File size: 30,051 Bytes
7718235
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from abc import ABC
from typing import Optional

import math
import torch
from torch import _dynamo
_dynamo.config.suppress_errors = True
import torch.nn.functional as F
from torch import nn
from torch.nn.functional import mse_loss, l1_loss, binary_cross_entropy, cross_entropy, kl_div, nll_loss
from pyro.distributions.conjugate import BetaBinomial
from pyro.distributions import Normal
from torch_geometric.nn import MessagePassing


class NeighborEmbedding(MessagePassing, ABC):
    def __init__(self, hidden_channels, num_rbf, cutoff_lower, cutoff_upper):
        super(NeighborEmbedding, self).__init__(aggr="add")
        self.distance_proj = nn.Linear(num_rbf, hidden_channels)
        self.combine = nn.Linear(hidden_channels * 2, hidden_channels)
        self.cutoff = CosineCutoff(cutoff_lower, cutoff_upper)

        self.reset_parameters()

    def reset_parameters(self):
        nn.init.xavier_uniform_(self.distance_proj.weight)
        nn.init.xavier_uniform_(self.combine.weight)
        self.distance_proj.bias.data.fill_(0)
        self.combine.bias.data.fill_(0)

    def forward(self, x, edge_index, edge_weight, edge_attr):
        # remove self loops
        mask = edge_index[0] != edge_index[1]
        if not mask.all():
            edge_index = edge_index[:, mask]
            edge_weight = edge_weight[mask]
            edge_attr = edge_attr[mask]

        C = self.cutoff(edge_weight)
        W = self.distance_proj(edge_attr) * C.view(-1, 1)

        x_neighbors = x
        # propagate_type: (x: Tensor, W: Tensor)
        x_neighbors = self.propagate(edge_index, x=x_neighbors, W=W, size=None)
        x_neighbors = self.combine(torch.cat([x, x_neighbors], dim=1))
        return x_neighbors

    def message(self, x_j, W):
        return x_j * W


class GaussianSmearing(nn.Module):
    def __init__(self, cutoff_lower=0.0, cutoff_upper=5.0, num_rbf=50, trainable=True):
        super(GaussianSmearing, self).__init__()
        self.cutoff_lower = cutoff_lower
        self.cutoff_upper = cutoff_upper
        self.num_rbf = num_rbf
        self.trainable = trainable

        offset, coeff = self._initial_params()
        if trainable:
            self.register_parameter("coeff", nn.Parameter(coeff))
            self.register_parameter("offset", nn.Parameter(offset))
        else:
            self.register_buffer("coeff", coeff)
            self.register_buffer("offset", offset)

    def _initial_params(self):
        offset = torch.linspace(self.cutoff_lower, self.cutoff_upper, self.num_rbf)
        coeff = -0.5 / (offset[1] - offset[0]) ** 2
        return offset, coeff

    def reset_parameters(self):
        offset, coeff = self._initial_params()
        self.offset.data.copy_(offset)
        self.coeff.data.copy_(coeff)

    def forward(self, dist):
        dist = dist.unsqueeze(-1) - self.offset
        return torch.exp(self.coeff * torch.pow(dist, 2))


class ExpNormalSmearing(nn.Module):
    def __init__(self, cutoff_lower=0.0, cutoff_upper=5.0, num_rbf=50, trainable=True):
        super(ExpNormalSmearing, self).__init__()
        self.cutoff_lower = cutoff_lower
        self.cutoff_upper = cutoff_upper
        self.num_rbf = num_rbf
        self.trainable = trainable

        self.cutoff_fn = CosineCutoff(0, cutoff_upper)
        self.alpha = 5.0 / (cutoff_upper - cutoff_lower)

        means, betas = self._initial_params()
        if trainable:
            self.register_parameter("means", nn.Parameter(means))
            self.register_parameter("betas", nn.Parameter(betas))
        else:
            self.register_buffer("means", means)
            self.register_buffer("betas", betas)

    def _initial_params(self):
        # initialize means and betas according to the default values in PhysNet
        # https://pubs.acs.org/doi/10.1021/acs.jctc.9b00181
        start_value = torch.exp(
            torch.scalar_tensor(-self.cutoff_upper + self.cutoff_lower)
        )
        means = torch.linspace(start_value, 1, self.num_rbf)
        betas = torch.tensor(
            [(2 / self.num_rbf * (1 - start_value)) ** -2] * self.num_rbf
        )
        return means, betas

    def reset_parameters(self):
        means, betas = self._initial_params()
        self.means.data.copy_(means)
        self.betas.data.copy_(betas)

    def forward(self, dist):
        dist = dist.unsqueeze(-1)
        return self.cutoff_fn(dist) * torch.exp(
            -self.betas
            * (torch.exp(self.alpha * (-dist + self.cutoff_lower)) - self.means) ** 2
        )


class ExpNormalSmearingUnlimited(nn.Module):
    def __init__(self, cutoff_lower=0.0, cutoff_upper=5.0, num_rbf=50, trainable=True):
        super(ExpNormalSmearingUnlimited, self).__init__()
        self.num_rbf = num_rbf
        self.trainable = trainable

        self.alpha = 1 / 20

        means, betas = self._initial_params()
        if trainable:
            self.register_parameter("means", nn.Parameter(means))
            self.register_parameter("betas", nn.Parameter(betas))
        else:
            self.register_buffer("means", means)
            self.register_buffer("betas", betas)

    def _initial_params(self):
        # initialize means and betas according to the default values in PhysNet
        # https://pubs.acs.org/doi/10.1021/acs.jctc.9b00181
        start_value = 0.1
        means = torch.linspace(start_value, 1, self.num_rbf)
        betas = torch.tensor(
            [(2 / self.num_rbf * (1 - start_value)) ** -2] * self.num_rbf
        )
        return means, betas

    def reset_parameters(self):
        means, betas = self._initial_params()
        self.means.data.copy_(means)
        self.betas.data.copy_(betas)

    def forward(self, dist):
        dist = dist.unsqueeze(-1)
        return torch.exp(
            -self.betas * (torch.exp(self.alpha * (-dist)) - self.means) ** 2
        )


class ShiftedSoftplus(nn.Module):
    def __init__(self):
        super(ShiftedSoftplus, self).__init__()
        self.shift = torch.log(torch.tensor(2.0)).item()

    def forward(self, x):
        return F.softplus(x) - self.shift


class CosineCutoff(nn.Module):
    def __init__(self, cutoff_lower=0.0, cutoff_upper=5.0):
        super(CosineCutoff, self).__init__()
        self.cutoff_lower = cutoff_lower
        self.cutoff_upper = cutoff_upper

    def forward(self, distances):
        if self.cutoff_lower > 0:
            cutoffs = 0.5 * (
                    torch.cos(
                        math.pi
                        * (
                                2
                                * (distances - self.cutoff_lower)
                                / (self.cutoff_upper - self.cutoff_lower)
                                + 1.0
                        )
                    )
                    + 1.0
            )
            # remove contributions below the cutoff radius
            cutoffs = cutoffs * (distances < self.cutoff_upper).float()
            cutoffs = cutoffs * (distances > self.cutoff_lower).float()
            return cutoffs
        else:
            cutoffs = 0.5 * (torch.cos(distances * math.pi / self.cutoff_upper) + 1.0)
            # remove contributions beyond the cutoff radius
            cutoffs = cutoffs * (distances < self.cutoff_upper).float()
            return cutoffs


class Distance(nn.Module):
    def __init__(
            self,
            cutoff_lower,
            cutoff_upper,
            return_vecs=False,
            loop=False,
    ):
        super(Distance, self).__init__()
        self.cutoff_lower = cutoff_lower
        self.cutoff_upper = cutoff_upper
        self.return_vecs = return_vecs
        self.loop = loop

    def forward(self, pos, edge_index):
        edge_vec = pos[edge_index[0]] - pos[edge_index[1]]

        mask: Optional[torch.Tensor] = None
        if self.loop:
            # mask out self loops when computing distances because
            # the norm of 0 produces NaN gradients
            # NOTE: might influence force predictions as self loop gradients are ignored
            mask = edge_index[0] != edge_index[1]
            edge_weight = torch.zeros(edge_vec.size(0), device=edge_vec.device, dtype=edge_vec.dtype)
            edge_weight[mask] = torch.norm(edge_vec[mask], dim=-1)
        else:
            edge_weight = torch.norm(edge_vec, dim=-1)

        lower_mask = edge_weight >= self.cutoff_lower
        if self.loop and mask is not None:
            # keep self loops even though they might be below the lower cutoff
            lower_mask = lower_mask | ~mask
        edge_index = edge_index[:, lower_mask]
        edge_weight = edge_weight[lower_mask]

        if self.return_vecs:
            edge_vec = edge_vec[lower_mask]
            return edge_index, edge_weight, edge_vec
        # TODO: return only `edge_index` and `edge_weight` once
        # Union typing works with TorchScript (https://github.com/pytorch/pytorch/pull/53180)
        return edge_index, edge_weight, None


class DistanceV2(nn.Module):
    def __init__(
            self,
            return_vecs=True,
            loop=False,
    ):
        super(DistanceV2, self).__init__()
        self.return_vecs = return_vecs
        self.loop = loop

    def forward(self, pos, coords, edge_index):
        # pos: [N, 3], coordinates of C_a
        # coords: [N, 3, 4], coordinates of C_b, C, N, O
        ca_ca = pos[edge_index[1]] - pos[edge_index[0]]
        cb_cb = coords[edge_index[1], :, [0]] - coords[edge_index[0], :, [0]]
        cb_N = coords[edge_index[1], :, [2]] - coords[edge_index[0], :, [0]]
        cb_O = coords[edge_index[1], :, [3]] - coords[edge_index[0], :, [0]]
        edge_vec = torch.cat([ca_ca.unsqueeze(-1), 
                              cb_cb.unsqueeze(-1),
                              cb_N.unsqueeze(-1), 
                              cb_O.unsqueeze(-1)], dim=-1)
        mask: Optional[torch.Tensor] = None
        if self.loop:
            mask = edge_index[0] != edge_index[1]
            edge_weight = torch.zeros(ca_ca.size(0), device=ca_ca.device, dtype=ca_ca.dtype)
            edge_weight[mask] = torch.norm(ca_ca[mask], dim=-1)
        else:
            edge_weight = torch.norm(ca_ca, dim=-1)

        return edge_index, edge_weight, edge_vec



rbf_class_mapping = {"gauss": GaussianSmearing, "expnorm": ExpNormalSmearing, "expnormunlim": ExpNormalSmearingUnlimited}


class AbsTanh(nn.Module):
    def __init__(self):
        super(AbsTanh, self).__init__()

    @staticmethod
    def forward(x: torch.Tensor) -> torch.Tensor:
        return torch.abs(torch.tanh(x))


class Tanh2(nn.Module):
    def __init__(self):
        super(Tanh2, self).__init__()

    @staticmethod
    def forward(x: torch.Tensor) -> torch.Tensor:
        return torch.square(torch.tanh(x))

def gelu(x):
    """Implementation of the gelu activation function.

    For information: OpenAI GPT's gelu is slightly different
    (and gives slightly different results):
    0.5 * x * (1 + torch.tanh(math.sqrt(2 / math.pi) * (x + 0.044715 * torch.pow(x, 3))))
    """
    return x * 0.5 * (1.0 + torch.erf(x / math.sqrt(2.0)))

act_class_mapping = {
    "ssp": ShiftedSoftplus,
    "softplus": nn.Softplus,
    "silu": nn.SiLU,
    "leaky_relu": nn.LeakyReLU,
    "tanh": nn.Tanh,
    "sigmoid": nn.Sigmoid,
    "pass": nn.Identity,
    "abs_tanh": AbsTanh,
    "tanh2": Tanh2,
    "softmax": nn.Softmax,
    "gelu": nn.GELU,
}


def cosin_contrastive_loss(input, target, margin=0):
    if target.ndim == 1:
        target = target.unsqueeze(1)
    if input.shape[0] == 1:
        return torch.tensor(0, dtype=input.dtype, device=input.device)
    # calculate distance of input
    dist = F.cosine_similarity(input.unsqueeze(1), input.unsqueeze(0), dim=2)
    # calculate similarity matrix
    sim = torch.eq(target, target.T)
    # change similarity matrix to -1 and 1
    sim = sim.float() * 2 - 1
    # calculate loss, but only for the upper triangle of the similarity matrix
    loss = - dist * sim + (sim + 1) / 2 + (sim - 1) * margin / 2
    # mean over all pairs
    loss = torch.clamp(loss.triu(diagonal=1), min=0).sum() / (target.shape[0] * (target.shape[0] - 1) / 2)
    return loss


def euclid_contrastive_loss(input, target):
    if target.ndim == 1:
        target = target.unsqueeze(1)
    if input.shape[0] == 1:
        return torch.tensor(0, dtype=input.dtype, device=input.device)
    # margin is set according to input dimension
    margin = 10 * input.shape[1]
    # calculate distance of input
    dist = torch.cdist(input, input)
    # calculate similarity matrix
    sim = torch.eq(target, target.T)
    # change similarity matrix to -1 and 1
    sim = sim.float() * 2 - 1
    # calculate loss, but only for the upper triangle of the similarity matrix
    mask = (dist > margin).float() * (sim == -1).float()
    loss = dist * sim * (1 - mask)
    # mean over all pairs
    loss = loss.triu(diagonal=1).sum() / (target.shape[0] * (target.shape[0] - 1) / 2)
    return loss


class WeightedCombinedLoss(nn.modules.loss._WeightedLoss):
    """
    Weighted combined loss function.
    Input weight should be a tensor of shape (5,).
    The first 2 weights are for the patho/beni loss
    The last 3 weights are for the beni/gof/lof loss
    """
    def __init__(self, weight: Optional[torch.Tensor] = None, 
                 task_weight: float = 10.0,
                 size_average=None, ignore_index: int = -100,
                 reduce=None, reduction: str = 'mean') -> None:
        super().__init__(weight, size_average, reduce, reduction)
        self.ignore_index = ignore_index
        self.task_weight = task_weight

    def forward(self, input: torch.Tensor, target: torch.Tensor) -> torch.Tensor:
        return combined_loss(input, target,
                             weight_1=self.weight[:2],
                             weight_2=self.weight[2:],
                             weight=self.task_weight,
                             reduction=self.reduction)


class WeightedLoss1(nn.modules.loss._WeightedLoss):
    """
    Weighted combined loss function.
    Input weight should be a tensor of shape (5,).
    The first 2 weights are for the patho/beni loss
    The last 3 weights are for the beni/gof/lof loss
    """
    def __init__(self, weight: Optional[torch.Tensor] = None, 
                 task_weight: float = 10.0,
                 size_average=None, ignore_index: int = -100,
                 reduce=None, reduction: str = 'mean') -> None:
        super().__init__(weight, size_average, reduce, reduction)
        self.ignore_index = ignore_index
        self.task_weight = task_weight

    def forward(self, input: torch.Tensor, target: torch.Tensor, weight: torch.Tensor=None, reduce=True, reduction=None) -> torch.Tensor:
        # my custom loss function, target should be -1, 0, 1, 3.
        # -1 represents LoF
        # 0 represents neutral
        # 1 represents GoF
        # 3 represents pathogenic, but unknown LoF or GoF
        # input should be the output of the model, which is a 2D tensor with shape [batch_size, 2]
        # the first column is the probability of neutral / pathogenic,
        # the second column is the probability of GoF / LoF,
        # reshape target to 1D tensor
        # first, we transfer the target to 0, 1, 0 is neutral, 1 is pathogenic
        # if target.ndim == 2:
        #     target = target.squeeze(1)
        if reduction is None:
            reduction = self.reduction
        weight_1 = self.weight[:2]
        target_1 = (target).float()
        weight_loss_1 = torch.ones_like(target_1, dtype=input.dtype, device=input.device)
        weight_loss_1[target == 1] = weight_1[1] / weight_1[0]
        if weight is not None:
            weight_loss_1 *= weight
        loss_1 = binary_cross_entropy(input=input,
                                      target=target_1,
                                      weight=weight_loss_1,
                                      reduce=reduce,
                                      reduction=reduction)
        return loss_1


class WeightedLoss2(nn.modules.loss._WeightedLoss):
    """
    Weighted combined loss function.
    Input weight should be a tensor of shape (5,).
    The first 2 weights are for the patho/beni loss
    The last 3 weights are for the beni/gof/lof loss
    """
    def __init__(self, weight: Optional[torch.Tensor] = None, 
                 task_weight: float = 10.0,
                 size_average=None, ignore_index: int = -100,
                 reduce=None, reduction: str = 'mean') -> None:
        super().__init__(weight, size_average, reduce, reduction)
        self.ignore_index = ignore_index
        self.task_weight = task_weight

    def forward(self, input: torch.Tensor, target: torch.Tensor, weight: torch.Tensor=None, reduce=True, reduction=None) -> torch.Tensor:
        # my custom loss function, target should be -1, 0, 1, 3.
        # -1 represents LoF
        # 0 represents neutral
        # 1 represents GoF
        # 3 represents pathogenic, but unknown LoF or GoF
        # input should be the output of the model, which is a 2D tensor with shape [batch_size, 2]
        # the first column is the probability of neutral / pathogenic,
        # the second column is the probability of GoF / LoF,
        # reshape target to 1D tensor
        # first, we transfer the target to 0, 1, 0 is neutral, 1 is pathogenic
        # if target.ndim == 2:
        #     target = target.squeeze(1)
        # weight is unused
        target_1 = (-1/3 * target**3 + target**2 + 1/3 * target).float()
        if reduction is None:
            reduction = self.reduction
        loss_1 = binary_cross_entropy(input=input,
                                      target=target_1,
                                      weight=weight,
                                      reduce=reduce,
                                      reduction=reduction)
        # only do the calculation if target equals -1 or 1
        filter = (target == -1) | (target == 1)
        # if filter is all False, then loss_2 is 0
        if not filter.any():
            return 0 * loss_1
        # 1 is !!GoF!!, 0 is !!LoF!!
        weight_2 = self.weight[2:]
        # then, we transfer the target to 0, 1, 0 is !!GoF!!, 1 is !!LoF!!
        target_2 = (1/2 * (-target + 1)).float()
        # loss_2 is the cross entropy loss on pathogenic / neutral / GoF
        weight_loss_2 = torch.ones_like(target_2, dtype=input.dtype, device=input.device)
        weight_loss_2[target == 1] = weight_2[1] / weight_2[0]
        if weight is not None:
            weight_loss_2 *= weight
        loss_2 = binary_cross_entropy(input=input[filter],
                                      target=target_2[filter],
                                      weight=weight_loss_2[filter],
                                      reduce=reduce,
                                      reduction=reduction)
        return loss_2


class WeightedLoss3(nn.modules.loss._WeightedLoss):
    """
    Weighted combined loss function.
    Input weight should be a tensor of shape (5,).
    The first 2 weights are for the patho/beni loss
    The last 3 weights are for the beni/gof/lof loss
    """
    def __init__(self, weight: Optional[torch.Tensor] = None, 
                 task_weight: float = 10.0,
                 size_average=None, ignore_index: int = -100,
                 reduce=None, reduction: str = 'mean') -> None:
        super().__init__(weight, size_average, reduce, reduction)
        self.ignore_index = ignore_index
        self.task_weight = task_weight

    def forward(self, input: torch.Tensor, target: torch.Tensor, weight: torch.Tensor=None, reduce=True, reduction=None) -> torch.Tensor:
        # my custom loss function, target should be -1, 0, 1, 3.
        # -1 represents LoF
        # 0 represents neutral
        # 1 represents GoF
        # 3 represents pathogenic, but unknown LoF or GoF
        # input should be the output of the model, which is a 2D tensor with shape [batch_size, 2]
        # the first column is the probability of neutral / pathogenic,
        # the second column is the probability of GoF / LoF,
        # reshape target to 1D tensor
        # first, we transfer the target to 0, 1, 0 is neutral, 1 is pathogenic
        # if target.ndim == 2:
        #     target = target.squeeze(1)
        # weight is unused
        # input should be the output of the model, which is a 2D tensor with shape [batch_size, 2]
        target_1 = (-1/3 * target**3 + target**2 + 1/3 * target).float()
        if reduction is None:
            reduction = self.reduction
        loss_1 = binary_cross_entropy(input=input[:, 0]/(input[:, 0] + input[:, 1]),
                                      target=target_1,
                                      weight=weight,
                                      reduce=reduce,
                                      reduction=reduction)
        # only do the calculation if target equals -1 or 1
        filter = (target == -1) | (target == 1)
        # if filter is all False, then loss_2 is 0
        if not filter.any():
            return 0 * loss_1
        # 1 is !!GoF!!, 0 is !!LoF!!
        weight_2 = self.weight[2:]
        # then, we transfer the target to 0, 1, 0 is !!GoF!!, 1 is !!LoF!!
        target_2 = (1/2 * (-target + 1)).float()
        # loss_2 is the cross entropy loss on pathogenic / neutral / GoF
        weight_loss_2 = torch.ones_like(target_2, dtype=input.dtype, device=input.device)
        weight_loss_2[target == 1] = weight_2[1] / weight_2[0]
        if weight is not None:
            weight_loss_2 *= weight
        loss_2 = -BetaBinomial(
                        concentration1=input[:, 0][filter],
                        concentration0=input[:, 1][filter],
                        total_count=1
                    ).log_prob(target_2[filter])
        # loss_2 times weights
        loss_2 *= weight_loss_2[filter]
        # mean over all pairs
        loss_2 = loss_2.mean()
        return loss_2


class RegressionWeightedLoss(nn.modules.loss._WeightedLoss):
    """
    Weighted combined loss function.
    Input weight should be a tensor of shape (5,).
    The first 2 weights are for the patho/beni loss
    The last 3 weights are for the beni/gof/lof loss
    """
    def __init__(self, weight: Optional[torch.Tensor] = None, 
                 task_weight: float = 10.0,
                 size_average=None, ignore_index: int = -100,
                 reduce=None, reduction: str = 'mean') -> None:
        super().__init__(weight, size_average, reduce, reduction)
        self.ignore_index = ignore_index
        self.task_weight = task_weight

    def forward(self, input, target) -> torch.Tensor:
        # my custom loss function, target should be -1, 0, 1, 3.
        # -1 represents LoF
        # 0 represents neutral
        # 1 represents GoF
        # 3 represents pathogenic, but unknown LoF or GoF
        # input should be the output of the model, which is a 2D tensor with shape [batch_size, 2]
        # the first column is the probability of neutral / pathogenic,
        # the second column is the probability of GoF / LoF,
        # reshape target to 1D tensor
        # first, we transfer the target to 0, 1, 0 is neutral, 1 is pathogenic
        # if target.ndim == 2:
        #     target = target.squeeze(1)
        regression_target = target[:, 1:]
        regression_input = input[:, 1:]
        regression_loss = mse_loss(input=regression_input,
                                   target=regression_target,
                                   reduction=self.reduction)
        target = target[:, [0]]
        input = input[:, [0]]
        target_1 = (-1/3 * target**3 + target**2 + 1/3 * target).float()
        loss_1 = binary_cross_entropy(input=input,
                                      target=target_1,
                                      reduction=self.reduction)
        # only do the calculation if target equals -1 or 1
        filter = (target == -1) | (target == 1)
        # if filter is all False, then loss_2 is 0
        if not filter.any():
            return 0 * loss_1
        # 1 is !!GoF!!, 0 is !!LoF!!
        weight_2 = self.weight[2:]
        # then, we transfer the target to 0, 1, 0 is !!GoF!!, 1 is !!LoF!!
        target_2 = (1/2 * (-target + 1)).float()
        # loss_2 is the cross entropy loss on pathogenic / neutral / GoF
        weight_loss_2 = torch.ones_like(target_2, dtype=input.dtype, device=input.device)
        weight_loss_2[target == 1] = weight_2[1] / weight_2[0]
        loss_2 = binary_cross_entropy(input=input[filter],
                                      target=target_2[filter],
                                      weight=weight_loss_2[filter],
                                      reduction=self.reduction)
        return loss_2 + regression_loss


class GPLoss(nn.modules.loss._WeightedLoss):
    def __init__():
        super().__init__()


def combined_loss(input: torch.Tensor, target: torch.Tensor,
                  weight: float=10.0, 
                  weight_1: Optional[torch.Tensor]=None, 
                  weight_2: Optional[torch.Tensor]=None,
                  reduction: str = 'mean') -> torch.Tensor:
    # my custom loss function, target should be -1, 0, 1, 3.
    # -1 represents LoF
    # 0 represents neutral
    # 1 represents GoF
    # 3 represents pathogenic, but unknown LoF or GoF
    # input should be the output of the model, which is a 2D tensor with shape [batch_size, 2]
    # the first column is the probability of neutral / pathogenic,
    # the second column is the probability of GoF / LoF,
    # reshape target to 1D tensor
    if target.ndim == 2:
        target = target.squeeze(1)
    # first, we transfer the target to 0, 1, 0 is neutral, 1 is pathogenic
    target_1 = (-1/3 * target**3 + target**2 + 1/3 * target).float()
    # then, we transfer the target to 0, 1, 0 is LoF, 1 is GoF
    target_2 = (1/2 * (target + 1)).float()
    # loss_1 is the cross entropy loss on pathogenic / neutral
    # define weight for loss_1
    weight_loss_1 = torch.ones_like(target_1, dtype=input.dtype, device=input.device)
    weight_loss_1[target_1 == 1] = weight_1[0] / weight_1[1]
    loss_1 = binary_cross_entropy(input=input[:, 0],
                                  target=target_1,
                                  weight=weight_loss_1,
                                  reduction=reduction)
    # loss_2 is the cross entropy loss on pathogenic / neutral / GoF
    # only do the calculation if target equals -1 or 1
    filter = (target == -1) | (target == 1)
    # if filter is all False, then loss_2 is 0
    if not filter.any():
        return loss_1
    weight_loss_2 = torch.ones_like(target_2, dtype=input.dtype, device=input.device)
    weight_loss_2[target_2 == 1] = weight_2[0] / weight_2[1]
    loss_2 = binary_cross_entropy(input=input[filter, 1],
                                  target=target_2[filter],
                                  weight=weight_loss_2[filter],
                                  reduction=reduction)
    # assume LoF / GoF task is more important, so we add a weight of 10 to loss_2
    # if no benign variants, ignore loss_1
    if not (target == 0).any():
        loss = loss_2
    else:
        loss = loss_1 + weight * loss_2
    return loss


def gaussian_loss(input: torch.Tensor, target: torch.Tensor):
    # input should be the output of the model, which is a 2D tensor with shape [batch_size, 2]
    # the first column is the mean of the gaussian distribution,
    # the second column is the standard deviation of the gaussian distribution,
    # we should add another loss to control the standard deviation
    loss = -Normal(loc=input[:, 0], scale=torch.nn.functional.softplus(input[:, 1])).log_prob(target).mean()
    loss += torch.nn.functional.softplus(input[:, 1]).mean()
    return loss


def mse_loss_weighted(input: torch.Tensor, target: torch.Tensor, weight: torch.Tensor=None, reduce=True, reduction=None) -> torch.Tensor:
    # calculate mean squared loss, but times weight
    mse = (input - target).pow(2)
    if weight is not None:
        mse *= weight
    if reduce:
        return mse.mean()
    else:
        return mse


loss_fn_mapping = {
    "mse_loss": mse_loss,
    "mse_loss_weighted": mse_loss_weighted,
    "l1_loss": l1_loss,
    "binary_cross_entropy": binary_cross_entropy,
    "cross_entropy": cross_entropy,
    "kl_div": kl_div,
    "cosin_contrastive_loss": cosin_contrastive_loss,
    "euclid_contrastive_loss": euclid_contrastive_loss,
    "combined_loss": combined_loss,
    "weighted_combined_loss": WeightedCombinedLoss,
    "weighted_loss": WeightedLoss2,
    "weighted_loss_betabinomial": WeightedLoss3,
    "gaussian_loss": gaussian_loss,
    "weighted_loss_pretrain": WeightedLoss1,   
    "regression_weighted_loss": RegressionWeightedLoss,
    "GP_loss": GPLoss,
}


def get_template_fn(template):
    if template == 'plain-distance':
        return plain_distance, 1
    elif template == 'exp-normal-smearing-distance':
        return exp_normal_smearing_distance, 50

def plain_distance(pos):
    eps=1e-10
    CA = pos[..., 3, :]  # [b, n_res, 5, 3] -> [b, n_res, 3]
    d = (eps + (CA[..., None, :, :] - CA[..., :, None, :]).pow(2).sum(dim=-1, keepdims=True)) ** 0.5
    return d

def exp_normal_smearing_distance(pos, cutoff_upper=100, cutoff_lower=0, num_rbf=50):
    alpha = 5.0 / (cutoff_upper - cutoff_lower)
    start_value = torch.exp(
        torch.scalar_tensor(-cutoff_upper + cutoff_lower)
    ).to(pos.device)
    means = torch.linspace(start_value, 1, num_rbf).to(pos.device)
    betas = torch.tensor(
        [(2 / num_rbf * (1 - start_value)) ** -2] * num_rbf
    ).to(pos.device)
    dist = plain_distance(pos)
    cutoffs = 0.5 * (torch.cos(dist * math.pi / cutoff_upper).to(pos.device) + 1.0)
    # remove contributions beyond the cutoff radius
    cutoffs = cutoffs * (dist < cutoff_upper).float()
    return cutoffs * torch.exp(
        -betas * (torch.exp(alpha * (-dist + cutoff_lower)) - means) ** 2
    )