From efda08481baa42bbdcdab4644f3666518cc55b2a Mon Sep 17 00:00:00 2001 From: Woosuk Kwon Date: Thu, 18 Sep 2025 16:31:01 -0700 Subject: [PATCH] minor Signed-off-by: Woosuk Kwon --- vllm/v1/worker/gpu/sampler.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/vllm/v1/worker/gpu/sampler.py b/vllm/v1/worker/gpu/sampler.py index cee666df22..d9f74b6f09 100644 --- a/vllm/v1/worker/gpu/sampler.py +++ b/vllm/v1/worker/gpu/sampler.py @@ -254,8 +254,8 @@ def compute_logprobs( ) # NOTE(woosuk): Here, to save GPU memory, we do not materialize the full - # logprobs tensor. Instead, we only compute the logprobs of the topk + 1 - # tokens. + # logprobs tensor. Instead, we only compute and return the logprobs of + # the topk + 1 tokens. BLOCK_SIZE = 1024 _topk_logprobs_kernel[(batch_size, )]( logprobs,