Initialize a 4-dim Tensor for inference about keras2cpp HOT 4 OPEN

Same issue, struggling to create multi-dimensional tensors

from keras2cpp.

Try Tensor in({1, 28, 28, 1})
That seemed to work for me but I am lost on how to input data into that tensor once you have formed it. Hope that helps a little.

from keras2cpp.

Hi again, after some more experimentation, I found a tedious but seemingly working solution for my problem. I'm using a 3d tensor of shape (1, 1, 16) for an LSTM, and formatted like this:
// Create a 3D Tensor on length 16 for input data.
Tensor in{1, 1, 16};
in.data_[0] = reflectivity;
in.data_[1] = echoDist;
in.data_[2] = redx;
in.data_[3] = redy;
in.data_[4] = redSize;
in.data_[5] = bluex;
in.data_[6] = bluey;
in.data_[7] = blueSize;
in.data_[8] = yellowx;
in.data_[9] = yellowy;
in.data_[10] = yellowSize;
in.data_[11] = imu_theta;
in.data_[12] = imu_accelx;
in.data_[13] = imu_accely;
in.data_[14] = liftpos;
in.data_[15] = hook_state;
// Run prediction.
Tensor out = model(in);
result = out.data_[0]
Note I get my result this way because my output is a single value. Your "output" code may be different. I know this is late but hope this helps you and any future users.

from keras2cpp.

And last thing, forgot to mention I removed the "kassert" and all related lines on around line 66 of the tensor.cc file.

from keras2cpp.