File size: 30,952 Bytes
8f87579
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
'''
Running statistics on the GPU using pytorch.

RunningTopK maintains top-k statistics for a set of channels in parallel.
RunningQuantile maintains (sampled) quantile statistics for a set of channels.
'''

import torch, math, numpy
from collections import defaultdict

class RunningTopK:
    '''
    A class to keep a running tally of the the top k values (and indexes)
    of any number of torch feature components.  Will work on the GPU if
    the data is on the GPU.

    This version flattens all arrays to avoid crashes.
    '''
    def __init__(self, k=100, state=None):
        if state is not None:
            self.set_state_dict(state)
            return
        self.k = k
        self.count = 0
        # This version flattens all data internally to 2-d tensors,
        # to avoid crashes with the current pytorch topk implementation.
        # The data is puffed back out to arbitrary tensor shapes on ouput.
        self.data_shape = None
        self.top_data = None
        self.top_index = None
        self.next = 0
        self.linear_index = 0
        self.perm = None

    def add(self, data):
        '''
        Adds a batch of data to be considered for the running top k.
        The zeroth dimension enumerates the observations.  All other
        dimensions enumerate different features.
        '''
        if self.top_data is None:
            # Allocation: allocate a buffer of size 5*k, at least 10, for each.
            self.data_shape = data.shape[1:]
            feature_size = int(numpy.prod(self.data_shape))
            self.top_data = torch.zeros(
                    feature_size, max(10, self.k * 5), out=data.new())
            self.top_index = self.top_data.clone().long()
            self.linear_index = 0 if len(data.shape) == 1 else torch.arange(
                feature_size, out=self.top_index.new()).mul_(
                        self.top_data.shape[-1])[:,None]
        size = data.shape[0]
        sk = min(size, self.k)
        if self.top_data.shape[-1] < self.next + sk:
            # Compression: if full, keep topk only.
            self.top_data[:,:self.k], self.top_index[:,:self.k] = (
                    self.result(sorted=False, flat=True))
            self.next = self.k
            free = self.top_data.shape[-1] - self.next
        # Pick: copy the top sk of the next batch into the buffer.
        # Currently strided topk is slow.  So we clone after transpose.
        # TODO: remove the clone() if it becomes faster.
        cdata = data.contiguous().view(size, -1).t().clone()
        td, ti = cdata.topk(sk, sorted=False)
        self.top_data[:,self.next:self.next+sk] = td
        self.top_index[:,self.next:self.next+sk] = (ti + self.count)
        self.next += sk
        self.count += size

    def result(self, sorted=True, flat=False):
        '''
        Returns top k data items and indexes in each dimension,
        with channels in the first dimension and k in the last dimension.
        '''
        k = min(self.k, self.next)
        # bti are top indexes relative to buffer array.
        td, bti = self.top_data[:,:self.next].topk(k, sorted=sorted)
        # we want to report top indexes globally, which is ti.
        ti = self.top_index.view(-1)[
                (bti + self.linear_index).view(-1)
                ].view(*bti.shape)
        if flat:
            return td, ti
        else:
            return (td.view(*(self.data_shape + (-1,))),
                    ti.view(*(self.data_shape + (-1,))))

    def to_(self, device):
        self.top_data = self.top_data.to(device)
        self.top_index = self.top_index.to(device)
        if isinstance(self.linear_index, torch.Tensor):
            self.linear_index = self.linear_index.to(device)

    def state_dict(self):
        return dict(
                constructor=self.__module__ + '.' +
                    self.__class__.__name__ + '()',
                k=self.k,
                count=self.count,
                data_shape=tuple(self.data_shape),
                top_data=self.top_data.cpu().numpy(),
                top_index=self.top_index.cpu().numpy(),
                next=self.next,
                linear_index=(self.linear_index.cpu().numpy()
                    if isinstance(self.linear_index, torch.Tensor)
                    else self.linear_index),
                perm=self.perm)

    def set_state_dict(self, dic):
        self.k = dic['k'].item()
        self.count = dic['count'].item()
        self.data_shape = tuple(dic['data_shape'])
        self.top_data = torch.from_numpy(dic['top_data'])
        self.top_index = torch.from_numpy(dic['top_index'])
        self.next = dic['next'].item()
        self.linear_index = (torch.from_numpy(dic['linear_index'])
                if len(dic['linear_index'].shape) > 0
                else dic['linear_index'].item())

class RunningQuantile:
    """
    Streaming randomized quantile computation for torch.

    Add any amount of data repeatedly via add(data).  At any time,
    quantile estimates (or old-style percentiles) can be read out using
    quantiles(q) or percentiles(p).

    Accuracy scales according to resolution: the default is to
    set resolution to be accurate to better than 0.1%,
    while limiting storage to about 50,000 samples.

    Good for computing quantiles of huge data without using much memory.
    Works well on arbitrary data with probability near 1.

    Based on the optimal KLL quantile algorithm by Karnin, Lang, and Liberty
    from FOCS 2016.  http://ieee-focs.org/FOCS-2016-Papers/3933a071.pdf
    """

    def __init__(self, resolution=6 * 1024, buffersize=None, seed=None,
            state=None):
        if state is not None:
            self.set_state_dict(state)
            return
        self.depth = None
        self.dtype = None
        self.device = None
        self.resolution = resolution
        # Default buffersize: 128 samples (and smaller than resolution).
        if buffersize is None:
            buffersize = min(128, (resolution + 7) // 8)
        self.buffersize = buffersize
        self.samplerate = 1.0
        self.data = None
        self.firstfree = [0]
        self.randbits = torch.ByteTensor(resolution)
        self.currentbit = len(self.randbits) - 1
        self.extremes = None
        self.size = 0

    def _lazy_init(self, incoming):
        self.depth = incoming.shape[1]
        self.dtype = incoming.dtype
        self.device = incoming.device
        self.data = [torch.zeros(self.depth, self.resolution,
            dtype=self.dtype, device=self.device)]
        self.extremes = torch.zeros(self.depth, 2,
                dtype=self.dtype, device=self.device)
        self.extremes[:,0] = float('inf')
        self.extremes[:,-1] = -float('inf')

    def to_(self, device):
        """Switches internal storage to specified device."""
        if device != self.device:
            old_data = self.data
            old_extremes = self.extremes
            self.data = [d.to(device) for d in self.data]
            self.extremes = self.extremes.to(device)
            self.device = self.extremes.device
            del old_data
            del old_extremes

    def add(self, incoming):
        if self.depth is None:
            self._lazy_init(incoming)
        assert len(incoming.shape) == 2
        assert incoming.shape[1] == self.depth, (incoming.shape[1], self.depth)
        self.size += incoming.shape[0]
        # Convert to a flat torch array.
        if self.samplerate >= 1.0:
            self._add_every(incoming)
            return
        # If we are sampling, then subsample a large chunk at a time.
        self._scan_extremes(incoming)
        chunksize = int(math.ceil(self.buffersize / self.samplerate))
        for index in range(0, len(incoming), chunksize):
            batch = incoming[index:index+chunksize]
            sample = sample_portion(batch, self.samplerate)
            if len(sample):
                self._add_every(sample)

    def _add_every(self, incoming):
        supplied = len(incoming)
        index = 0
        while index < supplied:
            ff = self.firstfree[0]
            available = self.data[0].shape[1] - ff
            if available == 0:
                if not self._shift():
                    # If we shifted by subsampling, then subsample.
                    incoming = incoming[index:]
                    if self.samplerate >= 0.5:
                        # First time sampling - the data source is very large.
                        self._scan_extremes(incoming)
                    incoming = sample_portion(incoming, self.samplerate)
                    index = 0
                    supplied = len(incoming)
                ff = self.firstfree[0]
                available = self.data[0].shape[1] - ff
            copycount = min(available, supplied - index)
            self.data[0][:,ff:ff + copycount] = torch.t(
                    incoming[index:index + copycount,:])
            self.firstfree[0] += copycount
            index += copycount

    def _shift(self):
        index = 0
        # If remaining space at the current layer is less than half prev
        # buffer size (rounding up), then we need to shift it up to ensure
        # enough space for future shifting.
        while self.data[index].shape[1] - self.firstfree[index] < (
                -(-self.data[index-1].shape[1] // 2) if index else 1):
            if index + 1 >= len(self.data):
                return self._expand()
            data = self.data[index][:,0:self.firstfree[index]]
            data = data.sort()[0]
            if index == 0 and self.samplerate >= 1.0:
                self._update_extremes(data[:,0], data[:,-1])
            offset = self._randbit()
            position = self.firstfree[index + 1]
            subset = data[:,offset::2]
            self.data[index + 1][:,position:position + subset.shape[1]] = subset
            self.firstfree[index] = 0
            self.firstfree[index + 1] += subset.shape[1]
            index += 1
        return True

    def _scan_extremes(self, incoming):
        # When sampling, we need to scan every item still to get extremes
        self._update_extremes(
                torch.min(incoming, dim=0)[0],
                torch.max(incoming, dim=0)[0])

    def _update_extremes(self, minr, maxr):
        self.extremes[:,0] = torch.min(
                torch.stack([self.extremes[:,0], minr]), dim=0)[0]
        self.extremes[:,-1] = torch.max(
                torch.stack([self.extremes[:,-1], maxr]), dim=0)[0]

    def _randbit(self):
        self.currentbit += 1
        if self.currentbit >= len(self.randbits):
            self.randbits.random_(to=2)
            self.currentbit = 0
        return self.randbits[self.currentbit]

    def state_dict(self):
        return dict(
                constructor=self.__module__ + '.' +
                    self.__class__.__name__ + '()',
                resolution=self.resolution,
                depth=self.depth,
                buffersize=self.buffersize,
                samplerate=self.samplerate,
                data=[d.cpu().numpy()[:,:f].T
                    for d, f in zip(self.data, self.firstfree)],
                sizes=[d.shape[1] for d in self.data],
                extremes=self.extremes.cpu().numpy(),
                size=self.size)

    def set_state_dict(self, dic):
        self.resolution = int(dic['resolution'])
        self.randbits = torch.ByteTensor(self.resolution)
        self.currentbit = len(self.randbits) - 1
        self.depth = int(dic['depth'])
        self.buffersize = int(dic['buffersize'])
        self.samplerate = float(dic['samplerate'])
        firstfree = []
        buffers = []
        for d, s in zip(dic['data'], dic['sizes']):
            firstfree.append(d.shape[0])
            buf = numpy.zeros((d.shape[1], s), dtype=d.dtype)
            buf[:,:d.shape[0]] = d.T
            buffers.append(torch.from_numpy(buf))
        self.firstfree = firstfree
        self.data = buffers
        self.extremes = torch.from_numpy((dic['extremes']))
        self.size = int(dic['size'])
        self.dtype = self.extremes.dtype
        self.device = self.extremes.device

    def minmax(self):
        if self.firstfree[0]:
            self._scan_extremes(self.data[0][:,:self.firstfree[0]].t())
        return self.extremes.clone()

    def median(self):
        return self.quantiles([0.5])[:,0]

    def mean(self):
        return self.integrate(lambda x: x) / self.size

    def variance(self):
        mean = self.mean()[:,None]
        return self.integrate(lambda x: (x - mean).pow(2)) / (self.size - 1)

    def stdev(self):
        return self.variance().sqrt()

    def _expand(self):
        cap = self._next_capacity()
        if cap > 0:
            # First, make a new layer of the proper capacity.
            self.data.insert(0, torch.zeros(self.depth, cap,
                dtype=self.dtype, device=self.device))
            self.firstfree.insert(0, 0)
        else:
            # Unless we're so big we are just subsampling.
            assert self.firstfree[0] == 0
            self.samplerate *= 0.5
        for index in range(1, len(self.data)):
            # Scan for existing data that needs to be moved down a level.
            amount = self.firstfree[index]
            if amount == 0:
                continue
            position = self.firstfree[index-1]
            # Move data down if it would leave enough empty space there
            # This is the key invariant: enough empty space to fit half
            # of the previous level's buffer size (rounding up)
            if self.data[index-1].shape[1] - (amount + position) >= (
                    -(-self.data[index-2].shape[1] // 2) if (index-1) else 1):
                self.data[index-1][:,position:position + amount] = (
                        self.data[index][:,:amount])
                self.firstfree[index-1] += amount
                self.firstfree[index] = 0
            else:
                # Scrunch the data if it would not.
                data = self.data[index][:,:amount]
                data = data.sort()[0]
                if index == 1:
                    self._update_extremes(data[:,0], data[:,-1])
                offset = self._randbit()
                scrunched = data[:,offset::2]
                self.data[index][:,:scrunched.shape[1]] = scrunched
                self.firstfree[index] = scrunched.shape[1]
        return cap > 0

    def _next_capacity(self):
        cap = int(math.ceil(self.resolution * (0.67 ** len(self.data))))
        if cap < 2:
            return 0
        # Round up to the nearest multiple of 8 for better GPU alignment.
        cap = -8 * (-cap // 8)
        return max(self.buffersize, cap)

    def _weighted_summary(self, sort=True):
        if self.firstfree[0]:
            self._scan_extremes(self.data[0][:,:self.firstfree[0]].t())
        size = sum(self.firstfree) + 2
        weights = torch.FloatTensor(size) # Floating point
        summary = torch.zeros(self.depth, size,
                dtype=self.dtype, device=self.device)
        weights[0:2] = 0
        summary[:,0:2] = self.extremes
        index = 2
        for level, ff in enumerate(self.firstfree):
            if ff == 0:
                continue
            summary[:,index:index + ff] = self.data[level][:,:ff]
            weights[index:index + ff] = 2.0 ** level
            index += ff
        assert index == summary.shape[1]
        if sort:
            summary, order = torch.sort(summary, dim=-1)
            weights = weights[order.view(-1).cpu()].view(order.shape)
        return (summary, weights)

    def quantiles(self, quantiles, old_style=False):
        if self.size == 0:
            return torch.full((self.depth, len(quantiles)), torch.nan)
        summary, weights = self._weighted_summary()
        cumweights = torch.cumsum(weights, dim=-1) - weights / 2
        if old_style:
            # To be convenient with torch.percentile
            cumweights -= cumweights[:,0:1].clone()
            cumweights /= cumweights[:,-1:].clone()
        else:
            cumweights /= torch.sum(weights, dim=-1, keepdim=True)
        result = torch.zeros(self.depth, len(quantiles),
                dtype=self.dtype, device=self.device)
        # numpy is needed for interpolation
        if not hasattr(quantiles, 'cpu'):
            quantiles = torch.Tensor(quantiles)
        nq = quantiles.cpu().numpy()
        ncw = cumweights.cpu().numpy()
        nsm = summary.cpu().numpy()
        for d in range(self.depth):
            result[d] = torch.tensor(numpy.interp(nq, ncw[d], nsm[d]),
                    dtype=self.dtype, device=self.device)
        return result

    def integrate(self, fun):
        result = None
        for level, ff in enumerate(self.firstfree):
            if ff == 0:
                continue
            term = torch.sum(
                    fun(self.data[level][:,:ff]) * (2.0 ** level),
                    dim=-1)
            if result is None:
                result = term
            else:
                result += term
        if result is not None:
            result /= self.samplerate
        return result

    def percentiles(self, percentiles):
        return self.quantiles(percentiles, old_style=True)

    def readout(self, count=1001, old_style=True):
        return self.quantiles(
                torch.linspace(0.0, 1.0, count), old_style=old_style)

    def normalize(self, data):
        '''
        Given input data as taken from the training distirbution,
        normalizes every channel to reflect quantile values,
        uniformly distributed, within [0, 1].
        '''
        assert self.size > 0
        assert data.shape[0] == self.depth
        summary, weights = self._weighted_summary()
        cumweights = torch.cumsum(weights, dim=-1) - weights / 2
        cumweights /= torch.sum(weights, dim=-1, keepdim=True)
        result = torch.zeros_like(data).float()
        # numpy is needed for interpolation
        ndata = data.cpu().numpy().reshape((data.shape[0], -1))
        ncw = cumweights.cpu().numpy()
        nsm = summary.cpu().numpy()
        for d in range(self.depth):
            normed = torch.tensor(numpy.interp(ndata[d], nsm[d], ncw[d]),
                dtype=torch.float, device=data.device).clamp_(0.0, 1.0)
            if len(data.shape) > 1:
                normed = normed.view(*(data.shape[1:]))
            result[d] = normed
        return result


class RunningConditionalQuantile:
    '''
    Equivalent to a map from conditions (any python hashable type)
    to RunningQuantiles.  The reason for the type is to allow limited
    GPU memory to be exploited while counting quantile stats on many
    different conditions, a few of which are common and which benefit
    from GPU, but most of which are rare and would not all fit into
    GPU RAM.

    To move a set of conditions to a device, use rcq.to_(device, conds).
    Then in the future, move the tallied data to the device before
    calling rcq.add, that is, rcq.add(cond, data.to(device)).

    To allow the caller to decide which conditions to allow to use GPU,
    rcq.most_common_conditions(n) returns a list of the n most commonly
    added conditions so far.
    '''
    def __init__(self, resolution=6 * 1024, buffersize=None, seed=None,
            state=None):
        self.first_rq = None
        self.call_stats = defaultdict(int)
        self.running_quantiles = {}
        if state is not None:
            self.set_state_dict(state)
            return
        self.rq_args = dict(resolution=resolution, buffersize=buffersize,
                seed=seed)

    def add(self, condition, incoming):
        if condition not in self.running_quantiles:
            self.running_quantiles[condition] = RunningQuantile(**self.rq_args)
            if self.first_rq is None:
                self.first_rq = self.running_quantiles[condition]
        self.call_stats[condition] += 1
        rq = self.running_quantiles[condition]
        # For performance reasons, the caller can move some conditions to
        # the CPU if they are not among the most common conditions.
        if rq.device is not None and (rq.device != incoming.device):
            rq.to_(incoming.device)
        self.running_quantiles[condition].add(incoming)

    def most_common_conditions(self, n):
        return sorted(self.call_stats.keys(),
                key=lambda c: -self.call_stats[c])[:n]

    def collected_add(self, conditions, incoming):
        for c in conditions:
            self.add(c, incoming)

    def conditional(self, c):
        return self.running_quantiles[c]

    def collected_quantiles(self, conditions, quantiles, old_style=False):
        result = torch.zeros(
                size=(len(conditions), self.first_rq.depth, len(quantiles)),
                dtype=self.first_rq.dtype,
                device=self.first_rq.device)
        for i, c in enumerate(conditions):
            if c in self.running_quantiles:
                result[i] = self.running_quantiles[c].quantiles(
                        quantiles, old_style)
        return result

    def collected_normalize(self, conditions, values):
        result = torch.zeros(
                size=(len(conditions), values.shape[0], values.shape[1]),
                dtype=torch.float,
                device=self.first_rq.device)
        for i, c in enumerate(conditions):
            if c in self.running_quantiles:
                result[i] = self.running_quantiles[c].normalize(values)
        return result

    def to_(self, device, conditions=None):
        if conditions is None:
            conditions = self.running_quantiles.keys()
        for cond in conditions:
            if cond in self.running_quantiles:
                self.running_quantiles[cond].to_(device)

    def state_dict(self):
        conditions = sorted(self.running_quantiles.keys())
        result = dict(
                constructor=self.__module__ + '.' +
                    self.__class__.__name__ + '()',
                rq_args=self.rq_args,
                conditions=conditions)
        for i, c in enumerate(conditions):
            result.update({
                '%d.%s' % (i, k): v
                for k, v in self.running_quantiles[c].state_dict().items()})
        return result

    def set_state_dict(self, dic):
        self.rq_args = dic['rq_args'].item()
        conditions = list(dic['conditions'])
        subdicts = defaultdict(dict)
        for k, v in dic.items():
            if '.' in k:
                p, s = k.split('.', 1)
                subdicts[p][s] = v
        self.running_quantiles = {
                c: RunningQuantile(state=subdicts[str(i)])
                for i, c in enumerate(conditions)}
        if conditions:
            self.first_rq = self.running_quantiles[conditions[0]]

    # example usage:
    # levels = rqc.conditional(()).quantiles(1 - fracs)
    # denoms = 1 - rqc.collected_normalize(cats, levels)
    # isects = 1 - rqc.collected_normalize(labels, levels)
    # unions = fracs + denoms[cats] - isects
    # iou = isects / unions




class RunningCrossCovariance:
    '''
    Running computation. Use this when an off-diagonal block of the
    covariance matrix is needed (e.g., when the whole covariance matrix
    does not fit in the GPU).

    Chan-style numerically stable update of mean and full covariance matrix.
    Chan, Golub. LeVeque. 1983. http://www.jstor.org/stable/2683386
    '''
    def __init__(self, state=None):
        if state is not None:
            self.set_state_dict(state)
            return
        self.count = 0
        self._mean = None
        self.cmom2 = None
        self.v_cmom2 = None

    def add(self, a, b):
        if len(a.shape) == 1:
            a = a[None, :]
            b = b[None, :]
        assert(a.shape[0] == b.shape[0])
        if len(a.shape) > 2:
            a, b = [d.view(d.shape[0], d.shape[1], -1).permute(0, 2, 1
                ).contiguous().view(-1, d.shape[1]) for d in [a, b]]
        batch_count = a.shape[0]
        batch_mean = [d.sum(0) / batch_count for d in [a, b]]
        centered = [d - bm for d, bm in zip([a, b], batch_mean)]
        # If more than 10 billion operations, divide into batches.
        sub_batch = -(-(10 << 30) // (a.shape[1] * b.shape[1]))
        # Initial batch.
        if self._mean is None:
            self.count = batch_count
            self._mean = batch_mean
            self.v_cmom2 = [c.pow(2).sum(0) for c in centered]
            self.cmom2 = a.new(a.shape[1], b.shape[1]).zero_()
            progress_addbmm(self.cmom2, centered[0][:,:,None],
                    centered[1][:,None,:], sub_batch)
            return
        # Update a batch using Chan-style update for numerical stability.
        oldcount = self.count
        self.count += batch_count
        new_frac = float(batch_count) / self.count
        # Update the mean according to the batch deviation from the old mean.
        delta = [bm.sub_(m).mul_(new_frac)
                for bm, m in zip(batch_mean, self._mean)]
        for m, d in zip(self._mean, delta):
            m.add_(d)
        # Update the cross-covariance using the batch deviation
        progress_addbmm(self.cmom2, centered[0][:,:,None],
                centered[1][:,None,:], sub_batch)
        self.cmom2.addmm_(alpha=new_frac * oldcount,
                mat1=delta[0][:,None], mat2=delta[1][None,:])
        # Update the variance using the batch deviation
        for c, vc2, d in zip(centered, self.v_cmom2, delta):
            vc2.add_(c.pow(2).sum(0))
            vc2.add_(d.pow_(2).mul_(new_frac * oldcount))

    def mean(self):
        return self._mean

    def variance(self):
        return [vc2 / (self.count - 1) for vc2 in self.v_cmom2]

    def stdev(self):
        return [v.sqrt() for v in self.variance()]

    def covariance(self):
        return self.cmom2 / (self.count - 1)

    def correlation(self):
        covariance = self.covariance()
        rstdev = [s.reciprocal() for s in self.stdev()]
        cor = rstdev[0][:,None] * covariance * rstdev[1][None,:]
        # Remove NaNs
        cor[torch.isnan(cor)] = 0
        return cor

    def to_(self, device):
        self._mean = [m.to(device) for m in self._mean]
        self.v_cmom2 = [vcs.to(device) for vcs in self.v_cmom2]
        self.cmom2 = self.cmom2.to(device)

    def state_dict(self):
        return dict(
                constructor=self.__module__ + '.' +
                    self.__class__.__name__ + '()',
                count=self.count,
                mean_a=self._mean[0].cpu().numpy(),
                mean_b=self._mean[1].cpu().numpy(),
                cmom2_a=self.v_cmom2[0].cpu().numpy(),
                cmom2_b=self.v_cmom2[1].cpu().numpy(),
                cmom2=self.cmom2.cpu().numpy())

    def set_state_dict(self, dic):
        self.count = dic['count'].item()
        self._mean = [torch.from_numpy(dic[k]) for k in ['mean_a', 'mean_b']]
        self.v_cmom2 = [torch.from_numpy(dic[k])
                for k in ['cmom2_a', 'cmom2_b']]
        self.cmom2 = torch.from_numpy(dic['cmom2'])

def progress_addbmm(accum, x, y, batch_size):
    '''
    Break up very large adbmm operations into batches so progress can be seen.
    '''
    from .progress import default_progress
    if x.shape[0] <= batch_size:
        return accum.addbmm_(x, y)
    progress = default_progress(None)
    for i in progress(range(0, x.shape[0], batch_size), desc='bmm'):
        accum.addbmm_(x[i:i+batch_size], y[i:i+batch_size])
    return accum


def sample_portion(vec, p=0.5):
    bits = torch.bernoulli(torch.zeros(vec.shape[0], dtype=torch.uint8,
        device=vec.device), p)
    return vec[bits]

if __name__ == '__main__':
    import warnings
    warnings.filterwarnings("error")
    import time
    import argparse
    parser = argparse.ArgumentParser(
        description='Test things out')
    parser.add_argument('--mode', default='cpu', help='cpu or cuda')
    parser.add_argument('--test_size', type=int, default=1000000)
    args = parser.parse_args()

    # An adverarial case: we keep finding more numbers in the middle
    # as the stream goes on.
    amount = args.test_size
    quantiles = 1000
    data = numpy.arange(float(amount))
    data[1::2] = data[-1::-2] + (len(data) - 1)
    data /= 2
    depth = 50
    test_cuda = torch.cuda.is_available()
    alldata = data[:,None] + (numpy.arange(depth) * amount)[None, :]
    actual_sum = torch.FloatTensor(numpy.sum(alldata * alldata, axis=0))
    amt = amount // depth
    for r in range(depth):
        numpy.random.shuffle(alldata[r*amt:r*amt+amt,r])
    if args.mode == 'cuda':
        alldata = torch.cuda.FloatTensor(alldata)
        dtype = torch.float
        device = torch.device('cuda')
    else:
        alldata = torch.FloatTensor(alldata)
        dtype = torch.float
        device = None
    starttime = time.time()
    qc = RunningQuantile(resolution=6 * 1024)
    qc.add(alldata)
    # Test state dict
    saved = qc.state_dict()
    # numpy.savez('foo.npz', **saved)
    # saved = numpy.load('foo.npz')
    qc = RunningQuantile(state=saved)
    assert not qc.device.type == 'cuda'
    qc.add(alldata)
    actual_sum *= 2
    ro = qc.readout(1001).cpu()
    endtime = time.time()
    gt = torch.linspace(0, amount, quantiles+1)[None,:] + (
            torch.arange(qc.depth, dtype=torch.float) * amount)[:,None]
    maxreldev = torch.max(torch.abs(ro - gt) / amount) * quantiles
    print("Maximum relative deviation among %d perentiles: %f" % (
        quantiles, maxreldev))
    minerr = torch.max(torch.abs(qc.minmax().cpu()[:,0] -
            torch.arange(qc.depth, dtype=torch.float) * amount))
    maxerr = torch.max(torch.abs((qc.minmax().cpu()[:, -1] + 1) -
            (torch.arange(qc.depth, dtype=torch.float) + 1) * amount))
    print("Minmax error %f, %f" % (minerr, maxerr))
    interr = torch.max(torch.abs(qc.integrate(lambda x: x * x).cpu()
            - actual_sum) / actual_sum)
    print("Integral error: %f" % interr)
    medianerr = torch.max(torch.abs(qc.median() -
        alldata.median(0)[0]) / alldata.median(0)[0]).cpu()
    print("Median error: %f" % interr)
    meanerr = torch.max(
            torch.abs(qc.mean() - alldata.mean(0)) / alldata.mean(0)).cpu()
    print("Mean error: %f" % meanerr)
    varerr = torch.max(
            torch.abs(qc.variance() - alldata.var(0)) / alldata.var(0)).cpu()
    print("Variance error: %f" % varerr)
    counterr = ((qc.integrate(lambda x: torch.ones(x.shape[-1]).cpu())
                - qc.size) / (0.0 + qc.size)).item()
    print("Count error: %f" % counterr)
    print("Time %f" % (endtime - starttime))
    # Algorithm is randomized, so some of these will fail with low probability.
    assert maxreldev < 1.0
    assert minerr == 0.0
    assert maxerr == 0.0
    assert interr < 0.01
    assert abs(counterr) < 0.001
    print("OK")