Chin Huang
Chin Huang
I suppose there is no errors converting from onnx to tf, and from tf to tflite. In fact, onnx-tf converter doesn't officially support tflite. So I am not sure how...
Have you tried the device = CUDA option, as described https://github.com/onnx/onnx-tensorflow/blob/master/doc/CLI.md? It will keep NCHW as much as possible, unless TF API doesn't support NCHW.
The current onnx-tf doesn't produce a frozen graph anymore. Instead, a SavedModel is created including a .pb file. However that pb file is to be used by Tensorflow SavedModel API;...
Agree. Hopefully someone can look into that option soon. @mdatres You could make a source build using the "tf-1.x" branch. I believe it exports into a frozen graph. I just...
The shape inference error is typically coming from onnx checker. That indicates the onnx model has some issues. Please try to run a simple test with your onnx file to...
Can you provide sample input data and code to reproduce the inference error?
Onnx model with external data is only supported using command line as described, https://github.com/onnx/onnx-tensorflow/blob/master/doc/CLI.md. Please use -e or --extdatadir in onnx-tf command and see if it helps.
The onnx-tf converter takes an onnx model graph and converts node by node to a Tensorflow graph. The converted graph/model can be saved as a Tensorflow SavedModel and used for...
The converted and saved model is for inference only by design. So unable to re-train is expected. Any contribution is welcome to enable training.
This issue is not trivial. It will take some time to investigate a fix. So I think the 1.8 release will not include it unless someone from community can jump...