmachinelearning/webnn#385.
-
Clamp algorithm webmachinelearning/webnn#399.
-
Add algorithm to concat() webmachinelearning/webnn#398.
-
Add the batch normalization algorithm webmachinelearning/webnn#397.
-
Add float16 to MLOperandType and ArrayBufferView compatibility table webmachinelearning/webnn#386.
-
-
6 pull requests closed.
-
Operand activation improvements webmachinelearning/webnn#400.
-
pull from main webmachinelearning/webnn#403.
-
Add internal slots to MLOperand and MLActivation webmachinelearning/webnn#337.
-
Add the clamp() algorithm webmachinelearning/webnn#348.
-
Add the concat algorithm webmachinelearning/webnn#366.
-
Fix #334: Improve the batch norm algorithm webmachinelearning/webnn#339.
-
-
45 pull requests received 108 new comments.
-
Align algorithms defined so far webmachinelearning/webnn#405 (13 comments by @anssiko, @huningxin, @zolkis)
-
Rework the sync async algorithms based on PR 323 webmachinelearning/webnn#329 (8 comments by @anssiko, @wchao1115, @zolkis)
-
Add the prelu() algorithm webmachinelearning/webnn#422 (6 comments by @anssiko, @dontcallmedom, @zolkis)
-
Fix typo error for lstm operations webmachinelearning/webnn#385 (4 comments by @anssiko, @BruceDai, @wchao1115)
-
Add float16 to MLOperandType and ArrayBufferView compatibility table webmachinelearning/webnn#386 (4 comments by @anssiko, @wchao1115)
-
Add the LSTM and LSTM Cell algorithms webmachinelearning/webnn#418 (2 comments by @anssiko, @zolkis)
-
Add internal slots to MLOperand and MLActivation webmachinelearning/webnn#337 (1 comment by @zolkis)
TR publications
-
3 documents published.
-
Web Neural Network API published 3 times as a Candidate Recommendation Draft. History of Web Neural Network API
-
Past meetings
Participants
-
3 participants joined.
-
Vivek Sekhar (Google LLC) joined.
-
Dwayne Robinson (Microsoft Corporation) joined.
-
Thomas Meier (NVIDIA) joined.
-
-
1 organization joined.
-
NVIDIA joined.
-
GitHub
webmachinelearning/webnn
-
5 issues opened.
-
`split` into sizes are not widely supported webmachinelearning/webnn#392, by @huningxin.
-
Allow 0 size dimensions (dimensions containing a 0 in the list of sizes, not a rank of 0 which is valid) webmachinelearning/webnn#391, by @huningxin.
-
0D scalars webmachinelearning/webnn#390, by @fdwr.
-
Remove the support of `null` value in `newShape` parameter of `reshape` build method webmachinelearning/webnn#388, by @huningxin.
-
Need to restrict the value of alpha to be positive for elu operation webmachinelearning/webnn#383, by @lisa0314.
-
-
9 issues received 23 new comments.
-
Need to define error handling of MLNamedArrayBufferViews transfer algorithm webmachinelearning/webnn#351 (4 comments by @domenic, @huningxin)
-
Need to restrict the value of alpha to be positive for elu operation webmachinelearning/webnn#383 (4 comments by @fdwr, @huningxin)
-
heads up re: proposal to add Float16Array to JavaScript webmachinelearning/webnn#373 (3 comments by @anssiko, @bakkot)
-
Support `depth_multiplier > 1` for a depthwise conv2d op webmachinelearning/webnn#353 (2 comments by @anssiko, @Honry)
-
Subclass MLGraph based on the context that creates it webmachinelearning/webnn#344 (2 comments by @anssiko, @zolkis)
-
Remove the support of `null` value in `newShape` parameter of `reshape` build method webmachinelearning/webnn#388 (2 comments by @fdwr, @wacky6)
-
Allow 0 size dimensions (dimensions containing a 0 in the list of sizes, not a rank of 0 which is valid) webmachinelearning/webnn#391 (1 comment by @fdwr)
-
Please clarify interpolation algorithm for resample2d webmachinelearning/webnn#358 (1 comment by @anssiko)
-
5 pull requests opened.
-
Acknowledge Jiewei Qian for Chromium implementation review and feedback webmachinelearning/webnn#389, by @huningxin.
-
Fix first-wave models table formatting webmachinelearning/webnn#387, by @anssiko.
-
Add float16 to MLOperandType and ArrayBufferView compatibility table webmachinelearning/webnn#386, by @anssiko.
-
Fix typo error for lstm operations webmachinelearning/webnn#385, by @BruceDai.
-
Fix a typo of elu sample code webmachinelearning/webnn#384, by @huningxin.
-
-
5 pull requests merged.
-
Acknowledge Jiewei Qian for Chromium implementation review and feedback webmachinelearning/webnn#389.
-
Add the builder.input() method steps webmachinelearning/webnn#364.
-
Fix a typo of elu sample code webmachinelearning/webnn#384.
-
Add the constant() method steps. webmachinelearning/webnn#365.
-
Fix first-wave models table formatting webmachinelearning/webnn#387.
-
-
12 pull requests received 44 new comments.
-
Add the constant() method steps. webmachinelearning/webnn#365 (16 comments by @huningxin, @wchao1115, @zolkis)
-
Add the builder.input() method steps webmachinelearning/webnn#364 (12 comments by @huningxin, @zolkis)
-
Acknowledge Jiewei Qian for Chromium implementation review and feedback webmachinelearning/webnn#389 (2 comments by @anssiko)
-
Add float16 to MLOperandType and ArrayBufferView compatibility table webmachinelearning/webnn#386 (2 comments by @anssiko)
-
Rework the sync async algorithms based on PR 323 webmachinelearning/webnn#329 (2 comments by @anssiko, @zolkis)
-
Fix first-wave models table formatting webmachinelearning/webnn#387 (2 comments by @anssiko, @ibelem)
-
Add internal slots to MLOperand and MLActivation webmachinelearning/webnn#337 (1 comment by @zolkis)
-
Lifecycle
-
1 charter started.
-
Chartered until 9 March 2025. Participants had to rejoin under this charter.
-
TR publications
-
1 document published.
-
Web Neural Network API published as a Candidate Recommendation Draft. History of Web Neural Network API
-
Past meetings
GitHub
webmachinelearning/webnn
-
8 issues opened.
-
Why is ml a member of Navigator? webmachinelearning/webnn#382, by @bathos.
-
MLOperand::dimensions() or MLGraphBuilder::shape() for debugging purposes and diagnostics webmachinelearning/webnn#379, by @fdwr.
-
Add samples of shape broadcasting webmachinelearning/webnn#378, by @huningxin.
-
Need clarify constraint of 'beginningPadding' and 'endingPadding' for pad operation in "reflection" and "symmetric" mode webmachinelearning/webnn#377, by @BruceDai.
-
Support for transformers webmachinelearning/webnn#375, by @dontcallmedom.
-
Simplify `MLPool2dOptions` by removing the `outputSizes` option webmachinelearning/webnn#374, by @huningxin.
-
heads up re: proposal to add Float16Array to JavaScript webmachinelearning/webnn#373, by @bakkot.
-
Need to update the spec links to latest published version webmachinelearning/webnn#371, by @ibelem.
-
-
4 issues closed.
-
Why is ml a member of Navigator? webmachinelearning/webnn#382.
-
Create a light-weight process to guide submitting new operator requests to WebNN webmachinelearning/webnn#231.
-
Make the `starts` and `sizes` definitions of `slice` op consistent and simplified webmachinelearning/webnn#369.
-
Need to update the spec links to latest published version webmachinelearning/webnn#371.
-
-
13 issues received 25 new comments.
-
Why is ml a member of Navigator? webmachinelearning/webnn#382 (5 comments by @anssiko, @bathos, @zolkis)
-
Make the `starts` and `sizes` definitions of `slice` op consistent and simplified webmachinelearning/webnn#369 (3 comments by @fdwr, @huningxin)
-
heads up re: proposal to add Float16Array to JavaScript webmachinelearning/webnn#373 (2 comments by @fdwr, @wchao1115)
-
Create a light-weight process to guide submitting new operator requests to WebNN webmachinelearning/webnn#231 (2 comments by @anssiko, @huningxin)
-
Specify the operand data type constraints of operation webmachinelearning/webnn#283 (2 comments by @huningxin)
-
MLOperand::dimensions() or MLGraphBuilder::shape() for debugging purposes and diagnostics webmachinelearning/webnn#379 (2 comments by @fdwr, @huningxin)
-
Clarify the usage of 32 bit floating point type and consider using double webmachinelearning/webnn#325 (1 comment by @wchao1115)
-
Define the algorithm of calculating the effective padding for "same-upper" and "same-lower" option webmachinelearning/webnn#326 (1 comment by @wchao1115)
-
Need clarify constraint of 'beginningPadding' and 'endingPadding' for pad operation in "reflection" and "symmetric" mode webmachinelearning/webnn#377 (1 comment by @fdwr)
-
Need to define the algorithm for the calculation of the output shape of Resample2d webmachinelearning/webnn#360 (1 comment by @fdwr)
-
Unsqueeze operator is missing (so add unsqueeze, or better yet, delete squeeze) webmachinelearning/webnn#296 (1 comment by @fdwr)
-
Simplify the operand layout support of conv2d and pooling 2d operations webmachinelearning/webnn#324 (1 comment by @wacky6)
-
-
3 pull requests opened.
-
Additional guidance for contributions webmachinelearning/webnn#381, by @wchao1115.
-
Simplify `slice` op webmachinelearning/webnn#376, by @huningxin.
-
Refer to the latest published version of the WebNN API webmachinelearning/webnn#372, by @ibelem.
-
-
4 pull requests merged.
-
Additional guidance for contributions webmachinelearning/webnn#381.
-
Simplify `slice` op webmachinelearning/webnn#376.
-
Update newShape with null value for reshape operation webmachinelearning/webnn#367.
-
Refer to the latest published version of the WebNN API webmachinelearning/webnn#372.
-
-
11 pull requests received 87 new comments.
-
Add the constant() method steps. webmachinelearning/webnn#365 (18 comments by @huningxin, @wchao1115, @zolkis)
-
Simplify `slice` op webmachinelearning/webnn#376 (17 comments by @anssiko, @BruceDai, @fdwr, @huningxin, @wacky6)
-
Fix #334: Improve the batch norm algorithm webmachinelearning/webnn#339 (11 comments by @wchao1115, @zolkis)
-
Add internal slots to MLOperand and MLActivation webmachinelearning/webnn#337 (11 comments by @wchao1115, @zolkis)
-
Additional guidance for contributions webmachinelearning/webnn#381 (9 comments by @anssiko, @huningxin, @wchao1115, @zolkis)
-
Rework the sync async algorithms based on PR 323 webmachinelearning/webnn#329 (7 comments by @huningxin, @wacky6, @wchao1115, @zolkis)
-
Update newShape with null value for reshape operation webmachinelearning/webnn#367 (4 comments by @anssiko, @dontcallmedom, @fdwr, @huningxin)
-
Refer to the latest published version of the WebNN API webmachinelearning/webnn#372 (3 comments by @anssiko, @ibelem)
-
TR publications
-
3 documents published.
-
Web Neural Network API published as a Candidate Recommendation Snapshot. History of Web Neural Network API
-
Web Neural Network API published twice as a Working Draft. History of Web Neural Network API
-
Past meetings
Participants
-
2 participants joined.
-
ChangSeok Oh (ByteDance) joined.
-
Pablo COCA (Fundacion CTIC) joined.
-
-
1 organization joined.
-
Fundacion CTIC joined.
-
GitHub
webmachinelearning/webnn
-
4 issues opened.
-
Make the `starts` and `sizes` definitions of `slice` op consistent and simplified webmachinelearning/webnn#369, by @huningxin.
-
The valid range of `output_padding` of MLConvTranspose2dOptions should be clarified webmachinelearning/webnn#368, by @miaobin.
-
Should we define IDL for MLActivation options as union of individual activation options? webmachinelearning/webnn#363, by @zolkis.
-
Need to define the algorithm for the calculation of the output shape of Resample2d webmachinelearning/webnn#360, by @lisa0314.
-
-
12 issues closed.
-
Should we define IDL for MLActivation options as union of individual activation options? webmachinelearning/webnn#363.
-
Propose to add Parametric ReLU into the operation list webmachinelearning/webnn#356.
-
Use static padding values for `pad` operation webmachinelearning/webnn#354.
-
Use unsigned long for axis of concat operation webmachinelearning/webnn#345.
-
[op compatibility] matMul webmachinelearning/webnn#27.
-
[op compatibility] conv2d webmachinelearning/webnn#28.
-
Define the set of operations and their specification webmachinelearning/webnn#17.
-
Handling unsupported OperandType webmachinelearning/webnn#36.
-
Include content filtering as use case webmachinelearning/webnn#236.
-
Self-Review Questionnaire: Security and Privacy webmachinelearning/webnn#119.
-
Define set of exceptions for WebNN API webmachinelearning/webnn#19.
-
Support CPU - WebAssembly scenario of the op level execution use case webmachinelearning/webnn#156.
-
-
21 issues received 43 new comments.
-
Use unsigned long for axis of concat operation webmachinelearning/webnn#345 (9 comments by @fdwr, @huningxin, @wacky6, @wchao1115)
-
Use static padding values for `pad` operation webmachinelearning/webnn#354 (3 comments by @fdwr, @huningxin, @zolkis)
-
Simplify the operand layout support of conv2d and pooling 2d operations webmachinelearning/webnn#324 (3 comments by @anssiko, @fdwr)
-
Please clarify interpolation algorithm for resample2d webmachinelearning/webnn#358 (3 comments by @anssiko, @BruceDai, @fdwr)
-
Clarify intermediate values (activations) vs MLActivation vs operations webmachinelearning/webnn#335 (3 comments by @huningxin, @zolkis)
-
Integration with real-time video processing webmachinelearning/webnn#226 (2 comments by @anssiko, @dontcallmedom)
-
Should we define IDL for MLActivation options as union of individual activation options? webmachinelearning/webnn#363 (2 comments by @zolkis)
-
Clarify the usage of 32 bit floating point type and consider using double webmachinelearning/webnn#325 (1 comment by @wchao1115)
-
Make the `starts` and `sizes` definitions of `slice` op consistent and simplified webmachinelearning/webnn#369 (1 comment by @fdwr)
-
webmachinelearning/webnn#27 (1 comment by @dontcallmedom)
-
webmachinelearning/webnn#28 (1 comment by @dontcallmedom)
-
webmachinelearning/webnn#119 (1 comment by @anssiko)
-
webmachinelearning/webnn#6 (1 comment by @anssiko)
-
webmachinelearning/webnn#19 (1 comment by @anssiko)
-
webmachinelearning/webnn#156 (1 comment by @anssiko)
-
-
7 pull requests opened.
-
Update newShape with null value for reshape operation webmachinelearning/webnn#367, by @BruceDai.
-
Add the concat algorithm webmachinelearning/webnn#366, by @zolkis.
-
Add the constant() method steps. webmachinelearning/webnn#365, by @zolkis.
-
Add the builder.input() method steps webmachinelearning/webnn#364, by @zolkis.
-
Update arch diagram to remove WebGL webmachinelearning/webnn#362, by @dontcallmedom.
-
update pad example webmachinelearning/webnn#361, by @BruceDai.
-
Use unsigned integer axis webmachinelearning/webnn#359, by @huningxin.
-
-
6 pull requests merged.
-
Add Status of this document note for Candidate Recommendation webmachinelearning/webnn#340.
-
Update arch diagram to remove WebGL webmachinelearning/webnn#362.
-
update pad example webmachinelearning/webnn#361.
-
Add prelu (Parametric ReLU) into the operation list webmachinelearning/webnn#357.
-
Use static padding values for `pad` operation webmachinelearning/webnn#355.
-
Use unsigned integer axis webmachinelearning/webnn#359.
-
-
1 pull request closed.
-
[WIP] Make axis definition of concat and split consistent webmachinelearning/webnn#352.
-
-
12 pull requests received 149 new comments.
-
Add the constant() method steps. webmachinelearning/webnn#365 (57 comments by @dontcallmedom, @huningxin, @zolkis)
-
Add internal slots to MLOperand and MLActivation webmachinelearning/webnn#337 (46 comments by @anssiko, @dontcallmedom, @huningxin, @zolkis)
-
Add Status of this document note for Candidate Recommendation webmachinelearning/webnn#340 (9 comments by @anssiko, @dontcallmedom)
-
Update arch diagram to remove WebGL webmachinelearning/webnn#362 (6 comments by @anssiko, @dontcallmedom, @huningxin)
-
Rework the sync async algorithms based on PR 323 webmachinelearning/webnn#329 (4 comments by @huningxin, @zolkis)
-
Add prelu (Parametric ReLU) into the operation list webmachinelearning/webnn#357 (4 comments by @huningxin, @wchao1115)
-
Update newShape with null value for reshape operation webmachinelearning/webnn#367 (3 comments by @anssiko, @BruceDai)
-
Use static padding values for `pad` operation webmachinelearning/webnn#355 (3 comments by @huningxin, @wchao1115)
-
[WIP] Make axis definition of concat and split consistent webmachinelearning/webnn#352 (2 comments by @huningxin)
-
Past meetings
Participants
-
1 participant joined.
-
Matthew Soulanille (Google LLC) joined.
-
GitHub
webmachinelearning/webnn
-
16 issues opened.
-
Please clarify interpolation algorithm for resample2d webmachinelearning/webnn#358, by @BruceDai.
-
Propose to add Parametric ReLU into the operation list webmachinelearning/webnn#356, by @huningxin.
-
Use static padding values for `pad` operation webmachinelearning/webnn#354, by @huningxin.
-
Support `depth_multiplier > 1` for a depthwise conv2d op webmachinelearning/webnn#353, by @Honry.
-
Need to define error handling of MLNamedArrayBufferViews transfer algorithm webmachinelearning/webnn#351, by @huningxin.
-
Need to understand how WebNN supports implementation that involves multiple devices and timelines webmachinelearning/webnn#350, by @wchao1115.
-
Add the clamp() algorithm webmachinelearning/webnn#347, by @zolkis.
-
WebNN and StableHLO opset compatibility webmachinelearning/webnn#346, by @anssiko.
-
Use unsigned long for axis of concat operation webmachinelearning/webnn#345, by @miaobin.
-
Subclass MLGraph based on the context that creates it webmachinelearning/webnn#344, by @huningxin.
-
Add sample code for WebGPU interop webmachinelearning/webnn#343, by @huningxin.
-
Clarify where GPU related states are stored for WebGPU interop (MLContext, MLCommandEncoder and MLGraph) webmachinelearning/webnn#342, by @huningxin.
-
Should validate MLGraph.[[context]] in MLContext.compute() and MLContext.computeSync() steps webmachinelearning/webnn#341, by @huningxin.
-
WPT tests tracker / 2 for remaining ops webmachinelearning/webnn#338, by @BruceDai.
-
Add internal slots to MLOperand, MLActivation and basic algorithms webmachinelearning/webnn#336, by @zolkis.
-
Clarify intermediate values (activations) vs MLActivation vs operations webmachinelearning/webnn#335, by @zolkis.
-
-
1 issue closed.
-
WPT tests tracker webmachinelearning/webnn#265.
-
-
12 issues received 21 new comments.
-
Should validate MLGraph.[[context]] in MLContext.compute() and MLContext.computeSync() steps webmachinelearning/webnn#341 (5 comments by @anssiko, @huningxin, @wacky6)
-
Use static padding values for `pad` operation webmachinelearning/webnn#354 (3 comments by @huningxin, @wacky6, @zolkis)
-
Use unsigned long for axis of concat operation webmachinelearning/webnn#345 (2 comments by @BruceDai, @huningxin)
-
WPT tests tracker / 2 for remaining ops webmachinelearning/webnn#338 (2 comments by @BruceDai, @fdwr)
-
Please clarify interpolation algorithm for resample2d webmachinelearning/webnn#358 (1 comment by @BruceDai)
-
Support `depth_multiplier > 1` for a depthwise conv2d op webmachinelearning/webnn#353 (1 comment by @Honry)
-
-
8 pull requests opened.
-
Add prelu (Parametric ReLU) into the operation list webmachinelearning/webnn#357, by @huningxin.
-
Use static padding values for `pad` operation webmachinelearning/webnn#355, by @huningxin.
-
[WIP] Make axis definition of concat and split consistent webmachinelearning/webnn#352, by @huningxin.
-
Fix a minor error in Example 7 webmachinelearning/webnn#349, by @Honry.
-
Add the clamp() algorithm webmachinelearning/webnn#348, by @zolkis.
-
Add Status of this document note for Candidate Recommendation webmachinelearning/webnn#340, by @anssiko.
-
Fix #334: Improve the batch norm algorithm webmachinelearning/webnn#339, by @zolkis.
-
Add internal slots to MLOperand and MLActivation webmachinelearning/webnn#337, by @zolkis.
-
-
1 pull request merged.
-
Fix a minor error in Example 7 webmachinelearning/webnn#349.
-
-
8 pull requests received 32 new comments.
-
Simplify MLContext creation webmachinelearning/webnn#322 (11 comments by @anssiko, @huningxin, @pyu10055, @wchao1115)
-
Add Status of this document note for Candidate Recommendation webmachinelearning/webnn#340 (8 comments by @anssiko, @dontcallmedom, @wchao1115)
-
Add internal slots to MLOperand and MLActivation webmachinelearning/webnn#337 (4 comments by @huningxin, @zolkis)
-
Rework the sync async algorithms based on PR 323 webmachinelearning/webnn#329 (1 comment by @zolkis)
-
TR publications
-
4 documents published.
-
Web Neural Network API published 4 times as a Working Draft. History of Web Neural Network API
-
Past meetings
-
1 meeting.
GitHub
webmachinelearning/webnn
-
6 issues opened.
-
Improve batchNorm webmachinelearning/webnn#334, by @zolkis.
-
Clarify/simplify graph execution on GPU / MLCommandEncoder webmachinelearning/webnn#333, by @zolkis.
-
Add an informative note for MLActivation that explains the performance benefit webmachinelearning/webnn#327, by @huningxin.
-
Define the algorithm of calculating the effective padding for "same-upper" and "same-lower" option webmachinelearning/webnn#326, by @huningxin.
-
Clarify the usage of 32 bit floating point type and consider using double webmachinelearning/webnn#325, by @huningxin.
-
Simplify the operand layout support of conv2d and pooling 2d operations webmachinelearning/webnn#324, by @huningxin.
-
-
2 issues closed.
-
Add an informative note for MLActivation that explains the performance benefit webmachinelearning/webnn#327.
-
The input and output resources race condition issue of asynchronous execution webmachinelearning/webnn#318.
-
-
5 issues received 7 new comments.
-
API simplification: context types, context options, createContext() webmachinelearning/webnn#302 (3 comments by @wchao1115, @zolkis)
-
Clarify/simplify graph execution on GPU / MLCommandEncoder webmachinelearning/webnn#333 (1 comment by @anssiko)
-
Conv2d and BatchNorm should support optional fused activation function webmachinelearning/webnn#185 (1 comment by @huningxin)
-
-
6 pull requests opened.
-
Fix #308: refer to MLContext validation steps from MLGraphBuilder webmachinelearning/webnn#332, by @zolkis.
-
Note fused activations can improve performance webmachinelearning/webnn#331, by @anssiko.
-
Move the MLGraph and MLCommandEncoder sections up for cleaner read webmachinelearning/webnn#330, by @zolkis.
-
Rework the sync async algorithms based on PR 323 webmachinelearning/webnn#329, by @zolkis.
-
Optimized compute method of effective filter size for conv2d and convTranspose2d. webmachinelearning/webnn#328, by @BruceDai.
-
Transfer the input and output views for asynchronous execution webmachinelearning/webnn#323, by @huningxin.
-
-
6 pull requests merged.
-
Fix #308: refer to MLContext validation steps from MLGraphBuilder webmachinelearning/webnn#332.
-
Note fused activations can improve performance webmachinelearning/webnn#331.
-
Optimized compute method of effective filter size for conv2d and convTranspose2d. webmachinelearning/webnn#328.
-
Move the MLGraph and MLCommandEncoder sections up for cleaner read webmachinelearning/webnn#330.
-
Transfer the input and output views for asynchronous execution webmachinelearning/webnn#323.
-
Add LSTM to the operator list webmachinelearning/webnn#321.
-
-
1 pull request closed.
-
WiP: Improve graph execution steps webmachinelearning/webnn#319.
-
-
9 pull requests received 90 new comments.
-
Simplify MLContext creation webmachinelearning/webnn#322 (30 comments by @anssiko, @huningxin, @wacky6, @wchao1115)
-
Transfer the input and output views for asynchronous execution webmachinelearning/webnn#323 (22 comments by @anssiko, @domenic, @huningxin, @wchao1115, @zolkis)
-
Add LSTM to the operator list webmachinelearning/webnn#321 (21 comments by @anssiko, @fdwr, @huningxin, @wchao1115)
-
Rework the sync async algorithms based on PR 323 webmachinelearning/webnn#329 (4 comments by @zolkis)
-
Fix #308: refer to MLContext validation steps from MLGraphBuilder webmachinelearning/webnn#332 (3 comments by @anssiko, @zolkis)
-
Optimized compute method of effective filter size for conv2d and convTranspose2d. webmachinelearning/webnn#328 (3 comments by @anssiko, @fdwr, @w3cbot)
-
Move the MLGraph and MLCommandEncoder sections up for cleaner read webmachinelearning/webnn#330 (3 comments by @anssiko, @zolkis)
-