From 1e43656ab1dec8f689b2d57122e66d7379cbaf06 Mon Sep 17 00:00:00 2001
From: amy-why-3459 <wuhaiyan17@huawei.com>
Date: Fri, 5 Jun 2026 11:39:07 +0800
Subject: [PATCH] [Bugfix] Update the value of --max-seed-tts-mean-wer in the
 accuracy test. (#4160)

Signed-off-by: amy-why-3459 <wuhaiyan17@huawei.com>
---
 .../qwen3_omni/run_qwen_omni_acc_benchmark.py |  2 +-
 vllm_omni/metrics/utils.py                    | 19 ++++++++++++++++++-
 2 files changed, 19 insertions(+), 2 deletions(-)

diff --git a/tests/e2e/accuracy/qwen3_omni/run_qwen_omni_acc_benchmark.py b/tests/e2e/accuracy/qwen3_omni/run_qwen_omni_acc_benchmark.py
index 5e872e73a..8ce6f1988 100644
--- a/tests/e2e/accuracy/qwen3_omni/run_qwen_omni_acc_benchmark.py
+++ b/tests/e2e/accuracy/qwen3_omni/run_qwen_omni_acc_benchmark.py
@@ -340,7 +340,7 @@ def build_arg_parser() -> argparse.ArgumentParser:
     p.add_argument(
         "--max-seed-tts-mean-wer",
         type=float,
-        default=0.25,
+        default=0.35,
         help="If set, fail when seed_tts_content_error_mean is strictly above this value.",
     )
     p.add_argument(
diff --git a/vllm_omni/metrics/utils.py b/vllm_omni/metrics/utils.py
index 1ed987a19..2062eb5ba 100644
--- a/vllm_omni/metrics/utils.py
+++ b/vllm_omni/metrics/utils.py
@@ -94,7 +94,13 @@ def _format_table(
         if isinstance(value, float):
             return f"{value:,.3f}"
         if isinstance(value, list):
-            return ", ".join(str(f"{v:,.3f}") for v in value)
+            if not value:
+                return ""
+            try:
+                avg = sum(float(v) for v in value) / len(value)
+                return f"{avg:,.3f} (n={len(value)})"
+            except (TypeError, ValueError):
+                return ", ".join(str(v) for v in value)
         return str(value)
 
     table = PrettyTable()
@@ -117,6 +123,17 @@ def _format_table(
         if column_key is None:
             raise ValueError("column_key is required for multi-column mode")
         col_headers = [f"{column_prefix}{row.get(column_key, '?')}" for row in data]
+        # PrettyTable requires unique field names.  When the same column key
+        # appears more than once (e.g. two events with the same stage_id),
+        # deduplicate by appending _2, _3, … to the later occurrences.
+        if len(col_headers) != len(set(col_headers)):
+            seen: dict[str, int] = {}
+            deduped: list[str] = []
+            for h in col_headers:
+                n = seen.get(h, 0)
+                seen[h] = n + 1
+                deduped.append(h if n == 0 else f"{h}_{n + 1}")
+            col_headers = deduped
         table.field_names = ["Field"] + col_headers
         table.align["Field"] = "l"
         for col in col_headers: