[Tencent FIX BUILD FLOW] stabilize Hygon DCU wheel build and install flow

README.md

@@ -65,22 +65,27 @@ pip install setuptools wheel
 ```shell
 git clone http://10.16.6.30/dcutoolkit/deeplearing/vllm.git # 根据需要的分支进行切换
 ```
-安装依赖:
-```shell
-pip install -r requirements/rocm.txt
-```
 - 提供2种源码编译方式(进入vllm目录):
 ```
-1. 编译whl包并安装
-python setup.py bdist_wheel 
+1. 编译whl包并安装（推荐）
+python tools/check_hygon_env.py
+VLLM_USE_HYGON=1 python setup.py bdist_wheel
+
 cd dist
 pip install vllm*
+python tools/check_hygon_env.py
 
 2. 源码编译安装
-python3 setup.py install （若调试，可使用python3 setup.py develop）
+VLLM_USE_HYGON=1 python3 setup.py install （若调试，可使用 VLLM_USE_HYGON=1 python3 setup.py develop）
 ```
 若需要添加git号，设置环境变量: export ADD_GIT_VERSION=1
 
+> 注意:
+> - 海光 DCU 环境必须设置 `VLLM_USE_HYGON=1`。
+> - 构建时会自动将海光定制包（如 `torch`、`triton`、`flash_attn`）改写为当前环境中的精确 `+das.*` 版本，避免 pip 在安装时替换为不兼容的 PyPI 版本。
+> - `hygon.txt` 只对高风险包做最小锁定：`numpy==1.26.4` 以及海光定制包，其余 Python 依赖保持正常解析。
+> - 海光路径会将 `fastapi[standard]`、`mistral_common[image]` 这类容易触发深度回溯的 extras 依赖替换为镜像中已验证的显式依赖版本。
+> - 可使用 `python tools/check_hygon_env.py` 在构建前和安装后校验关键包版本
 
 ### 运行基础环境准备
 1、使用上面基于光源pytorch2.9.0基础镜像环境

requirements/hygon.text

+# Hygon DCU dependencies (based on ROCm/HIP)
+# Use VLLM_USE_HYGON=1 to enable this file during build.
+#
+# Keep the lock set minimal:
+# - pin numpy explicitly to avoid ABI-breaking upgrades
+# - keep Hygon custom packages on base versions here, then rewrite them
+#   to exact +das.* installed versions during bdist_wheel
+-r common.txt
+
+# Pin numpy to 1.26.x: compatible with both custom torch ABI and
+# modern PyPI packages (pandas, opencv, etc.)
+numpy == 1.26.4
+
+# Replace heavy extras dependency chains with the explicit versions
+# already validated in the Hygon Docker image.
+fastapi == 0.135.3
+uvicorn == 0.42.0
+python-multipart == 0.0.22
+watchfiles == 1.1.1
+mistral_common == 1.11.0
+
+numba == 0.61.2 # Required for N-gram speculative decoding
+
+# Dependencies for Hygon DCU
+datasets
+ray[cgraph]>=2.48.0 # Ray Compiled Graph, required for pipeline parallelism in V1.
+peft
+pytest-asyncio
+tensorizer==2.10.1
+packaging>=24.2
+setuptools>=77.0.3,<80.0.0
+setuptools-scm>=8
+runai-model-streamer[s3,gcs]==0.15.3
+timm>=1.0.17
+grpcio-tools>=1.76.0
+
+numa
+quart
+fastrlock==0.8.3
+
+# Hygon DCU hardware-bound packages (base versions)
+torch == 2.9.0
+triton == 3.3.0
+flash_attn == 2.6.1
+flash_mla == 1.2.0
+lightop == 0.6.0
+lmslim == 0.3.1
+

requirements/rocm.txt

@@ -19,14 +19,8 @@ grpcio-tools>=1.76.0
 
 numa
 # pytrie
-cmake==3.29
+#cmake==3.29
 quart
 fastrlock==0.8.3
 # cupy==12.3.0
 
-torch == 2.9.0
-triton == 3.3.0
-flash_attn == 2.8.3
-flash_mla == 1.0.0
-lightop == 0.6.0
-lmslim == 0.3.1

setup.py

@@ -1046,6 +1046,117 @@ def get_vllm_version() -> str:
 
     return version
 
+def _pin_hygon_local_versions(requirements: list[str]) -> list[str]:
+    """Pin packages with +das.* local versions to their exact installed
+    versions, and deduplicate requirements (later entries win)."""
+    import importlib.metadata
+
+    # First pass: pin +das versions
+    pinned = []
+    for req in requirements:
+        match = re.match(r'^([a-zA-Z0-9_-]+)\s*==', req)
+        if match:
+            pkg_name = match.group(1)
+            try:
+                installed_ver = importlib.metadata.version(pkg_name)
+                if "+das" in installed_ver:
+                    pinned.append(f"{pkg_name} == {installed_ver}")
+                    continue
+            except importlib.metadata.PackageNotFoundError:
+                pass
+        pinned.append(req)
+
+    # Second pass: deduplicate by package name (later entries win)
+    seen = {}
+    for req in pinned:
+        match = re.match(r'^([a-zA-Z0-9_-]+)', req)
+        if match:
+            pkg_name = match.group(1).lower().replace("-", "_")
+            seen[pkg_name] = req
+        else:
+            seen[req] = req
+    return list(seen.values())
+
+
+def _validate_hygon_build_env() -> None:
+    """Fail fast with actionable errors for the Hygon build env."""
+    import importlib.metadata
+
+    exact_versions = {
+        "numpy": "1.26.4",
+    }
+    das_versions = {
+        "torch": "2.9.0",
+        "triton": "3.3.0",
+        "flash_attn": "2.6.1",
+        "flash_mla": "1.2.0",
+        "lightop": "0.6.0",
+        "lmslim": "0.3.1",
+    }
+
+    errors: list[str] = []
+
+    for pkg_name, expected_version in exact_versions.items():
+        try:
+            installed_ver = importlib.metadata.version(pkg_name)
+        except importlib.metadata.PackageNotFoundError:
+            errors.append(
+                f"- Missing required package `{pkg_name}`. Expected "
+                f"`{expected_version}`."
+            )
+            continue
+        if installed_ver != expected_version:
+            errors.append(
+                f"- `{pkg_name}` version mismatch: expected "
+                f"`{expected_version}`, got `{installed_ver}`."
+            )
+
+    for pkg_name, expected_base in das_versions.items():
+        try:
+            installed_ver = importlib.metadata.version(pkg_name)
+        except importlib.metadata.PackageNotFoundError:
+            errors.append(
+                f"- Missing required package `{pkg_name}`. Expected a "
+                f"Hygon build based on `{expected_base}`."
+            )
+            continue
+        if not installed_ver.startswith(expected_base):
+            errors.append(
+                f"- `{pkg_name}` version mismatch: expected base "
+                f"`{expected_base}`, got `{installed_ver}`."
+            )
+        if "+das" not in installed_ver:
+            errors.append(
+                f"- `{pkg_name}` is not a Hygon custom build: "
+                f"`{installed_ver}`."
+            )
+
+    if errors:
+        joined = "\n".join(errors)
+        raise RuntimeError(
+            "Hygon build environment validation failed.\n"
+            "Please fix the following packages before running "
+            "`VLLM_USE_HYGON=1 python setup.py bdist_wheel`:\n"
+            f"{joined}"
+        )
+
+
+def _rewrite_hygon_requirements(requirements: list[str]) -> list[str]:
+    """Replace heavy extras chains with explicit Hygon-validated packages."""
+    replaced_packages = {
+        "fastapi",
+        "mistral_common",
+        "opencv_python_headless",
+    }
+    rewritten: list[str] = []
+    for req in requirements:
+        match = re.match(r'^([a-zA-Z0-9_-]+)', req)
+        if match:
+            pkg_name = match.group(1).lower().replace("-", "_")
+            if pkg_name in replaced_packages:
+                continue
+        rewritten.append(req)
+    return rewritten
 
 def get_requirements() -> list[str]:
     """Get Python package dependencies from requirements.txt."""
@@ -1080,6 +1191,12 @@ def get_requirements() -> list[str]:
             modified_requirements.append(req)
         requirements = modified_requirements
     elif _is_hip():
+        if os.environ.get("VLLM_USE_HYGON", "").strip().lower() in ("1", "true"):
+            _validate_hygon_build_env()
+            requirements = _read_requirements("hygon.txt")
+            requirements = _rewrite_hygon_requirements(requirements)
+            requirements = _pin_hygon_local_versions(requirements)
+        else:
             requirements = _read_requirements("rocm.txt")
     elif _is_tpu():
         requirements = _read_requirements("tpu.txt")
@@ -1206,3 +1323,8 @@ setup(
     cmdclass=cmdclass,
     package_data=package_data,
 )
+if "bdist_wheel" in sys.argv:
+    egg_info_dir = ROOT_DIR / "vllm.egg-info"
+    if egg_info_dir.exists():
+        shutil.rmtree(egg_info_dir)
+        logger.info("Cleaned up %s", egg_info_dir)

tools/check_hygon_env.py

+#!/usr/bin/env python3
+
+"""Validate the Hygon DCU Python environment."""
+
+from __future__ import annotations
+
+import sys
+from importlib import metadata
+
+
+EXACT_VERSIONS = {
+    "numpy": "1.26.4",
+}
+
+DAS_VERSIONS = {
+    "torch": "2.9.0",
+    "triton": "3.3.0",
+    "flash_attn": "2.6.1",
+    "flash_mla": "1.2.0",
+    "lightop": "0.6.0",
+    "lmslim": "0.3.1",
+}
+
+
+def main() -> int:
+    errors: list[str] = []
+
+    print("Checking Hygon DCU Python environment...")
+
+    for pkg_name, expected_version in EXACT_VERSIONS.items():
+        try:
+            installed_ver = metadata.version(pkg_name)
+            print(f"  {pkg_name}: {installed_ver}")
+        except metadata.PackageNotFoundError:
+            errors.append(f"- Missing `{pkg_name}` (expected `{expected_version}`).")
+            continue
+        if installed_ver != expected_version:
+            errors.append(
+                f"- `{pkg_name}` mismatch: expected `{expected_version}`, "
+                f"got `{installed_ver}`."
+            )
+
+    for pkg_name, expected_base in DAS_VERSIONS.items():
+        try:
+            installed_ver = metadata.version(pkg_name)
+            print(f"  {pkg_name}: {installed_ver}")
+        except metadata.PackageNotFoundError:
+            errors.append(
+                f"- Missing `{pkg_name}` (expected Hygon build based on "
+                f"`{expected_base}`)."
+            )
+            continue
+        if not installed_ver.startswith(expected_base):
+            errors.append(
+                f"- `{pkg_name}` mismatch: expected base `{expected_base}`, "
+                f"got `{installed_ver}`."
+            )
+        if "+das" not in installed_ver:
+            errors.append(
+                f"- `{pkg_name}` is not a Hygon custom build: "
+                f"`{installed_ver}`."
+            )
+
+    if errors:
+        print("\nEnvironment check failed:")
+        for error in errors:
+            print(error)
+        return 1
+
+    print("\nEnvironment check passed.")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())
+