From 92f337faeba7486ffe23015788d6043d8319e142 Mon Sep 17 00:00:00 2001 From: Woosuk Kwon Date: Thu, 18 Sep 2025 12:44:21 -0700 Subject: [PATCH] minor Signed-off-by: Woosuk Kwon --- vllm/v1/worker/gpu/model_runner.py | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/vllm/v1/worker/gpu/model_runner.py b/vllm/v1/worker/gpu/model_runner.py index c1632dc8750ff..f2d0166724fc5 100644 --- a/vllm/v1/worker/gpu/model_runner.py +++ b/vllm/v1/worker/gpu/model_runner.py @@ -219,9 +219,10 @@ class GPUModelRunner: idx_mapping_list = [ self.req_states.req_id_to_index[req_id] for req_id in req_ids ] - self.input_buffers.idx_mapping.np[:num_reqs] = idx_mapping_list - idx_mapping_np = self.input_buffers.idx_mapping.np[:num_reqs] - idx_mapping = self.input_buffers.idx_mapping.copy_to_gpu(num_reqs) + idx_mapping = self.input_buffers.idx_mapping + idx_mapping.np[:num_reqs] = idx_mapping_list + idx_mapping_np = idx_mapping.np[:num_reqs] + idx_mapping = idx_mapping.copy_to_gpu(num_reqs) # Block tables: num_kv_cache_groups x [num_reqs, max_num_blocks] block_tables = self.block_tables.gather_block_tables(idx_mapping)