Onnx2tf

- Fixed a fatal bug that disabled parameter substitution when ONNX OP name starts with `/`.
- The problem occurred only when the `-osd` option was specified.
- [[DDNM] Support additional parameters to onnxsim 175](https://github.com/PINTO0309/onnx2tf/issues/175)

What's Changed
* Fixed a fatal bug that disabled parameter substitution when ONNX OP name starts with "/". by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/177


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.6.0...1.6.1

- Added `--no_large_tensor`, `-nlt` option for onnxsim

no_large_tensor: Optional[bool]
Suppresses constant bloat caused by Tile OP when optimizing models in onnxsim.\n
See: https://github.com/daquexian/onnx-simplifier/issues/178

- [Support additional parameters to onnxsim 175](https://github.com/PINTO0309/onnx2tf/issues/175)

What's Changed
* Added `--no_large_tensor` option for onnxsim by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/176


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.5.45...1.6.0

- `InstanceNormalization`
- Significantly revised `InstanceNormalization` process
- Limited to 3-D and 4-D processing
- Always transpose to NHWC before processing
- https://github.com/microsoft/onnxruntime/blob/main/onnxruntime/core/providers/cpu/nn/instance_norm.cc
- Ref: https://github.com/josephrocca/anime-gan-v2-web
- [[AnimeGanV2] InstanceNormalization accuracy is significantly low. 173](https://github.com/PINTO0309/onnx2tf/issues/173)

What's Changed
* Significantly revised `InstanceNormalization` process by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/174


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.5.44...1.5.45

- `Pad`
- Implement padding process with dynamic tensor
- https://tfhub.dev/tensorflow/lite-model/efficientdet/lite2/detection/default/1?lite-format=tflite
- https://github.com/PINTO0309/onnx2tf/releases/download/1.1.28/efficientdet_lite2_detection_1.onnx
- https://discuss.tensorflow.org/t/convert-efficientdet-lite-2-to-mlmodel/14542

bash
https://tfhub.dev/tensorflow/lite-model/efficientdet/lite2/detection/default/1?lite-format=tflite

python3 -m tf2onnx.convert \
--opset 12 \
--tflite lite-model_efficientdet_lite2_detection_default_1.tflite \
--output efficientdet_lite2_detection_1.onnx \
--dequantize \
--inputs-as-nchw serving_default_images:0

onnxsim efficientdet_lite2_detection_1.onnx efficientdet_lite2_detection_1.onnx
onnxsim efficientdet_lite2_detection_1.onnx efficientdet_lite2_detection_1.onnx
onnxsim efficientdet_lite2_detection_1.onnx efficientdet_lite2_detection_1.onnx

onnx2tf -i efficientdet_lite2_detection_1.onnx -osd


What's Changed
* [EfficientDet-lite2] Implement padding process with dynamic tensor by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/171


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.5.43...1.5.44

- `ConvTranspose`
- Shape Unmatch Error Mitigation Measures
- Search for and transpose shapes that do not cause shape unmatch errors
- [When Onnx Matmul inputs have different dimension 133](https://github.com/PINTO0309/onnx2tf/issues/133)

What's Changed
* `ConvTranspose`. Shape Unmatch Error Mitigation Measures by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/170


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.5.42...1.5.43

- Fixed dtype conversion error when outputting `Gather` for GPU Delegate

What's Changed
* Fixed dtype conversion error when outputting `Gather` for GPU Delegate by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/169


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.5.41...1.5.42