* Upgrade CI pipelines to support latest TF and ONNX runtime.

Signed-off-by: Jay Zhang <[email protected]>

* Upgrade CI pipelines to support latest TF and ONNX runtime.

Signed-off-by: Jay Zhang <[email protected]>

* Update the dtype since latest version of Numpy has removed str.
* Set Numpy to a fixed version in CI.
* Update the protobuf's version in CI.
* Keep h5py to a lower version for tf 1.15 in CI.

Signed-off-by: Jay Zhang <[email protected]>

This commit fixes a bug in the function compute_const_folding_using_tf
where an already-folded StridedSlice node could set the 'update'
flag to True, causing an infinite loop.

A regression test was also added. The test checks for the infinite loop
by setting a timeout using the timeout-decorator module from pypi.

Signed-off-by: Ilya Sherstyuk <[email protected]>
Co-authored-by: Ilya Sherstyuk <[email protected]>

…e. (#2122)

* Install onnxconverter-common in the beginning.
* Set onnxconverter-common to latest version.

Signed-off-by: Jay Zhang <[email protected]>

* Add tflite updates caused by tf 2.11.
* Fix SIGN op issue for tf2.11.
* Remove unused attribute of ConvTransposeWithDynamicPads op.
* Add CI tests for tf 2.11.0.

Signed-off-by: Jay Zhang <[email protected]>

Signed-off-by: Jay Zhang <[email protected]>

The quantization axis of QDQ nodes that are being inserted
before the kernel weights of all Conv nodes is currently 0.
This is incorrect; ConvTranspose requires axis=1.

Signed-off-by: Sirej Dua <[email protected]>
Co-authored-by: Jay Zhang <[email protected]>

* Add support to tf.strings.reduce_join
* Support for ReduceJoin op
* unit test for reduce_join

Signed-off-by: Salvetti, Francesco <[email protected]>

* A fix for seed field in the tf.keras random normal generator. The seed field was not passing to the converted onnx model due to that it exists in seed2 attribute instead of seed and its type is Integer and not float.

---------

Signed-off-by: inonbe <[email protected]>
Co-authored-by: Jay Zhang <[email protected]>

…#2130)

1) Fix tf ResizeBilinear and ResizeNearestNeighbor op with align_corners=True or half_pixel_centers=True
2) Update related tests to make input data include x.5 and add tests for align_corners=True

Signed-off-by: cosine <[email protected]>
Co-authored-by: Jay Zhang <[email protected]>

… defined (#2137)

Signed-off-by: Quentin Muller <[email protected]>
Co-authored-by: Jay Zhang <[email protected]>

…djust_contrast op (#2140)

Signed-off-by: cosine <[email protected]>
Co-authored-by: Jay Zhang <[email protected]>

* support cumprod op

---------

Signed-off-by: Francesco Salvetti <[email protected]>
Co-authored-by: Jay Zhang <[email protected]>

1. Update Split attributes for opset 18.
2. Update all Reduce* ops for opset 18.
3. Update transpose_optimizer for opset 18.
4. Enable opset 18 in CI tests.
---------

Signed-off-by: Jay Zhang <[email protected]>

* Change the default opset to 15.
* Update the ONNX version for one CI test.

Signed-off-by: Jay Zhang <[email protected]>

* Update version information for a new release 1.14.0.

---------

Signed-off-by: Jay Zhang <[email protected]>