tfjs-tflite-v0.0.1-alpha.7

tfjs-tflite (0.0.1-alpha.6 ==> 0.0.1-alpha.7)

This release compiled the underlying TFLite Web API WASM binaries with --xnn_wasmsimd_version=m91 which should increase the performance of certain quantized TFLite models.

• [tflite] update tflite web api version. (#5759).
• [tflite] convert output tensor types that are not supported by TFJS to compatible TFJS types (#5746).
• [tflite] Fix the broken demo (#5732).
• [tflite] Remove unneeded download step from readme (#5604).
• [tflite] Add README to npm package (#5581)..

tfjs-backend-webgpu-v0.0.1-alpha.8

tfjs-backend-webgpu (0.0.1-alpha.7 ==> 0.0.1-alpha.8)

Features

• [webgpu] Add support for scatterND (#5643). Thanks, @haoyunfeix.
• [webgpu] Add support for LogicalNot (#5644). Thanks, @haoyunfeix.
• [webgpu] Add support for split (#5645). Thanks, @haoyunfeix.
• [webgpu] Add support for sin, cos, sinh and cosh (#5642). Thanks, @haoyunfeix.
• [webgpu] Add depthToSpace op (#5585). Thanks, @haoyunfeix.
• [webgpu] Add halfPixelCenters support for resizeBilinear op (#5553). Thanks, @haoyunfeix.
• [webgpu] Enable bool and add op logicalAnd (#5190). Thanks, @axinging.

Bug fixes

• [webgpu] Fix wrongly used u32 (#5559). Thanks, @axinging.
• webgpu: Fix the memory leak for Cast (#5521). Thanks, @qjia7.
• Create a separate 'link-package-core' for packages that only need core (#5436).
• : [[webgpu] Add op tile] (#5186). Thanks, @axinging.

Performance

• webgpu: Optimize ScatterNd (#5667). Thanks, @qjia7.
• webgpu: Optimize depthwise conv2d (#5254). Thanks, @qjia7.
• webgpu: Optimize depthwise conv2d (#5209). Thanks, @qjia7.

Misc

• Update tfjs-backend-webgpu to 0.0.1-alpha.8. (#5681).
• [webgpu] Mark build-npm.sh as executable (#5676).
• [wasm] Fix kernels in wasm backend about how int32 inputs are handled (#5662).
• webgpu: Use a larger cpu forward threshold (#5665). Thanks, @qjia7.
• Use the release docker for all cloudbuild steps (#5669).
• [wgsl] Refine pow (#5622). Thanks, @axinging.
• webgpu: Batch several dispatching calls into one pass (#5598). Thanks, @qjia7.
• [webgpu] Fix matmul_small_output program assert error (#5646). Thanks, @haoyunfeix.
• webgpu: Remove glsl code (#5654). Thanks, @qjia7.
• [webgpu] Add support for SparseToDense (#5629). Thanks, @haoyunfeix.
• Add cases for Conv2DMMVec4Program and Conv2dNaiveProgram (#5637). Thanks, @axinging.
• webgpu: add flipLeftRight and rotate operators (#5649). Thanks, @xhcao.
• [webgpu] Add WGSL shader for Tanh (#5647). Thanks, @haoyunfeix.
• [webgpu] Use WGSL default (#5621). Thanks, @axinging.
• [webgpu] Use i32 as coords type in WGSL (#5628). Thanks, @axinging.
• [webgpu] Create FromPixles program as required (#5610). Thanks, @axinging.
• [webgpu] Only stride is used in PoolWithFilterSizeEqualsOneProgram (#5593). Thanks, @axinging.
• [webgpu] Fix WGSL pow case fail (#5589). Thanks, @axinging.
• [webgpu] Add MatMulSmallOutputSizeProgram WGSL support (#5567). Thanks, @axinging.
• [webgpu] Add conv2dBackprop WGSL support (#5565). Thanks, @axinging.
• [webgpu] Use module constant as workgroup size in WGSL (#5579). Thanks, @axinging.
• webgpu: Hide importExternalTexture behind flag (#5587). Thanks, @shaoboyan.
• Throw errors when any index is out of bound in GatherV2 (#5439).
• [webgpu] Add conv2d/conv2dNaive/matmul/im2col WGSL support (#5527). Thanks, @axinging.
• webgpu: Add wgsl support for FromPixels (#5552). Thanks, @shaoboyan.
• [webgpu] Add depthwise3x3/argminmax/reduce/resize_nearest_neighbor WGSL support (#5535). Thanks, @axinging.
• webgpu: optimize matmul with small output size (#5342). Thanks, @xhcao.
• [webgpu] Add time query profiling for fromPixels op (#5503). Thanks, @haoyunfeix.
• [webgpu] Fix exceed dispatch size on WGSL (#5519). Thanks, @axinging.
• [webgpu] update warning message to guide user enable timestamp-query extension (#5505). Thanks, @haoyunfeix.
• [webgpu] Fix warning that buffer binding size cannot be zero. (#5526). Thanks, @qjia7.
• [webgpu] Fix clipByValue vec4 error (#5511). Thanks, @axinging.
• webgpu: fix dispatch size exceeds the limits (#5432). Thanks, @xhcao.
• [webgpu] Add depthwise and other 21 programs WGSL support (#5403). Thanks, @axinging.
• [webgpu] Add conv2d vec4 and matmul vec4 WGSL support (#5369). Thanks, @axinging.
• Fix FromPixels type checker to avoid error in worker (#5472). Thanks, @shaoboyan.
• [webgpu] Fix wrongly modified NaN type (#5462). Thanks, @axinging.
• [webgpu] Use ImportToTexture to handle video in FromPixels (#5430). Thanks, @shaoboyan.
• [webgpu] Add binary WGSL support (#5322). Thanks, @axinging.
• [webgpu] Remove tfjs-backend-webgl and tfjs-converter dependency (#5428). Thanks, @haoyunfeix.
• [webgpu] Fix that exceeding limitation errors (#5385). Thanks, @qjia7.
• Build tfjs-backend-webgl with Bazel (#5339).
• [webgpu] Fix PadV2 with empty size (#5363). Thanks, @qjia7.
• Use Bazel and ts_library to build core and backend-cpu (#5133).
• Add 'WEBGPU_USE_PROFILE_TOOL' flag (#5300). Thanks, @shaoboyan.
• [webgpu] Fix nan fail dut to tint (#5321). Thanks, @axinging.
• [webgpu] Fix floorDiv vec4 compile fail (#5307). Thanks, @axinging.
• [webgpu] Update @webgpu/types to version 0.1.4 (#5290). Thanks, @haoyunfeix.
• [webgpu] Add WGSL support (#5252). Thanks, @axinging.
• [webgpu] Add case for matmul and conv2d to test special path (#5253). Thanks, @axinging.
• [webgpu] Check if runtime support WebGPU before initial a WebGPU backend (#5218). Thanks, @haoyunfeix.
• [webgpu] Upgrade clang-format version from 1.2.4 to 1.5.0 (#5225). Thanks, @haoyunfeix.
• [webgpu] Simplify shader key of binary op and unary op (#5222). Thanks, @axinging.
• [webgpu] Simplify activation shader key (#5208). Thanks, @axinging.
• Use CopyExternalImageToTexture() in FromPixels (#5202). Thanks, @shaoboyan.

tfjs-automl-v1.3.0

tfjs-automl (1.2.0 ==> 1.3.0)

Features

• Update dependencies to 3.6.0 (#5038). Thanks, @aravindvnair99.

Security

• Fix CVE-2021-23362 in tfjs-automl (#5078). Thanks, @aravindvnair99.

Misc

• Update tfjs-automl to 1.3.0. (#5614).
• [automl] Update tfjs deps to 3.9.0 (#5570).

tfjs-react-native-v0.7.0

tfjs-react-native (0.6.0 ==> 0.7.0)

• [react-native] Add option to TensorCamera to not use custom shaders to resize camera image texture (#5626). This means you don't need to specify cameraTextureWidth and cameraTextureHeight props in TensorCamera anymore.

tfjs-tflite-v0.0.1-alpha.6

tfjs-tflite now has better supports for quantized models. See XNNPACK doc for supported ops and limitations (the quantization related flags have been enabled for tfjs-tflite build). Check out this demo for running quantized CartoonGAN model with the latest tfjs-tflite package.

Other changes:

• [tflite] Update README and fix demo (#5545).
• Update tflite web api download script (#5543).
• Make Bazel npm outputs match current packages (#5451).
• [tflite] Update google storage location (#5457).
• Migrate tfjs-tflite package to bazel (#5358).
• [tflite] Fix script tags order and load example in README (#5376). Thanks, @Romaleks360.
• Use Bazel and ts_library to build core and backend-cpu (#5133).
• Update README to make things more clear to run the cartoonizer demo (#5271).

tfjs-v3.9.0

Core (3.8.0 ==> 3.9.0)

Features

• Add grayscaleToRGB ops (#5308). Thanks, @raffizulvian.

Bug fixes

• fix (#5523).
• [tfjs-core] fixed version test by require from both relative and absolute path (#5405).
• [tfjs-core] disable logging when prod flag is set (#5368).
• [core] fix: corrected dilations is not passed to the next func (#5364). Thanks, @sungongwei.
• [webgl]Fix PadV2 with empty size. (#5359).

Misc

• Update monorepo to 3.9.0. (#5544).
• Include README.md files in Bazel-generated npm packages (#5542).
• [wasm] Fix offset calculation in Conv2DBackpropInput (#5518).
• Fix string split call stack overflow (#5509).
• [webgl] Fixed nightly test failure (#5510).
• [webgpu] Add depthwise and other 21 programs WGSL support (#5403). Thanks, @axinging.
• fix depthwise conv2d bug (#5500).
• Add broadcast args (#4636). Thanks, @drubinstein.
• [core] Add fallback method for isMobile (#5478).
• fix(docs): elu wrong definition (#5474). Thanks, @raffizulvian.
• Make Bazel npm outputs match current packages (#5451).
• Run enumerate-tests with Bazel (#5361).
• Use absolute import for version_test (#5416).
• Add fused depthwiseConv2d case for filter 3x3 (#5333). Thanks, @axinging.
• tfjs_bundle related improvements (#5337).
• Use Bazel and ts_library to build core and backend-cpu (#5133).

Data (3.8.0 ==> 3.9.0)

Bug fixes

• [tfjs-data] Fix an issue caused by incorrectly cloning certain iterator (#5444).

Misc

• Update lockfiles branch tfjs_3.9.0_lockfiles lock files. (#5555).
• Update monorepo to 3.9.0. (#5544).
• Use Bazel and ts_library to build core and backend-cpu (#5133).

Layers (3.8.0 ==> 3.9.0)

Misc

• Update lockfiles branch tfjs_3.9.0_lockfiles lock files. (#5555).
• Update monorepo to 3.9.0. (#5544).
• improving error message when input or dataset output shape does not match the inputShape (#5524). Thanks, @sourabh112.
• Revert "Build tfjs-backend-webgl with Bazel " (#5476) (#5339).
• Make Bazel npm outputs match current packages (#5451).
• fix(typo): in comment; TensorFow -> TensorFlow (#5407). Thanks, @GoogleSites.
• Build tfjs-backend-webgl with Bazel (#5339).
• Use Bazel and ts_library to build core and backend-cpu (#5133).

Converter (3.8.0 ==> 3.9.0)

Bug fixes

• Create a separate 'link-package-core' for packages that only need core (#5436).

Misc

• Update lockfiles branch tfjs_3.9.0_lockfiles lock files. (#5555).
• Update monorepo to 3.9.0. (#5544).
• Add broadcast args (#4636). Thanks, @drubinstein.
• Use Bazel and ts_library to build core and backend-cpu (#5133).

Node (3.8.0 ==> 3.9.0)

Misc

• Update lockfiles branch tfjs_3.9.0_lockfiles lock files. (#5555).
• Update monorepo to 3.9.0. (#5544).
• [wasm] Fix offset calculation in Conv2DBackpropInput (#5518).
• Update package.json (#5499). Thanks, @rthadur.
• Add broadcast args (#4636). Thanks, @drubinstein.
• Update README.md (#5427). Thanks, @rthadur.
• Use Bazel and ts_library to build core and backend-cpu (#5133).

Wasm (3.8.0 ==> 3.9.0)

Features

• [wasm] add elu kernel (#5435). Thanks, @raffizulvian.

Bug fixes

• [webgl]Fix PadV2 with empty size. (#5359).

Security

• Bump path-parse from 1.0.6 to 1.0.7 in /tfjs-backend-wasm (#5464). Thanks, @dependabot[bot].

Misc

• Update lockfiles branch tfjs_3.9.0_lockfiles lock files. (#5555).
• Update monorepo to 3.9.0. (#5544).
• Update safari to 13 for wasm tests (#5540).
• [wasm] Fix offset calculation in Conv2DBackpropInput (#5518).
• [wasm] Make it more obvious for calling setWasmPaths when WASM backend is loaded from script tag (#5477).
• [wasm] Remove uses of 'build-deps-ci' (#5340).
• Use Bazel and ts_library to build core and backend-cpu (#5133).

Cpu (3.8.0 ==> 3.9.0)

Features

• [webgl] Allow Sqrt to run on packed data (#5396).

Misc

• Update monorepo to 3.9.0. (#5544).
• Include README.md files in Bazel-generated npm packages (#5542).
• Fix string split call stack overflow (#5509).
• Add broadcast args (#4636). Thanks, @drubinstein.
• Make Bazel npm outputs match current packages (#5451).
• Migrate tfjs-tflite package to bazel (#5358).
• Use absolute import for version_test (#5416).
• Allow Sigmoid to run on packed data (#5406).
• tfjs_bundle related improvements (#5337).
• Use Bazel and ts_library to build core and backend-cpu (#5133).

Webgl (3.8.0 ==> 3.9.0)

Features

• [webgl] Allow Sqrt to run on packed data (#5396).
• webgl: Use uniforms for im2col/encode[decode] matrix programs (#5297). Thanks, @qjia7.

Bug fixes

• #5479 [webgl: Fix the incorrect result with shapes uniforms] (#5502). Thanks, @qjia7.
• webgl: Use imod instead of mod (#5514). Thanks, @qjia7.
• #5447 [webgl: Use imod instead of mod] (#5514). Thanks, @qjia7.
• #5447 [webgl: Fix the incorrect result for blazeface] (#5483). Thanks, @qjia7.
• [webgl] Temporarily disable WEBGL_USE_SHAPES_UNIFORMS tests (#5480).
• [tfjs-data] Fix an issue caused by incorrectly cloning certain iterator (#5444).
• Create a separate 'link-package-core' for packages that only need core (#5436).
• Fix the incorrect benchmark result when using uniforms for im2col (#5404). Thanks, @qjia7.
• [webgl]Fix PadV2 with empty size. (#5359).
• [webgl]Fix conv2d when width is odd. (#5264).

Misc

• Update lockfiles branch tfjs_3.9.0_lockfiles lock files. (#5555).
• Update monorepo to 3.9.0. (#5544).
• webgl: Enable single run for yarn test (#5525). Thanks, @qjia7.
• [webgl] Fixed nightly test failure (#5510).
• fix depthwise conv2d bug (#5500).
• Unpack input in ArgMin/Max WebGL kernel (#5437).
• Add broadcast args (#4636). Thanks, @drubinstein.
• Revert "Build tfjs-backend-webgl with Bazel " (#5476) (#5339).
• Make Bazel npm outputs match current packa...

tfjs-v3.8.0

Core (3.7.0 ==> 3.8.0)

Bug fixes

• fixed depthwise conv2d packed shader issue where dilation is great than 2 (#5320).
• Fix #5273: Refactor IOHandlers to use same JSON parsing (#5274). Thanks, @MaximeKjaer.

Misc

• Update monorepo to 3.8.0. (#5331).
• Add flags for controlling when TopK GPU is used (#5328). Thanks, @ahmedsabie.
• [webgpu] Fix floorDiv vec4 compile fail (#5307). Thanks, @axinging.
• Don't try to load weights if there are no weights (#5310).
• Fix flipLeftRight op in various backends and add more tests (#5270).
• Add the missing backend initialization in runKernel (#5251).
• Switch CPU TopK Algorithm to use Floyd-Rivest (#5236).
• Switch WebGL TopK algorithm to use bitonic (#5233).
• Supported conv2dTranspose with explicit paddings. (#5185). Thanks, @BruceDai.
• Supported pooling operation with explicit padding. (#5163). Thanks, @BruceDai.

Data (3.7.0 ==> 3.8.0)

Misc

• Update lockfiles branch tfjs_3.8.0_lockfiles lock files. (#5332).
• Update monorepo to 3.8.0. (#5331).

Layers (3.7.0 ==> 3.8.0)

Bug fixes

• [tfjs-layers] remove deps of chained ops in layers (#5188).

Performance

• [tfjs-layers] remove deps of chained ops in layers (#5188).

Misc

• Update lockfiles branch tfjs_3.8.0_lockfiles lock files. (#5332).
• Update monorepo to 3.8.0. (#5331).

Converter (3.7.0 ==> 3.8.0)

Misc

• Update lockfiles branch tfjs_3.8.0_lockfiles lock files. (#5332).
• Update monorepo to 3.8.0. (#5331).
• fix converter build file for g3 (#5232).

Node (3.7.0 ==> 3.8.0)

Features

• libtensorflow path name replacing (#5171). Thanks, @kawauso.
• upgrade tensorflow deps to 2.5 and built raspberry pi 2.5 binary (#5204).

Bug fixes

• Fix #5273: Refactor IOHandlers to use same JSON parsing (#5274). Thanks, @MaximeKjaer.
• upgrade tensorflow deps to 2.5 and built raspberry pi 2.5 binary (#5204).

Misc

• Update lockfiles branch tfjs_3.8.0_lockfiles lock files. (#5332).
• Update monorepo to 3.8.0. (#5331).
• Update WINDOWS_TROUBLESHOOTING.md (#5327). Thanks, @rthadur.
• Add FlipLeftRight Node.JS kernel using reverse op. (#5311). Thanks, @DCtheTall.
• Revert "Add FlipLeftRight NodeJS kernel. " (#5263) (#5261).
• Add FlipLeftRight NodeJS kernel. (#5261). Thanks, @DCtheTall.
• Supported conv2dTranspose with explicit paddings. (#5185). Thanks, @BruceDai.
• Supported pooling operation with explicit padding. (#5163). Thanks, @BruceDai.

Wasm (3.7.0 ==> 3.8.0)

Features

• Add batchToSpaceND/spaceToBatchND kernels (#5288). Thanks, @kinsei0916.

Misc

• Update lockfiles branch tfjs_3.8.0_lockfiles lock files. (#5332).
• Update monorepo to 3.8.0. (#5331).
• [wasm] Mark 'os' as excluded in browser bundles (#5302).
• Fix flipLeftRight op in various backends and add more tests (#5270).
• Implement cosh in tfjs-backend-wasm (#5260). Thanks, @carrycooldude.
• Switch WASM TopK Algorithm to use Floyd-Rivest (#5244).
• Switch WebGL TopK algorithm to use bitonic (#5233).
• Update WASM README about setting up cross-origin isolation (#5169).

Cpu (3.7.0 ==> 3.8.0)

Misc

• Update lockfiles branch tfjs_3.8.0_lockfiles lock files. (#5332).
• Update monorepo to 3.8.0. (#5331).
• Fix flipLeftRight op in various backends and add more tests (#5270).
• Switch CPU TopK Algorithm to use Floyd-Rivest (#5236).
• Switch WebGL TopK algorithm to use bitonic (#5233).

Webgl (3.7.0 ==> 3.8.0)

Features

• webgl: Refactor the custom uniforms support (#5258). Thanks, @qjia7.
• webgl: Add shapes uniforms to reduce shader compilation time (#5240). Thanks, @qjia7.

Bug fixes

• fixed depthwise conv2d packed shader issue where dilation is great than 2 (#5320).
• Remove redeclared value uniform (#5312).

Performance

• webgl: Add shapes uniforms to reduce shader compilation time (#5240). Thanks, @qjia7.

Misc

• Update lockfiles branch tfjs_3.8.0_lockfiles lock files. (#5332).
• Update monorepo to 3.8.0. (#5331).
• Add flags for controlling when TopK GPU is used (#5328). Thanks, @ahmedsabie.
• Unpack input in TopK WebGL kernel (#5286).
• Fix flipLeftRight op in various backends and add more tests (#5270).
• Switch WebGL TopK algorithm to use bitonic (#5233).
• Fix webgl backend memory leak in firefox (#5193).

tfjs-react-native-v0.6.0

tfjs-react-native (0.5.0 ==> 0.6.0)

Documentation

• Update tfjs-react-native README to add steps to fix a build error when using expo for web (#4805).
• [tfjs-react-native] fix metro link (#4450). Thanks, @AmanRaj1608.

Security

• fix sec vulnerabilities for all packages (#4869).

Misc

• Update tfjs-react-native to 0.6.0. (#5247).
• Update react-native monorepo dependencies to 3.7.0 (#5217).
• Replace deprecated package '@react-native-community/async-storage' (#5228). Thanks, @gregor-mittag.
• Build tfjs-core and tfjs-backend-cpu with Bazel (#4856).
• Bump jpeg-js (#4718).
• Upgrade react native to 3.1.0 (#4610).
• Run browserstack tests with an allowance for flakiness (#4691).
• docs: remove duplicate documentation line (#4599). Thanks, @autimio.

tfjs-backend-webgpu-v0.0.1-alpha.7

tfjs-backend-webgpu (0.0.1-alpha.5 ==> 0.0.1-alpha.7)

Features

• expose CPU_HANDOFF_SIZE_THRESHOLD as a configurable flag (#5106).
• [webgpu] Add support for gatherND (#5057). Thanks, @haoyunfeix.
• [webgpu] Add support for sqrt (#5058). Thanks, @haoyunfeix.
• [webgpu] Add resizeNearestNeighbor support (#5059). Thanks, @haoyunfeix.
• added support for sigmoid activation function for fused ops (#4948).

Bug fixes

• webgpu: Fix a potential bug in vector*matrix (#5165). Thanks, @qjia7.
• webgpu: Fix the reduce shader key (#5148). Thanks, @qjia7.
• webgpu: Fix the mirrorPad error (#5131). Thanks, @qjia7.
• webgpu: Add CPU_HANDOFF_SIZE_THRESHOLD (#5119). Thanks, @qjia7.

Performance

• webgpu: Optimize Conv2DBackpropInput (#5155). Thanks, @qjia7.
• webgpu: Add vector*matrix kernel (#5129). Thanks, @qjia7.
• webgpu: Optimize getSamplerAtOutputCoords (#5105). Thanks, @qjia7.
• [webgpu] Use uniform instead of constant for shape (#4819). Thanks, @axinging.

Security

• Fix CVE-2021-23362 in tfjs-backend-webgpu (#5086). Thanks, @aravindvnair99.

Misc

• Bump version to 0.0.1-alpha.7 (#5176).
• Upgrade webgpu to 0.0.1-alpha.6 (#5175).
• [webgpu] Fix string comparison fail and add equal (#5149). Thanks, @axinging.
• webgpu: Fix the OOB error in transform/slice/stridedSlice (#5164). Thanks, @qjia7.
• webgpu: Add Pow kernel (#5130). Thanks, @qjia7.
• [webgpu] Add op StringNGrams (#5118). Thanks, @axinging.
• [webgpu] Fix nan errors (#5121). Thanks, @axinging.
• [webgpu] Update yarn.lock for package types/long (#5125). Thanks, @axinging.
• [webgpu] Add shared memory flag to binary op shared shader key (#5083). Thanks, @axinging.
• webgpu: Remove cpu backend (#5101). Thanks, @qjia7.
• webgpu: Add Conv2DBackpropInput kernel (#5104). Thanks, @qjia7.
• [webgpu] Fix invalid WGPUBindingType (#5082). Thanks, @axinging.
• webgpu: run concat on cpu if possible (#5076). Thanks, @qjia7.
• [webgpu] Reuse GPUBindGroupLayout and GPUPipelineLayout (#5001). Thanks, @axinging.
• webgpu: fix/suppress some failure cases (#5065). Thanks, @qjia7.
• webgpu: Add Transform kernel (#5061). Thanks, @qjia7.
• webgpu: upgrade @webgpu/types (#5062). Thanks, @qjia7.
• [webgpu] Simplify shader key of slice and cropAndResize (#4993). Thanks, @axinging.
• [WebGPU] Expand texture usage for fromPixels (#5016). Thanks, @shaoboyan.
• webgpu: Use outShapeStrides in more places (#4967). Thanks, @qjia7.
• webgpu: Use uniforms for PadV2 paddings (#4966). Thanks, @qjia7.
• [webgpu] Use uniform for program batchNorm, clip, im2col (#4958). Thanks, @axinging.

tfjs-v3.7.0

Core (3.6.0 ==> 3.7.0)

Features

• Added string support for logical binary ops, gatherND, and StridedSlice (#5135).
• Add kernel StringToHashBucketFast for CPU and WebGL backend (#5052).
• align allbacks with tensorflow on NaN propogation with min/max ops (#5028).

Bug fixes

• Fix CI failure for requiring the long package. (#5085).
• uses require instead of enable esmoduleinterop flag (#5074).

Security

• Bump ws from 7.4.4 to 7.4.6 in /tfjs-core (#5141). Thanks, @dependabot[bot].
• Bump hosted-git-info from 2.8.8 to 2.8.9 in /tfjs-core (#5048). Thanks, @dependabot[bot].

Misc

• Update monorepo to 3.7.0. (#5168).
• [webgpu] Add op StringNGrams (#5118). Thanks, @axinging.
• Allow string inputs in gatherND (#5157).
• Add kernel StringNGrams for CPU and WebGL backend (#5102).
• Add kernel StringSplit for CPU and WebGL backend (#5089).
• Add kernel SparseSegmentMean for CPU and WebGL backend (#5025).
• Add kernel SparseSegmentSum for CPU and WebGL backend (#5018).
• Fix SparseFillEmptyRows output shape bug (#5020).
• Upgrade rules_nodejs to 3.4.1 and karma to 6.3.2 (#5006).

Data (3.6.0 ==> 3.7.0)

Features

• Add kernel StringToHashBucketFast for CPU and WebGL backend (#5052).

Misc

• Update lockfiles branch tfjs_3.7.0_lockfiles lock files. (#5170).
• Update monorepo to 3.7.0. (#5168).

Layers (3.6.0 ==> 3.7.0)

Features

• use vanilla js for the math util methods (#5096).
• Add kernel StringToHashBucketFast for CPU and WebGL backend (#5052).

Bug fixes

• uses require instead of enable esmoduleinterop flag (#5074).

Misc

• Update lockfiles branch tfjs_3.7.0_lockfiles lock files. (#5170).
• Update monorepo to 3.7.0. (#5168).

Converter (3.6.0 ==> 3.7.0)

Features

• Fuse depthwise conv2d with activation function without biasAdd (#4910).
• [converter] Support function execution with named output indexing (#5127).
• Add support for StringSplit conversion (#5107).
• Add support for StringToHashBucketFast conversion (#5067).
• Add kernel StringToHashBucketFast for CPU and WebGL backend (#5052).
• [tfjs-converter] heuristically determine the signature name (#5043).

Bug fixes

• Allow up to TF-Hub==0.12 (#5152). Thanks, @chongkong.
• [converter] Support function execution with named output indexing (#5127).
• Add missing op categories in converter and tests (#5114).
• check function def for unsupported ops (#5090).
• uses require instead of enable esmoduleinterop flag (#5074).
• [tfjs-converter] heuristically determine the signature name (#5043).

Misc

• Update lockfiles branch tfjs_3.7.0_lockfiles lock files. (#5170).
• Update monorepo to 3.7.0. (#5168).
• Add support for StringNGrams conversion (#5109).
• Add support for SparseSegmentMean conversion (#5037).
• Add support for SparseSegmentSum conversion (#5026).
• Add support for SparseFillEmptyRows conversion (#5005).

Node (3.6.0 ==> 3.7.0)

Features

• [node] Add SummaryFileWriter.histogram TensorBoard logging (#5079). Thanks, @MaximeKjaer.
• Added string support for logical binary ops, gatherND, and StridedSlice (#5135).
• Make tfjs-node create a new TFJSBackend for each environment (#5108). Thanks, @bmcdorman.
• Add kernel StringToHashBucketFast for CPU and WebGL backend (#5052).
• align allbacks with tensorflow on NaN propogation with min/max ops (#5028).

Bug fixes

• Make tfjs-node create a new TFJSBackend for each environment (#5108). Thanks, @bmcdorman.
• uses require instead of enable esmoduleinterop flag (#5074).
• fix the node mem leak on saved model execution (#5012).

Misc

• Update lockfiles branch tfjs_3.7.0_lockfiles lock files. (#5170).
• Update monorepo to 3.7.0. (#5168).
• Add kernel StringNGrams for CPU and WebGL backend (#5102).
• Add kernel StringSplit for CPU and WebGL backend (#5089).
• update node-pre-gyp package (#5036).
• Add kernel SparseSegmentMean for CPU and WebGL backend (#5025).
• Add kernel SparseSegmentSum for CPU and WebGL backend (#5018).

Wasm (3.6.0 ==> 3.7.0)

Features

• Added string support for logical binary ops, gatherND, and StridedSlice (#5135).
• align allbacks with tensorflow on NaN propogation with min/max ops (#5028).

Bug fixes

• [wasm] fix #5021: wechat webassembly support (#5056). Thanks, @deepkolos.

Documentation

• Update README of WASM backend (#5097).

Security

• Bump ws from 7.4.4 to 7.4.6 in /tfjs-backend-wasm (#5140). Thanks, @dependabot[bot].
• Bump browserslist from 4.16.3 to 4.16.6 in /tfjs-backend-wasm (#5116). Thanks, @dependabot[bot].

Misc

• Update lockfiles branch tfjs_3.7.0_lockfiles lock files. (#5170).
• Update monorepo to 3.7.0. (#5168).

Cpu (3.6.0 ==> 3.7.0)

Features

• Added string support for logical binary ops, gatherND, and StridedSlice (#5135).
• Add kernel StringToHashBucketFast for CPU and WebGL backend (#5052).
• align allbacks with tensorflow on NaN propogation with min/max ops (#5028).

Bug fixes

• uses require instead of enable esmoduleinterop flag (#5074).

Security

• Fix CVE-2021-23362 in tfjs-backend-cpu (#5088). Thanks, @aravindvnair99.

Misc

• Update lockfiles branch tfjs_3.7.0_lockfiles lock files. (#5170).
• Update monorepo to 3.7.0. (#5168).
• [webgpu] Add op StringNGrams (#5118). Thanks, @axinging.
• Add kernel StringNGrams for CPU and WebGL backend (#5102).
• Add kernel StringSplit for CPU and WebGL backend (#5089).
• Add kernel SparseSegmentMean for CPU and WebGL backend (#5025).
• Add kernel SparseSegmentSum for CPU and WebGL backend (#5018).
• Fix SparseFillEmptyRows output shape bug (#5020).
• Upgrade rules_nodejs to 3.4.1 and karma to 6.3.2 (#5006).

Webgl (3.6.0 ==> 3.7.0)

Features

• Added string support for logical binary ops, gatherND, and StridedSlice (#5135).
• expose CPU_HANDOFF_SIZE_THRESHOLD as a configurable flag (#5106).
• Add kernel StringToHashBucketFast for CPU and WebGL backend (#5052).
• align allbacks with tensorflow on NaN propogation with min/max ops ([#5028]...