Update DenseAV/denseav/aggregators.py

DenseAV/denseav/aggregators.py

@@ -161,6 +161,7 @@ class BaseAggregator(torch.nn.Module):
                 audio_mask.to(torch.float32))
 
             if self.use_cls:
+                print(preds[AUDIO_CLS].shape)
                 audio_cls = preds[AUDIO_CLS].reshape(v, self.num_heads, new_c)
                 image_cls = preds[IMAGE_CLS].reshape(v, self.num_heads, new_c)
                 cls_sims = torch.einsum(