Open jamjambles opened 10 months ago
Hi all,
We are using the commonly adopted conversion steps in order to deploy models trained in PyTorch to TFLite for inference:
torch -> onnx -> tf -> tflite
Currently we rely on onnx-tensorflow to achieve the onnx to tf conversion step. In fact many dev blogs also point to this as the solution.
As mentioned in https://github.com/onnx/onnx-tensorflow/pull/1049, onnx-tensorflow is going to be deprecated and no longer maintained.
What is the recommended approach from the ONNX team/community for converting ONNX models to TF/TFLite moving forward?
In the meantime I will start looking into this actively maintained solution (onnx2tf): https://github.com/PINTO0309/onnx2tf
Thanks! James
Hi, have you found any solution? I am struggling with the same. Thank you!
Hi all,
We are using the commonly adopted conversion steps in order to deploy models trained in PyTorch to TFLite for inference:
torch -> onnx -> tf -> tflite
Currently we rely on onnx-tensorflow to achieve the onnx to tf conversion step. In fact many dev blogs also point to this as the solution.
As mentioned in https://github.com/onnx/onnx-tensorflow/pull/1049, onnx-tensorflow is going to be deprecated and no longer maintained.
What is the recommended approach from the ONNX team/community for converting ONNX models to TF/TFLite moving forward?
In the meantime I will start looking into this actively maintained solution (onnx2tf): https://github.com/PINTO0309/onnx2tf
Thanks! James