Replace division.py

division.py

 #!/usr/bin/env python3
 """
-universal_ratio.py
-一键批量对比任意文件夹下成对 Excel
-python universal_ratio.py folder1 folder2 [folder3 ...]
-输出：summary_ratio.xlsx（每张 Sheet 含 4 列百分比）
+universal_ratio_full.py
+一键批量对比成对 Excel（含首 token 延迟）
+python universal_ratio_full.py folder1 folder2 [folder3 ...]
+输出：summary_ratio.xlsx（每张 Sheet 含 5 列百分比）
 """
 
 import argparse
@@ -13,7 +13,6 @@ import pandas as pd
 from openpyxl import Workbook
 from openpyxl.styles import Alignment, Border, Side, Font, PatternFill
 from openpyxl.utils import get_column_letter
-from openpyxl.styles import Font
 DEFAULT_FONT = Font(name='Arial', size=10)
 # ------------------ 样式 ------------------
 thin = Side('thin')
@@ -21,6 +20,7 @@ border = Border(top=thin, bottom=thin, left=thin, right=thin)
 center = Alignment(horizontal='center', vertical='center', wrap_text=True)
 
 header_font = Font(bold=True, color='FFFFFF')
+black_font = Font(name='微软雅黑', size=11, bold=True, color='000000')  # 000000代表黑色
 blue_fill   = PatternFill('solid', fgColor='e0ffff')   # 纯蓝
 orange_fill = PatternFill('solid', fgColor='faf0e6')   # 橙色
 green_fill  = PatternFill('solid', fgColor='4de680')   # 绿色
@@ -49,18 +49,31 @@ def process_sheet(wb, prefix, path_A, path_B):
     min_rows = min(len(df_A), len(df_B))
     df_A, df_B = df_A.iloc[:min_rows], df_B.iloc[:min_rows]
 
-    # 要对比的列索引（总吞吐、生成吞吐、单路、不带首字）
-    pct_cols = [5, 6, 10, 11]
-    titles   = ['总吞吐量(%)', '生成吞吐量(%)', '单路生成吞吐(%)', '不带首字生成吞吐(%)']
+    # 列索引（总、生成、首token、单路、不带首）
+    pct_cols = [5, 6, 7, 10, 11]
+    titles   = [
+        '总吞吐量(%)',
+        '生成吞吐量(%)',
+        '首token延迟(%)',
+        '单路生成吞吐(%)',
+        '不带首字生成吞吐(%)'
+    ]
 
     pct_rows, pct_df = [], pd.DataFrame()
     for r in range(2, len(df_A)):
-        a = [df_A.iloc[r, c] for c in pct_cols]
-        b = [df_B.iloc[r, c] for c in pct_cols]
-        pct_rows.append([round(a[i] / b[i] * 100, 2) if b[i] else None for i in range(4)])
+        a_vals = [df_A.iloc[r, c] for c in pct_cols]
+        b_vals = [df_B.iloc[r, c] for c in pct_cols]
+        pct = []
+        for i, (a, b) in enumerate(zip(a_vals, b_vals)):
+            if i == 2:  # 首 token 延迟：倒序
+                pct.append(round(b / a * 100, 2) if a else None)
+            else:
+                pct.append(round(a / b * 100, 2) if b else None)
+        pct_rows.append(pct)
+
     pct_df = pd.DataFrame(pct_rows, columns=titles)
     avg = pct_df.mean().round(2).tolist()
-    max_row_ws = [idx + 3 for idx in pct_df.idxmax()]  # 第 3 行开始
+    max_row_ws = [idx + 3 for idx in pct_df.idxmax()]
     min_row_ws = [idx + 3 for idx in pct_df.idxmin()]
 
     ws = wb.create_sheet(title=prefix)
@@ -96,19 +109,21 @@ def process_sheet(wb, prefix, path_A, path_B):
     for r in range(1, len(rows_A) + 1):
         ws.cell(row=r, column=blank2, value=None)
 
-    # ---------- 百分比 ----------
-    pct_start = blank2 + 1
+
+    # 百分比区
+    pct_start = blank2 + 1  # 数据从 blank2+2 列开始
     for c_idx, title in enumerate(titles, 0):
-        cell = ws.cell(row=2, column=pct_start + c_idx, value=title)
-        cell.font = header_font
-        cell.fill = blue_fill
+        ws.cell(row=2, column=pct_start + c_idx, value=title).font = header_font
+        ws.cell(row=2, column=pct_start + c_idx).fill = blue_fill
+
+    # 写入数据
     for r_idx, vals in enumerate(pct_rows, 0):
         for c_idx, val in enumerate(vals, 0):
             ws.cell(row=3 + r_idx, column=pct_start + c_idx, value=val)
 
-    # 平均值
+    # 平均值行（关键修改：标签列向左移动）
     avg_row = 3 + len(pct_rows) + 1
-    ws.cell(row=avg_row, column=pct_start - 1, value='平均值').font = Font(bold=True)
+    ws.cell(row=avg_row, column=pct_start - 1, value='平均值').font = black_font  # 改为-2
     for c_idx, val in enumerate(avg, 0):
         ws.cell(row=avg_row, column=pct_start + c_idx, value=val)
 
@@ -130,7 +145,7 @@ def process_sheet(wb, prefix, path_A, path_B):
         ws.column_dimensions[get_column_letter(col[0].column)].width = max_len
 
 def main():
-    parser = argparse.ArgumentParser(description='批量生成比例对比表')
+    parser = argparse.ArgumentParser(description='批量生成比例对比表（含首 token 延迟）')
     parser.add_argument('folders', nargs='+', help='两个或多个文件夹路径')
     args = parser.parse_args()
 
@@ -140,7 +155,7 @@ def main():
         return
 
     wb = Workbook()
-    wb.remove(wb.active)  # 删除默认空 sheet
+    wb.remove(wb.active)
     for prefix, paths in pairs.items():
         process_sheet(wb, prefix, *paths)