Upload app.py with huggingface_hub

app.py

@@ -0,0 +1,63 @@
+import gradio as gr
+import spaces
+import torch
+import json
+
+model = None
+
+@spaces.GPU
+def predict_brain_response(video_path=None, audio_path=None, text_input=None):
+    """Predict fMRI brain responses to video, audio, or text stimuli."""
+    global model
+
+    if model is None:
+        from transformers import AutoModel
+        model = AutoModel.from_pretrained("facebook/tribev2", trust_remote_code=True)
+        model = model.cuda()
+        model.eval()
+
+    results = {}
+
+    if text_input:
+        with torch.no_grad():
+            output = model.predict_text(text_input)
+            results["text_response"] = {
+                "num_brain_regions": int(output.shape[-1]) if hasattr(output, "shape") else "N/A",
+                "mean_activation": float(output.mean()) if hasattr(output, "mean") else "N/A",
+                "max_activation": float(output.max()) if hasattr(output, "max") else "N/A",
+                "summary": "Brain response predicted across cortical regions"
+            }
+
+    if video_path:
+        with torch.no_grad():
+            output = model.predict_video(video_path)
+            results["video_response"] = {
+                "num_brain_regions": int(output.shape[-1]) if hasattr(output, "shape") else "N/A",
+                "mean_activation": float(output.mean()) if hasattr(output, "mean") else "N/A",
+                "summary": "Visual + temporal cortex response predicted"
+            }
+
+    if audio_path:
+        with torch.no_grad():
+            output = model.predict_audio(audio_path)
+            results["audio_response"] = {
+                "num_brain_regions": int(output.shape[-1]) if hasattr(output, "shape") else "N/A",
+                "mean_activation": float(output.mean()) if hasattr(output, "mean") else "N/A",
+                "summary": "Auditory cortex response predicted"
+            }
+
+    return results
+
+demo = gr.Interface(
+    fn=predict_brain_response,
+    inputs=[
+        gr.File(label="Video (optional)", type="filepath"),
+        gr.File(label="Audio (optional)", type="filepath"),
+        gr.Textbox(label="Text (optional)", placeholder="Enter text stimulus..."),
+    ],
+    outputs=gr.JSON(label="Brain Response Prediction"),
+    title="TRIBE V2 - Brain Response Prediction (Meta)",
+    description="Predicts fMRI brain responses to video, audio, and text stimuli using Meta TRIBE V2 model.",
+)
+
+demo.queue().launch()