Onnx2tf

- `Squeeze`
- Corrected a mistake in the decision statement
- [Cannot reshape a tensor with 41835024 elements to shape [3234,4] (12936 elements) 217](https://github.com/PINTO0309/onnx2tf/issues/217)

What's Changed
* Corrected a mistake in the decision statement by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/218


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.10...1.7.11

- `MatMul`
- Fixed a bug with insufficient transposition of accuracy checks.
- [tensorflow.python.framework.errors_impl.InvalidArgumentError: Exception encountered when calling layer 'tf.reshape_33' (type TFOpLambda) 213](https://github.com/PINTO0309/onnx2tf/issues/213)

What's Changed
* `MatMul` Fixed a bug with insufficient transposition of accuracy checks. by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/215


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.9...1.7.10

- `Gather`
- Support for substitution of `Gather` attribute parameter `axis`
- [Converting an ONNX model with static output to TfLite leads to a model that is not supported by the TfLite Java API 210](https://github.com/PINTO0309/onnx2tf/issues/210)

What's Changed
* Support for substitution of `Gather` attribute parameter `axis` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/214


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.8...1.7.9

- `BatchNormalization`
- Supports broadcast of parameters.
- `pre_explicit_broadcast`
- `explicit_broadcast`
- https://github.com/Vegetebird/MHFormer
- https://github.com/PINTO0309/PINTO_model_zoo/tree/main/355_MHFormer

https://user-images.githubusercontent.com/33194443/221724997-00d2e0da-bd3d-4238-85b5-68819e067cd3.mp4

- `BatchNormalization`, `MatMul`
- Added `np.ndarray` conversion process by `tf.convert_to_tensor` to prevent `saved_model` bloat

- CI
- MHFormer

- Correct existence check condition of `verification_data`.
- [Converting an ONNX model with static output to TfLite leads to a model that is not supported by the TfLite Java API 210](https://github.com/PINTO0309/onnx2tf/issues/210)

What's Changed
* Supports broadcast of parameters. `BatchNormalization` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/209
* Correct existence check condition of `verification_data`. by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/212


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.7...1.7.8

- `Softmax`
1. Skip error check if onnx and tensorflow input geometry is an exact match and `axis` is exactly the same.
2. Also, skip the process of forced replacement of `axis`.
3. This modification increases the speed of model conversion by skipping the unnecessary accuracy check process.
- With this modification, the output of `ShadowFormer` matches perfectly.
- [[ShadowFormer] Unnecessary Transpose extrapolated immediately after Softmax 205](https://github.com/PINTO0309/onnx2tf/issues/205)

What's Changed
* Skip error check if onnx and tensorflow input geometry is an exact match and axis is exactly the same. Also, skip the process of forced replacement of axis. by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/206


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.6...1.7.7

- Fixed a problem with `broadcast` and `shared_axes` calculations aborting when the following pattern `PReLU` is used.

input_tensor.shape: [1,2,3,4]
slope.shape: [1,1,1]
shared_axes: [1,2,3]

- `Conv`
- Move workaround processing of shape mismatch errors to before padding is performed.
- https://github.com/PINTO0309/PINTO_model_zoo/tree/main/353_ShadowFormer

What's Changed
* 1. Support for `PReLU` special patterns. 2. Move workaround processing of shape mismatch errors to before padding is performed. `Conv` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/204


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.5...1.7.6