# -*- encoding: utf-8 -*-
# @Author: Yihuai Lan
# @Time: 2021/08/21 04:36:25
# @File: groupatt.py
import random
import warnings
from typing import Tuple, Dict, Any
import torch
from torch import nn
from torch.nn import functional as F
from mwptoolkit.module.Encoder.rnn_encoder import GroupAttentionRNNEncoder
from mwptoolkit.module.Decoder.rnn_decoder import BasicRNNDecoder, AttentionalRNNDecoder
from mwptoolkit.module.Embedder.basic_embedder import BasicEmbedder
from mwptoolkit.utils.enum_type import SpecialTokens, NumMask
from mwptoolkit.loss.nll_loss import NLLLoss
[docs]class GroupATT(nn.Module):
"""
Reference:
Li et al. "Modeling Intra-Relation in Math Word Problems with Different Functional Multi-Head Attentions" in ACL 2019.
"""
def __init__(self, config, dataset):
super(GroupATT, self).__init__()
self.device = config['device']
self.bidirectional = config["bidirectional"]
self.hidden_size = config["hidden_size"]
self.decode_hidden_size = config['decode_hidden_size']
self.encoder_rnn_cell_type = config["encoder_rnn_cell_type"]
self.decoder_rnn_cell_type = config["decoder_rnn_cell_type"]
self.attention = config["attention"]
self.share_vocab = config["share_vocab"]
self.max_gen_len = config["max_output_len"]
self.teacher_force_ratio = config["teacher_force_ratio"]
self.embedding_size = config["embedding_size"]
self.num_layers = config["num_layers"]
self.dropout_ratio = config["dropout_ratio"]
self.vocab_size = len(dataset.in_idx2word)
self.symbol_size = len(dataset.out_idx2symbol)
self.mask_list = NumMask.number
self.out_symbol2idx = dataset.out_symbol2idx
self.out_idx2symbol = dataset.out_idx2symbol
self.in_word2idx = dataset.in_word2idx
self.in_idx2word = dataset.in_idx2word
if self.share_vocab:
self.sos_token_idx = self.in_word2idx[SpecialTokens.SOS_TOKEN]
else:
self.sos_token_idx = self.out_symbol2idx[SpecialTokens.SOS_TOKEN]
try:
self.out_sos_token = self.out_symbol2idx[SpecialTokens.SOS_TOKEN]
except:
self.out_sos_token = None
try:
self.out_eos_token = self.out_symbol2idx[SpecialTokens.EOS_TOKEN]
except:
self.out_eos_token = None
try:
self.out_pad_token = self.out_symbol2idx[SpecialTokens.PAD_TOKEN]
except:
self.out_pad_token = None
self.split_list = []
# chinese dataset
try:
self.split_list.append(self.in_word2idx['．'])
except:
pass
try:
self.split_list.append(self.in_word2idx["，"])
except:
pass
# english dataset
try:
self.split_list.append(self.in_word2idx["."])
except:
pass
try:
self.split_list.append(self.in_word2idx[","])
except:
pass
self.in_embedder = BasicEmbedder(self.vocab_size, self.embedding_size, self.dropout_ratio)
if self.share_vocab:
self.out_embedder = self.in_embedder
else:
self.out_embedder = BasicEmbedder(self.symbol_size, self.embedding_size, self.dropout_ratio)
self.encoder = GroupAttentionRNNEncoder(emb_size=self.embedding_size,
hidden_size=self.hidden_size,
n_layers=self.num_layers,
bidirectional=self.bidirectional,
rnn_cell=None,
rnn_cell_name=self.encoder_rnn_cell_type,
variable_lengths=False,
d_ff=2048,
dropout=self.dropout_ratio,
N=1)
self.decoder = AttentionalRNNDecoder(self.embedding_size, self.decode_hidden_size, self.hidden_size, self.num_layers, self.decoder_rnn_cell_type, self.dropout_ratio)
self.dropout = nn.Dropout(self.dropout_ratio)
self.generate_linear = nn.Linear(self.decode_hidden_size, self.symbol_size)
weight = torch.ones(self.symbol_size).to(config["device"])
pad = self.out_pad_token
self.loss = NLLLoss(weight, pad)
[docs] def process_gap_encoder_decoder(self, encoder_hidden):
if self.encoder_rnn_cell_type == 'lstm' and self.decoder_rnn_cell_type == 'lstm':
''' lstm -> lstm '''
encoder_hidden = self._init_state(encoder_hidden)
elif self.encoder_rnn_cell_type == 'gru' and self.decoder_rnn_cell_type == 'gru':
''' gru -> gru '''
encoder_hidden = self._init_state(encoder_hidden)
elif self.encoder_rnn_cell_type == 'gru' and self.decoder_rnn_cell_type == 'lstm':
''' gru -> lstm '''
encoder_hidden = (encoder_hidden, encoder_hidden)
encoder_hidden = self._init_state(encoder_hidden)
elif self.encoder_rnn_cell_type == 'lstm' and self.decoder_rnn_cell_type == 'gru':
''' lstm -> gru '''
encoder_hidden = encoder_hidden[0]
encoder_hidden = self._init_state(encoder_hidden)
return encoder_hidden
def _init_state(self, encoder_hidden):
if encoder_hidden is None:
return None
if isinstance(encoder_hidden, tuple):
encoder_hidden = tuple([self._cat_directions(h) for h in encoder_hidden])
else:
encoder_hidden = self._cat_directions(encoder_hidden)
return encoder_hidden
def _cat_directions(self, h):
if self.encoder.bidirectional:
h = torch.cat([h[0:h.size(0):2], h[1:h.size(0):2]], 2)
return h
[docs] def forward(self, seq, seq_length, target=None,output_all_layers=False) -> Tuple[
torch.Tensor, torch.Tensor, Dict[str, Any]]:
"""
:param torch.Tensor seq: input sequence, shape: [batch_size, seq_length].
:param torch.Tensor seq_length: the length of sequence, shape: [batch_size].
:param torch.Tensor | None target: target, shape: [batch_size, target_length], default None.
:param bool output_all_layers: return output of all layers if output_all_layers is True, default False.
:return : token_logits:[batch_size, output_length, output_size], symbol_outputs:[batch_size,output_length], model_all_outputs.
:rtype: tuple(torch.Tensor, torch.Tensor, dict)
"""
batch_size = seq.size(0)
device = seq.device
seq_emb = self.in_embedder(seq)
encoder_outputs, encoder_hidden, encoder_layer_outputs = self.encoder_forward(seq_emb, seq, seq_length,
output_all_layers)
decoder_inputs = self.init_decoder_inputs(target, device, batch_size)
token_logits, symbol_outputs, decoder_layer_outputs = self.decoder_forward(encoder_outputs, encoder_hidden,
decoder_inputs, target,
output_all_layers)
model_all_outputs = {}
if output_all_layers:
model_all_outputs['inputs_embedding'] = seq_emb
model_all_outputs.update(encoder_layer_outputs)
model_all_outputs.update(decoder_layer_outputs)
return token_logits, symbol_outputs, model_all_outputs
[docs] def calculate_loss(self, batch_data:dict) -> float:
"""Finish forward-propagating, calculating loss and back-propagation.
:param batch_data: one batch data. batch_data should include keywords 'question', 'ques len', 'equation'.
:return: loss value.
"""
seq = torch.tensor(batch_data['question']).to(self.device)
seq_length = torch.tensor(batch_data['ques len']).long()
target = torch.tensor(batch_data['equation']).to(self.device)
token_logits, _, _ = self.forward(seq,seq_length,target)
if self.share_vocab:
target = self.convert_in_idx_2_out_idx(target)
outputs = torch.nn.functional.log_softmax(token_logits, dim=-1)
self.loss.reset()
self.loss.eval_batch(outputs.view(-1, outputs.size(-1)), target.view(-1))
self.loss.backward()
return self.loss.get_loss()
[docs] def model_test(self, batch_data:dict) -> tuple:
"""Model test.
:param batch_data: one batch data.
:return: predicted equation, target equation.
batch_data should include keywords 'question', 'ques len', 'equation' and 'num list'.
"""
seq = torch.tensor(batch_data['question']).to(self.device)
seq_length = torch.tensor(batch_data['ques len']).long()
target = torch.tensor(batch_data['equation']).to(self.device)
num_list = batch_data['num list']
_, symbol_outputs, _ = self.forward(seq, seq_length)
if self.share_vocab:
target = self.convert_in_idx_2_out_idx(target)
all_outputs = self.convert_idx2symbol(symbol_outputs, num_list)
targets = self.convert_idx2symbol(target, num_list)
return all_outputs, targets
[docs] def predict(self,batch_data:dict,output_all_layers=False):
"""
predict samples without target.
:param dict batch_data: one batch data.
:param bool output_all_layers: return all layer outputs of model.
:return: token_logits, symbol_outputs, all_layer_outputs
"""
seq = torch.tensor(batch_data['question']).to(self.device)
seq_length = torch.tensor(batch_data['ques len']).long()
token_logits, symbol_outputs, model_all_outputs = self.forward(seq,seq_length,output_all_layers=output_all_layers)
return token_logits, symbol_outputs, model_all_outputs
[docs] def encoder_forward(self,seq_emb,seq,seq_length,output_all_layers=False):
encoder_outputs, encoder_hidden = self.encoder.forward(seq_emb,seq,self.split_list,seq_length)
encoder_hidden = self.process_gap_encoder_decoder(encoder_hidden)
all_layer_outputs={}
if output_all_layers:
all_layer_outputs['encoder_outputs']=encoder_outputs
all_layer_outputs['encoder_hidden']=encoder_hidden
return encoder_outputs,encoder_hidden,all_layer_outputs
[docs] def decoder_forward(self,encoder_outputs, encoder_hidden, decoder_inputs, target=None,output_all_layers=False):
if target is not None and random.random() < self.teacher_force_ratio:
if self.attention:
decoder_outputs, decoder_states = self.decoder(decoder_inputs, encoder_hidden, encoder_outputs)
else:
decoder_outputs, decoder_states = self.decoder(decoder_inputs, encoder_hidden)
token_logits = self.generate_linear(decoder_outputs)
outputs = token_logits.topk(1, dim=-1)[1]
else:
seq_len = decoder_inputs.size(1) if target is not None else self.max_gen_len
decoder_hidden = encoder_hidden
decoder_input = decoder_inputs[:, 0, :].unsqueeze(1)
decoder_outputs = []
token_logits = []
outputs = []
for idx in range(seq_len):
if self.attention:
decoder_output, decoder_hidden = self.decoder(decoder_input, decoder_hidden, encoder_outputs)
else:
decoder_output, decoder_hidden = self.decoder(decoder_input, decoder_hidden)
step_output = decoder_output.squeeze(1)
token_logit = self.generate_linear(step_output)
output = token_logit.topk(1, dim=-1)[1]
decoder_outputs.append(step_output)
token_logits.append(token_logit)
outputs.append(output)
if self.share_vocab:
output = self.convert_out_idx_2_in_idx(output)
decoder_input = self.out_embedder(output)
else:
decoder_input = self.out_embedder(output)
decoder_outputs = torch.stack(decoder_outputs, dim=1)
token_logits = torch.stack(token_logits, dim=1)
outputs = torch.stack(outputs,dim=1)
all_layer_outputs = {}
if output_all_layers:
all_layer_outputs['decoder_outputs']=decoder_outputs
all_layer_outputs['token_logits']=token_logits
all_layer_outputs['outputs']=outputs
return token_logits,outputs,all_layer_outputs
[docs] def decode(self, output):
device = output.device
batch_size = output.size(0)
decoded_output = []
for idx in range(batch_size):
decoded_output.append(self.in_word2idx[self.out_idx2symbol[output[idx]]])
decoded_output = torch.tensor(decoded_output).to(device).view(batch_size, -1)
return output
[docs] def convert_out_idx_2_in_idx(self, output):
device = output.device
batch_size = output.size(0)
seq_len = output.size(1)
decoded_output = []
for b_i in range(batch_size):
output_i = []
for s_i in range(seq_len):
output_i.append(self.in_word2idx[self.out_idx2symbol[output[b_i, s_i]]])
decoded_output.append(output_i)
decoded_output = torch.tensor(decoded_output).to(device).view(batch_size, -1)
return decoded_output
[docs] def convert_in_idx_2_out_idx(self, output):
device = output.device
batch_size = output.size(0)
seq_len = output.size(1)
decoded_output = []
for b_i in range(batch_size):
output_i = []
for s_i in range(seq_len):
output_i.append(self.out_symbol2idx[self.in_idx2word[output[b_i, s_i]]])
decoded_output.append(output_i)
decoded_output = torch.tensor(decoded_output).to(device).view(batch_size, -1)
return decoded_output
[docs] def convert_idx2symbol(self, output, num_list):
batch_size = output.size(0)
seq_len = output.size(1)
output_list = []
for b_i in range(batch_size):
num_len = len(num_list[b_i])
res = []
for s_i in range(seq_len):
idx = output[b_i][s_i]
if idx in [self.out_sos_token, self.out_eos_token, self.out_pad_token]:
break
symbol = self.out_idx2symbol[idx]
if "NUM" in symbol:
num_idx = self.mask_list.index(symbol)
if num_idx >= num_len:
res.append(symbol)
else:
res.append(num_list[b_i][num_idx])
else:
res.append(symbol)
output_list.append(res)
return output_list
def __str__(self):
info = super().__str__()
total = sum(p.numel() for p in self.parameters())
trainable = sum(p.numel() for p in self.parameters() if p.requires_grad)
parameters = "\ntotal parameters : {} \ntrainable parameters : {}".format(total, trainable)
return info + parameters