def predict(self, model, dataloader, exam, feats, prediction_file, need_sp_logit_file=False):
model.eval()
answer_dict = {}
sp_dict = {}
# dataloader.refresh()
total_test_loss = [0] * 5
# context_idxs, context_mask, segment_idxs,
# query_mapping, all_mapping,
# ids, y1, y2, q_type,
# start_mapping,
# is_support
tqdm_obj = tqdm(dataloader, ncols=80)
for step, batch in enumerate(tqdm_obj):
batch = tuple(t.to(self.device) for t in batch)
# batch['context_mask'] = batch['context_mask'].float()
start_logits, end_logits, type_logits, sp_logits, start_position, end_position = model(*batch)
loss1 = self.criterion(start_logits, batch[6]) + self.criterion(end_logits, batch[7])#y1,y2
loss2 = self.config.type_lambda * self.criterion(type_logits, batch[8])#q_type
loss3 = self.config.sp_lambda * self.sp_loss_fct(sp_logits.view(-1), batch[10].float().view(-1)).sum() / batch[9].sum() # start_mapping
loss = loss1 + loss2 + loss3
loss_list = [loss, loss1, loss2, loss3]
for i, l in enumerate(loss_list):
if not isinstance(l, int):
total_test_loss[i] += l.item()
batchsize = batch[0].size(0)
#ids
answer_dict_ = convert_to_tokens(exam, feats, batch[5], start_position.data.cpu().numpy().tolist(),
end_position.data.cpu().numpy().tolist(),
np.argmax(type_logits.data.cpu().numpy(), 1))
answer_dict.update(answer_dict_)
predict_support_np = torch.sigmoid(sp_logits).data.cpu().numpy()
for i in range(predict_support_np.shape[0]):
cur_sp_pred = []
cur_id = batch[5][i].item()
cur_sp_logit_pred = [] # for sp logit output
for j in range(predict_support_np.shape[1]):
if j >= len(exam[cur_id].sent_names):
break
if need_sp_logit_file:
temp_title, temp_id = exam[cur_id].sent_names[j]
cur_sp_logit_pred.append((temp_title, temp_id, predict_support_np[i, j]))
if predict_support_np[i, j] > self.config.sp_threshold:
cur_sp_pred.append(exam[cur_id].sent_names[j])
sp_dict.update({cur_id: cur_sp_pred})
new_answer_dict = {}
for key, value in answer_dict.items():
new_answer_dict[key] = value.replace(" ", "")
prediction = {'answer': new_answer_dict, 'sp': sp_dict}
with open(prediction_file, 'w', encoding='utf8') as f:
json.dump(prediction, f, indent=4, ensure_ascii=False)
for i, l in enumerate(total_test_loss):
print("Test Loss{}: {}".format(i, l / len(dataloader)))
def predict(model,
dataloader,
example_dict,
feature_dict,
prediction_file,
need_sp_logit_file=False):
model.eval()
answer_dict = {}
sp_dict = {}
dataloader.refresh()
total_test_loss = [0] * 5
for batch in tqdm(dataloader):
batch['context_mask'] = batch['context_mask'].float()
start_logits, end_logits, type_logits, sp_logits, start_position, end_position = model(
batch)
loss_list = compute_loss(batch, start_logits, end_logits, type_logits,
sp_logits, start_position, end_position)
for i, l in enumerate(loss_list):
if not isinstance(l, int):
total_test_loss[i] += l.item()
answer_dict_ = convert_to_tokens(
example_dict, feature_dict, batch['ids'],
start_position.data.cpu().numpy().tolist(),
end_position.data.cpu().numpy().tolist(),
np.argmax(type_logits.data.cpu().numpy(), 1))
answer_dict.update(answer_dict_)
predict_support_np = torch.sigmoid(sp_logits).data.cpu().numpy()
for i in range(predict_support_np.shape[0]):
cur_sp_pred = []
cur_id = batch['ids'][i]
cur_sp_logit_pred = [] # for sp logit output
for j in range(predict_support_np.shape[1]):
if j >= len(example_dict[cur_id].sent_names):
break
if need_sp_logit_file:
temp_title, temp_id = example_dict[cur_id].sent_names[j]
cur_sp_logit_pred.append(
(temp_title, temp_id, predict_support_np[i, j]))
if predict_support_np[i, j] > args.sp_threshold:
cur_sp_pred.append(example_dict[cur_id].sent_names[j])
sp_dict.update({cur_id: cur_sp_pred})
new_answer_dict = {}
for key, value in answer_dict.items():
new_answer_dict[key] = value.replace(" ", "")
prediction = {'answer': new_answer_dict, 'sp': sp_dict}
with open(prediction_file, 'w', encoding='utf8') as f:
json.dump(prediction, f, indent=4, ensure_ascii=False)
for i, l in enumerate(total_test_loss):
print("Test Loss{}: {}".format(i, l / len(dataloader)))
test_loss_record.append(sum(total_test_loss[:3]) / len(dataloader))
def bert_classification(self, content, question):
logger.info('1:{}'.format( content))
conv_dic = {}
conv_dic['_id'] = 0
conv_dic['context'] = self.process_context(content)
conv_dic['question'] = question
conv_dic["answer"] = ""
conv_dic['supporting_facts'] = []
rows = [conv_dic]
filename = "data/{}.json".format(time.time())
with open(filename, 'w', encoding='utf8') as fw:
json.dump(rows, fw, ensure_ascii=False, indent=4)
exam, feats, dataset = self.data.load_file(filename, False)
data_loader = DataLoader(dataset, batch_size=self.config.batch_size)
self.model.eval()
answer_dict = {}
sp_dict = {}
tqdm_obj = tqdm(data_loader, ncols=80)
for step, batch in enumerate(tqdm_obj):
batch = tuple(t.to(self.device) for t in batch)
start_logits, end_logits, type_logits, sp_logits, start_position, end_position = self.model(*batch)
batchsize = batch[0].size(0)
# ids
answer_dict_ = convert_to_tokens(exam, feats, batch[5], start_position.data.cpu().numpy().tolist(),
end_position.data.cpu().numpy().tolist(),
np.argmax(type_logits.data.cpu().numpy(), 1))
answer_dict.update(answer_dict_)
predict_support_np = torch.sigmoid(sp_logits).data.cpu().numpy()
for i in range(predict_support_np.shape[0]):
cur_sp_pred = []
cur_id = batch[5][i].item()
cur_sp_logit_pred = [] # for sp logit output
for j in range(predict_support_np.shape[1]):
if j >= len(exam[cur_id].sent_names):
break
if predict_support_np[i, j] > self.config.sp_threshold:
cur_sp_pred.append(exam[cur_id].sent_names[j])
sp_dict.update({cur_id: cur_sp_pred})
new_answer_dict = {}
for key, value in answer_dict.items():
new_answer_dict[key] = value.replace(" ", "")
prediction = {'answer': new_answer_dict, 'sp': sp_dict}
return {"data": prediction}
# loss2 = self.config.type_lambda * self.criterion(type_logits, batch[8]) # q_type
# # sp_value = self.sp_loss_fct(sp_logits.view(-1), batch[10].float().view(-1)).sum()
# # loss3 = self.config.sp_lambda * sp_value / batch[9].sum()
#
# loss = loss1 + loss2
# loss_list = [loss, loss1, loss2]
#
# for i, l in enumerate(loss_list):
# if not isinstance(l, int):
# total_test_loss[i] += l.item()
batchsize = batch[0].size(0)
# ids
answer_dict_ = convert_to_tokens(
exam, feats, batch[5],
start_position.data.cpu().numpy().tolist(),
end_position.data.cpu().numpy().tolist(),
np.argmax(type_logits.data.cpu().numpy(), 1))
answer_dict.update(answer_dict_)
predict_support_np = torch.sigmoid(sp_logits).data.cpu().numpy()
for i in range(predict_support_np.shape[0]):
cur_sp_pred = []
cur_id = batch[5][i].item()
cur_sp_logit_pred = [] # for sp logit output
for j in range(predict_support_np.shape[1]):
if j >= len(exam[cur_id].sent_names):
break
if predict_support_np[i, j] > config.sp_threshold: