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 6, 05:45 UTC
[ROCm][CI] get_cu_count was renamed to num_compute_units in #35042 (#37764)
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Andreas Karatzas committed
Mar 22, 2026 at 03:02 UTC
6ecba840d7c301899f87cda40921239482e65185
1 file changed
+3
-3
tests/model_executor/layers/test_rocm_unquantized_gemm.py
+3
-3
@@ -26,7 +26,7 @@ def test_rocm_unquantized_gemm_gfx1x_wvsplitk_path(monkeypatch):
26
monkeypatch.setattr("vllm.platforms.rocm.on_gfx1x", lambda: True)
27
monkeypatch.setattr("vllm.platforms.rocm.on_gfx9", lambda: False)
28
monkeypatch.setattr("vllm.platforms.rocm.on_gfx950", lambda: False)
29
- monkeypatch.setattr(utils, "get_cu_count", lambda: 120)
29
+ monkeypatch.setattr(utils, "num_compute_units", lambda: 120)
30
31
wvsplitk_mock = MagicMock(side_effect=lambda w, x_view, _, __: x_view @ w.t())
32
monkeypatch.setattr(utils.ops, "wvSplitK", wvsplitk_mock)
@@ -50,7 +50,7 @@ def test_rocm_unquantized_gemm_gfx1x_n_gt_4_falls_back(monkeypatch):
50
monkeypatch.setattr("vllm.platforms.rocm.on_gfx1x", lambda: True)
51
monkeypatch.setattr("vllm.platforms.rocm.on_gfx9", lambda: False)
52
monkeypatch.setattr("vllm.platforms.rocm.on_gfx950", lambda: False)
53
- monkeypatch.setattr(utils, "get_cu_count", lambda: 120)
53
+ monkeypatch.setattr(utils, "num_compute_units", lambda: 120)
54
55
wvsplitk_mock = MagicMock(side_effect=lambda w, x_view, _, __: x_view @ w.t())
56
monkeypatch.setattr(utils.ops, "wvSplitK", wvsplitk_mock)
@@ -74,7 +74,7 @@ def test_rocm_unquantized_gemm_gfx950_wvsplitkrc_path(monkeypatch):
74
monkeypatch.setattr("vllm.platforms.rocm.on_gfx1x", lambda: False)
75
monkeypatch.setattr("vllm.platforms.rocm.on_gfx9", lambda: False)
76
monkeypatch.setattr("vllm.platforms.rocm.on_gfx950", lambda: True)
77
- monkeypatch.setattr(utils, "get_cu_count", lambda: 120)
77
+ monkeypatch.setattr(utils, "num_compute_units", lambda: 120)
78
79
wvsplitkrc_mock = MagicMock(side_effect=lambda w, x_view, _, __: x_view @ w.t())
80
monkeypatch.setattr(utils.ops, "wvSplitKrc", wvsplitkrc_mock)