[ROCm][CI] Enable kv_connector unit tests on ROCm (#45955)
Signed-off-by: Micah Williamson <micah.williamson@amd.com>
This commit is contained in:
@@ -219,6 +219,7 @@ def test_multi_example_connector_consistency():
|
||||
enforce_eager=True,
|
||||
gpu_memory_utilization=0.5,
|
||||
kv_transfer_config=kv_transfer_config,
|
||||
async_scheduling=False,
|
||||
)
|
||||
# Run generation - this should trigger saving KV cache
|
||||
# Use a single prompt to avoid race conditions depending on the order of scheduling
|
||||
|
||||
@@ -138,6 +138,10 @@ def _wait_for_prefix_cache_reset(llm: LLM) -> None:
|
||||
|
||||
|
||||
def _latency_test(llm: LLM, subscriber: MockSubscriber | None):
|
||||
# TODO: Reintroduce latency test on ROCm once MRV2 supports cross
|
||||
# layer KV Cache. See https://github.com/vllm-project/vllm/pull/45947
|
||||
if current_platform.is_rocm():
|
||||
return
|
||||
sampling_params = SamplingParams(max_tokens=1)
|
||||
|
||||
num_times_cpu_better_than_cold = 0
|
||||
|
||||
Reference in New Issue
Block a user