From 172996d68e1f0263e7b4e1f2bff7a5fa336c4d0f Mon Sep 17 00:00:00 2001
From: luotao02 <luotao02@baidu.com>
Date: Thu, 21 May 2026 10:10:05 +0800
Subject: [PATCH] Agent: fix success count logic in extraction scripts to match
 actual log markers

- analyze_extraction_log.sh:
  - Success count: change from '[CPU X] OK' to 'Graph extracted to' (actual marker in code)
  - Success model list: also extract from 'Graph extracted to'
  - Update comments to reflect actual log marker format

- check_extraction_progress.sh:
  - Success count: change from 'Successfully extracted' to 'Graph extracted to'
  - Total attempts condition: sync to match new marker

The main success log output is 'INFO - Graph extracted to: ...', not
'Successfully extracted' or '[CPU X] OK'. Old logic missed most success records.

Example (batch7_safe_run_no_llm.log):
  - Old logic: 283 successes
  - New logic: 4,691 successes
---
 graph_net/agent/scripts/analyze_extraction_log.sh | 15 ++++++++-------
 .../agent/scripts/check_extraction_progress.sh    |  5 +++--
 2 files changed, 11 insertions(+), 9 deletions(-)

diff --git a/graph_net/agent/scripts/analyze_extraction_log.sh b/graph_net/agent/scripts/analyze_extraction_log.sh
index e88d5aeaf..7553efc34 100644
--- a/graph_net/agent/scripts/analyze_extraction_log.sh
+++ b/graph_net/agent/scripts/analyze_extraction_log.sh
@@ -76,13 +76,13 @@ echo ""
 # 1. 总体统计
 echo "--- 一、总体统计 ---"
 # 使用日志中实际的标记格式统计：
-# 成功: [CPU X] OK model_name 或 [GPU X] OK model_name
-# 失败: [CPU X] FAIL model_name 或 [GPU X] FAIL model_name
-# 注：同时支持 CPU-only 模式和 GPU 模式的日志
+# 成功: "Graph extracted to:"（主要成功路径）
+# 失败: "Extraction failed for"
+# 注：部分旧日志可能用 "Successfully extracted"，但当前代码主要输出 "Graph extracted to"
 # 部分日志含二进制数据，用 grep -a 强制按文本处理
 TOTAL=$(grep -ac "Starting extraction for model:" "$LOG_FILE" 2>/dev/null || echo 0)
-SUCCESS=$(grep -aoE '\[(CPU|GPU) [0-9]+\] OK [^ ]+' "$LOG_FILE" 2>/dev/null | wc -l | tr -d ' ')
-FAILED=$(grep -aoE '\[(CPU|GPU) [0-9]+\] FAIL [^ ]+' "$LOG_FILE" 2>/dev/null | wc -l | tr -d ' ')
+SUCCESS=$(grep -ac "Graph extracted to" "$LOG_FILE" 2>/dev/null || echo 0)
+FAILED=$(grep -ac "Extraction failed for" "$LOG_FILE" 2>/dev/null || echo 0)
 PROGRESS_LINES=$(grep -c "PROGRESS" "$LOG_FILE" 2>/dev/null || echo 0)
 
 echo "总尝试数: $TOTAL"
@@ -217,8 +217,9 @@ SUCCESS_LIST="/tmp/${BASE_NAME}_success.txt"
 grep "Starting extraction for model:" "$LOG_FILE" 2>/dev/null | \
     sed 's/.*Starting extraction for model: //' > "$PROCESSED"
 
-grep -aoE '\[(CPU|GPU) [0-9]+\] OK [^ ]+' "$LOG_FILE" 2>/dev/null | \
-    awk '{print $NF}' | sort -u > "$SUCCESS_LIST"
+grep -a "Graph extracted to" "$LOG_FILE" 2>/dev/null | \
+    sed 's/.*Graph extracted to: .*\//\//' | \
+    sed 's/.*samples\///' | sort -u > "$SUCCESS_LIST"
 
 echo "已处理模型列表: $PROCESSED ($(wc -l < "$PROCESSED" 2>/dev/null || echo 0) 个)"
 echo "成功模型列表:   $SUCCESS_LIST ($(wc -l < "$SUCCESS_LIST" 2>/dev/null || echo 0) 个)"
diff --git a/graph_net/agent/scripts/check_extraction_progress.sh b/graph_net/agent/scripts/check_extraction_progress.sh
index fbf1e21b5..05ac0de3d 100644
--- a/graph_net/agent/scripts/check_extraction_progress.sh
+++ b/graph_net/agent/scripts/check_extraction_progress.sh
@@ -137,8 +137,9 @@ echo ""
 echo "====== 统计汇总 ======"
 
 # 从日志统计
-TOTAL_IN_LOG=$(grep -cE "Starting extraction for|Successfully extracted|Extraction failed for" "$LOG_FILE" 2>/dev/null || echo 0)
-SUCCESS_IN_LOG=$(grep -c "Successfully extracted" "$LOG_FILE" 2>/dev/null || echo 0)
+# 注：日志中成功的主要标记是 "Graph extracted to"（而非 "Successfully extracted"）
+TOTAL_IN_LOG=$(grep -cE "Starting extraction for|Graph extracted to|Extraction failed for" "$LOG_FILE" 2>/dev/null || echo 0)
+SUCCESS_IN_LOG=$(grep -c "Graph extracted to" "$LOG_FILE" 2>/dev/null || echo 0)
 FAILED_IN_LOG=$(grep -c "Extraction failed for" "$LOG_FILE" 2>/dev/null || echo 0)
 
 # 从 PROGRESS 行提取进度