Onnx2tf

- `Exxxx` op - `Ixxxx` op
- pre_process_transpose
- post_process_transpose.
- `ConvTranspose`
- Support for `group > 1`
- [Problem in padding in ConvTranspose when stride is larger than 1 56](https://github.com/PINTO0309/onnx2tf/issues/56)

What's Changed
* pre_process_transpose, post_process_transpose. `E-I` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/66


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.2.23...1.2.24

- `Axxxx` op - `Dxxxx` op
- pre_process_transpose
- post_process_transpose
- `ConvTranspose`
- Direct replacement process for `ConvTranspose` to `tf.nn.convXd_transpose`. This is not an essential measure, but by specifying a parameter in the TensorFlow specification, it directly generates a TensorFlow `tf.nn.convXd_transpose` OP, ignoring the automatic transformation process.
https://www.tensorflow.org/api_docs/python/tf/nn/conv1d_transpose
https://www.tensorflow.org/api_docs/python/tf/nn/conv2d_transpose
https://www.tensorflow.org/api_docs/python/tf/nn/conv3d_transpose
- [Problem in padding in ConvTranspose when stride is larger than 1 56](https://github.com/PINTO0309/onnx2tf/issues/56)

What's Changed
* pre_process_transpose, post_process_transpose. `A-D` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/65


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.2.22...1.2.23

- `Add`, `Div`, `Mul`, `Sub`
- https://github.com/PINTO0309/onnx2tf/releases/download/1.1.28/sinet_320_op.onnx
Shrink a tensor whose `input_tensor_1` and `input_tensor_2` have the same rank and all but one dimension is 1. This is a workaround to avoid the pattern of failure of `explicit_broadcast` in subsequent processing.
https://github.com/gmalivenko/onnx2keras/issues/84

e.g.
input:
input_tensor_1: [1,80,80,12]
input_tensor_2: [1,12,1,1]

output:
input_tensor_1: [1,80,80,12]
input_tensor_2: [12]


What's Changed
* pre_explicit_broadcast. `Add`,`Div`,`Mul`,`Sub` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/64


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.2.21...1.2.22

- `Flatten`
- Support for undefined batch size

What's Changed
* Support for undefined batch size `Flatten` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/63


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.2.20...1.2.21

- `Transpose`
- Apply a workaround that avoids as much as possible that `Transpose` with more than 6 dimensions is converted to `FlexTranspose`
- fast_acvnet_generalization_opset16_192x320
1. https://github.com/google-coral/edgetpu/issues/274
2. https://github.com/tensorflow/tensorflow/issues/53702
- Before - 6D FlexTranspose
- After - 6D Transpose

What's Changed
* Apply a workaround that avoids as much as possible that `Transpose` with more than 6 dimensions is converted to `FlexTranspose` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/62


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.2.19...1.2.20

1. Addition of parameter replacement function. `Resize`, `Tile`, `Concat`
2. Added automatic transposition function to NHWC in `Resize`
3. Bug fixes for `ReduceL1`, `ReduceL2`

What's Changed
* Addition of parameter replacement function and added automatic transposition function to NHWC in `Resize` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/61


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.2.18...1.2.19