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 11, 23:45 UTC
[CI/Build][AMD] Use float16 in test_reset_prefix_cache_e2e to avoid accuracy issues (#29997)
Signed-off-by: Randall Smith <ransmith@amd.com> Co-authored-by: Randall Smith <ransmith@amd.com>
rasmith committed
Dec 5, 2025 at 02:42 UTC
feecba09afcd3c9a37c6b03b868dedf4cb851eb1
1 file changed
+1
tests/v1/core/test_reset_prefix_cache_e2e.py
+1
@@ -21,6 +21,7 @@ def test_reset_prefix_cache_e2e(monkeypatch):
21
max_num_batched_tokens=32,
22
max_model_len=2048,
23
compilation_config={"mode": 0},
24
+ dtype="float16",
25
)
26
engine = LLMEngine.from_engine_args(engine_args)
27
sampling_params = SamplingParams(