EpiNT pretrained backbone (authors' representations.bin)

README.md

@@ -0,0 +1,95 @@
+---
+library_name: braindecode
+license: mit
+tags:
+- EpiNT
+- braindecode
+- eeg
+- ieeg
+- seeg
+- ecog
+- epilepsy
+- foundation-model
+- model_hub_mixin
+- pytorch_model_hub_mixin
+---
+
+# EpiNT — Epilepsy Neurophysiological Transformer
+
+Pretrained weights for [`braindecode.models.EpiNT`](https://braindecode.org/stable/generated/braindecode.models.EpiNT.html),
+a braindecode port of **EpiNT** (Zhang et al., 2025), a channel-independent
+transformer pretrained on 2,700 hours of scalp EEG and intracranial EEG (ECoG,
+sEEG) from 1,199 patients.
+
+- Paper: [Cross-Modal Epileptic Signal Harmonization: Frequency Domain Mapping Quantization for Pre-training a Unified Neurophysiological Transformer](https://arxiv.org/abs/2506.17068) (arXiv:2506.17068)
+- Original code & weights: [RunKZhang/EpiNT](https://github.com/RunKZhang/EpiNT/tree/781e1d61e72899d59b4955b4c075404eef532a40)
+- braindecode docs: https://braindecode.org
+
+## Provenance & license
+
+These weights are the authors' pretrained backbone, `weights/representations.bin`
+at commit `781e1d61` (sha256 `b4449974058fd6d314debeef0a09c05730a70e822f17244e88d1ee4e7d9a8b1d`),
+converted into the braindecode `EpiNT` module through its `mapping` (all 102
+backbone tensors; the pretraining mask token `embed.mask_encoding` is dropped).
+The classification head is not part of the release and is randomly initialized.
+
+The original release is under the MIT license, which this repository preserves:
+
+> MIT License
+>
+> Copyright (c) 2024 Auton Lab, Carnegie Mellon University
+>
+> 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.
+
+## Model configuration
+
+This checkpoint uses the paper's configuration (~19.2M parameters):
+
+| | |
+|---|---|
+| `patch_size` | 256 |
+| `embed_dim` | 512 |
+| `ffn_dim` | 2048 |
+| `n_layers` | 6 |
+| `n_heads` | 8 |
+| `n_times` | 3072 (12 patches) |
+
+Each encoder `LayerNorm` spans tokens and features, so keep `n_times=3072`:
+12 s at 256 Hz for scalp EEG, or 3 s at 1024 Hz for iEEG. Signals are expected
+in microvolts. Every channel is encoded alone and the class tokens are averaged,
+so `n_chans` is free; `n_outputs` resets the task head.
+
+## Usage
+
+```python
+from braindecode.models import EpiNT
+
+model = EpiNT.from_pretrained("braindecode/epint-pretrained", n_chans=17, n_outputs=2)
+```
+
+## Citation
+
+```bibtex
+@article{zhang2025epint,
+  title   = {Cross-Modal Epileptic Signal Harmonization: Frequency Domain Mapping Quantization for Pre-training a Unified Neurophysiological Transformer},
+  author  = {Zhang, Runkai and Yu, Hua and Gan, John Q. and Wang, Haixian},
+  journal = {arXiv preprint arXiv:2506.17068},
+  year    = {2025}
+}
+```

config.json

@@ -0,0 +1,18 @@
+{
+  "patch_size": 256,
+  "embed_dim": 512,
+  "n_layers": 6,
+  "n_heads": 8,
+  "ffn_dim": 2048,
+  "drop_prob": 0.1,
+  "head_drop_prob": 0.1,
+  "rope_theta": 10000.0,
+  "activation": "torch.nn.modules.activation.ReLU",
+  "n_outputs": 2,
+  "n_chans": 1,
+  "chs_info": null,
+  "n_times": 3072,
+  "input_window_seconds": null,
+  "sfreq": 256,
+  "braindecode_version": "1.8.1"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9550cc54c2d0f8d103610db529a1046e91f056a5e3b10b3d33ec622ed6b4a5b
+size 76789328

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82979f06c4fd8765a845b8a7d3771d0e4664ffc5b7ec8cde46b3273897e98eb9
+size 76816422