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 0x7f9943169e50>
├── 'a' --> tensor([1, 2, 3])
└── 'b' --> <Tensor 0x7f9943169e80>
    └── 'x' --> tensor([[4, 5],
                        [6, 7]])

new random native tensor:
tensor([[ 0.9266, -1.1241, -0.2270],
        [-0.8881,  1.8911,  0.5266]])
new random tree tensor:
<Tensor 0x7f994262c070>
├── 'a' --> tensor([[ 0.1403, -0.4041,  0.6614],
│                   [ 1.0386,  1.4553,  1.0425]])
└── 'b' --> <Tensor 0x7f994357f190>
    └── 'x' --> tensor([[-0.3532,  0.3451,  0.3097,  0.2550],
                        [ 2.8158, -1.8183,  0.3966,  2.0692],
                        [ 1.8179,  2.5020, -0.8336,  0.4362]])