How to reshape a tensor in TensorFlowSharp

Question

TensorFlowSharp is the wrapper of TensorFlow in c# platform. click it to jump to TensorFlowSharp github

Now i need to reshape a tensor with shape [32,64,1] into a new tensor with shape [1, 2048].But as i refer to the official API document, the usage seems like that:

TFOutput Reshape (TensorFlow.TFOutput tensor, TensorFlow.TFOutput shape);

the problem is that i dont know how to express the shape i need in the way of TFOutput Any suggestions will be appreciated:)!

For the `shape` argument, can you construct a `Constant` of element type `int64` with the values `[1, 2048]` and pass that? — mrry, Aug 18 '17 at 17:29

score 2 · Accepted Answer · answered Aug 18 '17 at 22:27

In standard TensorFlowSharp, an example on how to do this operation can be given by:

tf.Reshape(x, tf.Const(shape));

where tf is the current default TFGraph in your TFSession.

Alternatively, if you use Keras Sharp, you might be able to do this operation using

using (var K = new TensorFlowBackend())
{
    double[,] input_array = new double[,] { { 1, 2 }, { 3, 4 } };
    Tensor variable = K.variable(array: input_array);
    Tensor variable_new_shape = K.reshape(variable, new int[] { 1, 4 });
    double[,] output = (double[,])variable_new_shape.eval();

    Assert.AreEqual(new double[,] { { 1, 2, 3, 4 } }, output);
}

as indicated in https://github.com/cesarsouza/keras-sharp/blob/efac7e34457ffb7cf6712793d5298b565549a1c2/Tests/TensorFlowBackendTest.cs#L45

How to reshape a tensor in TensorFlowSharp

1 Answers1