Update app.py

app.py

@@ -108,7 +108,20 @@ def inference_video(video):
         prediction = F.softmax(prediction, dim=1).flatten()
 
     return {kinetics_id_to_classname[str(i)]: float(prediction[i]) for i in range(400)}
+@spaces.GPU
+def ultra_inference_video(vid):
+    os.system('nvidia-smi')
+    # vid = load_video(video)
+
+    # The model expects inputs of shape: B x C x H x W
+    TC, H, W = vid.shape
+    inputs = vid.reshape(1, TC//3, 3, H, W).permute(0, 2, 1, 3, 4)
 
+    with torch.no_grad():
+        prediction = model_video(inputs.to(device))
+        prediction = F.softmax(prediction, dim=1).flatten()
+
+    return {kinetics_id_to_classname[str(i)]: float(prediction[i]) for i in range(400)}
 
 @spaces.GPU
 def inference_image(img):