From e1f96911203bcafb8aed81e04ea751b99b22bcdc Mon Sep 17 00:00:00 2001 From: Canlin Guo Date: Thu, 25 Dec 2025 23:48:34 +0800 Subject: [PATCH] [Clean] Remove the redundant decoding payloads logic (#404) Signed-off-by: gcanlin --- vllm_omni/worker/gpu_model_runner.py | 3 --- 1 file changed, 3 deletions(-) diff --git a/vllm_omni/worker/gpu_model_runner.py b/vllm_omni/worker/gpu_model_runner.py index 9b326a652..4ca0a6a0e 100644 --- a/vllm_omni/worker/gpu_model_runner.py +++ b/vllm_omni/worker/gpu_model_runner.py @@ -739,9 +739,6 @@ class OmniGPUModelRunner(GPUModelRunner): intermediate_tensors: IntermediateTensors | None = None, ): """Align with v0.12 preprocess and omni's additional information handling.""" - # Decode payload first, ensure request state has prompt_embeds / additional_information - self._decode_and_store_request_payloads(scheduler_output) - num_scheduled_tokens = scheduler_output.total_num_scheduled_tokens is_first_rank = get_pp_group().is_first_rank is_encoder_decoder = self.model_config.is_encoder_decoder