I’m trying to convert a CRNN model which consists of both Conv and LSTM layers. Everything works fine. However, when I’m using onnx.checker.check_model(model) I’m getting a warning:
Warning: ConstantFill was a removed experimental ops. In the future, we may directly reject this operator. Please update your model as soon as possible.
PyTorch saves the model in ai.onnx v9 version. When I try to convert it to version of 10, I’m getting:
Assertion false failed: No Adapter For Current Version $9 for MaxPool
Everything would be okay, but when I try to run the onnx model on ML dotnet, I’m getting RunTimeExeption due to this ConstantFill.
What is ConstantFill?
Is there any solution?
PS: I know that it is possible to use LSTM layers in Onnx.