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 12, 02:15 UTC
[Rocm][CI] Fix test_speculator_eagle3 by skipping the CompressedTensorw4a16 Model (#30001)
Signed-off-by: charlifu <charlifu@amd.com> Co-authored-by: Alexei-V-Ivanov-AMD <156011006+Alexei-V-Ivanov-AMD@users.noreply.github.com>
Charlie Fu committed
Dec 4, 2025 at 01:52 UTC
9aa33a74b00b2db7d7da22a59ed64b44ebbabe14
1 file changed
+5
tests/v1/spec_decode/test_speculators_eagle3.py
+5
@@ -5,6 +5,7 @@ import torch
5
6
from vllm.config import SpeculativeConfig
7
from vllm.model_executor.models.interfaces import supports_eagle3
8
+from vllm.platforms import current_platform
9
10
11
@pytest.mark.parametrize(
@@ -21,6 +22,10 @@ from vllm.model_executor.models.interfaces import supports_eagle3
22
pytest.param(
23
"nm-testing/Speculator-Qwen3-8B-Eagle3-converted-071-quantized-w4a16",
24
id="qwen3-eagle3-speculator-w4a16-verifier",
25
+ marks=pytest.mark.skipif(
26
+ current_platform.is_rocm(),
27
+ reason="The tests are skipped on rocm platform.",
28
+ ),
29
),
30
],
31
)