-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathcommands.sh
829 lines (624 loc) · 86.9 KB
/
commands.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:3 -partial -partial_hidden_zero --sigma 0.35 2>&1 | tee logs/partial_dp/20210411/1201/hiddenzero_lr0.1_sigma0.35
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:3 --sigma 0.5 -norm 0.02 2>&1 | tee logs/dp/20210411/1511/lr0.1_sigma0.5_norm0.02
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:3 -partial --sigma 0.5 -norm 0.02 --epochs 1 2>&1 | tee logs/partial_dp/20210413/1147/lr0.1_sigma0.5_norm0.02_epoch1
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:3 -partial --sigma 0.5 -norm 0.02 --epochs 1 2>&1 | tee logs/partial_dp/20210413/1457/lr0.1_sigma0.5_norm0.02_epoch1_save_every_batch
# repeat 4 times, partial dp
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:2 -partial -norm 0.02 --seed 0 2>&1 | tee logs/partial_dp/20210414/1211/nohidden_lr0.1_norm0.02_seed0 # screen partialdp2
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:3 -partial -norm 0.02 --seed 123 2>&1 | tee logs/partial_dp/20210414/1211/nohidden_lr0.1_norm0.02_seed123 # screen partialdp
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:4 -partial -norm 0.02 --seed 22 2>&1 | tee logs/partial_dp/20210414/1211/nohidden_lr0.1_norm0.02_seed22 # screen partialdp3
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:5 -partial -norm 0.02 --seed 300 2>&1 | tee logs/partial_dp/20210414/1211/nohidden_lr0.1_norm0.02_seed300 # screen partialdp4
# repeat 2 times, dp
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:6 -norm 0.02 --seed 0 2>&1 | tee logs/dp/20210414/1224/lr0.1_sigma0.5_norm0.02_seed0 # screen dp
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:1 -norm 0.02 --seed 123 2>&1 | tee logs/dp/20210414/1224/lr0.1_sigma0.5_norm0.02_seed123 # screen dp2
# partial dp, norm=0.01
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:0 -partial -norm 0.01 2>&1 | tee logs/partial_dp/20210414/1128/nohidden_lr0.1_norm0.01 # screen partialdp4
# join canary insertion csv with privacy
python attacks/join_epsilon_with_attack.py -ckpt model/partialdp/20210411/122315/ -log logs/partial_dp/20210411/1219/nohidden_lr0.1_norm0.02 -csv attacks/canary_insertion/partial_dp_sigma05_lr01_norm002.csv
python attacks/join_epsilon_with_attack.py -ckpt model/partialdp/20210413/145857/ -log logs/partial_dp/20210413/1457/lr0.1_sigma0.5_norm0.02_epoch1_save_every_batch -csv attacks/canary_insertion/partial_dp_sigma05_lr01_norm002_epoch1_save_every_batch.csv
python attacks/join_epsilon_with_attack.py -ckpt model/dp/20210409/223157/ -log logs/dp/20210409/2329/lstm.log -csv attacks/canary_insertion/dp_sigma05_lr005_norm01.csv
python attacks/join_epsilon_with_attack.py -ckpt model/dp/20210411/151450/ -log logs/dp/20210411/1511/lr0.1_sigma0.5_norm0.02 -csv attacks/canary_insertion/dp_sigma05_lr01_norm002.csv
python attacks/join_epsilon_with_attack.py -ckpt model/partialdp/20210413/145857/ -log logs/partial_dp/20210413/1457/lr0.1_sigma0.5_norm0.02_epoch1_save_every_batch -csv attacks/membership_inference/partial_dp_lr01_sigma05_norm002_1000_epoch1_save_every_batch.csv
python attacks/join_epsilon_with_attack.py -ckpt model/nodp/20210409/185850/ -log logs/nodp/20210409/1855/lstm.log -csv attacks/canary_insertion/nodp_10insertion_6digits.csv
python attacks/join_epsilon_with_attack.py -ckpt model/nodp/20210413/123134/ -log logs/nodp/20210413/1107/epoch1.log -csv attacks/canary_insertion/nodp_10insertion_epoch1_bs32.csv
python attacks/join_epsilon_with_attack.py -ckpt model/nodp/20210413/115956/ -log logs/nodp/20210413/1121/bs16.log -csv attacks/canary_insertion/nodp_10insertion_epoch1_bs16.csv
python attacks/join_epsilon_with_attack.py -ckpt model/dp/20210409/222642/ -log logs/dp/20210409/2225/lstm.log -csv attacks/canary_insertion/dp_sigma05_lr01_norm01.csv
python attacks/join_epsilon_with_attack.py -ckpt model/dp/20210411/123315/ -log logs/dp/20210411/1232/sigma0.25_lr0.1_norm0.1 -csv attacks/canary_insertion/dp_sigma025_lr01_norm01.csv
python attacks/join_epsilon_with_attack.py -ckpt model/partialdp/20210410/093833/ -log logs/partial_dp/20210410/0937/sigma0.25 -csv attacks/canary_insertion/partial_dp_sigma025_lr01_norm01.csv
# join membership csv with privacy
python attacks/join_epsilon_with_attack.py -ckpt model/partialdp/20210410/093833/ -log logs/partial_dp/20210410/0937/sigma0.25 -csv attacks/membership_inference/partialdp_lr01_sigma025_norm01_1000.csv
python attacks/join_epsilon_with_attack.py -ckpt model/partialdp/20210411/122315/ -log logs/partial_dp/20210411/1219/nohidden_lr0.1_norm0.02 -csv attacks/membership_inference/partial_dp_lr01_sigma05_norm002_1000_epoch50.csv
python attacks/join_epsilon_with_attack.py -ckpt model/partialdp/20210413/145857/ -log logs/partial_dp/20210413/1457/lr0.1_sigma0.5_norm0.02_epoch1_save_every_batch -csv attacks/membership_inference/partial_dp_lr01_sigma05_norm002_1000_epoch1_save_every_batch_with_privacy.csv
python attacks/join_epsilon_with_attack.py -ckpt model/dp/20210409/223157/ -log logs/dp/20210409/2329/lstm.log -csv attacks/membership_inference/dp_lr005_sigma05_norm01_1000.csv
python attacks/join_epsilon_with_attack.py -ckpt model/dp/20210411/151450/ -log logs/dp/20210411/1511/lr0.1_sigma0.5_norm0.02 -csv attacks/membership_inference/dp_lr05_sigma05_norm002_1000.csv
python attacks/join_epsilon_with_attack.py -ckpt model/nodp/20210409/185850/ -log logs/nodp/20210409/1855/lstm.log -csv attacks/membership_inference/nodp_1000.csv
# membership inference attack
python attacks/mem_inference.py -ckpt model/nodp/20210409/185850/ --outputf attacks/membership_inference/nodp_1000.csv --cuda cuda:5 --N 1000 -bs 64
python attacks/mem_inference.py -ckpt model/dp/20210409/223157/ --outputf attacks/membership_inference/dp_lr005_sigma05_norm01_1000.csv --cuda cuda:5 --N 1000 -bs 64
python attacks/mem_inference.py -ckpt model/dp/20210411/151450/ --outputf attacks/membership_inference/dp_lr05_sigma05_norm002_1000.csv --cuda cuda:5 --N 1000 -bs 64
python attacks/mem_inference.py -ckpt model/partialdp/20210410/093833/ --outputf attacks/membership_inference/partialdp_lr01_sigma025_norm01_1000.csv --cuda cuda:5 --N 1000 -bs 64
# no dp, repeat
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b --cuda cuda:0 2>&1 | tee logs/nodp/20210416/1710/bs16.log
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b --cuda cuda:1 --seed 0 2>&1 | tee logs/nodp/20210416/1710/bs16_see0.log
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b --cuda cuda:2 --seed 123 2>&1 | tee logs/nodp/20210416/1710/bs16_seed123.log
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b --cuda cuda:4 --seed 22 2>&1 | tee logs/nodp/20210416/1710/bs16_seed22.log
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b --cuda cuda:6 --seed 300 2>&1 | tee logs/nodp/20210416/1710/bs16_seed300.log
# canary insertion for partial-dp repeat
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210414/001535/ --cuda cuda:5 --outputf attacks/canary_insertion/partial_dp_sigma05_lr01_norm002_seed0.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210414/002043/ --cuda cuda:5 --outputf attacks/canary_insertion/partial_dp_sigma05_lr01_norm002_seed123.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210414/002122/ --cuda cuda:5 --outputf attacks/canary_insertion/partial_dp_sigma05_lr01_norm002_seed22.csv
# canary insertion for dp repeat
python attacks/canary_insertion.py -bs 256 --checkpoint model/dp/20210414/002943/ --cuda cuda:0 --outputf attacks/canary_insertion/dp_sigma05_lr01_norm002_seed0.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/dp/20210414/002746/ --cuda cuda:0 --outputf attacks/canary_insertion/p_sigma05_lr01_norm002_seed123.csv
# partial dp, parameter search
# screen -r partialdp
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:0 -partial -norm 0.01 --sigma 0.45 --seed 1111 2>&1 | tee logs/partial_dp/20210416/2351/nohidden_lr0.1_norm0.01_sigma0.45
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b --cuda cuda:0 2>&1 | tee logs/nodp/20210416/2354/bs16.log
# screen -r partialdp2
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:1 -partial -norm 0.005 --sigma 0.45 --seed 1111 2>&1 | tee logs/partial_dp/20210416/2351/nohidden_lr0.1_norm0.005_sigma0.45
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b --cuda cuda:1 --seed 0 2>&1 | tee logs/nodp/20210416/2354/bs16_see0.log
# screen -r partialdp3
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:2 -partial -norm 0.01 --sigma 0.5 --seed 1111 2>&1 | tee logs/partial_dp/20210416/2351/nohidden_lr0.1_norm0.01_sigma0.5
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b --cuda cuda:2 --seed 123 2>&1 | tee llogs/nodp/20210416/2354/bs16_seed123.log
# screen -r partialdp4
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:5 -partial -norm 0.005 --sigma 0.5 --seed 1111 2>&1 | tee logs/partial_dp/20210416/2351/nohidden_lr0.1_norm0.005_sigma0.5
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b --cuda cuda:5 --seed 22 2>&1 | tee logs/nodp/20210416/2354/bs16_seed22.log
# screen -R partialdp5
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:4 -partial -norm 0.01 --sigma 0.55 --seed 1111 2>&1 | tee logs/partial_dp/20210416/2351/nohidden_lr0.1_norm0.01_sigma0.55
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b --cuda cuda:4 --seed 300 2>&1 | tee logs/nodp/20210416/2354/bs16_seed300.log
# screen -R partialdp6
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:6 -partial -norm 0.005 --sigma 0.55 --seed 1111 2>&1 | tee logs/partial_dp/20210416/2351/nohidden_lr0.1_norm0.005_sigma0.55
# parameter tunning on the norm
# screen -R partialdp6
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:6 -partial -norm 1e-3 --sigma 0.5 --seed 1111 2>&1 | tee logs/partial_dp/20210418/1912/nohidden_lr0.1_norm1e-3_sigma0.5
# screen -R partialdp5
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:1 -partial -norm 5e-4 --sigma 0.5 --seed 1111 2>&1 | tee logs/partial_dp/20210418/1912/nohidden_lr0.1_norm5e-4_sigma0.5
# screen -R partialdp4 # rerun
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:2 -partial -norm 1e-4 --sigma 0.5 --seed 1111 2>&1 | tee logs/partial_dp/20210418/1912/nohidden_lr0.1_norm1e-4_sigma0.5
# screen -R partialdp3
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:4 -partial -norm 5e-5 --sigma 0.5 --seed 1111 2>&1 | tee logs/partial_dp/20210418/1912/nohidden_lr0.1_norm5e-5_sigma0.5
# screen -R partialdp2 # rerun
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:5 -partial -norm 1e-5 --sigma 0.5 --seed 1111 2>&1 | tee logs/partial_dp/20210418/1912/nohidden_lr0.1_norm1e-5_sigma0.5
# screen -R partialdp # rerun
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:0 -partial -norm 5e-6 --sigma 0.5 --seed 1111 2>&1 | tee logs/partial_dp/20210418/1912/nohidden_lr0.1_norm5e-6_sigma0.5
# repeat 5 times for norm=1e-3
# screen -R partialdp6
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:6 -partial -norm 1e-3 --sigma 0.5 --seed 0 2>&1 | tee logs/partial_dp/20210421/1021/nohidden_lr0.1_norm1e-3_sigma0.5_seed0
# screen -R partialdp5
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:1 -partial -norm 1e-3 --sigma 0.5 --seed 123 2>&1 | tee logs/partial_dp/20210421/1021/nohidden_lr0.1_norm1e-3_sigma0.5_seed123
# screen -R partialdp4
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:2 -partial -norm 1e-3 --sigma 0.5 --seed 22 2>&1 | tee logs/partial_dp/20210421/1021/nohidden_lr0.1_norm1e-3_sigma0.5_seed22
# screen -R partialdp3
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:4 -partial -norm 1e-3 --sigma 0.5 --seed 300 2>&1 | tee logs/partial_dp/20210421/1021/nohidden_lr0.1_norm1e-3_sigma0.5_seed300
# repeat 5 times for norm=1e-3, dp
# screen -R dp # have run
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:5 -norm 1e-3 --seed 1111 2>&1 | tee logs/dp/20210421/1029/lr0.1_sigma0.5_norm1e-3_seed1111 # screen dp
# repeat 5 times for norm=1e-2, dp
# screen -R dp2
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:5 -norm 1e-2 --seed 1111 2>&1 | tee logs/dp/20210422/1437/lr0.1_sigma0.5_norm1e-2_seed1111 # screen dp
# # screen -R dp3
# python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:6 -norm 1e-2 --seed 123 2>&1 | tee logs/dp/20210421/1029/lr0.1_sigma0.5_norm1e-3_seed123 # screen dp
# # screen -R dp4
# python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:6 -norm 1e-2 --seed 22 2>&1 | tee logs/dp/20210421/1029/lr0.1_sigma0.5_norm1e-3_seed22 # screen dp
# # screen -R dp5
# python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:6 -norm 1e-2 --seed 300 2>&1 | tee logs/dp/20210421/1029/lr0.1_sigma0.5_norm1e-3_seed300 # screen dp
# canary for no-dp, new runs
# canary insertion for dp repeat
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210418/181252 --cuda cuda:0 --outputf attacks/canary_insertion/nodp/nodp_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210418/192226 --cuda cuda:0 --outputf attacks/canary_insertion/nodp/nodp_seed0.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210417/144949 --cuda cuda:0 --outputf attacks/canary_insertion/nodp/nodp_seed123.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210419/052248 --cuda cuda:0 --outputf attacks/canary_insertion/nodp/nodp_seed22.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210418/210231 --cuda cuda:0 --outputf attacks/canary_insertion/nodp/nodp_seed300.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210417/000742 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.45_norm0.01_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210417/000240 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.45_norm0.005_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210417/000251 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.5_norm0.01_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210417/111003 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.5_norm0.005_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210417/000327 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.55_norm0.01_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210417/000344 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.55_norm0.005_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210418/191438 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.5_norm0.001_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210419/063839 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.5_norm0.0005_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210419/134334 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.5_norm0.0001_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210419/064057 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.5_norm5e-05_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210419/134345 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.5_norm1e-05_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210419/134357 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.5_norm5e-06_seed1111.csv
# membership
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/nodp/20210418/181252 --cuda cuda:0 --outputf attacks/membership_inference/nodp/nodp_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/nodp/20210418/192226 --cuda cuda:0 --outputf attacks/membership_inference/nodp/nodp_seed0.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/nodp/20210417/144949 --cuda cuda:0 --outputf attacks/membership_inference/nodp/nodp_seed123.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/nodp/20210419/052248 --cuda cuda:0 --outputf attacks/membership_inference/nodp/nodp_seed22.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/nodp/20210418/210231 --cuda cuda:0 --outputf attacks/membership_inference/nodp/nodp_seed300.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210417/000742 --cuda cuda:0 --outputf attacks/membership_inference/partialdp/lr0.1_sigma0.45_norm0.01_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210417/000240 --cuda cuda:0 --outputf attacks/membership_inference/partialdp/lr0.1_sigma0.45_norm0.005_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210417/000251 --cuda cuda:0 --outputf attacks/membership_inference/partialdp/lr0.1_sigma0.5_norm0.01_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210417/111003 --cuda cuda:0 --outputf attacks/membership_inference/partialdp/lr0.1_sigma0.5_norm0.005_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210417/000327 --cuda cuda:0 --outputf attacks/membership_inference/partialdp/lr0.1_sigma0.55_norm0.01_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210417/000344 --cuda cuda:0 --outputf attacks/membership_inference/partialdp/lr0.1_sigma0.55_norm0.005_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210418/191438 --cuda cuda:0 --outputf attacks/membership_inference/partialdp/lr0.1_sigma0.5_norm0.001_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210419/063839 --cuda cuda:0 --outputf attacks/membership_inference/partialdp/lr0.1_sigma0.5_norm0.0005_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210419/134334 --cuda cuda:0 --outputf attacks/membership_inference/partialdp/lr0.1_sigma0.5_norm0.0001_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210419/064057 --cuda cuda:0 --outputf attacks/membership_inference/partialdp/lr0.1_sigma0.5_norm5e-05_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210419/134345 --cuda cuda:0 --outputf attacks/membership_inference/partialdp/lr0.1_sigma0.5_norm1e-05_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210419/134357 --cuda cuda:0 --outputf attacks/membership_inference/partialdp/lr0.1_sigma0.5_norm5e-06_seed1111.csv
# parameter search on the sigma and norm
# screen -R partialdp
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:1 -partial -norm 5e-3 --sigma 0.1 --seed 1111 2>&1 | tee logs/partial_dp/20210423/nohidden_lr0.1_norm5e-3_sigma0.1_seed1111
# screen -R partialdp5
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:2 -partial -norm 0.25 --sigma 0.1 --seed 1111 2>&1 | tee logs/partial_dp/20210423/nohidden_lr0.1_norm0.25_sigma0.1_seed1111
# screen -R partialdp4
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:4 -partial -norm 5e-3 --sigma 0.05 --seed 1111 2>&1 | tee logs/partial_dp/20210423/nohidden_lr0.1_norm5e-3_sigma0.05_seed1111
# screen -R partialdp3
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:6 -partial -norm 0.25 --sigma 0.05 --seed 1111 2>&1 | tee logs/partial_dp/20210423/nohidden_lr0.1_norm0.25_sigma0.05_seed1111
# screen -R partialdp2 # notyet
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:5 -partial -norm 5e-3 --sigma 0.01 --seed 1111 2>&1 | tee logs/partial_dp/20210423/nohidden_lr0.1_norm5e-3_sigma0.01_seed1111
# screen -R partialdp6 # not yet
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:0 -partial -norm 0.25 --sigma 0.01 --seed 1111 2>&1 | tee logs/partial_dp/20210423/nohidden_lr0.1_norm0.25_sigma0.01_seed1111
# resume 50 epochs for sigma=0.5, on dialog
# screen -R resume1
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:1 -partial -norm 1e-3 --sigma 0.5 --seed 1111 -resume -resume_from_epoch_num 50 -resume_from model/partialdp/20210418/191438/data-wikitext-2-add10b_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.5_norm-0.001_dl-8e-05.pt_ppl-161.1260678_acc-0.33143_epoch-50_ep-5.376_dl-8e-05_ap-3.60 2>&1 | tee logs/partial_dp/20210423/resume/nohidden_lr0.1_norm1e-3_sigma0.5_seed1111
# screen -R resume2
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:1 -partial -norm 1e-3 --sigma 0.5 --seed 0 -resume -resume_from_epoch_num 50 -resume_from model/partialdp/20210421/123500/data-wikitext-2-add10b_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.5_norm-0.001_dl-8e-05.pt_ppl-166.0325998_acc-0.30857_epoch-50_ep-5.376_dl-8e-05_ap-3.60 2>&1 | tee logs/partial_dp/20210423/resume/nohidden_lr0.1_norm1e-3_sigma0.5_seed0
# dp param search, on dialog
# screen -R dp1
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:0 -norm 0.25 --seed 1111 2>&1 | tee logs/dp/20210423/param_search/lr0.1_sigma0.5_norm0.25_seed1111
# screen -R dp2
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:0 -norm 0.5 --seed 1111 2>&1 | tee logs/dp/20210423/param_search/lr0.1_sigma0.5_norm0.5_seed1111
# dp repeat 5 times, on interaction
# screen -R dp1
python -u main.py --epochs 100 -bs 7 --lr 0.05 -dp --cuda cuda:1 -norm 0.1 --seed 1111 2>&1 | tee logs/dp/20210424/repeat/lr0.05_sigma0.5_norm0.1_seed1111
# screen -R dp2
python -u main.py --epochs 100 -bs 7 --lr 0.05 -dp --cuda cuda:2 -norm 0.1 --seed 0 2>&1 | tee logs/dp/20210424/repeat/lr0.05_sigma0.5_norm0.1_seed0
# screen -R dp3
python -u main.py --epochs 100 -bs 7 --lr 0.05 -dp --cuda cuda:4 -norm 0.1 --seed 123 2>&1 | tee logs/dp/20210424/repeat/lr0.05_sigma0.5_norm0.1_seed123
# screen -R dp4
python -u main.py --epochs 100 -bs 7 --lr 0.05 -dp --cuda cuda:5 -norm 0.1 --seed 22 2>&1 | tee logs/dp/20210424/repeat/lr0.05_sigma0.5_norm0.1_seed22
# screen -R dp4
python -u main.py --epochs 100 -bs 7 --lr 0.05 -dp --cuda cuda:6 -norm 0.1 --seed 300 2>&1 | tee logs/dp/20210424/repeat/lr0.05_sigma0.5_norm0.1_seed300
# parameter search on the much smaller sigma and norm, dialog
# screen -R partialdp1 #
python -u main.py -bs 7 --lr 0.1 -dp --epochs 100 --cuda cuda:1 -partial -norm 1e-3 --sigma 0.1 --seed 1111 2>&1 | tee logs/partial_dp/20210425/param_search_new/nohidden_lr0.1_norm1e-3_sigma0.1_seed1111
# screen -R partialdp2 #
python -u main.py -bs 7 --lr 0.1 -dp --epochs 100 --cuda cuda:1 -partial -norm 2.5e-3 --sigma 0.1 --seed 1111 2>&1 | tee logs/partial_dp/20210425/param_search_new/nohidden_lr0.1_norm2.5e-3_sigma0.1_seed1111
# screen -R partialdp3 # not yet
python -u main.py -bs 7 --lr 0.1 -dp --epochs 100 --cuda cuda:1 -partial -norm 7.5e-3 --sigma 0.1 --seed 1111 2>&1 | tee logs/partial_dp/20210425/param_search_new/nohidden_lr0.1_norm7.5e-3_sigma0.1_seed1111
# screen -R partialdp4 # not yet
python -u main.py -bs 7 --lr 0.1 -dp --epochs 100 --cuda cuda:1 -partial -norm 1e-2 --sigma 0.1 --seed 1111 2>&1 | tee logs/partial_dp/20210425/param_search_new/nohidden_lr0.1_norm1e-2_sigma0.1_seed1111
python -u main.py -bs 7 --lr 0.1 -dp --epochs 100 --cuda cuda:1 -partial -norm 0.25 --sigma 0.005 --seed 1111 2>&1 | tee logs/partial_dp/20210425/param_search/nohidden_lr0.1_norm0.25_sigma0.005_seed1111
# resume 50 epochs for sigma=0.01, norm=0.25 on interaction
# screen -R resume1
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:0 -partial -norm 5e-3 --sigma 0.1 --seed 1111 -resume -resume_from_epoch_num 50 -resume_from model/partialdp/20210423/111019/data-wikitext-2-add10b_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.1_norm-0.005_dl-8e-05.pt_ppl-151.1701144_acc-0.33714_epoch-50_ep-132047.094_dl-8e-05_ap-1.10 2>&1 | tee logs/partial_dp/20210425/resume/lr0.1_norm1e-3_sigma0.1_seed1111
# missing digit, baseline, on interaction
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-normalized/missing_digits --cuda cuda:3 2>&1 | tee logs/nodp/normalized/20210426/lstm.log
# screen -R nodp2
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-normalized/missing_digits --cuda cuda:3 --seed 0 2>&1 | tee logs/nodp/normalized/20210426/lstm_seed0.log
# screen -R nodp3
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-normalized/missing_digits --cuda cuda:3 --seed 123 2>&1 | tee logs/nodp/normalized/20210426/lstm_seed123.log
# screen -R nodp4
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-normalized/missing_digits --cuda cuda:3 --seed 22 2>&1 | tee logs/nodp/normalized/20210426/lstm_seed22.log
# screen -R nodp5
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-normalized/missing_digits --cuda cuda:3 --seed 300 2>&1 | tee logs/nodp/normalized/20210426/lstm_seed300.log
# missing digit, partial dp, on dialog
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:0 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add10b --epochs 100 --seed 1111 2>&1 | tee logs/partial_dp/missed/20210426/lr0.1_sigm0.5_norm1e-3_seed1111_miss10.log
# missing digit, partial dp, on interaction
### screen -R partialdp1
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:0 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add10b --epochs 100 --seed 0 2>&1 | tee logs/partial_dp/missed/20210426/lr0.1_sigm0.5_norm1e-3_seed0_miss10.log
### screen -R partialdp2
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:1 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add10b --epochs 100 --seed 123 2>&1 | tee logs/partial_dp/missed/20210426/lr0.1_sigm0.5_norm1e-3_seed123_miss10.log
### screen -R partialdp3
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:2 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add10b --epochs 100 --seed 22 2>&1 | tee logs/partial_dp/missed/20210426/lr0.1_sigm0.5_norm1e-3_seed22_miss10.log
### screen -R partialdp4
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:4 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add10b --epochs 100 --seed 300 2>&1 | tee logs/partial_dp/missed/20210426/lr0.1_sigm0.5_norm1e-3_seed300_miss10.log
# dialog, test, on dialog server
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:0 -dp -partial -bs 1 --sigma 0.5 -norm 1e-3 2>&1 | tee logs/partial_dp/dialog/20210426/sigma0.5_norm1e-3
# resume 50 epochs for sigma=0.5, on interaction
# screen -R resume1
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:5 -partial -norm 1e-3 --sigma 0.5 --seed 123 -resume -resume_from_epoch_num 50 -resume_from model/partialdp/20210421/123511/data-wikitext-2-add10b_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.5_norm-0.001_dl-8e-05.pt_ppl-159.3748673_acc-0.33714_epoch-50_ep-5.376_dl-8e-05_ap-3.60 2>&1 | tee logs/partial_dp/20210423/resume/nohidden_lr0.1_norm1e-3_sigma0.5_seed123
# screen -R resume2
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:6 -partial -norm 1e-3 --sigma 0.5 --seed 22 -resume -resume_from_epoch_num 50 -resume_from model/partialdp/20210421/123522/data-wikitext-2-add10b_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.5_norm-0.001_dl-8e-05.pt_ppl-164.0274903_acc-0.32571_epoch-50_ep-5.375_dl-8e-05_ap-3.60 2>&1 | tee logs/partial_dp/20210423/resume/nohidden_lr0.1_norm1e-3_sigma0.5_seed22
# resume 50 epochs for sigma=0.5, on dialog
# screen -R resume3
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:0 -partial -norm 1e-3 --sigma 0.5 --seed 300 -resume -resume_from_epoch_num 50 -resume_from model/partialdp/20210421/123530/data-wikitext-2-add10b_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.5_norm-0.001_dl-8e-05.pt_ppl-187.3671091_acc-0.31429_epoch-50_ep-5.376_dl-8e-05_ap-3.60 2>&1 | tee logs/partial_dp/20210423/resume/nohidden_lr0.1_norm1e-3_sigma0.5_seed300
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:1 -partial -norm 1e-3 --sigma 0.5 --seed 300 -resume -resume_from_epoch_num 50 -resume_from model/partialdp/20210421/123530/data-wikitext-2-add10b_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.5_norm-0.001_dl-8e-05.pt_ppl-187.3671091_acc-0.31429_epoch-50_ep-5.376_dl-8e-05_ap-3.60 2>&1 | tee logs/partial_dp/20210423/resume/nohidden_lr0.1_norm1e-3_sigma0.5_seed300
# dialog experiment, baseline
# screen -R dialog_nodp
python -u main.py -bs 16 --lr 20 --data data/simdial --data_type dial --cuda cuda:3 2>&1 | tee logs/nodp/dialog/20210427/dialog_bs16.log
# dialog, parameter tuning, on dialog server
# screen -R dialog_partialdp
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:1 -dp -partial -bs 3 --sigma 0.5 -norm 1e-3 --epochs 100 2>&1 | tee logs/partial_dp/dialog/20210428/sigma0.5_norm1e-3
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:1 -dp -partial -bs 3 --sigma 0.5 -norm 1e-2 --epochs 100 2>&1 | tee logs/partial_dp/dialog/20210428/sigma0.5_norm1e-2
# canary, for normalized experiment, baseline, on interaction
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210426/182422 --cuda cuda:3 --outputf attacks/canary_insertion/nodp_normalized/nodp_normalized_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210426/234106 --cuda cuda:3 --outputf attacks/canary_insertion/nodp_normalized/nodp_normalized_seed0.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210426/234125 --cuda cuda:3 --outputf attacks/canary_insertion/nodp_normalized/nodp_normalized_seed123.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210426/234146 --cuda cuda:3 --outputf attacks/canary_insertion/nodp_normalized/nodp_normalized_seed22.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210426/234201 --cuda cuda:3 --outputf attacks/canary_insertion/nodp_normalized/nodp_normalized_seed300.csv
# canary, for param search, on interaction
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210425/150805 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.1_norm0.005_seed1111_resume_50epochs.csv
# canary, dp, on interaction
python attacks/canary_insertion.py -bs 256 --checkpoint model/dp/20210425/151315 --cuda cuda:3 --outputf attacks/canary_insertion/dp/lr0.05_sigma0.5_norm0.1_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/dp/20210425/151340 --cuda cuda:3 --outputf attacks/canary_insertion/dp/lr0.05_sigma0.5_norm0.1_seed0.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/dp/20210425/151359 --cuda cuda:3 --outputf attacks/canary_insertion/dp/lr0.05_sigma0.5_norm0.1_seed123.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/dp/20210425/151417 --cuda cuda:3 --outputf attacks/canary_insertion/dp/lr0.05_sigma0.5_norm0.1_seed22.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/dp/20210425/151429 --cuda cuda:5 --outputf attacks/canary_insertion/dp/lr0.05_sigma0.5_norm0.1_seed300.csv
# canary insertion, partial dp, on interaction
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210421/123500 --cuda cuda:5 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.5_norm0.001_seed0.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210421/123511 --cuda cuda:5 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.5_norm0.001_seed123.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210421/123522 --cuda cuda:5 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.5_norm0.001_seed22.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210421/123530 --cuda cuda:6 --outputf attacks/canary_insertion/partialdp/lr0.1_sigma0.5_norm0.001_seed300.csv
# canary insertion, partial dp, on interaction, resume
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210427/213802 --cuda cuda:6 --outputf attacks/canary_insertion/partialdp/resume/lr0.1_sigma0.5_norm0.001_seed123_resume50epochs.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210427/213754 --cuda cuda:6 --outputf attacks/canary_insertion/partialdp/resume/lr0.1_sigma0.5_norm0.001_seed22_resume50epochs.csv
# dialog, parameter tuning, on dialog server
# screen -R dialog_partialdp
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:1 -dp -partial -bs 3 --sigma 0.7 -norm 1e-1 --epochs 50 2>&1 | tee logs/partial_dp/dialog/20210429/sigma0.7_norm1e-1
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:1 -dp -partial -bs 3 --sigma 0.7 -norm 1e-2 --epochs 50 2>&1 | tee logs/partial_dp/dialog/20210429/sigma0.7_norm1e-2
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:0 -dp -partial -bs 3 --sigma 0.6 -norm 1e-1 --epochs 50 2>&1 | tee logs/partial_dp/dialog/20210429/sigma0.6_norm1e-1
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:0 -dp -partial -bs 3 --sigma 0.6 -norm 1e-2 --epochs 50 2>&1 | tee logs/partial_dp/dialog/20210429/sigma0.6_norm1e-2
# canary, on dialog, resume
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210423/221533 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/resume/lr0.1_sigma0.5_norm0.001_seed0_resume50epochs.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210427/213900 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/resume/lr0.1_sigma0.5_norm0.001_seed300_resume50epochs.csv
# dialog, parameter tuning, on dialog server, full dp
# screen -R dialog_dp
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:1 -dp -bs 3 --sigma 0.6 -norm 1e-2 --epochs 50 2>&1 | tee logs/dp/dialog/20210430/sigma0.6_norm1e-2
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:1 -dp -bs 1 --sigma 0.6 -norm 1e-2 --epochs 50 2>&1 | tee logs/dp/dialog/20210430/sigma0.6_norm1e-2_bs1
# screen -R dialog_partialdp
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:0 -dp -partial -bs 3 --sigma 0.7 -norm 5e-3 --epochs 50 2>&1 | tee logs/partial_dp/dialog/20210430/sigma0.7_norm5e-3
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:0 -dp -partial -bs 3 --sigma 0.7 -norm 5e-2 --epochs 50 2>&1 | tee logs/partial_dp/dialog/20210430/sigma0.7_norm5e-2
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:0 -dp -partial -bs 3 --sigma 0.7 -norm 1e-3 --epochs 50 2>&1 | tee logs/partial_dp/dialog/20210430/sigma0.7_norm1e-3
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:6 -dp -partial -bs 3 --sigma 0.7 -norm 5e-4 --epochs 50 2>&1 | tee logs/partial_dp/dialog/20210430/sigma0.7_norm5e-4
# dialog, parameter tuning, on dialog server, full dp
# screen -R dialog_dp
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:1 -dp -bs 3 --sigma 0.6 -norm 1e-2 --epochs 50 2>&1 | tee logs/dp/dialog/20210430/sigma0.6_norm1e-2
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:1 -dp -bs 3 --sigma 0.6 -norm 5e-2 --epochs 50 2>&1 | tee logs/dp/dialog/20210430/sigma0.6_norm5e-2
# dialog, full dp, 100 epochs
# python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:1 -dp -bs 1 --sigma 0.6 -norm 1e-2 --epochs 50 -resume -resume_from_epoch_num 50 -resume_from model/dp/20210501/095703/data-simdial_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50260_bs-1_bptt-35_lr-0.1_dp-True_partial-False_0hidden-False_sigma-0.6_norm-0.01_dl-8e-05.pt_ppl-20.8570108_acc-0.61681_epoch-50_ep-2.232_dl-8e-05_ap-5.90 2>&1 | tee logs/dp/dialog/20210430/sigma0.6_norm1e-2_bs1_resume50epochs
python -u main.py --lr 0.1 --data data/simdial --data_type dial --cuda cuda:1 -dp -bs 1 --sigma 0.6 -norm 1e-2 --epochs 100 2>&1 | tee logs/dp/dialog/20210430/sigma0.6_norm1e-2_bs1_100epochs
# on interaction server, canary, for missed partial dp
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210427/211240 --cuda cuda:1 --outputf attacks/canary_insertion/partialdp_missed/lr0.1_sigma0.5_norm0.001_seed0.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210427/211322 --cuda cuda:2 --outputf attacks/canary_insertion/partialdp_missed/lr0.1_sigma0.5_norm0.001_seed123.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210427/211339 --cuda cuda:4 --outputf attacks/canary_insertion/partialdp_missed/lr0.1_sigma0.5_norm0.001_seed22.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210427/211411 --cuda cuda:5 --outputf attacks/canary_insertion/partialdp_missed/lr0.1_sigma0.5_norm0.001_seed300.csv
# dialog experiment, baseline, on interaction server
# screen -R dialog_nodp
python -u main.py -bs 16 --lr 20 --data data/simdial --data_type dial --cuda cuda:1 --log-interval 10 --seed 1111 2>&1 | tee logs/nodp/dialog/20210501/dialog_bs16_seed1111.log
python -u main.py -bs 16 --lr 20 --data data/simdial --data_type dial --cuda cuda:1 --log-interval 10 --seed 0 2>&1 | tee logs/nodp/dialog/20210501/dialog_bs16_seed0.log
python -u main.py -bs 16 --lr 20 --data data/simdial --data_type dial --cuda cuda:2 --log-interval 10 --seed 123 2>&1 | tee logs/nodp/dialog/20210501/dialog_bs16_seed123.log
python -u main.py -bs 16 --lr 20 --data data/simdial --data_type dial --cuda cuda:4 --log-interval 10 --seed 22 2>&1 | tee logs/nodp/dialog/20210501/dialog_bs16_seed22.log
python -u main.py -bs 16 --lr 20 --data data/simdial --data_type dial --cuda cuda:5 --log-interval 10 --seed 300 2>&1 | tee logs/nodp/dialog/20210501/dialog_bs16_seed300.log
# dialog canary insertion
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210429/234246 --cuda cuda:0 --data_type dial --outputf attacks/canary_insertion/dialog/partialdp/lr0.1_sigma0.7_norm0.01_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210430/230016 --cuda cuda:0 --data_type dial --outputf attacks/canary_insertion/dialog/partialdp/lr0.1_sigma0.7_norm0.005_seed1111.csv
# dialog, canary insertion
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210501/192118 --cuda cuda:0 --data_type dial --outputf attacks/canary_insertion/dialog/nodp/nodp_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210502/124723 --cuda cuda:0 --data_type dial --outputf attacks/canary_insertion/dialog/nodp/nodp_seed0.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210501/192209 --cuda cuda:0 --data_type dial --outputf attacks/canary_insertion/dialog/nodp/nodp_seed123.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210501/192230 --cuda cuda:0 --data_type dial --outputf attacks/canary_insertion/dialog/nodp/nodp_seed22.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210501/192246 --cuda cuda:0 --data_type dial --outputf attacks/canary_insertion/dialog/nodp/nodp_seed300.csv
# dialog, nodp, on interaction
# add 20
python -u main.py -bs 16 --lr 20 --data data/simdial-20 --data_type dial --cuda cuda:0 --log-interval 10 --seed 1111 2>&1 | tee logs/nodp/dialog/20210503/add20/dialog_bs16_seed1111.log
python -u main.py -bs 16 --lr 20 --data data/simdial-20 --data_type dial --cuda cuda:1 --log-interval 10 --seed 0 2>&1 | tee logs/nodp/dialog/20210503/add20/dialog_bs16_seed0.log
python -u main.py -bs 16 --lr 20 --data data/simdial-20 --data_type dial --cuda cuda:2 --log-interval 10 --seed 123 2>&1 | tee logs/nodp/dialog/20210503/add20/dialog_bs16_seed123.log
python -u main.py -bs 16 --lr 20 --data data/simdial-20 --data_type dial --cuda cuda:4 --log-interval 10 --seed 22 2>&1 | tee logs/nodp/dialog/20210503/add20/dialog_bs16_seed22.log
python -u main.py -bs 16 --lr 20 --data data/simdial-20 --data_type dial --cuda cuda:5 --log-interval 10 --seed 300 2>&1 | tee logs/nodp/dialog/20210503/add20/dialog_bs16_seed300.log
# add 30
python -u main.py -bs 16 --lr 20 --data data/simdial-30 --data_type dial --cuda cuda:0 --log-interval 10 --seed 1111 2>&1 | tee logs/nodp/dialog/20210503/add30/dialog_bs16_seed1111.log
python -u main.py -bs 16 --lr 20 --data data/simdial-30 --data_type dial --cuda cuda:1 --log-interval 10 --seed 0 2>&1 | tee logs/nodp/dialog/20210503/add30/dialog_bs16_seed0.log
python -u main.py -bs 16 --lr 20 --data data/simdial-30 --data_type dial --cuda cuda:2 --log-interval 10 --seed 123 2>&1 | tee logs/nodp/dialog/20210503/add30/dialog_bs16_seed123.log
python -u main.py -bs 16 --lr 20 --data data/simdial-30 --data_type dial --cuda cuda:4 --log-interval 10 --seed 22 2>&1 | tee logs/nodp/dialog/20210503/add30/dialog_bs16_seed22.log
python -u main.py -bs 16 --lr 20 --data data/simdial-30 --data_type dial --cuda cuda:6 --log-interval 10 --seed 300 2>&1 | tee logs/nodp/dialog/20210503/add30/dialog_bs16_seed300.log
# add 10 inside
python -u main.py -bs 16 --lr 20 --data data/simdial-in-10 --data_type dial --cuda cuda:0 --log-interval 10 --seed 1111 2>&1 | tee logs/nodp/dialog/20210503/add10inside/dialog_bs16_seed1111.log
python -u main.py -bs 16 --lr 20 --data data/simdial-in-10 --data_type dial --cuda cuda:1 --log-interval 10 --seed 0 2>&1 | tee logs/nodp/dialog/20210503/add10inside/dialog_bs16_seed0.log
python -u main.py -bs 16 --lr 20 --data data/simdial-in-10 --data_type dial --cuda cuda:2 --log-interval 10 --seed 123 2>&1 | tee logs/nodp/dialog/20210503/add10inside/dialog_bs16_seed123.log
python -u main.py -bs 16 --lr 20 --data data/simdial-in-10 --data_type dial --cuda cuda:4 --log-interval 10 --seed 22 2>&1 | tee logs/nodp/dialog/20210503/add10inside/dialog_bs16_seed22.log
python -u main.py -bs 16 --lr 20 --data data/simdial-in-10 --data_type dial --cuda cuda:5 --log-interval 10 --seed 300 2>&1 | tee logs/nodp/dialog/20210503/add10inside/dialog_bs16_seed300.log
################################################
# membership running, wiki
################################################
# nodp
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210418/181252/ --cuda cuda:0 --outputf attacks/membership_inference/nodp/final_fix/nodp_seed1111.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210418/192226/ --cuda cuda:0 --outputf attacks/membership_inference/nodp/final_fix/nodp_seed0.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210417/144949/ --cuda cuda:0 --outputf attacks/membership_inference/nodp/final_fix/nodp_seed123.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210419/052248/ --cuda cuda:0 --outputf attacks/membership_inference/nodp/final_fix/nodp_seed22.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210418/210231/ --cuda cuda:0 --outputf attacks/membership_inference/nodp/final_fix/nodp_seed300.csv
# dp
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/dp/20210425/151315 --cuda cuda:0 --outputf attacks/membership_inference/dp/final_fix/lr0.05_sigma0.5_norm0.1_seed1111.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/dp/20210425/151340 --cuda cuda:0 --outputf attacks/membership_inference/dp/final_fix/lr0.05_sigma0.5_norm0.1_seed0.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/dp/20210425/151359 --cuda cuda:0 --outputf attacks/membership_inference/dp/final_fix/lr0.05_sigma0.5_norm0.1_seed123.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/dp/20210425/151417 --cuda cuda:0 --outputf attacks/membership_inference/dp/final_fix/lr0.05_sigma0.5_norm0.1_seed22.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/dp/20210425/151429 --cuda cuda:0 --outputf attacks/membership_inference/dp/final_fix/lr0.05_sigma0.5_norm0.1_seed300.csv
# partial dp, first 50
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210418/191438 --cuda cuda:1 --outputf attacks/membership_inference/partialdp/final_fix/lr0.1_sigma0.5_norm0.001_seed1111.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210421/123500 --cuda cuda:1 --outputf attacks/membership_inference/partialdp/final_fix/lr0.1_sigma0.5_norm0.001_seed0.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210421/123511 --cuda cuda:1 --outputf attacks/membership_inference/partialdp/final_fix/lr0.1_sigma0.5_norm0.001_seed123.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210421/123522 --cuda cuda:1 --outputf attacks/membership_inference/partialdp/final_fix/lr0.1_sigma0.5_norm0.001_seed22.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210421/123530 --cuda cuda:1 --outputf attacks/membership_inference/partialdp/final_fix/lr0.1_sigma0.5_norm0.001_seed300.csv
# partial dp, resume 50
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210427/213802 --cuda cuda:1 --outputf attacks/membership_inference/partialdp/final_fix/resume/lr0.1_sigma0.5_norm0.001_seed123.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210427/213754 --cuda cuda:1 --outputf attacks/membership_inference/partialdp/final_fix/resume/lr0.1_sigma0.5_norm0.001_seed22.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210427/213900 --cuda cuda:1 --outputf attacks/membership_inference/partialdp/final_fix/resume/lr0.1_sigma0.5_norm0.001_seed300.csv
# no dp normalized,
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210426/182422 --cuda cuda:2 --outputf attacks/membership_inference/nodp_normalized/final_fix/nodp_seed1111.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210426/234106 --cuda cuda:2 --outputf attacks/membership_inference/nodp_normalized/final_fix/nodp_seed0.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210426/234125 --cuda cuda:2 --outputf attacks/membership_inference/nodp_normalized/final_fix/nodp_seed123.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210426/234146 --cuda cuda:2 --outputf attacks/membership_inference/nodp_normalized/final_fix/nodp_seed22.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210426/234201 --cuda cuda:2 --outputf attacks/membership_inference/nodp_normalized/final_fix/nodp_seed300.csv
# missed dp normalized
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210427/211240 --cuda cuda:2 --outputf attacks/membership_inference/partialdp_missed/final_fix/lr0.1_sigma0.5_norm0.001_seed0.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210427/211322 --cuda cuda:2 --outputf attacks/membership_inference/partialdp_missed/final_fix/lr0.1_sigma0.5_norm0.001_seed123.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210427/211411 --cuda cuda:2 --outputf attacks/membership_inference/partialdp_missed/final_fix/lr0.1_sigma0.5_norm0.001_seed300.csv
# orange ones, param search
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210417/000742 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.45_norm0.01_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210417/000240 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.45_norm0.005_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210417/000251 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.5_norm0.01_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210417/111003 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.5_norm0.005_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210417/000327 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.55_norm0.01_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210417/000344 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.55_norm0.005_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210419/063839 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.5_norm0.0005_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210419/134334 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.5_norm0.0001_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210419/064057 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.5_norm5e-05_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210419/134345 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.5_norm1e-05_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210419/134357 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.5_norm5e-06_seed1111.csv
# more param tunning
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210423/111019 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.1_norm0.005_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210423/111038 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.1_norm0.25_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210423/111051 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.05_norm0.005_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210423/111126 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.05_norm0.25_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210423/134414 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.01_norm0.005_seed1111.csv
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210423/183122 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/lr0.1_sigma0.01_norm0.25_seed1111.csv
# more param tunning, resume
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210425/150805 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/final_fix/param_search/resume/lr0.1_sigma0.1_norm0.005_seed1111.csv
##################
# on dialog server
##################
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210423/221533 --cuda cuda:1 --outputf attacks/membership_inference/partialdp/final_fix/resume/lr0.1_sigma0.5_norm0.001_seed0.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210423/221514 --cuda cuda:1 --outputf attacks/membership_inference/partialdp/final_fix/resume/lr0.1_sigma0.5_norm0.001_seed1111.csv
# missed dp normalized, on dialog server
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210426/223009 --cuda cuda:1 --outputf attacks/membership_inference/partialdp_missed/final_fix/lr0.1_sigma0.5_norm0.001_seed22.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210427/211339 --cuda cuda:1 --outputf attacks/membership_inference/partialdp_missed/final_fix/lr0.1_sigma0.5_norm0.001_seed300.csv
# param tunning, on dialog server
# dp
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/dp/20210423/224844 --cuda cuda:1 --outputf attacks/membership_inference/dp/final_fix/lr0.1_sigma0.25_norm0.1_seed1111.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/dp/20210423/224945 --cuda cuda:1 --outputf attacks/membership_inference/dp/final_fix/lr0.1_sigma0.5_norm0.1_seed1111.csv
# dialog experiments
################################################
# membership running, dialog
################################################
# dialog
python attacks/mem_inference.py --data_type dial --data data/simdial --path0 attacks/membership_inference/candidates/dialog/test --path1 attacks/membership_inference/candidates/dialog/train -bs 64 --N 1000 --checkpoint model/nodp/20210503/220336 --cuda cuda:0 --outputf attacks/membership_inference/dialog/nodp/final_fix/nodp_seed1111.csv
python attacks/mem_inference.py --data_type dial --data data/simdial --path0 attacks/membership_inference/candidates/dialog/test --path1 attacks/membership_inference/candidates/dialog/train -bs 64 --N 1000 --checkpoint model/partialdp/20210503/230904 --cuda cuda:0 --outputf attacks/membership_inference/dialog/partialdp/final_fix/nodp_seed1111.csv
# canary insertion for partial dp, param search
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210423/111019 --cuda cuda:0 --outputf attacks/canary_insertion/partialdp/param_search/lr0.1_sigma0.1_norm0.005_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210423/111038 --cuda cuda:1 --outputf attacks/canary_insertion/partialdp/param_search/lr0.1_sigma0.1_norm0.25_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210423/111051 --cuda cuda:2 --outputf attacks/canary_insertion/partialdp/param_search/lr0.1_sigma0.05_norm0.005_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210423/111126 --cuda cuda:3 --outputf attacks/canary_insertion/partialdp/param_search/lr0.1_sigma0.05_norm0.25_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210423/134414 --cuda cuda:4 --outputf attacks/canary_insertion/partialdp/param_search/lr0.1_sigma0.01_norm0.005_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210423/183122 --cuda cuda:5 --outputf attacks/canary_insertion/partialdp/param_search/lr0.1_sigma0.01_norm0.25_seed1111.csv
#=============================
#membership
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210423/111019 --cuda cuda:0 --outputf attacks/membership_inference/partialdp/param_search/lr0.1_sigma0.1_norm0.005_seed1111.csv --data_type doc --data data/wikitext-2-add10b
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210423/111038 --cuda cuda:1 --outputf attacks/membership_inference/partialdp/param_search/lr0.1_sigma0.1_norm0.25_seed1111.csv --data_type doc --data data/wikitext-2-add10b
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210423/111051 --cuda cuda:2 --outputf attacks/membership_inference/partialdp/param_search/lr0.1_sigma0.05_norm0.005_seed1111.csv --data_type doc --data data/wikitext-2-add10b
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210423/111126 --cuda cuda:3 --outputf attacks/membership_inference/partialdp/param_search/lr0.1_sigma0.05_norm0.25_seed1111.csv --data_type doc --data data/wikitext-2-add10b
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210423/134414 --cuda cuda:4 --outputf attacks/membership_inference/partialdp/param_search/lr0.1_sigma0.01_norm0.005_seed1111.csv --data_type doc --data data/wikitext-2-add10b
python attacks/mem_inference.py -bs 64 --N 1000 --checkpoint model/partialdp/20210423/183122 --cuda cuda:5 --outputf attacks/membership_inference/partialdp/param_search/lr0.1_sigma0.01_norm0.25_seed1111.csv --data_type doc --data data/wikitext-2-add10b
######################################
# missing digits, miss only 1
######################################
# missing digit, baseline, on interaction
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add1b-normalized/missing_digits --cuda cuda:3 2>&1 | tee logs/nodp/normalized/miss_1/lstm.log
# screen -R nodp2
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add1b-normalized/missing_digits --cuda cuda:3 --seed 0 2>&1 | tee logs/nodp/normalized/miss_1/lstm_seed0.log
# screen -R nodp3
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add1b-normalized/missing_digits --cuda cuda:3 --seed 123 2>&1 | tee logs/nodp/normalized/miss_1/lstm_seed123.log
# screen -R nodp4
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add1b-normalized/missing_digits --cuda cuda:3 --seed 22 2>&1 | tee logs/nodp/normalized/miss_1/lstm_seed22.log
# screen -R nodp5
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add1b-normalized/missing_digits --cuda cuda:3 --seed 300 2>&1 | tee logs/nodp/normalized/miss_1/lstm_seed300.log
# missing digit, partial dp, on interaction
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:1 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add1b --epochs 100 --seed 1111 2>&1 | tee logs/partial_dp/missed/miss_1/lr0.1_sigm0.5_norm1e-3_seed1111_miss1.log
# missing digit, partial dp, on interaction
### screen -R partialdp1
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:2 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add1b --epochs 100 --seed 0 2>&1 | tee logs/partial_dp/missed/miss_1/lr0.1_sigm0.5_norm1e-3_seed0_miss1.log
### screen -R partialdp2
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:4 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add1b --epochs 100 --seed 123 2>&1 | tee logs/partial_dp/missed/miss_1/lr0.1_sigm0.5_norm1e-3_seed123_miss1.log
### screen -R partialdp3
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:5 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add1b --epochs 100 --seed 22 2>&1 | tee logs/partial_dp/missed/miss_1/lr0.1_sigm0.5_norm1e-3_seed22_miss1.log
### screen -R partialdp4
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:6 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add1b --epochs 100 --seed 300 2>&1 | tee logs/partial_dp/missed/miss_1/lr0.1_sigm0.5_norm1e-3_seed300_miss1.log
# no dp normalized, miss 1
# screen -r d7
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210506/234706 --cuda cuda:0 --outputf attacks/membership_inference/nodp_normalized_miss1/nodp_seed1111.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210507/044232 --cuda cuda:0 --outputf attacks/membership_inference/nodp_normalized_miss1/nodp_seed0.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210507/094710 --cuda cuda:0 --outputf attacks/membership_inference/nodp_normalized_miss1/nodp_seed123.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210507/141548 --cuda cuda:3 --outputf attacks/membership_inference/nodp_normalized_miss1/nodp_seed22.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210507/192746 --cuda cuda:3 --outputf attacks/membership_inference/nodp_normalized_miss1/nodp_seed300.csv
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:0 -model_dir model/nodp/20210506/234706
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:0 -model_dir model/nodp/20210507/044232
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:0 -model_dir model/nodp/20210507/094710
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:3 -model_dir model/nodp/20210507/141548
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:3 -model_dir model/nodp/20210507/192746
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210506/234706 --cuda cuda:0 --outputf attacks/canary_insertion/nodp_normalized_miss1/nodp_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210507/044232 --cuda cuda:0 --outputf attacks/canary_insertion/nodp_normalized_miss1/nodp_seed0.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210507/094710 --cuda cuda:0 --outputf attacks/canary_insertion/nodp_normalized_miss1/nodp_seed123.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210507/141548 --cuda cuda:3 --outputf attacks/canary_insertion/nodp_normalized_miss1/nodp_seed22.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210507/192746 --cuda cuda:3 --outputf attacks/canary_insertion/nodp_normalized_miss1/nodp_seed300.csv
######### missing experiments
# 1. append another secret,
# already run on cuda:0
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:1 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add10b-missed-append10 --epochs 100 --seed 1111 2>&1 | tee logs/partial_dp/missed/append/lr0.1_sigm0.5_norm1e-3_seed1111_miss10.log
### already run screen -r d6
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:2 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add10b-missed-append10 --epochs 100 --seed 22 2>&1 | tee logs/partial_dp/missed/append/lr0.1_sigm0.5_norm1e-3_seed22_miss10.log
### screen -R partialdp2,
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:3 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add10b-missed-append10 --epochs 100 --seed 123 2>&1 | tee logs/partial_dp/missed/append/lr0.1_sigm0.5_norm1e-3_seed123_miss10.log
# 2. use both unk and digits as secrets
### screen -R partialdp3
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:4 -partial -norm 1e-3 --sigma 0.5 -missing_digits -digits_unk_as_private --data data/wikitext-2-add10b --epochs 100 --seed 22 2>&1 | tee logs/partial_dp/missed/both_unk_digits/lr0.1_sigm0.5_norm1e-3_seed22_miss10.log
# run already on cuda:5
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:5 -partial -norm 1e-3 --sigma 0.5 -missing_digits -digits_unk_as_private --data data/wikitext-2-add10b --epochs 100 --seed 300 2>&1 | tee logs/partial_dp/missed/both_unk_digits/lr0.1_sigm0.5_norm1e-3_seed300_miss10.log
# run already on cuda: 1
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:1 -partial -norm 1e-3 --sigma 0.5 -missing_digits -digits_unk_as_private --data data/wikitext-2-add10b --epochs 100 --seed 1111 2>&1 | tee logs/partial_dp/missed/both_unk_digits/lr0.1_sigm0.5_norm1e-3_seed1111_miss10.log
### missing 20
# screen -r d3
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:4 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add20b --epochs 100 --seed 300 2>&1 | tee logs/partial_dp/missed/miss_20/lr0.1_sigm0.5_norm1e-3_seed300_miss20.log
# screen -r d5
python -u main.py -bs 7 --lr 0.1 -dp --cuda cuda:6 -partial -norm 1e-3 --sigma 0.5 -missing_digits --data data/wikitext-2-add20b --epochs 100 --seed 1111 2>&1 | tee logs/partial_dp/missed/miss_20/lr0.1_sigm0.5_norm1e-3_seed1111_miss20.log
# missing digit, baseline, on interaction
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-missed-append10-normalized/missing_digits --cuda cuda:3 2>&1 | tee logs/nodp/normalized-append10/20210511/lstm.log
# screen -R nodp2
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-missed-append10-normalized/missing_digits --cuda cuda:3 --seed 0 2>&1 | tee logs/nodp/normalized-append10/20210511/lstm_seed0.log
# screen -R nodp3
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-missed-append10-normalized/missing_digits --cuda cuda:3 --seed 123 2>&1 | tee logs/nodp/normalized-append10/20210511/lstm_seed123.log
# screen -R nodp4
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-missed-append10-normalized/missing_digits --cuda cuda:3 --seed 22 2>&1 | tee logs/nodp/normalized-append10/20210511/lstm_seed22.log
# screen -R nodp5
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-missed-append10-normalized/missing_digits --cuda cuda:3 --seed 300 2>&1 | tee logs/nodp/normalized-append10/20210511/lstm_seed300.log
# missing digit, baseline, on interaction
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add20b-normalized/missing_digits --cuda cuda:3 2>&1 | tee logs/nodp/normalized-missing20/20210511/lstm.log
# screen -R nodp2
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add20b-normalized/missing_digits --cuda cuda:3 --seed 0 2>&1 | tee logs/nodp/normalized-missing20/20210511/lstm_seed0.log
# screen -R nodp3
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add20b-normalized/missing_digits --cuda cuda:3 --seed 123 2>&1 | tee logs/nodp/normalized-missing20/20210511/lstm_seed123.log
# screen -R nodp4
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add20b-normalized/missing_digits --cuda cuda:3 --seed 22 2>&1 | tee logs/nodp/normalized-missing20/20210511/lstm_seed22.log
# screen -R nodp5
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add20b-normalized/missing_digits --cuda cuda:3 --seed 300 2>&1 | tee logs/nodp/normalized-missing20/20210511/lstm_seed300.log
#
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210506/234027 --cuda cuda:3 --outputf attacks/membership_inference/partialdp_missed_1/lr0.1_sigma0.5_norm0.001_seed1111.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210506/234042 --cuda cuda:3 --outputf attacks/membership_inference/partialdp_missed_1/lr0.1_sigma0.5_norm0.001_seed123.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210506/234049 --cuda cuda:3 --outputf attacks/membership_inference/partialdp_missed_1/lr0.1_sigma0.5_norm0.001_seed22.csv
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210506/234057 --cuda cuda:3 --outputf attacks/membership_inference/partialdp_missed_1/lr0.1_sigma0.5_norm0.001_seed300.csv
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:3 -model_dir model/partialdp/20210506/234027
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:3 -model_dir model/partialdp/20210506/234042
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:3 -model_dir model/partialdp/20210506/234049
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:3 -model_dir model/partialdp/20210506/234057
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210506/234027 --cuda cuda:3 --outputf attacks/canary_insertion/partialdp_missed_1/lr0.1_sigma0.5_norm0.001_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210506/234042 --cuda cuda:3 --outputf attacks/canary_insertion/partialdp_missed_1/lr0.1_sigma0.5_norm0.001_seed123.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210506/234049 --cuda cuda:3 --outputf attacks/canary_insertion/partialdp_missed_1/lr0.1_sigma0.5_norm0.001_seed22.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210506/234057 --cuda cuda:3 --outputf attacks/canary_insertion/partialdp_missed_1/lr0.1_sigma0.5_norm0.001_seed300.csv
python -u scripts/adjust_ppl_acc.py -bs 64 --cuda cuda:3 -model_dir model/partialdp/20210507/221838
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210507/221838 --cuda cuda:3 --outputf attacks/membership_inference/partialdp_missed_1/lr0.1_sigma0.5_norm0.001_seed100.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210507/221838 --cuda cuda:3 --outputf attacks/canary_insertion/partialdp_missed_1/lr0.1_sigma0.5_norm0.001_seed100.csv
# screen -r d0
python -u main.py -bs 4 --lr 2 --data data/simdial --data_type dial --cuda cuda:0 --seed 1111 --epochs 50 -save_epoch_num 1 --log-interval 50 2>&1 | tee logs/nodp/dialog/20210515/add10_lr2_bs4/dialog_bs4_lr2_seed1111.log
# screen -r d00
python -u main.py -bs 4 --lr 2 --data data/simdial --data_type dial --cuda cuda:0 --seed 0 --epochs 50 -save_epoch_num 1 --log-interval 50 2>&1 | tee logs/nodp/dialog/20210515/add10_lr2_bs4/dialog_bs4_lr2_seed0.log
# screen -r d1seed123
python -u main.py -bs 4 --lr 2 --data data/simdial --data_type dial --cuda cuda:1 --seed 123 --epochs 50 -save_epoch_num 1 --log-interval 50 2>&1 | tee logs/nodp/dialog/20210515/add10_lr2_bs4/dialog_bs4_lr2_seed123.log
# screen -r d2seed22
python -u main.py -bs 4 --lr 2 --data data/simdial --data_type dial --cuda cuda:2 --seed 22 --epochs 50 -save_epoch_num 1 --log-interval 50 2>&1 | tee logs/nodp/dialog/20210515/add10_lr2_bs4/dialog_bs4_lr2_seed22.log
# screen -r d2seed300
python -u main.py -bs 4 --lr 2 --data data/simdial --data_type dial --cuda cuda:3 --seed 300 --epochs 50 -save_epoch_num 1 --log-interval 50 2>&1 | tee logs/nodp/dialog/20210515/add10_lr2_bs4/dialog_bs4_lr2_seed300.log
# manual check
# both
# compare with attacks/canary_insertion/nodp_normalized/nodp_normalized_seed0.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210510/101057/data-wikitext-2-add10b_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.5_norm-0.001_dl-8e-05.pt_ppl-257.9607115_acc-0.28376_epoch-16_ep-4.343_dl-8e-05_ap-3.80 --cuda cuda:3 --outputf attacks/canary_insertion/partialdp_missed_both_unk_digits/test/ppl258.csv
# append
# nodp
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210511/230530/data-missing_digits_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-16_bptt-35_lr-20.0_dp-False_partial-False_0hidden-False.pt_ppl-204.8520007_acc-0.28234_epoch-1_ep-0.000_dl-0_ap-0.00 --cuda cuda:3 --outputf attacks/canary_insertion/nodp_normalized_append10/test/ppl204.csv
# nodp
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210511/230530/data-missing_digits_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-16_bptt-35_lr-20.0_dp-False_partial-False_0hidden-False.pt_ppl-252.3595410_acc-0.25790_epoch-1_ep-0.000_dl-0_ap-0.00 --cuda cuda:3 --outputf attacks/canary_insertion/nodp_normalized_append10/test/ppl252.csv
# sdp
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210511/164815/data-wikitext-2-add10b-missed-append10_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.5_norm-0.001_dl-8e-05.pt_ppl-254.9234443_acc-0.27307_epoch-8_ep-3.624_dl-8e-05_ap-4.00 --cuda cuda:3 --outputf attacks/canary_insertion/partialdp_missed_append/test/ppl252.csv
#### for the missing experiments
# append
python -u scripts/adjust_ppl_acc.py -bs 64 --cuda cuda:0 -model_dir model/partialdp/20210511/164815/
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210511/164815/ --cuda cuda:0 --outputf attacks/membership_inference/partialdp_missed_append/lr0.1_sigma0.5_norm0.001_seed22.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/partialdp/20210511/164815/ --cuda cuda:0 --outputf attacks/canary_insertion/partialdp_missed_append/lr0.1_sigma0.5_norm0.001_seed22.csv
python -u scripts/adjust_ppl_acc.py --cuda cuda:1 -bs 64 -model_dir model/partialdp/20210510/110828/
python attacks/mem_inference.py --cuda cuda:1 --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210510/110828/ --outputf attacks/membership_inference/partialdp_missed_append/lr0.1_sigma0.5_norm0.001_seed1111.csv
python attacks/canary_insertion.py --cuda cuda:1 -bs 256 --checkpoint model/partialdp/20210510/110828/ --outputf attacks/canary_insertion/partialdp_missed_append/lr0.1_sigma0.5_norm0.001_seed1111.csv
# both unk and digit
python -u scripts/adjust_ppl_acc.py --cuda cuda:2 -bs 64 -model_dir model/partialdp/20210510/131908/
python attacks/mem_inference.py --cuda cuda:2 --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210510/131908/ --outputf attacks/membership_inference/partialdp_missed_both_unk_digits/lr0.1_sigma0.5_norm0.001_seed300.csv
python attacks/canary_insertion.py --cuda cuda:2 -bs 256 --checkpoint model/partialdp/20210510/131908/ --outputf attacks/canary_insertion/partialdp_missed_both_unk_digits/lr0.1_sigma0.5_norm0.001_seed300.csv
python -u scripts/adjust_ppl_acc.py --cuda cuda:4 -bs 64 -model_dir model/partialdp/20210510/101057/
python attacks/mem_inference.py --cuda cuda:4 --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210510/101057/ --outputf attacks/membership_inference/partialdp_missed_both_unk_digits/lr0.1_sigma0.5_norm0.001_seed1111.csv
python attacks/canary_insertion.py --cuda cuda:4 -bs 256 --checkpoint model/partialdp/20210510/101057/ --outputf attacks/canary_insertion/partialdp_missed_both_unk_digits/lr0.1_sigma0.5_norm0.001_seed1111.csv
# miss 20
python -u scripts/adjust_ppl_acc.py --cuda cuda:5 -bs 64 -model_dir model/partialdp/20210511/112916/
python attacks/mem_inference.py --cuda cuda:5 --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210511/112916/ --outputf attacks/membership_inference/partialdp_missed_20/lr0.1_sigma0.5_norm0.001_seed300.csv
python attacks/canary_insertion.py --cuda cuda:5 -bs 256 --checkpoint model/partialdp/20210511/112916/ --outputf attacks/canary_insertion/partialdp_missed_20/lr0.1_sigma0.5_norm0.001_seed300.csv
python -u scripts/adjust_ppl_acc.py --cuda cuda:6 -bs 64 -model_dir model/partialdp/20210511/113022/
python attacks/mem_inference.py --cuda cuda:6 --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/partialdp/20210511/113022/ --outputf attacks/membership_inference/partialdp_missed_20/lr0.1_sigma0.5_norm0.001_seed1111.csv
python attacks/canary_insertion.py --cuda cuda:6 -bs 256 --checkpoint model/partialdp/20210511/113022/ --outputf attacks/canary_insertion/partialdp_missed_20/lr0.1_sigma0.5_norm0.001_seed1111.csv
####################################################
# no dp, normalized, baselines
# no dp, normalized, missing 20
python -u scripts/adjust_ppl_acc.py -bs 64 --cuda cuda:0 -model_dir model/nodp/20210514/002133/
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210514/002133/ --cuda cuda:0 --outputf attacks/membership_inference/nodp_normalized_miss20/nodp_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210514/002133/ --cuda cuda:0 --outputf attacks/canary_insertion/nodp_normalized_miss20/nodp_seed1111.csv
python -u scripts/adjust_ppl_acc.py --cuda cuda:1 -bs 64 -model_dir model/nodp/20210511/230530/
python attacks/mem_inference.py --cuda cuda:1 --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210511/230530/ --outputf attacks/membership_inference/nodp_normalized_append10/nodp_seed0.csv
python attacks/canary_insertion.py --cuda cuda:1 -bs 256 --checkpoint model/nodp/20210511/230530/ --outputf attacks/canary_insertion/nodp_normalized_append10/nodp_seed0.csv
# both unk and digit
python -u scripts/adjust_ppl_acc.py --cuda cuda:2 -bs 64 -model_dir model/nodp/20210512/234722/
python attacks/mem_inference.py --cuda cuda:2 --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210512/234722/ --outputf attacks/membership_inference/nodp_normalized_append10/nodp_seed22.csv
python attacks/canary_insertion.py --cuda cuda:2 -bs 256 --checkpoint model/nodp/20210512/234722/ --outputf attacks/canary_insertion/nodp_normalized_append10/nodp_seed22.csv
python -u scripts/adjust_ppl_acc.py --cuda cuda:4 -bs 64 -model_dir model/nodp/20210512/112006/
python attacks/mem_inference.py --cuda cuda:4 --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210512/112006/ --outputf attacks/membership_inference/nodp_normalized_append10/nodp_seed123.csv
python attacks/canary_insertion.py --cuda cuda:4 -bs 256 --checkpoint model/nodp/20210512/112006/ --outputf attacks/canary_insertion/nodp_normalized_append10/nodp_seed123.csv
# miss 20
python -u scripts/adjust_ppl_acc.py --cuda cuda:5 -bs 64 -model_dir model/nodp/20210513/114701/
python attacks/mem_inference.py --cuda cuda:5 --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210513/114701/ --outputf attacks/membership_inference/nodp_normalized_append10/nodp_seed300.csv
python attacks/canary_insertion.py --cuda cuda:5 -bs 256 --checkpoint model/nodp/20210513/114701/ --outputf attacks/canary_insertion/nodp_normalized_append10/nodp_seed300.csv
python -u scripts/adjust_ppl_acc.py --cuda cuda:6 -bs 64 -model_dir model/nodp/20210511/112425/
python attacks/mem_inference.py --cuda cuda:6 --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210511/112425/ --outputf attacks/membership_inference/nodp_normalized_append10/nodp_seed1111.csv
python attacks/canary_insertion.py --cuda cuda:6 -bs 256 --checkpoint model/nodp/20210511/112425/ --outputf attacks/canary_insertion/nodp_normalized_append10/nodp_seed1111.csv
# not missing digit, baseline, sanitization on interaction
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-normalized/not_missing_digits --cuda cuda:0 2>&1 | tee logs/nodp/normalized/not_miss/20210515/lstm.log
# screen -R nodp2
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-normalized/not_missing_digits --cuda cuda:1 --seed 0 2>&1 | tee logs/nodp/normalized/not_miss/20210515/lstm_seed0.log
# screen -R nodp3
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-normalized/not_missing_digits --cuda cuda:2 --seed 123 2>&1 | tee logs/nodp/normalized/not_miss/20210515/lstm_seed123.log
# screen -R nodp4
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-normalized/not_missing_digits --cuda cuda:4 --seed 22 2>&1 | tee logs/nodp/normalized/not_miss/20210515/lstm_seed22.log
# screen -R nodp5
python -u main.py -bs 16 --lr 20 --data data/wikitext-2-add10b-normalized/not_missing_digits --cuda cuda:5 --seed 300 2>&1 | tee logs/nodp/normalized/not_miss/20210515/lstm_seed300.log
python -u main.py -bs 4 --lr 2 --data data/simdial --data_type dial --cuda cuda:3 --seed 1111 --epochs 50 -save_epoch_num 1 --log-interval 50 2>&1 | tee logs/nodp/dialog/20210515/add10_lr2_bs4/dialog_bs4_lr2_seed1111.log
python -u main.py -bs 4 --lr 2 --data data/simdial --data_type dial --cuda cuda:3 --seed 0 --epochs 50 -save_epoch_num 1 --log-interval 50 2>&1 | tee logs/nodp/dialog/20210515/add10_lr2_bs4/dialog_bs4_lr2_seed0.log
python -u main.py -bs 4 --lr 2 --data data/simdial --data_type dial --cuda cuda:3 --seed 123 --epochs 50 -save_epoch_num 1 --log-interval 50 2>&1 | tee logs/nodp/dialog/20210515/add10_lr2_bs4/dialog_bs4_lr2_seed123.log
python -u main.py -bs 4 --lr 2 --data data/simdial --data_type dial --cuda cuda:3 --seed 22 --epochs 50 -save_epoch_num 1 --log-interval 50 2>&1 | tee logs/nodp/dialog/20210515/add10_lr2_bs4/dialog_bs4_lr2_seed22.log
python -u main.py -bs 4 --lr 2 --data data/simdial --data_type dial --cuda cuda:6 --seed 300 --epochs 50 -save_epoch_num 1 --log-interval 50 2>&1 | tee logs/nodp/dialog/20210515/add10_lr2_bs4/dialog_bs4_lr2_seed300.log
##### no dp, normalized, not miss
#### for the missing experiments
model/nodp/20210515/122150
model/nodp/20210515/122210
model/nodp/20210515/122226
model/nodp/20210515/122246
model/nodp/20210515/122306
# append
# screen -r nomiss0
python -u scripts/adjust_ppl_acc.py -bs 64 --cuda cuda:0 -model_dir model/nodp/20210515/122150
python attacks/mem_inference.py --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210515/122150 --cuda cuda:0 --outputf attacks/membership_inference/nodp_normalized_nomiss/nodp_seed1111.csv
python attacks/canary_insertion.py -bs 256 --checkpoint model/nodp/20210515/122150 --cuda cuda:0 --outputf attacks/canary_insertion/nodp_normalized_nomiss/nodp_seed1111.csv
# screen -r nomiss1
python -u scripts/adjust_ppl_acc.py --cuda cuda:1 -bs 64 -model_dir model/nodp/20210515/122210
python attacks/mem_inference.py --cuda cuda:1 --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210515/122210 --outputf attacks/membership_inference/nodp_normalized_nomiss/nodp_seed0.csv
python attacks/canary_insertion.py --cuda cuda:1 -bs 256 --checkpoint model/nodp/20210515/122210 --outputf attacks/canary_insertion/nodp_normalized_nomiss/nodp_seed0.csv
# screen -r nomiss2
python -u scripts/adjust_ppl_acc.py --cuda cuda:2 -bs 64 -model_dir model/nodp/20210515/122226
python attacks/mem_inference.py --cuda cuda:2 --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210515/122226 --outputf attacks/membership_inference/nodp_normalized_nomiss/nodp_seed123.csv
python attacks/canary_insertion.py --cuda cuda:2 -bs 256 --checkpoint model/nodp/20210515/122226 --outputf attacks/canary_insertion/nodp_normalized_nomiss/nodp_seed123.csv
# screen -r nomiss3
python -u scripts/adjust_ppl_acc.py --cuda cuda:3 -bs 64 -model_dir model/nodp/20210515/122246
python attacks/mem_inference.py --cuda cuda:3 --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210515/122246 --outputf attacks/membership_inference/nodp_normalized_nomiss/nodp_seed22.csv
python attacks/canary_insertion.py --cuda cuda:3 -bs 256 --checkpoint model/nodp/20210515/122246 --outputf attacks/canary_insertion/nodp_normalized_nomiss/nodp_seed22.csv
# screen -r nomiss4
python -u scripts/adjust_ppl_acc.py --cuda cuda:4 -bs 64 -model_dir model/nodp/20210515/122306
python attacks/mem_inference.py --cuda cuda:4 --data_type doc --data data/wikitext-2-add10b -bs 64 --N 1000 --checkpoint model/nodp/20210515/122306 --outputf attacks/membership_inference/nodp_normalized_nomiss/nodp_seed300.csv
python attacks/canary_insertion.py --cuda cuda:4 -bs 256 --checkpoint model/nodp/20210515/122306 --outputf attacks/canary_insertion/nodp_normalized_nomiss/nodp_seed300.csv
#### calculate test ppl, for epsilon=4.91, epoch=37
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:0 -outputf data/adjusted_metrics/eps4.91_epoch37 -model_dir model/partialdp/20210421/123500/data-wikitext-2-add10b_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.5_norm-0.001_dl-8e-05.pt_ppl-177.2112399_acc-0.30286_epoch-37_ep-4.911_dl-8e-05_ap-3.70
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:0 -outputf data/adjusted_metrics/eps4.91_epoch37 -model_dir model/partialdp/20210421/123522/data-wikitext-2-add10b_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.5_norm-0.001_dl-8e-05.pt_ppl-174.1074169_acc-0.32000_epoch-37_ep-4.911_dl-8e-05_ap-3.70
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:0 -outputf data/adjusted_metrics/eps4.91_epoch37 -model_dir model/partialdp/20210421/123511/data-wikitext-2-add10b_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.5_norm-0.001_dl-8e-05.pt_ppl-173.9384952_acc-0.33143_epoch-37_ep-4.912_dl-8e-05_ap-3.70
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:0 -outputf data/adjusted_metrics/eps4.91_epoch37 -model_dir model/partialdp/20210421/123530/data-wikitext-2-add10b_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.5_norm-0.001_dl-8e-05.pt_ppl-187.3829027_acc-0.31429_epoch-37_ep-4.912_dl-8e-05_ap-3.70
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:0 -outputf data/adjusted_metrics/eps4.91_epoch37 -model_dir model/partialdp/20210418/191438/data-wikitext-2-add10b_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.5_norm-0.001_dl-8e-05.pt_ppl-176.7718065_acc-0.32000_epoch-37_ep-4.912_dl-8e-05_ap-3.70
| end of epoch 37 | time: 3118.57s | valid loss 5.16 | valid ppl 174.11 | valid acc 0.320
(ε = 4.91, δ = 8e-05) for α = 3.7
-----------------------------------------------------------------------------------------
model saved to model/partialdp/20210421/123522/data-wikitext-2-add10b_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50258_bs-7_bptt-35_lr-0.1_dp-True_partial-True_0hidden-False_sigma-0.5_norm-0.001_dl-8e-05.pt_ppl-174.1074169_acc-0.32000_epoch-37_ep-4.911_dl-8e-05_ap-3.70, ppl: 174.1074169055034
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:0 -outputf data/adjusted_metrics/eps4.91_epoch37 -model_dir model/nodp/20210501/192118/data-simdial_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50260_bs-16_bptt-35_lr-20.0_dp-False_partial-False_0hidden-False.pt_ppl-3.0253297_acc-0.75599_epoch-36_ep-0.000_dl-0_ap-0.00
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:0 -outputf data/adjusted_metrics/eps4.91_epoch37 -model_dir model/nodp/20210501/192118/data-simdial_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50260_bs-16_bptt-35_lr-20.0_dp-False_partial-False_0hidden-False.pt_ppl-3.0253297_acc-0.75599_epoch-36_ep-0.000_dl-0_ap-0.00
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:0 -outputf data/adjusted_metrics/eps4.91_epoch37 -model_dir model/nodp/20210501/192118/data-simdial_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50260_bs-16_bptt-35_lr-20.0_dp-False_partial-False_0hidden-False.pt_ppl-3.0253297_acc-0.75599_epoch-36_ep-0.000_dl-0_ap-0.00
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:0 -outputf data/adjusted_metrics/eps4.91_epoch37 -model_dir model/nodp/20210501/192118/data-simdial_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50260_bs-16_bptt-35_lr-20.0_dp-False_partial-False_0hidden-False.pt_ppl-3.0253297_acc-0.75599_epoch-36_ep-0.000_dl-0_ap-0.00
python -u scripts/adjust_ppl_acc.py -bs 256 --cuda cuda:0 -outputf data/adjusted_metrics/eps4.91_epoch37 -model_dir model/nodp/20210501/192118/data-simdial_model-LSTM_ebd-200_hid-200_bi-False_lay-1_tie-False_tok-50260_bs-16_bptt-35_lr-20.0_dp-False_partial-False_0hidden-False.pt_ppl-3.0253297_acc-0.75599_epoch-36_ep-0.000_dl-0_ap-0.00