Onnx2tf

1. Fixed a fatal bug in `MatMul` shape unmatch error avoidance handling.
2. Fixed fatal bug in axes correction process for `ReduceXX` with `opset=18,17,16,15,14,13,12,11`.
- There is no compatibility guaranteed between `ReduceSum` and other `ReduceXX` processes due to differences in opset.
- https://github.com/onnx/onnx/blob/main/docs/Operators.md#aionnx-default
- https://github.com/fateshelled/unimatch_onnx
- https://github.com/PINTO0309/onnx2tf/releases/download/1.1.28/gmflow-scale1-mixdata-train320x576-4c3a6e9a_1x3x480x640_bidir_flow_sim.onnx

What's Changed
* 1.Fixed a fatal bug in `MatMul` shape unmatch error avoidance handling. 2.Fixed fatal bug in axes correction process for `ReduceXX` with `opset=18,17,16,15,14,13,12,11`. by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/201


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.4...1.7.5

- Re-release `flatc` with some customizations of our own to address the lack of arithmetic precision of the quantization parameters.
- [[TODO] Re-release flatc with some customizations of our own to address the lack of arithmetic precision of the quantization parameters](https://github.com/PINTO0309/onnx2tf/issues/196)
- [tflite to JSON to tflite quantization error 1](https://github.com/PINTO0309/tflite2json2tflite/issues/1)

What's Changed
* Re-release `flatc` with some customizations of our own to address the lack of arithmetic precision of the quantization parameters by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/199


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.3...1.7.4

1. Improved processing stability of `Gather`
2. Corrected error judgment condition of `noop_with_empty_axes` in `RedeceXX`.
3. Fixed a bug that parameter replacement specification in `Reshape` OP was ignored.
4. Dealing with the absence of bias in `Gemm`.

What's Changed
* 1. Improved processing stability of `Gather` and corrected error judgment condition of `noop_with_empty_axes` in `RedeceXX`. 2. Fixed a bug that parameter replacement specification in `Reshape` OP was ignored. 3. Dealing with the absence of bias in `Gemm`. by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/197


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.2...1.7.3

- `FlexTranspose` generation suppression and `Transpose` dimensionality reduction applied to almost all `Transpose`
- `onnx2tf -i xxxx.onnx -nodafc 2`
- `Concat`
- `Constant`
- `Conv`
- `ConvTranspose`
- `Flatten`
- `GatherElements`
- `GridSample`
- `Hardmax`
- `InstanceNormalization`
- `MatMul`
- `NonMaxSuppression`
- `Reshape`
- `Resize`
- `ScatterElements`
- `common_functions`
- [Remove transpose before reshape 193](https://github.com/PINTO0309/onnx2tf/issues/193)

What's Changed
* `FlexTranspose` generation suppression and `Transpose` dimensionality reduction applied to almost all `Transpose` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/195


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.1...1.7.2

- Added a process to suppress the creation of `FlexTranspose` for extrapolation of the `Transpose` immediately before the `Reshape`
- If the `-nodafc` option has a value less than `6`, compress `Transpose` to the specified numeric dimension.
- If `-nodafc` option is set to `2`, `Transpose` is completely lost.
- [model_2.onnx.zip](https://github.com/PINTO0309/onnx2tf/files/10769390/model_2.onnx.zip)


onnx2tf -i model_2.onnx -nodafc 2


- [Remove transpose before reshape 193](https://github.com/PINTO0309/onnx2tf/issues/193)

What's Changed
* Added a process to suppress the creation of `FlexTranspose` for extrapolation of the `Transpose` immediately before the `Reshape` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/194


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.0...1.7.1

[Note] This release implements an extensive additional mechanism that will be needed in the future to avoid post-conversion accuracy errors, so problems may occur that were not reproduced in v1.6.7. If you have any inconvenience in using the tool, I recommend using v1.6.7.

- [Experimental / WIP] Generate a miniscule model for each OP only, retain the results of test inference, and propagate to all OPs.
- [x] First step
- [x] `Softmax`
- [x] `Sub`
- [x] `ReduceMax`
- [x] `Reshape`
In the figure below, the output from the first `Reshape` only model is propagated to `ReduceMax` and `Sub`. Next, a `ReduceMax` only model is generated and test inference is performed based on the inference values propagated from the `Reshape` model.
- In other words, it generates models for only one OP each and stores all test inference results for each OP alone.
- By using a method that sequentially stores inference results in a model with only one OP, it becomes easy to backward from the point where an inference accuracy error is detected and rerun the inference test from the OP where the problem is assumed to have occurred.
- [x] Second step
- To attempt to speed up the inference checks in [ddnm.onnx](https://s3.ap-northeast-2.wasabisys.com/temp-models/onnx2tf_175/ddnm.onnx), the OPs in the following list will be addressed.
- Eliminate `Softmax`'s redundant multiple inference accuracy correction process.
- Implement on a trial basis in the following OPs for verification.
- [x] `Add`
- [x] `Concat`
- [x] `Conv`
- [x] `Cos`
- [x] `Gemm`
- [x] `InstanceNormalization`
- [x] `MatMul`
- [x] `Mul`
- [x] `Div`
- [x] `Mod`
- [x] `ReduceMax`
- [x] `Reshape`
- [x] `Resize`
- [x] `Sigmoid`
- [x] `Sin`
- [x] `Softmax`
- [x] `Sub`
- [x] `Transpose`
- [x] `Unsqueeze`

ssc4onnx -if ddnm.onnx
┏━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ OP Type ┃ OPs ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Add │ 141 │
│ Concat │ 19 │
│ Conv │ 120 │
│ Cos │ 1 │
│ Gemm │ 34 │
│ InstanceNormalization │ 71 │
│ MatMul │ 12 │
│ Mul │ 145 │
│ Reshape │ 166 │
│ Resize │ 5 │
│ Sigmoid │ 67 │
│ Sin │ 1 │
│ Softmax │ 6 │
│ Transpose │ 12 │
│ Unsqueeze │ 65 │
│ ---------------------- │ ---------- │
│ Total number of OPs │ 865 │
│ ====================== │ ========== │
│ Model Size │ 433.8MiB │
└────────────────────────┴────────────┘

- [ ] Third step - Will be separated into separate pull requests
- Add a process to perform the calculation of the maximum absolute error on the inference results of all OPs.
- [ ] Fourth step - Will be separated into separate pull requests
- Implement propagation of inference results for all operations other than the above.
- [ ] Fifth step - Will be separated into separate pull requests
- Instead of a processing format for inferring TensorFlow models, primitive operations are replaced with Numpy operations to make the process as fast as possible.
- [ ] Sixth step - Will be separated into separate pull requests
- Seek ways to eliminate accuracy errors without creating JSON files.
- https://github.com/PINTO0309/onnx2tf/issues/175

- TF v2.12.0rc0
1. `Transpose` now supports 6D tensors.
2. Support for `Atan2`.

- Support for `keras_v3` format