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 0x7fa8d853b210>
├── 'a' --> tensor([1, 2, 3])
└── 'b' --> <Tensor 0x7fa8d853b250>
    └── 'x' --> tensor([[4, 5],
                        [6, 7]])

new random native tensor:
tensor([[-0.2441,  0.4233,  1.6141],
        [ 0.6953, -1.9220,  0.0342]])
new random tree tensor:
<Tensor 0x7fa876df68d0>
├── 'a' --> tensor([[-0.1558, -0.6478, -0.2780],
│                   [-1.3952, -1.5896, -1.8346]])
└── 'b' --> <Tensor 0x7fa876d1efd0>
    └── 'x' --> tensor([[ 0.3287, -0.1604, -0.2906,  1.5804],
                        [-0.6113,  0.3921,  0.8725,  0.8264],
                        [-0.9002,  0.9896,  0.0084, -0.4874]])