Update README.md

README.md

@@ -2,9 +2,94 @@
 tags:
 - model_hub_mixin
 - pytorch_model_hub_mixin
+- vision
+- perceiver
+- adaptive-computation
+- image-classification
+license: mit
+datasets:
+- timm/imagenet-1k-wds
 ---
 
-This model has been pushed to the Hub using the [PytorchModelHubMixin](https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin) integration:
-- Code: https://github.com/pjjajal/adaperceiver-public
-- Paper: https://arxiv.org/abs/2511.18105
-- Docs: [More Information Needed]
+# AdaPerceiver (ImageNet-1K Fine-Tuned)
+
+This repository hosts the **ImageNet-1K fine-tuned AdaPerceiver model**, introduced in  
+**“AdaPerceiver: Transformers with Adaptive Width, Depth, and Tokens”**.
+
+📄 Paper: https://arxiv.org/abs/2511.18105  
+📦 Code: https://github.com/pjajal/AdaPerceiver  
+📚 Model Collection: https://huggingface.co/collections/pjajal/adaperceiver-v1  
+
+This model is fine-tuned from the **logit + feature distilled AdaPerceiver backbone** trained on ImageNet-12K (found [here](https://huggingface.co/pjajal/adaperceiver-v1)).
+
+---
+
+## Model Description
+
+**AdaPerceiver** is a Perceiver-style transformer architecture designed for **runtime-adaptive computation**.  
+A single trained model can dynamically trade off **accuracy and compute** by adjusting:
+
+- the **number of latent tokens**,
+- the **effective depth**, and
+- the **embedding dimension**.
+
+This specific checkpoint corresponds to the **ImageNet-1K classification fine-tuned AdaPerceiver model**, described in Appendix D.2 of the paper.
+
+---
+
+## Training Details
+
+- **Fine-Tuning Data:** ImageNet-1K
+- **Initialization:** Logit + feature distilled AdaPerceiver (ImageNet-12K)
+- **Objective:** Supervised classification fine-tuning
+- **Architecture:** Adaptive Perceiver with block-masked attention and Matryoshka FFNs
+- **Adaptivity Axes:** Tokens, Depth, Width
+
+During fine-tuning, the AdaPerceiver backbone is frozen and only the classification head, output tokens, and output cross-attention layers are updated.
+
+For full training details, see Appendix D of the paper.
+
+---
+
+## How to Use
+
+This model can be loaded using the AdaPerceiver Hub-compatible classification class.
+
+```python
+import torch
+from hub.networks.adaperceiver_classification import ClassificationAdaPerceiver
+
+model = ClassificationAdaPerceiver.from_pretrained(
+    "pjajal/adaperceiver-v1-in1k-ft"
+)
+
+# forward(
+#   x: input image tensor (B, C, H, W)
+#   num_tokens: number of latent tokens to process (optional)
+#   mat_dim: embedding dimension (optional)
+#   depth: early-exit depth (optional)
+#   depth_tau: confidence threshold for early exit (optional)
+#   token_grans: block-mask granularities (optional)
+# )
+out = model(
+    torch.randn(1, 3, 224, 224),
+    num_tokens=256,
+    mat_dim=192,
+    depth=12,
+)
+
+print(out.logits.shape)
+```
+
+## Reference
+
+If you use this models please cite the AdaPerceiver paper:
+
+```bibtex
+@article{jajal2025adaperceiver,
+  title={AdaPerceiver: Transformers with Adaptive Width, Depth, and Tokens},
+  author={Jajal, Purvish and Eliopoulos, Nick John and Chou, Benjamin Shiue-Hal and Thiruvathukal, George K and Lu, Yung-Hsiang and Davis, James C},
+  journal={arXiv preprint arXiv:2511.18105},
+  year={2025}
+}
+```