[doc] add missing imports (#15699)

Signed-off-by: reidliu41 <reid201711@gmail.com> Co-authored-by: reidliu41 <reid201711@gmail.com>
2025-03-28 23:56:48 +08:00
parent 7329ff5468
commit 2914006fe0
5 changed files with 30 additions and 0 deletions
--- a/docs/source/performance/optimization.md
+++ b/docs/source/performance/optimization.md
@@ -31,6 +31,8 @@ vLLM supports an experimental feature chunked prefill. Chunked prefill allows to
 You can enable the feature by specifying `--enable-chunked-prefill` in the command line or setting `enable_chunked_prefill=True` in the LLM constructor.

 ```python
+from vllm import LLM
+
 llm = LLM(model="meta-llama/Llama-2-7b-hf", enable_chunked_prefill=True)
 # Set max_num_batched_tokens to tune performance.
 # NOTE: 2048 is the default max_num_batched_tokens for chunked prefill.