Add architecture-only model card

README.md

@@ -0,0 +1,229 @@
+---
+license: bsd-3-clause
+library_name: braindecode
+pipeline_tag: feature-extraction
+tags:
+  - eeg
+  - biosignal
+  - pytorch
+  - neuroscience
+  - braindecode
+  - foundation-model
+  - convolutional
+---
+
+# SignalJEPA_Contextual
+
+Contextual downstream architecture introduced in signal-JEPA Guetschel, P et al (2024) .
+
+> **Architecture-only repository.** This repo documents the
+> `braindecode.models.SignalJEPA_Contextual` class. **No pretrained weights are
+> distributed here** — instantiate the model and train it on your own
+> data, or fine-tune from a published foundation-model checkpoint
+> separately.
+
+## Quick start
+
+```bash
+pip install braindecode
+```
+
+```python
+from braindecode.models import SignalJEPA_Contextual
+
+model = SignalJEPA_Contextual(
+    n_chans=22,
+    sfreq=250,
+    input_window_seconds=4.0,
+    n_outputs=4,
+)
+```
+
+The signal-shape arguments above are example defaults — adjust them
+to match your recording.
+
+## Documentation
+
+- Full API reference (parameters, references, architecture figure):
+  <https://braindecode.org/stable/generated/braindecode.models.SignalJEPA_Contextual.html>
+- Interactive browser with live instantiation:
+  <https://huggingface.co/spaces/braindecode/model-explorer>
+- Source on GitHub: <https://github.com/braindecode/braindecode/blob/master/braindecode/models/signal_jepa.py#L489>
+
+## Architecture description
+
+The block below is the rendered class docstring (parameters,
+references, architecture figure where available).
+
+<div class='bd-doc'><main>
+<p>Contextual downstream architecture introduced in signal-JEPA Guetschel, P et al (2024) [1]_.</p>
+  <span style="display:inline-block;padding:2px 8px;border-radius:4px;background:#5cb85c;color:white;font-size:11px;font-weight:600;margin-right:4px;">Convolution</span>
+
+:bdg-dark-line:`Channel`<span style="display:inline-block;padding:2px 8px;border-radius:4px;background:#d9534f;color:white;font-size:11px;font-weight:600;margin-right:4px;">Foundation Model</span>
+
+
+
+ This architecture is one of the variants of :class:`SignalJEPA`
+ that can be used for classification purposes.
+
+ .. figure:: https://braindecode.org/dev/_static/model/sjepa_contextual.jpg
+     :align: center
+     :alt: sJEPA Contextual.
+
+ .. versionadded:: 0.9
+
+ .. rubric:: Pretrained Weights
+
+ Loads the pre-trained SSL encoder (feature encoder + transformer +
+ channel embedding) from the shared checkpoints. Two hub IDs are
+ available:
+
+ - ``braindecode/signal-jepa``: with pre-trained channel embeddings
+   (62 channels, ``channel_embedding='pretrain_aligned'``).
+ - ``braindecode/signal-jepa_without-chans``: without channel embeddings.
+   The embedding table is re-initialized from your channel locations
+   (``channel_embedding='scratch'``, the default).
+
+ .. important::
+    **Pre-trained Weights Available**
+
+    .. code:: python
+        from braindecode.models import SignalJEPA_Contextual
+
+        # Use pre-trained channel embeddings (subset of the 62 channels):
+        model = SignalJEPA_Contextual.from_pretrained(
+            "braindecode/signal-jepa",
+            n_times=2048,  # required: the SSL checkpoint is n_times-agnostic
+            n_outputs=4,
+            strict=False,  # classifier head is not in the SSL checkpoint
+        )
+
+        # Arbitrary channel set:
+        model = SignalJEPA_Contextual.from_pretrained(
+            "braindecode/signal-jepa_without-chans",
+            chs_info=[{"ch_name": "A", "loc": [...]}, ...],
+            n_times=2048,
+            n_outputs=4,
+            strict=False,
+        )
+
+    Requires installing ``braindecode[hub]`` for Hub integration.
+
+ .. rubric:: Usage
+
+ .. code:: python
+     from braindecode.models import SignalJEPA_Contextual
+
+     model = SignalJEPA_Contextual(
+         chs_info=[{"ch_name": "Fp1", "loc": [...]}, ...],
+         input_window_seconds=16.0,
+         sfreq=128,
+         n_outputs=4,  # e.g., 4-class classification
+     )
+
+     # Forward: (batch, n_chans, n_times) -> (batch, n_outputs)
+     output = model(eeg_data)
+
+ .. warning::
+
+     Pre-trained at **128 Hz** on EEG bandpass-filtered between
+     **0.5 and 40 Hz** and rescaled by a factor of :math:`10^{6}`
+     (volts to microvolts). Apply the same preprocessing to your
+     data to match the pre-training distribution.
+
+ Parameters
+ ----------
+ n_spat_filters : int
+     Number of spatial filters.
+
+ References
+ ----------
+ .. [1] Guetschel, P., Moreau, T., & Tangermann, M. (2024).
+     S-JEPA: towards seamless cross-dataset transfer through dynamic spatial attention.
+     In 9th Graz Brain-Computer Interface Conference, https://www.doi.org/10.3217/978-3-99161-014-4-003
+
+ .. rubric:: Hugging Face Hub integration
+
+ When the optional ``huggingface_hub`` package is installed, all models
+ automatically gain the ability to be pushed to and loaded from the
+ Hugging Face Hub. Install with::
+
+     pip install braindecode[hub]
+
+ **Pushing a model to the Hub:**
+
+ .. code::
+     from braindecode.models import SignalJEPA_Contextual
+
+     # Train your model
+     model = SignalJEPA_Contextual(n_chans=22, n_outputs=4, n_times=1000)
+     # ... training code ...
+
+     # Push to the Hub
+     model.push_to_hub(
+         repo_id="username/my-signaljepa_contextual-model",
+         commit_message="Initial model upload",
+     )
+
+ **Loading a model from the Hub:**
+
+ .. code::
+     from braindecode.models import SignalJEPA_Contextual
+
+     # Load pretrained model
+     model = SignalJEPA_Contextual.from_pretrained("username/my-signaljepa_contextual-model")
+
+     # Load with a different number of outputs (head is rebuilt automatically)
+     model = SignalJEPA_Contextual.from_pretrained("username/my-signaljepa_contextual-model", n_outputs=4)
+
+ **Extracting features and replacing the head:**
+
+ .. code::
+     import torch
+
+     x = torch.randn(1, model.n_chans, model.n_times)
+     # Extract encoder features (consistent dict across all models)
+     out = model(x, return_features=True)
+     features = out["features"]
+
+     # Replace the classification head
+     model.reset_head(n_outputs=10)
+
+ **Saving and restoring full configuration:**
+
+ .. code::
+     import json
+
+     config = model.get_config()            # all __init__ params
+     with open("config.json", "w") as f:
+         json.dump(config, f)
+
+     model2 = SignalJEPA_Contextual.from_config(config)    # reconstruct (no weights)
+
+ All model parameters (both EEG-specific and model-specific such as
+ dropout rates, activation functions, number of filters) are automatically
+ saved to the Hub and restored when loading.
+
+ See :ref:`load-pretrained-models` for a complete tutorial.</main>
+</div>
+
+## Citation
+
+Please cite both the original paper for this architecture (see the
+*References* section above) and braindecode:
+
+```bibtex
+@article{aristimunha2025braindecode,
+  title   = {Braindecode: a deep learning library for raw electrophysiological data},
+  author  = {Aristimunha, Bruno and others},
+  journal = {Zenodo},
+  year    = {2025},
+  doi     = {10.5281/zenodo.17699192},
+}
+```
+
+## License
+
+BSD-3-Clause for the model code (matching braindecode).
+Pretraining-derived weights, if you fine-tune from a checkpoint,
+inherit the licence of that checkpoint and its training corpus.