A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[CI/Build] do not enforce precompilation on tpu ci tests (#25992)
Signed-off-by: Xiang Si <sixiang@google.com>
Xiang Si committed
Oct 3, 2025 at 06:38 UTC
adae0c1f439263906b5da7b4d9a06b8cdc38719e
1 file changed
+1
-4
tests/v1/entrypoints/llm/test_struct_output_generate.py
+1
-4
@@ -123,13 +123,10 @@ def test_structured_output(
123
if current_platform.is_tpu() and speculative_config:
124
pytest.skip("TPU does not support speculative decoding")
125
126
- # Don't use eager execution on TPUs because we want to test for no
127
- # recompilation at runtime
128
- enforce_eager = bool(not current_platform.is_tpu())
126
# Use a single LLM instance for several scenarios to
127
# speed up the test suite.
128
llm = LLM(model=model_name,
132
- enforce_eager=enforce_eager,
129
+ enforce_eager=True,
130
max_model_len=1024,
131
structured_outputs_config=dict(backend=backend,
132
disable_any_whitespace=backend