Skip to content

test: Add trtllm kv router sanity tests #1988

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Jul 17, 2025
Merged

test: Add trtllm kv router sanity tests #1988

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d2d4dd7
Add kv router examples, increase startup delay for more consistency
rmccorm4 Jul 17, 2025
34bc961
Add note for future improvement of kv router tests
rmccorm4 Jul 17, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
33 changes: 31 additions & 2 deletions tests/serve/test_trtllm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -192,7 +192,7 @@ def wait_for_ready(self, payload, logger=logging.getLogger()):
completions_response_handler,
],
model="deepseek-ai/DeepSeek-R1-Distill-Llama-8B",
delayed_start=45,
delayed_start=60,
),
"disaggregated": TRTLLMConfig(
name="disaggregated",
Expand All @@ -205,7 +205,36 @@ def wait_for_ready(self, payload, logger=logging.getLogger()):
completions_response_handler,
],
model="deepseek-ai/DeepSeek-R1-Distill-Llama-8B",
delayed_start=45,
delayed_start=60,
),
# TODO: These are sanity tests that the kv router examples launch
# and inference without error, but do not do detailed checks on the
# behavior of KV routing.
"aggregated_router": TRTLLMConfig(
name="aggregated_router",
directory="/workspace/components/backends/trtllm",
script_name="agg_router.sh",
marks=[pytest.mark.gpu_1, pytest.mark.tensorrtllm],
endpoints=["v1/chat/completions", "v1/completions"],
response_handlers=[
chat_completions_response_handler,
completions_response_handler,
],
model="deepseek-ai/DeepSeek-R1-Distill-Llama-8B",
delayed_start=60,
),
"disaggregated_router": TRTLLMConfig(
name="disaggregated_router",
directory="/workspace/components/backends/trtllm",
script_name="disagg_router.sh",
marks=[pytest.mark.gpu_2, pytest.mark.tensorrtllm],
endpoints=["v1/chat/completions", "v1/completions"],
response_handlers=[
chat_completions_response_handler,
completions_response_handler,
],
model="deepseek-ai/DeepSeek-R1-Distill-Llama-8B",
delayed_start=60,
),
}

Expand Down
Loading