From a02f2b3e08291ef900e44e3f4ce7cc7f63302279 Mon Sep 17 00:00:00 2001 From: youkaichao Date: Fri, 1 Nov 2024 14:44:24 -0700 Subject: [PATCH] fix Signed-off-by: youkaichao --- vllm/worker/enc_dec_model_runner.py | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/vllm/worker/enc_dec_model_runner.py b/vllm/worker/enc_dec_model_runner.py index 1cd2cdcd6ec0b..4c7593d24edce 100644 --- a/vllm/worker/enc_dec_model_runner.py +++ b/vllm/worker/enc_dec_model_runner.py @@ -28,7 +28,7 @@ ModelInputForGPUWithSamplingMetadata, _get_graph_batch_size) from vllm.worker.model_runner_base import ( - ModelRunnerBase, _add_attn_metadata_broadcastable_dict, + _add_attn_metadata_broadcastable_dict, _add_sampling_metadata_broadcastable_dict) from vllm.worker.utils import assert_enc_dec_mr_supported_scenario @@ -91,8 +91,6 @@ def __init__( models) but these arguments are present here for compatibility with the base-class constructor. ''' - ModelRunnerBase.__init__(self, vllm_config=vllm_config) - self._maybe_force_supported_attention_backend() super().__init__(