Tensors
Overview
The tensor-flow library is an external library that is not pre-loaded on the davinci platform.
Tensors
Tensors are a generalization of the matrix/vector concept. If you think of a vector as being a one-dimensioanl array of numbers, and matrices as
being a two-dimensional array of numbers, then you can think of a tensor as being a n-dimensional array of numbers. That is, its a collection of numbers
that are indexed by n indices.
tf.tensor( value, shape, dataType)
- value - the value of the tensor stated as an array
- shape - (optional) if not present, the shape will be inferred from the value
- dataType - (optional) specifies the type of data contained in the tensor. Will be inferred if not present
Simple Examples
Create a simple scalar
let input = tf.tensor(2);
Try it!
Create a simple 1-dimensional tensor
let input = tf.tensor([1, 2, 1, 1]);
Try it!
Create a simple 2-dimensional tensor
let input = tf.tensor([[1, 2, 1, 1]]);
Try it!
Reshaping a Tensor
Once a tensor has been created, you can create other tensors with the same values, but different shape by
using the reshape method.
let x = tf.tensor([1, 2, 1, 1]);
let y = tf.reshape(x, [2,2]);
Try it!
Sample Code
