arange

nvtripy.arange(*args, **kwargs) → Tensor

This function has multiple overloads:

---

nvtripy.arange(start: numbers.Number | nvtripy.DimensionSize, stop: numbers.Number | nvtripy.DimensionSize, step: numbers.Number | nvtripy.DimensionSize = 1, dtype: nvtripy.dtype = float32) -> nvtripy.Tensor

Returns a 1D tensor containing a sequence of numbers in the half-open interval \([\text{start}, \text{stop})\) incrementing by \(\text{step}\).

Parameters:

• start – The inclusive lower bound of the values to generate. If a tensor is provided, it must be a scalar tensor.

• stop – The exclusive upper bound of the values to generate. If a tensor is provided, it must be a scalar tensor.

• step – The spacing between values. If a tensor is provided, it must be a scalar tensor.

• dtype – [dtype=T1] The desired data type of the tensor.

Returns:

[dtype=T1] A tensor of shape \([\frac{\text{stop}-\text{start}}{\text{step}}]\).

Return type:

Tensor

TYPE CONSTRAINTS:

• T1: bool, bfloat16, float16, float32, int8, int32, int64

Example

output = tp.arange(0.5, 2.5)

Local Variables

>>> output
tensor([0.5000, 1.5000], dtype=float32, loc=gpu:0, shape=(2,))

Example: Custom ``step`` Value

output = tp.arange(2.3, 0.8, -0.2)

Local Variables

>>> output
tensor([2.3000, 2.1000, 1.9000, 1.7000, 1.5000, 1.3000, 1.1000, 0.9000], dtype=float32, loc=gpu:0, shape=(8,))

---

nvtripy.arange(stop: numbers.Number | nvtripy.DimensionSize, dtype: nvtripy.dtype = float32) -> nvtripy.Tensor

Returns a 1D tensor containing a sequence of numbers in the half-open interval \([0, \text{stop})\) incrementing by 1.

Parameters:

• stop – The exclusive upper bound of the values to generate.

• dtype – The desired datatype of the tensor.

Returns:

A tensor of shape \([\text{stop}]\).

Return type:

Tensor

Example

output = tp.arange(5)

Local Variables

>>> output
tensor([0.0000, 1.0000, 2.0000, 3.0000, 4.0000], dtype=float32, loc=gpu:0, shape=(5,))