SunSec commited on
Commit
c76e731
·
verified ·
1 Parent(s): 8e312b7

Add files using upload-large-folder tool

Browse files
Files changed (50) hide show
  1. deep_search/data_for_rl/MultiHopRAG_only_qa.json +0 -0
  2. deep_search/data_for_rl/domain_counts.json +546 -0
  3. deep_search/data_for_rl/jsonl_to_json.py +27 -0
  4. deep_search/data_for_rl/merge_data.py +47 -0
  5. deep_search/data_for_rl/musique_ans_v1.0_train_only_qa.json +0 -0
  6. deep_search/data_for_rl/musique_ans_v1.0_train_only_qa.jsonl +0 -0
  7. deep_search/data_for_rl/select_source.py +35 -0
  8. deep_search/data_for_rl/source_counts.json +6 -0
  9. deep_search/plot/draw_1.py +44 -0
  10. deep_search/plot/draw_2.py +96 -0
  11. deep_search/plot/draw_3.py +48 -0
  12. deep_search/search_o1/3-26_selected_data_1174_sft_doc_by_itself_QwQ-32B_1.log +2 -0
  13. deep_search/search_o1/bf_run_gen_data_mix_last_8000.sh +197 -0
  14. deep_search/search_o1/infer_wo_search_1.sh +32 -0
  15. deep_search/search_o1/llm_as_judge.sh +53 -0
  16. deep_search/search_o1/nohup.out +72 -0
  17. deep_search/search_o1/run_eval.sh +220 -0
  18. deep_search/search_o1/run_eval_benchmark_2wiki.sh +45 -0
  19. deep_search/search_o1/run_eval_benchmark_frames.sh +31 -0
  20. deep_search/search_o1/run_eval_benchmark_gaia.sh +31 -0
  21. deep_search/search_o1/run_eval_benchmark_train.sh +15 -0
  22. deep_search/search_o1/run_eval_sft.sh +246 -0
  23. deep_search/search_o1/run_eval_sft_3.log +1 -0
  24. deep_search/search_o1/run_gen_data (1).sh +53 -0
  25. deep_search/search_o1/run_gen_data_2.sh +54 -0
  26. deep_search/search_o1/run_gen_data_acc_ratios0.sh +62 -0
  27. deep_search/search_o1/run_gen_data_for_chinese_simple_qa.sh +80 -0
  28. deep_search/search_o1/run_gen_data_selected_2k.sh +216 -0
  29. deep_search/search_o1/run_syn_rollout_filter.sh +32 -0
  30. deep_search/search_o1/search_ex.json +138 -0
  31. deep_search/search_o1/search_without_wiki.json +78 -0
  32. deep_search/search_o1/vllm_serve.sh +10 -0
  33. deep_search/sft_logs/3-12-qa1200_qwq_qwen32b.log +0 -0
  34. deep_search/sft_logs/3-13_r1_32b_1200.log +0 -0
  35. deep_search/sft_logs/3-15_sft_qwq.log +0 -0
  36. deep_search/sft_logs/3-16_difficult_rollout_data_sft_add_prompt.log +0 -0
  37. deep_search/sft_logs/3-16_r1_32b_1478.log +0 -0
  38. deep_search/sft_logs/3-18_2k_qwen32b_inst.log +0 -0
  39. deep_search/sft_logs/3-18_2k_qwq_r1_32b.log +0 -0
  40. deep_search/sft_logs/3-19-1.2k_remove_exp_qwq.log +0 -0
  41. deep_search/sft_logs/3-20_1161_remove_exp_qwen7b_inst.log +0 -0
  42. deep_search/sft_logs/3-20_merged_2462_remove_exp_qwq.log +0 -0
  43. deep_search/sft_logs/3-22-merged_2462_nq_237_total_2699.log +0 -0
  44. deep_search/sft_logs/3-25_strict_selected_1526.log +0 -0
  45. deep_search/sft_logs/3-27.log +0 -0
  46. deep_search/sft_logs/4-2.log +0 -0
  47. deep_search/sft_logs/re.log +0 -0
  48. deep_search/sft_logs/sft_re.log +0 -0
  49. deep_search/sft_logs/test.log +0 -0
  50. ssh_node.sh +6 -0
deep_search/data_for_rl/MultiHopRAG_only_qa.json ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/data_for_rl/domain_counts.json ADDED
@@ -0,0 +1,546 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "film": 65685,
3
+ "history": 21913,
4
+ "music": 21613,
5
+ "geography": 15521,
6
+ "sports": 7353,
7
+ "television": 6374,
8
+ "politics": 5583,
9
+ "literature": 2876,
10
+ "biography": 2521,
11
+ "business": 2506,
12
+ "technology": 1964,
13
+ "education": 1835,
14
+ "people": 1710,
15
+ "genealogy": 977,
16
+ "football": 937,
17
+ "entertainment": 827,
18
+ "media": 826,
19
+ "military": 638,
20
+ "religion": 610,
21
+ "aviation": 587,
22
+ "biology": 507,
23
+ "gaming": 494,
24
+ "law": 493,
25
+ "publishing": 459,
26
+ "royalty": 413,
27
+ "art": 405,
28
+ "comics": 393,
29
+ "theater": 342,
30
+ "family": 332,
31
+ "theatre": 306,
32
+ "government": 290,
33
+ "transportation": 290,
34
+ "motorsport": 270,
35
+ "baseball": 257,
36
+ "physics": 253,
37
+ "crime": 219,
38
+ "architecture": 218,
39
+ "automotive": 216,
40
+ "economics": 195,
41
+ "opera": 192,
42
+ "finance": 189,
43
+ "american football": 180,
44
+ "basketball": 175,
45
+ "philosophy": 153,
46
+ "astronomy": 153,
47
+ "automobile": 152,
48
+ "video games": 147,
49
+ "games": 147,
50
+ "science": 143,
51
+ "mythology": 141,
52
+ "motorsports": 136,
53
+ "university": 132,
54
+ "dance": 128,
55
+ "sport": 121,
56
+ "wrestling": 118,
57
+ "chemistry": 113,
58
+ "fashion": 112,
59
+ "linguistics": 107,
60
+ "olympics": 105,
61
+ "academia": 104,
62
+ "radio": 99,
63
+ "hockey": 98,
64
+ "medicine": 94,
65
+ "boxing": 94,
66
+ "journalism": 78,
67
+ "space": 76,
68
+ "organization": 72,
69
+ "culture": 72,
70
+ "food": 69,
71
+ "retail": 68,
72
+ "geology": 68,
73
+ "cricket": 62,
74
+ "mma": 60,
75
+ "cuisine": 59,
76
+ "psychology": 56,
77
+ "society": 53,
78
+ "mathematics": 53,
79
+ "company": 52,
80
+ "meteorology": 51,
81
+ "nobility": 49,
82
+ "language": 48,
83
+ "energy": 48,
84
+ "engineering": 47,
85
+ "events": 45,
86
+ "manga": 44,
87
+ "archaeology": 41,
88
+ "culinary": 38,
89
+ "tennis": 38,
90
+ "anime": 37,
91
+ "animation": 37,
92
+ "martial arts": 37,
93
+ "brewing": 37,
94
+ "astronautics": 35,
95
+ "beer": 35,
96
+ "banking": 33,
97
+ "industry": 33,
98
+ "health": 32,
99
+ "soccer": 32,
100
+ "nobel prize": 31,
101
+ "demographics": 31,
102
+ "poker": 31,
103
+ "sports betting": 31,
104
+ "folklore": 30,
105
+ "nfl": 30,
106
+ "anthropology": 30,
107
+ "sociology": 30,
108
+ "fiction": 28,
109
+ "ice hockey": 28,
110
+ "person": 27,
111
+ "ballet": 26,
112
+ "firearms": 26,
113
+ "arts": 25,
114
+ "comedy": 25,
115
+ "casino": 24,
116
+ "telecommunications": 24,
117
+ "healthcare": 24,
118
+ "theme parks": 24,
119
+ "athletics": 23,
120
+ "pageant": 23,
121
+ "amusement parks": 23,
122
+ "beverage": 23,
123
+ "environment": 22,
124
+ "theme park": 22,
125
+ "agriculture": 22,
126
+ "law enforcement": 21,
127
+ "formula one": 20,
128
+ "gambling": 20,
129
+ "real estate": 19,
130
+ "navy": 19,
131
+ "rugby": 19,
132
+ "design": 18,
133
+ "books": 17,
134
+ "marketing": 17,
135
+ "amusement_parks": 17,
136
+ "cryptocurrency": 17,
137
+ "advertising": 16,
138
+ "museums": 16,
139
+ "college football": 16,
140
+ "computer science": 15,
141
+ "alcohol": 15,
142
+ "restaurants": 14,
143
+ "golf": 14,
144
+ "horse racing": 13,
145
+ "photography": 13,
146
+ "pharmaceuticals": 12,
147
+ "philanthropy": 12,
148
+ "hospitality": 12,
149
+ "space exploration": 12,
150
+ "tv": 12,
151
+ "lottery": 12,
152
+ "broadcasting": 12,
153
+ "celebrity": 11,
154
+ "beauty pageant": 11,
155
+ "country music": 11,
156
+ "wwe": 11,
157
+ "military history": 10,
158
+ "event": 10,
159
+ "botany": 10,
160
+ "environmental science": 10,
161
+ "video game": 10,
162
+ "aerospace": 9,
163
+ "legal": 9,
164
+ "construction": 9,
165
+ "labor": 9,
166
+ "confectionery": 9,
167
+ "brewery": 8,
168
+ "railway": 8,
169
+ "hinduism": 8,
170
+ "internet": 8,
171
+ "ufc": 8,
172
+ "intelligence": 8,
173
+ "awards": 8,
174
+ "martial_arts": 8,
175
+ "theology": 8,
176
+ "ufology": 7,
177
+ "podcast": 7,
178
+ "stock market": 7,
179
+ "islam": 7,
180
+ "cycling": 7,
181
+ "criminal justice": 7,
182
+ "science fiction": 7,
183
+ "beverages": 7,
184
+ "racing": 7,
185
+ "bible": 7,
186
+ "terrorism": 7,
187
+ "christianity": 7,
188
+ "theme_parks": 6,
189
+ "accounting": 6,
190
+ "combat sports": 6,
191
+ "railroad": 6,
192
+ "chess": 6,
193
+ "disney": 6,
194
+ "transport": 6,
195
+ "musical": 6,
196
+ "currency": 6,
197
+ "art history": 5,
198
+ "spirituality": 5,
199
+ "amusement": 5,
200
+ "organized crime": 5,
201
+ "motorcycle": 5,
202
+ "horseracing": 5,
203
+ "spirits": 5,
204
+ "corporation": 5,
205
+ "esports": 5,
206
+ "motor racing": 5,
207
+ "manufacturing": 5,
208
+ "book": 5,
209
+ "fragrance": 5,
210
+ "international organization": 5,
211
+ "international organizations": 5,
212
+ "modeling": 4,
213
+ "toy": 4,
214
+ "telecom": 4,
215
+ "paleontology": 4,
216
+ "broadway": 4,
217
+ "charity": 4,
218
+ "role-playing games": 4,
219
+ "eurovision": 4,
220
+ "professional wrestling": 4,
221
+ "calendar": 4,
222
+ "organizations": 4,
223
+ "materials science": 4,
224
+ "figure skating": 4,
225
+ "automobile racing": 4,
226
+ "library": 4,
227
+ "australian football": 4,
228
+ "newspaper": 4,
229
+ "adult entertainment": 4,
230
+ "numismatics": 4,
231
+ "corporate": 4,
232
+ "nonprofit": 4,
233
+ "hotel": 4,
234
+ "museum": 4,
235
+ "social media": 4,
236
+ "magazine": 4,
237
+ "holiday": 4,
238
+ "careers": 4,
239
+ "standards": 4,
240
+ "mountaineering": 4,
241
+ "auto racing": 3,
242
+ "food industry": 3,
243
+ "performing arts": 3,
244
+ "amusement rides": 3,
245
+ "pornography": 3,
246
+ "statistics": 3,
247
+ "brands": 3,
248
+ "management": 3,
249
+ "nutrition": 3,
250
+ "maritime": 3,
251
+ "festival": 3,
252
+ "mixed martial arts": 3,
253
+ "ecology": 3,
254
+ "amusement park": 3,
255
+ "military technology": 3,
256
+ "research": 3,
257
+ "mafia": 3,
258
+ "defense": 3,
259
+ "natural disaster": 3,
260
+ "toys": 3,
261
+ "restaurant": 3,
262
+ "competition": 3,
263
+ "conservation": 3,
264
+ "zoology": 3,
265
+ "k-pop": 3,
266
+ "afl": 3,
267
+ "census": 3,
268
+ "mining": 3,
269
+ "nba": 3,
270
+ "paranormal": 3,
271
+ "naval": 3,
272
+ "consumer goods": 3,
273
+ "police": 3,
274
+ "music industry": 3,
275
+ "news": 3,
276
+ "cigars": 3,
277
+ "fast food": 2,
278
+ "urban planning": 2,
279
+ "insurance": 2,
280
+ "animals": 2,
281
+ "pharmacy": 2,
282
+ "occupation": 2,
283
+ "profession": 2,
284
+ "skateboarding": 2,
285
+ "shipping": 2,
286
+ "beauty pageants": 2,
287
+ "encyclopedia": 2,
288
+ "game theory": 2,
289
+ "oil and gas": 2,
290
+ "electronics": 2,
291
+ "parapsychology": 2,
292
+ "materials": 2,
293
+ "wine": 2,
294
+ "cosmetics": 2,
295
+ "psychiatry": 2,
296
+ "railroads": 2,
297
+ "cryptozoology": 2,
298
+ "cryptography": 2,
299
+ "text analysis": 2,
300
+ "cognitive science": 2,
301
+ "adult film": 2,
302
+ "bakery": 2,
303
+ "prize": 2,
304
+ "roller coasters": 2,
305
+ "non-profit": 2,
306
+ "logistics": 2,
307
+ "reality tv": 2,
308
+ "exploration": 2,
309
+ "professions": 2,
310
+ "australian rules football": 2,
311
+ "humanitarian": 2,
312
+ "heraldry": 2,
313
+ "disaster": 2,
314
+ "air force": 2,
315
+ "cooking": 2,
316
+ "dogs": 2,
317
+ "ethnography": 2,
318
+ "comic books": 2,
319
+ "gender": 2,
320
+ "natural disasters": 2,
321
+ "fantasy": 2,
322
+ "animal welfare": 2,
323
+ "feminism": 2,
324
+ "environmental protection": 2,
325
+ "nascar": 2,
326
+ "universities": 2,
327
+ "video_games": 2,
328
+ "economy": 2,
329
+ "occult": 2,
330
+ "libraries": 2,
331
+ "social science": 2,
332
+ "organisation": 2,
333
+ "records": 2,
334
+ "space technology": 2,
335
+ "development": 2,
336
+ "public health": 2,
337
+ "lifestyle": 2,
338
+ "writing systems": 2,
339
+ "languages": 2,
340
+ "european union": 2,
341
+ "work": 1,
342
+ "writing": 1,
343
+ "climate": 1,
344
+ "honors": 1,
345
+ "archery": 1,
346
+ "rail": 1,
347
+ "cinema": 1,
348
+ "space science": 1,
349
+ "philology": 1,
350
+ "retailing": 1,
351
+ "commerce": 1,
352
+ "nhl": 1,
353
+ "microfinance": 1,
354
+ "cookbook": 1,
355
+ "measurement": 1,
356
+ "professional_wrestling": 1,
357
+ "activism": 1,
358
+ "bodybuilding": 1,
359
+ "theme_park": 1,
360
+ "f1": 1,
361
+ "stock car racing": 1,
362
+ "social work": 1,
363
+ "star wars": 1,
364
+ "semiotics": 1,
365
+ "lgbtq": 1,
366
+ "clothing": 1,
367
+ "etymology": 1,
368
+ "sufism": 1,
369
+ "food safety": 1,
370
+ "rail transport": 1,
371
+ "african american studies": 1,
372
+ "propaganda": 1,
373
+ "conspiracy_theory": 1,
374
+ "cars": 1,
375
+ "mechanics": 1,
376
+ "reference": 1,
377
+ "astrology": 1,
378
+ "sailing": 1,
379
+ "travel": 1,
380
+ "videogames": 1,
381
+ "international law": 1,
382
+ "conspiracy": 1,
383
+ "philippines": 1,
384
+ "mechanical engineering": 1,
385
+ "water industry": 1,
386
+ "logic": 1,
387
+ "oil industry": 1,
388
+ "shooting": 1,
389
+ "tourism": 1,
390
+ "circus": 1,
391
+ "apparel": 1,
392
+ "boatbuilding": 1,
393
+ "computer security": 1,
394
+ "bookstores": 1,
395
+ "alcoholic beverages": 1,
396
+ "cocktail": 1,
397
+ "social services": 1,
398
+ "gymnastics": 1,
399
+ "general knowledge": 1,
400
+ "religious studies": 1,
401
+ "cosmology": 1,
402
+ "honours": 1,
403
+ "cannabis": 1,
404
+ "social justice": 1,
405
+ "collegiate athletics": 1,
406
+ "drama": 1,
407
+ "firefighting": 1,
408
+ "human rights": 1,
409
+ "material science": 1,
410
+ "acoustics": 1,
411
+ "pet care": 1,
412
+ "dairy": 1,
413
+ "publications": 1,
414
+ "supply chain management": 1,
415
+ "humanitarian aid": 1,
416
+ "nuclear": 1,
417
+ "security": 1,
418
+ "hip hop": 1,
419
+ "oilfield services": 1,
420
+ "biotechnology": 1,
421
+ "humanities": 1,
422
+ "social development": 1,
423
+ "aquarium": 1,
424
+ "video_game": 1,
425
+ "lgbt": 1,
426
+ "pageants": 1,
427
+ "forensics": 1,
428
+ "badminton": 1,
429
+ "gaelic football": 1,
430
+ "lgbtq+": 1,
431
+ "ice_hockey": 1,
432
+ "pokemon": 1,
433
+ "american_football": 1,
434
+ "packaging": 1,
435
+ "formula1": 1,
436
+ "traditions": 1,
437
+ "plasma science": 1,
438
+ "firearm": 1,
439
+ "political economy": 1,
440
+ "astronaut": 1,
441
+ "korean language": 1,
442
+ "molecular biology": 1,
443
+ "electrical engineering": 1,
444
+ "peacebuilding": 1,
445
+ "sumo": 1,
446
+ "coffee": 1,
447
+ "utility": 1,
448
+ "astronauts": 1,
449
+ "kickboxing": 1,
450
+ "whiskey": 1,
451
+ "ethnology": 1,
452
+ "ethics": 1,
453
+ "cultural heritage": 1,
454
+ "sculpture": 1,
455
+ "watchmaking": 1,
456
+ "motorcycle clubs": 1,
457
+ "video gaming": 1,
458
+ "herbal medicine": 1,
459
+ "kpop": 1,
460
+ "taxation": 1,
461
+ "consulting": 1,
462
+ "earth science": 1,
463
+ "spaceflight": 1,
464
+ "wildlife": 1,
465
+ "dog breeds": 1,
466
+ "buddhism": 1,
467
+ "darts": 1,
468
+ "differential geometry": 1,
469
+ "indigenous studies": 1,
470
+ "boating": 1,
471
+ "snooker": 1,
472
+ "amusements": 1,
473
+ "water": 1,
474
+ "branding": 1,
475
+ "cartoon": 1,
476
+ "regulation": 1,
477
+ "whisky": 1,
478
+ "sports, film": 1,
479
+ "fitness": 1,
480
+ "occupations": 1,
481
+ "nuclear science": 1,
482
+ "wildlife conservation": 1,
483
+ "pharmacology": 1,
484
+ "crafts": 1,
485
+ "metallurgy": 1,
486
+ "gastronomy": 1,
487
+ "physiology": 1,
488
+ "social entrepreneurship": 1,
489
+ "human sexuality": 1,
490
+ "beauty": 1,
491
+ "anarchism": 1,
492
+ "infrastructure": 1,
493
+ "tradition": 1,
494
+ "history, music": 1,
495
+ "civil rights": 1,
496
+ "birding": 1,
497
+ "communication theory": 1,
498
+ "ethnobotany": 1,
499
+ "holidays": 1,
500
+ "college sports": 1,
501
+ "franchising": 1,
502
+ "tabletop games": 1,
503
+ "soap opera": 1,
504
+ "auction": 1,
505
+ "sexuality": 1,
506
+ "author": 1,
507
+ "actors": 1,
508
+ "graphic design": 1,
509
+ "pseudoscience": 1,
510
+ "hindustani music": 1,
511
+ "pottery": 1,
512
+ "behavioral science": 1,
513
+ "grocery": 1,
514
+ "environmental disasters": 1,
515
+ "environmentalism": 1,
516
+ "online gaming": 1,
517
+ "epidemiology": 1,
518
+ "alpine skiing": 1,
519
+ "gardening": 1,
520
+ "paralympics": 1,
521
+ "food and drink": 1,
522
+ "urban legend": 1,
523
+ "online betting": 1,
524
+ "streaming platforms": 1,
525
+ "artificial intelligence": 1,
526
+ "betting": 1,
527
+ "fantasy sports": 1,
528
+ "climate change": 1,
529
+ "climate science": 1,
530
+ "parenting": 1,
531
+ "global health": 1,
532
+ "environmental science and technology": 1,
533
+ "blues": 1,
534
+ "typography": 1,
535
+ "ontology": 1,
536
+ "japanese language": 1,
537
+ "genetics": 1,
538
+ "perfume": 1,
539
+ "international relations": 1,
540
+ "country": 1,
541
+ "publication": 1,
542
+ "public_relations": 1,
543
+ "peace": 1,
544
+ "demography": 1,
545
+ "tax": 1
546
+ }
deep_search/data_for_rl/jsonl_to_json.py ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import json
2
+
3
+ def jsonl_to_json(jsonl_file, json_file):
4
+ """
5
+ 将 JSONL 文件转换为 JSON 文件。
6
+
7
+ :param jsonl_file: 输入的 JSONL 文件路径
8
+ :param json_file: 输出的 JSON 文件路径
9
+ """
10
+ # 用于存储所有的 JSON 对象
11
+ data_list = []
12
+
13
+ # 打开 JSONL 文件并逐行读取
14
+ with open(jsonl_file, 'r', encoding='utf-8') as f:
15
+ for line in f:
16
+ # 解析每一行的 JSON 数据
17
+ json_obj = json.loads(line.strip())
18
+ data_list.append(json_obj)
19
+
20
+ # 将数据写入 JSON 文件
21
+ with open(json_file, 'w', encoding='utf-8') as f:
22
+ json.dump(data_list, f, ensure_ascii=False, indent=4)
23
+
24
+ # 示例调用
25
+ input_file = "/share/project/sunshuang/deep_search/data_for_rl/musique_ans_v1.0_train_only_qa.jsonl"
26
+ output_file = "/share/project/sunshuang/deep_search/data_for_rl/musique_ans_v1.0_train_only_qa.json"
27
+ jsonl_to_json(input_file, output_file)
deep_search/data_for_rl/merge_data.py ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import json
2
+
3
+ def merge_json_files(file_paths, sources, output_file):
4
+ """
5
+ 合并多个 JSON 文件,并为每条数据添加 source 字段。
6
+
7
+ :param file_paths: 输入的 JSON 文件路径列表
8
+ :param sources: 每个文件对应的 source 字段值
9
+ :param output_file: 输出的合并后的 JSON 文件路径
10
+ """
11
+ # 检查输入参数是否匹配
12
+ if len(file_paths) != len(sources):
13
+ raise ValueError("file_paths 和 sources 的长度必须相同")
14
+
15
+ # 存储合并后的数据
16
+ merged_data = []
17
+
18
+ # 遍历每个文件及其对应的 source
19
+ for file_path, source in zip(file_paths, sources):
20
+ with open(file_path, 'r', encoding='utf-8') as f:
21
+ data = json.load(f) # 加载 JSON 数据
22
+ for item in data:
23
+ item['source'] = source # 添加 source 字段
24
+ merged_data.extend(data) # 合并数据
25
+
26
+ # 将合并后的数据写入输出文件
27
+ with open(output_file, 'w', encoding='utf-8') as f:
28
+ json.dump(merged_data, f, ensure_ascii=False, indent=4)
29
+
30
+ # 示例调用
31
+ file_paths = [
32
+ '/share/project/sunshuang/deep_search/data_for_rl/2wiki_train_only_compositional_qa.json',
33
+ '/share/project/sunshuang/deep_search/data_for_rl/hotpot_train_v1.1_only_qa.json',
34
+ '/share/project/sunshuang/deep_search/data_for_rl/MultiHopRAG_only_qa.json',
35
+ '/share/project/sunshuang/deep_search/data_for_rl/musique_ans_v1.0_train_only_qa.json'
36
+ ]
37
+
38
+ sources = [
39
+ '2wiki_compositional',
40
+ 'hotpot',
41
+ 'MultiHopRAG',
42
+ 'musique'
43
+ ]
44
+
45
+ output_file = '/share/project/sunshuang/deep_search/data_for_rl/merged_data.json'
46
+
47
+ merge_json_files(file_paths, sources, output_file)
deep_search/data_for_rl/musique_ans_v1.0_train_only_qa.json ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/data_for_rl/musique_ans_v1.0_train_only_qa.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/data_for_rl/select_source.py ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import json
2
+
3
+ def read_json(file_path):
4
+ """从 JSON 文件中读取数据"""
5
+ with open(file_path, "r", encoding="utf-8") as f:
6
+ data = json.load(f)
7
+ print(f"load from {file_path}, total item: {len(data)}")
8
+ return data
9
+
10
+ def read_jsonl(file_path):
11
+ """从 JSON Lines 文件中读取数据"""
12
+ data = []
13
+ with open(file_path, "r", encoding="utf-8") as f:
14
+ for line in f:
15
+ data.append(json.loads(line.strip()))
16
+ print(f"load from {file_path}, total item: {len(data)}")
17
+ return data
18
+
19
+ def write_json(data, file_path):
20
+ """将数据写入 JSON 文件"""
21
+ with open(file_path, "w", encoding="utf-8") as f:
22
+ json.dump(data, f, ensure_ascii=False, indent=4)
23
+ print(f"write to {file_path}, total item: {len(data)}")
24
+
25
+
26
+ input_file = "/share/project/sunshuang/deep_search/data_for_rl/tagged_domain_keypoints/merged_data_tagged_domain_keypoints_keywords_count.json"
27
+ output_file = "/share/project/sunshuang/deep_search/data_for_rl/musique_tagged/musique_tagged_domain_keypoints_keywords_count.json"
28
+ data = read_json(input_file)
29
+
30
+ selected_data = []
31
+ for item in data:
32
+ if item["source"] == "musique":
33
+ selected_data.append(item)
34
+
35
+ write_json(selected_data, output_file)
deep_search/data_for_rl/source_counts.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "2wiki_compositional": 80860,
3
+ "hotpot": 72991,
4
+ "MultiHopRAG": 2556,
5
+ "musique": 19938
6
+ }
deep_search/plot/draw_1.py ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import matplotlib.pyplot as plt
2
+ import numpy as np
3
+
4
+ # 数据
5
+ categories = ['2Wiki', 'Bamboogle', 'Musique', 'Frames', 'GAIA']
6
+ DG = [26.8, 17.6, 6.2, 10.1, 6.8]
7
+ RAG = [34.8, 31.2, 14.6, 13.5, 0]
8
+ Search_o1 = [50.4, 52.8, 18.8, 30.2, 17.5]
9
+ R1_searcher = [66.4, 68.8, 26.8, 40.3, 20.4]
10
+ cyber = [76.4, 76.8, 29.4, 50.1, 33.7]
11
+
12
+ # 设置柱状图的位置
13
+ x = np.arange(len(categories))
14
+ width = 0.12
15
+
16
+ # 创建图表
17
+ fig, ax = plt.subplots(figsize=(12, 9))
18
+
19
+ ax.bar(x - 2*width, DG, width, label='Directly Gen', color='#8092C4')
20
+ ax.bar(x - width, RAG, width, label='Standard RAG', color='#D0DCE8')
21
+ ax.bar(x, Search_o1, width, label='Search-o1', color='#F7EABB')
22
+ ax.bar(x + width, R1_searcher, width, label='R1-Searcher', color='#EFD6D1')
23
+ ax.bar(x + 2*width, cyber, width, label='SimpleDeepSearcher', color='#C4B7CB')
24
+
25
+ ax.set_ylabel('Accuracy / Percentile (%)', fontsize=12)
26
+ ax.set_xticks(x)
27
+ ax.set_xticklabels(categories, fontsize=10)
28
+ ax.legend(loc='upper right', fontsize=10)
29
+
30
+ # 不再添加数据标签
31
+
32
+ # 调整布局
33
+ plt.ylim(0, 80)
34
+ plt.tight_layout()
35
+
36
+ # 增加顶部空间
37
+ plt.subplots_adjust(top=0.94)
38
+
39
+ # 添加标题
40
+ plt.title("Backbone: Qwen-2.5-7B-Instruct, evaluation: LLM-as-Judge", fontsize=18, ha="center")
41
+
42
+ output_path = 'benchmarks_cyber_7B.png'
43
+ plt.savefig(output_path, dpi=1000)
44
+ print(output_path)
deep_search/plot/draw_2.py ADDED
@@ -0,0 +1,96 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # import matplotlib.pyplot as plt
2
+ # import numpy as np
3
+
4
+ # # 数据
5
+ # categories = ['2Wiki', 'Bamboogle', 'Musique', 'Frames', 'GAIA']
6
+ # DG = [26.8, 17.6, 6.2, 10.1, 6.8]
7
+ # RAG = [34.8, 31.2, 14.6, 13.5, 0]
8
+ # Search_o1 = [50.4, 52.8, 18.8, 30.2, 17.5]
9
+ # R1_searcher = [66.4, 68.8, 26.8, 40.3, 20.4]
10
+ # deep = [66.6, 72.8, 29.3, 0, 0]
11
+ # cyber = [76.4, 76.8, 29.4, 50.1, 33.7]
12
+
13
+ # # 设置柱状图的位置
14
+ # x = np.arange(len(categories))
15
+ # width = 0.12
16
+
17
+ # # 创建图表
18
+ # fig, ax = plt.subplots(figsize=(12, 8))
19
+
20
+ # ax.bar(x - 2*width, DG, width, label='Directly Gen', color='#8092C4')
21
+ # ax.bar(x - width, RAG, width, label='Standard RAG', color='#D0DCE8')
22
+ # ax.bar(x, Search_o1, width, label='Search-o1', color='#F7EABB')
23
+ # ax.bar(x + width, R1_searcher, width, label='R1-Searcher', color='#EFD6D1')
24
+ # ax.bar(x + 2*width, deep, width, label='DeepResearcher', color='#E8BE74')
25
+ # ax.bar(x + 3*width, cyber, width, label='CyberSearcher', color='#C4B7CB')
26
+
27
+
28
+ # ax.set_ylabel('Accuracy / Percentile (%)', fontsize=12)
29
+ # ax.set_xticks(x)
30
+ # ax.set_xticklabels(categories, fontsize=10)
31
+ # ax.legend(loc='upper right', fontsize=10)
32
+
33
+ # # 添加数据标签
34
+ # for i in range(len(categories)):
35
+ # ax.text(x[i] - 2*width, DG[i] + 0.5, f'{DG[i]:.1f}', ha='center', va='bottom', fontsize=10)
36
+ # ax.text(x[i] - width, RAG[i] + 0.5, f'{RAG[i]:.1f}', ha='center', va='bottom', fontsize=10)
37
+ # ax.text(x[i], Search_o1[i] + 0.5, f'{Search_o1[i]:.1f}', ha='center', va='bottom', fontsize=10)
38
+ # ax.text(x[i] + width, R1_searcher[i] + 0.5, f'{R1_searcher[i]:.1f}', ha='center', va='bottom', fontsize=10)
39
+ # ax.text(x[i] + 2*width, deep[i] + 0.5, f'{deep[i]:.1f}', ha='center', va='bottom', fontsize=10)
40
+ # ax.text(x[i] + 3*width, cyber[i] + 0.5, f'{cyber[i]:.1f}', ha='center', va='bottom', fontsize=10)
41
+
42
+ # # 调整布局
43
+ # plt.ylim(0, 80)
44
+ # plt.tight_layout()
45
+
46
+ # output_path = '/opt/aps/workdir/benchmarks_cyber_7B.png'
47
+ # plt.savefig(output_path, dpi=1000)
48
+ # print(output_path)
49
+ import matplotlib.pyplot as plt
50
+ import numpy as np
51
+
52
+ # 数据
53
+ categories = ['2Wiki', 'Bamboogle', 'Musique', 'Frames', 'GAIA']
54
+ DG = [39.8,29.6,17.4,31.3,11.7]
55
+ RAG = [50.6,46.4,19.6,31.6,0]
56
+ Search_o1 = [78.0,78.4,36.4,64.4,37.9]
57
+ cyber = [82.8,86.4,34.4,68.7,47.6]
58
+ webthinker= [0,0,0,0,44.7]
59
+ # 设置柱状图的位置
60
+ x = np.arange(len(categories))
61
+ width = 0.12
62
+
63
+ # 创建图表
64
+ fig, ax = plt.subplots(figsize=(12, 9))
65
+
66
+ ax.bar(x - 2*width, DG, width, label='Directly Gen', color='#8092C4')
67
+ ax.bar(x - width, RAG, width, label='Standard RAG', color='#D0DCE8')
68
+ ax.bar(x, Search_o1, width, label='Search-o1', color='#F7EABB')
69
+ ax.bar(x + 2*width, webthinker, width, label='WebThinker', color='#EFD6D1')
70
+ # ax.bar(x + 2*width, deep, width, label='DeepResearcher', color='#E8BE74')
71
+ ax.bar(x + 1*width, cyber, width, label='SimpleDeepSearcher', color='#C4B7CB')
72
+
73
+
74
+ ax.set_ylabel('Accuracy / Percentile (%)', fontsize=12)
75
+ ax.set_xticks(x)
76
+ ax.set_xticklabels(categories, fontsize=10)
77
+ ax.legend(loc='upper right', fontsize=10)
78
+
79
+ # 不再添加数据标签
80
+
81
+ # 调整布局
82
+ plt.ylim(0, 90)
83
+ plt.tight_layout()
84
+
85
+ # 增加顶部空间
86
+ plt.subplots_adjust(top=0.94)
87
+
88
+ # 添加标题
89
+ # plt.title("Use Qwen-2.5-32B-Instruct as the backbone and use LLM-as-Judge as the metric.", fontsize=18, ha="center")
90
+ plt.title("Backbone: Qwen-2.5-32B-Instruct, evaluation: LLM-as-Judge", fontsize=18, ha="center")
91
+
92
+
93
+
94
+ output_path = 'benchmarks_cyber_32B.png'
95
+ plt.savefig(output_path, dpi=1000)
96
+ print(output_path)
deep_search/plot/draw_3.py ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import matplotlib.pyplot as plt
2
+ import numpy as np
3
+
4
+ # 数据
5
+ categories = ['2Wiki', 'Bamboogle', 'Musique', 'Frames', 'GAIA']
6
+ DG = [39.8,29.6,17.4,31.3,11.7]
7
+ RAG = [50.6,46.4,19.6,31.6,0]
8
+ Search_o1 = [78.0,78.4,36.4,64.4,37.9]
9
+ cyber = [82.8,86.4,34.4,68.7,47.6]
10
+ webthinker= [0,0,0,0,44.7]
11
+
12
+ # 设置柱状图的位置
13
+ x = np.arange(len(categories))
14
+ width = 0.12
15
+
16
+ # 创建图表
17
+ fig, ax = plt.subplots(figsize=(12, 9))
18
+
19
+ # 处理RAG数据,过滤掉0值 [[2]][[5]]
20
+ rag_mask = np.array(RAG) != 0
21
+ rag_x = x[rag_mask] - width
22
+ rag_heights = np.array(RAG)[rag_mask]
23
+
24
+ # 绘制各模型柱状图
25
+ ax.bar(x - 2*width, DG, width, label='Directly Gen', color='#8092C4')
26
+ ax.bar(rag_x, rag_heights, width, label='Standard RAG', color='#D0DCE8')
27
+ ax.bar(x, Search_o1, width, label='Search-o1', color='#F7EABB')
28
+ ax.bar(x + 1*width, cyber, width, label='SimpleDeepSearcher', color='#C4B7CB')
29
+ ax.bar(x + 2*width, webthinker, width, label='WebThinker', color='#EFD6D1')
30
+
31
+ # 扩展x轴范围以包含所有柱子 [[10]]
32
+ ax.set_xlim(x[0]-2.5*width, x[-1]+2.5*width)
33
+
34
+ # 设置图表属性
35
+ ax.set_ylabel('Accuracy / Percentile (%)', fontsize=12)
36
+ ax.set_xticks(x)
37
+ ax.set_xticklabels(categories, fontsize=10)
38
+ ax.legend(loc='upper right', fontsize=10)
39
+ plt.ylim(0, 90)
40
+ plt.tight_layout()
41
+ plt.subplots_adjust(top=0.94)
42
+ plt.title("Use Qwen-2.5-32B-Instruct as the backbone and use LLM-as-Judge as the metric.",
43
+ fontsize=18, ha="center")
44
+
45
+ # 保存图表
46
+ output_path = 'benchmarks_cyber_32B.png'
47
+ plt.savefig(output_path, dpi=1000)
48
+ print(output_path)
deep_search/search_o1/3-26_selected_data_1174_sft_doc_by_itself_QwQ-32B_1.log ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ nohup: ignoring input
2
+ /share/project/miniconda/envs/search_o1/bin/python: can't open file '/share/project/sunshuang/deep_search/search_o1/wait_eval_use_one_model_for_ckpt.py': [Errno 2] No such file or directory
deep_search/search_o1/bf_run_gen_data_mix_last_8000.sh ADDED
@@ -0,0 +1,197 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # worker 1
2
+ export CUDA_VISIBLE_DEVICES=0,1
3
+ python /share/project/sunshuang/deep_search/search_o1/scripts/search_o1_sum_all_webpage_rollout.py\
4
+ --dataset_name bf_merged_data_tagged_domain_keypoints_keywords_count_add_source_answer \
5
+ --cache_dir_base cache_eval_sum_all_webpage \
6
+ --output_dir_base /share/project/baifei/filter-syn-ai-search/rollout/0321/result_154/search/roll-32b-inst-search-154items \
7
+ --split gen \
8
+ --max_search_limit 10 \
9
+ --max_turn 10 \
10
+ --top_k 10 \
11
+ --rollout_num 5 \
12
+ --max_doc_len 5000 \
13
+ --model_path "/share/project/zhipengchen/model/Qwen2.5-32B-Instruct" \
14
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
15
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/baifei/filter-syn-ai-search/rollout/0321/result_154/search/search-154.log 2>&1 &
16
+
17
+ cat <<test1
18
+ export CUDA_VISIBLE_DEVICES=6,7
19
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
20
+ --dataset_name mixed_data_left_last_8000_split_2 \
21
+ --cache_dir_base cache_eval_sum_all_webpage \
22
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_mixed_data_left_last_8000_split_2 \
23
+ --split gen \
24
+ --max_search_limit 10 \
25
+ --max_turn 10 \
26
+ --top_k 10 \
27
+ --rollout_num 1 \
28
+ --max_doc_len 5000 \
29
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
30
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
31
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/mixed_data_left_last_8000_split_2.log 2>&1 &
32
+
33
+
34
+
35
+ export CUDA_VISIBLE_DEVICES=4,5
36
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
37
+ --dataset_name mixed_data_left_last_8000_split_3 \
38
+ --cache_dir_base cache_eval_sum_all_webpage \
39
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_mixed_data_left_last_8000_split_3 \
40
+ --split gen \
41
+ --max_search_limit 10 \
42
+ --max_turn 10 \
43
+ --top_k 10 \
44
+ --rollout_num 3 \
45
+ --max_doc_len 5000 \
46
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
47
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
48
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/mixed_data_left_last_8000_split_3.log 2>&1 &
49
+
50
+
51
+
52
+
53
+ export CUDA_VISIBLE_DEVICES=6,7
54
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
55
+ --dataset_name mixed_data_left_last_8000_split_4 \
56
+ --cache_dir_base cache_eval_sum_all_webpage \
57
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_mixed_data_left_last_8000_split_4 \
58
+ --split gen \
59
+ --max_search_limit 10 \
60
+ --max_turn 10 \
61
+ --top_k 10 \
62
+ --rollout_num 3 \
63
+ --max_doc_len 5000 \
64
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
65
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
66
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/mixed_data_left_last_8000_split_4.log 2>&1 &
67
+
68
+
69
+ # worker 2
70
+ export CUDA_VISIBLE_DEVICES=4,5
71
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
72
+ --dataset_name mixed_data_left_last_8000_split_5 \
73
+ --cache_dir_base cache_eval_sum_all_webpage \
74
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_mixed_data_left_last_8000_split_5 \
75
+ --split gen \
76
+ --max_search_limit 10 \
77
+ --max_turn 10 \
78
+ --top_k 10 \
79
+ --rollout_num 3 \
80
+ --max_doc_len 5000 \
81
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
82
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
83
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/mixed_data_left_last_8000_split_5.log 2>&1 &
84
+
85
+
86
+ export CUDA_VISIBLE_DEVICES=0,1
87
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
88
+ --dataset_name mixed_data_left_last_8000_split_6 \
89
+ --cache_dir_base cache_eval_sum_all_webpage \
90
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_mixed_data_left_last_8000_split_6 \
91
+ --split gen \
92
+ --max_search_limit 10 \
93
+ --max_turn 10 \
94
+ --top_k 10 \
95
+ --rollout_num 1 \
96
+ --max_doc_len 5000 \
97
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
98
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
99
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/mixed_data_left_last_8000_split_6.log 2>&1 &
100
+
101
+
102
+ export CUDA_VISIBLE_DEVICES=4,5
103
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
104
+ --dataset_name mixed_data_left_last_8000_split_7 \
105
+ --cache_dir_base cache_eval_sum_all_webpage \
106
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_mixed_data_left_last_8000_split_7 \
107
+ --split gen \
108
+ --max_search_limit 10 \
109
+ --max_turn 10 \
110
+ --top_k 10 \
111
+ --rollout_num 3 \
112
+ --max_doc_len 5000 \
113
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
114
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
115
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/mixed_data_left_last_8000_split_7.log 2>&1 &
116
+
117
+ export CUDA_VISIBLE_DEVICES=6,7
118
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
119
+ --dataset_name mixed_data_left_last_8000_split_8 \
120
+ --cache_dir_base cache_eval_sum_all_webpage \
121
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_mixed_data_left_last_8000_split_8 \
122
+ --split gen \
123
+ --max_search_limit 10 \
124
+ --max_turn 10 \
125
+ --top_k 10 \
126
+ --rollout_num 3 \
127
+ --max_doc_len 5000 \
128
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
129
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
130
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/mixed_data_left_last_8000_split_8.log 2>&1 &
131
+
132
+
133
+ # worker 3
134
+ export CUDA_VISIBLE_DEVICES=0,1
135
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
136
+ --dataset_name mixed_data_left_last_8000_split_9 \
137
+ --cache_dir_base cache_eval_sum_all_webpage \
138
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_mixed_data_left_last_8000_split_9 \
139
+ --split gen \
140
+ --max_search_limit 10 \
141
+ --max_turn 10 \
142
+ --top_k 10 \
143
+ --rollout_num 3 \
144
+ --max_doc_len 5000 \
145
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
146
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
147
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/mixed_data_left_last_8000_split_9.log 2>&1 &
148
+
149
+
150
+
151
+ export CUDA_VISIBLE_DEVICES=2,3
152
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
153
+ --dataset_name mixed_data_left_last_8000_split_10 \
154
+ --cache_dir_base cache_eval_sum_all_webpage \
155
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_mixed_data_left_last_8000_split_10 \
156
+ --split gen \
157
+ --max_search_limit 10 \
158
+ --max_turn 10 \
159
+ --top_k 10 \
160
+ --rollout_num 3 \
161
+ --max_doc_len 5000 \
162
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
163
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
164
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/mixed_data_left_last_8000_split_10.log 2>&1 &
165
+
166
+
167
+ export CUDA_VISIBLE_DEVICES=4,5
168
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
169
+ --dataset_name mixed_data_left_last_8000_split_11 \
170
+ --cache_dir_base cache_eval_sum_all_webpage \
171
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_mixed_data_left_last_8000_split_11 \
172
+ --split gen \
173
+ --max_search_limit 10 \
174
+ --max_turn 10 \
175
+ --top_k 10 \
176
+ --rollout_num 3 \
177
+ --max_doc_len 5000 \
178
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
179
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
180
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/mixed_data_left_last_8000_split_11.log 2>&1 &
181
+
182
+
183
+ export CUDA_VISIBLE_DEVICES=6,7
184
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
185
+ --dataset_name mixed_data_left_last_8000_split_12 \
186
+ --cache_dir_base cache_eval_sum_all_webpage \
187
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_mixed_data_left_last_8000_split_12 \
188
+ --split gen \
189
+ --max_search_limit 10 \
190
+ --max_turn 10 \
191
+ --top_k 10 \
192
+ --rollout_num 3 \
193
+ --max_doc_len 5000 \
194
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
195
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
196
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/mixed_data_left_last_8000_split_12.log 2>&1 &
197
+ test1
deep_search/search_o1/infer_wo_search_1.sh ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ ############################
3
+ # musique
4
+
5
+
6
+ export CUDA_VISIBLE_DEVICES=6,7
7
+ python scripts/inference_wo_search.py \
8
+ --model_name QwQ-32B \
9
+ --model_path /share/project/zhipengchen/model/QwQ-32B \
10
+ --dataset_name musique_syn \
11
+ --input_file /share/project/sunshuang/deep_search/search_o1/data/test/musique_syn.json
12
+
13
+
14
+ python scripts/inference_wo_search.py \
15
+ --model_name DeepSeek-R1-Distill-Qwen-32B \
16
+ --model_path /share/project/zhipengchen/model/DeepSeek-R1-Distill-Qwen-32B \
17
+ --dataset_name musique_syn \
18
+ --input_file /share/project/sunshuang/deep_search/search_o1/data/test/musique_syn.json
19
+
20
+
21
+ python scripts/inference_wo_search.py \
22
+ --model_name Qwen2.5-7B-Instruct \
23
+ --model_path /share/project/zhipengchen/model/Qwen2.5-7B-Instruct \
24
+ --dataset_name musique_syn \
25
+ --input_file /share/project/sunshuang/deep_search/search_o1/data/test/musique_syn.json
26
+
27
+ python scripts/inference_wo_search.py \
28
+ --model_name Qwen2.5-32B-Instruct \
29
+ --model_path /share/project/zhipengchen/model/Qwen2.5-32B-Instruct \
30
+ --dataset_name musique_syn \
31
+ --input_file /share/project/sunshuang/deep_search/search_o1/data/test/musique_syn.json
32
+
deep_search/search_o1/llm_as_judge.sh ADDED
@@ -0,0 +1,53 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ export CUDA_VISIBLE_DEVICES=0,1
3
+ nohup python -u scripts/llm_as_judge.py \
4
+ --model_path /share/project/zhipengchen/model/QwQ-32B/ \
5
+ --input_file /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:9578#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_new_instruction_2k_sft_remove_exp/73/syn_zh/turn_9.json \
6
+ --language zh > /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:9578#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_new_instruction_2k_sft_remove_exp/73/syn_zh/inf.log 2>&1 &
7
+
8
+
9
+
10
+
11
+ export CUDA_VISIBLE_DEVICES=2,3
12
+ nohup python -u scripts/llm_as_judge.py \
13
+ --model_path /share/project/zhipengchen/model/QwQ-32B/ \
14
+ --input_file /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:12665#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_merged_new_instruction_1161_1301_sft_remove_exp_2462/154/syn_zh/turn_11.json \
15
+ --language zh > /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:12665#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_merged_new_instruction_1161_1301_sft_remove_exp_2462/154/syn_zh/inf.log 2>&1 &
16
+
17
+
18
+ export CUDA_VISIBLE_DEVICES=2,3
19
+ nohup python -u scripts/llm_as_judge.py \
20
+ --model_path /share/project/zhipengchen/model/QwQ-32B/ \
21
+ --input_file /share/project/sunshuang/deep_search/search_o1/output/output_eval/outputs_sum_all_webpage_qwq_new_setting_syn_zh/turn_8.json \
22
+ --language zh > /share/project/sunshuang/deep_search/search_o1/output/output_eval/outputs_sum_all_webpage_qwq_new_setting_syn_zh/inf.log 2>&1 &
23
+
24
+
25
+ # /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:12665#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_merged_new_instruction_1161_1301_sft_remove_exp_2462/228/syn_zh/turn_10.json
26
+
27
+ export CUDA_VISIBLE_DEVICES=2,3
28
+ nohup python -u scripts/llm_as_judge.py \
29
+ --model_path /share/project/zhipengchen/model/QwQ-32B/ \
30
+ --input_file /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:12665#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_merged_new_instruction_1161_1301_sft_remove_exp_2462/228/syn_zh/turn_10.json \
31
+ --language zh > /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:12665#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_merged_new_instruction_1161_1301_sft_remove_exp_2462/228/syn_zh/inf_llm.log 2>&1 &
32
+
33
+
34
+
35
+
36
+ nohup python -u scripts/llm_as_judge.py \
37
+ --model_name gpt-4o \
38
+ --model_path gpt-4o \
39
+ --input_file /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:9578#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_new_instruction_2k_sft_remove_exp/73/musique_syn/turn_13.json \
40
+ --language en > /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:9578#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_new_instruction_2k_sft_remove_exp/73/musique_syn/inf_gpt-4o.log 2>&1 &
41
+
42
+
43
+ nohup python -u scripts/llm_as_judge.py \
44
+ --model_name gpt-4o \
45
+ --model_path gpt-4o \
46
+ --input_file /share/project/sunshuang/deep_search/search_o1/output/output_eval/outputs_sum_all_webpage_qwq_new_setting_musique_syn/turn_13.json \
47
+ --language en > /share/project/sunshuang/deep_search/search_o1/output/output_eval/outputs_sum_all_webpage_qwq_new_setting_musique_syn/inf_gpt-4o.log 2>&1 &
48
+
49
+ nohup python -u scripts/llm_as_judge.py \
50
+ --model_name gpt-4o \
51
+ --model_path gpt-4o \
52
+ --input_file /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:12665#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_merged_new_instruction_1161_1301_sft_remove_exp_2462/228/musique_syn/turn_13.json \
53
+ --language en > /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:12665#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_merged_new_instruction_1161_1301_sft_remove_exp_2462/228/musique_syn/inf_gpt-4o.log 2>&1 &
deep_search/search_o1/nohup.out ADDED
@@ -0,0 +1,72 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ usage: search_o1_sum_all_webpage_rollout.py [-h] --dataset_name DATASET_NAME
2
+ --split SPLIT
3
+ [--subset_num SUBSET_NUM]
4
+ [--max_search_limit MAX_SEARCH_LIMIT]
5
+ [--max_turn MAX_TURN]
6
+ [--top_k TOP_K]
7
+ [--max_doc_len MAX_DOC_LEN]
8
+ [--use_jina]
9
+ [--jina_api_key JINA_API_KEY]
10
+ --model_path MODEL_PATH
11
+ [--temperature TEMPERATURE]
12
+ [--top_p TOP_P]
13
+ [--top_k_sampling TOP_K_SAMPLING]
14
+ [--repetition_penalty REPETITION_PENALTY]
15
+ [--max_tokens MAX_TOKENS]
16
+ --bing_subscription_key
17
+ BING_SUBSCRIPTION_KEY
18
+ [--bing_endpoint BING_ENDPOINT]
19
+ --cache_dir_base CACHE_DIR_BASE
20
+ --output_dir_base OUTPUT_DIR_BASE
21
+ [--is_exclude_urls]
22
+ [--rollout_num ROLLOUT_NUM]
23
+ search_o1_sum_all_webpage_rollout.py: error: unrecognized arguments:
24
+ Traceback (most recent call last):
25
+ File "/share/project/sunshuang/deep_search/search_o1/scripts/search_o1_sum_all_webpage.py", line 13, in <module>
26
+ from transformers import AutoTokenizer
27
+ File "/share/project/miniconda/envs/search_o1/lib/python3.9/site-packages/transformers/__init__.py", line 26, in <module>
28
+ from . import dependency_versions_check
29
+ File "/share/project/miniconda/envs/search_o1/lib/python3.9/site-packages/transformers/dependency_versions_check.py", line 16, in <module>
30
+ from .utils.versions import require_version, require_version_core
31
+ File "/share/project/miniconda/envs/search_o1/lib/python3.9/site-packages/transformers/utils/__init__.py", line 21, in <module>
32
+ from huggingface_hub import get_full_repo_name # for backward compatibility
33
+ File "<frozen importlib._bootstrap>", line 1055, in _handle_fromlist
34
+ File "/share/project/miniconda/envs/search_o1/lib/python3.9/site-packages/huggingface_hub/__init__.py", line 962, in __getattr__
35
+ submod = importlib.import_module(submod_path)
36
+ File "/share/project/miniconda/envs/search_o1/lib/python3.9/importlib/__init__.py", line 127, in import_module
37
+ return _bootstrap._gcd_import(name[level:], package, level)
38
+ File "/share/project/miniconda/envs/search_o1/lib/python3.9/site-packages/huggingface_hub/hf_api.py", line 63, in <module>
39
+ from ._inference_endpoints import InferenceEndpoint, InferenceEndpointType
40
+ File "/share/project/miniconda/envs/search_o1/lib/python3.9/site-packages/huggingface_hub/_inference_endpoints.py", line 9, in <module>
41
+ from .inference._client import InferenceClient
42
+ File "/share/project/miniconda/envs/search_o1/lib/python3.9/site-packages/huggingface_hub/inference/_client.py", line 45, in <module>
43
+ from huggingface_hub.inference._common import (
44
+ File "/share/project/miniconda/envs/search_o1/lib/python3.9/site-packages/huggingface_hub/inference/_common.py", line 53, in <module>
45
+ from ._generated.types import ChatCompletionStreamOutput, TextGenerationStreamOutput
46
+ File "/share/project/miniconda/envs/search_o1/lib/python3.9/site-packages/huggingface_hub/inference/_generated/types/__init__.py", line 62, in <module>
47
+ from .feature_extraction import FeatureExtractionInput, FeatureExtractionInputTruncationDirection
48
+ File "<frozen importlib._bootstrap>", line 1007, in _find_and_load
49
+ File "<frozen importlib._bootstrap>", line 986, in _find_and_load_unlocked
50
+ File "<frozen importlib._bootstrap>", line 680, in _load_unlocked
51
+ File "<frozen importlib._bootstrap_external>", line 846, in exec_module
52
+ File "<frozen importlib._bootstrap_external>", line 941, in get_code
53
+ File "<frozen importlib._bootstrap_external>", line 1040, in get_data
54
+ KeyboardInterrupt
55
+ usage: search_o1_sum_all_webpage.py [-h] --dataset_name DATASET_NAME --split
56
+ SPLIT [--subset_num SUBSET_NUM]
57
+ [--max_search_limit MAX_SEARCH_LIMIT]
58
+ [--max_turn MAX_TURN] [--top_k TOP_K]
59
+ [--max_doc_len MAX_DOC_LEN] [--use_jina]
60
+ [--jina_api_key JINA_API_KEY] --model_path
61
+ MODEL_PATH [--temperature TEMPERATURE]
62
+ [--top_p TOP_P]
63
+ [--top_k_sampling TOP_K_SAMPLING]
64
+ [--repetition_penalty REPETITION_PENALTY]
65
+ [--max_tokens MAX_TOKENS]
66
+ --bing_subscription_key
67
+ BING_SUBSCRIPTION_KEY
68
+ [--bing_endpoint BING_ENDPOINT]
69
+ --cache_dir_base CACHE_DIR_BASE
70
+ --output_dir_base OUTPUT_DIR_BASE
71
+ [--is_exclude_urls]
72
+ search_o1_sum_all_webpage.py: error: the following arguments are required: --split, --model_path, --bing_subscription_key
deep_search/search_o1/run_eval.sh ADDED
@@ -0,0 +1,220 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ export http_proxy=http://127.0.0.1:7880
2
+ export https_proxy=http://127.0.0.1:7880
3
+
4
+ export CUDA_VISIBLE_DEVICES=6,7
5
+
6
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/reason_two_model_1.py \
7
+ # --dataset_name hotpotqa \
8
+ # --cache_dir_base cache/eval \
9
+ # --output_dir_base outputs/eval_qwen_2_5_7b \
10
+ # --split test \
11
+ # --max_search_limit 5 \
12
+ # --max_turn 10 \
13
+ # --top_k 5 \
14
+ # --max_doc_len 3000 \
15
+ # --subset_num 100 \
16
+ # --model_path "/opt/aps/workdir/output/checkpoint/qwen_2_5_7b_test" \
17
+ # --model_doc_reason_path "/capacity/userdata/models/DeepSeek-R1-Distill-Qwen-32" \
18
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
19
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
20
+ # --bing_endpoint "https://google.serper.dev/search"
21
+
22
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_eval.py \
23
+ # --dataset_name hotpotqa \
24
+ # --cache_dir_base cache_modify_extract_answer/eval \
25
+ # --output_dir_base outputs_modify_extract_answer/eval_qwen_2_5_32b_data_1217 \
26
+ # --split test \
27
+ # --max_search_limit 5 \
28
+ # --max_turn 10 \
29
+ # --top_k 5 \
30
+ # --max_doc_len 3000 \
31
+ # --subset_num 100 \
32
+ # --model_path "/opt/aps/workdir/output/checkpoint/qwen_2_5_32b_data_1217" \
33
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
34
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
35
+ # --bing_endpoint "https://google.serper.dev/search"
36
+
37
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_eval.py \
38
+ # --dataset_name hotpotqa \
39
+ # --cache_dir_base cache_modify_extract_answer/eval \
40
+ # --output_dir_base outputs_modify_extract_answer/eval_original_qwen_2_5_32b \
41
+ # --split test \
42
+ # --max_search_limit 5 \
43
+ # --max_turn 10 \
44
+ # --top_k 5 \
45
+ # --max_doc_len 3000 \
46
+ # --subset_num 100 \
47
+ # --model_path "/capacity/userdata/models/Qwen2.5-32B-Instruct" \
48
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
49
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
50
+ # --bing_endpoint "https://google.serper.dev/search"
51
+
52
+
53
+
54
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_eval.py \
55
+ # --dataset_name hotpotqa \
56
+ # --cache_dir_base cache_modify_extract_answer/eval \
57
+ # --output_dir_base outputs_modify_extract_answer/eval_qwen_7b_inst_data_1217 \
58
+ # --split test \
59
+ # --max_search_limit 5 \
60
+ # --max_turn 10 \
61
+ # --top_k 5 \
62
+ # --max_doc_len 3000 \
63
+ # --subset_num 100 \
64
+ # --model_path "/opt/aps/workdir/output/checkpoint/qwen_7b_inst_data_1217" \
65
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
66
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
67
+ # --bing_endpoint "https://google.serper.dev/search"
68
+
69
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_eval.py \
70
+ # --dataset_name hotpotqa \
71
+ # --cache_dir_base cache_modify_extract_answer/eval \
72
+ # --output_dir_base outputs_modify_extract_answer/eval_original_qwen_7b_inst \
73
+ # --split test \
74
+ # --max_search_limit 5 \
75
+ # --max_turn 10 \
76
+ # --top_k 5 \
77
+ # --max_doc_len 3000 \
78
+ # --subset_num 100 \
79
+ # --model_path "/opt/aps/workdir/home/.cache/modelscope/hub/models/qwen/Qwen2.5-7B-Instruct" \
80
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
81
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
82
+ # --bing_endpoint "https://google.serper.dev/search"
83
+
84
+ # qwen32b
85
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_eval.py \
86
+ # --dataset_name hotpotqa \
87
+ # --cache_dir_base cache_eval/eval_hotpotqa_dev_500 \
88
+ # --output_dir_base output_eval/eval_hopotqa_dev_500_original_qwen_32b_inst \
89
+ # --split test \
90
+ # --max_search_limit 5 \
91
+ # --max_turn 10 \
92
+ # --top_k 5 \
93
+ # --max_doc_len 3000 \
94
+ # --subset_num 500 \
95
+ # --model_path "/capacity/userdata/models/Qwen2.5-32B-Instruct" \
96
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
97
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
98
+ # --bing_endpoint "https://google.serper.dev/search"
99
+
100
+
101
+ datasets=('2wiki' 'hotpotqa' 'musique')
102
+
103
+
104
+ for dataset in "${datasets[@]}"; do
105
+ echo "Running evaluation for dataset: $dataset"
106
+
107
+ cache_dir_base="cache_eval/eval_${dataset}_dev_500_1"
108
+ output_dir_base="output_eval/eval_${dataset}_dev_500_qwen_32b_original_tokenizer_inst_data_1217_1"
109
+
110
+ mkdir -p "$cache_dir_base"
111
+ mkdir -p "$output_dir_base"
112
+
113
+ /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_eval.py \
114
+ --dataset_name "$dataset" \
115
+ --cache_dir_base "$cache_dir_base" \
116
+ --output_dir_base "$output_dir_base" \
117
+ --split test \
118
+ --max_search_limit 5 \
119
+ --max_turn 10 \
120
+ --top_k 5 \
121
+ --max_doc_len 3000 \
122
+ --subset_num 100 \
123
+ --model_path "/opt/aps/workdir/output/sft_use_original_tokenizer/qwen_32b_original_tokenizer_inst_data_1217" \
124
+ --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
125
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
126
+ --bing_endpoint "https://google.serper.dev/search"
127
+ done
128
+
129
+
130
+ /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_eval.py \
131
+ --dataset_name hotpotqa \
132
+ --cache_dir_base cache_eval/eval_hotpotqa_dev_500_1 \
133
+ --output_dir_base output_eval/eval_hopotqa_dev_500_qwen_32b_original_tokenizer_inst_data_1217_1 \
134
+ --split test \
135
+ --max_search_limit 5 \
136
+ --max_turn 10 \
137
+ --top_k 5 \
138
+ --max_doc_len 3000 \
139
+ --subset_num 500 \
140
+ --model_path "/opt/aps/workdir/output/sft_use_original_tokenizer/qwen_32b_original_tokenizer_inst_data_1217" \
141
+ --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
142
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
143
+ --bing_endpoint "https://google.serper.dev/search"
144
+
145
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_eval.py \
146
+ # --dataset_name hotpotqa \
147
+ # --cache_dir_base cache_eval/eval_hotpotqa_dev_500 \
148
+ # --output_dir_base output_eval/eval_hopotqa_dev_500_original_r1_32b \
149
+ # --split test \
150
+ # --max_search_limit 5 \
151
+ # --max_turn 10 \
152
+ # --top_k 5 \
153
+ # --max_doc_len 3000 \
154
+ # --subset_num 500 \
155
+ # --model_path "/capacity/userdata/models/DeepSeek-R1-Distill-Qwen-32" \
156
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
157
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
158
+ # --bing_endpoint "https://google.serper.dev/search"
159
+
160
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_eval.py \
161
+ # --dataset_name hotpotqa \
162
+ # --cache_dir_base cache_eval/eval_hotpotqa_dev_500 \
163
+ # --output_dir_base output_eval/eval_hopotqa_dev_500_original_qwen_7b_inst \
164
+ # --split test \
165
+ # --max_search_limit 5 \
166
+ # --max_turn 10 \
167
+ # --top_k 5 \
168
+ # --max_doc_len 3000 \
169
+ # --subset_num 500 \
170
+ # --model_path "/opt/aps/workdir/home/.cache/modelscope/hub/models/qwen/Qwen2.5-7B-Instruct" \
171
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
172
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
173
+ # --bing_endpoint "https://google.serper.dev/search"
174
+
175
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_eval.py \
176
+ # --dataset_name hotpotqa \
177
+ # --cache_dir_base cache_modify_extract_answer/eval \
178
+ # --output_dir_base outputs_modify_extract_answer/eval_original_qwq \
179
+ # --split test \
180
+ # --max_search_limit 5 \
181
+ # --max_turn 10 \
182
+ # --top_k 5 \
183
+ # --max_doc_len 3000 \
184
+ # --subset_num 100 \
185
+ # --model_path "/capacity/userdata/models/QwQ-32B-Preview" \
186
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
187
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
188
+ # --bing_endpoint "https://google.serper.dev/search"
189
+
190
+
191
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_eval.py \
192
+ # --dataset_name hotpotqa \
193
+ # --cache_dir_base cache/eval \
194
+ # --output_dir_base outputs/eval_llama3_8b \
195
+ # --split test \
196
+ # --max_search_limit 5 \
197
+ # --max_turn 10 \
198
+ # --top_k 5 \
199
+ # --max_doc_len 3000 \
200
+ # --subset_num 100 \
201
+ # --model_path "/opt/aps/workdir/output/checkpoint/llama3_8b_data_1217" \
202
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
203
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
204
+ # --bing_endpoint "https://google.serper.dev/search"
205
+
206
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_eval.py \
207
+ # --dataset_name hotpotqa \
208
+ # --cache_dir_base cache/eval \
209
+ # --output_dir_base outputs/eval_original_llama3_8b \
210
+ # --split test \
211
+ # --max_search_limit 5 \
212
+ # --max_turn 10 \
213
+ # --top_k 5 \
214
+ # --max_doc_len 3000 \
215
+ # --subset_num 100 \
216
+ # --model_path "/capacity/userdata/models/QwQ-32B-Preview" \
217
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
218
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
219
+ # --bing_endpoint "https://google.serper.dev/search"
220
+
deep_search/search_o1/run_eval_benchmark_2wiki.sh ADDED
@@ -0,0 +1,45 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ export CUDA_VISIBLE_DEVICES=0,1
2
+ /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
3
+ --dataset_name eval \
4
+ --cache_dir_base cache_eval_sum_all_webpage_wiki \
5
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:12665#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_merged_new_instruction_1161_1301_sft_remove_exp_2462/228/eval_wiki \
6
+ --split test \
7
+ --max_search_limit 10 \
8
+ --max_turn 10 \
9
+ --top_k 10 \
10
+ --max_doc_len 5000 \
11
+ --model_path "/share/project/sunshuang/deep_search/output/JOB:12665#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_merged_new_instruction_1161_1301_sft_remove_exp_2462/checkpoint-228" \
12
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
13
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:12665#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_merged_new_instruction_1161_1301_sft_remove_exp_2462/228/eval_wiki/inf.log 2>&1
14
+
15
+
16
+ /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
17
+ --dataset_name eval \
18
+ --cache_dir_base cache_eval_sum_all_webpage_wiki \
19
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:9578#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_new_instruction_2k_sft_remove_exp/73/eval_wiki \
20
+ --split test \
21
+ --max_search_limit 10 \
22
+ --max_turn 10 \
23
+ --top_k 10 \
24
+ --max_doc_len 5000 \
25
+ --model_path "/share/project/sunshuang/deep_search/output/JOB:9578#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_new_instruction_2k_sft_remove_exp/checkpoint-73" \
26
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
27
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:9578#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_new_instruction_2k_sft_remove_exp/73/eval_wiki/inf.log 2>&1
28
+
29
+
30
+
31
+ # /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:15076#LR:1e-5#BASE:Qwen2.5-7B-Instruct#TOKEN:Qwen2.5-7B-Instruct#BSZ:2#ACC:4_new_instruction_2k_sft_remove_exp/91/eval_wiki
32
+
33
+ /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
34
+ --dataset_name eval \
35
+ --cache_dir_base cache_eval_sum_all_webpage_wiki \
36
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:15076#LR:1e-5#BASE:Qwen2.5-7B-Instruct#TOKEN:Qwen2.5-7B-Instruct#BSZ:2#ACC:4_new_instruction_2k_sft_remove_exp/91/eval_wiki \
37
+ --split test \
38
+ --max_search_limit 10 \
39
+ --max_turn 10 \
40
+ --top_k 10 \
41
+ --max_doc_len 5000 \
42
+ --model_path "/share/project/sunshuang/deep_search/output/JOB:15076#LR:1e-5#BASE:Qwen2.5-7B-Instruct#TOKEN:Qwen2.5-7B-Instruct#BSZ:2#ACC:4_new_instruction_2k_sft_remove_exp/checkpoint-91" \
43
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
44
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:15076#LR:1e-5#BASE:Qwen2.5-7B-Instruct#TOKEN:Qwen2.5-7B-Instruct#BSZ:2#ACC:4_new_instruction_2k_sft_remove_exp/91/eval_wiki/inf.log 2>&1
45
+
deep_search/search_o1/run_eval_benchmark_frames.sh ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ export CUDA_VISIBLE_DEVICES=2,3
3
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
4
+ --dataset_name frames \
5
+ --cache_dir_base cache_eval_sum_all_webpage \
6
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval/outputs_sum_all_webpage_qwq_new_prompt_frames \
7
+ --split test \
8
+ --max_search_limit 10 \
9
+ --max_turn 10 \
10
+ --top_k 10 \
11
+ --max_doc_len 5000 \
12
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
13
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
14
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval/outputs_sum_all_webpage_qwq_new_prompt_frames/inf.log 2>&1 &
15
+
16
+
17
+ export CUDA_VISIBLE_DEVICES=2,3
18
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
19
+ --dataset_name frames \
20
+ --cache_dir_base cache_eval_sum_all_webpage_frames \
21
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/91/frames \
22
+ --split test \
23
+ --max_search_limit 10 \
24
+ --max_turn 10 \
25
+ --top_k 10 \
26
+ --max_doc_len 5000 \
27
+ --model_path "/share/project/sunshuang/deep_search/output/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/checkpoint-91" \
28
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
29
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/91/frames/inf.log 2>&1 &
30
+
31
+
deep_search/search_o1/run_eval_benchmark_gaia.sh ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ export CUDA_VISIBLE_DEVICES=0,1,2,3
3
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
4
+ --dataset_name gaia \
5
+ --cache_dir_base cache_eval_sum_all_webpage_1w \
6
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/91/gaia_1w \
7
+ --split test \
8
+ --max_search_limit 10 \
9
+ --max_turn 10 \
10
+ --top_k 10 \
11
+ --max_doc_len 5000 \
12
+ --model_path "/share/project/sunshuang/deep_search/output/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/checkpoint-91" \
13
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
14
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/91/gaia_1w/inf.log 2>&1 &
15
+
16
+
17
+ export CUDA_VISIBLE_DEVICES=2,3
18
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
19
+ --dataset_name frames \
20
+ --cache_dir_base cache_eval_sum_all_webpage_frames \
21
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/91/frames \
22
+ --split test \
23
+ --max_search_limit 10 \
24
+ --max_turn 10 \
25
+ --top_k 10 \
26
+ --max_doc_len 5000 \
27
+ --model_path "/share/project/sunshuang/deep_search/output/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/checkpoint-91" \
28
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
29
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/91/frames/inf.log 2>&1 &
30
+
31
+
deep_search/search_o1/run_eval_benchmark_train.sh ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ export CUDA_VISIBLE_DEVICES=0,1,6,7
2
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
3
+ --dataset_name merged_data_1174_test \
4
+ --cache_dir_base cache_eval_sum_all_webpage \
5
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/91/merged_data_1174_test \
6
+ --split test \
7
+ --max_search_limit 10 \
8
+ --max_turn 10 \
9
+ --top_k 10 \
10
+ --max_doc_len 5000 \
11
+ --model_path "/share/project/sunshuang/deep_search/output/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/checkpoint-91" \
12
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
13
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/91/merged_data_1174_test/inf.log 2>&1 &
14
+
15
+
deep_search/search_o1/run_eval_sft.sh ADDED
@@ -0,0 +1,246 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ export CUDA_VISIBLE_DEVICES=0,1
2
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
3
+ --dataset_name eval \
4
+ --cache_dir_base cache_eval_sum_all_webpage_1w \
5
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwq/gaia \
6
+ --split test \
7
+ --max_search_limit 10 \
8
+ --max_turn 10 \
9
+ --top_k 10 \
10
+ --max_doc_len 5000 \
11
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
12
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
13
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwq/gaia/inf.log 2>&1 &
14
+
15
+
16
+ export CUDA_VISIBLE_DEVICES=2,3
17
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
18
+ --dataset_name eval \
19
+ --cache_dir_base cache_eval_sum_all_webpage_1w \
20
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen7b_inst_sft_871/41/eval \
21
+ --split test \
22
+ --max_search_limit 10 \
23
+ --max_turn 10 \
24
+ --top_k 10 \
25
+ --max_doc_len 5000 \
26
+ --model_path "/share/project/sunshuang/deep_search/q1/JOB:2569#LR:1e-5#BASE:Qwen2.5-7B-Instruct#TOKEN:Qwen2.5-7B-Instruct#BSZ:2#ACC:4_new_800/checkpoint-41" \
27
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
28
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen7b_inst_sft_871/41/eval/inf.log 2>&1 &
29
+
30
+
31
+ export CUDA_VISIBLE_DEVICES=4,5
32
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
33
+ --dataset_name frames \
34
+ --cache_dir_base cache_eval_sum_all_webpage_1w \
35
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen7b_inst_sft_871/41/frames \
36
+ --split test \
37
+ --max_search_limit 10 \
38
+ --max_turn 10 \
39
+ --top_k 10 \
40
+ --max_doc_len 5000 \
41
+ --model_path "/share/project/sunshuang/deep_search/q1/JOB:2569#LR:1e-5#BASE:Qwen2.5-7B-Instruct#TOKEN:Qwen2.5-7B-Instruct#BSZ:2#ACC:4_new_800/checkpoint-41" \
42
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
43
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen7b_inst_sft_871/41/frames/inf.log 2>&1 &
44
+
45
+
46
+ export CUDA_VISIBLE_DEVICES=6,7
47
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
48
+ --dataset_name gaia \
49
+ --cache_dir_base cache_eval_sum_all_webpage_1w \
50
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen7b_inst_sft_871/41/gaia \
51
+ --split test \
52
+ --max_search_limit 10 \
53
+ --max_turn 10 \
54
+ --top_k 10 \
55
+ --max_doc_len 5000 \
56
+ --model_path "/share/project/sunshuang/deep_search/q1/JOB:2569#LR:1e-5#BASE:Qwen2.5-7B-Instruct#TOKEN:Qwen2.5-7B-Instruct#BSZ:2#ACC:4_new_800/checkpoint-41" \
57
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
58
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen7b_inst_sft_871/41/gaia/inf.log 2>&1 &
59
+
60
+
61
+
62
+ export CUDA_VISIBLE_DEVICES=2,3
63
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
64
+ --dataset_name eval \
65
+ --cache_dir_base cache_eval_sum_all_webpage_1w \
66
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen7b_inst_sft_871/78/eval \
67
+ --split test \
68
+ --max_search_limit 10 \
69
+ --max_turn 10 \
70
+ --top_k 10 \
71
+ --max_doc_len 5000 \
72
+ --model_path "/share/project/sunshuang/deep_search/q1/JOB:2569#LR:1e-5#BASE:Qwen2.5-7B-Instruct#TOKEN:Qwen2.5-7B-Instruct#BSZ:2#ACC:4_new_800/checkpoint-78" \
73
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
74
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen7b_inst_sft_871/78/eval/inf.log 2>&1 &
75
+
76
+
77
+ export CUDA_VISIBLE_DEVICES=4,5
78
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
79
+ --dataset_name frames \
80
+ --cache_dir_base cache_eval_sum_all_webpage_1w \
81
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen7b_inst_sft_871/78/frames \
82
+ --split test \
83
+ --max_search_limit 10 \
84
+ --max_turn 10 \
85
+ --top_k 10 \
86
+ --max_doc_len 5000 \
87
+ --model_path "/share/project/sunshuang/deep_search/q1/JOB:2569#LR:1e-5#BASE:Qwen2.5-7B-Instruct#TOKEN:Qwen2.5-7B-Instruct#BSZ:2#ACC:4_new_800/checkpoint-78" \
88
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
89
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen7b_inst_sft_871/78/frames/inf.log 2>&1 &
90
+
91
+
92
+ export CUDA_VISIBLE_DEVICES=6,7
93
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
94
+ --dataset_name gaia \
95
+ --cache_dir_base cache_eval_sum_all_webpage_1w \
96
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen7b_inst_sft_871/78/gaia \
97
+ --split test \
98
+ --max_search_limit 10 \
99
+ --max_turn 10 \
100
+ --top_k 10 \
101
+ --max_doc_len 5000 \
102
+ --model_path "/share/project/sunshuang/deep_search/q1/JOB:2569#LR:1e-5#BASE:Qwen2.5-7B-Instruct#TOKEN:Qwen2.5-7B-Instruct#BSZ:2#ACC:4_new_800/checkpoint-78" \
103
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
104
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen7b_inst_sft_871/78/gaia/inf.log 2>&1 &
105
+
106
+
107
+
108
+
109
+
110
+ export CUDA_VISIBLE_DEVICES=2,3
111
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
112
+ --dataset_name eval \
113
+ --cache_dir_base cache_eval_sum_all_webpage_1w \
114
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen_32b/eval \
115
+ --split test \
116
+ --max_search_limit 10 \
117
+ --max_turn 10 \
118
+ --top_k 10 \
119
+ --max_doc_len 5000 \
120
+ --model_path "/share/project/zhipengchen/model/Qwen2.5-32B-Instruct" \
121
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
122
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen_32b/eval/inf.log 2>&1 &
123
+
124
+
125
+ export CUDA_VISIBLE_DEVICES=4,5
126
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
127
+ --dataset_name frames \
128
+ --cache_dir_base cache_eval_sum_all_webpage_1w \
129
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen_32b/frames \
130
+ --split test \
131
+ --max_search_limit 10 \
132
+ --max_turn 10 \
133
+ --top_k 10 \
134
+ --max_doc_len 5000 \
135
+ --model_path "/share/project/zhipengchen/model/Qwen2.5-32B-Instruct" \
136
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
137
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen_32b/frames/inf.log 2>&1 &
138
+
139
+
140
+ export CUDA_VISIBLE_DEVICES=6,7
141
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
142
+ --dataset_name gaia \
143
+ --cache_dir_base cache_eval_sum_all_webpage_1w \
144
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen_32b/gaia \
145
+ --split test \
146
+ --max_search_limit 10 \
147
+ --max_turn 10 \
148
+ --top_k 10 \
149
+ --max_doc_len 5000 \
150
+ --model_path "/share/project/zhipengchen/model/Qwen2.5-32B-Instruct" \
151
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
152
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwen_32b/gaia/inf.log 2>&1 &
153
+
154
+ ######
155
+ # r1 32b
156
+ export CUDA_VISIBLE_DEVICES=2,3
157
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
158
+ --dataset_name eval \
159
+ --cache_dir_base cache_eval_sum_all_webpage \
160
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/dpsk_qwen_32b/eval \
161
+ --split test \
162
+ --max_search_limit 10 \
163
+ --max_turn 10 \
164
+ --top_k 10 \
165
+ --max_doc_len 5000 \
166
+ --model_path "/share/project/zhipengchen/model/DeepSeek-R1-Distill-Qwen-32B" \
167
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
168
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/dpsk_qwen_32b/eval/inf.log 2>&1 &
169
+
170
+
171
+ export CUDA_VISIBLE_DEVICES=4,5
172
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
173
+ --dataset_name frames \
174
+ --cache_dir_base cache_eval_sum_all_webpage \
175
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/dpsk_qwen_32b/frames \
176
+ --split test \
177
+ --max_search_limit 10 \
178
+ --max_turn 10 \
179
+ --top_k 10 \
180
+ --max_doc_len 5000 \
181
+ --model_path "/share/project/zhipengchen/model/DeepSeek-R1-Distill-Qwen-32B" \
182
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
183
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/dpsk_qwen_32b/frames/inf.log 2>&1 &
184
+
185
+
186
+ export CUDA_VISIBLE_DEVICES=6,7
187
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
188
+ --dataset_name gaia \
189
+ --cache_dir_base cache_eval_sum_all_webpage \
190
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/dpsk_qwen_32b/gaia \
191
+ --split test \
192
+ --max_search_limit 10 \
193
+ --max_turn 10 \
194
+ --top_k 10 \
195
+ --max_doc_len 5000 \
196
+ --model_path "/share/project/zhipengchen/model/DeepSeek-R1-Distill-Qwen-32B" \
197
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
198
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/dpsk_qwen_32b/gaia/inf.log 2>&1 &
199
+
200
+
201
+ ################
202
+ # /share/project/sunshuang/deep_search/output/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/checkpoint-91
203
+ export CUDA_VISIBLE_DEVICES=2,3
204
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
205
+ --dataset_name eval \
206
+ --cache_dir_base cache_eval_sum_all_webpage \
207
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwq_1.1k_91/eval \
208
+ --split test \
209
+ --max_search_limit 10 \
210
+ --max_turn 10 \
211
+ --top_k 10 \
212
+ --max_doc_len 5000 \
213
+ --model_path "/share/project/sunshuang/deep_search/output/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/checkpoint-91" \
214
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
215
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwq_1.1k_91/eval/inf.log 2>&1 &
216
+
217
+
218
+ export CUDA_VISIBLE_DEVICES=4,5
219
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
220
+ --dataset_name frames \
221
+ --cache_dir_base cache_eval_sum_all_webpage \
222
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwq_1.1k_91/frames \
223
+ --split test \
224
+ --max_search_limit 10 \
225
+ --max_turn 10 \
226
+ --top_k 10 \
227
+ --max_doc_len 5000 \
228
+ --model_path "/share/project/sunshuang/deep_search/output/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/checkpoint-91" \
229
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
230
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwq_1.1k_91/frames/inf.log 2>&1 &
231
+
232
+
233
+ export CUDA_VISIBLE_DEVICES=6,7
234
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
235
+ --dataset_name gaia \
236
+ --cache_dir_base cache_eval_sum_all_webpage \
237
+ --output_dir_base /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwq_1.1k_91/gaia \
238
+ --split test \
239
+ --max_search_limit 10 \
240
+ --max_turn 10 \
241
+ --top_k 10 \
242
+ --max_doc_len 5000 \
243
+ --model_path "/share/project/sunshuang/deep_search/output/JOB:11783#LR:1e-5#BASE:QwQ-32B#TOKEN:QwQ-32B#BSZ:2#ACC:4_selected_data_1174_sft/checkpoint-91" \
244
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
245
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/sunshuang/deep_search/search_o1/output/output_eval_sft/qwq_1.1k_91/gaia/inf.log 2>&1 &
246
+
deep_search/search_o1/run_eval_sft_3.log ADDED
@@ -0,0 +1 @@
 
 
1
+ nohup: ignoring input
deep_search/search_o1/run_gen_data (1).sh ADDED
@@ -0,0 +1,53 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ export http_proxy=http://127.0.0.1:7880
2
+ export https_proxy=http://127.0.0.1:7880
3
+
4
+ export CUDA_VISIBLE_DEVICES=0,1,2,3
5
+
6
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_5.py \
7
+ # --dataset_name hotpotqa \
8
+ # --cache_dir_base cache_gen_data_hotpot_qa \
9
+ # --output_dir_base outputs_gen_data_hotpot_qa \
10
+ # --split test \
11
+ # --max_search_limit 5 \
12
+ # --max_turn 10 \
13
+ # --top_k 5 \
14
+ # --max_doc_len 3000 \
15
+ # --batch_size 5000 \
16
+ # --model_path "/capacity/userdata/models/DeepSeek-R1-Distill-Qwen-32" \
17
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
18
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
19
+ # --bing_endpoint "https://google.serper.dev/search"
20
+
21
+
22
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python -u scripts/run_search_o1_5.py \
23
+ # --dataset_name musique \
24
+ # --cache_dir_base /opt/aps/workdir/sunshuang/search_o1/cache_reason_two_model/eval_reason_two_model \
25
+ # --output_dir_base outputs_gen_data_musique_qa_by_r1_32b \
26
+ # --split train \
27
+ # --max_search_limit 5 \
28
+ # --max_turn 10 \
29
+ # --top_k 5 \
30
+ # --max_doc_len 3000 \
31
+ # --batch_size 2000 \
32
+ # --batch_nums 2 \
33
+ # --model_path "/capacity/userdata/models/DeepSeek-R1-Distill-Qwen-32" \
34
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
35
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
36
+ # --bing_endpoint "https://google.serper.dev/search" > logs/gen_data_musique_qa_by_r1_32b.log 2>&1
37
+
38
+ /opt/aps/workdir/miniforge3/envs/search_o1/bin/python -u scripts/run_search_o1_5_with_start_id.py \
39
+ --dataset_name musique \
40
+ --cache_dir_base /opt/aps/workdir/sunshuang/search_o1/cache_reason_two_model/eval_reason_two_model \
41
+ --output_dir_base outputs_gen_data_musique_qa_by_r1_32b_batch2_3 \
42
+ --split train \
43
+ --max_search_limit 5 \
44
+ --max_turn 10 \
45
+ --top_k 5 \
46
+ --max_doc_len 3000 \
47
+ --batch_size 2000 \
48
+ --start_batch_id 3 \
49
+ --batch_nums 1 \
50
+ --model_path "/capacity/userdata/models/DeepSeek-R1-Distill-Qwen-32" \
51
+ --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
52
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
53
+ --bing_endpoint "https://google.serper.dev/search" > logs/gen_data_musiqueqa_by_r1_32b_batch2_3.log 2>&1
deep_search/search_o1/run_gen_data_2.sh ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ export http_proxy=http://127.0.0.1:7880
2
+ export https_proxy=http://127.0.0.1:7880
3
+
4
+ export CUDA_VISIBLE_DEVICES=4,5,6,7
5
+
6
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_5.py \
7
+ # --dataset_name hotpotqa \
8
+ # --cache_dir_base cache_eval_hotpotqa \
9
+ # --output_dir_base outputs_eval_hotpotqa \
10
+ # --split test \
11
+ # --max_search_limit 5 \
12
+ # --max_turn 10 \
13
+ # --top_k 5 \
14
+ # --max_doc_len 3000 \
15
+ # --batch_size 5000 \
16
+ # --subset_num 500 \
17
+ # --model_path "/opt/aps/workdir/output/checkpoint/debug_1" \
18
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
19
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
20
+ # --bing_endpoint "https://google.serper.dev/search"
21
+
22
+ # /opt/aps/workdir/miniforge3/envs/search_o1/bin/python scripts/run_search_o1_5.py \
23
+ # --dataset_name hotpotqa \
24
+ # --cache_dir_base cache_eval_hotpotqa_1 \
25
+ # --output_dir_base outputs_eval_hotpotqa_1 \
26
+ # --split test \
27
+ # --max_search_limit 5 \
28
+ # --max_turn 10 \
29
+ # --top_k 5 \
30
+ # --max_doc_len 3000 \
31
+ # --batch_size 2000 \
32
+ # --subset_num 100 \
33
+ # --model_path "/capacity/userdata/models/Qwen2.5-7B" \
34
+ # --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
35
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
36
+ # --bing_endpoint "https://google.serper.dev/search"
37
+
38
+
39
+ /opt/aps/workdir/miniforge3/envs/search_o1/bin/python -u scripts/run_search_o1_5_with_start_id.py \
40
+ --dataset_name musique \
41
+ --cache_dir_base /opt/aps/workdir/sunshuang/search_o1/cache_reason_two_model/eval_reason_two_model \
42
+ --output_dir_base outputs_gen_data_musique_qa_by_r1_32b_batch4_5 \
43
+ --split train \
44
+ --max_search_limit 5 \
45
+ --max_turn 10 \
46
+ --top_k 5 \
47
+ --max_doc_len 3000 \
48
+ --batch_size 2000 \
49
+ --start_batch_id 4 \
50
+ --batch_nums 2 \
51
+ --model_path "/capacity/userdata/models/DeepSeek-R1-Distill-Qwen-32" \
52
+ --jina_api_key "jina_04d684ee4cc54d2ebe7c43bb7ad4aff0qlkdZGwm14NFBtm5BDkgK9KNf6vQ" \
53
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
54
+ --bing_endpoint "https://google.serper.dev/search" > logs/gen_data_musiqueqa_by_r1_32b_batch4_5.log 2>&1
deep_search/search_o1/run_gen_data_acc_ratios0.sh ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # worker 1
2
+ export CUDA_VISIBLE_DEVICES=0,1
3
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
4
+ --dataset_name acc_ratios_0_total_1608_rollout_split1 \
5
+ --cache_dir_base cache_eval_sum_all_webpage \
6
+ --output_dir_base output/output_sum_all_webpage_gen_data/acc_ratios_0_total_1608_rollout_split1 \
7
+ --split gen \
8
+ --max_search_limit 10 \
9
+ --max_turn 10 \
10
+ --top_k 10 \
11
+ --rollout_num 10 \
12
+ --max_doc_len 5000 \
13
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
14
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
15
+ --bing_endpoint "https://google.serper.dev/search" > output/output_sum_all_webpage_gen_data/acc_ratios_0_total_1608_rollout_split1/inf.log 2>&1 &
16
+
17
+
18
+ export CUDA_VISIBLE_DEVICES=2,3
19
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
20
+ --dataset_name acc_ratios_0_total_1608_rollout_split2 \
21
+ --cache_dir_base cache_eval_sum_all_webpage \
22
+ --output_dir_base output/output_sum_all_webpage_gen_data/acc_ratios_0_total_1608_rollout_split2 \
23
+ --split gen \
24
+ --max_search_limit 10 \
25
+ --max_turn 10 \
26
+ --top_k 10 \
27
+ --rollout_num 10 \
28
+ --max_doc_len 5000 \
29
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
30
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
31
+ --bing_endpoint "https://google.serper.dev/search" > output/output_sum_all_webpage_gen_data/acc_ratios_0_total_1608_rollout_split2/inf.log 2>&1 &
32
+
33
+
34
+ export CUDA_VISIBLE_DEVICES=2,3
35
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
36
+ --dataset_name acc_ratios_0_total_1608_rollout_split3 \
37
+ --cache_dir_base cache_eval_sum_all_webpage \
38
+ --output_dir_base output/output_sum_all_webpage_gen_data/acc_ratios_0_total_1608_rollout_split3 \
39
+ --split gen \
40
+ --max_search_limit 10 \
41
+ --max_turn 10 \
42
+ --top_k 10 \
43
+ --rollout_num 10 \
44
+ --max_doc_len 5000 \
45
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
46
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
47
+ --bing_endpoint "https://google.serper.dev/search" > output/output_sum_all_webpage_gen_data/acc_ratios_0_total_1608_rollout_split3/inf.log 2>&1 &
48
+
49
+ export CUDA_VISIBLE_DEVICES=6,7
50
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
51
+ --dataset_name acc_ratios_0_total_1608_rollout_split4 \
52
+ --cache_dir_base cache_eval_sum_all_webpage \
53
+ --output_dir_base output/output_sum_all_webpage_gen_data/acc_ratios_0_total_1608_rollout_split4 \
54
+ --split gen \
55
+ --max_search_limit 10 \
56
+ --max_turn 10 \
57
+ --top_k 10 \
58
+ --rollout_num 10 \
59
+ --max_doc_len 5000 \
60
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
61
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
62
+ --bing_endpoint "https://google.serper.dev/search" > output/output_sum_all_webpage_gen_data/acc_ratios_0_total_1608_rollout_split4/inf.log 2>&1 &
deep_search/search_o1/run_gen_data_for_chinese_simple_qa.sh ADDED
@@ -0,0 +1,80 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ export http_proxy=http://127.0.0.1:7880
2
+ export https_proxy=http://127.0.0.1:7880
3
+
4
+ export CUDA_VISIBLE_DEVICES=4,5,6,7
5
+
6
+ /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
7
+ --dataset_name chinese_simpleqa \
8
+ --cache_dir_base cache \
9
+ --output_dir_base output/outputs_sum_all_webpage_chinese_simpleqa500_qwq \
10
+ --split test \
11
+ --max_search_limit 10 \
12
+ --max_turn 10 \
13
+ --top_k 10 \
14
+ --max_doc_len 5000 \
15
+ --subset_num 500 \
16
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
17
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
18
+ --bing_endpoint "https://google.serper.dev/search"
19
+
20
+ # /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
21
+ # --dataset_name chinese_simpleqa \
22
+ # --cache_dir_base cache_simpleqa_exurls_qwq \
23
+ # --output_dir_base output/outputs_sum_all_webpage_simpleqa500_ex_urls_qwq \
24
+ # --split test \
25
+ # --max_search_limit 10 \
26
+ # --max_turn 10 \
27
+ # --top_k 10 \
28
+ # --max_doc_len 5000 \
29
+ # --subset_num 500 \
30
+ # --is_exclude_urls \
31
+ # --model_path "/share/project/zhipengchen/model/QwQ-32B" \
32
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
33
+ # --bing_endpoint "https://google.serper.dev/search"
34
+
35
+
36
+
37
+
38
+ /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
39
+ --dataset_name chinese_simpleqa \
40
+ --cache_dir_base cache \
41
+ --output_dir_base output/outputs_sum_all_webpage_chinese_simpleqa500_ckpt176 \
42
+ --split test \
43
+ --max_search_limit 10 \
44
+ --max_turn 10 \
45
+ --top_k 10 \
46
+ --max_doc_len 5000 \
47
+ --subset_num 500 \
48
+ --model_path "/share/project/sunshuang/deep_research/search_o1/model/checkpoint-176" \
49
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
50
+ --bing_endpoint "https://google.serper.dev/search"
51
+
52
+
53
+ /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
54
+ --dataset_name simpleqa \
55
+ --cache_dir_base cache_simpleqa_exurls_ckpt176 \
56
+ --output_dir_base output/outputs_sum_all_webpage_simpleqa500_ex_urls_ckpt176 \
57
+ --split test \
58
+ --max_search_limit 10 \
59
+ --max_turn 10 \
60
+ --top_k 10 \
61
+ --max_doc_len 5000 \
62
+ --subset_num 500 \
63
+ --is_exclude_urls \
64
+ --model_path "/share/project/sunshuang/deep_research/search_o1/model/checkpoint-176" \
65
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
66
+ --bing_endpoint "https://google.serper.dev/search"
67
+
68
+ /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage.py \
69
+ --dataset_name hotpotqa \
70
+ --cache_dir_base cache \
71
+ --output_dir_base output/outputs_sum_all_webpage_hotpotqa500_ckpt176 \
72
+ --split test \
73
+ --max_search_limit 10 \
74
+ --max_turn 10 \
75
+ --top_k 10 \
76
+ --max_doc_len 5000 \
77
+ --subset_num 500 \
78
+ --model_path "/share/project/sunshuang/deep_research/search_o1/model/checkpoint-176" \
79
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
80
+ --bing_endpoint "https://google.serper.dev/search"
deep_search/search_o1/run_gen_data_selected_2k.sh ADDED
@@ -0,0 +1,216 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # worker 1
2
+ export CUDA_VISIBLE_DEVICES=0,1
3
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
4
+ --dataset_name selected_2k_part_1 \
5
+ --cache_dir_base cache_eval_sum_all_webpage \
6
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_selected_2k_part_1 \
7
+ --split gen \
8
+ --max_search_limit 10 \
9
+ --max_turn 10 \
10
+ --top_k 10 \
11
+ --rollout_num 3 \
12
+ --max_doc_len 5000 \
13
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
14
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
15
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/selected_2k_part_1_1.log 2>&1 &
16
+
17
+
18
+ export CUDA_VISIBLE_DEVICES=2,3
19
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
20
+ --dataset_name selected_2k_part_2 \
21
+ --cache_dir_base cache_eval_sum_all_webpage \
22
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_selected_2k_part_2 \
23
+ --split gen \
24
+ --max_search_limit 10 \
25
+ --max_turn 10 \
26
+ --top_k 10 \
27
+ --rollout_num 1 \
28
+ --max_doc_len 5000 \
29
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
30
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
31
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/selected_2k_part_2_1.log 2>&1 &
32
+
33
+
34
+ export CUDA_VISIBLE_DEVICES=6,7
35
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
36
+ --dataset_name selected_2k_part_3 \
37
+ --cache_dir_base cache_eval_sum_all_webpage \
38
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_selected_2k_part_3 \
39
+ --split gen \
40
+ --max_search_limit 10 \
41
+ --max_turn 10 \
42
+ --top_k 10 \
43
+ --rollout_num 3 \
44
+ --max_doc_len 5000 \
45
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
46
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
47
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/selected_2k_part_3_1.log 2>&1 &
48
+
49
+
50
+ export CUDA_VISIBLE_DEVICES=4,5
51
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
52
+ --dataset_name selected_2k_part_4 \
53
+ --cache_dir_base cache_eval_sum_all_webpage \
54
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_selected_2k_part_4 \
55
+ --split gen \
56
+ --max_search_limit 10 \
57
+ --max_turn 10 \
58
+ --top_k 10 \
59
+ --rollout_num 3 \
60
+ --max_doc_len 5000 \
61
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
62
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
63
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/selected_2k_part_4_1.log 2>&1 &
64
+
65
+
66
+ export CUDA_VISIBLE_DEVICES=6,7
67
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
68
+ --dataset_name selected_2k_part_5 \
69
+ --cache_dir_base cache_eval_sum_all_webpage \
70
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_selected_2k_part_5 \
71
+ --split gen \
72
+ --max_search_limit 10 \
73
+ --max_turn 10 \
74
+ --top_k 10 \
75
+ --rollout_num 3 \
76
+ --max_doc_len 5000 \
77
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
78
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
79
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/selected_2k_part_5_1.log 2>&1 &
80
+
81
+
82
+
83
+ export CUDA_VISIBLE_DEVICES=0,1
84
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
85
+ --dataset_name selected_2k_part_6 \
86
+ --cache_dir_base cache_eval_sum_all_webpage \
87
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_selected_2k_part_6 \
88
+ --split gen \
89
+ --max_search_limit 10 \
90
+ --max_turn 10 \
91
+ --top_k 10 \
92
+ --rollout_num 3 \
93
+ --max_doc_len 5000 \
94
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
95
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
96
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/selected_2k_part_6_1.log 2>&1 &
97
+
98
+
99
+
100
+ export CUDA_VISIBLE_DEVICES=2,3
101
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
102
+ --dataset_name selected_2k_part_7 \
103
+ --cache_dir_base cache_eval_sum_all_webpage \
104
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_selected_2k_part_7 \
105
+ --split gen \
106
+ --max_search_limit 10 \
107
+ --max_turn 10 \
108
+ --top_k 10 \
109
+ --rollout_num 2 \
110
+ --max_doc_len 5000 \
111
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
112
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
113
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/selected_2k_part_7_1.log 2>&1 &
114
+
115
+
116
+ export CUDA_VISIBLE_DEVICES=4,5
117
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
118
+ --dataset_name selected_2k_part_8 \
119
+ --cache_dir_base cache_eval_sum_all_webpage \
120
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_selected_2k_part_8 \
121
+ --split gen \
122
+ --max_search_limit 10 \
123
+ --max_turn 10 \
124
+ --top_k 10 \
125
+ --rollout_num 2 \
126
+ --max_doc_len 5000 \
127
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
128
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
129
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/selected_2k_part_8_1.log 2>&1 &
130
+
131
+
132
+
133
+ export CUDA_VISIBLE_DEVICES=6,7
134
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
135
+ --dataset_name selected_2k_part_9 \
136
+ --cache_dir_base cache_eval_sum_all_webpage \
137
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_selected_2k_part_9 \
138
+ --split gen \
139
+ --max_search_limit 10 \
140
+ --max_turn 10 \
141
+ --top_k 10 \
142
+ --rollout_num 7 \
143
+ --max_doc_len 5000 \
144
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
145
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
146
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/selected_2k_part_9_1.log 2>&1 &
147
+
148
+
149
+
150
+
151
+
152
+
153
+ export CUDA_VISIBLE_DEVICES=0,1
154
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
155
+ --dataset_name selected_2k_part_10 \
156
+ --cache_dir_base cache_eval_sum_all_webpage \
157
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_selected_2k_part_10 \
158
+ --split gen \
159
+ --max_search_limit 10 \
160
+ --max_turn 10 \
161
+ --top_k 10 \
162
+ --rollout_num 3 \
163
+ --max_doc_len 5000 \
164
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
165
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
166
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/selected_2k_part_10_1.log 2>&1 &
167
+
168
+
169
+
170
+ export CUDA_VISIBLE_DEVICES=2,3
171
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
172
+ --dataset_name selected_2k_part_11 \
173
+ --cache_dir_base cache_eval_sum_all_webpage \
174
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_selected_2k_part_11 \
175
+ --split gen \
176
+ --max_search_limit 10 \
177
+ --max_turn 10 \
178
+ --top_k 10 \
179
+ --rollout_num 2 \
180
+ --max_doc_len 5000 \
181
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
182
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
183
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/selected_2k_part_11_1.log 2>&1 &
184
+
185
+
186
+ export CUDA_VISIBLE_DEVICES=4,5
187
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
188
+ --dataset_name selected_2k_part_12 \
189
+ --cache_dir_base cache_eval_sum_all_webpage \
190
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_selected_2k_part_12 \
191
+ --split gen \
192
+ --max_search_limit 10 \
193
+ --max_turn 10 \
194
+ --top_k 10 \
195
+ --rollout_num 2 \
196
+ --max_doc_len 5000 \
197
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
198
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
199
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/selected_2k_part_12_1.log 2>&1 &
200
+
201
+
202
+
203
+ export CUDA_VISIBLE_DEVICES=2,3
204
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
205
+ --dataset_name selected_2k_part_13 \
206
+ --cache_dir_base cache_eval_sum_all_webpage \
207
+ --output_dir_base output/output_sum_all_webpage_gen_data/outputs_selected_2k_part_13 \
208
+ --split gen \
209
+ --max_search_limit 10 \
210
+ --max_turn 10 \
211
+ --top_k 10 \
212
+ --rollout_num 2 \
213
+ --max_doc_len 5000 \
214
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
215
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
216
+ --bing_endpoint "https://google.serper.dev/search" > logs/log_sum_all_webpage_gen_data/selected_2k_part_13_1.log 2>&1 &
deep_search/search_o1/run_syn_rollout_filter.sh ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # worker 1
2
+ # export CUDA_VISIBLE_DEVICES=0,1
3
+ # python /share/project/sunshuang/deep_search/search_o1/scripts/search_o1_sum_all_webpage_rollout.py\
4
+ # --dataset_name bf_merged_data_tagged_domain_keypoints_keywords_count_add_source_answer \
5
+ # --cache_dir_base cache_eval_sum_all_webpage \
6
+ # --output_dir_base /share/project/baifei/filter-syn-ai-search/rollout/0321/result_154/search/roll-32b-inst-search-154items \
7
+ # --split gen \
8
+ # --max_search_limit 10 \
9
+ # --max_turn 10 \
10
+ # --top_k 10 \
11
+ # --rollout_num 5 \
12
+ # --max_doc_len 5000 \
13
+ # --model_path "/share/project/zhipengchen/model/Qwen2.5-32B-Instruct" \
14
+ # --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
15
+ # --bing_endpoint "https://google.serper.dev/search" > /share/project/baifei/filter-syn-ai-search/rollout/0321/result_154/search/search-154.log 2>&1 &
16
+
17
+ export CUDA_VISIBLE_DEVICES=0,1
18
+ nohup /share/project/miniconda/envs/search_o1/bin/python -u scripts/search_o1_sum_all_webpage_rollout.py \
19
+ --dataset_name syn_data_test_wyh \
20
+ --cache_dir_base cache_eval_sum_all_webpage_wyhtest \
21
+ --output_dir_base /share/project/wuyuhuan/still-Syn/data/rollout_out/ \
22
+ --split gen \
23
+ --max_search_limit 10 \
24
+ --max_turn 10 \
25
+ --top_k 10 \
26
+ --rollout_num 1 \
27
+ --max_doc_len 5000 \
28
+ --model_path "/share/project/zhipengchen/model/QwQ-32B" \
29
+ --bing_subscription_key "cb0d28279a826d7e5cf22d71f683c77ffd4ba27d" \
30
+ --bing_endpoint "https://google.serper.dev/search" > /share/project/wuyuhuan/still-Syn/logs/rollout_output_test.log 2>&1 &
31
+
32
+ echo 1
deep_search/search_o1/search_ex.json ADDED
@@ -0,0 +1,138 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "searchParameters": {
3
+ "q": "when does season 14 of grey's anatomy come out",
4
+ "type": "search",
5
+ "num": 11,
6
+ "mkt": "en-US",
7
+ "setLang": "en",
8
+ "engine": "google"
9
+ },
10
+ "answerBox": {
11
+ "snippet": "Grey's Anatomy season 14\nGrey's Anatomy\n\nOriginal network\nABC\nOriginal release\nSeptember 28, 2017 – May 17, 2018\nSeason chronology\n\nList of episodes\n",
12
+ "title": "Grey's Anatomy season 14 - Wikipedia",
13
+ "link": "https://en.wikipedia.org/wiki/Grey%27s_Anatomy_season_14"
14
+ },
15
+ "organic": [
16
+ {
17
+ "title": "Season 14 (Grey's Anatomy)",
18
+ "link": "https://greysanatomy.fandom.com/wiki/Season_14_(Grey%27s_Anatomy)",
19
+ "snippet": "The fourteenth season of Grey's Anatomy premiered September 28, 2017, with a two-hour premiere. It was broadcast on Thursday nights at 8 PM on ABC.",
20
+ "position": 1
21
+ },
22
+ {
23
+ "title": "Grey's Anatomy Season 14 Premiere Date Announced - ABC",
24
+ "link": "https://abc.com/news/9fce69b2-3b93-40f3-9cd3-d75b1d90bb05/category/738075",
25
+ "snippet": "Grey's Anatomy fans, get ready for the Season 14 two-hour premiere on Thursday, September 28 at 8|7c!",
26
+ "date": "Jul 31, 2017",
27
+ "position": 2
28
+ },
29
+ {
30
+ "title": "Grey's Anatomy (TV Series 2005– ) - Episode list - IMDb",
31
+ "link": "https://www.imdb.com/title/tt0413573/episodes/?season=14",
32
+ "snippet": "S14.E1 ∙ Break Down the House. Thu, Sep 28, 2017.",
33
+ "position": 3
34
+ },
35
+ {
36
+ "title": "Season 14 – Grey's Anatomy - Rotten Tomatoes",
37
+ "link": "https://www.rottentomatoes.com/tv/greys_anatomy/s14",
38
+ "snippet": "Watch Grey's Anatomy — Season 14 with a subscription on Disney+, Hulu, Netflix, or buy it on Fandango at Home. Grey's Anatomy — Season 14. What to Know.",
39
+ "position": 4
40
+ },
41
+ {
42
+ "title": "1-800-799-7233 | Grey's Anatomy Universe Wiki | Fandom",
43
+ "link": "https://greysanatomy.fandom.com/wiki/1-800-799-7233",
44
+ "snippet": "1-800-799-7233 is the ninth episode and mid-season premiere of the fourteenth season and the 302nd overall episode of Grey's Anatomy.",
45
+ "date": "Jan 18, 2018",
46
+ "position": 5
47
+ },
48
+ {
49
+ "title": "Grey's Anatomy Season 14 [DVD] [2018] - Amazon.com",
50
+ "link": "https://www.amazon.com/Greys-Anatomy-Season-14-DVD/dp/B07H2VPXY1",
51
+ "snippet": "Grey's Anatomy Season 14 [DVD] [2018] ; Language, ‎English (Dolby Digital 5.1) ; Aspect Ratio, ‎Unknown ; Release date, ‎October 22, 2018 ; ASIN, ‎B07H2VPXY1.",
52
+ "rating": 3.9,
53
+ "ratingCount": 1134,
54
+ "position": 6
55
+ },
56
+ {
57
+ "title": "Grey's Anatomy Episodes",
58
+ "link": "https://greysanatomy.fandom.com/wiki/Grey%27s_Anatomy_Episodes",
59
+ "snippet": "We sat down with the cast of Electric State, coming March 14th on Netflix. This page lists all episodes of Grey's Anatomy.",
60
+ "position": 7
61
+ },
62
+ {
63
+ "title": "Grey's Anatomy - Wikipedia",
64
+ "link": "https://en.wikipedia.org/wiki/Grey%27s_Anatomy",
65
+ "snippet": "The series premiered on March 27, 2005, on ABC as a mid-season replacement. The show's title is a reference to Gray's Anatomy, a classic human anatomy textbook.",
66
+ "sitelinks": [
67
+ {
68
+ "title": "Season 14",
69
+ "link": "https://en.wikipedia.org/wiki/Grey%27s_Anatomy_season_14"
70
+ },
71
+ {
72
+ "title": "List of Grey's Anatomy episodes",
73
+ "link": "https://en.wikipedia.org/wiki/List_of_Grey%27s_Anatomy_episodes"
74
+ },
75
+ {
76
+ "title": "Season 1",
77
+ "link": "https://en.wikipedia.org/wiki/Grey%27s_Anatomy_season_1"
78
+ },
79
+ {
80
+ "title": "Season 15",
81
+ "link": "https://en.wikipedia.org/wiki/Grey%27s_Anatomy_season_15"
82
+ }
83
+ ],
84
+ "position": 8
85
+ },
86
+ {
87
+ "title": "Grey's Anatomy Season 14 Air Dates & Countdown - EpisoDate.com",
88
+ "link": "https://www.episodate.com/tv-show/grey-s-anatomy?season=14",
89
+ "snippet": "Grey's Anatomy Season 14 Air Dates ; Thursday Sep 28, 2017 · S14E01 - Break Down the House ; Thursday Sep 28, 2017 · S14E02 - Get Off on the Pain ; Thursday Oct 05, ...",
90
+ "rating": 9.5,
91
+ "ratingMax": 10,
92
+ "ratingCount": 109,
93
+ "position": 9
94
+ },
95
+ {
96
+ "title": "'Grey's Anatomy' Is Obviously Returning For Season 14 - Bustle",
97
+ "link": "https://www.bustle.com/p/when-does-greys-anatomy-season-14-premiere-the-show-will-go-on-as-long-as-meredith-does-57795",
98
+ "snippet": "UPDATE: ABC announced that Grey's Anatomy Season 14 will have a two-hour season premiere on Sept. 28. EARLIER: Grey's Anatomy is set to air on ...",
99
+ "date": "May 18, 2017",
100
+ "position": 10
101
+ }
102
+ ],
103
+ "peopleAlsoAsk": [
104
+ {
105
+ "question": "Is Grey's Anatomy season 19 out?",
106
+ "snippet": "It premiered on October 6, 2022, for the 2022–23 broadcast television season. Krista Vernoff returned for the season as executive producer and showrunner alongside her production company Trip the Light Productions.",
107
+ "title": "Grey's Anatomy season 19 - Wikipedia",
108
+ "link": "https://en.wikipedia.org/wiki/Grey%27s_Anatomy_season_19"
109
+ }
110
+ ],
111
+ "relatedSearches": [
112
+ {
113
+ "query": "Grey's Anatomy season 14 cast Guest Stars"
114
+ },
115
+ {
116
+ "query": "Grey's Anatomy Season 15 summary"
117
+ },
118
+ {
119
+ "query": "How many Episodes in Grey's Anatomy season 15"
120
+ },
121
+ {
122
+ "query": "When did season 15 of Grey's Anatomy come out"
123
+ },
124
+ {
125
+ "query": "Grey's Anatomy season 14 Episode 11 full cast"
126
+ },
127
+ {
128
+ "query": "Who dies in season 15 of Grey's Anatomy"
129
+ },
130
+ {
131
+ "query": "Grey's Anatomy season 14 Episode 16 cast astronaut"
132
+ },
133
+ {
134
+ "query": "Grey's anatomy season 14 Episode 1 full cast"
135
+ }
136
+ ],
137
+ "credits": 2
138
+ }
deep_search/search_o1/search_without_wiki.json ADDED
@@ -0,0 +1,78 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "searchParameters": {
3
+ "q": "when does season 14 of grey's anatomy come out -site:en.wikipedia.org",
4
+ "type": "search",
5
+ "mkt": "en-US",
6
+ "setLang": "en",
7
+ "engine": "google"
8
+ },
9
+ "organic": [
10
+ {
11
+ "title": "Season 14 (Grey's Anatomy)",
12
+ "link": "https://greysanatomy.fandom.com/wiki/Season_14_(Grey%27s_Anatomy)",
13
+ "snippet": "The fourteenth season of Grey's Anatomy premiered September 28, 2017, with a two-hour premiere. It was broadcast on Thursday nights at 8 PM on ABC.",
14
+ "position": 1
15
+ },
16
+ {
17
+ "title": "Season 14 – Grey's Anatomy",
18
+ "link": "https://www.rottentomatoes.com/tv/greys_anatomy/s14",
19
+ "snippet": "Watch Grey's Anatomy — Season 14 with a subscription on Disney+, Hulu, Netflix, or buy it on Fandango at Home. Grey's Anatomy — Season 14. What to Know.",
20
+ "position": 2
21
+ },
22
+ {
23
+ "title": "Grey's Anatomy (TV Series 2005– ) - Episode list - ...",
24
+ "link": "https://www.imdb.com/title/tt0413573/episodes/?season=14",
25
+ "snippet": "S14.E21 ∙ Bad Reputation. Thu, Apr 26, 2018. Crisis management is brought in to help Grey Sloan sort out allegations against Harper Avery; Alex and Jo are ...",
26
+ "position": 3
27
+ },
28
+ {
29
+ "title": "Grey's Anatomy Season 14 Premiere Date Announced",
30
+ "link": "https://abc.com/news/9fce69b2-3b93-40f3-9cd3-d75b1d90bb05/category/738075",
31
+ "snippet": "Grey's Anatomy fans, get ready for the Season 14 two-hour premiere on Thursday, September 28 at 8|7c!",
32
+ "date": "Jul 31, 2017",
33
+ "position": 4
34
+ },
35
+ {
36
+ "title": "Grey's Anatomy Season 14 [DVD] [2018]",
37
+ "link": "https://www.amazon.com/Greys-Anatomy-Season-14-DVD/dp/B07H2VPXY1",
38
+ "snippet": "Grey's Anatomy Season 14 [DVD] [2018] ; Language, ‎English (Dolby Digital 5.1) ; Aspect Ratio, ‎Unknown ; Release date, ‎October 22, 2018 ; ASIN, ‎B07H2VPXY1.",
39
+ "rating": 3.9,
40
+ "ratingCount": 1134,
41
+ "currency": "$",
42
+ "price": 24.26,
43
+ "position": 5
44
+ },
45
+ {
46
+ "title": "Buy & Watch Grey's Anatomy: Season 14 - Vudu - Fandango",
47
+ "link": "https://athome.fandango.com/content/browse/details/Greys-Anatomy-You-Really-Got-a-Hold-on-Me/891354?cmp=rt_leaderboard",
48
+ "snippet": "Watch Grey's Anatomy: Season 14 on Fandango at Home! Buy and binge watch the entire series on Fandango at Home today.",
49
+ "position": 6
50
+ },
51
+ {
52
+ "title": "Timeline - Grey's Anatomy Universe Wiki - Fandom",
53
+ "link": "https://greysanatomy.fandom.com/wiki/Timeline",
54
+ "snippet": "Where Do We Go From Here takes place on November 14, 2014. The Bed's Too Big ... Season 14 starts on or around mid to late-2017, immediately after the ...",
55
+ "position": 7
56
+ },
57
+ {
58
+ "title": "Grey's Anatomy Episodes",
59
+ "link": "https://greysanatomy.fandom.com/wiki/Grey%27s_Anatomy_Episodes",
60
+ "snippet": "Season 7 (2010-11) ; 13, \"Don't Deceive Me (Please Don't Go)\", February 3, 2011, 11.18 ; 14, \"P.Y.T. (Pretty Young Thing)\", February 10, 2011, 10.47.",
61
+ "position": 8
62
+ },
63
+ {
64
+ "title": "'Grey's Anatomy' Is Obviously Returning For Season 14",
65
+ "link": "https://www.bustle.com/p/when-does-greys-anatomy-season-14-premiere-the-show-will-go-on-as-long-as-meredith-does-57795",
66
+ "snippet": "UPDATE: ABC announced that Grey's Anatomy Season 14 will have a two-hour season premiere on Sept. 28. EARLIER: Grey's Anatomy is set to air on ...",
67
+ "date": "May 18, 2017",
68
+ "position": 9
69
+ },
70
+ {
71
+ "title": "Grey's Anatomy Season 14 Air Dates & Countdown",
72
+ "link": "https://www.episodate.com/tv-show/grey-s-anatomy?season=14",
73
+ "snippet": "Grey's Anatomy Season 14 Air Dates ; Thursday Sep 28, 2017 · S14E01 - Break Down the House ; Thursday Sep 28, 2017 · S14E02 - Get Off on the Pain ; Thursday Oct 05, ...",
74
+ "position": 10
75
+ }
76
+ ],
77
+ "credits": 1
78
+ }
deep_search/search_o1/vllm_serve.sh ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ export CUDA_VISIBLE_DEVICES=1,2,3,4
2
+ nohup vllm serve /share/project/zhipengchen/model/Qwen2.5-72B-Instruct \
3
+ --tensor-parallel-size=4 \
4
+ --gpu-memory-utilization 0.95 \
5
+ --port 8003 > /share/project/sunshuang/deep_search/search_o1/logs/vlm_serve_qwen72b_1.log 2>&1 &
6
+
7
+ export CUDA_VISIBLE_DEVICES=3
8
+ vllm serve /opt/aps/workdir/home/.cache/modelscope/hub/models/qwen/Qwen2.5-7B-Instruct \
9
+ --tensor-parallel-size=1 \
10
+ --gpu-memory-utilization 0.95
deep_search/sft_logs/3-12-qa1200_qwq_qwen32b.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/3-13_r1_32b_1200.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/3-15_sft_qwq.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/3-16_difficult_rollout_data_sft_add_prompt.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/3-16_r1_32b_1478.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/3-18_2k_qwen32b_inst.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/3-18_2k_qwq_r1_32b.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/3-19-1.2k_remove_exp_qwq.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/3-20_1161_remove_exp_qwen7b_inst.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/3-20_merged_2462_remove_exp_qwq.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/3-22-merged_2462_nq_237_total_2699.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/3-25_strict_selected_1526.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/3-27.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/4-2.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/re.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/sft_re.log ADDED
The diff for this file is too large to render. See raw diff
 
deep_search/sft_logs/test.log ADDED
The diff for this file is too large to render. See raw diff
 
ssh_node.sh ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ ssh job-86a69723-0cea-434e-8ebd-135aba19ceb4-master-0
2
+ ssh job-86a69723-0cea-434e-8ebd-135aba19ceb4-worker-0
3
+ ssh job-86a69723-0cea-434e-8ebd-135aba19ceb4-worker-1
4
+ ssh job-86a69723-0cea-434e-8ebd-135aba19ceb4-worker-2
5
+ ssh job-86a69723-0cea-434e-8ebd-135aba19ceb4-worker-3
6
+ ssh job-86a69723-0cea-434e-8ebd-135aba19ceb4-worker-4