Create and modify tensors

Tensor methods in Sentis are similar to methods in frameworks like numpy, TensorFlow, and PyTorch.

Create a basic tensor

Create a basic tensor using the methods in the TensorFloat or TensorInt APIs. For example use TensorFloat.Zeros to create a tensor filled with 0.

Refer to Create input for a model for more information.

Reshape a tensor

You can't change the shape of a tensor, but you can use ShallowReshape to create a reshaped copy of a tensor.

For example:

// Create a 2 × 2 × 2 tensor
TensorShape shape = new TensorShape(2, 2, 2);
int[] data = new int[] { 1, 2, 3, 4, 10, 20, 30, 40 };
TensorInt tensor = new TensorInt(shape, data);

// Create a copy of the tensor with the new shape 2 × 1 × 4
TensorShape newShape = new TensorShape(2, 1, 4);
Tensor reshapedTensor = tensor.ShallowReshape(newShape);

The shape of the tensor and the new shape must have the same number of elements. You can use the length property of a tensor to check the number of elements.

The result of ShallowReshape is a shallow copy that points to the same memory as the original output.

Note: Don't use ShallowReshapeif you use BackendType.GPUPixel, because ShallowReshape isn't compatible with the way Sentis stores tensor data in textures.

Additional resources