Card: list the training data (datasets:, base_model:) and add a Training data section
Browse filesFollow-up to braindecode PR #1493: `from_pretrained` reads `datasets:` into `model.card_data`. Ids: MOABB names, NEMAR nm/on ids, OpenNeuro ds ids; evidence per checkpoint.
README.md
CHANGED
|
@@ -10,8 +10,9 @@ tags:
|
|
| 10 |
- population-transformer
|
| 11 |
- pytorch_model_hub_mixin
|
| 12 |
- model_hub_mixin
|
| 13 |
-
base_model: PopulationTransformer/popt_brainbert_stft
|
| 14 |
pipeline_tag: feature-extraction
|
|
|
|
|
|
|
| 15 |
---
|
| 16 |
|
| 17 |
# PopulationTransformer (PopT) — braindecode weights
|
|
@@ -61,6 +62,25 @@ layer on the `CLS` token, as upstream's `PtDownstreamModel.linear_out`) is
|
|
| 61 |
| activation | GELU |
|
| 62 |
| parameters (encoder + spec head) | ~20.0M |
|
| 63 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 64 |
## Usage
|
| 65 |
|
| 66 |
```python
|
|
|
|
| 10 |
- population-transformer
|
| 11 |
- pytorch_model_hub_mixin
|
| 12 |
- model_hub_mixin
|
|
|
|
| 13 |
pipeline_tag: feature-extraction
|
| 14 |
+
datasets:
|
| 15 |
+
- nm000253
|
| 16 |
---
|
| 17 |
|
| 18 |
# PopulationTransformer (PopT) — braindecode weights
|
|
|
|
| 62 |
| activation | GELU |
|
| 63 |
| parameters (encoder + spec head) | ~20.0M |
|
| 64 |
|
| 65 |
+
<!-- braindecode:training-data -->
|
| 66 |
+
## Training data
|
| 67 |
+
|
| 68 |
+
The data that shaped these weights. The ids are the `datasets:` field of this card,
|
| 69 |
+
which braindecode reads into `model.card_data.datasets` when you call
|
| 70 |
+
`from_pretrained` (MOABB class names for MOABB datasets, OpenNeuro / NEMAR
|
| 71 |
+
accessions, Hub ids otherwise). `nm`/`on` ids are datasets on [NEMAR](https://nemar.org)
|
| 72 |
+
(`on` = NEMAR's copy of the OpenNeuro dataset with the same number). Rows marked
|
| 73 |
+
`private:` are data the authors did not release; they are not in `datasets:`.
|
| 74 |
+
|
| 75 |
+
| `datasets:` id | Dataset | Stage | Modality |
|
| 76 |
+
|---|---|---|---|
|
| 77 |
+
| [`nm000253`](https://nemar.org/dataset/nm000253) | [Brain Treebank: large-scale intracranial recordings from naturalistic language stimuli (Wang et al., NeurIPS 2024 D&B; 10 subjects, 26 movie sessions, ~43.5 h sEEG)](https://braintreebank.dev/) | pretraining, pretraining | iEEG |
|
| 78 |
+
|
| 79 |
+
- frozen per-channel front end: BrainBERT stft_large_pretrained (itself pretrained on Brain Treebank)
|
| 80 |
+
- PopT self-supervised pretraining on BrainBERT features; 19 of 26 movies (same 10 subjects), 7 held out for downstream
|
| 81 |
+
- Sources: https://huggingface.co/PopulationTransformer/popt_brainbert_stft README; arXiv:2406.03044 (ICLR 2025) p.5, Section 4; https://github.com/czlwang/PopulationTransformer README.
|
| 82 |
+
|
| 83 |
+
|
| 84 |
## Usage
|
| 85 |
|
| 86 |
```python
|