TFJS 4.4.0

Core (4.3.0 ==> 4.4.0)

Misc

• Update monorepo to 4.4.0. (#7549).
• Support string flag (#7509). Thanks, @axinging.
• webgpu: upgrade webgpu/types (#7528). Thanks, @qjia7.
• Fix missing isTypedArray when mixing versions of @tensorflow packages (#7489).

Data (4.3.0 ==> 4.4.0)

Misc

• Update monorepo to 4.4.0. (#7549).

Layers (4.3.0 ==> 4.4.0)

Misc

• Update monorepo to 4.4.0. (#7549).
• RandomHeight Preprocessing Layer (#7483). Thanks, @RWallie.
• BaseRandomLayer Abstract Layer and RandomWidth Preprocessing Layer (#7345). Thanks, @RWallie.

Converter (4.3.0 ==> 4.4.0)

Misc

• Update monorepo to 4.4.0. (#7549).
• Fix graph executor to not dispose output tensors (#7505). Thanks, @chunnienc.

Node (4.3.0 ==> 4.4.0)

Bug fixes

• [tfjs-node] fixed summary writer memory leak (#7490).

Misc

• Update monorepo to 4.4.0. (#7549).
• Fix missing isTypedArray when mixing versions of @tensorflow packages (#7489).
• [tfjs-node] Encode jpeg (#7484).

Wasm (4.3.0 ==> 4.4.0)

Security

• build(deps-dev): bump webpack in /tfjs-backend-wasm/starter/webpack (#7476). Thanks, @dependabot[bot].

Misc

• Update monorepo to 4.4.0. (#7549).

Cpu (4.3.0 ==> 4.4.0)

Misc

• Update monorepo to 4.4.0. (#7549).
• [cpu] Cast pixel to number in maxpool utils (#7534).

Webgl (4.3.0 ==> 4.4.0)

Misc

• Update monorepo to 4.4.0. (#7549).

TFJS 4.3.0

Update 2023-03-21: This Release is Deprecated

This release contains a regression that causes some graph models to dispose their tensors too early. See #7504 for details.

Core (4.2.0 ==> 4.3.0)

Features

• Update flags.ts (#7367). Thanks, @Linchenn.
• [op] added support for TensorScatterUpdate op (#7189).

Bug fixes

• Revert "Add a case for atanh when input is 1 " (#7325) (#6964). Thanks, @Linchenn.
• Fix eluGrad's error (#7251). Thanks, @xhcao.

Performance

• Improve assertAndGetBroadcastShape (#7413). Thanks, @Linchenn.

Documentation

• Update tensor.ts (#7366). Thanks, @Linchenn.

Misc

• Update monorepo to 4.3.0. (#7488).
• [wasm] Add Multinomial kernel (#7468). Thanks, @chunnienc.
• Fix gather signature in chained ops (#7465).
• disable offscreen canvas for safari, since gl fencing is not working yet for its technology preview (#7436).
• Propagate errors from indexedDB database.put (#7322).
• Add StaticRegexReplace Op (#7379).
• SharedArrayBuffer is not GPU data (#7344). Thanks, @axinging.
• Support MatrixBandPart and allow numLower to be a tensor (#7351). Thanks, @Linchenn.
• Add a case for atanh when input is 1 (#6964). Thanks, @xhcao.
• [tfjs-node] support einsum in tfjs node (#7321).
• fix comment in tile.ts (#7298). Thanks, @sonmh79.

Data (4.2.0 ==> 4.3.0)

Misc

• Update monorepo to 4.3.0. (#7488).
• build(deps): bump vega in /tfjs-data/demo/boston-housing (#7444). Thanks, @dependabot[bot].
• build(deps): bump vega-functions in /tfjs-data/demo/boston-housing (#7442). Thanks, @dependabot[bot].
• build(deps): bump json5 in /tfjs-data/demo/boston-housing (#7219). Thanks, @dependabot[bot].
• build(deps): bump fast-json-patch in /tfjs-data/demo/boston-housing (#7211). Thanks, @dependabot[bot].

Layers (4.2.0 ==> 4.3.0)

Misc

• Update monorepo to 4.3.0. (#7488).
• Typo fix (#7393). Thanks, @dmotz.
• #7376 Allow seeding the Orthogonal Initializer and use QR instead of gramSchmidt (#7377). Thanks, @cosminc98.

Converter (4.2.0 ==> 4.3.0)

Features

• [converter] add TensorScatterUpdate op to converter and executor (#7246).

Bug fixes

• allow weight only keras H5 file that lacks keras version info to be parsed without error (#7359).

Misc

• Update tfjs_4.3.0 lockfiles (#7492).
• Update monorepo to 4.3.0. (#7488).
• Never pass a NoOp value to an op (#7460).
• Fix topological sort for sync nodes execution order (#7431). Thanks, @chunnienc.
• Remove array.at (#7409). Thanks, @chunnienc.
• Add new tensor diposal algorithm for sync graph execution (#7392). Thanks, @chunnienc.
• Add cache for parseNodeName (#7391). Thanks, @chunnienc.
• Update rules_python to 0.18.1 (#7383).
• Add StaticRegexReplace Op (#7379).
• added converter support for inf and infinite ops (#7365).
• Support MatrixBandPart and allow numLower to be a tensor (#7351). Thanks, @Linchenn.
• Support tfjs converter options in JAX conversion (#7331).
• Update TypeScript to 4.9.4 (#7300).
• Debug converter python code with debugpy (#7283).
• Refactor python converter functions to not call each other (#7281).
• Rename bazel BUILD files (#7253). Thanks, @chunnienc.
• Make converter hashtable test not overly specific (#7249).
• build(deps): bump json5 from 2.1.0 to 2.2.3 in /tfjs-converter/demo (#7240). Thanks, @dependabot[bot].

Node (4.2.0 ==> 4.3.0)

Features

• [op] added support for TensorScatterUpdate op (#7189).

Misc

• Update tfjs_4.3.0 lockfiles (#7492).
• Update monorepo to 4.3.0. (#7488).
• [wasm] Add Multinomial kernel (#7468). Thanks, @chunnienc.
• Add StaticRegexReplace Op (#7379).
• [tfjs-node] support einsum in tfjs node (#7321).
• Update TypeScript to 4.9.4 (#7300).
• Improve npm publishing infra (#7287).
• build(deps): bump json5 from 2.2.0 to 2.2.3 in /tfjs-node-gpu (#7243). Thanks, @dependabot[bot].
• build(deps): bump json5 from 2.2.0 to 2.2.3 in /tfjs-node (#7242). Thanks, @dependabot[bot].
• Do not run docker interactively in tfjs-node publish (#7227).

Wasm (4.2.0 ==> 4.3.0)

Features

• [wasm] Add atan kernel (#7252). Thanks, @chunnienc.
• [op] added support for TensorScatterUpdate op (#7189).

Security

• build(deps): bump json5 in /tfjs-backend-wasm/starter/webpack (#7218). Thanks, @dependabot[bot].

Misc

• Update tfjs_4.3.0 lockfiles (#7492).
• Update monorepo to 4.3.0. (#7488).
• [wasm] Add ArgMin (#7306). Thanks, @chunnienc.
• [wasm] Add Conv3D kernels (#7466). Thanks, @chunnienc.
• Add unique (#7469). Thanks, @chunnienc.
• [wasm] Add Multinomial kernel (#7468). Thanks, @chunnienc.
• Add expm1 (#7467). Thanks, @chunnienc.
• [wasm] Add Bincount kernel (#7412). Thanks, @chunnienc.
• Add LinSpace (#7403). Thanks, @chunnienc.
• Fix license headers added in 2023 (#7353). Thanks, @chunnienc.
• [wasm] Add assorted unary and binary kernels (#7350). Thanks, @chunnienc.
• [wasm] Add ResizeBilinearGrad and ResizeNearestNeighborGrad kernels (#7346). Thanks, @chunnienc.
• Add Log1p (#7328). Thanks, @chunnienc.
• [wasm] Add LRN kernels (#7335). Thanks, @chunnienc.
• [wasm] Add Dilation2D backprop kernels (#7334). Thanks, @chunnienc.
• [wasm] Add Dilation2D (#7327). Thanks, @chunnienc.
• [wasm] Add sparseToDense kernel (#6528). Thanks, @kon72.
• [wasm] Add EluGrad (#7307). Thanks, @chunnienc.
• [wasm] Enable lowerBound and upperBound op tests (#7301). Thanks, @chunnienc.
• [wasm] Add AvgPool3DGrad and MaxPool3DGrad kernels (#7295). Thanks, @chunnienc.
• Add SearchSorted (#7299). Thanks, @chunnienc.
• [wasm] Add AvgPool3D and MaxPool3D kernels (#7294). Thanks, @chunnienc.
• [wasm] Add BroadcastArgs kernel (#7290). Thanks, @chunnienc.
• Upgrade global build param from C++11 to C++17 (#7268). Thanks, @chunnienc.
• [wasm] Add DenseBincount kernel (#7267). Thanks, @chunnienc.
• Add Diag kernel (#7277). Thanks, @chunnienc.
• [wasm] Add asin and asinh kernels ...

tfjs-v4.2.0

Core (4.1.0 ==> 4.2.0)

Bug fixes

• Fix avgPool3d (#7133). Thanks, @Linchenn.
• Add depth check for dilation2d (#7137). Thanks, @Linchenn.
• Fix flatten implementation on objects (#7059).
• Fix batch matching for batch mat mul (#7062). Thanks, @Linchenn.

Misc

• Update monorepo to 4.2.0. (#7226).
• Use node's util.types.isUint8Array etc for isTypedArray (#7181).
• Use static getters to get optimizer class names (#7168).
• Simplify how Optimizers are re-exported in train.ts (#7156).
• Register optimizers in a centralized location (#7157).
• Mark all calls to 'op()' as pure (#7155).
• Refactor type conversion for read back (#7044). Thanks, @axinging.
• fix tensor.print() for 0 shaped tensor (#7127).
• Switch from android 9 to 10 for tests (#7132).
• Fix the doc error of step operator (#7120). Thanks, @xhcao.
• doc: Fix diag() example to make is less confusing. (#7113). Thanks, @robhybrid.
• Cast to 'unknown' instead of '{}' in double assertions (#7116).
• [webgpu] Create tensor from GPUBuffer (#7034). Thanks, @axinging.
• Add positive dilation and strides check (#7063).

Data (4.1.0 ==> 4.2.0)

Misc

• Update monorepo to 4.2.0. (#7226).

Layers (4.1.0 ==> 4.2.0)

Misc

• Update monorepo to 4.2.0. (#7226).
• Switch from android 9 to 10 for tests (#7132).
• Fix #7104 - tf.initializers.<random | glorot | he | leCunn>Uniform() ignores seed argument & add tests that replicated the issue, fix wrong serialization name registered for LeCunUniform initializer class (#7108). Thanks, @adrian-branescu.
• Bump minimatch from 3.0.4 to 3.1.2 in /tfjs-layers (#7086). Thanks, @dependabot[bot].

Converter (4.1.0 ==> 4.2.0)

Misc

• Update lockfiles branch tfjs_4.2.0_lockfiles lock files. (#7230).
• Update monorepo to 4.2.0. (#7226).
• Update rules_python to 0.16.1 (#7160).
• Mark tfdf ops as supported by tfjs-converter (#7141).
• Change dependency from saved_model_cli to saved_model_utils (#7039). Thanks, @BlaziusMaximus.

Node (4.1.0 ==> 4.2.0)

Bug fixes

• Fix avgPool3d (#7133). Thanks, @Linchenn.

Misc

• Update lockfiles branch tfjs_4.2.0_lockfiles lock files. (#7230).
• Update monorepo to 4.2.0. (#7226).
• [tfjs-node] replace Exception with Error constructor (#7131).
• Cast to 'unknown' instead of '{}' in double assertions (#7116).
• Bump minimatch from 3.0.4 to 3.1.2 in /tfjs-node-gpu (#7085). Thanks, @dependabot[bot].
• Bump minimatch from 3.0.4 to 3.1.2 in /tfjs-node (#7084). Thanks, @dependabot[bot].

Wasm (4.1.0 ==> 4.2.0)

Bug fixes

• Fix batch matching for batch mat mul (#7062). Thanks, @Linchenn.

Performance

• Optimize BatchMatMul's for loop (#7069). Thanks, @Linchenn.

Misc

• Update lockfiles branch tfjs_4.2.0_lockfiles lock files. (#7230).
• Update monorepo to 4.2.0. (#7226).
• Switch from android 9 to 10 for tests (#7132).
• Cast to 'unknown' instead of '{}' in double assertions (#7116).
• Bump minimatch from 3.0.4 to 3.1.2 in /tfjs-backend-wasm (#7088). Thanks, @dependabot[bot].

Cpu (4.1.0 ==> 4.2.0)

Bug fixes

• Fix avgPool3d (#7133). Thanks, @Linchenn.
• Fix batch matching for batch mat mul (#7062). Thanks, @Linchenn.

Performance

• Optimize BatchMatMul's for loop (#7069). Thanks, @Linchenn.

Misc

• Update monorepo to 4.2.0. (#7226).
• Refactor type conversion for read back (#7044). Thanks, @axinging.
• Cast to 'unknown' instead of '{}' in double assertions (#7116).

Webgl (4.1.0 ==> 4.2.0)

Bug fixes

• Fix avgPool3d (#7133). Thanks, @Linchenn.
• Fix batch matching for batch mat mul (#7062). Thanks, @Linchenn.

Performance

• Optimize BatchMatMul's for loop (#7069). Thanks, @Linchenn.

Documentation

• webgl: fix a typing nit (#7199). Thanks, @xhcao.

Misc

• Update lockfiles branch tfjs_4.2.0_lockfiles lock files. (#7230).
• Update monorepo to 4.2.0. (#7226).
• Switch from android 9 to 10 for tests (#7132).
• Cast to 'unknown' instead of '{}' in double assertions (#7116).
• [webgpu] Create tensor from GPUBuffer (#7034). Thanks, @axinging.

tfjs-v4.1.0

Core (4.0.0 ==> 4.1.0)

Bug fixes

• [ops ] guard shape for decimal and negative values (#7014).

Misc

• Update monorepo to 4.1.0. (#7058).
• Move flag ENGINE_COMPILE_ONLY (#7019). Thanks, @Linchenn.
• Make Tensor explicitly implement TensorInfo (#7007).
• Turn on the noImplicitOverride tsconfig option (#7009).
• Cleanup and standardize package.json scripts of Bazel packages (#7005).
• Use a template literal type for kernel registry key (#6976).
• [wasm] Fix AvgPool and MaxPool for 1x1 kernels (#6969).

Data (4.0.0 ==> 4.1.0)

Misc

• Update monorepo to 4.1.0. (#7058).
• Turn on the noImplicitOverride tsconfig option (#7009).
• Cleanup and standardize package.json scripts of Bazel packages (#7005).

Layers (4.0.0 ==> 4.1.0)

Misc

• Update monorepo to 4.1.0. (#7058).
• Turn on the noImplicitOverride tsconfig option (#7009).
• Cleanup and standardize package.json scripts of Bazel packages (#7005).
• cleanup conv class definitions (#6947). Thanks, @vladmandic.
• Move fitLoop and fitTensors into the LayersModel class (#6859).
• Fix incorrect import path in center crop layer (#6946).
• CenterCrop Layer (#6875). Thanks, @AdamLang96.

Converter (4.0.0 ==> 4.1.0)

Misc

• Update lockfiles branch tfjs_4.1.0_lockfiles lock files. (#7068).
• Update monorepo to 4.1.0. (#7058).
• Support dump with graphmodel.execute (#6953). Thanks, @axinging.
• Cleanup and standardize package.json scripts of Bazel packages (#7005).
• Implement InitializeTableV2 (#6985).
• Add support for RaggedGather/Range/TensorToTensor conversion (#6968).
• Add null check for signature inputs outputs (#6978).
• Add predictAsync function to support async structured outputs (#6975).
• Use assertIn instead of assertTrue in python tests (#6958).

Node (4.0.0 ==> 4.1.0)

Misc

• Update lockfiles branch tfjs_4.1.0_lockfiles lock files. (#7068).
• Update monorepo to 4.1.0. (#7058).

Wasm (4.0.0 ==> 4.1.0)

Misc

• Update lockfiles branch tfjs_4.1.0_lockfiles lock files. (#7068).
• Update monorepo to 4.1.0. (#7058).
• Bump loader-utils in /tfjs-backend-wasm/starter/webpack (#7052). Thanks, @dependabot[bot].
• Bump loader-utils in /tfjs-backend-wasm/starter/webpack (#7023). Thanks, @dependabot[bot].
• Turn on the noImplicitOverride tsconfig option (#7009).
• Cleanup and standardize package.json scripts of Bazel packages (#7005).
• implement isNaN for wasm (#6998).
• added reciprocal op support for wasm (#6992).
• [wasm] Fix AvgPool and MaxPool for 1x1 kernels (#6969).

Cpu (4.0.0 ==> 4.1.0)

Misc

• Update monorepo to 4.1.0. (#7058).
• Turn on the noImplicitOverride tsconfig option (#7009).
• Cleanup and standardize package.json scripts of Bazel packages (#7005).

Webgl (4.0.0 ==> 4.1.0)

Bug fixes

• Use VAOs for save+restore of vertexAttribPointer state between different webgl programs. (#6913). Thanks, @kdashg.
• fix (#6930). Thanks, @Linchenn.

Misc

• Update lockfiles branch tfjs_4.1.0_lockfiles lock files. (#7068).
• Update monorepo to 4.1.0. (#7058).
• [webgl] Donot release tensor texture at reading (#6932). Thanks, @axinging.
• Remove @types/webgl2 since they are included in lib/dom (#7038).
• Move flag ENGINE_COMPILE_ONLY (#7019). Thanks, @Linchenn.
• Turn on the noImplicitOverride tsconfig option (#7009).
• Cleanup and standardize package.json scripts of Bazel packages (#7005).

tfjs-v4.0.0

This release upgrades TypeScript to 4.8.4 and @webgpu/types to 0.1.21. If you don't use TypeScript, you can update to 4.0 without reading this.

Breaking changes

This release is breaking for projects that use typescript<4.4. All other projects should be unaffected.

Read our guide to upgrading to 4.0 here.

Core (3.21.0 ==> 4.0.0)

Features

• [WebGL] Texture to tensor API (#6853). Thanks, @Linchenn.

Bug fixes

• #6922 [Fix setTimeoutCustom in worker thread] (#6923). Thanks, @gyagp.

Misc

• Update monorepo to 4.0.0. (#6929).
• Upgrade tfjs repo to 4.0.0 (#6927).
• Add kernel RaggedRange for CPU and WebGL backend (#6871).
• Upstream google3 changes (#6912).
• Add resource initializer support (#6904).

Data (3.21.0 ==> 4.0.0)

Misc

• Update monorepo to 4.0.0. (#6929).
• Upgrade tfjs repo to 4.0.0 (#6927).

Layers (3.21.0 ==> 4.0.0)

Misc

• Update monorepo to 4.0.0. (#6929).
• Upgrade tfjs repo to 4.0.0 (#6927).
• Resizing Layer (#6879). Thanks, @koyykdy.
• [layers] Consolidate loadLayersModel and loadLayersModelInternal (#6874).
• Category Encoding Layer (#6855). Thanks, @AdamLang96.

Converter (3.21.0 ==> 4.0.0)

Misc

• Update lockfiles branch tfjs_4.0.0_lockfiles lock files. (#6935).
• Update monorepo to 4.0.0. (#6929).
• Upgrade tfjs repo to 4.0.0 (#6927).
• Add asset zipping functionality to TFJS converter (#6915).
• Add resource initializer support (#6904).

Node (3.21.0 ==> 4.0.0)

Misc

• Update lockfiles branch tfjs_4.0.0_lockfiles lock files. (#6935).
• Update monorepo to 4.0.0. (#6929).
• Upgrade tfjs repo to 4.0.0 (#6927).
• Build tfjs-node node addon with the release docker (#6925).
• Add kernel RaggedRange for CPU and WebGL backend (#6871).

Wasm (3.21.0 ==> 4.0.0)

Misc

• Update lockfiles branch tfjs_4.0.0_lockfiles lock files. (#6935).
• Update monorepo to 4.0.0. (#6929).
• Add resource initializer support (#6904).

Cpu (3.21.0 ==> 4.0.0)

Misc

• Update monorepo to 4.0.0. (#6929).
• Add kernel RaggedRange for CPU and WebGL backend (#6871).

Webgl (3.21.0 ==> 4.0.0)

Features

• [WebGL] Texture to tensor API (#6853). Thanks, @Linchenn.

Misc

• Update lockfiles branch tfjs_4.0.0_lockfiles lock files. (#6935).
• Update monorepo to 4.0.0. (#6929).
• Upgrade tfjs repo to 4.0.0 (#6927).
• Add kernel RaggedRange for CPU and WebGL backend (#6871).

tfjs-v3.21.0

Core (3.20.0 ==> 3.21.0)

Bug fixes

• : #6687 [Customize setTimeout] (#6694). Thanks, @gyagp.
• Fix TFJS TF dissrepancy in concat errors (#6824).

Documentation

• fix(docs): wrong selu definition (#5489). Thanks, @raffizulvian.

Misc

• Update monorepo to 3.21.0. (#6905).
• Revert "Add resource initializer support " (#6900) (#6826).
• [e2e] Fix incorrect reference to link-package (#6895).
• Add resource initializer support (#6826).
• Support loading models from ModelArtifacts and ModelJSON in loadGraphModelSync (#6847).
• Update jasmine_util.ts (#6872). Thanks, @Linchenn.
• webgpu: Expand DepthwiseConv2DVec4Program to support any stride (#6820). Thanks, @qjia7.
• Fix clamp bug when min = max (#6825).
• Add kernel RaggedGather for CPU and WebGL backend (#6765).
• Fix mistakes (#6793). Thanks, @ivan-pan.

Data (3.20.0 ==> 3.21.0)

Misc

• Update monorepo to 3.21.0. (#6905).
• Fix mistakes (#6793). Thanks, @ivan-pan.

Layers (3.20.0 ==> 3.21.0)

Misc

• Update monorepo to 3.21.0. (#6905).
• Use a predetermined seed for tfjs-layers tests (#6878).
• Avoid testing webgl1 in webgl2 layers test (#6849).
• Rescaling Preprocessing Layer (#6840). Thanks, @AdamLang96.
• Fix mistakes (#6793). Thanks, @ivan-pan.

Converter (3.20.0 ==> 3.21.0)

Misc

• Update lockfiles branch tfjs_3.21.0_lockfiles lock files. (#6906).
• Update monorepo to 3.21.0. (#6905).
• Revert "Add resource initializer support " (#6900) (#6826).
• Add resource initializer support (#6826).
• Support loading models from ModelArtifacts and ModelJSON in loadGraphModelSync (#6847).
• fixed prelu fusing code that pre-maturely neg the const on multiply (#6876).
• Upgrade nodejs to 18.7.0 (#6863).
• Fix kernels_to_ops generating in the wrong location (#6823).
• Fix mistakes (#6793). Thanks, @ivan-pan.
• adding missing license header (#6792).

Node (3.20.0 ==> 3.21.0)

Misc

• Update lockfiles branch tfjs_3.21.0_lockfiles lock files. (#6906).
• Update monorepo to 3.21.0. (#6905).
• [e2e] Fix incorrect reference to link-package (#6895).
• Run Verdaccio tests in nightly (#6660).
• Add kernel RaggedGather for CPU and WebGL backend (#6765).
• [tfjs-node]: upgrade TF to 2.9.1 (#6786). Thanks, @rthadur.

Wasm (3.20.0 ==> 3.21.0)

Bug fixes

• Fix TFJS TF dissrepancy in concat errors (#6824).

Misc

• Update lockfiles branch tfjs_3.21.0_lockfiles lock files. (#6906).
• Update monorepo to 3.21.0. (#6905).
• [e2e] Fix incorrect reference to link-package (#6895).
• add exception handling to wasm feature detection (#6753). Thanks, @vladmandic.

Cpu (3.20.0 ==> 3.21.0)

Bug fixes

• Fix TFJS TF dissrepancy in concat errors (#6824).

Misc

• Update monorepo to 3.21.0. (#6905).
• Add kernel RaggedGather for CPU and WebGL backend (#6765).

Webgl (3.20.0 ==> 3.21.0)

Features

• [WebGL backend] Add max 1D texture dimension flag (#6808). Thanks, @Linchenn.

Bug fixes

• : #6687 [Customize setTimeout] (#6694). Thanks, @gyagp.
• Fix TFJS TF dissrepancy in concat errors (#6824).

Misc

• Update lockfiles branch tfjs_3.21.0_lockfiles lock files. (#6906).
• Update monorepo to 3.21.0. (#6905).
• webgl: Fix NaN issue (#6828). Thanks, @qjia7.
• Fix gatherND for 5D inputs (#6832). Thanks, @Linchenn.
• Add kernel RaggedGather for CPU and WebGL backend (#6765).
• shape uniform (#6809). Thanks, @Linchenn.

tfjs-tflite-v0.0.1-alpha.9

Bumped the underlying tflite version to newest (2.10.0)

tfjs-v3.20.0

What's Changed

• [Jasmine] Add expectation to tests in jasmine test framework by @haoyunfeix in #6656
• Set willReadFrequently to true for GPU correctness test by @gyagp in #6665
• [WebGL] Implement packed NHWC Conv2d by @Linchenn in #6639
• [WebGL] Fix NHWC packed depthwise conv2d for dilation=3 by @Linchenn in #6662
• webgpu: dont use mod as it is a reserved keyword by @crowlKats in #6667
• [converter] updated the dev readme for converter to cover python tests by @pyu10055 in #6675
• [core] Fix undefined expect in test_util by @haoyunfeix in #6671
• [WebGL] Fix the test for Depthwise with dilation=3 by @Linchenn in #6679
• webgpu: fix notEqual error by @xhcao in #6669
• [local-benchmark tool] Enable populate modelUrl and input shapes via URL by @Linchenn in #6680
• Add iOS and Android to WebGL presubmit tests by @mattsoulanille in #6681
• [webgpu] Remove extra device stuff passing to the constructor by @haoyunfeix in #6689
• Fix browserstack-benchmark tool by @Linchenn in #6703
• [webgpu] Set WEBGPU_IMPORT_EXTERNAL_TEXTURE to true by default by @axinging in #6706
• [webgpu] Fix alignment missing for 5D and 6D by @haoyunfeix in #6701
• [browserstack benchmark tool] Enable customizing local builds by @Linchenn in #6705
• Use the firebase-admin package instead of firebase to push benchmark results by @mattsoulanille in #6711
• [browserstack benchmark tool] Enable code snippet benchmark by @Linchenn in #6704
• [browserstack benchmark tool] Update browser list with new devices by @Linchenn in #6713
• Make link-package's build-deps-for script check transitive deps by @mattsoulanille in #6712
• Add structured outputs support to tfjs-inference by @ahmedsabie in #6696
• [browserstack benchmark tool] fix build script by @Linchenn in #6716
• Run the next BrowserStack benchmark immediately instead of in batches by @mattsoulanille in #6717
• Add kernel RaggedTensorToTensor for CPU and WebGL backend by @ahmedsabie in #6686
• Fix tf.image.transform on CPU and WASM by @andrusza2 in #6658
• [automl] Use the same karma timeout settings as Bazel by @mattsoulanille in #6719
• De-flake fromPixels test failure by @hujiajie in #6626
• [tfjs-converter] Revert "fix protobuf version to 3.20.0 (#6612)" by @pyu10055 in #6721
• [browserstack benchmark tool] End firebase instance only if it exists by @Linchenn in #6727
• Setup optimization target deices/browsers by @Linchenn in #6722
• Remove fromPixelsAsync test cases by @gyagp in #6723
• webgpu: Skip raggedTensorToTensor test cases by @gyagp in #6732
• webgpu: export webgpu utilities by @xhcao in #6707
• [benchmark tools] Update mobilenet models by @Linchenn in #6729
• [benchmark tools] Update model lists for benchmarking by @Linchenn in #6730
• [tfjs-converter] fixed protobuf version mismatch error on cloudbuild by @pyu10055 in #6737
• [local benchmark tool] Add convertedBy filed by @Linchenn in #6745
• Add use_structured_outputs flag to README by @ahmedsabie in #6739
• [browserstack benchmark tool] Add a html to present benchmark results by @Linchenn in #6740
• [browserstack benchmark tool] Move CodeSnippet from backend to frontend by @Linchenn in #6738
• webgpu: Merge MatMulPackedProgram and MatMulPackedVec4Program by @qjia7 in #6688
• [browserstack benchmark tool] batch benchmark code snippet by @Linchenn in #6742
• [webgpu] Add Atan2, IsNaN, Reciprocal by @axinging in #6743
• [tfjs-converter] remove kept flag after tensor is replaced or popped from the tensor list by @pyu10055 in #6759
• Strengthen frame readiness while testing fromPixels with video element by @hujiajie in #6751
• webgpu: enlarge the splitted dimInner size by @qjia7 in #6755
• [webgpu] s/ScatterOptimizedProgram/ScatterProgram/g by @hujiajie in #6761
• Adds JAX-->TFjs converter by @marcvanzee in #6744
• [browserstack benchmark tool] Use stream to write to outfile by @Linchenn in #6750
• [tfjs-converter ]include jax_conversion target to fix nightly test by @pyu10055 in #6767
• [WebGL backend] Add a flag to enable Software WebGL by @Linchenn in #6772
• [webgpu] Tweak WGSL entry point emission so '{' isn't visually swallowed by @hujiajie in #6764
• Add cpu forward for Cast by @qjia7 in #6760
• webgpu: Optimize AvgPool when filter size = input size by @qjia7 in #6762
• [op] allow dtype secification for onehot op by @pyu10055 in #6781
• webgpu: Add non-shared argminmax program by @qjia7 in #6778
• [converter] added dtype support for oneHot for converter by @pyu10055 in #6782
• Update monorepo to 3.20.0. by @pyu10055 in #6788
• Update lockfiles branch tfjs_3.20.0_lockfiles lock files. by @pyu10055 in #6789

New Contributors

• @andrusza2 made their first contribution in #6658
• @marcvanzee made their first contribution in #6744

Full Changelog: tfjs-v3.19.0...tfjs-v3.20.0

tfjs-v3.19.0

Core (3.18.0 ==> 3.19.0)

Performance

• Enable WebGL's conv2dWithIm2Row to process multiple batches (#6442). Thanks, @Linchenn.

Misc

• Update lockfiles branch tfjs_3.19.0_lockfiles lock files. (#6666).
• Update monorepo to 3.19.0. (#6659).
• Pin webgpu types to an exact version (#6640).
• Remove unused files (#6586).
• Add string support to squeeze type (#6617).
• [webgpu] Support convTranspose vec4 (#6603). Thanks, @axinging.
• Fix iOS and macOS Safari CI Failures (#6608).
• [core] Remove accidental console.log (#6587).
• Update Jasmine from 3.1.0 to 4.2.1 (#6520).
• Add an experimental remote build execution config based on Angular's (#6576).
• Convert CRLF to LF (#6574).
• Bump seedrandom to avoid unsafe-eval issues (#6440). Thanks, @s-mustafa.
• Upgrade @bazel/bazelisk and @bazel/ibazel (#6565).
• Add a Bazel test_suite to organize tests (#6550).
• Add randomStandardNormal op (#6533). Thanks, @kinsei0916.
• [wasm] Build tfjs-backend-wasm entirely with Bazel (#6458).
• webgpu: Support depthwise conv2d with nchw format (#6084). Thanks, @qjia7.
• Fix kernel_registry test flakiness for async backend (#6461). Thanks, @hujiajie.
• Set willReadFrequently to true (#6513). Thanks, @axinging.
• [core]Fix compilation error. (#6516).
• Fixed nightly failures (#6488).
• Add hint willReadFrequently for canvas (#6445). Thanks, @axinging.
• Use a more simple type construction for IOHandlerSync (#6481).
• wasm backend add LogicalNot, LogicalOr, LogicalXor (#6431). Thanks, @homfen.
• [converter] Fix fusedMatMul bug. (#6455).
• Remove unneeded devDependencies from packages that build with Bazel (#6452).
• [secu] update package.json files to resolve security alerts (#6448).

Data (3.18.0 ==> 3.19.0)

Misc

• Update monorepo to 3.19.0. (#6659).
• Bump terser from 5.12.1 to 5.14.2 in /tfjs-data/demo/boston-housing (#6648). Thanks, @dependabot[bot].
• Remove unused files (#6586).
• Update Jasmine from 3.1.0 to 4.2.1 (#6520).
• Bump seedrandom to avoid unsafe-eval issues (#6440). Thanks, @s-mustafa.
• Upgrade @bazel/bazelisk and @bazel/ibazel (#6565).
• Add a Bazel test_suite to organize tests (#6550).
• Add string_decoder to tfjs-data (#6538).
• [tfjs-data] Fix test files list missing setup_tests.ts (#6482).
• Remove unneeded devDependencies from packages that build with Bazel (#6452).
• [secu] update package.json files to resolve security alerts (#6448).

Layers (3.18.0 ==> 3.19.0)

Misc

• Update monorepo to 3.19.0. (#6659).
• Remove unused files (#6586).
• Fix iOS and macOS Safari CI Failures (#6608).
• Update Jasmine from 3.1.0 to 4.2.1 (#6520).
• Convert CRLF to LF (#6574).
• Upgrade @bazel/bazelisk and @bazel/ibazel (#6565).
• Add a Bazel test_suite to organize tests (#6550).
• Remove unneeded devDependencies from packages that build with Bazel (#6452).
• [secu] update package.json files to resolve security alerts (#6448).

Converter (3.18.0 ==> 3.19.0)

Bug fixes

• remove python2_wheel target (#6597).
• Add '--bazel_patch_module_resolver' to 'tfjs-converter_test' (#6593).

Misc

• Update lockfiles branch tfjs_3.19.0_lockfiles lock files. (#6666).
• Update monorepo to 3.19.0. (#6659).
• Bump terser from 5.12.1 to 5.14.2 in /tfjs-converter/demo (#6642). Thanks, @dependabot[bot].
• Some numpy type usages are deprecated (#6625). Thanks, @rthadur.
• Add structured_outputs support in TFJS converter (#6559).
• fix protobuf version to 3.20.0 (#6612).
• Update Jasmine from 3.1.0 to 4.2.1 (#6520).
• Add an experimental remote build execution config based on Angular's (#6576).
• [converter] Remove unused files (#6583).
• Remove python2 from tfjs-converter (#6575).
• [converter] Avoid spying on properties of the 'tfOps' module (#6563).
• Upgrade @bazel/bazelisk and @bazel/ibazel (#6565).
• Add a Bazel test_suite to organize tests (#6550).
• Add randomStandardNormal op (#6533). Thanks, @kinsei0916.
• fix tensorlist reserve op (#6523).
• Bump protobufjs from 6.8.9 to 6.11.3 in /tfjs-converter (#6489). Thanks, @dependabot[bot].
• [converter] allow python converter to store intermediate frozen graph (#6456).
• [converter] Fix fusedMatMul bug. (#6455).
• Remove unneeded devDependencies from packages that build with Bazel (#6452).
• [secu] update package.json files to resolve security alerts (#6448).

Node (3.18.0 ==> 3.19.0)

Documentation

• add doc for tfjs-node on mac with m1 (#6630).

Misc

• Update lockfiles branch tfjs_3.19.0_lockfiles lock files. (#6666).
• Update monorepo to 3.19.0. (#6659).
• [webgpu] Support convTranspose vec4 (#6603). Thanks, @axinging.
• Update Jasmine from 3.1.0 to 4.2.1 (#6520).
• Support Linux arm64 in tfjs-node (#6564).
• [wasm] Build tfjs-backend-wasm entirely with Bazel (#6458).
• [node] Add broadcasting support for BatchMatMul (#6477). Thanks, @Linchenn.
• [secu] update package.json files to resolve security alerts (#6448).

Wasm (3.18.0 ==> 3.19.0)

Misc

• Update lockfiles branch tfjs_3.19.0_lockfiles lock files. (#6666).
• Update monorepo to 3.19.0. (#6659).
• Bump terser from 5.12.1 to 5.14.2 in /tfjs-backend-wasm/starter/webpack (#6652). Thanks, @dependabot[bot].
• Bump terser from 5.12.1 to 5.14.2 in /tfjs-backend-wasm/starter/parcel (#6643). Thanks, @dependabot[bot].
• [wasm] Add Object.fromEntries polyfill to tests to fix Safari 11 (#6621).
• Forward String op WASM implementations to CPU (#5715).
• [wasm] Increase beforeAll timeout that loads wasm files (#6618).
• [wasm] Cache wasm files when running Karma tests (#6613).
• Fix iOS and macOS Safari CI Failures (#6608).
• [wasm] Remove unused test (#6588).
• Update Jasmine from 3.1.0 to 4.2.1 ([#6...

tfjs-v3.18.0

Core (3.17.0 ==> 3.18.0)

Features

• Support NCHW dataFormat for fusedConv2d (Core) (#6373). Thanks, @Linchenn.

Bug fixes

• Fix CPU fusedConv2d (activation) for NCHW format (#6400). Thanks, @Linchenn.

Misc

• Update monorepo to 3.18.0. (#6433).
• [core] Restore makeTensorFromDataId (#6432).
• Implement transpose for complex64 type (#6410). Thanks, @DCtheTall.
• Implement loadGraphModelSync (#6428).
• Fix WebGL fusedConv2d's bias and PReLU computation (the third/BF branch) (#6423). Thanks, @Linchenn.
• [core] Add a synchronous I/O handler (#6417).
• webgpu: remove custom size parameter of dataToGPU (#6393). Thanks, @xhcao.

Data (3.17.0 ==> 3.18.0)

Misc

• Update lockfiles branch tfjs_3.18.0_lockfiles lock files. (#6434).
• Update monorepo to 3.18.0. (#6433).

Layers (3.17.0 ==> 3.18.0)

Misc

• Update lockfiles branch tfjs_3.18.0_lockfiles lock files. (#6434).
• Update monorepo to 3.18.0. (#6433).

Converter (3.17.0 ==> 3.18.0)

Misc

• Update lockfiles branch tfjs_3.18.0_lockfiles lock files. (#6434).
• Update monorepo to 3.18.0. (#6433).
• Implement loadGraphModelSync (#6428).
• [converter] Add missing cpu dependency to test_snippets (#6416).

Node (3.17.0 ==> 3.18.0)

Features

• Support NCHW dataFormat for fusedConv2d (Core) (#6373). Thanks, @Linchenn.

Bug fixes

• disable new tests (#6425). Thanks, @Linchenn.

Misc

• Update lockfiles branch tfjs_3.18.0_lockfiles lock files. (#6434).
• Update monorepo to 3.18.0. (#6433).

Wasm (3.17.0 ==> 3.18.0)

Misc

• Update lockfiles branch tfjs_3.18.0_lockfiles lock files. (#6434).
• Update monorepo to 3.18.0. (#6433).
• Implement transpose for complex64 type (#6410). Thanks, @DCtheTall.

Cpu (3.17.0 ==> 3.18.0)

Bug fixes

• Fix CPU fusedConv2d (activation) for NCHW format (#6400). Thanks, @Linchenn.

Misc

• Update lockfiles branch tfjs_3.18.0_lockfiles lock files. (#6434).
• Update monorepo to 3.18.0. (#6433).

Webgl (3.17.0 ==> 3.18.0)

Features

• Support NCHW dataFormat for fusedConv2d (Core) (#6373). Thanks, @Linchenn.

Bug fixes

• Fix CPU fusedConv2d (activation) for NCHW format (#6400). Thanks, @Linchenn.

Misc

• Update lockfiles branch tfjs_3.18.0_lockfiles lock files. (#6434).
• Update monorepo to 3.18.0. (#6433).
• Use parameter properties and avoid implicit construction (#6421).
• Fix WebGL fusedConv2d's bias and PReLU computation (the third/BF branch) (#6423). Thanks, @Linchenn.