bruAristimunha commited on
Commit
5b1a9e4
·
verified ·
1 Parent(s): 7553ef8

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

Browse files

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.

Files changed (1) hide show
  1. README.md +23 -0
README.md CHANGED
@@ -1,3 +1,8 @@
 
 
 
 
 
1
  # plot_bcic_iv_2a_moabb_cropped
2
 
3
  Pretrained artifacts for the Braindecode tutorial `examples/model_building/plot_bcic_iv_2a_moabb_cropped.py`.
@@ -9,3 +14,21 @@ These files are meant to be loaded by the tutorial so the docs can show stable p
9
  - `params.safetensors`: classifier parameters
10
  - `history.json`: Skorch training history used by the tutorial plots
11
  - `metadata.json`: summary metrics for the stored checkpoint
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets:
3
+ - nm000139
4
+ ---
5
+
6
  # plot_bcic_iv_2a_moabb_cropped
7
 
8
  Pretrained artifacts for the Braindecode tutorial `examples/model_building/plot_bcic_iv_2a_moabb_cropped.py`.
 
14
  - `params.safetensors`: classifier parameters
15
  - `history.json`: Skorch training history used by the tutorial plots
16
  - `metadata.json`: summary metrics for the stored checkpoint
17
+
18
+ <!-- braindecode:training-data -->
19
+ ## Training data
20
+
21
+ The data that shaped these weights. The ids are the `datasets:` field of this card,
22
+ which braindecode reads into `model.card_data.datasets` when you call
23
+ `from_pretrained` (MOABB class names for MOABB datasets, OpenNeuro / NEMAR
24
+ accessions, Hub ids otherwise). `nm`/`on` ids are datasets on [NEMAR](https://nemar.org)
25
+ (`on` = NEMAR's copy of the OpenNeuro dataset with the same number). Rows marked
26
+ `private:` are data the authors did not release; they are not in `datasets:`.
27
+
28
+ | `datasets:` id | Dataset | Stage | Modality |
29
+ |---|---|---|---|
30
+ | [`nm000139`](https://nemar.org/dataset/nm000139) | [BCI Competition IV dataset 2a (4-class motor imagery, 9 subjects, 22 EEG channels), MOABB BNCI2014_001](https://moabb.neurotechx.com/docs/generated/moabb.datasets.BNCI2014_001.html) | training | EEG |
31
+
32
+ - Supervised cropped-decoding ShallowFBCSPNet (dense prediction, 1000-sample windows), subject 3 only. Gradient updates on session '0train' (MOABB session name); session '1test' is the skorch validation set and EarlyStopping(load_best=True) restores the epoch with the best validation loss, so the evaluation session selected the saved epoch (no gradient updates on it). Up to 100 epochs, patience 10. Same preprocessing as the trialwise tutorial.
33
+ - Sources: braindecode git 05e606dd^:scripts/train_and_push_tutorial_checkpoints.py (offline trainer that produced the tutorial checkpoints; removed from the repo in 05e606dd, 2026-04-19, and moved to https://gist.github.com/bru...; https://huggingface.co/braindecode/plot_bcic_iv_2a_moabb_cropped/resolve/main/metadata.json; W/bd/examples/model_building/plot_bcic_iv_2a_moabb_cropped.py (gallery example of the same name).
34
+