Hugging Face's logo

Datasets:
shoumenchougou
/
RWKV-7-Arithmetic

Modalities:
Text
Formats:
json
Size:
10M - 100M
Libraries:
Datasets
Dask
License:
Dataset card Data Studio Files
xet
Community
4
shoumenchougou commited on
Commit
0cd561e
·
verified ·
1 Parent(s): 22fe502

Delete Test.ipynb

Browse files
Files changed (1) hide show
  1. Test.ipynb +0 -612
Test.ipynb DELETED
@@ -1,612 +0,0 @@
1
- {
2
- "cells": [
3
- {
4
- "cell_type": "markdown",
5
- "id": "a2a88bfa",
6
- "metadata": {},
7
- "source": [
8
- "# 测试脚本\n",
9
- "\n",
10
- "使用此脚本需要打开 RWKV Runner，通过调用 API 接口进行批量测试。\n",
11
- "\n",
12
- "使用该脚本时，被测试的 jsonl 文件需要是和训练集相同的单论问答对话；\n",
13
- "\n",
14
- "测试时，会同时生成一个‘测试结果’和‘正确答案’的对比到指定 jsonl 中。"
15
- ]
16
- },
17
- {
18
- "cell_type": "markdown",
19
- "id": "139fd74a",
20
- "metadata": {},
21
- "source": [
22
- "## 测试指定 jsonl 文件"
23
- ]
24
- },
25
- {
26
- "cell_type": "code",
27
- "execution_count": null,
28
- "id": "11af4de0",
29
- "metadata": {},
30
- "outputs": [],
31
- "source": [
32
- "# %% [markdown]\n",
33
- "# # 模型测试脚本 (单单元格, 简化逐行结果, 更新请求参数)\n",
34
- "#\n",
35
- "# 请按以下步骤操作：\n",
36
- "# **1.** **修改配置部分**: 找到下面的 `--- 配置 ---` 部分，并更新 `JSONL_FILE_PATH`, `OUTPUT_JSONL_PATH`, `API_URL`, `HEADERS`。`REQUEST_PARAMS` 已根据您的要求更新。\n",
37
- "# **2.** **检查 API 响应解析**: 在 `get_model_completion` 函数内部，找到标记为 `!!! 重要 !!!` 的部分，确保代码能正确解析你的模型 API 返回的 JSON 数据以提取文本输出。\n",
38
- "# **3.** **运行此单元格**: 执行这个单元格开始测试。结果将逐行写入指定的 `OUTPUT_JSONL_PATH` 文件。\n",
39
- "\n",
40
- "# %%\n",
41
- "import requests\n",
42
- "import json\n",
43
- "import sys\n",
44
- "import os\n",
45
- "from tqdm.notebook import tqdm # 使用 notebook 版本的 tqdm\n",
46
- "from datetime import datetime # 用于添加时间戳\n",
47
- "\n",
48
- "# --- 配置 ---\n",
49
- "# vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv\n",
50
- "# vvvvvvvvvvvvvvvvv 请在这里修改你的配置 vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv\n",
51
- "\n",
52
- "JSONL_FILE_PATH = \"./Test/ADD_base_test.jsonl\" # <--- 替换为你的 **输入** JSONL 文件路径\n",
53
- "OUTPUT_JSONL_PATH = \"./Test/result/ADD_base_test-20250526.jsonl\" # <--- 设置 **输出** JSONL 文件的路径\n",
54
- "API_URL = \"http://19**2.**168.0.103:8022/v1/completions\"\n",
55
- "\n",
56
- "# API 请求的 Headers (如果需要身份验证等，在这里添加)\n",
57
- "HEADERS = {\n",
58
- " 'Content-Type': 'application/json',\n",
59
- " # 'Authorization': 'Bearer YOUR_API_KEY' # 如果需要 API Key\n",
60
- "}\n",
61
- "\n",
62
- "# API 请求的参数 (根据您的要求更新)\n",
63
- "# !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!\n",
64
- "# !!! 这里是更新后的请求参数 !!!\n",
65
- "# !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!\n",
66
- "REQUEST_PARAMS = {\n",
67
- " \"max_tokens\": 100,\n",
68
- " \"temperature\": 0.4,\n",
69
- " \"top_p\": 0, # 注意：top_p=0 比较少见，通常接近 1 或等于 1，请确认是否符合预期\n",
70
- " \"presence_penalty\": 0,\n",
71
- " \"frequency_penalty\": 0,\n",
72
- " \"stop\": [\"\\n\", \"User:\"]\n",
73
- "}\n",
74
- "# !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!\n",
75
- "\n",
76
- "# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
77
- "# ^^^^^^^^^^^^^^^^^^^^^^^^^ 配置结束 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
78
- "# --- 配置结束 ---\n",
79
- "\n",
80
- "print(\"--- 配置加载 ---\")\n",
81
- "print(f\"输入测试文件路径: {JSONL_FILE_PATH}\")\n",
82
- "print(f\"输出结果文件路径: {OUTPUT_JSONL_PATH} (仅包含: expected, actual, is_correct)\")\n",
83
- "print(f\"API 地址: {API_URL}\")\n",
84
- "print(f\"请求参数 (已更新): {REQUEST_PARAMS}\") # 确认参数已更新\n",
85
- "print(\"-\" * 30)\n",
86
- "\n",
87
- "# --- 辅助函数 (与之前相同) ---\n",
88
- "\n",
89
- "def process_jsonl_line(line):\n",
90
- " \"\"\"解析 JSONL 行，提取 prompt 和 expected_answer\"\"\"\n",
91
- " try:\n",
92
- " data = json.loads(line)\n",
93
- " full_text = data.get(\"text\")\n",
94
- " if not full_text:\n",
95
- " return None, None\n",
96
- "\n",
97
- " parts = full_text.split(\"\\n\\nAssistant:\", 1)\n",
98
- " if len(parts) != 2:\n",
99
- " return None, None\n",
100
- "\n",
101
- " prompt = parts[0] + \"\\n\\nAssistant:\"\n",
102
- " expected_answer = parts[1].strip()\n",
103
- "\n",
104
- " return prompt, expected_answer\n",
105
- "\n",
106
- " except (json.JSONDecodeError, Exception):\n",
107
- " return None, None\n",
108
- "\n",
109
- "def get_model_completion(prompt):\n",
110
- " \"\"\"向模型 API 发送请求并获取输出\"\"\"\n",
111
- " payload = {\n",
112
- " \"prompt\": prompt,\n",
113
- " **REQUEST_PARAMS # 这里会自动使用上面更新后的参数\n",
114
- " }\n",
115
- " try:\n",
116
- " response = requests.post(API_URL, headers=HEADERS, json=payload, timeout=60)\n",
117
- " response.raise_for_status()\n",
118
- " response_data = response.json()\n",
119
- "\n",
120
- " # --- 解析模型输出 ---\n",
121
- " # !!! 重要: 这里需要根据你的 API 返回的具体格式来调整 !!!\n",
122
- " model_output = None\n",
123
- " try:\n",
124
- " model_output = response_data.get('choices', [{}])[0].get('text', '').strip()\n",
125
- " # ===> 如果上面这行无效，你需要根据你的 API 返回调整这里 <===\n",
126
- " if model_output is None: model_output = \"\"\n",
127
- " except (KeyError, IndexError, AttributeError, TypeError):\n",
128
- " return None # 表示解析失败\n",
129
- "\n",
130
- " return model_output\n",
131
- "\n",
132
- " except (requests.exceptions.RequestException, json.JSONDecodeError, Exception):\n",
133
- " return None\n",
134
- "\n",
135
- "# --- 主测试逻辑 (与之前相同) ---\n",
136
- "\n",
137
- "# 检查输入文件是否存在\n",
138
- "if not os.path.exists(JSONL_FILE_PATH):\n",
139
- " print(f\"错误：输入文件未找到 '{JSONL_FILE_PATH}'。请确保路径正确并重新运行单元格。\")\n",
140
- "else:\n",
141
- " total_count = 0\n",
142
- " correct_count = 0\n",
143
- " lines_processed = 0\n",
144
- " invalid_format_count = 0\n",
145
- " api_errors = 0\n",
146
- "\n",
147
- " print(f\"\\n开始使用文件进行模型测试: {JSONL_FILE_PATH}\")\n",
148
- " print(f\"简化结果将逐行写入: {OUTPUT_JSONL_PATH}\")\n",
149
- " print(\"-\" * 30)\n",
150
- "\n",
151
- " try:\n",
152
- " with open(OUTPUT_JSONL_PATH, 'a', encoding='utf-8') as outfile:\n",
153
- " try:\n",
154
- " num_lines = sum(1 for line in open(JSONL_FILE_PATH, 'r', encoding='utf-8'))\n",
155
- " except Exception:\n",
156
- " num_lines = None\n",
157
- "\n",
158
- " with open(JSONL_FILE_PATH, 'r', encoding='utf-8') as infile:\n",
159
- " file_iterator = tqdm(infile, total=num_lines, desc=\"测试进度\", unit=\" 行\", bar_format='{l_bar}{bar}| {n_fmt}/{total_fmt} [{elapsed}<{remaining}, {rate_fmt}{postfix}]')\n",
160
- "\n",
161
- " for line_num, line in enumerate(file_iterator):\n",
162
- " lines_processed += 1\n",
163
- " line = line.strip()\n",
164
- " if not line: continue\n",
165
- "\n",
166
- " prompt, expected_answer = process_jsonl_line(line)\n",
167
- "\n",
168
- " if prompt is None or expected_answer is None:\n",
169
- " invalid_format_count += 1\n",
170
- " accuracy = (correct_count / total_count * 100) if total_count > 0 else 0\n",
171
- " file_iterator.set_postfix_str(f\"正确:{correct_count}/{total_count} Acc:{accuracy:.1f}% APIErr:{api_errors} FormatErr:{invalid_format_count}\")\n",
172
- " continue\n",
173
- "\n",
174
- " model_output = get_model_completion(prompt)\n",
175
- "\n",
176
- " if model_output is not None:\n",
177
- " total_count += 1\n",
178
- " is_correct = (model_output == expected_answer)\n",
179
- " if is_correct: correct_count += 1\n",
180
- "\n",
181
- " result_data = {\n",
182
- " \"expected_answer\": expected_answer,\n",
183
- " \"model_output\": model_output,\n",
184
- " \"is_correct\": is_correct\n",
185
- " }\n",
186
- " json_string = json.dumps(result_data, ensure_ascii=False)\n",
187
- " outfile.write(json_string + '\\n')\n",
188
- " outfile.flush()\n",
189
- " else:\n",
190
- " api_errors += 1\n",
191
- "\n",
192
- " accuracy = (correct_count / total_count * 100) if total_count > 0 else 0\n",
193
- " file_iterator.set_postfix_str(f\"正确:{correct_count}/{total_count} Acc:{accuracy:.1f}% APIErr:{api_errors} FormatErr:{invalid_format_count}\")\n",
194
- "\n",
195
- " except FileNotFoundError:\n",
196
- " print(f\"错误：处理期间未找到输入文件 '{JSONL_FILE_PATH}'。\")\n",
197
- " except IOError as e:\n",
198
- " print(f\"错误: 读写文件时发生错误: {e}\")\n",
199
- " except Exception as e:\n",
200
- " print(f\"\\n处理文件时发生意外错误: {e}\")\n",
201
- " import traceback\n",
202
- " traceback.print_exc()\n",
203
- "\n",
204
- " print(\"\\n测试完成!\")\n",
205
- " print(f\"总共处理了 {lines_processed} 行输入文件。\")\n",
206
- " print(f\"{total_count} 个有效测试结果已写入/追加到 {OUTPUT_JSONL_PATH}\")\n",
207
- "\n",
208
- " # --- 显示最终总结 ---\n",
209
- " print(\"\\n\" + \"=\" * 30)\n",
210
- " print(\"测试总结\")\n",
211
- " print(\"=\" * 30)\n",
212
- " print(f\"输入文件中格式无效的行数: {invalid_format_count}\")\n",
213
- " print(f\"API 调用或响应解析错误数: {api_errors}\")\n",
214
- " print(f\"成功获取模型输出的测试用例总数: {total_count}\")\n",
215
- " print(f\"模型正确预测数: {correct_count}\")\n",
216
- "\n",
217
- " if total_count > 0:\n",
218
- " accuracy = (correct_count / total_count) * 100\n",
219
- " print(f\"准确率 (AC Rate) [正确数 / 成功获取结果数]: {accuracy:.2f}% ({correct_count}/{total_count})\")\n",
220
- " elif lines_processed > 0:\n",
221
- " print(\"准确率 (AC Rate): N/A (没有成功获取任何模型输出)\")\n",
222
- " else:\n",
223
- " print(\"准确率 (AC Rate): N/A (没有处理任何行)\")"
224
- ]
225
- },
226
- {
227
- "cell_type": "markdown",
228
- "id": "ae91c93a",
229
- "metadata": {},
230
- "source": [
231
- "## 测试整个文件夹中的全部 jsonl"
232
- ]
233
- },
234
- {
235
- "cell_type": "code",
236
- "execution_count": null,
237
- "id": "43660c81",
238
- "metadata": {},
239
- "outputs": [],
240
- "source": [
241
- "# %% [markdown]\n",
242
- "# # 模型测试脚本 (批量处理文件夹中的jsonl文件)\n",
243
- "#\n",
244
- "# 请按以下步骤操作：\n",
245
- "# **1.** **修改配置部分**: 找到下面的 `--- 配置 ---` 部分，并更新 `INPUT_FOLDER_PATH`, `OUTPUT_FOLDER_PATH`, `API_URL`, `HEADERS`。`REQUEST_PARAMS` 已根据您的要求更新。\n",
246
- "# **2.** **检查 API 响应解析**: 在 `get_model_completion` 函数内部，找到标记为 `!!! 重要 !!!` 的部分，确保代码能正确解析你的模型 API 返回的 JSON 数据以提取文本输出。\n",
247
- "# **3.** **运行此单元格**: 执行这个单元格开始测试。结果将逐文件写入输出文件夹。\n",
248
- "\n",
249
- "# %%\n",
250
- "import requests\n",
251
- "import json\n",
252
- "import sys\n",
253
- "import os\n",
254
- "import csv # 导入csv模块\n",
255
- "from tqdm.notebook import tqdm # 使用 notebook 版本的 tqdm\n",
256
- "from datetime import datetime # 用于添加时间戳\n",
257
- "from IPython import get_ipython # 用于在 Jupyter Notebook 中检测环境\n",
258
- "\n",
259
- "# --- 配置 ---\n",
260
- "# vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv\n",
261
- "# vvvvvvvvvvvvvvvvv 请在这里修改你的配置 vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv\n",
262
- "\n",
263
- "INPUT_FOLDER_PATH = \"./Test\" # 输入路径 \n",
264
- "OUTPUT_FOLDER_PATH = \"./Test/Results/20250526/15\" # 输出路径 \n",
265
- "API_URL = \"http://19**2.**168.0.103:8022/v1/completions\" # API 地址\n",
266
- "\n",
267
- "HEADERS = {\n",
268
- " 'Content-Type': 'application/json',\n",
269
- "}\n",
270
- "\n",
271
- "REQUEST_PARAMS = {\n",
272
- " \"max_tokens\": 100,\n",
273
- " \"temperature\": 0.4,\n",
274
- " \"top_p\": 0,\n",
275
- " \"presence_penalty\": 0,\n",
276
- " \"frequency_penalty\": 0,\n",
277
- " \"stop\": [\"\\n\", \"User:\"]\n",
278
- "}\n",
279
- "# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
280
- "# ^^^^^^^^^^^^^^^^^^^^^^^^^ 配置结束 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n",
281
- "# --- 配置结束 ---\n",
282
- "\n",
283
- "print(\"--- 配置加载 ---\")\n",
284
- "print(f\"输入文件夹路径: {INPUT_FOLDER_PATH}\")\n",
285
- "print(f\"输出文件夹路径: {OUTPUT_FOLDER_PATH}\")\n",
286
- "print(f\"API 地址: {API_URL}\")\n",
287
- "print(f\"请求参数 (已更新): {REQUEST_PARAMS}\")\n",
288
- "print(\"-\" * 30)\n",
289
- "\n",
290
- "# --- 辅助函数 ---\n",
291
- "def process_jsonl_line(line):\n",
292
- " try:\n",
293
- " data = json.loads(line)\n",
294
- " full_text = data.get(\"text\")\n",
295
- " if not full_text: return None, None\n",
296
- " parts = full_text.split(\"\\n\\nAssistant:\", 1)\n",
297
- " if len(parts) != 2: return None, None\n",
298
- " prompt = parts[0] + \"\\n\\nAssistant:\"\n",
299
- " expected_answer = parts[1].strip()\n",
300
- " return prompt, expected_answer\n",
301
- " except: return None, None\n",
302
- "\n",
303
- "def get_model_completion(prompt):\n",
304
- " payload = {\"prompt\": prompt, **REQUEST_PARAMS}\n",
305
- " try:\n",
306
- " response = requests.post(API_URL, headers=HEADERS, json=payload, timeout=60)\n",
307
- " response.raise_for_status()\n",
308
- " response_data = response.json()\n",
309
- " # !!! 重要: 这里需要根据你的 API 返回的具体格式来调整 !!!\n",
310
- " model_output = response_data.get('choices', [{}])[0].get('text', '').strip()\n",
311
- " return model_output if model_output is not None else \"\"\n",
312
- " except (requests.exceptions.RequestException, json.JSONDecodeError, KeyError, IndexError, AttributeError, TypeError) as e:\n",
313
- " # print(f\"API请求或解析错误: {e}\") # 可以取消注释以调试API问题\n",
314
- " return None\n",
315
- "\n",
316
- "def process_single_file(input_file_path, output_file_path):\n",
317
- " total_count, correct_count, lines_processed, invalid_format_count, api_errors = 0, 0, 0, 0, 0\n",
318
- " print(f\"\\n开始处理文件: {input_file_path}\")\n",
319
- " print(f\"详细结果将写入: {output_file_path}\")\n",
320
- " # print(\"-\" * 30) # 减少重复打印分隔线\n",
321
- " try:\n",
322
- " with open(output_file_path, 'w', encoding='utf-8') as outfile:\n",
323
- " try:\n",
324
- " with open(input_file_path, 'r', encoding='utf-8') as f_count: num_lines = sum(1 for _ in f_count)\n",
325
- " except: num_lines = None\n",
326
- "\n",
327
- " with open(input_file_path, 'r', encoding='utf-8') as infile:\n",
328
- " # tqdm的bar_format可以保持简洁一些\n",
329
- " file_iterator = tqdm(infile, total=num_lines, desc=f\"测试 {os.path.basename(input_file_path)}\", unit=\" 行\", bar_format='{l_bar}{bar}| {n_fmt}/{total_fmt} [{elapsed}<{remaining}, {rate_fmt}{postfix}]')\n",
330
- " for line in file_iterator:\n",
331
- " lines_processed += 1\n",
332
- " line = line.strip()\n",
333
- " if not line: continue\n",
334
- " prompt, expected_answer = process_jsonl_line(line)\n",
335
- " if prompt is None or expected_answer is None:\n",
336
- " invalid_format_count += 1\n",
337
- " else:\n",
338
- " model_output = get_model_completion(prompt)\n",
339
- " if model_output is not None:\n",
340
- " total_count += 1 # 有效的API响应和测试对\n",
341
- " is_correct = (model_output == expected_answer)\n",
342
- " if is_correct: correct_count += 1\n",
343
- " result_data = {\"expected_answer\": expected_answer, \"model_output\": model_output, \"is_correct\": is_correct}\n",
344
- " outfile.write(json.dumps(result_data, ensure_ascii=False) + '\\n')\n",
345
- " else:\n",
346
- " api_errors += 1\n",
347
- " # 更新进度条后缀\n",
348
- " accuracy = (correct_count / total_count * 100) if total_count > 0 else 0.0\n",
349
- " file_iterator.set_postfix_str(f\"正确:{correct_count}/{total_count} ({accuracy:.1f}%) APIErr:{api_errors} FormatErr:{invalid_format_count}\")\n",
350
- " outfile.flush()\n",
351
- " except FileNotFoundError:\n",
352
- " print(f\"错误：处理期间未找到输入文件 '{input_file_path}'。\")\n",
353
- " return None\n",
354
- " except IOError as e:\n",
355
- " print(f\"错误: 读写文件 '{output_file_path}' 时发生错误: {e}\")\n",
356
- " return None\n",
357
- " except Exception as e:\n",
358
- " print(f\"\\n处理文件 '{os.path.basename(input_file_path)}' 时发生意外错误: {e}\")\n",
359
- " import traceback; traceback.print_exc()\n",
360
- " return None\n",
361
- " \n",
362
- " final_accuracy = (correct_count / total_count * 100) if total_count > 0 else 0.0\n",
363
- " # 确保返回的字典键名清晰\n",
364
- " return {\n",
365
- " \"filename\": os.path.basename(input_file_path), # 文件全名\n",
366
- " \"lines_processed\": lines_processed,\n",
367
- " \"invalid_format_count\": invalid_format_count,\n",
368
- " \"api_errors\": api_errors,\n",
369
- " \"total_valid_tests\": total_count, # 测试数据条数\n",
370
- " \"correct_predictions\": correct_count, # 正确条数\n",
371
- " \"accuracy_percent\": final_accuracy # 正确率\n",
372
- " }\n",
373
- "\n",
374
- "# --- 主处理逻辑 ---\n",
375
- "def main():\n",
376
- " os.makedirs(OUTPUT_FOLDER_PATH, exist_ok=True)\n",
377
- " today_date = datetime.now().strftime(\"%Y%m%d\")\n",
378
- " \n",
379
- " input_files_paths = [os.path.join(INPUT_FOLDER_PATH, item) for item in os.listdir(INPUT_FOLDER_PATH) if item.endswith('.jsonl') and os.path.isfile(os.path.join(INPUT_FOLDER_PATH, item))]\n",
380
- " \n",
381
- " if not input_files_paths:\n",
382
- " print(f\"错误：在输入文件夹 '{INPUT_FOLDER_PATH}' 中未找到任何jsonl文件。\")\n",
383
- " return\n",
384
- " \n",
385
- " print(f\"\\n找到 {len(input_files_paths)} 个jsonl文件待处理:\")\n",
386
- " for file_path_item in input_files_paths: print(f\" - {os.path.basename(file_path_item)}\")\n",
387
- " \n",
388
- " all_stats_collected = []\n",
389
- " \n",
390
- " summary_txt_filename = f\"accuracy_summary_{today_date}.txt\"\n",
391
- " summary_txt_filepath = os.path.join(OUTPUT_FOLDER_PATH, summary_txt_filename)\n",
392
- " \n",
393
- " summary_csv_filename = f\"accuracy_summary_{today_date}.csv\"\n",
394
- " summary_csv_filepath = os.path.join(OUTPUT_FOLDER_PATH, summary_csv_filename)\n",
395
- "\n",
396
- " try:\n",
397
- " with open(summary_txt_filepath, 'w', encoding='utf-8') as summary_txt_file, \\\n",
398
- " open(summary_csv_filepath, 'w', encoding='utf-8', newline='') as summary_csv_file:\n",
399
- " \n",
400
- " csv_writer = csv.writer(summary_csv_file)\n",
401
- "\n",
402
- " # 写入TXT文件头 (保持不变)\n",
403
- " summary_txt_file.write(f\"测试日期: {today_date}\\n\")\n",
404
- " summary_txt_file.write(\"=\"*40 + \"\\n\")\n",
405
- " summary_txt_file.write(f\"{'文件名':<30} | {'正确率':>7}\\n\")\n",
406
- " summary_txt_file.write(\"=\"*40 + \"\\n\")\n",
407
- " summary_txt_file.flush()\n",
408
- "\n",
409
- " # --- 修改CSV文件头 ---\n",
410
- " csv_header = ['文件名', '测试数据条数', '正确条数', '正确率 (%)']\n",
411
- " csv_writer.writerow(csv_header)\n",
412
- " summary_csv_file.flush()\n",
413
- "\n",
414
- " for current_input_file_path in input_files_paths:\n",
415
- " base_name = os.path.basename(current_input_file_path)\n",
416
- " name_without_ext = os.path.splitext(base_name)[0]\n",
417
- " output_jsonl_file = os.path.join(OUTPUT_FOLDER_PATH, f\"{name_without_ext}_{today_date}.jsonl\")\n",
418
- " \n",
419
- " stats_data = process_single_file(current_input_file_path, output_jsonl_file)\n",
420
- " \n",
421
- " if stats_data:\n",
422
- " all_stats_collected.append(stats_data)\n",
423
- " \n",
424
- " # 写入TXT文件 (文件名缩短逻辑保持)\n",
425
- " filename_display_txt = stats_data['filename']\n",
426
- " if len(filename_display_txt) > 28: filename_display_txt = filename_display_txt[:25] + \"...\"\n",
427
- " summary_txt_file.write(f\"{filename_display_txt:<30} | {stats_data['accuracy_percent']:>6.2f}%\\n\")\n",
428
- " summary_txt_file.flush()\n",
429
- "\n",
430
- " # --- 修改写入CSV文件的数据行 ---\n",
431
- " csv_row = [\n",
432
- " stats_data['filename'], # 文件全名\n",
433
- " stats_data['total_valid_tests'],\n",
434
- " stats_data['correct_predictions'],\n",
435
- " f\"{stats_data['accuracy_percent']:.2f}%\" # 格式化正确率\n",
436
- " ]\n",
437
- " csv_writer.writerow(csv_row)\n",
438
- " summary_csv_file.flush()\n",
439
- " \n",
440
- " # 所有文件处理完毕后，写入总体统计\n",
441
- " if all_stats_collected:\n",
442
- " # 使用 process_single_file 返回的键名\n",
443
- " grand_total_lines = sum(s[\"lines_processed\"] for s in all_stats_collected)\n",
444
- " grand_total_tests = sum(s[\"total_valid_tests\"] for s in all_stats_collected)\n",
445
- " grand_total_correct = sum(s[\"correct_predictions\"] for s in all_stats_collected)\n",
446
- " overall_accuracy_percent = (grand_total_correct / grand_total_tests * 100) if grand_total_tests > 0 else 0.0\n",
447
- " \n",
448
- " # 写入TXT总体统计 (保持不变)\n",
449
- " summary_txt_file.write(\"=\"*40 + \"\\n\")\n",
450
- " summary_txt_file.write(f\"{'总体准确率':<30} | {overall_accuracy_percent:>6.2f}%\\n\")\n",
451
- " summary_txt_file.write(\"=\"*40 + \"\\n\")\n",
452
- " summary_txt_file.flush()\n",
453
- "\n",
454
- " # --- 修改写入CSV的总体统计行 ---\n",
455
- " csv_writer.writerow([]) # 可选：写入一个空行作为分隔\n",
456
- " overall_csv_row = [\n",
457
- " 'TOTAL / OVERALL',\n",
458
- " grand_total_tests,\n",
459
- " grand_total_correct,\n",
460
- " f\"{overall_accuracy_percent:.2f}%\"\n",
461
- " ]\n",
462
- " csv_writer.writerow(overall_csv_row)\n",
463
- " summary_csv_file.flush()\n",
464
- " \n",
465
- " # 控制台打印总结信息\n",
466
- " print(\"\\n\" + \"=\" * 50 + \"\\n所有文件处理完成!\\n\" + \"=\" * 50)\n",
467
- " print(f\"\\n总计处理了 {len(all_stats_collected)} 个文件，{grand_total_lines} 行输入\")\n",
468
- " print(f\"总计 {grand_total_tests} 个有效测试，{grand_total_correct} 个正确预测\")\n",
469
- " print(f\"总体准确率: {overall_accuracy_percent:.2f}%\")\n",
470
- " \n",
471
- " print(\"\\n各文件详细统计 (控制台):\")\n",
472
- " for s_item in all_stats_collected:\n",
473
- " print(f\"\\n文件: {s_item['filename']}\")\n",
474
- " print(f\" 处理行数: {s_item['lines_processed']}\")\n",
475
- " print(f\" 格式错误行数: {s_item['invalid_format_count']}\")\n",
476
- " print(f\" API错误数: {s_item['api_errors']}\")\n",
477
- " print(f\" 有效测试数 (total_valid_tests): {s_item['total_valid_tests']}\")\n",
478
- " print(f\" 正确预测数 (correct_predictions): {s_item['correct_predictions']}\")\n",
479
- " print(f\" 准确率 (accuracy_percent): {s_item['accuracy_percent']:.2f}%\")\n",
480
- " \n",
481
- " print(f\"\\n已将准确率摘要写入到 TXT: {summary_txt_filepath}\")\n",
482
- " print(f\"已将准确率摘要写入到 CSV: {summary_csv_filepath}\")\n",
483
- " \n",
484
- " else: \n",
485
- " message = \"所有文件的处理均未成功生成统计数据。\"\n",
486
- " summary_txt_file.write(\"=\"*40 + \"\\n\" + f\"{message}\\n\")\n",
487
- " # CSV中也可以记录此信息\n",
488
- " csv_writer.writerow([message, 'N/A', 'N/A', 'N/A'])\n",
489
- " summary_csv_file.flush()\n",
490
- " print(f\"\\n{message} 摘要文件已更新。\")\n",
491
- " \n",
492
- " except IOError as e:\n",
493
- " print(f\"\\n错误：处理摘要文件时发生IO错误: {e}\")\n",
494
- " if all_stats_collected: # 尝试打印已收集的数据\n",
495
- " print(\"\\n注意：摘要文件写入可能存在问题，但以下是控制台的统计信息。\")\n",
496
- " # (可以复用上面的控制台打印逻辑，但为了简洁此处省略)\n",
497
- " print(\"请检查控制台输出获取部分结果。\")\n",
498
- "\n",
499
- " if not all_stats_collected and input_files_paths:\n",
500
- " print(\"\\n所有文件的处理均未成功生成统计数据（未在摘要文件中记录此信息，若摘要文件创建失败）。\")\n",
501
- "\n",
502
- "# 执行主函数\n",
503
- "if __name__ == \"__main__\":\n",
504
- " try:\n",
505
- " shell = get_ipython().__class__.__name__\n",
506
- " if shell != 'ZMQInteractiveShell': raise NameError\n",
507
- " except NameError:\n",
508
- " try:\n",
509
- " from tqdm import tqdm as std_tqdm\n",
510
- " globals()['tqdm'] = std_tqdm \n",
511
- " print(\"信息：非Jupyter Notebook环境，使用标准tqdm。\")\n",
512
- " except ImportError:\n",
513
- " print(\"警告：标准tqdm库未安装。进度条可能无法正常显示或仅简单打印。\")\n",
514
- " class dummy_tqdm: # 改进的dummy_tqdm\n",
515
- " def __init__(self, iterable=None, desc=\"\", total=None, unit=\"\", bar_format=None, **kwargs):\n",
516
- " self.iterable, self.desc, self.total, self.current, self.unit = iterable, desc, total, 0, unit\n",
517
- " self.postfix_text = \"\"\n",
518
- " if self.total:\n",
519
- " print(f\"{self.desc}: 开始处理 {self.total} {self.unit}...\")\n",
520
- " else:\n",
521
- " print(f\"{self.desc}: 开始处理...\")\n",
522
- "\n",
523
- " def __iter__(self):\n",
524
- " for i, obj in enumerate(self.iterable):\n",
525
- " yield obj\n",
526
- " self.update(1)\n",
527
- " if self.total and (i + 1) % (self.total // 10 if self.total >=10 else 1) == 0: # 每10%或每项打印\n",
528
- " self.print_status()\n",
529
- " elif not self.total and (i+1) % 50 == 0: # 如果没有total，每50项打印\n",
530
- " self.print_status()\n",
531
- "\n",
532
- "\n",
533
- " def set_postfix_str(self, s):\n",
534
- " self.postfix_text = s\n",
535
- " # 不立即打印，由 __iter__ 中的逻辑控制打印频率\n",
536
- "\n",
537
- " def print_status(self):\n",
538
- " total_str = str(self.total) if self.total else \"?\"\n",
539
- " sys.stdout.write(f\"\\r{self.desc}: {self.current}/{total_str} {self.unit} | {self.postfix_text} \")\n",
540
- " sys.stdout.flush()\n",
541
- "\n",
542
- " def update(self, n=1):\n",
543
- " self.current += n\n",
544
- "\n",
545
- " def close(self):\n",
546
- " self.print_status() # 确保最后的状态被打印\n",
547
- " sys.stdout.write(f\"\\n{self.desc}: 处理完成 {self.current} {self.unit}。\\n\")\n",
548
- " sys.stdout.flush()\n",
549
- " globals()['tqdm'] = dummy_tqdm\n",
550
- " main()"
551
- ]
552
- },
553
- {
554
- "cell_type": "markdown",
555
- "id": "2d12c839",
556
- "metadata": {},
557
- "source": [
558
- "整理一下这个markdown表格，把除了后两列的单独作为一个表格，把后两列重新放到新建的表格里，新建的��格为逐行排列的：\n",
559
- "\n",
560
- "文件名\n",
561
- "数据说明\n",
562
- "示例\n",
563
- "\n",
564
- "表格：\n",
565
- "|数据文件名|\t数据条数|\t数据说明|示例|\n",
566
- "|--------------|----------|------------------------------------|-----------| \n",
567
- "| ADD_4M\t |3997733\t | **1.** 使用‘全角’、‘中文数字’、‘大写中文数字’随机替换整个数字<br>**2.** 运算符附近有 1~2 个随机空格<br>**3.** 含简单自然语言描述/自然语言噪声\t | {\"text\": \"User: 249476576 减 796580834 还剩多少？\\n\\nAssistant: -547104258\"} |\n",
568
- "| ADD_2M\t |1999673\t | **1.** 使用‘全角’、‘中文数字’、‘大写中文数字’随机替换整个数字<br>**2.** 运算符附近有 1~2 个随机空格<br>**3.** 含简单自然语言描述/自然语言噪声 | {\"text\": \"User: 捌仟玖佰捌拾叁点肆贰陆玖+9334160.73357\\n\\nAssistant: 9343144.16047\"} |\n",
569
- "| ADD_base_8M\t |7992002\t | -999~999的全部两两组合，仅算术式子，无任何随机更改\t\t\t | {\"text\": \"User: -999 + -991 = ?\\n\\nAssistant: -1990\"} |\n",
570
- "| ADD_n1m0\t |270887\t | **1.** 使用‘全角’、‘中文数字’、‘大写中文数字’随机替换整个数字；<br>**2.** 运算符附近有 1~2 个随机空格<br>**3.** 额外限制最少有 1 次进/退位 | {\"text\": \"User: 78041304-805555\\n\\nAssistant: 77235749\"} |\n",
571
- "| ADD_n1m1\t |199900\t | **1.** 使用‘全角’、‘中文数字’、‘大写中文数字’随机替换整个数字<br>**2.** 运算符附近有 1~2 个随机空格<br>**3.** 含简单自然语言描述/自然语言噪声<br>4. 限制最少 1 次进/退位\t | {\"text\": \"User: ３９２１８５７２１４+玖仟柒佰肆拾壹萬肆仟叁佰玖拾肆 = ?\\n\\nAssistant: 4019271608\"} | \n",
572
- "| ADD_n2m0\t |270798\t | **1.** 使用‘全角’、‘中文数字’、‘大写中文数字’随机替换整个数字；<br>**2.** 运算符附近有 1~2 个随机空格<br>**3.** 额外限制最少有 2 次进/退位\t\t\t | {\"text\": \"User: 9420731235+陆仟柒佰伍拾陆萬肆仟贰佰玖拾玖\\n\\nAssistant: 9488295534\"} |\n",
573
- "| ADD_n2m1\t |199847\t | **1.** 使用‘全角’、‘中文数字’、‘大写中文数字’随机替换整个数字<br>**2.** 运算符附近有 1~2 个随机空格<br>**3.** 含简单自然语言描述/自然语言噪声<br>4. 限制至少 2 次进/退位\t | {\"text\": \"User: ３０６８５３６２+柒仟壹佰肆拾萬壹仟壹佰贰拾捌 = ?\\n\\nAssistant: 102086490\"} | \n",
574
- "| ADD_n3m0\t |270853\t | **1.** 使用‘全角’、‘中文数字’、‘大写中文数字’随机替换整个数字；<br>**2.** 运算符附近有 1~2 个随机空格<br>**3.** 额外限制最少有 3 次进/退位\t\t\t | {\"text\": \"User: 肆拾叁億肆仟伍佰零肆萬贰仟肆佰捌拾玖加78759970978\\n\\nAssistant: 83105013467\"} |\n",
575
- "| ADD_n4m3\t |271345\t | **1.** 使用‘全角’、‘中文数字’、‘大写中文数字’随机替换整个数字；<br>**2.** 运算符附近有 1~2 个随机空格<br>**3.** 额外限制最少有 4 次进/退位，且至少有一个连续进/退位长度为 3 的连续段 \t| {\"text\": \"User: 六亿两千三百二十八万一千零二十四点五四五二六减２３２９５５８．３３３１７\\n\\nAssistant: 620951466.21209\"} | \n",
576
- "| ADD_random_0.25M\t |249735\t | **1.** 逐位随机进行替换<br>**2.** 替换内容为：大小写中文、全半角数字、（例如：叁2五）<br>**3.** 每个数字和运算符直接随机0~2个空格\t\t\t | {\"text\": \"User: ４６４９５５６９．６７８８６-四百６十亿8千六百４十二万贰千陆百8十玖点九３四五伍=？\\n\\nAssistant: -46039927120.25569\"} |\n",
577
- "| ADD_random_4M\t |4494484\t | **1.** 使用‘全角’、‘中文数字’、‘大写中文数字’随机替换数字的某个位，例如：八4肆<br>**2.** 数字内部和运算符附近有 1~2 个随机空格 | {\"text\": \"User: ８千１百陆十六万捌千8百捌十捌-2仟５佰壹拾八 億2仟六佰１拾八萬陆仟 六佰５拾五\\n\\nAssistant: -251744517767\"} |\n",
578
- "| ADD_many0_50k\t |50000\t | 强化数据，如**2.**000+1\t\t\t | {\"text\": \"User: -315052474 4.00 和 ８２５８２４１４９１７６．００００ 等于？\\n\\nAssistant: 822673624432\"} |\n",
579
- "| ADD_en\t |976394\t | **1.** 使用‘全角’、‘中文数字’、‘大写中文数字’、‘英文数字’随机替换整个数字；<br>例如： 叁3三３three THREE Three\t\t\t | {\"text\": \"User: seven Nine FOUR six six One seven NINE SIX one ZERO zero point Six SIX ONE seven NINE six One NINE+2=?\\n\\nAssistant: 794661796102\"} |\n",
580
- "| ADD_en_base_1M-v1\t |1000000\t | 英语基础数据，从qa_add_base_8M.jsonl下采样直译而来\t\t\t | {\"text\": \"User: 275MINUS six Hundred sixty five= , solve this.\\n\\nAssistant: -390\"} |\n",
581
- "| ADD_en_base_1M-v2\t |1000000\t | 英语基础数据，从qa_add_base_8M.jsonl下采样直译而来\t\t\t | {\"text\": \"User: -684 plusfour Hundred eighty Five=equals WHAT?\\n\\nAssistant: -199\"} |\n",
582
- "| ADD_en_random_mix\t |1464727\t | **1.** 使用‘全角’、‘中文数字’、‘大写中文数字’、‘英文数字’随机替换整个数字；<br>**2.** 对每个单词逐个字母进行大小写随机<br>例如： 叁3三３three THREE Three tHrEe\t\t\t | {\"text\": \"User: ６９４１+Six MILLION Three HUNDRED FIFTY four THOUSAND TWENTY three=?\\n\\nAssistant: 6360964\"} | \n",
583
- "| ADD_en_by-desc\t |489961\t | **1.** 使用‘全角’、‘英文数字’随机替换整个数字<br>**2.** 运算符附近有 1~2 个随机空格<br>**3.** 含简单自然语言描述/自然语言噪声\t\t\t | {\"text\": \"User: What is Eight three five Eight TWO four zero TWO Point Four added To 10994.06?\\n\\nAssistant: 83593396.46\"} |\n",
584
- "| ADD_use-connect\t |1542000\t | 强制使用标准化的英文连接符\t\t\t | {\"text\": \"User: FIVE-BILLION-EIGHT-HUNDRED-NINETY-THREE-MILLION-NINE-HUNDRED-FIFTY-FIVE-THOUSAND-EIGHT-HUNDRED-FORTY-THREE+ NiNE-HUndred-FoRTY-sIx-BiLlIon-sIx-hUnDReD-Eighty-foUr-miLLiON-Eight-HunDRed-seveNty-two-THoUSAND-FOUR-HUndREd-FiFty-nIne-PoInT-EIGHT-nINe-siX-SIX = ?\\n\\nAssistant: 95257882830**2.**8966\"} |\n",
585
- "| X_ch_mix\t |2000000\t | **1.** 使用‘全角’、‘中文数字’、‘大写中文数字’随机替换整个数字<br>**2.** 运算符附近有 1~2 个随机空格<br>**3.** 随机中英问号<br>4. 解方程格式\t\t\t | {\"text\": \"User: x+514= 585.028, 求解 x。\\n\\nAssistant: x = 585.028 - 514 = 7**1.**028\"} | \n",
586
- "| X_ch_en_mix\t |250000\t | **1.** 使用‘全角’、‘英文数字’、‘中文数字’、‘大写中文数字'随机替换数字的某个位<br>**2.** 运算符附近有 1~2 个随机空格<br>**3.** 随机中英问号<br>4. 解方程格式\t\t\t | {\"text\": \"User: x-６８５９１９一７１８５=－68伍９1玖1692捌, 求 x 的值。\\n\\nAssistant: x = -68591916928 + 68591917185 = 257\"} | \n",
587
- "| X_en_random_mix\t |1000000\t | **1.** 使用‘全角’、‘中文数字’、‘大写中文数字’随机替换整个数字<br>**2.** 运算符附近有 1~2 个随机空格<br>**3.** 含简单自然语言描述/自然语言噪声<br>4. 对每个单词逐个字母进行大小写随机<br>5. 解方程格式 |{\"text\": \"User: miNus FouR PoINT TWO fOUr fIve FOUR zERo zerO sIx TiMeS tEN tO the pOWEr OF EigHT+x=－４２４５３９８５４．８ , x = ?\\n\\nAssistant: x = -424539854.8 + 424540060 = 205.2\"} | \n",
588
- "| qa_error\t |248\t | 噪声数据\t\t\t | {\"text\": \"User: 椅子 + 桌子 =\\n\\nAssistant: 无法相加\"} |\n",
589
- "| qa_gibberish_unanswerable\t|10000\t | 乱码，抗干扰，答案为：无法回答\t\t\t | |\n"
590
- ]
591
- },
592
- {
593
- "cell_type": "markdown",
594
- "id": "a233fb46",
595
- "metadata": {},
596
- "source": []
597
- }
598
- ],
599
- "metadata": {
600
- "kernelspec": {
601
- "display_name": "torch",
602
- "language": "python",
603
- "name": "python3"
604
- },
605
- "language_info": {
606
- "name": "python",
607
- "version": "3.12.7"
608
- }
609
- },
610
- "nbformat": 4,
611
- "nbformat_minor": 5
612
- }