torch-nested
Easily manipulate torch.Tensors inside highly nested data-structures.
You may want to consider using torch.nested,
but if you are working with nested dicts, lists, tuples, etc. of torch.Tensors,
here is the package for you.
A proper documentation is coming. Until then, a basic example is shown below, and you can look at the docstrings or tests of this package for more information.
Basic usage
Given a nested structure that contains torch.Tensor, this package makes it easy to access these Tensors and
work with them:
import torch
from torch_nested import NestedTensors
INPUT_DATA = [
(
torch.ones(3),
torch.zeros(2)
),
torch.ones((2, 2, 2)),
{
"foo": torch.ones(2),
"bar": [],
"har": "rar"
},
1
]
tensors = NestedTensors(INPUT_DATA)
# Original data preserved in .data-member
assert tensors.data == INPUT_DATA
# Simple accessing and setting
for i, tensor in enumerate(tensors):
tensors[i] = tensor + i
# Has basic dunders
assert len(tensors) == 4
assert torch.all(next(tensors) == torch.ones(3))
Calling print(tensors.shape()) would yield:
torch_nested.Size(
[
(
torch.Size([3]),
torch.Size([2])
),
torch.Size([2, 2, 2]),
{
foo: torch.Size([2]),
bar: None,
har: None
},
None
]
)
Supported data-structures
The following data-structures are supported so far:
torch.TensordictlisttupleNone- Any class with a
.tensors-attribute - Any class with a
.data-attribute, even if it isn't atorch.Tensor
For example
class ObjWithTensors:
tensors = [torch.ones(2), torch.zeros(2)]
class ObjWithData:
data = [torch.ones(2), torch.zeros(2)]
tensors = NestedTensors([ObjWithTensors(), ObjWithData()])
Running print(tensors.size()) would result in the following output:
NestedSize(
[
ObjWithTensors(
tensors: [
torch.Size([2]),
torch.Size([2])
]
),
ObjWithData(
data: [
torch.Size([2]),
torch.Size([2])
]
)
]
)
More data-structures will be supported in the future. Any data that is of an unsupported type
will not have its Tensors readable or writable, and NestedShape will show None there.
Metadata
Release files for torch-nested 0.0.5
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| torch_nested-0.0.5.tar.gz | 8.8 kB | Without provenance | Details for torch_nested-0.0.5.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| torch_nested-0.0.5-py3-none-any.whl | Python 3 | none | any | Details for torch_nested-0.0.5-py3-none-any.whl |
Total release size: 17.3 kB
Release files / torch_nested-0.0.5.tar.gz
| Download URL | torch_nested-0.0.5.tar.gz |
|---|---|
| Size | 8.8 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
4d2c57a03029ed220401258f0c1dbce2a726fb45f6fc8630c7a45645f9bab692
|
|
BLAKE2b-256 checksum How to use checksums |
c3f6c13b8ff10003ad1d31be8df1ebb34379439c451b55e60be401d47177045c
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/4.0.2 CPython/3.9.16
|
Release files / torch_nested-0.0.5-py3-none-any.whl
| Download URL | torch_nested-0.0.5-py3-none-any.whl |
|---|---|
| Size | 8.5 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
f9bd50da3ad67d933388c33f0c6be16e6efc69458c453a5f0e48184c64f656db
|
|
BLAKE2b-256 checksum How to use checksums |
0df42271ed66da64eb15d98f58d77dd6bf92b3615b79beedff24c9329fb17a66
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/4.0.2 CPython/3.9.16
|
Security for torch-nested 0.0.5
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.