Upload folder using huggingface_hub
Browse files
README.md
CHANGED
|
@@ -14,21 +14,6 @@ This repository contains a PyTorch checkpoint for a basic V-JEPA2-based offline
|
|
| 14 |
|
| 15 |
The checkpoint stores decoder weights and metadata for an offline encoder. This repository includes a custom `transformers.AutoModel` wrapper and does not require the original training codebase.
|
| 16 |
|
| 17 |
-
## Files
|
| 18 |
-
|
| 19 |
-
- `vjepa2_offline_encoder.pth`: PyTorch checkpoint containing decoder weights, decoding-unit selection metadata, feature-extractor configuration, and registered attributes.
|
| 20 |
-
- `config.json`, `configuration_vjepa2_fmri_encoder.py`, `modeling_vjepa2_fmri_encoder.py`: custom Transformers files for `AutoModel` loading.
|
| 21 |
-
- `requirements.txt`: minimal Python dependencies.
|
| 22 |
-
|
| 23 |
-
## Data
|
| 24 |
-
|
| 25 |
-
This checkpoint was trained using data from:
|
| 26 |
-
|
| 27 |
-
- **BOLD Moments Dataset (BMD)**: whole-brain fMRI responses to short naturalistic videos.
|
| 28 |
-
- **Social interaction video fMRI dataset from Emalie McMahon and collaborators**: fMRI responses to naturalistic two-person social action videos.
|
| 29 |
-
|
| 30 |
-
This repository does not include the underlying fMRI datasets or stimulus videos.
|
| 31 |
-
|
| 32 |
## Input/Output Contract
|
| 33 |
|
| 34 |
The intended input is a short video clip corresponding to the training stimulus duration:
|
|
@@ -85,6 +70,21 @@ with torch.no_grad():
|
|
| 85 |
prediction = model.forward_features(features)
|
| 86 |
```
|
| 87 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 88 |
## Citations
|
| 89 |
|
| 90 |
If you use this checkpoint, please cite the source datasets:
|
|
|
|
| 14 |
|
| 15 |
The checkpoint stores decoder weights and metadata for an offline encoder. This repository includes a custom `transformers.AutoModel` wrapper and does not require the original training codebase.
|
| 16 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
## Input/Output Contract
|
| 18 |
|
| 19 |
The intended input is a short video clip corresponding to the training stimulus duration:
|
|
|
|
| 70 |
prediction = model.forward_features(features)
|
| 71 |
```
|
| 72 |
|
| 73 |
+
## Data
|
| 74 |
+
|
| 75 |
+
This checkpoint was trained using data from:
|
| 76 |
+
|
| 77 |
+
- **BOLD Moments Dataset (BMD)**: whole-brain fMRI responses to short naturalistic videos.
|
| 78 |
+
- **Social interaction video fMRI dataset from Emalie McMahon and collaborators**: fMRI responses to naturalistic two-person social action videos.
|
| 79 |
+
|
| 80 |
+
This repository does not include the underlying fMRI datasets or stimulus videos.
|
| 81 |
+
|
| 82 |
+
## Files
|
| 83 |
+
|
| 84 |
+
- `vjepa2_offline_encoder.pth`: PyTorch checkpoint containing decoder weights, decoding-unit selection metadata, feature-extractor configuration, and registered attributes.
|
| 85 |
+
- `config.json`, `configuration_vjepa2_fmri_encoder.py`, `modeling_vjepa2_fmri_encoder.py`: custom Transformers files for `AutoModel` loading.
|
| 86 |
+
- `requirements.txt`: minimal Python dependencies.
|
| 87 |
+
|
| 88 |
## Citations
|
| 89 |
|
| 90 |
If you use this checkpoint, please cite the source datasets:
|