From c74b9bb546c0faa45b56042a89807cb16ebfc09a Mon Sep 17 00:00:00 2001 From: Sourashis Roy Date: Fri, 20 Dec 2024 05:26:39 +0000 Subject: [PATCH] Rerun tests Signed-off-by: Sourashis Roy --- vllm/v1/worker/gpu_input_batch.py | 1 + 1 file changed, 1 insertion(+) diff --git a/vllm/v1/worker/gpu_input_batch.py b/vllm/v1/worker/gpu_input_batch.py index dcbb7b7825aa5..fcb1df1d27928 100644 --- a/vllm/v1/worker/gpu_input_batch.py +++ b/vllm/v1/worker/gpu_input_batch.py @@ -395,6 +395,7 @@ def _construct_prompt_tokens_tensor( # TODO - Add a method in vllm/utils.py to pad a numpy array similar # to make_tensor_with_pad which takes a list and move the logic # there. + padded_prompts = np.full((self.num_reqs, max_prompt_len), vocab_size, dtype=np.int64)