def score(self, examples, return_enc_state=False):
args = self.args
if not isinstance(examples, list):
examples = [examples]
src_words = [e.src for e in examples]
tgt_words = [e.tgt for e in examples]
src_length = [len(c) for c in src_words]
seqs_x = to_input_variable(src_words, self.vocab.src, cuda=args.cuda, batch_first=True)
seqs_y = to_input_variable(tgt_words, self.vocab.tgt, max_len=self.max_len,
cuda=args.cuda,
append_boundary_sym=True,
batch_first=True)
# log_probs = self.forward(seqs_x, seqs_y, x_length=src_length, to_word=False)['prob']
# # y_label = seqs_y[:, 1:].contiguous()
# y_label = seqs_y.contiguous()
#
# words_norm = y_label.ne(-1).float().sum(1)
#
# loss = self.critic(inputs=log_probs, labels=y_label, reduce=False, normalization=self.normalization)
#
# if self.norm_by_words:
# loss = loss.div(words_norm).sum()
# else:
# loss = loss.sum()
# return loss
enc_out = self.encode(seqs_x, seqs_length=src_length)
# check: batch_size, hidden_size
return self.decoder.score(raw_score=enc_out, tgt_var=seqs_y)
def score(self, examples, return_enc_state=False):
args = self.args
if not isinstance(examples, list):
examples = [examples]
src_words = [e.src for e in examples]
src_length = [len(c) for c in src_words]
src_var = to_input_variable(src_words,
self.vocab.src,
cuda=args.cuda,
batch_first=True)
tgt_var = to_input_variable(src_words,
self.vocab.tgt,
cuda=args.cuda,
append_boundary_sym=True,
batch_first=True)
encoder_outputs, encoder_hidden = self.encode_var(
src_var=src_var, src_length=src_length)
scores = self.decoder.score(inputs=tgt_var,
encoder_hidden=encoder_hidden,
encoder_outputs=encoder_outputs)
enc_states = self.decoder.init_state(encoder_hidden)
if return_enc_state:
return scores, enc_states
else:
return scores
def encode_to_hidden(self, examples, need_sort=False):
if not isinstance(examples, list):
examples = [examples]
if not need_sort:
sent_words = [e.src for e in examples]
length = [len(e.src) for e in examples]
src_var = to_input_variable(sent_words,
self.vocab.src,
training=False,
cuda=self.args.cuda,
batch_first=True)
encoder_output, encoder_hidden = self.encode(input_var=src_var,
length=length)
return {
"outputs": encoder_output,
"hidden": encoder_hidden,
'length': length,
'batch_size': len(examples)
}
sent_words = [e.src for e in examples]
batch_size = len(sent_words)
sent_lengths = [len(sent_word) for sent_word in sent_words]
sorted_example_ids = sorted(range(batch_size),
key=lambda x: -sent_lengths[x])
example_old_pos_map = [-1] * batch_size
for new_pos, old_pos in enumerate(sorted_example_ids):
example_old_pos_map[old_pos] = new_pos
sorted_sent_words = [sent_words[i] for i in sorted_example_ids]
sorted_sent_var = to_input_variable(sorted_sent_words,
self.vocab.src,
cuda=self.args.cuda,
batch_first=True)
if self.training and self.args.src_wd:
sorted_sent_var = unk_replace(sorted_sent_var, self.step_unk_rate,
self.vocab.src)
sorted_sent_lengths = [
len(sent_word) for sent_word in sorted_sent_words
]
_, sent_hidden = self.encode(sorted_sent_var, sorted_sent_lengths)
if sent_hidden.dim() > 2:
hidden = sent_hidden[:, example_old_pos_map, :]
else:
hidden = sent_hidden[example_old_pos_map, :]
return {
"outputs": None,
"hidden": hidden,
'length': sent_lengths,
'batch_size': batch_size
}
def greedy_search(self, examples, to_word=True):
args = self.args
if not isinstance(examples, list):
examples = [examples]
src_words = [e.src for e in examples]
src_var = to_input_variable(src_words,
self.vocab.src,
cuda=args.cuda,
batch_first=True)
src_length = [len(c) for c in src_words]
encoder_outputs, encoder_hidden = self.encode_var(
src_var=src_var, src_length=src_length)
decoder_output, decoder_hidden, ret_dict, _ = self.decoder.forward(
encoder_hidden=encoder_hidden,
encoder_outputs=encoder_outputs,
teacher_forcing_ratio=0.0)
result = torch.stack(ret_dict['sequence']).squeeze()
final_result = []
example_nums = result.size(1)
if to_word:
for i in range(example_nums):
hyp = result[:, i].data.tolist()
res = id2word(hyp, self.vocab)
seems = [[res], [len(res)]]
final_result.append(seems)
return final_result
def beam_search(self, src_sent, beam_size=5, dmts=None):
if dmts is None:
dmts = self.args.tgt_max_time_step
src_var = to_input_variable(src_sent,
self.vocab.src,
cuda=self.args.cuda,
training=False,
append_boundary_sym=False,
batch_first=True)
src_length = [len(src_sent)]
encoder_outputs, encoder_hidden = self.encode_var(
src_var=src_var, src_length=src_length)
meta_data = self.beam_decoder.beam_search(
encoder_hidden=encoder_hidden,
encoder_outputs=encoder_outputs,
beam_size=beam_size,
decode_max_time_step=dmts)
topk_sequence = meta_data['sequence']
topk_score = meta_data['score'].squeeze()
completed_hypotheses = torch.cat(topk_sequence, dim=-1)
number_return = completed_hypotheses.size(0)
final_result = []
final_scores = []
for i in range(number_return):
hyp = completed_hypotheses[i, :].data.tolist()
res = id2word(hyp, self.vocab.tgt)
final_result.append(res)
final_scores.append(topk_score[i].item())
return final_result, final_scores
def sentence_encode(self, sent_words):
batch_size = len(sent_words)
sent_lengths = [len(sent_word) for sent_word in sent_words]
sorted_example_ids = sorted(range(batch_size),
key=lambda x: -sent_lengths[x])
example_old_pos_map = [-1] * batch_size
for new_pos, old_pos in enumerate(sorted_example_ids):
example_old_pos_map[old_pos] = new_pos
sorted_sent_words = [sent_words[i] for i in sorted_example_ids]
sorted_sent_var = to_input_variable(sorted_sent_words,
self.vocab.src,
cuda=self.args.cuda,
batch_first=True)
if self.training and self.args.src_wd:
sorted_sent_var = unk_replace(sorted_sent_var, self.step_unk_rate,
self.vocab.src)
sorted_sent_lengths = [
len(sent_word) for sent_word in sorted_sent_words
]
_, sent_hidden = self.encoder.forward(sorted_sent_var,
sorted_sent_lengths)
hidden = sent_hidden[:, example_old_pos_map, :]
return hidden
def predict(self, examples, to_word=True):
args = self.args
if not isinstance(examples, list):
examples = [examples]
src_words = [e.src for e in examples]
src_length = [len(c) for c in src_words]
seqs_x = to_input_variable(src_words, self.vocab.src, cuda=args.cuda, batch_first=True)
predict = self.forward(seqs_x, x_length=src_length, to_word=to_word)
return predict['pred']
def encode_to_hidden(self, examples):
if not isinstance(examples, list):
examples = [examples]
batch_size = len(examples)
sorted_example_ids = sorted(range(batch_size),
key=lambda x: -len(examples[x].tgt))
example_old_pos_map = [-1] * batch_size
sorted_examples = [examples[i] for i in sorted_example_ids]
syntax_word = [e.tgt for e in sorted_examples]
syntax_var = to_input_variable(syntax_word,
self.vocab.tgt,
training=False,
cuda=self.args.cuda,
batch_first=True)
length = [len(e.tgt) for e in sorted_examples]
syntax_output, syntax_hidden = self.syntax_encode(syntax_var, length)
sent_words = [e.src for e in sorted_examples]
sentence_hidden = self.sentence_encode(sent_words)
tgt_var = to_input_variable(sent_words,
self.vocab.src,
training=False,
cuda=self.args.cuda,
append_boundary_sym=True,
batch_first=True)
for new_pos, old_pos in enumerate(sorted_example_ids):
example_old_pos_map[old_pos] = new_pos
return {
'hidden': sentence_hidden,
"syn_output": syntax_output,
"syn_hidden": syntax_hidden,
'tgt_var': tgt_var,
'old_pos': example_old_pos_map
}
def score(self, examples, return_enc_state=False):
"""
Used for teacher-forcing training,
return the log_probability of <input,output>.
"""
args = self.args
if isinstance(examples, list):
src_words = [e.src for e in examples]
tgt_words = [e.tgt for e in examples]
else:
src_words = examples.src
tgt_words = examples.tgt
seqs_x = to_input_variable(src_words,
self.src_vocab,
cuda=args.cuda,
batch_first=True)
seqs_y = to_input_variable(tgt_words,
self.tgt_vocab,
cuda=args.cuda,
append_boundary_sym=True,
batch_first=True)
y_inp = seqs_y[:, :-1].contiguous()
y_label = seqs_y[:, 1:].contiguous()
words_norm = y_label.ne(self.pad).float().sum(1)
log_probs = self.forward(seqs_x, y_inp)
loss = self.critic(inputs=log_probs,
labels=y_label,
reduce=False,
normalization=self.normalization)
if self.norm_by_words:
loss = loss.div(words_norm).sum()
else:
loss = loss.sum()
return loss
def encode(self, examples):
args = self.args
if isinstance(examples, list):
src_words = [e.src for e in examples]
else:
src_words = examples.src
src_var = to_input_variable(src_words,
self.vocab.src,
cuda=args.cuda,
batch_first=True)
src_length = [len(c) for c in src_words]
encoder_outputs, encoder_hidden = self.encode_var(src_var, src_length)
return encoder_hidden
def forward(self, examples):
if not isinstance(examples, list):
examples = [examples]
batch_size = len(examples)
sent_words = [e.src for e in examples]
ret = self.encode_to_hidden(examples)
ret = self.hidden_to_latent(ret=ret, is_sampling=self.training)
ret = self.latent_for_init(ret=ret)
decode_init = ret['decode_init']
tgt_var = to_input_variable(sent_words,
self.vocab.src,
training=False,
cuda=self.args.cuda,
append_boundary_sym=True,
batch_first=True)
decode_init = self.bridger.forward(decode_init)
if self.training and self.args.tgt_wd > 0.:
input_var = unk_replace(tgt_var, self.step_unk_rate,
self.vocab.src)
tgt_token_scores = self.decoder.generate(
con_inputs=input_var,
encoder_hidden=decode_init,
encoder_outputs=None,
teacher_forcing_ratio=1.0,
)
reconstruct_loss = -torch.sum(
self.decoder.score_decoding_results(tgt_token_scores, tgt_var))
else:
reconstruct_loss = -torch.sum(
self.decoder.score(
inputs=tgt_var,
encoder_outputs=None,
encoder_hidden=decode_init,
))
return {
"mean": ret['mean'],
"logv": ret['logv'],
"z": ret['latent'],
'nll_loss': reconstruct_loss,
'batch_size': batch_size
}
def forward(self, examples, is_dis=False):
if not isinstance(examples, list):
examples = [examples]
batch_size = len(examples)
words = [e.src for e in examples]
tgt_var = to_input_variable(words,
self.vocab.src,
training=False,
cuda=self.args.cuda,
append_boundary_sym=True,
batch_first=True)
syn_seqs = [e.tgt for e in examples]
syn_var = to_input_variable(syn_seqs,
self.vocab.tgt,
training=False,
cuda=self.args.cuda,
append_boundary_sym=True,
batch_first=True)
ret = self.encode_to_hidden(examples)
ret = self.hidden_to_latent(ret=ret, is_sampling=self.training)
ret = self.latent_for_init(ret=ret)
syn_hidden = ret['syn_hidden']
sem_hidden = ret['sem_hidden']
if is_dis:
dis_syn_loss, dis_sem_loss = self.get_dis_loss(
syntax_hidden=syn_hidden,
semantic_hidden=sem_hidden,
syn_tgt=syn_var,
sem_tgt=tgt_var)
ret['dis syn'] = dis_syn_loss
ret['dis sem'] = dis_sem_loss
return ret
decode_init = ret['decode_init']
sentence_decode_init = self.bridger.forward(decode_init)
if self.training and self.args.tgt_wd:
input_var = unk_replace(tgt_var, self.step_unk_rate,
self.vocab.src)
tgt_log_score = self.decoder.generate(
con_inputs=input_var,
encoder_hidden=sentence_decode_init,
encoder_outputs=None,
teacher_forcing_ratio=1.0)
reconstruct_loss = -torch.sum(
self.decoder.score_decoding_results(tgt_log_score, tgt_var))
else:
reconstruct_loss = -torch.sum(
self.decoder.score(inputs=tgt_var,
encoder_outputs=None,
encoder_hidden=sentence_decode_init))
mul_syn_loss, mul_sem_loss = self.get_mul_loss(
syntax_hidden=syn_hidden,
semantic_hidden=sem_hidden,
syn_tgt=syn_var,
sem_tgt=tgt_var)
adv_syn_loss, adv_sem_loss = self.get_adv_loss(
syntax_hidden=syn_hidden,
semantic_hidden=sem_hidden,
syn_tgt=syn_var,
sem_tgt=tgt_var)
ret['adv'] = adv_syn_loss + adv_sem_loss
ret['mul'] = mul_syn_loss + mul_sem_loss
ret['nll_loss'] = reconstruct_loss
ret['sem_loss'] = mul_sem_loss
ret['syn_loss'] = mul_syn_loss
ret['batch_size'] = batch_size
return ret