Release 2.8.0-rc1

Major Features and Improvements

• No major features or improvements.

Breaking Changes

• No breaking changes

Bug Fixes and Other Changes

• Update version for 2.8.0-rc1 release. (commit: 90d2cec9a837ca51f6fd6ea89481cbc918174f7d)

Release 2.8.0-rc0

Major Features and Improvements

• No major features or improvements.

Breaking Changes

• No breaking changes

Bug Fixes and Other Changes

• Replace int64 with int64_t and uint64 with uint64_t. (commit: 21360c763767823b82768ce42c5c90c0c9012601)
• Replace remaining deprecated int64 with int64_t. (commit: d34dee26cde50f6195e7aed8793c8b51a240f387)
• update to latest benchmark API changes (commit: 860e1013385289ad3f9eb4d854b55c23e7cb8087)
• Internal changes. (commit: 916b252d6fba3e1143c6ab6f98d9f37d6e8d4088)
• Remove unused "using" declarations. (commit: 3e05381ee0fe790b368072e0148b35ed21d8a995)
• Force-allow labels for the models which are not yet loaded. The feature is meant to be used for non-prod environments only as it may break the serving until relevant models are not loaded yet. (commit: 988bbce80038ac0b7141dcb4413124ba486344cf)
• Update CreateRPC API interface. (commit: 2e7ca90c18f310c542ed0dcde92d676db6454285)
• Update a precondition failure log message. (commit: 4f63dbf86deb2f6cd2a3d680bf6bfc7f086cfdcc)
• Internal change. (commit: 6ece2f5af9baf33f1efb90ff35495186f0e636de)
• Add --tensorflow_session_config_file flag to tf serving model server to support custom SessionConfig (commit: 342a8088524c33f68e3eb4d66800f01a777ceb38)
• Add --experimental_cc_shared_library by default to all builds. (commit: 49b138fdd4c0fb7170736193063c6f03dfb4dba4)
• Add --num_request_iterations_for_warmup flag (fixes #1949) (commit: 2a55aec18cc1923ece84c7fcf701185306ef99b1)
• Update version for 2.8.0-rc0 release. (#1956) (commit: 297818a55a20477c035ddb2feda54c30780fab41)

Release 2.6.2

Major Features and Improvements

• No major features or improvements.

Breaking Changes

• No breaking changes

Bug Fixes and Other Changes

• This release is based on TF version 2.6.2

Release 2.7.0

Major Features and Improvements

• No major features or improvements.

Breaking Changes

• No breaking changes

Bug Fixes and Other Changes

• This release is based on TF version 2.7.0

Release 2.6.1

Major Features and Improvements

• No major features or improvements.

Breaking Changes

• No breaking changes

Bug Fixes and Other Changes

• This release is based on TF version 2.6.1

Release 2.5.3

Major Features and Improvements

• No major features or improvements.

Breaking Changes

• No breaking changes

Bug Fixes and Other Changes

• This release is based on TF version 2.5.2

Release 2.4.4

TensorFlow Serving using TensorFlow 2.4.4

Major Features and Improvements

• No Major Features and Improvements

Breaking Changes

• No Breaking changes

Bug Fixes and Other Changes

• Improve error message when version directory is not found (commit: d687d3e8827c82f4f1b68337c67b2cbe6e4126e7)
• Update TF Text to v2.3.0. (commit: 390732ace75ca9110c706f0aeb65d35a5390ca81)
• Make Internal Changes (commit: d599e9513b00b0a83b7038659ba6ad9cafd18e9d)
• Add release notes for TF Serving 2.3.0 (commit: 85a6a23e3ad2e68767b44468f9040f8085bccbb2)
• In shared batch scheduler, rename 'max_batch_size' to 'input_batch_size_limit'. (commit: ead408b9ac1a7491ea410e59bec4b1bd91cd4a00)
• Implement unit test for external version of remote predict op. (commit: 9d4ae676c67acb100913f0682a9332aaa3303e55)
• 1. Add an example which call the Remote Predict Op directly. (commit: d5b980f487996aa1f890a559eae968735dfebf5d)
• Update default remote model name. (commit: a069a434cfd27fc64e21bc43a11aaffd8847376b)
• Adds user guide for Remote Predict Op (commit: fc82463d197dde52a613b1717f8705964c57178f)
• Add support for serving regress/classify (native) TF2 models. (commit: b724ceda57b4fdc9b5bd09da231c7186435cd807)
• Use LOG(WARNING) instead of LOG(INFO) for actual allocation problems. (commit: 4a826d0f2f2ee46d240408297616711612156ddd)
• Make Internal Change (commit: e24c50b9599eae89af74439fcc1fc6badce036d5)
• Move profiler services behind profiler namespace. (commit: d8faed196f40703881556a33ee86503f7a4a237e)
• Temporarily remove sm_80 to fix gpu image build failure. (commit: 6eab5efaa1acaef452795841badaa7c8663b50e6)
• Upgrade to CUDA Version 11.0. Build docker images on a local machine successfully with command: docker build --pull -t $USER/tensorflow-serving-devel-gpu -f Dockerfile.devel-gpu . docker build --pull -t $USER/tensorflow-serving-gpu -f Dockerfile.gpu . (commit: b291c4df771b3a6767782bdcfa086f93512ca182)
• Internal fix on TFLite support. (commit: 23a5e8ba8a854e8dfa56eec94e48d77730ec8576)
• Broaden net_http client visibility and use HTTPStatusCode in ClientResponse status (commit: 72927141f29d83614cf8304725b1a7e6357ce6ac)
• Fixit: fix order dependency in batching_session_test. (commit: 0eadb8f80ed9c222ca64cc74df95af797bfd180d)
• Update CUDNN_VERSION to 8.0.4.30. (commit: 84e0189d1b9828813125efef9ff2d541f101705a)
• Move some utilities from batching_session to batching_util for upcoming TFRT SavedModel batching supports. (commit: 43e7cfaa39357b69d54a3a91b8696d7aed7fa7e4)
• Temporarily disable failing test, until we root cause and fix it. (commit: aa6c5f5b6175fb17064c0afbd999da9eccae6ea3)
• Remove deprecated flag fail_if_no_model_versions_found from tensorflow serving binary (commit: 005f0f6ad762ef11ede9d04d9a256f7967a4c31f)
• Internal Change. (commit: a05d5245e6a9132444c7b2cf989f58ceb08527f6)
• Update version for 2.4.0-rc2 release. (commit: 2388d4253351811d8a3b2fb924095cbf2ba53214)
• Adds LICENSE file to tensorflow-serving-api python package. (commit: 93668685296f7ab70b8ebdc477684e808806bec9)
• Update Dockerfile.devel (commit: e3afe78c1cf7f64b9724229817f9e31b6baf1da7)
• Update Dockerfile.devel-gpu (commit: 5229c271116c820525cde5670e70d8a406397fb1)
• temporary fix for 2.4.0rc2 docker image (commit: af0ddc005e136c9cc43fb9ad14a9562266ad6ed0)
• Update Dockerfile.devel (commit: c07d0cc4117ff82a32cddd0fc9498a423d822827)
• Update Dockerfile.devel-gpu (commit: 4dae5974a8ebf7b638dcf4f63838c76d7a5030fa)
• Update setup.py (commit: 5b1ea00c68fa0e174de1428ecd472b562f2aed7d)
• Update setup.py (commit: b54e566cc92d167b005753b9ec6f5700c46ab3a3)
• Update version for 2.4.0-rc3 release. (commit: abe4ffee01a3790655126f4dd35576dd8dc46722)
• Update version for 2.4.0-rc4 release. (commit: 16e661a66a00b7e67d10ca7a5e90cf23f04e279d)
• Update version for 2.4.0 release. (commit: 48c4da3a31abb17dfc26b5e1edb185de5f9da5c0)
• Update version for 2.4.1 release. (commit: 03b2bedc14e2f18aefd073f3bf885d4c33464712)
• Make Internal Change (commit: 644baeb482741e96d2ca0a5acdc2d7d2ab038710)
• Enable download of TF Serving sources at arbitrary commit for GPU/MKL docker images. (commit: 4725d9095919c64bb3f6c60b55bfbb103fe6a52e)
• Fix TensorFlow serving build issue for MKL dockerfiles (commit: 424608049a5441bb8e0d8befec16b008ebda35d0)
• Update TF Text to v2.4.3. (commit: 3bd841791b2b25f999a0ad78ccf3594000fc09ea)
• Update version for 2.4.2 release. (commit: a6d405e771a0bd3d4349784afb3cd65a1920cb92)
• Update version for 2.4.3 release. (commit: 9a377cdd047c970f6ded05830e2e59e4c359aaf3)
• Update version for 2.4.4 release. (#1931) (commit: d5e1076412b414c2c61ad19174d09fc8fb43e2a6)

Thanks to our Contributors

This release contains contributions from many people at Google, as well as:

Abolfazl Shahbazi, Adarshreddy Adelli

Release 2.7.0-rc1

Bug Fixes and Other Changes

• Update Dockerfile.devel* with py3.7 installed. (commit: 498196b1869d84e2f0c5a4a206545084916d2b0a)
• Update version for 2.7.0-rc1 release. (#1927) (commit: f4d06b21083ba916315b69ccb29dab3eaa721dea)

TensorFlow Serving using TensorFlow 2.7.0-rc0.

Major Features and Improvements

Breaking Changes

Bug Fixes and Other Changes

• Fix nvidia-docker link at Serving Docker using your GPU (commit: 10216d4)
• Replace --runtime=nvidia flag with --gpus all flag as --runtime=nvidia has been dpreciated from Docker19.03 (commit: 9eb2ea0)
• Update resnet_client_grpc.py example (commit: bb1428d)
• Update resnet_client.py example (commit: cb28fca)
• Update resnet_warmup.py example (commit: 3b52852)
• Update resnet_client.cc example to send image bytes as opposed to JPEG blob. (commit: 91ce5e1)
• Allow model batch size to override num_tflite_interpreters_per_pool (commit: 582dec8)
• Change ThreadPoolFactory API so that it can create thread pools not owned by the factory itself. (commit: 01b3811)
• Fix wrong header for tensorflow Env. (commit: d0c7490)
• Update benchmarks to newer API. (commit: ad08f48)
• Make Internal Change. (commit: 844ff1c)
• Fix aspired_versions_manager_benchmark test failure - eliminate redundant notification (commit: b18406b)
• Make tensorflow_serving's test_main behave similarly to tensorflow/core's test_main, except that the serving's test_main cannot access the benchmark library functions directly. Therefore it has to use wrappers. (commit: 5c6be5d)
• Remove dependency on TF error code string (commit: 40f6281)
• Return early if shared status sees error (commit: e3a8d8e)
• Add more debugging information for RoundToLowestAllowedBatch (commit: 1b049c6)
• Add release notes for tf-serving 2.3.4. (commit: 3a47647)
• Update TF Text to v2.6.0. (commit: ea21333)
• Add release notes for tf-serving 2.4.3 (commit: a5184ce)
• propagate thread_pool in input-split case. (commit: 67e660e)
• Add release notes for tf-serving 2.5.2 (commit: a4f0e56)
• Use InputSplitMetadata in batching-session's SplitInputTask, so it handles the lazy splits case. (commit: a461c55)
• Internal change (commit: 9ec4243)
• Add release notes for tf-serving 2.6.0 (commit: b4c2152)
• Internal build cleanup (commit: e67bc35)
• Internal build cleanup (commit: 6b1a02b)
• Remove references to deleted targets. (commit: d683fa1)
• internal refactor (commit: 7ce29f6)
• Update version for 2.7.0-rc0 release. (commit: c9e2e34)

Release 2.6.0

Major Features and Improvements

• Update TF Text to v2.5.0. (commit: 48e5a9e23a1e0b2951b77c3e8f9832193d9b1851)
• Add support for Google ALTS. (commit: ceba636bb7c5c98bde35d1818fd033b36300fffe)
• Enable HTTP PATCH support in HTTP/REST server (commit: 6cbc4a9eb419c8078c3a4e791381cda70dd8fc78)

Breaking Changes

• No breaking changes

Bug Fixes and Other Changes

• Enable tensor output filters with batching. (commit: 448dbe14624538ab76fd6aeb2a456344e7f41c78)
• Update tf.io import in warmup example doc. (commit: 6579d2d056530565cd6606a39c82b2f6c1d3799e)
• Resize tensors if the dimensions of the tflite and tensorflow inputs mismatch, even if the number of elements are the same (commit: 8293f44bd5c5ecc68636cd0d036234f891d29366)
• Add basic batch scheduler for tflite models to improve batch parallelism. (commit: 0ffd6080437ca8175b067be7cc00f5b3df9ea92a)
• Reserve Unavailable error to TF communication ops only. (commit: db9aca187affd0453627a1729916acfea98ae800)
• Add the flag thread_pool_factory_config_file to model server and fix a typo. (commit: efc445f416f8cb20606ca0d2aaf44c13fae7ea4c)

Thanks to our Contributors

This release contains contributions from many people at Google.

Release 2.5.2

Major Features and Improvements

• No major features or improvements.

Breaking Changes

• No breaking changes

Bug Fixes and Other Changes

• This release is based on TF version 2.5.1

Release 2.4.3

Major Features and Improvements

• No major features or improvements.

Breaking Changes

• No breaking changes

Bug Fixes and Other Changes

• This release is based on TF version 2.4.3

Release 2.3.4

Major Features and Improvements

• No major features or improvements.

Breaking Changes

• No breaking changes

Bug Fixes and Other Changes

• TensorFlow Serving using TensorFlow 2.3.4

TensorFlow Serving using TensorFlow 2.6.0-rc2

Bug Fixes and Other Changes

• Enable tensor output filters with batching. (commit: 448dbe14624538ab76fd6aeb2a456344e7f41c78)
• Resize tensors if the dimensions of the tflite and tensorflow inputs mismatch, even if the number of elements are the same (commit: 8293f44bd5c5ecc68636cd0d036234f891d29366)
• Update TF Text to v2.5.0. (commit: 48e5a9e23a1e0b2951b77c3e8f9832193d9b1851)
• Add basic batch scheduler for tflite models to improve batch parallelism. (commit: 0ffd6080437ca8175b067be7cc00f5b3df9ea92a)
• Enable HTTP PATCH support in evhttp_server (commit: 6cbc4a9eb419c8078c3a4e791381cda70dd8fc78)

Thanks to our Contributors

Release 2.3.3

Major Features and Improvements

• TensorFlow Serving using TensorFlow 2.3.3.

Breaking Changes

Bug Fixes and Other Changes

TensorFlow Serving using TensorFlow 2.6.0-rc1

Release 2.6.0-rc1

Bug Fixes and Other Changes

• Enable tensor output filters with batching. (commit: 448dbe14624538ab76fd6aeb2a456344e7f41c78)
• Update tf.io import in warmup example doc. (commit: 6579d2d056530565cd6606a39c82b2f6c1d3799e)
• Resize tensors if the dimensions of the tflite and tensorflow inputs mismatch, even if the number of elements are the same
  (commit: 8293f44bd5c5ecc68636cd0d036234f891d29366)
• Update TF Text to v2.5.0. (commit: 48e5a9e23a1e0b2951b77c3e8f9832193d9b1851)
• Add basic batch scheduler for tflite models to improve batch parallelism. (commit: 0ffd6080437ca8175b067be7cc00f5b3df9ea92a
  )
• Add new --use_alts_credentials flag, to enable building secure credentials using Google ALTS. (commit: ceba636bb7c5c98bde35
  d1818fd033b36300fffe)
• Enable HTTP PATCH support in evhttp_server (commit: 6cbc4a9eb419c8078c3a4e791381cda70dd8fc78)

Release 2.4.2

Major Features and Improvements

• TensorFlow Serving using TensorFlow 2.4.2.

Breaking Changes

Bug Fixes and Other Changes

• Enable download of TF Serving sources at arbitrary commit for GPU/MKL docker images. (commit: 4725d9095919c64bb3f6c60b55bfbb103fe6a52e)
• Fix TensorFlow serving build issue for MKL dockerfiles (commit: 424608049a5441bb8e0d8befec16b008ebda35d0)
• Update version for 2.4.2 release. (commit: a6d405e771a0bd3d4349784afb3cd65a1920cb92)

TensorFlow Serving using TensorFlow 2.2.3.

Major Features and Improvements

Breaking Changes

Bug Fixes and Other Changes

• Internal changes. (commit: d044f1e632a80bafa0de2797ee813c68361ec9b7)
• Explicitly export files needed by other packages (commit: 90cb1c0215c7b8d851b92b644f12ded3db494941)
• internal BUILD file cleanup. (commit: 12c0f496229033a1e3467ca08fe3e59e867c729c)
• Add a SourceAdapter that adds a prefix to StoragePath. (commit: f337623da81521eefd8cdc2da1c4a450ecf1d028)
• Internal change. (commit: b052238fe06a03b82cefd82b4245fc7d425bbdc8)
• Switch users of tensorflow::Env::Now*() to EnvTime::Now*(). (commit: 8a0895eb8127941b2e9dada20718dd28f3dbaee1)
• Update the serving model test to use profiler_client.start_tracing instead of (commit: 495e15bbd0f431fa2e93c79cc216f6dbcfc9b6a6)
• moves tf.app to use compat.v1 (commit: fb3c70dc8a44720df870d733f885f6a29e5b6a88)
• Migrates tf.app to use compat.v1 (commit: e8d5ec5ae939b37144420f1dcabc00eedc316bcc)
• Automated rollback of commit e8d5ec5ae939b37144420f1dcabc00eedc316bcc (commit: 46fbcb2ca93f46ff18dee4be9dcde74f9a43df6f)
• Add test for flex op. (commit: 83ca96f1fa9344e381ce2aa95742e4d877031127)
• Automated rollback of commit 83ca96f1fa9344e381ce2aa95742e4d877031127 (commit: b853dfef53d0b766a44d600824d5c7fa73845fff)
• Add release note for TFS 2.1 (commit: 74107ae603011852f916ac0e24e1da3bc804c92f)
• Remove SessionBundle usage from regressor/classifier. (commit: b1b6ffd7dfa654a4f4629519d902d958c07ee7c8)
• Remove SessionBundle support from Predictor. (commit: 2090d67f4e5e8ee5aa7faf8437bea096a438450a)
• Replace the error_codes.proto references in tf serving. (commit: ab475bf6c5e5e4b3b42ffa2aecf18b39fd481ad3)
• In ServableStateMonitor, introduce a helper method 'GetAvailableServables' in ServableStateMonitor. (commit: 4dcfa25b2b65f5a288c774489ad114a2744b56f9)
• Automated rollback of commit b853dfef53d0b766a44d600824d5c7fa73845fff (commit: 1750599bb57625747d810c7c06207fb25784a960)
• Remove unused GetCreator() method. (commit: 114c0352f3b5dac08353f932f1b950f060d988e8)
• Adds performance guide and documentation for TensorBoard integration (commit: f1e4eb2259da90bb9c5fe028ba824ac18a436f67)
• Use SavedModelBundleFactory for session source adapter instead (commit: a59e0dcb56c635e344599b2a7cf948373019b03b)
• Update TF Text to 2.0.1 (commit: bd206e419be2b0c5b1c6306ccab80cf38496875b)
• Remove SessionBundleSourceAdapter as we load Session bundles via (commit: d50aa2b0b986b11368ddcf6b6eb20b9381af474c)
• Remove option to create session bundle based platform config. (commit: 04d0f7e5b76ec0aa0c39b8e8d53db861c9f209bd)
• Cleaning up BUILD visibility for tf_pyclif_proto_library intermediate targets, part 2. (commit: 4774fd838bd34605f24110ad0b9820b17a542687)
• Use SavedModelBundleSourceAdapterConfig instead of (commit: 8ed3ceea985529a350290cf782cb34c3c66827d4)
• Automated rollback of commit 4774fd838bd34605f24110ad0b9820b17a542687 (commit: 8aa38bd41bed02d45a55de4be9fb3b1e67ef79ed)
• Fix docs to refer to SavedModelBundleSourceAdatperConfig as (commit: 4de4c7b6d2f5932f8f2be939335e38c94f1f2dac)
• Update minimum bazel version to 1.2.1. (commit: 1a36026198df5f7dec1e626ef9b112fecdd2916b)
• Automated rollback of commit 8aa38bd41bed02d45a55de4be9fb3b1e67ef79ed (commit: befd6c65a28f6499a459533bbc1ecdc45d1fbb0b)
• Drop support for beta gRPC APIs. (commit: 13d01fc64330ff883bd1553122d9fd114a5a7368)
• API spec for httpserver response-streaming (with flow-control). (commit: fd597f074ce127056515bc52ee3a3d4ff4b727bb)
• Use @rules_pkg to replace @bazel_tools to build deb package (commit: 1e7f349a33992f67ee8ca260b32ddd238e74018c)
• Change Python version to PY3. (commit: 7516746a311f96b57a60598feba40cbdd3989e73)
• Update Python tests in PY3. (commit: 0cf65d2105c191c00fba8918ba75fc955bbeace3)
• Re-format Python test. (commit: ae4de8b2607eaa007079d8f9f02ac79744363cb0)
• Upgrade bazel version for Dockerfiles. (commit: e507aa193b9f3520d40e3da5e4d2263280ff35e4)
• Change dockerfile for PY3. (commit: 7cbd06e8b7720b82b1d2dfae54c3a828d3a52eb4)
• Update TF Text to v2.1.1 which removes ICU variable conflict with core TF. (commit: 2fd33e18a3019d90a4588eeebd364d64b37605ac)
• Reduce contention in FastReadDynamicPtr by sharding the ReadPtrs, by default one per CPU. (commit: d3b374bc70348f2e5e22b7e9ebb191ee9d5b3268)
• Ensure that all outstanding ReadPtrs are destroyed before allowing a (commit: e41ee40826652b6aa5a3f071107074923d6ff6c7)
• fix typo in flag (commit: 138d9b799606ee4a211775699d285206e10a837d)
• Automated rollback of commit 46fbcb2ca93f46ff18dee4be9dcde74f9a43df6f (commit: 385e7962ceae7e30e58d1e4ad54b0502795b9ef5)
• Fixes outdated instructions in kubernetes tutorial (commit: ee665381efc6450bd8dc1499d5f6c939263228fe)
• Refined the httpserver flow-control API for streaming writes. (commit: e085cb37b4b5c62154aa063dcbb781b24d5b3fd4)
• Update bazel genfile directory after bazel version upgrade. (commit: a328ed619b823052bdecfe9f619fbdf3c2a99dbe)
• Use package python-distutils-extra instead of python3-distutils for ubuntu16.04. (commit: b9602b3820dddea0d9fa9423b1ae9eaaf2aec977)
• Explicitly disable Tensorflow v2 behaviors for all TF1.x binaries and tests (commit: 3ffd96983af578cc2d215cd5a7076364dc009f29)
• Internal changes. (commit: fdbbe4804611e22f693f9aba81a863f17b40a312)
• Makes clear that version labels are not supported by REST API surface (commit: 9781ed1857ab0b1195107821ad0f3720872546a1)
• Fix deps on tensorflow proto targets. (commit: 0b83b7e0a442d641a84d24afb278e48c8f1c24d5)
• Allow splitting fields from batched session metadata into individual sessions (commit: caf2a92ba07ca4d10515f0b018c920e9b325c6c8)
• For google-internal, optimize proto parse code for lazy scenario, by annotate byte-type fields with Cord. (commit: 726ca19322680330223a7c9762b1e866b44926aa)
• Fix tensorflow_serving build when building against a recent hash of TensorFlow (commit: 3d5b44a95a2c458affeb284f1141a8612427dd8c)
• Allow passing ThreadPoolOptions in various Session implementations. (commit: 2b6212cf0aa88b719ee00267f83c89d4f7599ef1)
• Fix test broken by CL 301607174: some Session subclasses may not have implementation for Run() with thread pool options. (commit: ff10e691a753011b17208c98eeafaa52c002794c)
• Inline deps variables in BUILD files. (commit: dfc7a5c3de9f2621715cbc37588bac72eb33b79d)
• Update bazel version used in the docker images. (commit: 162f72949c6ecbe9e610182c923dec0aa5924cf2)
• Bump TF version. (commit: 6d6792cd172a486ca4a568455c182f944c47c669)
• Bump TF version. (commit: 08a1ecf9429ae4811dbeb0f255cad08cf746ab73)
• Small performance optimization. (commit: 4196a724f2669df31d0a11848cb51d64273a9fe5)
• Update version for 2.2.0-rc0 release. (commit: ecf3328e5b1ea1e3d2fae312a737155eb3f5ec80)
• Update version for 2.2.0-rc1 release. (commit: 6e254ffa92fe536269519a529fcb2f2f0ee1db3c)
• Update version for 2.2.0-rc2 release. (commit: 3ad9a987b79eaa5315eb44874f6c07ef243d4fc0)
• Format error strings correctly in JSON response (Fixes #1600). (commit: 1ff4d31cd9a0a736162813c149139cce0ccaaa2c)
• Merge pull request #1615 from peakji:master (commit: 22635f61110045f27c433a5fcf336138f86df4a6)
• Fix broken GetModelMetadata request processing (#1612) (commit: 55c40374b548b89e8de6d899ef2b0b355c0fa9e5)
• Fix icu dependency checksum. (commit: 322c9ee6ad9eed2765beaeb2fc884547d1ebfbaa)
• Support Python 3.7 in tensorflow-serving-api package (Fixes #1640) (commit: f775bb25e80a6c7b3c66842eb9085d44d9752ec2)
• Update ICU library to include knowledge of built-in data files. (commit: 774f2489384cf985c534298d1303474c268efe5c)
• Update TF Text regression model to catch errors thrown from within ops. (commit: 11e49b09f18919a9a0914b3ac8795854f6d58b2d)
• Adds storage.googleapis.com as the primary download location for the ICU, and resets the sha256 to match this archive. (commit: 028d05095c4e302c06096e5ea32917718828ea47)
• Upgrade to CUDA Version 10.1. (commit: 3ab70a7811f63b994da076e2688ccc66feccee96)
• Update TF Text to v2.2.0. (commit: fd9842816eddb4782579eadd119156190d6d2fec)
• Update version for 2.2.0 release. (commit: 066c50a189f647bef3d813f2f765b47ab53085c3)
• Install python3-distutils in gpu docker config. (commit: 145b3374b536fe8c12c14ce2ca5765acfa69dff6)
• Install cuda-nvrtc 10.1 into release tf serving gpu docker image. (commit: 9b74939d6bd45486a3e10168f9863a33e6a488eb)
• Specifying compute capabilities in the TF 2.2 .bazelrc file to reduce startup time on GPUs; this is in line with TF 2.3 changes (commit: 54de1065213e5a8a888203a2795eb556881fc08a)
• Update version for 2.2.2 release. (commit: d0b8fdf8cd1f5a31c698e51c62ce80e9ce0ffdac)
• Update .bazelrc (commit: 3a9b6971540e43c430fc08d979af0f590d08668b)
• Update version for 2.2.3 release. (commit: e89e0f51f403b52edc2c474546aeb4f09b874535)

Thanks to our Contributors

This release contains contributions from many people at Google, as well as:

broken, Haiming, Minglotus-6, rhaertel80, Shyam Saladi, Tensorflow-Copybara

TensorFlow Serving using TensorFlow 2.6.0-rc0.

Major Features and Improvements

No major features or improvements.

Breaking Changes

No Breaking changes.

Bug Fixes and Other Changes

• Enable tensor output filters with batching. (commit: 448dbe14624538ab76fd6aeb2a456344e7f41c78)
• Update tf.io import in warmup example doc. (commit: 6579d2d056530565cd6606a39c82b2f6c1d3799e)
• Rollback incompatible C++17 changes. (commit: ba0fa72b61bc2c42388b815253ba72e6830f03cf)
• Roll forward with std::optional -> absl::optional. (commit: 009dac683bf84165b84702d177bb9a021ebef728)
• Replace STL algorithm call with a container method (performance-inefficient-algorithm). (commit: f5bc09998e0043ce72d34b14104379163048406c)
• Internal change. (commit: 61ca084e5823977973bf31e72841fc423a8f8027)
• Remove unused lambda capture. (commit: c80db5ab64f378420c00b52b773cfcea712b8f8d)
• Use absl::StrContains instead of find() != npos. (commit: 9b3d4d7c8ab650925e081d295b51bf96c1c89d9d)
• Remove unused "using" decl. (commit: ffcc4a16c76c4fa1189edd9682fc486442a33e52)
• Resize tensors if the dimensions of the tflite and tensorflow inputs mismatch, even if the number of elements are the same (commit: 8293f44bd5c5ecc68636cd0d036234f891d29366)
• Make Internal Change. (commit: 12bbcedb6a30f66c4bf05816aabdac977e0871b4)
• Move status_proto to public visible apis/ (it being used by public API protos) (commit: 7f894c79fce5e58758f3cb49e858a16e3602ae80)
• Move core/logging.proto -> apis/logging.proto (commit: 37c64b8820a923aafc1b5c8bf264fd5cce5224f3)
• Update TF Text to v2.5.0. (commit: 48e5a9e23a1e0b2951b77c3e8f9832193d9b1851)
• Fix argument name vs comment mismatch (commit: c9bd6717e0efdeda241599279c4f8f25e887549d)
• cleanup unused vars+functions. (commit: 4ce9c6b98753cbdcffbacf3767ae2157eb664069)
• Internal change. (commit: f2958313f4a29bc107fb345d04a2c1788beb743a)
• Internal change. (commit: ddc342189e506d3c0500f9a01d831cb2df9589f0)
• Remove unused lambda captures in loader_harness_test.cc (commit: bb34b1d4b221d2ea02f854a3d84a64c7bcab56e2)
• Avoid using 'version' after it was moved. (commit: 23b0755182aca025a6594f1461d3a57baab5889a)
• Adding python targets for config protos (commit: 757c3a6b6c8a03731dc73ff758f69a61aeddcf67)
• Add basic batch scheduler for tflite models to improve batch parallelism. (commit: 0ffd6080437ca8175b067be7cc00f5b3df9ea92a)
• Remove experimental tags from uses of gRPC C++ callback API. (commit: 97aa2db6dd0819dadf07e4bd2e8e49ca39c6c6f9)
• Remove experimental tags from uses of gRPC C++ callback API. (commit: b355023b034ca6ef72b507920d9de2a02e0f4a2a)
• Reserve Unavailable error to TF communication ops only. (commit: db9aca187affd0453627a1729916acfea98ae800)
• Add the flag thread_pool_factory_config_file to model server and fix a typo. (commit: efc445f416f8cb20606ca0d2aaf44c13fae7ea4c)
• Add new --use_alts_credentials flag, to enable building secure credentials using Google ALTS. (commit: ceba636bb7c5c98bde35d1818fd033b36300fffe)
• Enable HTTP PATCH support in evhttp_server (commit: 6cbc4a9eb419c8078c3a4e791381cda70dd8fc78)
• Make Internal Change (commit: 6a54e2a806cdae494e302b28318dec2f14892a06)
• Update version for 2.6.0-rc0 release. (commit: ae853e4175074c82fff04d15952b3095ff662375)

Thanks to our Contributors

This release contains contributions from many people at Google, as well as:

nniuzft

Release 2.1.4

Major Features and Improvements

Breaking Changes

Bug Fixes and Other Changes

• Fix link for TFRecord in Saved Model Warmup documentation. (commit: 127a112a91bda3d7d3c3a56802632376bbe3e36e)
• Be able to discard aspired-versions request from SourceRouter (commit: 10e4987502ee91fe74c6c179ed4ba52f17cc75b4)
• Fix copying of string tensor outputs by explicitly copying each (commit: 9544077bdb6eef9b20a0688a042155ee6dea011a)
• Adds integration with TensorBoard profiler service (commit: 91adea9716b57dd58427714427b944fbe9b3f89e)
• Migrate from std::string to tensorflow::tstring. (commit: e24571ac9ce390733f3b02188c7d740f08fff62d)

Thanks to our Contributors

This release contains contributions from many people at Google.