Update README.md

README.md

@@ -7,3 +7,38 @@ license: mit
 ![Example training predictions](example.png)
 
 A Vision Transformer (ViT) model trained on BOLD activation maps from [NSD-Flat](https://huggingface.co/datasets/clane9/NSD-Flat). The training objective was to auto-regressively predict the next patch with shuffled patch order.
+
+## Dependencies
+
+- [boldGPT](https://github.com/clane9/boldGPT)
+- [huggingface_hub](https://huggingface.co/docs/huggingface_hub/index)
+- [safetensors](https://huggingface.co/docs/safetensors/index)
+
+## Usage
+
+```python
+from boldgpt.data import ActivityTransform
+from boldgpt.models import create_model
+from datasets import load_dataset
+from huggingface_hub import hf_hub_download
+from safetensors.torch import load_model
+
+model = create_model("boldgpt_small_patch10")
+
+load_model(
+    model,
+    hf_hub_download(
+        repo_id="clane9/boldgpt_small_patch10", filename="model.safetensors"
+    ),
+)
+
+dataset = load_dataset("clane9/NSD-Flat", split="train")
+dataset.set_format("torch")
+batch = dataset[:1]
+
+transform = ActivityTransform()
+batch["activity"] = transform(batch["activity"])
+
+# output: (B, N, K) predicted next token logits
+output, state = model(batch)
+```