File size: 12,402 Bytes
dcacefd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# Copyright (c) Facebook, Inc. and its affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.

from copy import deepcopy
import math
from typing import Union

import torch
import torch.nn as nn
import torch.nn.functional as F
from omegaconf import OmegaConf

import esm
from esm.modules import (
    TransformerLayer,
    LearnedPositionalEmbedding,
    SinusoidalPositionalEmbedding,
    RobertaLMHead,
    ESM1bLayerNorm,
    ContactPredictionHead,
    ESM1LayerNorm,
    FeedForwardNetwork,
    NormalizedResidualBlock,
    gelu,
)
from esm.multihead_attention import MultiheadAttention


def Cfg(**kwds):
    return OmegaConf.create(kwds)


def merge_config(default_cfg, override_cfg):
    return OmegaConf.merge(default_cfg, override_cfg)


class ESM2WithStructuralAdatper(nn.Module):
    @classmethod
    def from_pretrained(cls, args, override_args=None, name='esm2_t33_650M_UR50D'):
        import esm
        pretrained_model, alphabet = esm.pretrained.load_model_and_alphabet_hub(name)

        pretrained_args = Cfg(
            num_layers=pretrained_model.num_layers,
            embed_dim=pretrained_model.embed_dim,
            attention_heads=pretrained_model.attention_heads,
            token_dropout=pretrained_model.token_dropout,
        )
        args = merge_config(pretrained_args, args)
        # args.adapter_layer_indices = getattr(args, 'adapter_layer_indices', [6, 20, 32])

        args.adapter_layer_indices = [-1]
        args.adapter_layer_indices = list(
            map(lambda x: (args.num_layers + x) % args.num_layers,
                args.adapter_layer_indices)
        )
        #args.adapter_layer_indices = [6, 20, 32]

        model = cls(args, deepcopy(alphabet))
        model.load_state_dict(pretrained_model.state_dict(), strict=False)

        del pretrained_model

        # freeze pretrained parameters
        for pname, param in model.named_parameters():
            if 'adapter' not in pname:
                param.requires_grad = False
        return model

    def __init__(
        self,
        args,
        alphabet: Union[esm.data.Alphabet, str] = "ESM-1b",
        # num_layers: int = 33,
        # embed_dim: int = 1280,
        # attention_heads: int = 20,
        # token_dropout: bool = True,
    ):
        super().__init__()
        self.args = args
        self.num_layers = args.num_layers
        self.embed_dim = args.embed_dim
        self.attention_heads = args.attention_heads
        if not isinstance(alphabet, esm.data.Alphabet):
            alphabet = esm.data.Alphabet.from_architecture(alphabet)
        self.alphabet = alphabet
        self.alphabet_size = len(alphabet)
        self.padding_idx = alphabet.padding_idx
        self.mask_idx = alphabet.mask_idx
        self.cls_idx = alphabet.cls_idx
        self.eos_idx = alphabet.eos_idx
        self.prepend_bos = alphabet.prepend_bos
        self.append_eos = alphabet.append_eos
        self.token_dropout = args.token_dropout

        self._init_submodules()

    def _init_submodules(self):
        self.embed_scale = 1
        self.embed_tokens = nn.Embedding(
            self.alphabet_size,
            self.embed_dim,
            padding_idx=self.padding_idx,
        )

        self.layers = nn.ModuleList(
            [
                self._init_layer(_)
                for _ in range(self.num_layers)
            ]
        )

        self.contact_head = ContactPredictionHead(
            self.num_layers * self.attention_heads,
            self.prepend_bos,
            self.append_eos,
            eos_idx=self.eos_idx,
        )
        self.emb_layer_norm_after = ESM1bLayerNorm(self.embed_dim)

        self.lm_head = RobertaLMHead(
            embed_dim=self.embed_dim,
            output_dim=self.alphabet_size,
            weight=self.embed_tokens.weight,
        )

    def _init_layer(self, layer_idx):
        if layer_idx in self.args.adapter_layer_indices:
            layer = TransforerLayerWithStructralAdapter(
                self.embed_dim,
                4 * self.embed_dim,
                self.attention_heads,
                add_bias_kv=False,
                use_esm1b_layer_norm=True,
                use_rotary_embeddings=True,
                encoder_embed_dim=self.args.encoder.d_model,
                dropout=self.args.dropout
            )
        else:
            layer = TransformerLayer(
                self.embed_dim,
                4 * self.embed_dim,
                self.attention_heads,
                add_bias_kv=False,
                use_esm1b_layer_norm=True,
                use_rotary_embeddings=True,
            )
        return layer

    def forward_layers(self, x, encoder_out, padding_mask, repr_layers=[], hidden_representations=[], need_head_weights=False, attn_weights=[]):
        for layer_idx, layer in enumerate(self.layers):
            if layer_idx in self.args.adapter_layer_indices:
                x, attn = layer(
                    x, encoder_out, self_attn_padding_mask=padding_mask, need_head_weights=need_head_weights
                )
            else:
                x, attn = layer(
                    x, self_attn_padding_mask=padding_mask, need_head_weights=need_head_weights
                )
            if (layer_idx + 1) in repr_layers:
                hidden_representations[layer_idx + 1] = x.transpose(0, 1)
            if need_head_weights:
                # (H, B, T, T) => (B, H, T, T)
                attn_weights.append(attn.transpose(1, 0))

        return x, hidden_representations, attn_weights, layer_idx

    def forward(self, tokens, encoder_out, repr_layers=[], need_head_weights=False, return_contacts=False):
        if return_contacts:
            need_head_weights = True

        assert tokens.ndim == 2
        padding_mask = tokens.eq(self.padding_idx)  # B, T

        x = self.embed_scale * self.embed_tokens(tokens)

        if self.token_dropout:
            x.masked_fill_((tokens == self.mask_idx).unsqueeze(-1), 0.0)
            # x: B x T x C
            mask_ratio_train = 0.15 * 0.8
            src_lengths = (~padding_mask).sum(-1)
            mask_ratio_observed = (tokens == self.mask_idx).sum(-1).to(x.dtype) / src_lengths
            x = x * (1 - mask_ratio_train) / (1 - mask_ratio_observed)[:, None, None]

        if padding_mask is not None:
            x = x * (1 - padding_mask.unsqueeze(-1).type_as(x))

        repr_layers = set(repr_layers)
        hidden_representations = {}
        if 0 in repr_layers:
            hidden_representations[0] = x

        if need_head_weights:
            attn_weights = []

        # (B, T, E) => (T, B, E)
        x = x.transpose(0, 1)

        if not padding_mask.any():
            padding_mask = None

        # for layer_idx, layer in enumerate(self.layers):
        #     x, attn = layer(
        #         x,
        #         self_attn_padding_mask=padding_mask,
        #         need_head_weights=need_head_weights,
        #     )
        #     if (layer_idx + 1) in repr_layers:
        #         hidden_representations[layer_idx + 1] = x.transpose(0, 1)
        #     if need_head_weights:
        #         # (H, B, T, T) => (B, H, T, T)
        #         attn_weights.append(attn.transpose(1, 0))

        x, hidden_representations, attn_weights, layer_idx = self.forward_layers(
            x, encoder_out, padding_mask,
            repr_layers=repr_layers,
            hidden_representations=hidden_representations,
            need_head_weights=need_head_weights,
            attn_weights=attn_weights if need_head_weights else None
        )


        x = self.emb_layer_norm_after(x)
        x = x.transpose(0, 1)  # (T, B, E) => (B, T, E)

        # last hidden representation should have layer norm applied
        if (layer_idx + 1) in repr_layers:
            hidden_representations[layer_idx + 1] = x
        x = self.lm_head(x)

        result = {"logits": x, "representations": hidden_representations}
        if need_head_weights:
            # attentions: B x L x H x T x T
            attentions = torch.stack(attn_weights, 1)
            if padding_mask is not None:
                attention_mask = 1 - padding_mask.type_as(attentions)
                attention_mask = attention_mask.unsqueeze(1) * attention_mask.unsqueeze(2)
                attentions = attentions * attention_mask[:, None, None, :, :]
            result["attentions"] = attentions
            if return_contacts:
                contacts = self.contact_head(tokens, attentions)
                result["contacts"] = contacts

        return result

    def predict_contacts(self, tokens):
        return self(tokens, return_contacts=True)["contacts"]


class TransforerLayerWithStructralAdapter(nn.Module):
    def __init__(
        self,
        embed_dim,
        ffn_embed_dim,
        attention_heads,
        encoder_embed_dim,
        add_bias_kv=True,
        use_esm1b_layer_norm=False,
        use_rotary_embeddings: bool = False,
        dropout=0.1,
    ):
        super().__init__()
        self.embed_dim = embed_dim
        self.ffn_embed_dim = ffn_embed_dim
        self.attention_heads = attention_heads
        self.use_rotary_embeddings = use_rotary_embeddings

        self.encoder_embed_dim = encoder_embed_dim
        self.dropout = dropout
        self._init_submodules(add_bias_kv, use_esm1b_layer_norm)


    def _init_submodules(self, add_bias_kv, use_esm1b_layer_norm):
        BertLayerNorm = ESM1bLayerNorm if use_esm1b_layer_norm else ESM1LayerNorm

        self.self_attn = MultiheadAttention(
            self.embed_dim,
            self.attention_heads,
            add_bias_kv=add_bias_kv,
            add_zero_attn=False,
            use_rotary_embeddings=self.use_rotary_embeddings,
        )
        self.self_attn_layer_norm = BertLayerNorm(self.embed_dim)

        self.fc1 = nn.Linear(self.embed_dim, self.ffn_embed_dim)
        self.fc2 = nn.Linear(self.ffn_embed_dim, self.embed_dim)

        self.final_layer_norm = BertLayerNorm(self.embed_dim)

        # structural adapter
        self.structural_adapter_attn = NormalizedResidualBlock(
            layer=MultiheadAttention(
                self.embed_dim,
                self.attention_heads,
                kdim=self.encoder_embed_dim,
                vdim=self.encoder_embed_dim,
                add_bias_kv=add_bias_kv,
                add_zero_attn=False,
                use_rotary_embeddings=True,
            ),
            embedding_dim=self.embed_dim,
            dropout=self.dropout
        )
        self.structural_adapter_ffn = NormalizedResidualBlock(
            layer=FeedForwardNetwork(
                self.embed_dim,
                self.embed_dim // 2, # NOTE: bottleneck FFN is important
                # self.ffn_embed_dim,
                activation_dropout=self.dropout
            ),
            embedding_dim=self.embed_dim,
            dropout=self.dropout
        )

    def forward(
        self, x, encoder_out, self_attn_mask=None, self_attn_padding_mask=None, need_head_weights=False
    ):
        residual = x
        x = self.self_attn_layer_norm(x)
        x, attn = self.self_attn(
            query=x,
            key=x,
            value=x,
            key_padding_mask=self_attn_padding_mask,
            need_weights=True,
            need_head_weights=need_head_weights,
            attn_mask=self_attn_mask,
        )
        x = residual + x

        # x = self.forward_adapter(x, encoder_out, attn_mask=self_attn_mask, attn_padding_mask=self_attn_padding_mask)

        residual = x
        x = self.final_layer_norm(x)
        x = gelu(self.fc1(x))
        x = self.fc2(x)
        x = residual + x

        x = x + self.forward_adapter(x, encoder_out, attn_mask=self_attn_mask, attn_padding_mask=self_attn_padding_mask)
        return x, attn

    def forward_adapter(self, x, encoder_out, attn_mask, attn_padding_mask):
        encoder_feats = encoder_out['feats']
        encoder_feats = encoder_feats.transpose(0, 1)

        x = self.structural_adapter_attn(
            x,
            key=encoder_feats,
            value=encoder_feats,
            key_padding_mask=attn_padding_mask,
            attn_mask=attn_mask,
            need_weights=False
        )[0]

        x = self.structural_adapter_ffn(x)
        # x = x.transpose(0, 1)
        return x