[CI][ROCm] Re-enable the DSv4.1 decoder replay graph test on ROCm (#60392)

Signed-off-by: Stefan Koncarevic <Stefan.Koncarevic@amd.com>
This commit is contained in:
stefankoncarevic
2026-10-07 09:57:52 -05:00
committed by GitHub
parent 14e3902d51
commit 71f9a74920
@@ -21,7 +21,6 @@ from vllm.models.deepseek_v41.decoder_replay_layers import (
from vllm.models.deepseek_v41.nvidia.decoder_replay_cudagraph import (
DecoderReplayCudaGraphManager,
)
from vllm.platforms import current_platform
from vllm.v1.worker.gpu import cudagraph_utils
pytestmark = pytest.mark.skipif(not torch.cuda.is_available(), reason="needs CUDA")
@@ -126,9 +125,6 @@ def _graph_context(num_tokens):
)
@pytest.mark.skipif(
current_platform.is_rocm(), reason="decoder replay CUDA graphs are NVIDIA-only"
)
def test_replay_graph_matches_eager(monkeypatch):
"""The replay graph of the next captured size pads and runs the rows."""
monkeypatch.setattr(cudagraph_utils, "get_pp_group", MagicMock)
@@ -136,6 +132,8 @@ def test_replay_graph_matches_eager(monkeypatch):
monkeypatch.setattr(cudagraph_utils, "graph_capture", lambda device: capture)
compilation = CompilationConfig(decoder_replay_cudagraph_capture_sizes=[4])
cfg = MagicMock(compilation_config=compilation, speculative_config=None)
# Mocking it true would capture into the cuMem pool, which breaks on ROCm.
cfg.use_cumem_cudagraph_pool = False
cfg.scheduler_config = MagicMock(max_num_seqs=2, max_num_batched_tokens=64)
cfg.cache_config.use_kda_recoverssm = False
cfg.model_config.hf_config = MagicMock(hc_mult=HC, hidden_size=HIDDEN)