# Braindecode > Open-source PyTorch library for decoding raw EEG, MEG, iEEG/ECoG, fNIRS, and sEMG time-series signals. Braindecode provides a unified PyTorch interface (`EEGModuleMixin`) for neural and biosignal decoding: - **103 neural decoding architectures**, including **48 foundation models** (e.g., REVE, CBraMod, LaBraM, BIOT, Bendr, SeizureTransformer, EEGPT) alongside classic baselines (`EEGNet`, `ShallowFBCSPNet`, `Deep4Net`, `ATCNet`, `EEGConformer`), with Hugging Face Hub weight loading and export. - **`ChannelLayer` universal kriging** for continuous 3D sensor-position interpolation and cross-montage transfer across arbitrary channel layouts. - **`SzCORE` clinical evaluation** for standardized seizure detection scoring, alongside trialwise and cropped continuous-recording training and inference. - **MNE / MOABB / EEGDash ecosystem integration**: train directly on MNE `Raw` and `Epochs` objects, benchmark reproducibly with MOABB, stream large-scale recordings via EEGDash, and customize training loops with PyTorch or skorch. Canonical user documentation is served under `/stable/`. Machine-readable Markdown mirrors for LLM agents are indexed at `/dev/llms.txt`. ## Canonical Documentation (`/stable/` HTML) - [Stable Documentation Home](https://braindecode.org/stable/index.html): Overview, quickstart, and navigation across stable release docs. - [Installation Guide](https://braindecode.org/stable/install/install_pip.html): PyPI, conda, optional extras, and GPU/CPU setup. - [API Reference](https://braindecode.org/stable/api.html): Complete reference for `braindecode.models`, `modules`, `datasets`, `preprocessing`, `augmentation`, `training`, and `functional`. - [The Decoding Problem & Input Conventions](https://braindecode.org/stable/models/models.html): Standardized `(batch, n_chans, n_times)` tensor contracts and `EEGModuleMixin` metadata. - [Model Summary Table](https://braindecode.org/stable/models/models_table.html): Filterable catalog of neural decoding architectures, foundation models, parameter counts, and input requirements. - [Model Categorization](https://braindecode.org/stable/models/models_categorization.html): Architectural taxonomy spanning convolutional, recurrent, attention/transformer, Riemannian/SPD, graph, and foundation models. - [Channel Strategies & `ChannelLayer`](https://braindecode.org/stable/user_guide/channel_strategies.html): Universal kriging and cross-montage channel adaptation (`braindecode.modules.ChannelLayer`). - [`ChannelLayer` API Reference](https://braindecode.org/stable/generated/channels/braindecode.modules.ChannelLayer.html): Module parameters and 3D electrode coordinate resolution. - [Cropped Training & Inference](https://braindecode.org/stable/user_guide/cropped_inference.html): Dense sliding-window training and prediction on continuous recordings. - [Tutorials & Applied Examples](https://braindecode.org/stable/auto_examples/index.html): End-to-end gallery covering MNE data loading, MOABB benchmarks, foundation-model fine-tuning, channel interpolation, and clinical decoding. - [Agent & LLM Documentation Guide](https://braindecode.org/stable/agent_docs.html): Guidance for coding agents working with Braindecode APIs. - [Citing Braindecode](https://braindecode.org/stable/cite.html): Primary DOI and publication references. ## Markdown Documentation for LLMs - [Development LLM Index (`llms.txt`)](https://braindecode.org/dev/llms.txt): Version-stamped Markdown index for the latest development build. - [Stable LLM Index (`llms.txt`)](https://braindecode.org/stable/llms.txt): Version-stamped Markdown index for the current stable build. - [API Reference (Markdown)](https://braindecode.org/dev/api.md): Plain-Markdown API overview and module index. - [Decoding Problem & Models Overview (Markdown)](https://braindecode.org/dev/models/models.md): Input shapes, metadata arguments, and usage patterns in Markdown. - [Model Catalog (Markdown)](https://braindecode.org/dev/models/models_table.md): Markdown summary of architectures and foundation models. - [Tutorials Index (Markdown)](https://braindecode.org/dev/auto_examples/index.md): Markdown index of runnable examples. - [Installation from PyPI (Markdown)](https://braindecode.org/dev/install/install_pip.md): Installation instructions in Markdown. - [Citing Braindecode (Markdown)](https://braindecode.org/dev/cite.md): BibTeX and citation details in Markdown. ## Ecosystem & Source - [GitHub Repository](https://github.com/braindecode/braindecode): Source code, issue tracker, and pull requests. - [Hugging Face Organization](https://huggingface.co/braindecode): Pretrained foundation-model checkpoints and model cards. - [MNE-Python](https://mne.tools/stable/index.html): Neurophysiological signal structures and preprocessing. - [MOABB](https://moabb.neurotechx.com/docs/index.html): Mother of All BCI Benchmarks reproducible evaluation suite. - [EEGDash](https://eegdash.org/): Large-scale EEG/MEG dataset hub and BIDS data access.