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 0x7f8b0a9bd250>
├── 'a' --> tensor([1, 2, 3])
└── 'b' --> <Tensor 0x7f8b0a93ba30>
    └── 'x' --> tensor([[4, 5],
                        [6, 7]])

new random native tensor:
tensor([[-0.6972,  0.1486,  2.3812],
        [-0.5509,  0.3637,  1.7873]])
new random tree tensor:
<Tensor 0x7f8b0a5e9ee0>
├── 'a' --> tensor([[ 1.6505, -1.6343,  0.7266],
│                   [-1.1697,  0.9796,  0.3503]])
└── 'b' --> <Tensor 0x7f8b0a9ab460>
    └── 'x' --> tensor([[ 0.4693, -0.0212,  1.3313,  0.3956],
                        [-0.3996,  1.0222, -1.4236, -1.2373],
                        [ 0.1592, -1.7313, -1.9491,  2.0583]])