- `Add`, `Sub`, `Mul`, `Div`, `Mod`
- Added automatic correction function for conversion accuracy errors
- Conversion speed is reduced instead of compensating for accuracy errors
- `Add`, `Sub`, `Mul`, `Div`, `Mod`, `MatMul`, `Softmax`
- Strict conversion mode implemented
- However, if the input tensor has undefined dimensions, the automatic correction will not work perfectly.
- As of 2023.05.07, this is a work in progress and is an experimental feature.
- Added option to disable strict conversion mode
- `-dsm`, `--disable_strict_mode`
-dsm, --disable_strict_mode
If specified, the conversion speed is greatly accelerated because the strict accuracy
correction process is skipped, but the frequency of transposition errors increases
and accuracy errors are more likely to occur. Strict mode is enabled by default.
As of 2023.05.07, this is a work in progress and is an experimental feature.
Therefore, only some OPs are converted in exact mode for accuracy correction.
- Enable `-dsm` to greatly reduce the time to complete testing, as CI testing only checks the conversion structure.
- [[InSPyReNet] Swin Transformer Support question 312](https://github.com/PINTO0309/onnx2tf/issues/312)
What's Changed
* Added automatic correction function for conversion accuracy errors by PINTO0309 in https://github.com/PINTO0309/onnx2tf/pull/343
**Full Changelog**: https://github.com/PINTO0309/onnx2tf/compare/1.10.3...1.11.0