From 42f99150c1ad98318e8af142fb31fb2316387bcf Mon Sep 17 00:00:00 2001 From: Woosuk Kwon Date: Tue, 23 Sep 2025 09:23:21 -0700 Subject: [PATCH] fix Signed-off-by: Woosuk Kwon --- vllm/v1/worker/gpu/model_runner.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/vllm/v1/worker/gpu/model_runner.py b/vllm/v1/worker/gpu/model_runner.py index 1399fd2e88c2b..70485676bfce9 100644 --- a/vllm/v1/worker/gpu/model_runner.py +++ b/vllm/v1/worker/gpu/model_runner.py @@ -293,7 +293,7 @@ class GPUModelRunner: query_start_loc_gpu = query_start_loc.gpu[:num_reqs + 1] query_start_loc_cpu = query_start_loc.cpu[:num_reqs + 1] seq_lens_gpu = self.input_buffers.seq_lens.gpu[:num_reqs] - seq_lens_cpu = self.input_buffers.seq_lens.np[:num_reqs] + seq_lens_cpu = self.input_buffers.seq_lens.cpu[:num_reqs] seq_lens_np = self.input_buffers.seq_lens.np[:num_reqs] # Some input token ids are directly read from the last sampled tokens.