x

docker/scripts/entrypoint.sh

@@ -19,5 +19,10 @@ if [ -f ${SCRIPTS_DIR}/apply_mineru_patches.py ]; then
     ${VENV}/bin/python ${SCRIPTS_DIR}/apply_mineru_patches.py || echo "[entrypoint] apply_mineru_patches.py failed (non-fatal)"
 fi
 
+# 确保 vllm 源码目录在 PYTHONPATH 中
+# patch_vllm_platform.py 中的 os.environ 修改不持久化到父进程
+# 这里无条件导出，防止 editable install 被破坏后 vllm 无法导入
+export PYTHONPATH="/opt/vllm:${PYTHONPATH}"
+
 echo "[entrypoint] starting: $*"
 exec "$@"

docker/scripts/patch_vllm_platform.py

@@ -6,7 +6,9 @@
 """
 
 import os
+import re
 import sys
+import sysconfig
 import traceback
 
 VLLM_DIR = '/opt/vllm/vllm'
@@ -38,26 +40,88 @@ def patch6_init_platform_fallback():
 def patch7_rocm_break_import_cycle():
     """补丁 7：platforms/rocm.py —— logger.warning_once → sys.stderr.write
 
-    这个补丁会直接改写 vLLM 源码中的函数调用。部分 vLLM 版本里
-    logger.warning_once(...) 参数不一定兼容 sys.stderr.write(...)，改写后可能
-    导致 import vllm / import vllm.platforms.rocm 失败。MinerU 会把这种导入失败
-    包装成 "Please install vllm"，因此默认禁用，只在显式设置环境变量时启用。
+    默认启用安全的“整段 except 块”替换，避免 logger.warning_once() 在平台检测
+    期间触发循环导入，导致 current_platform 落到 UnspecifiedPlatform。
+    如需禁用，设置 VLLM_PATCH_ROCM_WARNING_ONCE=0。
     """
-    if os.environ.get('VLLM_PATCH_ROCM_WARNING_ONCE') != '1':
-        print('Patch 7: skipped (set VLLM_PATCH_ROCM_WARNING_ONCE=1 to enable).')
+    if os.environ.get('VLLM_PATCH_ROCM_WARNING_ONCE') == '0':
+        print('Patch 7: skipped (VLLM_PATCH_ROCM_WARNING_ONCE=0).')
         return
     f = os.path.join(VLLM_DIR, 'platforms', 'rocm.py')
     c = open(f).read()
-    old = 'logger.warning_once('
-    new = 'import sys as _sys\n        _sys.stderr.write('
-    c2 = c.replace(old, new)
+    if 'amdsmi unavailable, using torch.cuda fallback' in c:
+        print('Patch 7: already applied.')
+        return
+    pattern = re.compile(
+        r'(\n\s*except Exception as e:\n'
+        r'\s*logger\.debug\("Failed to get GCN arch via amdsmi: %s", e\)\n'
+        r'\s*logger\.warning_once\(\n'
+        r'(?:\s*"[^"]*"\n)+'
+        r'\s*\)\n)',
+        re.MULTILINE,
+    )
+
+    def repl(match):
+        indent = re.search(r'\n(\s*)except Exception as e:', match.group(1)).group(1)
+        body_indent = indent + '    '
+        return (
+            f'\n{indent}except Exception as e:\n'
+            f'{body_indent}import sys as _sys\n'
+            f'{body_indent}_sys.stderr.write('
+            '"vLLM ROCm: amdsmi unavailable, using torch.cuda fallback for GPU detection\\n")\n'
+        )
+
+    c2, n = pattern.subn(repl, c, count=1)
     if c2 != c:
         open(f, 'w').write(c2)
-        print('Patch 7: rocm.py circular import broken.')
+        print('Patch 7: rocm.py logger.warning_once circular import patch applied.')
     else:
         print('Patch 7: already applied or pattern not found.')
 
 
+def ensure_vllm_dist_info():
+    """为 /opt/vllm 源码导入创建最小 dist-info。
+
+    vLLM 平台检测会通过 importlib.metadata 查询 vllm 分发元数据/entry points。
+    如果只靠 PYTHONPATH 导入源码，没有 dist-info，就会出现：
+    "The vLLM package was not found..."，并可能得到 UnspecifiedPlatform。
+    """
+    try:
+        import vllm
+    except Exception:
+        return
+
+    version = getattr(vllm, '__version__', '0.1.dev1') or '0.1.dev1'
+    site_packages = sysconfig.get_paths().get('purelib')
+    if not site_packages:
+        return
+    dist_info = os.path.join(site_packages, f'vllm-{version}.dist-info')
+    os.makedirs(dist_info, exist_ok=True)
+
+    metadata = os.path.join(dist_info, 'METADATA')
+    if not os.path.exists(metadata):
+        with open(metadata, 'w') as fp:
+            fp.write(
+                'Metadata-Version: 2.1\n'
+                'Name: vllm\n'
+                f'Version: {version}\n'
+                'Summary: vLLM source tree mounted at /opt/vllm\n'
+            )
+
+    with open(os.path.join(dist_info, 'top_level.txt'), 'w') as fp:
+        fp.write('vllm\n')
+    with open(os.path.join(dist_info, 'INSTALLER'), 'w') as fp:
+        fp.write('mineru-rocm-runtime\n')
+    with open(os.path.join(dist_info, 'entry_points.txt'), 'w') as fp:
+        fp.write(
+            '[vllm.platform_plugins]\n'
+            'rocm = vllm.platforms.rocm:rocm_platform_plugin\n'
+        )
+    with open(os.path.join(dist_info, 'RECORD'), 'w') as fp:
+        fp.write('')
+    print(f'vllm dist-info ensured: {dist_info}')
+
+
 def ensure_vllm_installed():
     """确保 vLLM Python 包在当前虚拟环境中可导入。
 
@@ -94,9 +158,16 @@ def main():
     patch7_rocm_break_import_cycle()
     try:
         ensure_vllm_installed()
+        ensure_vllm_dist_info()
         import vllm
         from vllm.platforms import current_platform
         print(f'vllm runtime import OK: {vllm.__version__}, platform={type(current_platform).__name__}')
+        if type(current_platform).__name__ == 'UnspecifiedPlatform':
+            raise RuntimeError(
+                'vLLM platform detection returned UnspecifiedPlatform; '
+                'ROCm backend is not usable. Check amdsmi, /dev/kfd, /dev/dri, '
+                'and vllm.platform_plugins metadata.'
+            )
     except Exception:
         print('ERROR: vllm runtime import failed after platform patches:')
         traceback.print_exc()