TorchSig is an open-source signal processing machine learning toolkit based on the PyTorch data handling pipeline. The user-friendly toolkit simplifies common digital signal processing operations, augmentations, and transformations when dealing with both real and complex-valued signals. TorchSig streamlines the integration process of these signals processing tools building on PyTorch, enabling faster and easier development and research for machine learning techniques applied to signals data, particularly within (but not limited to) the radio frequency domain.
- Ubuntu ≥ 22.04
- Hard drive storage with ≥ 1 TB
- CPU with ≥ 4 cores
- GPU with ≥ 16 GB storage (recommended)
- Python ≥ 3.10
We highly reccomend Ubuntu or using a Docker container.
Use PyPI to install:
pip install torchsig
Or clone the torchsig repository and install using the following commands:
git clone https://github.com/TorchDSP/torchsig.git
cd torchsig
# Runtime only (for normal usage)
pip install -e .
# Full development environment (for contributors)
pip install -e .[dev]
# Documentation build environment
pip install -e ".[docs]"
# To run the notebooks
pip install -e .[notebooks]
# To run the notebooks, including examples
pip install -e .[examples]
When to use which command?
-
pip install -e .— pulls in only the runtime dependencies (PyTorch, OpenCV, etc.).
Use this if you just want to run the library in your own projects. -
pip install -e .[dev]— adds the extra dev group, installing testing, linting, and coverage tools.
It also includes the documentation dependencies. Choose this when you plan to develop, run the test suite, build the documentation, or contribute code back to the project. -
pip install -e .[docs]— installs the Sphinx documentation toolchain without the testing and linting tools. Choose this when you only need to build the documentation. -
pip install -e .[notebooks]— adds the extra notebook group, jupyter notebook tools. Choose this when you plan to develop or run notebooks for the project. -
pip install -e .[examples]— adds the extra examples group, including optional dependencies to run all examples in the notebooks and scripts. Choose this when you plan to run the examples from the project.
TorchSig has a series of Jupyter notebooks in the examples/ directory. View the README inside examples/ to learn more.
TorchSig uses a unified dataset architecture. Create datasets using the Python API:
# define dataset metadata, can override defaults
dataset_metadata = TorchSigDefaults().default_dataset_metadata
# optionally, apply impairments
impairments = Impairments(level=0)
burst_impairments = impairments.signal_transforms
whole_signal_impairments = impairments.dataset_transforms
# create the dataset
dataset = TorchSigIterableDataset(
metadata=dataset_metadata,
transforms=[whole_signal_impairments, Spectrogram(fft_size=dataset_metadata["fft_size"])],
component_transforms=[burst_impairments],
)
# create a dataloader (reproducible)
dataloader = WorkerSeedingDataLoader(dataset, batch_size=2)
# save the dataset to disk
dataset_creator = DatasetCreator(
dataset_length=20,
dataloader=dataloader,
root="./sample_dataset",
overwrite=True,
multithreading=False,
)
dataset_creator.create()
# load the dataset in from disk
static_dataset = StaticTorchSigDataset(
root="./sample_dataset",
)
print(static_dataset[0])One option for running TorchSig is within Docker. Start by building the Docker container:
docker build -t torchsig -f docker/Dockerfile .And then you can launch a Docker instance:
docker run -it torchsigSee docker/README.md to learn more.
To contribute to our library, please make sure to run the following:
# pytests all pass
pytest
# pylint score > 9/10
pylint --rcfile=.pylintrc torchsig
# not required
# but helpful for maintaining PEP 8 Style Guide
ruff check torchsigBoth need to pass in order to contribute to our Github.
TorchSig provides many useful tools to facilitate and accelerate research on signals processing machine learning technologies:
- Unified Dataset Architecture: TorchSig features a single, flexible dataset system that supports both signal classification (single signal) and signal detection (multiple signals) tasks through configuration.
- Comprehensive Signal Library: Support for 60+ signal types across all major modulation families (FSK, QAM, PSK, ASK, OFDM, Analog) with realistic impairments and channel effects.
- Advanced Transform System: Numerous signals processing transforms enable existing ML techniques to be employed on signals data, with unified impairment models supporting perfect, cabled, and wireless channel conditions.
SignalandSignalMetadataObject: Enable signal objects and metadata to be seamlessly handled and operated on throughout the TorchSig infrastructure.TorchSigIterableDataset: Unified dataset class that synthetically creates, augments, and transforms signals datasets. Behavior (classification vs detection) is determined by configuration parameters.- Can generate samples infinitely when
num_samples=None, or finite datasets whennum_samplesis specified. - Dataset type determined by
num_signals_max: 1 for classification, >1 for detection tasks.
- Can generate samples infinitely when
DatasetCreator: Writes a PyTorchDataLoadercontaining aTorchSigIterableDatasetobjects to disk with progress tracking and memory optimization.StaticTorchSigDataset: Loads previously generated datasets from disk back into memory.- Can access previously generated samples efficiently.
- Supports both classification and detection datasets through unified interface.
Documentation can be found online or built locally by following the instructions below.
pip install -e ".[docs]"
make docs
firefox docs/build/html/index.html
To simplify environment setup and maintain code quality, this project uses a Makefile. This provides a standardized set of shortcuts for common development tasks, ensuring consistency across different environments.
| Command | Description | Tool Used |
|---|---|---|
make install |
Installs dependencies and the package in editable mode. | pip |
make test |
Runs the fast test suite by default, skipping tests marked slow and CPU-only tests marked slow_no_gpu. |
pytest |
make test TEST_MODE=full |
Runs the full test suite, including slow and slow_no_gpu tests. |
pytest |
make test TEST_MODE=fast |
Runs the default fast test suite explicitly. | pytest |
make test-cov |
Runs tests and generates a detailed coverage report. | pytest-cov |
make test-notebooks |
Executes all Jupyter notebooks to verify they run without errors. | jupyter |
make test-notebooks-clean |
Removes stamp files created by notebook execution. | shell |
make clean-notebooks |
Removes all output from executed notebooks. | jupyter |
make lint |
Performs static analysis to find bugs and style issues. | ruff |
make format |
Automatically formats the codebase to project standards. | ruff |
make fix |
Automatically fixes linting errors and formats the code. | ruff |
make clean |
Wipes __pycache__, test caches, and /tmp artifacts. |
shell |
make build |
Builds source distribution (sdist) and wheel for PyPI. | build |
make verify |
Validates distribution files and lists them (pre-publish check). | twine, shell |
make publish |
Uploads distribution files to PyPI. | twine |
make docs |
Builds the HTML documentation. | sphinx |
make open-docs |
Opens the built documentation in the default browser. | shell |
make benchmarks |
Runs the bencharks. | pytest-benchmarks |
make benchmarks-clean |
Removes previous benchmark results. | shell |
For a full list of available targets and descriptions, run:
make helpNote for Windows Users: make is a Unix utility. To use these commands on Windows, please use WSL (Windows Subsystem for Linux), Git Bash, or install make via Chocolatey.
TorchSig uses pre-commit to run automated checks before commits. After installing the development dependencies, install the Git hooks with:
pre-commit installTo run all pre-commit checks manually:
pre-commit run --all-filesThe hooks will automatically run on staged files when committing. If a hook modifies a file, review and stage the changes before committing again.
TorchSig is released under the MIT License. The MIT license is a popular open-source software license enabling free use, redistribution, and modifications, even for commercial purposes, provided the license is included in all copies or substantial portions of the software. TorchSig has no connection to MIT, other than through the use of this license.
| Title | Year | Cite (APA) |
|---|---|---|
| TorchSig 2.0: Dataset Customization, New Transforms and Future Plans | 2025 | Oh, E., Mullins, J., Carrick, M., Vondal, M., Hoffman, J., Leonardo, F., Toliver, P., Miller, R. (2025, September). TorchSig 2.0: Dataset Customization, New Transforms and Future Plans. In Proceedings of the GNU Radio Conference (Vol. 10, No. 1). |
| TorchSig: A GNU Radio Block and New Spectrogram Tools for Augmenting ML Training | 2024 | Vallance, P., Oh, E., Mullins, J., Gulati, M., Hoffman, J., & Carrick, M. (2024, September). TorchSig: A GNU Radio Block and New Spectrogram Tools for Augmenting ML Training. In Proceedings of the GNU Radio Conference (Vol. 9, No. 1). |
| Large Scale Radio Frequency Wideband Signal Detection & Recognition | 2022 | Boegner, L., Vanhoy, G., Vallance, P., Gulati, M., Feitzinger, D., Comar, B., & Miller, R. D. (2022). Large Scale Radio Frequency Wideband Signal Detection & Recognition. arXiv preprint arXiv:2211.10335. |
| Large Scale Radio Frequency Signal Classification | 2022 | Boegner, L., Gulati, M., Vanhoy, G., Vallance, P., Comar, B., Kokalj-Filipovic, S., ... & Miller, R. D. (2022). Large Scale Radio Frequency Signal Classification. arXiv preprint arXiv:2207.09918. |
Please cite TorchSig if you use it for your research or business.
@misc{torchsig,
title={Large Scale Radio Frequency Signal Classification},
author={Luke Boegner and Manbir Gulati and Garrett Vanhoy and Phillip Vallance and Bradley Comar and Silvija Kokalj-Filipovic and Craig Lennon and Robert D. Miller},
year={2022},
archivePrefix={arXiv},
eprint={2207.09918},
primaryClass={cs-LG},
note={arXiv:2207.09918}
url={https://arxiv.org/abs/2207.09918}
}