Today
-
Complete training-data provenance and Braindecode software citation
README-only metadata audit, 2026-11-09. Braindecode Zenodo citation. No weights, config, license, or upstream paper citations changed.
Yesterday
-
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.
-
Upload braindecode_labram_base.pt
-
Update README.md
-
Update README.md
Shallow convertion for braindecode:
```
#!/usr/bin/env python3
"""
Complete LaBraM Weight Transfer Script
Combines explicit weight mapping with full backbone transfer.
Uses precise key renaming to transfer all compatible parameters.
Transfers weights from LaBraM checkpoint to Braindecode Labram model.
"""
import torch
import argparse
from braindecode.models import Labram
def create_weight_mapping():
"""
Create comprehensive weight mapping from LaBraM to Braindecode.
Includes:
- Temporal convolution layers (patch_embed)
- All transformer blocks
- Position embeddings
- Other backbone components
"""
return {
# Temporal Convolution Layers
'student.patch_embed.conv1.weight': 'patch_embed.temporal_conv.conv1.weight',
'student.patch_embed.conv1.bias': 'patch_embed.temporal_conv.conv1.bias',
'student.patch_embed.norm1.weight': 'patch_embed.temporal_conv.norm1.weight',
'student.patch_embed.norm1.bias': 'patch_… -
initial commit
