Experimenting

modeling_hinvec.py

@@ -278,10 +278,6 @@ class HinvecAttention(nn.Module):
 
             attn = (attention_mask < 0)*torch.finfo(query_vectors.dtype).min
             attn = attn[:, None, :, None].expand(-1, 1, attn.size(1), attn.size(1))
-            print(query_vectors.shape)
-            print(key_vectors.shape)
-            print(value_vectors.shape)
-            print("#"*10)
             attn_output, attn_weights = attention_interface(
                 self,
                 query_vectors,