Card: list the training data (datasets:, base_model:) and add a Training data section

Follow-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

@@ -2,6 +2,25 @@
 license: mit
 tags:
 - braindecode
+datasets:
+- BCIC-IV-1
+- Emobrain
+- WAY-EEG-GAL
+- Inria-BCI-Challenge
+- on004362
+- Trujillo2020
+- Trujillo2017
+- SEED
+- SEED-IV
+- SEED-GER
+- SEED-FRA
+- nm000385
+- SPIS-Resting-State
+- nm000216
+- TUAR
+- TUEP
+- TUSZ
+- TUSL
 ---
 
 Shallow conversion from the original weight for braindecode.
@@ -303,4 +322,41 @@ Examples:
     print(f"Skipped:               {stats['skipped']}")
     print("="*70)
 
-```
+```
+
+<!-- braindecode:training-data -->
+## Training data
+
+The data that shaped these weights. The ids are the `datasets:` field of this card,
+which braindecode reads into `model.card_data.datasets` when you call
+`from_pretrained` (MOABB class names for MOABB datasets, OpenNeuro / NEMAR
+accessions, Hub ids otherwise). `nm`/`on` ids are datasets on [NEMAR](https://nemar.org)
+(`on` = NEMAR's copy of the OpenNeuro dataset with the same number). Rows marked
+`private:` are data the authors did not release; they are not in `datasets:`.
+
+| `datasets:` id | Dataset | Stage | Modality |
+|---|---|---|---|
+| `BCIC-IV-1` | [BCI Competition IV dataset 1 (Blankertz et al. 2007; 59 ch, 7 subjects, motor imagery)](https://www.bbci.de/competition/iv/desc_1.html) | pretraining | EEG |
+| `Emobrain` | Emobrain, eNTERFACE'06 multimodal emotion dataset (Savran et al. 2006; 64-ch EEG + fNIRS, 16 subjects) | pretraining | EEG |
+| `WAY-EEG-GAL` | [Grasp-and-Lift EEG Challenge (Luciw et al. 2014; 32 ch, 12 subjects)](https://www.kaggle.com/c/grasp-and-lift-eeg-detection) | pretraining | EEG |
+| `Inria-BCI-Challenge` | [Inria BCI Challenge / Kaggle ERN (Margaux et al. 2012; P300 speller error-related potentials, 26 subjects)](https://www.kaggle.com/c/inria-bci-challenge) | pretraining | EEG |
+| [`on004362`](https://nemar.org/dataset/on004362) | [EEG Motor Movement/Imagery Dataset (EEGMMIDB, Schalk et al. 2004; 109 subjects)](https://physionet.org/content/eegmmidb/1.0.0/) | pretraining | EEG |
+| `Trujillo2020` | [Raw EEG Data (Trujillo 2020; information-integration and rule-based categorization tasks, 64 ch)](https://doi.org/10.18738/T8/SS2NHB) | pretraining | EEG |
+| `Trujillo2017` | [Resting State EEG Data (Trujillo et al. 2017; 22 subjects, 64 ch)](https://doi.org/10.3389/fnins.2017.00425) | pretraining | EEG |
+| `SEED` | [SJTU Emotion EEG Dataset (15 subjects)](https://bcmi.sjtu.edu.cn/home/seed/) | pretraining | EEG |
+| `SEED-IV` | [SJTU Emotion EEG Dataset IV (15 subjects)](https://bcmi.sjtu.edu.cn/home/seed/) | pretraining | EEG |
+| `SEED-GER` | [SEED-GER, German-subject emotion EEG (8 subjects; Liu et al. 2022)](https://bcmi.sjtu.edu.cn/home/seed/) | pretraining | EEG |
+| `SEED-FRA` | [SEED-FRA, French-subject emotion EEG (8 subjects; Liu et al. 2022)](https://bcmi.sjtu.edu.cn/home/seed/) | pretraining | EEG |
+| [`nm000385`](https://nemar.org/dataset/nm000385) | [Siena Scalp EEG Database (Detti et al. 2020; 14 patients)](https://physionet.org/content/siena-scalp-eeg/1.0.0/) | pretraining | EEG |
+| `SPIS-Resting-State` | [SPIS Resting State Dataset (Torkamani-Azar et al. 2020; 10 subjects)](https://github.com/mastaneht/SPIS-Resting-State-Dataset) | pretraining | EEG |
+| [`nm000216`](https://nemar.org/dataset/nm000216) | [Brain Invaders calibration-less P300 BCI with modulation of flash duration (bi2015a; Korczowski et al. 2019) - LaBraM's 'Target Versus Non-Target'](https://zenodo.org/record/3266930) | pretraining | EEG |
+| `TUAR` | [TUH EEG Artifact Corpus](https://isip.piconepress.com/projects/nedc/html/tuh_eeg/) | pretraining | EEG |
+| `TUEP` | [TUH EEG Epilepsy Corpus](https://isip.piconepress.com/projects/nedc/html/tuh_eeg/) | pretraining | EEG |
+| `TUSZ` | [TUH EEG Seizure Corpus](https://isip.piconepress.com/projects/nedc/html/tuh_eeg/) | pretraining | EEG |
+| `TUSL` | [TUH EEG Slowing Corpus](https://isip.piconepress.com/projects/nedc/html/tuh_eeg/) | pretraining | EEG |
+| `private:LaBraM-self-collected-EEG` | LaBraM authors' self-collected EEG (>140 subjects, 62 ch, 1000 Hz, ESI NeuroScan; 342.23 h; not released) | pretraining | EEG |
+
+- same corpus for the vector-quantized neural tokenizer (vqnsp) and masked-code pretraining; 2,534.78 h total; TUAB, TUEV, SEED-V and MoBI (downstream) excluded per paper
+- **Incomplete list.** Same as braindecode/labram-pretrained: private self-collected EEG not itemised; released-checkpoint cohort unconfirmed. A dataset missing here is therefore not proof that these weights never saw it.
+- Sources: arXiv:2405.18765 Appendix D (Pre-training Dataset Description); arXiv:2405.18765 Appendix D; arXiv:2405.18765 References; arXiv:2405.18765 Sec. 3 Experiments, paragraph 'Pre-training & Fine-tuning'.
+