def predict(self, text_list):
result = []
print(text_list)
test_examples = self.processor.get_ifrn_examples(text_list)
print("test_examples", test_examples[0].text_a)
test_features = convert_examples_to_features(test_examples,
self.label_list,
self.args.max_seq_length,
self.tokenizer,
show_exp=False)
all_input_ids = torch.tensor([f.input_ids for f in test_features],
dtype=torch.long)
all_input_mask = torch.tensor([f.input_mask for f in test_features],
dtype=torch.long)
all_segment_ids = torch.tensor([f.segment_ids for f in test_features],
dtype=torch.long)
test_data = TensorDataset(all_input_ids, all_input_mask,
all_segment_ids)
# Run prediction for full data
test_sampler = SequentialSampler(test_data)
test_dataloader = DataLoader(test_data,
sampler=test_sampler,
batch_size=self.args.eval_batch_size)
for idx, (input_ids, input_mask,
segment_ids) in enumerate(test_dataloader):
item = {}
input_ids = input_ids.to(self.device)
input_mask = input_mask.to(self.device)
segment_ids = segment_ids.to(self.device)
text = test_examples[idx].text_a
with torch.no_grad():
logits = self.model(input_ids, segment_ids, input_mask)
print("logits: ", logits)
logits = F.softmax(logits, dim=1)
print("logits2: ", logits)
pred = logits.max(1)[1]
print("pred: ", pred)
logits = logits.detach().cpu().numpy()[0].tolist()
print("logits3", logits)
if return_text:
item['text'] = text
item['label'] = pred.item()
item['scores'] = {0: logits[0], 1: logits[1]}
result.append(item)
return result
def get_matrix_of_similarities(arguments=[]):
tokenizer = BertTokenizer.from_pretrained(model_path, do_lower_case=True)
if len(arguments) == 0:
arguments = arguments_init[:]
input_examples, output_examples, index_mapping = create_examples(arguments)
eval_features = convert_examples_to_features(input_examples,
max_seq_length, tokenizer)
all_input_ids = torch.tensor([f.input_ids for f in eval_features],
dtype=torch.long)
all_input_mask = torch.tensor([f.input_mask for f in eval_features],
dtype=torch.long)
all_segment_ids = torch.tensor([f.segment_ids for f in eval_features],
dtype=torch.long)
eval_data = TensorDataset(all_input_ids, all_input_mask, all_segment_ids)
eval_sampler = SequentialSampler(eval_data)
eval_dataloader = DataLoader(eval_data,
sampler=eval_sampler,
batch_size=eval_batch_size)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = SigmoidBERT.from_pretrained(model_path, )
model.to(device)
model.eval()
predicted_logits = []
with torch.no_grad():
for input_ids, input_mask, segment_ids in eval_dataloader:
input_ids = input_ids.to(device)
input_mask = input_mask.to(device)
segment_ids = segment_ids.to(device)
logits = model(input_ids, segment_ids, input_mask)
logits = logits.detach().cpu().numpy()
predicted_logits.extend(logits[:, 0])
distance_matrix = np.zeros(shape=(len(arguments), len(arguments)))
for idx in range(len(predicted_logits)):
output_examples[idx].append(predicted_logits[idx])
for i in range(len(arguments)):
for j in range(len(arguments)):
if i == j:
distance_matrix[i, j] = 0.0
elif (i, j) in index_mapping:
distance_matrix[i,
j] = 1.0 - predicted_logits[index_mapping[(i,
j)]]
else:
distance_matrix[i,
j] = 1.0 - predicted_logits[index_mapping[(j,
i)]]
#Sort by similarity
output_examples = sorted(output_examples, key=lambda x: x[2], reverse=True)
# print("Predicted similarities (sorted by similarity):")
# for idx in range(len(output_examples)):
# example = output_examples[idx]
# print("Sentence A:", example[0])
# print("Sentence B:", example[1])
# print("Similarity:", example[2])
# print("")
return distance_matrix
'Overwhelming scientific consensus says human activity is primarily responsible for global climate change.',
'Rising levels of human-produced gases released into the atmosphere create a greenhouse effect that traps heat and causes global warming.'
]
#Compare every argument with each other
input_examples = []
output_examples = []
for i in range(0, len(arguments) - 1):
for j in range(i + 1, len(arguments)):
input_examples.append(
InputExample(text_a=arguments[i], text_b=arguments[j], label=-1))
output_examples.append([arguments[i], arguments[j]])
tokenizer = BertTokenizer.from_pretrained(model_path, do_lower_case=True)
eval_features = convert_examples_to_features(input_examples, max_seq_length,
tokenizer)
all_input_ids = torch.tensor([f.input_ids for f in eval_features],
dtype=torch.long)
all_input_mask = torch.tensor([f.input_mask for f in eval_features],
dtype=torch.long)
all_segment_ids = torch.tensor([f.segment_ids for f in eval_features],
dtype=torch.long)
eval_data = TensorDataset(all_input_ids, all_input_mask, all_segment_ids)
eval_sampler = SequentialSampler(eval_data)
eval_dataloader = DataLoader(eval_data,
sampler=eval_sampler,
batch_size=eval_batch_size)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")