fix: only response [DONE] once when streaming response. (#378)

b396cb49 · Ricardo Lu · GitHub · 1c395b4e · b396cb49
Unverified Commit b396cb49 authored Jul 07, 2023 by Ricardo Lu Committed by GitHub Jul 06, 2023
Hide whitespace changes
Inline Side-by-side

Showing with 2 additions and 2 deletions

vllm/entrypoints/openai/api_server.py vllm/entrypoints/openai/api_server.py +2 -2

No files found.
--- a/vllm/entrypoints/openai/api_server.py
+++ b/vllm/entrypoints/openai/api_server.py
@@ -269,7 +269,7 @@ async def create_chat_completion(raw_request: Request):
                        finish_reason=output.finish_reason,
                    )
                    yield f"data: {response_json}\n\n"
-            yield "data: [DONE]\n\n"
+        yield "data: [DONE]\n\n"

    # Streaming response
    if request.stream:
@@ -465,7 +465,7 @@ async def create_completion(raw_request: Request):
                        finish_reason=output.finish_reason,
                    )
                    yield f"data: {response_json}\n\n"
-            yield "data: [DONE]\n\n"
+        yield "data: [DONE]\n\n"

    # Streaming response
    if stream: