Add Simulusion 1.1.0 demo checkpoint

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+jepa_neighbours.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,76 @@
+---
+library_name: simulusion
+license: mit
+pipeline_tag: image-feature-extraction
+datasets:
+- uoft-cs/cifar10
+tags:
+- simulusion
+- pytorch
+- ijepa
+- jepa
+- self-supervised-learning
+- vision-transformer
+- cifar10
+---
+
+# I-JEPA ViT-Tiny on CIFAR-10
+
+![Test images and their nearest training images in I-JEPA feature space](jepa_neighbours.png)
+
+A demo checkpoint from [Simulusion](https://github.com/ParamThakkar123/simulusion), an open-source PyTorch library that puts world models behind one API. It implements the paper **[Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture (Assran et al., CVPR 2023)](https://arxiv.org/abs/2301.08243)**.
+
+I-JEPA learns image features without labels and without reconstructing pixels: it predicts the representations of masked image blocks from a visible context block. In Simulusion it is a representation-learning component for world models.
+
+## Results
+
+A ViT-Tiny (patch size 4) trained for 30 epochs on CIFAR-10, about an hour on a 4 GB RTX 3050 laptop GPU. Classifying test images by their 20 nearest training images in feature space scores **41.1%**, against 29.6% for the same network at random initialisation (chance is 10%).
+
+This is a short demo run sized for a laptop GPU, meant to show what the model does. It is **not** a reproduction of the paper's results. More clips and learning curves are in the [gallery](https://paramthakkar123.github.io/simulusion/gallery.html).
+
+## Usage
+
+```bash
+pip install simulusion huggingface_hub
+```
+
+```python
+from simulusion.models.jepa_agent import JEPAAgent
+
+agent = JEPAAgent.from_pretrained("simillusion/ijepa-vit-tiny-cifar10")
+```
+
+Checkpoints are plain state dicts and load with `torch.load(..., weights_only=True)`. They include optimizer state, so you can continue training from them.
+
+## Training
+
+Reproduce this checkpoint from a clone of the repository:
+
+```bash
+python demos/jepa_demo.py train --epochs 30
+```
+
+## Citation
+
+If you use this model, please cite the original paper:
+
+```bibtex
+@inproceedings{assran2023ijepa,
+  title     = {Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture},
+  author    = {Assran, Mahmoud and Duval, Quentin and Misra, Ishan and Bojanowski, Piotr and Vincent, Pascal and Rabbat, Michael and LeCun, Yann and Ballas, Nicolas},
+  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
+  year      = {2023},
+  url       = {https://arxiv.org/abs/2301.08243},
+}
+```
+
+and Simulusion:
+
+```bibtex
+@software{simulusion,
+  title  = {Simulusion: a PyTorch library for world models},
+  author = {Thakkar, Param},
+  year   = {2026},
+  url    = {https://github.com/ParamThakkar123/simulusion},
+}
+```

config.yaml

@@ -0,0 +1,65 @@
+data:
+  batch_size: 128
+  color_jitter_strength: 0.0
+  crop_scale:
+  - 0.3
+  - 1.0
+  crop_size: 32
+  dataset: cifar10
+  download: false
+  image_folder: train
+  num_workers: 0
+  pin_mem: true
+  root_path: data
+  use_color_distortion: false
+  use_gaussian_blur: false
+  use_horizontal_flip: false
+  val_split: null
+logging:
+  enable_sweep: false
+  enable_wandb: false
+  folder: runs/jepa
+  sweep_config: {}
+  wandb_entity: ''
+  wandb_project: simulusion
+  write_tag: jepa
+mask:
+  allow_overlap: false
+  aspect_ratio:
+  - 0.75
+  - 1.5
+  enc_mask_scale:
+  - 0.85
+  - 1.0
+  min_keep: 4
+  num_enc_masks: 1
+  num_pred_masks: 4
+  patch_size: 4
+  pred_mask_scale:
+  - 0.15
+  - 0.2
+meta:
+  copy_data: false
+  load_checkpoint: false
+  loss_type: l2
+  model_name: vit_tiny
+  pred_depth: null
+  pred_emb_dim: 384
+  read_checkpoint: null
+  use_bfloat16: true
+optimization:
+  early_stopping: false
+  ema:
+  - 0.996
+  - 1.0
+  epochs: 30
+  final_lr: 1.0e-06
+  final_weight_decay: 0.4
+  ipe_scale: 1.0
+  lr: 0.001
+  lr_reference_batch_size: null
+  min_delta: 0.0001
+  patience: 10
+  start_lr: 0.0002
+  warmup: 2
+  weight_decay: 0.04

jepa-latest.pth.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cfb7e81011ada5cd7cae813f0ce985786b8e1d21a09309e471eefb2bc78868e
+size 215643004