debug

app.py

@@ -54,17 +54,18 @@ def load_video(video_path):
         img = Image.fromarray(vr[frame_index].asnumpy())
         images_group.append(img)
     torch_imgs = transform(images_group)
+    
+    # The model expects inputs of shape: B x C x T x H x W
+    TC, H, W = torch_imgs.shape
+    torch_imgs = torch_imgs.reshape(1, TC//3, 3, H, W).permute(0, 2, 1, 3, 4)
+
     return torch_imgs
     
 
 def inference(video):
     vid = load_video(video)
     
-    # The model expects inputs of shape: B x C x T x H x W
-    TC, H, W = vid.shape
-    inputs = vid.reshape(1, TC//3, 3, H, W).permute(0, 2, 1, 3, 4)
-    
-    prediction = model(inputs)
+    prediction = model(vid)
     prediction = F.softmax(prediction, dim=1).flatten()
 
     return {kinetics_id_to_classname[str(i)]: float(prediction[i]) for i in range(400)}