Add vLLM v0.18.1 source tree with KV transfer abort fix

third_party/vllm/ now tracked in git for direct patch management.
Based on vLLM v0.18.1 release with one patch applied:

  vllm/v1/core/sched/scheduler.py:
    Replace fatal assert with graceful skip when KV transfer callback
    arrives for an already-aborted request during PD disaggregated serving.

Future vLLM modifications should be made directly in third_party/vllm/
and committed normally. The patches/ directory is kept as documentation
of what changed from upstream.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

third_party/vllm/tests/quantization/test_cpu_offload.py

@@ -0,0 +1,73 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+# Expanded quantized model tests for CPU offloading
+# Base tests: tests/basic_correctness/test_cpu_offload.py
+
+import pytest
+
+from tests.quantization.utils import is_quant_method_supported
+
+from ..utils import compare_two_settings
+
+
+@pytest.mark.skipif(
+    not is_quant_method_supported("fp8"),
+    reason="fp8 is not supported on this GPU type.",
+)
+def test_cpu_offload_fp8():
+    # Test loading a quantized checkpoint
+    compare_two_settings(
+        "neuralmagic/Qwen2-1.5B-Instruct-FP8",
+        ["--enforce_eager"],
+        ["--enforce_eager", "--cpu-offload-gb", "1"],
+        max_wait_seconds=480,
+    )
+
+
+@pytest.mark.skipif(
+    not is_quant_method_supported("gptq_marlin"),
+    reason="gptq_marlin is not supported on this GPU type.",
+)
+def test_cpu_offload_gptq(monkeypatch):
+    # This quant method is sensitive to dummy weights, so we force real weights
+    monkeypatch.setenv("VLLM_TEST_FORCE_LOAD_FORMAT", "auto")
+    # Test GPTQ Marlin
+    compare_two_settings(
+        "Qwen/Qwen2-1.5B-Instruct-GPTQ-Int4",
+        ["--enforce_eager"],
+        ["--enforce_eager", "--cpu-offload-gb", "1"],
+        max_wait_seconds=480,
+    )
+
+
+@pytest.mark.skipif(
+    not is_quant_method_supported("awq_marlin"),
+    reason="awq_marlin is not supported on this GPU type.",
+)
+def test_cpu_offload_awq(monkeypatch):
+    # This quant method is sensitive to dummy weights, so we force real weights
+    monkeypatch.setenv("VLLM_TEST_FORCE_LOAD_FORMAT", "auto")
+    # Test AWQ Marlin
+    compare_two_settings(
+        "Qwen/Qwen2-1.5B-Instruct-AWQ",
+        ["--enforce_eager"],
+        ["--enforce_eager", "--cpu-offload-gb", "1"],
+        max_wait_seconds=480,
+    )
+
+
+@pytest.mark.skipif(
+    not is_quant_method_supported("gptq_marlin"),
+    reason="gptq_marlin is not supported on this GPU type.",
+)
+def test_cpu_offload_compressed_tensors(monkeypatch):
+    # This quant method is sensitive to dummy weights, so we force real weights
+    monkeypatch.setenv("VLLM_TEST_FORCE_LOAD_FORMAT", "auto")
+    # Test wNa16
+    compare_two_settings(
+        "nm-testing/Qwen1.5-MoE-A2.7B-Chat-quantized.w4a16",
+        ["--enforce_eager"],
+        ["--enforce_eager", "--cpu-offload-gb", "1"],
+        max_wait_seconds=480,
+    )