Fix the bug in constructing the all_hidden_states of DeBERTa v2 (#10466)

* fix all_hidden_states

* use output_states instead of next_kv

src/transformers/models/deberta_v2/modeling_deberta_v2.py

@@ -450,10 +450,11 @@ class DebertaV2Encoder(nn.Module):
         else:
             next_kv = hidden_states
         rel_embeddings = self.get_rel_embedding()
+        output_states = next_kv
         for i, layer_module in enumerate(self.layer):
 
             if output_hidden_states:
-                all_hidden_states = all_hidden_states + (hidden_states,)
+                all_hidden_states = all_hidden_states + (output_states,)
 
             output_states = layer_module(
                 next_kv,