kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-01-13 22:18:36 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

Fix tests to expect RequestError for max_num_tokens validation in PyTorch backend

Browse Source 
Signed-off-by: tzulingk@nvidia.com <tzulingk@nvidia.com>
This commit is contained in:
tzulingk@nvidia.com 2025-12-14 18:02:11 -08:00 committed by Pengyun Lin
parent 92411c4348
commit febc86ced0
1 changed files with 5 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
tests/unittest/llmapi/test_llm.py
View File

@ -2393,7 +2393,8 @@ def test_llm_chunked_prefill():
enable_chunked_prefill=False,
fast_build=True)
with pytest.raises(ValueError):
# max_num_tokens validation now raises RequestError consistently
with pytest.raises(RequestError):
output = llm.generate_async(
"A " * build_config.max_num_tokens,
sampling_params=sampling_params,
@ -2436,13 +2437,9 @@ def _test_llm_capture_request_error(pytorch_backend: bool, tp_size: int = 1):
)
prompt = 'A ' * 65 # the minimum max_num_tokens is 64
if pytorch_backend:
# pytorch backend will raise ValueError for max_num_tokens
with pytest.raises(ValueError):
llm.generate(prompt)
else:
with pytest.raises(RequestError):
llm.generate(prompt)
# Both backends now consistently raise RequestError for max_num_tokens validation
with pytest.raises(RequestError):
llm.generate(prompt)
def test_llm_capture_request_error():