You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is there a way to get the non-tensor types from the onnx model?
For example, if the model input is of type: std::vector<std::map<int64_t, float>>, I could get the input type of ONNX_TYPE_SEQUENCE by running session->GetInputTypeInfo(i).GetONNXType();
However, I'm confused how I could continue to get the type of the map while creating Ort::Value using the CXX api?
Another question is when loading the Bert model from Bert_model, one of the input shapes is [-1, 256] (Actually this model converted from Tensorflow and the inputs shape all starts with -1). Is the negative index supported when calling session run function? I'm getting the error of out of data bound.
For example, if the model input is of type: std::vector<std::map<int64_t, float>>, I could get the input type of
ONNX_TYPE_SEQUENCEby runningsession->GetInputTypeInfo(i).GetONNXType();However, I'm confused how I could continue to get the type of the map while creating Ort::Value using the CXX api?