deepscreen/models/predictors/mol_trans.py

import math
import copy

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


class MolTrans(nn.Module):
    """
        Interaction Network with 2D interaction map
    """
    def __init__(
            self,
            input_dim_drug: 23532,
            input_dim_target: 16693,
            max_drug_seq,
            max_protein_seq,
            emb_size: 384,
            dropout_rate: 0.1,
            # DenseNet
            scale_down_ratio: 0.25,
            growth_rate: 20,
            transition_rate: 0.5,
            num_dense_blocks: 4,
            kernal_dense_size: 3,
            # Encoder
            intermediate_size: 1536,
            num_attention_heads: 12,
            attention_probs_dropout_prob: 0.1,
            hidden_dropout_prob: 0.1,
            # flatten_dim: 78192,
            # batch_size
    ):
        super().__init__()
        self.max_d = max_drug_seq
        self.max_p = max_protein_seq
        self.emb_size = emb_size
        self.dropout_rate = dropout_rate

        # densenet
        self.scale_down_ratio = scale_down_ratio
        self.growth_rate = growth_rate
        self.transition_rate = transition_rate
        self.num_dense_blocks = num_dense_blocks
        self.kernal_dense_size = kernal_dense_size
        # self.batch_size = batch_size
        self.input_dim_drug = input_dim_drug
        self.input_dim_target = input_dim_target
        self.n_layer = 2

        # encoder
        self.hidden_size = emb_size
        self.intermediate_size = intermediate_size
        self.num_attention_heads = num_attention_heads
        self.attention_probs_dropout_prob = attention_probs_dropout_prob
        self.hidden_dropout_prob = hidden_dropout_prob

        # self.flatten_dim = flatten_dim

        # specialized embedding with positional one
        self.demb = Embeddings(self.input_dim_drug, self.emb_size, self.max_d, self.dropout_rate)
        self.pemb = Embeddings(self.input_dim_target, self.emb_size, self.max_p, self.dropout_rate)

        self.d_encoder = EncoderMultipleLayers(self.n_layer, self.hidden_size, self.intermediate_size,
                                               self.num_attention_heads, self.attention_probs_dropout_prob,
                                               self.hidden_dropout_prob)
        self.p_encoder = EncoderMultipleLayers(self.n_layer, self.hidden_size, self.intermediate_size,
                                               self.num_attention_heads, self.attention_probs_dropout_prob,
                                               self.hidden_dropout_prob)

        self.icnn = nn.Conv2d(1, 3, 3, padding=0)

        self.decoder = nn.Sequential(
            # nn.Linear(self.flatten_dim, 512),
            nn.LazyLinear(512),
            nn.ReLU(True),

            nn.BatchNorm1d(512),
            nn.Linear(512, 64),
            nn.ReLU(True),

            nn.BatchNorm1d(64),
            nn.Linear(64, 32),
            nn.ReLU(True),

            # # output layer
            # nn.Linear(32, 1)
        )

    def forward(self, v_d, v_p):
        d, d_mask = v_d
        p, p_mask = v_p
        ex_d_mask = d_mask.unsqueeze(1).unsqueeze(2)
        ex_p_mask = p_mask.unsqueeze(1).unsqueeze(2)

        ex_d_mask = (1.0 - ex_d_mask) * -10000.0
        ex_p_mask = (1.0 - ex_p_mask) * -10000.0

        d_emb = self.demb(d)  # batch_size x seq_length x embed_size
        p_emb = self.pemb(p)

        batch_size = d_emb.size(0)

        # set output_all_encoded_layers be false, to obtain the last layer hidden states only.
        d_encoded_layers = self.d_encoder(d_emb.float(), ex_d_mask.float())
        # print(d_encoded_layers.shape)
        p_encoded_layers = self.p_encoder(p_emb.float(), ex_p_mask.float())
        # print(p_encoded_layers.shape)

        # repeat to have the same tensor size for aggregation
        d_aug = torch.unsqueeze(d_encoded_layers, 2).repeat(1, 1, self.max_p, 1)  # repeat along protein size
        p_aug = torch.unsqueeze(p_encoded_layers, 1).repeat(1, self.max_d, 1, 1)  # repeat along drug size

        i = d_aug * p_aug  # interaction
        i_v = i.view(batch_size, -1, self.max_d, self.max_p)
        # batch_size x embed size x max_drug_seq_len x max_protein_seq_len
        i_v = torch.sum(i_v, dim=1)
        i_v = torch.unsqueeze(i_v, 1)

        i_v = F.dropout(i_v, p=self.dropout_rate)

        i = self.icnn(i_v).view(batch_size, -1)
        score = self.decoder(i)
        return score


class LayerNorm(nn.Module):
    def __init__(self, hidden_size, variance_epsilon=1e-12):
        super(LayerNorm, self).__init__()
        self.gamma = nn.Parameter(torch.ones(hidden_size))
        self.beta = nn.Parameter(torch.zeros(hidden_size))
        self.variance_epsilon = variance_epsilon

    def forward(self, x):
        u = x.mean(-1, keepdim=True)
        s = (x - u).pow(2).mean(-1, keepdim=True)
        x = (x - u) / torch.sqrt(s + self.variance_epsilon)
        return self.gamma * x + self.beta


class Embeddings(nn.Module):
    """Construct the embeddings from protein/target, position embeddings.
    """

    def __init__(self, vocab_size, hidden_size, max_position_size, dropout_rate):
        super(Embeddings, self).__init__()
        self.word_embeddings = nn.Embedding(vocab_size, hidden_size)
        self.position_embeddings = nn.Embedding(max_position_size, hidden_size)

        self.LayerNorm = LayerNorm(hidden_size)
        self.dropout = nn.Dropout(dropout_rate)

    def forward(self, input_ids):
        seq_length = input_ids.size(1)
        position_ids = torch.arange(seq_length, dtype=torch.long, device=input_ids.device)
        position_ids = position_ids.unsqueeze(0).expand_as(input_ids)

        words_embeddings = self.word_embeddings(input_ids)
        position_embeddings = self.position_embeddings(position_ids)

        embeddings = words_embeddings + position_embeddings
        embeddings = self.LayerNorm(embeddings)
        embeddings = self.dropout(embeddings)
        return embeddings


class SelfAttention(nn.Module):
    def __init__(self, hidden_size, num_attention_heads, attention_probs_dropout_prob):
        super(SelfAttention, self).__init__()
        if hidden_size % num_attention_heads != 0:
            raise ValueError(
                "The hidden size (%d) is not a multiple of the number of attention "
                "heads (%d)" % (hidden_size, num_attention_heads))
        self.num_attention_heads = num_attention_heads
        self.attention_head_size = int(hidden_size / num_attention_heads)
        self.all_head_size = self.num_attention_heads * self.attention_head_size

        self.query = nn.Linear(hidden_size, self.all_head_size)
        self.key = nn.Linear(hidden_size, self.all_head_size)
        self.value = nn.Linear(hidden_size, self.all_head_size)

        self.dropout = nn.Dropout(attention_probs_dropout_prob)

    def transpose_for_scores(self, x):
        new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.attention_head_size)
        x = x.view(*new_x_shape)
        return x.permute(0, 2, 1, 3)

    def forward(self, hidden_states, attention_mask):
        mixed_query_layer = self.query(hidden_states)
        mixed_key_layer = self.key(hidden_states)
        mixed_value_layer = self.value(hidden_states)

        query_layer = self.transpose_for_scores(mixed_query_layer)
        key_layer = self.transpose_for_scores(mixed_key_layer)
        value_layer = self.transpose_for_scores(mixed_value_layer)

        # Take the dot product between "query" and "key" to get the raw attention scores.
        attention_scores = torch.matmul(query_layer, key_layer.transpose(-1, -2))
        attention_scores = attention_scores / math.sqrt(self.attention_head_size)

        attention_scores = attention_scores + attention_mask

        # Normalize the attention scores to probabilities.
        attention_probs = nn.Softmax(dim=-1)(attention_scores)

        # This is actually dropping out entire tokens to attend to, which might
        # seem a bit unusual, but is taken from the original Transformer paper.
        attention_probs = self.dropout(attention_probs)

        context_layer = torch.matmul(attention_probs, value_layer)
        context_layer = context_layer.permute(0, 2, 1, 3).contiguous()
        new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,)
        context_layer = context_layer.view(*new_context_layer_shape)
        return context_layer


class SelfOutput(nn.Module):
    def __init__(self, hidden_size, hidden_dropout_prob):
        super(SelfOutput, self).__init__()
        self.dense = nn.Linear(hidden_size, hidden_size)
        self.LayerNorm = LayerNorm(hidden_size)
        self.dropout = nn.Dropout(hidden_dropout_prob)

    def forward(self, hidden_states, input_tensor):
        hidden_states = self.dense(hidden_states)
        hidden_states = self.dropout(hidden_states)
        hidden_states = self.LayerNorm(hidden_states + input_tensor)
        return hidden_states


class Attention(nn.Module):
    def __init__(self, hidden_size, num_attention_heads, attention_probs_dropout_prob, hidden_dropout_prob):
        super(Attention, self).__init__()
        self.self = SelfAttention(hidden_size, num_attention_heads, attention_probs_dropout_prob)
        self.output = SelfOutput(hidden_size, hidden_dropout_prob)

    def forward(self, input_tensor, attention_mask):
        self_output = self.self(input_tensor, attention_mask)
        attention_output = self.output(self_output, input_tensor)
        return attention_output


class Intermediate(nn.Module):
    def __init__(self, hidden_size, intermediate_size):
        super(Intermediate, self).__init__()
        self.dense = nn.Linear(hidden_size, intermediate_size)

    def forward(self, hidden_states):
        hidden_states = self.dense(hidden_states)
        hidden_states = F.relu(hidden_states)
        return hidden_states


class Output(nn.Module):
    def __init__(self, intermediate_size, hidden_size, hidden_dropout_prob):
        super(Output, self).__init__()
        self.dense = nn.Linear(intermediate_size, hidden_size)
        self.LayerNorm = LayerNorm(hidden_size)
        self.dropout = nn.Dropout(hidden_dropout_prob)

    def forward(self, hidden_states, input_tensor):
        hidden_states = self.dense(hidden_states)
        hidden_states = self.dropout(hidden_states)
        hidden_states = self.LayerNorm(hidden_states + input_tensor)
        return hidden_states


class Encoder(nn.Module):
    def __init__(self, hidden_size, intermediate_size, num_attention_heads, attention_probs_dropout_prob,
                 hidden_dropout_prob):
        super(Encoder, self).__init__()
        self.attention = Attention(hidden_size, num_attention_heads, attention_probs_dropout_prob, hidden_dropout_prob)
        self.intermediate = Intermediate(hidden_size, intermediate_size)
        self.output = Output(intermediate_size, hidden_size, hidden_dropout_prob)

    def forward(self, hidden_states, attention_mask):
        attention_output = self.attention(hidden_states, attention_mask)
        intermediate_output = self.intermediate(attention_output)
        layer_output = self.output(intermediate_output, attention_output)
        return layer_output


class EncoderMultipleLayers(nn.Module):
    def __init__(self, n_layer, hidden_size, intermediate_size, num_attention_heads, attention_probs_dropout_prob,
                 hidden_dropout_prob):
        super().__init__()
        layer = Encoder(hidden_size, intermediate_size, num_attention_heads, attention_probs_dropout_prob,
                        hidden_dropout_prob)
        self.layer = nn.ModuleList([copy.deepcopy(layer) for _ in range(n_layer)])

    def forward(self, hidden_states, attention_mask, output_all_encoded_layers=True):
        all_encoder_layers = []
        for layer_module in self.layer:
            hidden_states = layer_module(hidden_states, attention_mask)
            # if output_all_encoded_layers:
            #    all_encoder_layers.append(hidden_states)
        # if not output_all_encoded_layers:
        #    all_encoder_layers.append(hidden_states)
        return hidden_states