"examples/vscode:/vscode.git/clone" did not exist on "44a76f96b32afbccb7ddd41fa1a0cce2a21aa8a5"
Unverified Commit 0fa3d079 authored by Richard Huo's avatar Richard Huo Committed by GitHub
Browse files

test: [DYN-1988] skip TRTLLM KVBM disagg test for now until upstream's fix is released (#5974)

parent 8c8680b1
...@@ -336,7 +336,7 @@ class DeterminismTester(ApiTester): ...@@ -336,7 +336,7 @@ class DeterminismTester(ApiTester):
top_k = -1 top_k = -1
if check_module_available("tensorrt_llm"): if check_module_available("tensorrt_llm"):
top_k = 0 top_k = 1 # TensorRT-LLM requires top_k>=0 and dynamo frontend does not support top_k=0
# For determinism: use temperature=0 which should trigger greedy decoding in vLLM # For determinism: use temperature=0 which should trigger greedy decoding in vLLM
# Setting top_p=1.0 and top_k=-1 to avoid any sampling/filtering # Setting top_p=1.0 and top_k=-1 to avoid any sampling/filtering
return super().make_request( return super().make_request(
......
...@@ -547,6 +547,10 @@ def tester(llm_server): ...@@ -547,6 +547,10 @@ def tester(llm_server):
class TestDeterminismDisagg(BaseTestDeterminism): class TestDeterminismDisagg(BaseTestDeterminism):
"""Test class for determinism validation.""" """Test class for determinism validation."""
@pytest.mark.skipif(
check_module_available("tensorrt_llm"),
reason="Skipping test until the TRT-LLM disagg hang issue is fixed. (https://github.com/NVIDIA/TensorRT-LLM/pull/11247)",
)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"llm_server", "llm_server",
[ [
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment