chore: import upstream snapshot with attribution

2026-07-13 12:55:37 +08:00
commit 7ce4c8e27e
5900 changed files with 1668062 additions and 0 deletions
@@ -0,0 +1,29 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import pytest
+import torch
+
+from tests.kernels.utils import opcheck
+from vllm import _custom_ops as ops  # noqa: F401
+
+
+@pytest.mark.skipif(
+    not hasattr(torch.ops._C, "awq_dequantize"),
+    reason="AWQ is not supported on this GPU type.",
+)
+def test_awq_dequantize_opcheck(monkeypatch: pytest.MonkeyPatch):
+    with monkeypatch.context() as m:
+        m.setenv("VLLM_USE_TRITON_AWQ", "0")
+        qweight = torch.randint(
+            -2000000000, 2000000000, (8192, 256), device="cuda", dtype=torch.int32
+        )
+        scales = torch.rand((64, 2048), device="cuda", dtype=torch.float16)
+        zeros = torch.empty((64, 256), device="cuda", dtype=torch.int32)
+        split_k_iters = 0
+        thx = 0
+        thy = 0
+        opcheck(
+            torch.ops._C.awq_dequantize,
+            (qweight, scales, zeros, split_k_iters, thx, thy),
+        )