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/fp_quant.py

@@ -0,0 +1,32 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""Test model set-up and inference for quantized HF models supported
+on the GPU backend using FPQuant.
+
+Validating the configuration and printing results for manual checking.
+
+Run `pytest tests/quantization/test_fp_quant.py`.
+"""
+
+import pytest
+
+from tests.quantization.utils import is_quant_method_supported
+
+MODELS = [
+    "ISTA-DASLab/Qwen3-0.6B-RTN-NVFP4",
+    "ISTA-DASLab/Qwen3-0.6B-RTN-MXFP4",
+]
+DTYPE = ["bfloat16"]
+EAGER = [True, False]
+
+
+@pytest.mark.skipif(
+    not is_quant_method_supported("fp_quant"),
+    reason="FPQuant is not supported on this GPU type.",
+)
+@pytest.mark.parametrize("model", MODELS)
+@pytest.mark.parametrize("eager", EAGER)
+def test_fpquant(vllm_runner, model, eager):
+    with vllm_runner(model, enforce_eager=eager) as llm:
+        output = llm.generate_greedy(["1 2 3 4 5"], max_tokens=2)
+    assert output[0][1] == "1 2 3 4 5 6"