File size: 26,620 Bytes
e1aaaac
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# Adapted from https://github.com/fra31/auto-attack/blob/master/autoattack/autopgd_base.py

import time
import torch
import torch.nn as nn
import torch.nn.functional as F
import math
import random

from autoattack.other_utils import L0_norm, L1_norm, L2_norm
from autoattack.checks import check_zero_gradients


def L1_projection(x2, y2, eps1):
    '''
    x2: center of the L1 ball (bs x input_dim)
    y2: current perturbation (x2 + y2 is the point to be projected)
    eps1: radius of the L1 ball

    output: delta s.th. ||y2 + delta||_1 <= eps1
    and 0 <= x2 + y2 + delta <= 1
    '''

    x = x2.clone().float().view(x2.shape[0], -1)
    y = y2.clone().float().view(y2.shape[0], -1)
    sigma = y.clone().sign()
    u = torch.min(1 - x - y, x + y)
    #u = torch.min(u, epsinf - torch.clone(y).abs())
    u = torch.min(torch.zeros_like(y), u)
    l = -torch.clone(y).abs()
    d = u.clone()

    bs, indbs = torch.sort(-torch.cat((u, l), 1), dim=1)
    bs2 = torch.cat((bs[:, 1:], torch.zeros(bs.shape[0], 1).to(bs.device)), 1)

    inu = 2*(indbs < u.shape[1]).float() - 1
    size1 = inu.cumsum(dim=1)

    s1 = -u.sum(dim=1)

    c = eps1 - y.clone().abs().sum(dim=1)
    c5 = s1 + c < 0
    c2 = c5.nonzero().squeeze(1)

    s = s1.unsqueeze(-1) + torch.cumsum((bs2 - bs) * size1, dim=1)

    if c2.nelement != 0:

        lb = torch.zeros_like(c2).float()
        ub = torch.ones_like(lb) *(bs.shape[1] - 1)

        #print(c2.shape, lb.shape)

        nitermax = torch.ceil(torch.log2(torch.tensor(bs.shape[1]).float()))
        counter2 = torch.zeros_like(lb).long()
        counter = 0

        while counter < nitermax:
            counter4 = torch.floor((lb + ub) / 2.)
            counter2 = counter4.type(torch.LongTensor)

            c8 = s[c2, counter2] + c[c2] < 0
            ind3 = c8.nonzero().squeeze(1)
            ind32 = (~c8).nonzero().squeeze(1)
            #print(ind3.shape)
            if ind3.nelement != 0:
                lb[ind3] = counter4[ind3]
            if ind32.nelement != 0:
                ub[ind32] = counter4[ind32]

            #print(lb, ub)
            counter += 1

        lb2 = lb.long()
        alpha = (-s[c2, lb2] -c[c2]) / size1[c2, lb2 + 1] + bs2[c2, lb2]
        d[c2] = -torch.min(torch.max(-u[c2], alpha.unsqueeze(-1)), -l[c2])

    return (sigma * d).view(x2.shape)





class APGDAttack():
    """
    AutoPGD
    https://arxiv.org/abs/2003.01690

    :param predict:       forward pass function
    :param norm:          Lp-norm of the attack ('Linf', 'L2', 'L0' supported)
    :param n_restarts:    number of random restarts
    :param n_iter:        number of iterations
    :param eps:           bound on the norm of perturbations
    :param seed:          random seed for the starting point
    :param loss:          loss to optimize ('ce', 'dlr' supported)
    :param eot_iter:      iterations for Expectation over Trasformation
    :param rho:           parameter for decreasing the step size
    """

    def __init__(
            self,
            predict,
            n_iter=100,
            norm='Linf',
            n_restarts=1,
            eps=None,
            seed=0,
            loss='ce',
            eot_iter=1,
            rho=.75,
            topk=None,
            verbose=False,
            device=None,
            use_largereps=False,
            is_tf_model=False,
            logger=None):
        """
        AutoPGD implementation in PyTorch
        """

        self.model = predict
        self.n_iter = n_iter
        self.eps = eps
        self.norm = norm
        self.n_restarts = n_restarts
        self.seed = seed
        self.loss = loss
        self.eot_iter = eot_iter
        self.thr_decr = rho
        self.topk = topk
        self.verbose = verbose
        self.device = device
        self.use_rs = True
        #self.init_point = None
        self.use_largereps = use_largereps
        #self.larger_epss = None
        #self.iters = None
        self.n_iter_orig = n_iter + 0
        self.eps_orig = eps + 0.
        self.is_tf_model = is_tf_model
        self.y_target = None
        self.logger = logger

        assert self.norm in ['Linf', 'L2', 'L1']
        assert not self.eps is None

        ### set parameters for checkpoints
        self.n_iter_2 = max(int(0.22 * self.n_iter), 1)
        self.n_iter_min = max(int(0.06 * self.n_iter), 1)
        self.size_decr = max(int(0.03 * self.n_iter), 1)

    def init_hyperparam(self, x):

        if self.device is None:
            self.device = x.device
        self.orig_dim = list(x.shape[1:])
        self.ndims = len(self.orig_dim)
        if self.seed is None:
            self.seed = time.time()

    def check_oscillation(self, x, j, k, y5, k3=0.75):
        t = torch.zeros(x.shape[1]).to(self.device)
        for counter5 in range(k):
            t += (x[j - counter5] > x[j - counter5 - 1]).float()

        return (t <= k * k3 * torch.ones_like(t)).float()

    def check_shape(self, x):
        return x if len(x.shape) > 0 else x.unsqueeze(0)

    def normalize(self, x):
        if self.norm == 'Linf':
            t = x.abs().view(x.shape[0], -1).max(1)[0]

        elif self.norm == 'L2':
            t = (x ** 2).view(x.shape[0], -1).sum(-1).sqrt()

        elif self.norm == 'L1':
            try:
                t = x.abs().view(x.shape[0], -1).sum(dim=-1)
            except:
                t = x.abs().reshape([x.shape[0], -1]).sum(dim=-1)

        return x / (t.view(-1, *([1] * self.ndims)) + 1e-12)

    def dlr_loss(self, x, y):
        x_sorted, ind_sorted = x.sort(dim=1)
        ind = (ind_sorted[:, -1] == y).float()
        u = torch.arange(x.shape[0])

        return -(x[u, y] - x_sorted[:, -2] * ind - x_sorted[:, -1] * (
                1. - ind)) / (x_sorted[:, -1] - x_sorted[:, -3] + 1e-12)

    #

    def attack_single_run(self, x, y, x_init=None):
        if len(x.shape) < self.ndims:
            x = x.unsqueeze(0)
            y = y.unsqueeze(0)

        if self.norm == 'Linf':
            t = 2 * torch.rand(x.shape).to(self.device).detach() - 1
            x_adv = x + self.eps * torch.ones_like(x
                                                   ).detach() * self.normalize(t)
        elif self.norm == 'L2':
            t = torch.randn(x.shape).to(self.device).detach()
            x_adv = x + self.eps * torch.ones_like(x
                                                   ).detach() * self.normalize(t)
        elif self.norm == 'L1':
            t = torch.randn(x.shape).to(self.device).detach()
            delta = L1_projection(x, t, self.eps)
            x_adv = x + t + delta





        if not x_init is None:
            x_adv = x_init.clone()
            if self.norm == 'L1' and self.verbose:
                print('[custom init] L1 perturbation {:.5f}'.format(
                    (x_adv - x).abs().view(x.shape[0], -1).sum(1).max()))


        x_adv = x_adv.clamp(0., 1.)
        x_best = x_adv.clone()
        x_best_adv = x_adv.clone()
        loss_steps = torch.zeros([self.n_iter, x.shape[0]]
                                 ).to(self.device)
        loss_best_steps = torch.zeros([self.n_iter + 1, x.shape[0]]
                                      ).to(self.device)
        acc_steps = torch.zeros_like(loss_best_steps)

        if not self.is_tf_model:
            if self.loss == 'ce':
                criterion_indiv = nn.CrossEntropyLoss(reduction='none')
            elif self.loss == 'ce-targeted-cfts':
                criterion_indiv = lambda x, y: -1. * F.cross_entropy(x, y,
                                                                     reduction='none')
            elif self.loss == 'dlr':
                criterion_indiv = self.dlr_loss
            elif self.loss == 'dlr-targeted':
                criterion_indiv = self.dlr_loss_targeted
            elif self.loss == 'ce-targeted':
                criterion_indiv = self.ce_loss_targeted
            else:
                raise ValueError('unknowkn loss')
        else:
            if self.loss == 'ce':
                criterion_indiv = self.model.get_logits_loss_grad_xent
            elif self.loss == 'dlr':
                criterion_indiv = self.model.get_logits_loss_grad_dlr
            elif self.loss == 'dlr-targeted':
                criterion_indiv = self.model.get_logits_loss_grad_target
            else:
                raise ValueError('unknowkn loss')


        x_adv.requires_grad_()
        grad = torch.zeros_like(x)
        for _ in range(self.eot_iter):
            if not self.is_tf_model:
                with torch.enable_grad():
                    logits = self.model(x_adv)
                    loss_indiv = criterion_indiv(logits, y)
                    loss = loss_indiv.sum()

                grad += torch.autograd.grad(loss, [x_adv])[0].detach()
            else:
                if self.y_target is None:
                    logits, loss_indiv, grad_curr = criterion_indiv(x_adv, y)
                else:
                    logits, loss_indiv, grad_curr = criterion_indiv(x_adv, y,
                                                                    self.y_target)
                grad += grad_curr

        grad /= float(self.eot_iter)
        grad_best = grad.clone()

        if self.loss in ['dlr', 'dlr-targeted']:
            # check if there are zero gradients
            check_zero_gradients(grad, logger=self.logger)

        acc = logits.detach().max(1)[1] == y
        acc_steps[0] = acc + 0
        loss_best = loss_indiv.detach().clone()

        alpha = 2. if self.norm in ['Linf', 'L2'] else 1. if self.norm in ['L1'] else 2e-2
        step_size = alpha * self.eps * torch.ones([x.shape[0], *(
                [1] * self.ndims)]).to(self.device).detach()
        x_adv_old = x_adv.clone()
        counter = 0
        k = self.n_iter_2 + 0
        n_fts = math.prod(self.orig_dim)
        if self.norm == 'L1':
            k = max(int(.04 * self.n_iter), 1)
            if x_init is None:
                topk = .2 * torch.ones([x.shape[0]], device=self.device)
                sp_old =  n_fts * torch.ones_like(topk)
            else:
                topk = L0_norm(x_adv - x) / n_fts / 1.5
                sp_old = L0_norm(x_adv - x)
            #print(topk[0], sp_old[0])
            adasp_redstep = 1.5
            adasp_minstep = 10.
            #print(step_size[0].item())
        counter3 = 0

        loss_best_last_check = loss_best.clone()
        reduced_last_check = torch.ones_like(loss_best)
        n_reduced = 0

        u = torch.arange(x.shape[0], device=self.device)
        for i in range(self.n_iter):
            ### gradient step
            with torch.no_grad():
                x_adv = x_adv.detach()
                grad2 = x_adv - x_adv_old
                x_adv_old = x_adv.clone()

                a = 0.75 if i > 0 else 1.0

                if self.norm == 'Linf':
                    x_adv_1 = x_adv + step_size * torch.sign(grad)
                    x_adv_1 = torch.clamp(torch.min(torch.max(x_adv_1,
                                                              x - self.eps), x + self.eps), 0.0, 1.0)
                    x_adv_1 = torch.clamp(torch.min(torch.max(
                        x_adv + (x_adv_1 - x_adv) * a + grad2 * (1 - a),
                        x - self.eps), x + self.eps), 0.0, 1.0)

                elif self.norm == 'L2':
                    x_adv_1 = x_adv + step_size * self.normalize(grad)
                    x_adv_1 = torch.clamp(x + self.normalize(x_adv_1 - x
                                                             ) * torch.min(self.eps * torch.ones_like(x).detach(),
                                                                           L2_norm(x_adv_1 - x, keepdim=True)), 0.0, 1.0)
                    x_adv_1 = x_adv + (x_adv_1 - x_adv) * a + grad2 * (1 - a)
                    x_adv_1 = torch.clamp(x + self.normalize(x_adv_1 - x
                                                             ) * torch.min(self.eps * torch.ones_like(x).detach(),
                                                                           L2_norm(x_adv_1 - x, keepdim=True)), 0.0, 1.0)

                elif self.norm == 'L1':
                    grad_topk = grad.abs().view(x.shape[0], -1).sort(-1)[0]
                    topk_curr = torch.clamp((1. - topk) * n_fts, min=0, max=n_fts - 1).long()
                    grad_topk = grad_topk[u, topk_curr].view(-1, *[1]*(len(x.shape) - 1))
                    sparsegrad = grad * (grad.abs() >= grad_topk).float()
                    x_adv_1 = x_adv + step_size * sparsegrad.sign() / (
                            L1_norm(sparsegrad.sign(), keepdim=True) + 1e-10)

                    delta_u = x_adv_1 - x
                    delta_p = L1_projection(x, delta_u, self.eps)
                    x_adv_1 = x + delta_u + delta_p


                x_adv = x_adv_1 + 0.

            ### get gradient
            x_adv.requires_grad_()
            grad = torch.zeros_like(x)
            for _ in range(self.eot_iter):
                if not self.is_tf_model:
                    with torch.enable_grad():
                        logits = self.model(x_adv)
                        loss_indiv = criterion_indiv(logits, y)
                        loss = loss_indiv.sum()

                    grad += torch.autograd.grad(loss, [x_adv])[0].detach()
                else:
                    if self.y_target is None:
                        logits, loss_indiv, grad_curr = criterion_indiv(x_adv, y)
                    else:
                        logits, loss_indiv, grad_curr = criterion_indiv(x_adv, y, self.y_target)
                    grad += grad_curr

            grad /= float(self.eot_iter)

            pred = logits.detach().max(1)[1] == y
            acc = torch.min(acc, pred)
            acc_steps[i + 1] = acc + 0
            ind_pred = (pred == 0).nonzero().squeeze()
            x_best_adv[ind_pred] = x_adv[ind_pred] + 0.
            if self.verbose:
                str_stats = ' - step size: {:.5f} - topk: {:.2f}'.format(
                    step_size.mean(), topk.mean() * n_fts) if self.norm in ['L1'] else ''
                print('[m] iteration: {} - best loss: {:.6f} - robust accuracy: {:.2%}{}'.format(
                    i, loss_best.sum(), acc.float().mean(), str_stats))
                #print('pert {}'.format((x - x_best_adv).abs().view(x.shape[0], -1).sum(-1).max()))

            ### check step size
            with torch.no_grad():
                y1 = loss_indiv.detach().clone()
                loss_steps[i] = y1 + 0
                ind = (y1 > loss_best).nonzero().squeeze()
                x_best[ind] = x_adv[ind].clone()
                grad_best[ind] = grad[ind].clone()
                loss_best[ind] = y1[ind] + 0
                loss_best_steps[i + 1] = loss_best + 0

                counter3 += 1

                if counter3 == k:
                    if self.norm in ['Linf', 'L2']:
                        fl_oscillation = self.check_oscillation(loss_steps, i, k,
                                                                loss_best, k3=self.thr_decr)
                        fl_reduce_no_impr = (1. - reduced_last_check) * (
                                loss_best_last_check >= loss_best).float()
                        fl_oscillation = torch.max(fl_oscillation,
                                                   fl_reduce_no_impr)
                        reduced_last_check = fl_oscillation.clone()
                        loss_best_last_check = loss_best.clone()

                        if fl_oscillation.sum() > 0:
                            ind_fl_osc = (fl_oscillation > 0).nonzero().squeeze()
                            step_size[ind_fl_osc] /= 2.0
                            n_reduced = fl_oscillation.sum()

                            x_adv[ind_fl_osc] = x_best[ind_fl_osc].clone()
                            grad[ind_fl_osc] = grad_best[ind_fl_osc].clone()

                        k = max(k - self.size_decr, self.n_iter_min)

                    elif self.norm == 'L1':
                        sp_curr = L0_norm(x_best - x)
                        fl_redtopk = (sp_curr / sp_old) < .95
                        topk = sp_curr / n_fts / 1.5
                        step_size[fl_redtopk] = alpha * self.eps
                        step_size[~fl_redtopk] /= adasp_redstep
                        step_size.clamp_(alpha * self.eps / adasp_minstep, alpha * self.eps)
                        sp_old = sp_curr.clone()

                        x_adv[fl_redtopk] = x_best[fl_redtopk].clone()
                        grad[fl_redtopk] = grad_best[fl_redtopk].clone()

                    counter3 = 0
                    #k = max(k - self.size_decr, self.n_iter_min)

        #

        return (x_best, acc, loss_best, x_best_adv)

    def perturb(self, x, y=None, best_loss=False, x_init=None):
        """
        :param x:           clean images
        :param y:           clean labels, if None we use the predicted labels
        :param best_loss:   if True the points attaining highest loss
                            are returned, otherwise adversarial examples
        """

        assert self.loss in ['ce', 'dlr'] #'ce-targeted-cfts'
        if not y is None and len(y.shape) == 0:
            x.unsqueeze_(0)
            y.unsqueeze_(0)
        self.init_hyperparam(x)

        x = x.detach().clone().float().to(self.device)
        if not self.is_tf_model:
            y_pred = self.model(x).max(1)[1]
        else:
            y_pred = self.model.predict(x).max(1)[1]
        if y is None:
            #y_pred = self.predict(x).max(1)[1]
            y = y_pred.detach().clone().long().to(self.device)
        else:
            y = y.detach().clone().long().to(self.device)

        adv = x.clone()
        if self.loss != 'ce-targeted':
            acc = y_pred == y
        else:
            acc = y_pred != y
        loss = -1e10 * torch.ones_like(acc).float()
        if self.verbose:
            print('-------------------------- ',
                  'running {}-attack with epsilon {:.5f}'.format(
                      self.norm, self.eps),
                  '--------------------------')
            print('initial accuracy: {:.2%}'.format(acc.float().mean()))



        if self.use_largereps:
            epss = [3. * self.eps_orig, 2. * self.eps_orig, 1. * self.eps_orig]
            iters = [.3 * self.n_iter_orig, .3 * self.n_iter_orig,
                     .4 * self.n_iter_orig]
            iters = [math.ceil(c) for c in iters]
            iters[-1] = self.n_iter_orig - sum(iters[:-1]) # make sure to use the given iterations
            if self.verbose:
                print('using schedule [{}x{}]'.format('+'.join([str(c
                                                                    ) for c in epss]), '+'.join([str(c) for c in iters])))

        startt = time.time()
        if not best_loss:
            torch.random.manual_seed(self.seed)
            torch.cuda.random.manual_seed(self.seed)

            for counter in range(self.n_restarts):
                ind_to_fool = acc.nonzero().squeeze()
                if len(ind_to_fool.shape) == 0:
                    ind_to_fool = ind_to_fool.unsqueeze(0)
                if ind_to_fool.numel() != 0:
                    x_to_fool = x[ind_to_fool].clone()
                    y_to_fool = y[ind_to_fool].clone()


                    if not self.use_largereps:
                        res_curr = self.attack_single_run(x_to_fool, y_to_fool)
                    else:
                        res_curr = self.decr_eps_pgd(x_to_fool, y_to_fool, epss, iters)
                    best_curr, acc_curr, loss_curr, adv_curr = res_curr
                    ind_curr = (acc_curr == 0).nonzero().squeeze()

                    acc[ind_to_fool[ind_curr]] = 0
                    adv[ind_to_fool[ind_curr]] = adv_curr[ind_curr].clone()
                    if self.verbose:
                        print('restart {} - robust accuracy: {:.2%}'.format(
                            counter, acc.float().mean()),
                            '- cum. time: {:.1f} s'.format(
                                time.time() - startt))

            return adv

        else:
            adv_best = x.detach().clone()
            loss_best = torch.ones([x.shape[0]]).to(
                self.device) * (-float('inf'))
            for counter in range(self.n_restarts):
                best_curr, _, loss_curr, _ = self.attack_single_run(x, y)
                ind_curr = (loss_curr > loss_best).nonzero().squeeze()
                adv_best[ind_curr] = best_curr[ind_curr] + 0.
                loss_best[ind_curr] = loss_curr[ind_curr] + 0.

                if self.verbose:
                    print('restart {} - loss: {:.5f}'.format(
                        counter, loss_best.sum()))

            return adv_best

    def decr_eps_pgd(self, x, y, epss, iters, use_rs=True):
        assert len(epss) == len(iters)
        assert self.norm in ['L1']
        self.use_rs = False
        if not use_rs:
            x_init = None
        else:
            x_init = x + torch.randn_like(x)
            x_init += L1_projection(x, x_init - x, 1. * float(epss[0]))
        eps_target = float(epss[-1])
        if self.verbose:
            print('total iter: {}'.format(sum(iters)))
        for eps, niter in zip(epss, iters):
            if self.verbose:
                print('using eps: {:.2f}'.format(eps))
            self.n_iter = niter + 0
            self.eps = eps + 0.
            #
            if not x_init is None:
                x_init += L1_projection(x, x_init - x, 1. * eps)
            x_init, acc, loss, x_adv = self.attack_single_run(x, y, x_init=x_init)

        return (x_init, acc, loss, x_adv)

class APGDAttack_targeted(APGDAttack):
    def __init__(
            self,
            predict,
            n_iter=100,
            norm='Linf',
            n_restarts=1,
            eps=None,
            seed=0,
            eot_iter=1,
            rho=.75,
            topk=None,
            n_target_classes=9,
            verbose=False,
            device=None,
            use_largereps=False,
            is_tf_model=False,
            logger=None):
        """
        AutoPGD on the targeted DLR loss
        """
        super(APGDAttack_targeted, self).__init__(predict, n_iter=n_iter, norm=norm,
                                                  n_restarts=n_restarts, eps=eps, seed=seed, loss='dlr-targeted',
                                                  eot_iter=eot_iter, rho=rho, topk=topk, verbose=verbose, device=device,
                                                  use_largereps=use_largereps, is_tf_model=is_tf_model, logger=logger)

        self.y_target = None
        self.n_target_classes = n_target_classes

    def dlr_loss_targeted(self, x, y):
        x_sorted, ind_sorted = x.sort(dim=1)
        u = torch.arange(x.shape[0])

        return -(x[u, y] - x[u, self.y_target]) / (x_sorted[:, -1] - .5 * (
                x_sorted[:, -3] + x_sorted[:, -4]) + 1e-12)

    def ce_loss_targeted(self, x, y):
        return -1. * F.cross_entropy(x, self.y_target, reduction='none')


    def perturb(self, x, y=None, x_init=None, best_loss=False):
        """
        :param x:           clean images
        :param y:           clean labels, if None we use the predicted labels
        """

        assert self.loss in ['dlr-targeted'] #'ce-targeted'
        assert not best_loss, 'not implemented yet'

        if not y is None and len(y.shape) == 0:
            x.unsqueeze_(0)
            y.unsqueeze_(0)
        self.init_hyperparam(x)

        x = x.detach().clone().float().to(self.device)
        if not self.is_tf_model:
            y_pred = self.model(x).max(1)[1]
        else:
            y_pred = self.model.predict(x).max(1)[1]
        if y is None:
            #y_pred = self._get_predicted_label(x)
            y = y_pred.detach().clone().long().to(self.device)
        else:
            y = y.detach().clone().long().to(self.device)

        adv = x.clone()
        acc = y_pred == y
        if self.verbose:
            print('-------------------------- ',
                  'running {}-attack with epsilon {:.5f}'.format(
                      self.norm, self.eps),
                  '--------------------------')
            print('initial accuracy: {:.2%}'.format(acc.float().mean()))

        startt = time.time()

        torch.random.manual_seed(self.seed)
        torch.cuda.random.manual_seed(self.seed)

        #

        if self.use_largereps:
            epss = [3. * self.eps_orig, 2. * self.eps_orig, 1. * self.eps_orig]
            iters = [.3 * self.n_iter_orig, .3 * self.n_iter_orig,
                     .4 * self.n_iter_orig]
            iters = [math.ceil(c) for c in iters]
            iters[-1] = self.n_iter_orig - sum(iters[:-1])
            if self.verbose:
                print('using schedule [{}x{}]'.format('+'.join([str(c
                                                                    ) for c in epss]), '+'.join([str(c) for c in iters])))

        for target_class in range(2, self.n_target_classes + 2):
            for counter in range(self.n_restarts):
                ind_to_fool = acc.nonzero().squeeze()
                if len(ind_to_fool.shape) == 0:
                    ind_to_fool = ind_to_fool.unsqueeze(0)
                if ind_to_fool.numel() != 0:
                    x_to_fool = x[ind_to_fool].clone()
                    y_to_fool = y[ind_to_fool].clone()

                    if not self.is_tf_model:
                        output = self.model(x_to_fool)
                    else:
                        output = self.model.predict(x_to_fool)
                    self.y_target = output.sort(dim=1)[1][:, -target_class]

                    if not self.use_largereps:
                        res_curr = self.attack_single_run(x_to_fool, y_to_fool)
                    else:
                        res_curr = self.decr_eps_pgd(x_to_fool, y_to_fool, epss, iters)
                    best_curr, acc_curr, loss_curr, adv_curr = res_curr
                    ind_curr = (acc_curr == 0).nonzero().squeeze()

                    acc[ind_to_fool[ind_curr]] = 0
                    adv[ind_to_fool[ind_curr]] = adv_curr[ind_curr].clone()
                    if self.verbose:
                        print('target class {}'.format(target_class),
                              '- restart {} - robust accuracy: {:.2%}'.format(
                                  counter, acc.float().mean()),
                              '- cum. time: {:.1f} s'.format(
                                  time.time() - startt))

        return adv