Fix 1D query issue from `_prune_hidden_states` (#3539)

vllm/model_executor/layers/logits_processor.py

@@ -77,7 +77,6 @@ def _prune_hidden_states(
     hidden_states: torch.Tensor,
     sampling_metadata: SamplingMetadata,
 ) -> torch.Tensor:
-    hidden_states = hidden_states.view(-1, hidden_states.shape[-1])
     return hidden_states.index_select(0,
                                       sampling_metadata.selected_token_indices)