diff --git a/vllm/config.py b/vllm/config.py index 60a32fb182dbb..e8475efcb56e1 100644 --- a/vllm/config.py +++ b/vllm/config.py @@ -701,7 +701,7 @@ def get_num_layers_by_block_type( if self.is_attention_free and attn_block_type: return 0 - if is_full_attn_model and attn_block_type: + if is_full_attn_model and not attn_block_type: return 0 start, end = self.get_layers_start_end_indices(parallel_config)