Metadata-Version: 2.4
Name: ml-pipes-torch
Version: 0.1.0rc2
Summary: Torch tensor payloads and torch-domain operators for ml-pipes
Project-URL: Homepage, https://github.com/trained-by-humans/ml-pipes
Project-URL: Documentation, https://github.com/trained-by-humans/ml-pipes/blob/main/packages/torch/docs/INDEX.md
Project-URL: Repository, https://github.com/trained-by-humans/ml-pipes
Project-URL: Issues, https://github.com/trained-by-humans/ml-pipes/issues
Project-URL: Changelog, https://github.com/trained-by-humans/ml-pipes/releases
License-Expression: Apache-2.0
License-File: LICENSE
Keywords: inference,machine-learning,pipelines,pytorch,torch
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: Apache Software License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Programming Language :: Python :: 3.13
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Classifier: Topic :: Software Development :: Libraries :: Application Frameworks
Requires-Python: >=3.10
Requires-Dist: ml-pipes-core==0.1.0rc2
Requires-Dist: ml-pipes-tensor==0.1.0rc2
Requires-Dist: numpy>=1.26
Requires-Dist: torch
Requires-Dist: torchvision
Description-Content-Type: text/markdown

# ml-pipes-torch

`ml-pipes-torch` brings Torch-backed execution into `ml-pipes`.

Use this package when a pipeline needs to cross into Torch for model
inference, device-aware execution, or postprocess that is worth keeping on the
Torch side. In practice, it often sits between NumPy-oriented preparation or
postprocess stages so mixed pipelines can move in and out of the Torch domain
explicitly. It also mirrors the generic tensor registry helpers that are often
worth keeping on-device during Torch-side postprocess, together with
Torch-native vision-adjacent registry helpers such as box-format conversion,
mask reconstruction, filtering, resizing, and NMS.

Operators can be used as plain callables, but this package is best used inside
an `ml-pipes` pipeline.

## Package Reference

| Field          | Value                                                                                                     |
|----------------|-----------------------------------------------------------------------------------------------------------|
| Package        | `ml-pipes-torch`                                                                                          |
| Guide          | [`docs/README.md`](https://github.com/trained-by-humans/ml-pipes/blob/main/packages/torch/docs/README.md) |
| Reference      | [`docs/INDEX.md`](https://github.com/trained-by-humans/ml-pipes/blob/main/packages/torch/docs/INDEX.md)   |
| Depends on     | `ml-pipes-core`, `ml-pipes-tensor`                                                                        |
| Public modules | `ml_pipes.torch`                                                                                          |
| Content        | Torch execution stages; NumPy/Torch boundary crossing; device placement; mirrored Torch-native tensor postprocess |

See [`docs/PACKAGES.md`](https://github.com/trained-by-humans/ml-pipes/blob/main/docs/PACKAGES.md)
for direct package installs, umbrella profiles, and the full package matrix.
