DimensionSize

class nvtripy.DimensionSize(data: int, name: str | None = None)

Bases: Tensor

A 0D, int32 tensor that represents a scalar value extracted from the shape of a tensor.

Parameters:

• data (int) – The value of the DimensionSize, which should be a scalar integer.

• name (str | None) – An optional name.

eval() → Tensor

Immediately evaluates this tensor. By default, tensors are evaluated lazily.

Note that an evaluated tensor will always reside in device memory.

Returns:

The evaluated tensor.

Return type:

Tensor

Example

import time

start = time.perf_counter()
tensor = tp.ones((3, 3))
init_time = time.perf_counter()
tensor.eval()
eval_time = time.perf_counter()

print(f"Tensor init_time took: {(init_time - start)  * 1000.0:.3f} ms")
print(f"Tensor evaluation took: {(eval_time - init_time)  * 1000.0:.3f} ms")

Local Variables

>>> tensor
tensor(
    [[1, 1, 1],
     [1, 1, 1],
     [1, 1, 1]], 
    dtype=float32, loc=gpu:0, shape=(3, 3))

Output

Tensor init_time took: 9.500 ms
Tensor evaluation took: 3957.263 ms

tolist() → List | Number

Returns the tensor as a nested list. If the tensor is a scalar, returns a python number.

Returns:

The tensor represented as a nested list or a python number.

Return type:

List | Number

Example: Ranked tensor

tensor = tp.ones((2, 2))
tensor_list = tensor.tolist()

Local Variables

>>> tensor_list
[[1.0, 1.0], [1.0, 1.0]]

Example: Scalar

tensor = tp.Tensor(2.0, dtype=tp.float32)
tensor_scalar = tensor.tolist()

Local Variables

>>> tensor_scalar
2.0