[CI] Soft fail if translate raise error (vllm-project#11625)

Do not always set to failure when running translate job. Just return the
right one instead.


- vLLM version: v0.23.0
- vLLM main:
vllm-project/vllm@1f486d9

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>

Author: wangxiyuan

.github/workflows/schedule_doc_translate.yaml

@@ -110,6 +110,7 @@ jobs:
       - name: Validate translation coverage
         if: steps.translate.outcome == 'success'
         id: validate
+        continue-on-error: true
         run: |
           python .github/workflows/scripts/po_translate.py \
             --files "${{ steps.detect.outputs.files }}" \

.github/workflows/scripts/detect_po_changes.py

@@ -48,12 +48,11 @@
 DETAILS_RE = __import__("re").compile(r'^\?\?\?(\s+"[^"]*")?\s*$')
 TAB_RE = __import__("re").compile(r'^===\s+"[^"]*"\s*$')
 
-# Characters that indicate a line is purely structural (not translatable):
-#   - Markdown headings and list markers
-#   - Table separators
-#   - HTML tags
-#   - Blank lines (handled separately)
-NO_CN_CHARS = set(".,:;!?-_*+~|/<>[](){}#@$%^&= \t\r\n0123456789")
+# Table rows that are purely structural or data-only (no natural language
+# prose).  Matches rows that look like markdown table rows where every
+# cell is either a number, a date, a link, a single word, a checkmark,
+# or a short identifier.
+TABLE_ROW_RE = __import__("re").compile(r"^\|.*\|$")
 
 
 def _is_translatable_paragraph(paragraph: str) -> bool:
@@ -71,6 +70,13 @@ def _is_translatable_paragraph(paragraph: str) -> bool:
     if ADMONITION_RE.match(text) and '"' not in text:
         return False
 
+    # Skip pure table data rows (contributor tables, feature matrices,
+    # supported models, etc.) that contain structured data but no
+    # natural language sentences.  Only skip rows that are clearly
+    # data-only: every cell is a token, number, date, link, or emoji.
+    if TABLE_ROW_RE.match(text) and _is_table_data_row(text):
+        return False
+
     # Count characters that typically appear in natural language.
     alpha = sum(1 for c in text if c.isalpha())
     if alpha == 0:
@@ -92,6 +98,40 @@ def _is_translatable_paragraph(paragraph: str) -> bool:
     return not (unique <= {"-", "*", "=", "_", "~", "|", ":", "+"})
 
 
+def _is_table_data_row(row: str) -> bool:
+    """Return True if *row* is a markdown table row with structured data only.
+
+    A row is considered data-only when it has at least one cell that
+    looks like structured data (number, date, URL, checkmark, etc.).
+    Pure-text header rows like "| Feature | Support | Note |" are
+    left for translation.
+    """
+    cells = [c.strip() for c in row.split("|")[1:-1]]
+    if not cells:
+        return False
+
+    has_data_cell = False
+    for cell in cells:
+        cell = cell.strip()
+        if not cell:
+            continue
+        if cell in {"✅", "❌", "❔", "✔", "✘", "🟠", "🔵", "—"}:
+            has_data_cell = True
+            continue
+        if __import__("re").match(r"^[\d\s\.\/\-,:]+$", cell):
+            has_data_cell = True
+            continue
+        if __import__("re").match(r"^\[.*\]\(.*\)$", cell):
+            has_data_cell = True
+            continue
+        # Pure-text cell with a sentence (period/question/exclamation)
+        # means this is a prose row, not a data row.
+        if __import__("re").search(r"[.!?]", cell):
+            return False
+
+    return has_data_cell
+
+
 def _extract_paragraphs(content: str) -> list[str]:
     """Extract translatable paragraphs from a markdown source.
 

.github/workflows/scripts/po_translate.py

@@ -50,11 +50,12 @@
 --- OUTPUT FORMAT ---
 1. Return ONLY the same list of msgid/msgstr pairs with msgstr filled in.
    No markdown code fences (```), no explanations, no summaries, no greetings.
-2. Every msgid from the input MUST appear exactly once in the output with its
-   corresponding msgstr filled in. Do not drop, merge, split, or reorder entries.
-3. Keep msgid lines COMPLETELY UNCHANGED — never modify source text.
+2. The number of entries in your output MUST EQUAL the number in the input.
+   Count them: if you received N entries, you must return exactly N entries.
+   Do NOT drop, merge, split, reorder, or add entries under any circumstance.
+3. Keep msgid lines COMPLETELY UNCHANGED — copy them exactly as-is from input.
 
---- WHAT TO PRESERVE ---
+--- WHAT TO PRESERVE (keep EXACTLY as in msgid) ---
 4. All format specifiers: %s, %d, %f, {{}}, {{{{}}}}, {{name}}, etc.
 5. All markdown syntax: **bold**, *italic*, `inline code`, ```code blocks```,
    [links](urls), ![images](urls), # headings, - lists, 1. ordered lists,
@@ -64,41 +65,49 @@
 8. Proper nouns: person names, contributor names, author names, company names,
    product names (vLLM, Ascend, CANN, Huawei, etc.).
 
+--- CONTENT THAT SHOULD NOT BE TRANSLATED ---
+9. DO NOT translate contributor names, GitHub usernames, or dates.
+   These should be copied verbatim from msgid to msgstr.
+   Example: "Xiyuan Wang [@wangxiyuan] 2025-01-15" → keep as-is.
+
+10. DO NOT translate table headers or table separator rows that are purely
+    structural. Copy them verbatim.
+    Example: "| Name | GitHub ID | Date |" → keep as-is.
+    Example: "|:-----------:|:-----:|:-----:|" → keep as-is.
+
+11. DO NOT translate code identifiers, variable names, CLI flags, or shell
+    commands. Copy them verbatim.
+    Example: "--data-parallel-size" → keep as-is.
+    Example: "vllm serve /path/to/model" → keep as-is.
+
+12. DO NOT translate URLs, email addresses, or paths.
+    Copy them verbatim from msgid to msgstr.
+
 --- MkDocs MATERIAL EXTENSIONS ---
-These are special MkDocs syntax elements. Keep the KEYWORDS and STRUCTURE
-exactly as-is; only translate the human-readable TEXT parts.
-
-9. ADMONITIONS: Lines starting with "!!! type" or "!!! type \"title\"".
-   The type keyword (note, warning, tip, danger, etc.) and the "!!!" marker
-   MUST stay in English.
-   Examples:
-     msgid "!!! note"               → msgstr "!!! note"  (no translatable text)
-     msgid "!!! warning"            → msgstr "!!! warning"
-     msgid "!!! note \"Important\""  → msgstr "!!! note \"重要\""
-
-10. COLLAPSIBLE BLOCKS: Lines starting with "??? \"title\"".
-    Keep "???" and the quote syntax; translate only the title text inside quotes.
-    Example:
-      msgid "??? \"Click here to see 'Build from Dockerfile'\""  → msgstr "??? \"点击这里查看'从Dockerfile构建'\""
-
-11. CONTENT TABS: Lines starting with "=== \"label\"".
-    Keep "===" and the quote syntax; translate only the label text.
-    Example:
-      msgid "=== \"Before using pip\""  → msgstr "=== \"使用pip之前\""
+13. ADMONITIONS (!!! type): Keep "!!!" and type keyword (note, warning, tip)
+    in English. Only translate the title text after type.
+    Example: msgid "!!! note" → msgstr "!!! note"
+    Example: msgid "!!! note \"Important\"" → msgstr "!!! note \"重要\""
+
+14. COLLAPSIBLE BLOCKS (???): Keep "???" and quote syntax. Translate only
+    the title text inside quotes.
+    Example: msgid "??? \"Click here...\"" → msgstr "??? \"点击这里...\""
+
+15. CONTENT TABS (===): Keep "===" and quote syntax. Translate only the label.
+    Example: msgid "=== \"Before using pip\"" → msgstr "=== \"使用pip之前\""
 
 --- TRANSLATION QUALITY ---
-12. Use natural, fluent Chinese technical documentation style. Avoid word-by-word
-    literal translation. Restructure long English sentences into natural Chinese
-    sentence flow.
-13. Use standard Chinese technical terminology consistently.
-14. For markdown links [text](url): translate the display text in [] but keep the
-    URL in () exactly as-is. Example: [Quick Start](quick_start.md) → [快速开始](quick_start.md)
-15. For headings (# Title): translate the heading text.
-16. DO NOT add "#, fuzzy" markers.
-17. If a msgid is purely structural (symbols, code, file paths only), copy it
-    verbatim to msgstr — do not attempt to translate.
-18. Never invent or guess content. If genuinely unsure about a term, leave it in
-    English rather than creating a wrong translation.
+16. Use natural, fluent Chinese technical documentation style. Avoid word-by-word
+    literal translation.
+17. Use standard Chinese technical terminology consistently.
+18. For markdown links [text](url): translate the display text in [] but keep
+    the URL in () exactly as-is.
+    Example: [Quick Start](quick_start.md) → [快速开始](quick_start.md)
+19. For headings (# Title): translate the heading text.
+20. DO NOT add "#, fuzzy" markers.
+21. If a msgid is purely structural (symbols, code, file paths only), copy it
+    verbatim to msgstr.
+22. Never invent or guess content. If unsure about a term, leave it in English.
 
 {content}"""
 
@@ -166,13 +175,17 @@ async def translate_file(self, po_path: str) -> bool:
                     return False
 
             # Parse the translated snippet and merge back.
-            if not self._merge_translations(po, untranslated, translated_snippet):
+            merged = self._merge_translations(po, untranslated, translated_snippet)
+            if merged == 0:
                 shutil.copy2(backup, po_path)
                 print("FAILED (merge)")
                 return False
 
             po.save(str(path))
-            print("OK")
+            if merged < len(untranslated):
+                print(f"OK ({merged}/{len(untranslated)} merged)")
+            else:
+                print("OK")
             return True
         except Exception as e:
             print(f"ERROR: {e}")
@@ -252,26 +265,31 @@ async def do_chunk(idx: int) -> tuple[int, str | None, str | None]:
         return translated
 
     @staticmethod
-    def _merge_translations(po, untranslated: list[POEntry], translated_snippet: str) -> bool:
-        """Parse translated snippet and merge msgstr values back into *po*."""
+    def _merge_translations(po, untranslated: list[POEntry], translated_snippet: str) -> int:
+        """Parse translated snippet and merge msgstr values back into *po*.
+
+        Returns the number of entries that were successfully merged.
+        If zero entries could be merged, the translation is considered failed.
+        """
         try:
             translated_po = pofile(translated_snippet)
         except Exception:
-            return False
+            return 0
 
         translated_map: dict[str, str] = {}
         for entry in translated_po:
             if entry.msgid and entry.msgstr:
                 translated_map[entry.msgid] = entry.msgstr
 
+        merged = 0
         for entry in untranslated:
             if entry.msgid in translated_map:
                 entry.msgstr = translated_map[entry.msgid]
+                merged += 1
             else:
                 print(f"\n    Missing translation for: {entry.msgid[:60]}...")
-                return False
 
-        return True
+        return merged
 
     @staticmethod
     def _clean_response(response: str) -> str: