From 3e49479c4b766a601804f0c6f5f1c9a3def5ad0c Mon Sep 17 00:00:00 2001 From: Ekagra Ranjan <3116519+ekagra-ranjan@users.noreply.github.com> Date: Fri, 1 May 2026 23:19:07 -0400 Subject: [PATCH] Limit concurrency on `test_transcription_api_correctness.py` (#41478) Signed-off-by: Ekagra Ranjan <3116519+ekagra-ranjan@users.noreply.github.com> --- .../openai/correctness/test_transcription_api_correctness.py | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/tests/entrypoints/openai/correctness/test_transcription_api_correctness.py b/tests/entrypoints/openai/correctness/test_transcription_api_correctness.py index f17f6f5f90..a3df30fb02 100644 --- a/tests/entrypoints/openai/correctness/test_transcription_api_correctness.py +++ b/tests/entrypoints/openai/correctness/test_transcription_api_correctness.py @@ -26,6 +26,9 @@ from vllm.tokenizers import get_tokenizer from ....models.registry import HF_EXAMPLE_MODELS from ....utils import RemoteOpenAIServer +# Tuned to prevent OOM on 18GB GPUs in transcription correctness tests. +MAX_SEQS_FOR_TRANSCRIPTION_TEST = 32 + def to_bytes(y, sr): buffer = io.BytesIO() @@ -184,6 +187,7 @@ def test_wer_correctness( server_args = [ "--enforce-eager", f"--tokenizer_mode={model_info.tokenizer_mode}", + f"--max_num_seqs={MAX_SEQS_FOR_TRANSCRIPTION_TEST}", ] if model_info.trust_remote_code: server_args.append("--trust-remote-code")