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 0x7f8c5f556310>
├── 'a' --> tensor([1, 2, 3])
└── 'b' --> <Tensor 0x7f8c5f556d50>
    └── 'x' --> tensor([[4, 5],
                        [6, 7]])

new random native tensor:
tensor([[-0.3538, -1.1970,  0.1106],
        [ 1.0215, -1.2022, -0.6034]])
new random tree tensor:
<Tensor 0x7f8c5f4ff650>
├── 'a' --> tensor([[-0.4780, -0.1017,  0.2546],
│                   [-1.6638, -1.7531, -0.8387]])
└── 'b' --> <Tensor 0x7f8c5f556250>
    └── 'x' --> tensor([[-1.9670, -0.8534,  0.3448,  1.8033],
                        [ 0.1344, -1.2076,  1.4108,  2.7070],
                        [-0.6261, -1.9774, -0.1637, -0.4172]])