Onnx2tf

- Fix onnxsim calls using `--no_large_tensor`

onnxsim xxx.onnx yyy.onnx --no_large_tensor 10MB


What's Changed
* Fix onnxsim calls using `--no_large_tensor` by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/256


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.31...1.7.32

- `RandomNormalLike`, `RandomUniformLike`
- Improved stability of `RandomNormalLike` and `RandomUniformLike`
- `Split`
- Suppression of `FlexSplitV` generation
- https://github.com/isletennos/MMVC_Trainer
- `D_180000.onnx`
- `G_180000.onnx`


onnx2tf -i D_180000.onnx -ois specs:1,257,100 -cotof -cotoa 1e-1


- [Incorrect tensor shape when converting a model with "torch.randn_like" 253](https://github.com/PINTO0309/onnx2tf/issues/253)

What's Changed
* 1.Improved stability of `RandomNormalLike` and `RandomUniformLike`. 2.Suppression of `FlexSplitV` generation by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/255


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.30...1.7.31

- Significantly reduced RAM consumption during model conversion when the `-cotof` option is not specifie
- https://github.com/PINTO0309/PINTO_model_zoo/tree/main/352_MAXIM

What's Changed
* Significantly reduced RAM consumption during model conversion when the `-cotof` option is not specifie by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/254


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.29...1.7.30

- Fixed lack of sanitizing of OP names when outputting INT8 models

What's Changed
* Fixed lack of sanitizing of OP names when outputting INT8 models by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/252


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.28...1.7.29

- `Add`, `Sub`, `Div`, `Mul`, `Mod`
- Fixed a problem with the `shape_unmatched_special_avoidance_workaround` judgment condition to improve the Transformer conversion success rate.
- Ref: https://github.com/zhangyi-3/KBNet

What's Changed
* Fixed a problem with the `shape_unmatched_special_avoidance_workaround` judgment condition to improve the Transformer conversion success rate. by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/250


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.27...1.7.28

- Improved INT8 quantization to be performed based on `saved_model` signature information
- Ref: https://www.tensorflow.org/lite/performance/post_training_quantization#full_integer_quantization
- Improved matching of the input OP name specified in the dataset for calibration with the input OP name in the signature to prevent input order discrepancies.
- When quantizing INT8 for models with multiple inputs, you no longer need to be aware of the order in which the calibration data sets are specified.
- Only when performing INT8 quantization, the `saved_model` signature information of the converted model is displayed in the log as reference information, as shown in the figure below.
- The input OP name in ONNX and the input OP name after conversion to `saved_model` may mismatch. This is due to automatic sanitization of strings that cannot be used in the input OP name of `saved_model`. e.g. `:`, `/`
- [[BERT-Squad] INT8 quantization: The input data type must be Float32. 248](https://github.com/PINTO0309/onnx2tf/issues/248)

What's Changed
* Improved INT8 quantization to be performed based on `saved_model` signature information by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/249


**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.7.26...1.7.27