torchwrench

Name	torchwrench JSON
Version	0.7.3 JSON
	download
home_page	None
Summary	Collection of functions and modules to help development in PyTorch.
upload_time	2025-07-18 22:01:46
maintainer	None
docs_url	None
author	None
requires_python	>=3.8
license	MIT License Copyright (c) 2025 Labbeti Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
keywords	pytorch deep-learning
VCS
bugtrack_url
requirements	No requirements were recorded.
Travis-CI	No Travis.
coveralls test coverage	No coveralls.

            # torchwrench

<center>

<a href="https://www.python.org/">
    <img alt="Python" src="https://img.shields.io/badge/-Python 3.8+-blue?style=for-the-badge&logo=python&logoColor=white">
</a>
<a href="https://github.com/Labbeti/torchwrench/actions">
    <img alt="Build" src="https://img.shields.io/github/actions/workflow/status/Labbeti/torchwrench/test.yaml?branch=main&style=for-the-badge&logo=github">
</a>
<a href='https://torchwrench.readthedocs.io/'>
    <img src='https://readthedocs.org/projects/torchwrench/badge/?version=stable&style=for-the-badge' alt='Documentation Status' />
</a>
<a href="https://pytorch.org/get-started/locally/">
    <img alt="PyTorch" src="https://img.shields.io/badge/-PyTorch 1.10+-ee4c2c?style=for-the-badge&logo=pytorch&logoColor=white">
</a>

Collection of functions and modules to help development in PyTorch.

</center>


## Installation

With pip:
```bash
pip install torchwrench
```

With uv:
```bash
uv add torchwrench
```

The main requirement is **[PyTorch](https://pytorch.org/)**.

To check if the package is installed and show the package version, you can use the following command in your terminal:
```bash
torchwrench-info
```

This library has been tested on all Python versions **3.8 - 3.13**, all PyTorch versions **1.10 - 2.6**, and on **Linux, Mac and Windows** systems.

## Examples

`torchwrench` functions and modules can be used like `torch` ones. The default acronym for `torchwrench` is `tw`.

### Label conversions
Supports **multiclass** labels conversions between probabilities, classes indices, classes names and onehot encoding.

```python
import torchwrench as tw

probs = tw.as_tensor([[0.9, 0.1], [0.4, 0.6]])
names = tw.probs_to_name(probs, idx_to_name={0: "Cat", 1: "Dog"})
# ["Cat", "Dog"]
```

This package also supports **multilabel** labels conversions between probabilities, classes multi-indices, classes multi-names and multihot encoding.

```python
import torchwrench as tw

multihot = tw.as_tensor([[1, 0, 0], [0, 1, 1], [0, 0, 0]])
indices = tw.multihot_to_indices(multihot)
# [[0], [1, 2], []]
```

Finally, this packages includes the **powerset multilabel** conversions :

```python
import torchwrench as tw

multihot = tw.as_tensor([[1, 0, 0], [0, 1, 1], [0, 0, 0]])
indices = tw.multilabel_to_powerset(multihot, num_classes=3, max_set_size=2)
# tensor([[0, 1, 0, 0, 0, 0, 0],
#         [0, 0, 0, 0, 0, 0, 1],
#         [1, 0, 0, 0, 0, 0, 0]])
```

### Typing

Typing with number of dimensions :

```python
import torchwrench as tw

x1 = tw.as_tensor([1, 2])
print(isinstance(x1, tw.Tensor2D))  # False
x2 = tw.as_tensor([[1, 2], [3, 4]])
print(isinstance(x2, tw.Tensor2D))  # True
```

Typing with tensor dtype :

```python
import torchwrench as tw

x1 = tw.as_tensor([1, 2], dtype=tw.int)
print(isinstance(x1, tw.SignedIntegerTensor))  # True

x2 = tw.as_tensor([1, 2], dtype=tw.long)
print(isinstance(x2, tw.SignedIntegerTensor1D))  # True

x3 = tw.as_tensor([1, 2], dtype=tw.float)
print(isinstance(x3, tw.SignedIntegerTensor))  # False
```

### Padding & cropping

Pad a specific dimension :

```python
import torchwrench as tw

x = tw.rand(10, 3, 1)
padded = tw.pad_dim(x, target_length=5, dim=1, pad_value=-1)
# x2 has shape (10, 5, 1), padded with -1
```

Pad nested list of tensors to a single one :

```python
import torchwrench as tw

tensors = [tw.rand(10, 2), [tw.rand(3)] * 5, tw.rand(0, 5)]
padded = tw.pad_and_stack_rec(tensors, pad_value=0)
# padded has shape (3, 10, 5), padded with 0
```

Remove values at a specific dimension :

```python
import torchwrench as tw

x = tw.rand(10, 5, 3)
cropped = tw.crop_dim(x, dim=1, target_length=2)
# cropped has shape (10, 2, 3)
```

### Masking

```python
import torchwrench as tw

x = tw.as_tensor([3, 1, 2])
mask = tw.lengths_to_non_pad_mask(x, max_len=4)
# Each row i contains x[i] True values for non-padding mask
# tensor([[True, True, True, False],
#         [True, False, False, False],
#         [True, True, False, False]])
```

```python
import torchwrench as tw

x = tw.as_tensor([1, 2, 3, 4])
mask = tw.as_tensor([True, True, False, False])
result = tw.masked_mean(x, mask)
# result contains the mean of the values marked as True: 1.5
```

### Others tensors manipulations!

```python
import torchwrench as tw

x = tw.as_tensor([1, 2, 3, 4])
result = tw.insert_at_indices(x, indices=[0, 2], values=5)
# result contains tensor with inserted values: tensor([5, 1, 2, 5, 3, 4])
```

```python
import torchwrench as tw

perm = tw.randperm(10)
inv_perm = tw.get_inverse_perm(perm)

x1 = tw.rand(10)
x2 = x1[perm]
x3 = x2[inv_perm]
# inv_perm are indices that allow us to get x3 from x2, i.e. x1 == x3 here
```

### Extra: pre-compute datasets to HDF files

Here is an example of pre-computing spectrograms of torchaudio `SPEECHCOMMANDS` dataset, using `pack_dataset` function:

```python
from torchaudio.datasets import SPEECHCOMMANDS
from torchaudio.transforms import Spectrogram
from torchwrench import nn
from torchwrench.extras.hdf import pack_to_hdf

speech_commands_root = "path/to/speech_commands"
packed_root = "path/to/packed_dataset.hdf"

dataset = SPEECHCOMMANDS(speech_commands_root, download=True, subset="validation")
# dataset[0] is a tuple, contains waveform and other metadata

class MyTransform(nn.Module):
    def __init__(self) -> None:
        super().__init__()
        self.spectrogram_extractor = Spectrogram()

    def forward(self, item):
        waveform = item[0]
        spectrogram = self.spectrogram_extractor(waveform)
        return (spectrogram,) + item[1:]

pack_to_hdf(dataset, packed_root, MyTransform())
```

Then you can load the pre-computed dataset using `HDFDataset`:
```python
from torchwrench.extras.hdf import HDFDataset

packed_root = "path/to/packed_dataset.hdf"
packed_dataset = HDFDataset(packed_root)
packed_dataset[0]  # == first transformed item, i.e. transform(dataset[0])
```

## Contact
Maintainer:
- [Étienne Labbé](https://labbeti.github.io/) "Labbeti": labbeti.pub@gmail.com

            

Raw data

            {
    "_id": null,
    "home_page": null,
    "name": "torchwrench",
    "maintainer": null,
    "docs_url": null,
    "requires_python": ">=3.8",
    "maintainer_email": "\"\u00c9tienne Labb\u00e9 (Labbeti)\" <labbeti.pub@gmail.com>",
    "keywords": "pytorch, deep-learning",
    "author": null,
    "author_email": "\"\u00c9tienne Labb\u00e9 (Labbeti)\" <labbeti.pub@gmail.com>",
    "download_url": "https://files.pythonhosted.org/packages/58/79/243992598ee2e04d5a25c77bd6ba7f3390b8bb4b55591eda0ba3bee5e22f/torchwrench-0.7.3.tar.gz",
    "platform": null,
    "description": "# torchwrench\n\n<center>\n\n<a href=\"https://www.python.org/\">\n    <img alt=\"Python\" src=\"https://img.shields.io/badge/-Python 3.8+-blue?style=for-the-badge&logo=python&logoColor=white\">\n</a>\n<a href=\"https://github.com/Labbeti/torchwrench/actions\">\n    <img alt=\"Build\" src=\"https://img.shields.io/github/actions/workflow/status/Labbeti/torchwrench/test.yaml?branch=main&style=for-the-badge&logo=github\">\n</a>\n<a href='https://torchwrench.readthedocs.io/'>\n    <img src='https://readthedocs.org/projects/torchwrench/badge/?version=stable&style=for-the-badge' alt='Documentation Status' />\n</a>\n<a href=\"https://pytorch.org/get-started/locally/\">\n    <img alt=\"PyTorch\" src=\"https://img.shields.io/badge/-PyTorch 1.10+-ee4c2c?style=for-the-badge&logo=pytorch&logoColor=white\">\n</a>\n\nCollection of functions and modules to help development in PyTorch.\n\n</center>\n\n\n## Installation\n\nWith pip:\n```bash\npip install torchwrench\n```\n\nWith uv:\n```bash\nuv add torchwrench\n```\n\nThe main requirement is **[PyTorch](https://pytorch.org/)**.\n\nTo check if the package is installed and show the package version, you can use the following command in your terminal:\n```bash\ntorchwrench-info\n```\n\nThis library has been tested on all Python versions **3.8 - 3.13**, all PyTorch versions **1.10 - 2.6**, and on **Linux, Mac and Windows** systems.\n\n## Examples\n\n`torchwrench` functions and modules can be used like `torch` ones. The default acronym for `torchwrench` is `tw`.\n\n### Label conversions\nSupports **multiclass** labels conversions between probabilities, classes indices, classes names and onehot encoding.\n\n```python\nimport torchwrench as tw\n\nprobs = tw.as_tensor([[0.9, 0.1], [0.4, 0.6]])\nnames = tw.probs_to_name(probs, idx_to_name={0: \"Cat\", 1: \"Dog\"})\n# [\"Cat\", \"Dog\"]\n```\n\nThis package also supports **multilabel** labels conversions between probabilities, classes multi-indices, classes multi-names and multihot encoding.\n\n```python\nimport torchwrench as tw\n\nmultihot = tw.as_tensor([[1, 0, 0], [0, 1, 1], [0, 0, 0]])\nindices = tw.multihot_to_indices(multihot)\n# [[0], [1, 2], []]\n```\n\nFinally, this packages includes the **powerset multilabel** conversions :\n\n```python\nimport torchwrench as tw\n\nmultihot = tw.as_tensor([[1, 0, 0], [0, 1, 1], [0, 0, 0]])\nindices = tw.multilabel_to_powerset(multihot, num_classes=3, max_set_size=2)\n# tensor([[0, 1, 0, 0, 0, 0, 0],\n#         [0, 0, 0, 0, 0, 0, 1],\n#         [1, 0, 0, 0, 0, 0, 0]])\n```\n\n### Typing\n\nTyping with number of dimensions :\n\n```python\nimport torchwrench as tw\n\nx1 = tw.as_tensor([1, 2])\nprint(isinstance(x1, tw.Tensor2D))  # False\nx2 = tw.as_tensor([[1, 2], [3, 4]])\nprint(isinstance(x2, tw.Tensor2D))  # True\n```\n\nTyping with tensor dtype :\n\n```python\nimport torchwrench as tw\n\nx1 = tw.as_tensor([1, 2], dtype=tw.int)\nprint(isinstance(x1, tw.SignedIntegerTensor))  # True\n\nx2 = tw.as_tensor([1, 2], dtype=tw.long)\nprint(isinstance(x2, tw.SignedIntegerTensor1D))  # True\n\nx3 = tw.as_tensor([1, 2], dtype=tw.float)\nprint(isinstance(x3, tw.SignedIntegerTensor))  # False\n```\n\n### Padding & cropping\n\nPad a specific dimension :\n\n```python\nimport torchwrench as tw\n\nx = tw.rand(10, 3, 1)\npadded = tw.pad_dim(x, target_length=5, dim=1, pad_value=-1)\n# x2 has shape (10, 5, 1), padded with -1\n```\n\nPad nested list of tensors to a single one :\n\n```python\nimport torchwrench as tw\n\ntensors = [tw.rand(10, 2), [tw.rand(3)] * 5, tw.rand(0, 5)]\npadded = tw.pad_and_stack_rec(tensors, pad_value=0)\n# padded has shape (3, 10, 5), padded with 0\n```\n\nRemove values at a specific dimension :\n\n```python\nimport torchwrench as tw\n\nx = tw.rand(10, 5, 3)\ncropped = tw.crop_dim(x, dim=1, target_length=2)\n# cropped has shape (10, 2, 3)\n```\n\n### Masking\n\n```python\nimport torchwrench as tw\n\nx = tw.as_tensor([3, 1, 2])\nmask = tw.lengths_to_non_pad_mask(x, max_len=4)\n# Each row i contains x[i] True values for non-padding mask\n# tensor([[True, True, True, False],\n#         [True, False, False, False],\n#         [True, True, False, False]])\n```\n\n```python\nimport torchwrench as tw\n\nx = tw.as_tensor([1, 2, 3, 4])\nmask = tw.as_tensor([True, True, False, False])\nresult = tw.masked_mean(x, mask)\n# result contains the mean of the values marked as True: 1.5\n```\n\n### Others tensors manipulations!\n\n```python\nimport torchwrench as tw\n\nx = tw.as_tensor([1, 2, 3, 4])\nresult = tw.insert_at_indices(x, indices=[0, 2], values=5)\n# result contains tensor with inserted values: tensor([5, 1, 2, 5, 3, 4])\n```\n\n```python\nimport torchwrench as tw\n\nperm = tw.randperm(10)\ninv_perm = tw.get_inverse_perm(perm)\n\nx1 = tw.rand(10)\nx2 = x1[perm]\nx3 = x2[inv_perm]\n# inv_perm are indices that allow us to get x3 from x2, i.e. x1 == x3 here\n```\n\n### Extra: pre-compute datasets to HDF files\n\nHere is an example of pre-computing spectrograms of torchaudio `SPEECHCOMMANDS` dataset, using `pack_dataset` function:\n\n```python\nfrom torchaudio.datasets import SPEECHCOMMANDS\nfrom torchaudio.transforms import Spectrogram\nfrom torchwrench import nn\nfrom torchwrench.extras.hdf import pack_to_hdf\n\nspeech_commands_root = \"path/to/speech_commands\"\npacked_root = \"path/to/packed_dataset.hdf\"\n\ndataset = SPEECHCOMMANDS(speech_commands_root, download=True, subset=\"validation\")\n# dataset[0] is a tuple, contains waveform and other metadata\n\nclass MyTransform(nn.Module):\n    def __init__(self) -> None:\n        super().__init__()\n        self.spectrogram_extractor = Spectrogram()\n\n    def forward(self, item):\n        waveform = item[0]\n        spectrogram = self.spectrogram_extractor(waveform)\n        return (spectrogram,) + item[1:]\n\npack_to_hdf(dataset, packed_root, MyTransform())\n```\n\nThen you can load the pre-computed dataset using `HDFDataset`:\n```python\nfrom torchwrench.extras.hdf import HDFDataset\n\npacked_root = \"path/to/packed_dataset.hdf\"\npacked_dataset = HDFDataset(packed_root)\npacked_dataset[0]  # == first transformed item, i.e. transform(dataset[0])\n```\n\n## Contact\nMaintainer:\n- [\u00c9tienne Labb\u00e9](https://labbeti.github.io/) \"Labbeti\": labbeti.pub@gmail.com\n",
    "bugtrack_url": null,
    "license": "MIT License\n        \n        Copyright (c) 2025 Labbeti\n        \n        Permission is hereby granted, free of charge, to any person obtaining a copy\n        of this software and associated documentation files (the \"Software\"), to deal\n        in the Software without restriction, including without limitation the rights\n        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell\n        copies of the Software, and to permit persons to whom the Software is\n        furnished to do so, subject to the following conditions:\n        \n        The above copyright notice and this permission notice shall be included in all\n        copies or substantial portions of the Software.\n        \n        THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\n        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\n        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\n        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\n        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\n        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\n        SOFTWARE.\n        ",
    "summary": "Collection of functions and modules to help development in PyTorch.",
    "version": "0.7.3",
    "project_urls": {
        "Changelog": "https://github.com/Labbeti/torchwrench/blob/main/CHANGELOG.md",
        "Documentation": "https://torchwrench.readthedocs.io/",
        "Homepage": "https://pypi.org/project/torchwrench/",
        "Repository": "https://github.com/Labbeti/torchwrench.git",
        "Tracker": "https://github.com/Labbeti/torchwrench/issues"
    },
    "split_keywords": [
        "pytorch",
        " deep-learning"
    ],
    "urls": [
        {
            "comment_text": null,
            "digests": {
                "blake2b_256": "5879243992598ee2e04d5a25c77bd6ba7f3390b8bb4b55591eda0ba3bee5e22f",
                "md5": "c71e97fa2bf3ca1957eb006d6bf6e88f",
                "sha256": "0e87a4da240d54de27c5447d5cac41011590bdee1bf5d075335e5e12cc0b21fc"
            },
            "downloads": -1,
            "filename": "torchwrench-0.7.3.tar.gz",
            "has_sig": false,
            "md5_digest": "c71e97fa2bf3ca1957eb006d6bf6e88f",
            "packagetype": "sdist",
            "python_version": "source",
            "requires_python": ">=3.8",
            "size": 116079,
            "upload_time": "2025-07-18T22:01:46",
            "upload_time_iso_8601": "2025-07-18T22:01:46.307735Z",
            "url": "https://files.pythonhosted.org/packages/58/79/243992598ee2e04d5a25c77bd6ba7f3390b8bb4b55591eda0ba3bee5e22f/torchwrench-0.7.3.tar.gz",
            "yanked": false,
            "yanked_reason": null
        }
    ],
    "upload_time": "2025-07-18 22:01:46",
    "github": true,
    "gitlab": false,
    "bitbucket": false,
    "codeberg": false,
    "github_user": "Labbeti",
    "github_project": "torchwrench",
    "travis_ci": false,
    "coveralls": false,
    "github_actions": true,
    "lcname": "torchwrench"
}