Releases: onnx/tensorflow-onnx
Releases · onnx/tensorflow-onnx
v1.16.1
V1.16.0
What's Changed
- Fix CI nightly test issues. by @fatcat-z in #2229
- Bump ONNX 1.14.1 by @jcwchen in #2220
- Support tf ResizeArea op. by @fatcat-z in #2241
- Keep activations in bidirectional LSTM (fixes: #2211) by @MaximeChurin in #2234
- Support bool inputs for MatrixBandPart by @Gerstenberger in #2269
- Fix unsupported ops TF 2.14.0: OnesLike by @Gerstenberger in #2270
- Add a condition when remove a transpose node in TransposeOptimizer. by @fatcat-z in #2272
- Change the way to set node name. by @fatcat-z in #2278
- Enable support to ONNX 1.15. by @fatcat-z in #2279
- Adjust CI tests coverage in different pipelines. by @fatcat-z in #2281
- Enable support to ONNXRUNTIME 1.16.3. by @fatcat-z in #2283
A huge thank you to our community for this release!
Full Changelog: v1.15.1...v1.16.0
Contributors
MaximeChurin, Gerstenberger, and 2 other contributors
Assets 2
v1.15.1
v1.15.0
What's Changed
- Fix None shape error when one input to ConcatV2 has a shape. by @klasma in #2135
- Change Equal 11 for string input by @mikeessen in #2149
- Update tf_loader.py to remove unnecessary slices by @DLumi in #2159
- Fix fusing of Conv and Mul in TransposeOptimizer (#2088) by @cosineFish in #2152
- Support for more TensorScatter* operations by @javidcf in #2179
- Added support for int64 -> string CategoryMapper updated by @pagran in #2181
- Fixed group attribute in convolution op by @javidcf in #2090
- Fix a bug that multiple (conv, batch_norm) ops could not be optimized. by @fatcat-z in #2187
- Support bitwise ops by @f-salvetti in #2192
- Update parent graph in loop "cond" subgraphs by @f-salvetti in #2201
- Fix wrong shapes in loop body inputs if shape invariances are set in TF by @f-salvetti in #2203
- Support unique_with_counts by @f-salvetti in #2195
- Some minor fixes for keras-resnet50 tutorial by @shreyas-bk in #2207
- Use TFLite Interpreter's public instead of private API to get tensor details by @q-ycong-p in #2204
- Support latest version of TensorFlow (both 2.12 and 2.13) and ONNX opset 18. by @fatcat-z in #2216
A huge thank you to our community for this release!
Full Changelog: v1.14.0...v1.15.0
Contributors
javidcf, klasma, and 8 other contributors
Assets 2
tf2onnx-1.14.0
What's Changed
- Update the dtype since latest version of Numpy has removed str. by @fatcat-z in #2112
- Fix Infinite Loop in constant folding by @ilyasher in #2119
- Update tf2onnx to support TensorFlow 2.11. by @fatcat-z in #2131
- Allow --extra_opset to accept a list of extra opsets. by @fatcat-z in #2136
- Fix the axis of inserted QDQ for ConvTranspose by @sdua-nv in #2134
- Adding tf.strings.reduce_join mapping by @fsalv in #2091
- A fix for seed attribute in the keras random normal generator by @inonbe in #2126
- Fix ResizeBilinear and ResizeNearestNeighbor and update related tests by @cosineFish in #2130
- Update Resize (opset 11) layer to support scales option when dims are… by @qant-um in #2137
- Fix issue #2102: Set reduction axis of mean to height and width for tf adjust_contrast op by @cosineFish in #2140
- support cumprod op by @f-salvetti in #2133
- Support ONNXRUNTIME 1.14 and opset 18. by @fatcat-z in #2143
- Change the default opset to 15. by @fatcat-z in #2145
A huge thank you to our community for this release!
ilyasher
sdua-nv
fsalv
inonbe
cosineFish
qant-um
f-salvetti
Full Changelog: v1.13.0...v1.14.0
Contributors
qant-um, cosineFish, and 6 other contributors
Assets 2
tf2onnx-1.13.0
What's Changed
- Update CLI reference doc, adjust perameters order and add output-as-nchw by @hwangdeyu in #2028
- Update the way to get opset version for running tests. by @fatcat-z in #2033
- Update left tests in test_subclassing for the opset version issue. by @fatcat-z in #2034
- Fix Select op with Mul inf input to nan error by @hwangdeyu in #2035
- Improve several issue templates by @hwangdeyu in #2043
- Upgrade the required version for flatbuffers. by @fatcat-z in #2050
- Remove extract_sub_graph function from tf_optimize method. by @fatcat-z in #2051
- Replace distutils.version with packaging.version namespace. by @fatcat-z in #2056
- Add Support for MatMulInteger by @GregoryMorse in #2072
A huge thank you to our community for this release!
Full Changelog: v1.12.1...v1.13.0
Contributors
GregoryMorse, hwangdeyu, and fatcat-z
Assets 2
tf2onnx-1.12.1
Summary
- Enabling opset 17 support.
- Focusing on bug fix.
What's Changed
- ONNX opset 17 with IR version 8 support by @hwangdeyu in #2014
- Upgrade test opsets and remove deprecated numpy and version usage by @hwangdeyu in #2018
- Turn on graph tf optimize grappler dependency by @hwangdeyu in #2020
- Fix output node in rewrite random normal by @hwangdeyu in #2023
- Skip failed RFFT tests for TF-2.4.x by @q-ycong-p in #2026
- Add lstm sigmoid dropout test for dependency optimization by @hwangdeyu in #2027
- Update CLI reference doc, adjust perameters order and add output-as-nchw @hwangdeyu in #2028
- Fix Select op with Mul inf input to nan error by @hwangdeyu in #2035
A huge thank you to our community for this release!
Contributors
hwangdeyu and q-ycong-p
Assets 2
tf2onnx-1.12.0
Summary
- added support for python 3.10, onnx 1.12, onnxruntime 1.12, tensorflow 2.9
- added --outputs_as_nchw option to transpose output to from nhwc to nchw
- enhanced optimizer and exception handing
- improved keras pattern and ci
- misc fixes
What's Changed
- Transpose optimization for Softmax and LogSoftmax (fixes #1716) by @janbernloehr in #1964
- The from_tflite() function should accept None as default value of input_names and output_names. by @fatcat-z in #1967
- Change Loop op with maximum iterations input M equals to empty string by @hwangdeyu in #1971
- Update README.md by @andife in #1976
- Update CONTRIBUTING.md by @andife in #1978
- Update Keras related tests to support latest TF version. by @fatcat-z in #1980
- Fix a test issue in keras2onnx_application_tests.yml. by @fatcat-z in #1982
- L2_NORMALIZATION support for tflite by @shesung in #1989
- Replace deprecated
np.objectwithobjectby @vvolhejn in #1990 - Implement --outputs_as_nchw option to transpose output to from nhwc to nchw by @hwangdeyu in #1979
- Fix transpose split optimize attr when opset >=13 by @hwangdeyu in #1996
- Skip existing const node in _parse_graph_input by @q-ycong-p in #2000
- Add handling of HardSigmoid recurrent activation for Keras LSTM by @q-ycong-p in #2001
- Add more tests for tf 2.9.x into CI pipelines. by @fatcat-z in #2009
- Fix problem with adding more than one tf.newaxis at the same time by @southfreebird in #2007
- Improve ZerosLike implementation and optimize for opset >= 9 by @hwangdeyu in #2003
- Add newly required dependencies for latest ORT version. by @fatcat-z in #2012
- Add support for Python310 and ORT 1.12 by @hwangdeyu in #1975
- Increment main to 1.12, TF to 2.9 by @hwangdeyu in #2015
A huge thank you to our community for this release!
janbernloehr, andife , shesung, vvolhejn, southfreebird
Full Changelog: v1.11.1...v1.12.0
Contributors
janbernloehr, shesung, and 6 other contributors
Assets 2
tf2onnx-1.11.1
Summary
- added support for onnx opset 16, upgraded default opset to 13
- removed python 3.6 check from CI tests
- enhanced api and added new ops
- misc fixes
Changes since tf2onnx-1.10.1
- Minor doc tf 2.8 change by @hwangdeyu in #1934
- Use make_tensor_sequence_value_info instead of a deprecated function by @jcwchen in #1930
- Update tensorflowjs 3.17 tests by @hwangdeyu in #1942
- Bug fix for concatenating node instead of str. by @FrankD412 in #1933
- Remove python 3.6 support, upgrade CI by @hwangdeyu in #1940
- Add opset 16 support and check ci by @hwangdeyu in #1937
- Update the default opset version for tf2onnx. by @fatcat-z in #1946
- Update the way to check input_signature in from_function(). by @fatcat-z in #1947
- Add mapping for TF lite op TFL_BATCH_MATMUL and TFL_MATMUL. by @fatcat-z in #1950
- Update protobuf version in ci by @hwangdeyu in #1951
- Add TensorScatterAdd op for opset 16 by @hwangdeyu in #1949
- Add a new api from_tflite to improve user experience. by @fatcat-z in #1954
- Upgrade opset 16 version related doc by @hwangdeyu in #1953
- Remove useless sync workflow. by @fatcat-z in #1955
Full Changelog: v1.10.1...v1.11.1
A huge thank you to our community for this release!
Contributors
FrankD412, hwangdeyu, and 2 other contributors
Assets 2
tf2onnx-1.10.1
Summary
- added support for tf-2.8
- enhanced optimizer
- added custom op example and load_op_libraries in doc
- fixed ci to support ort version 1.11
- misc fixes
Changes since tf2onnx-1.9.3
- Update opset 15 version supported ops doc by @hwangdeyu in #1773
- add tf-2.7 in readme by @guschmue in #1772
- Add support for one hot encoding of 0-rank tensor by @LoicDagnas in #1771
- Fix onnxruntime-nightly-unittest-matrix CI failed by @hwangdeyu in #1784
- Handle 'Neg' nodes in TransposeOptimizer by @jan-haug in #1785
- Fix tfjs CI failure and add tfjs convert log by @hwangdeyu in #1797
- Fix misspelling by @mshr-h in #1800
- Skip the blob failed model tests by @hwangdeyu in #1806
- Add unit32 unit64 type support by @hwangdeyu in #1808
- Fix keras Conv2D BiasAdd fuse by @hwangdeyu in #1796
- Add Rint op conversion by @hwangdeyu in #1833
- Specify the version scope of tensorflow-text. by @fatcat-z in #1843
- Add used opset into opset_imports argument to create an onnx model. by @fatcat-z in #1855
- feat: For QDQ models, disable constant folding and ensure the conv weights are transposed without using an explicit op by @peri044 in #1856
- Enable tensorflow 2.8.0 in CI pipeline. by @fatcat-z in #1863
- Remove --fold_const parameter by @hwangdeyu in #1861
- Add load_op_libraries parameter in docs by @hwangdeyu in #1853
- Force to upgrade typing-extensions to a version above 4.0.0. by @fatcat-z in #1873
- Fix onnx-weekly CI job failure. by @fatcat-z in #1874
- Add tf custom op conversion example by @hwangdeyu in #1878
- Upgrade Windows image to windows-2019. by @fatcat-z in #1879
- Rename the default branch to main by @hwangdeyu in #1880
- Add onnxruntime 1.10 into normaly run. by @fatcat-z in #1887
- Change package author by @garymm in #1885
- Create ado-sync-issue.yml by @fatcat-z in #1892
- Upgrade some keras jobs to use ort 1.11.0. by @fatcat-z in #1896
- Fix nightly CI job failures caused by numpy version. by @fatcat-z in #1898
- Add commit value to onnx file and clean up some code by @hwangdeyu in #1897
- Fix keras bidirectional merge failures by @kota-row in #1869
- Keras nightly CI jobs update. by @fatcat-z in #1904
- Add the absolute path condition when loads the tf custom op library by @hwangdeyu in #1909
- Change custom op example to new API and add the doc by @hwangdeyu in #1883
- Fix split transpose optimize axes issue by @hwangdeyu in #1918
- Replace the description exporting to conversion by @hwangdeyu in #1917
- Allow passing optimizers to from_keras conversion function by @shelkesagar29 in #1907
- Add split transpose optimizer tests by @hwangdeyu in #1923
Full Changelog: v1.9.3...v1.10.1
A huge thank you to our community for this release!
LoicDagnas, jan-haug, mshr-h, peri044, kota-row, shelkesagar29
Contributors
garymm, kota-row, and 8 other contributors