Quick Start

Create a Tree-based Tensor

You can create a tree-based tensor or a native tensor like the following example code.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
import builtins
import os
from functools import partial

import treetensor.torch as torch

print = partial(builtins.print, sep=os.linesep)

if __name__ == '__main__':
    t1 = torch.tensor([[1, 2, 3],
                       [4, 5, 6]])
    print('new native tensor:', t1)

    t2 = torch.tensor({
        'a': [1, 2, 3],
        'b': {'x': [[4, 5], [6, 7]]},
    })
    print('new tree tensor:', t2)

    t3 = torch.randn(2, 3)
    print('new random native tensor:', t3)

    t4 = torch.randn({
        'a': (2, 3),
        'b': {'x': (3, 4)},
    })
    print('new random tree tensor:', t4)

The output should be like below.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
new native tensor:
tensor([[1, 2, 3],
        [4, 5, 6]])
new tree tensor:
<Tensor 0x7fe7ecae71d0>
├── a --> tensor([1, 2, 3])
└── b --> <Tensor 0x7fe7ecae7410>
    └── x --> tensor([[4, 5],
                      [6, 7]])

new random native tensor:
tensor([[ 0.6374,  0.5619,  1.2875],
        [-0.1674,  1.4108, -0.1149]])
new random tree tensor:
<Tensor 0x7fe7e2ec7410>
├── a --> tensor([[-1.0148,  0.7347, -1.6171],
│                 [-1.1030, -2.0437,  0.1575]])
└── b --> <Tensor 0x7fe7e2ec74d0>
    └── x --> tensor([[ 1.0864,  0.0259, -0.2414,  0.0771],
                      [ 0.4995,  0.4199,  0.1481,  0.0871],
                      [-0.5012,  0.1868,  1.5312, -0.5145]])