def get_prediction(in_sentences): labels = ["Negative", "Positive"] input_examples = [run_classifier.InputExample(guid="", text_a = x, text_b = None, label = 0) for x in in_sentences] # here, "" is just a dummy label input_features = run_classifier.convert_examples_to_features(input_examples, label_list, MAX_SEQ_LENGTH, tokenizer) predict_input_fn = run_classifier.input_fn_builder(features=input_features, seq_length=MAX_SEQ_LENGTH, is_training=False, drop_remainder=False) predictions = estimator.predict(predict_input_fn, checkpoint_path=checkpoint_path) return [(sentence, prediction['probabilities'], labels[prediction['labels']]) for sentence, prediction in zip(in_sentences, predictions)]
def run_train(data_file_path, output_dir):
print('***** Model output directory: {} *****'.format(output_dir))
# get data from data loader
train, _, _ = ContextualRelevance(data_file_path).get_data()
print(train.columns)
# Use the InputExample class from BERT's run_classifier code to create examples from the data
train_InputExamples = train.apply(
lambda x: run_classifier.InputExample(
guid=
None, # Globally unique ID for bookkeeping, unused in this example
text_a=x[DATA_COLUMN],
text_b=x[ANSWER_COLUMN],
label=x[LABEL_COLUMN]),
axis=1)
# get bert_code tokenizer form hub model
tokenizer = create_tokenizer_from_hub_module(BERT_MODEL_HUB, False)
print(tokenizer.tokenize("מריצים אימון..."))
# Convert our train and test features to InputFeatures that BERT understands.
train_features = run_classifier.convert_examples_to_features(
train_InputExamples, label_list, MAX_SEQ_LENGTH, tokenizer)
# Compute # train and warmup steps from batch size
num_train_steps = int(len(train_features) / BATCH_SIZE * NUM_TRAIN_EPOCHS)
num_warmup_steps = int(num_train_steps * WARMUP_PROPORTION)
# Specify outpit directory and number of checkpoint steps to save
run_config = tf.compat.v1.estimator.RunConfig(
model_dir=output_dir,
save_summary_steps=SAVE_SUMMARY_STEPS,
save_checkpoints_steps=SAVE_CHECKPOINTS_STEPS)
model_fn = model_fn_builder(num_labels=len(label_list),
learning_rate=LEARNING_RATE,
num_train_steps=num_train_steps,
num_warmup_steps=num_warmup_steps,
bert_model_hub=BERT_MODEL_HUB)
estimator = tf.compat.v1.estimator.Estimator(
model_fn=model_fn,
config=run_config,
params={"batch_size": BATCH_SIZE})
# Create an input function for training. drop_remainder = True for using TPUs.
train_input_fn = run_classifier.input_fn_builder(features=train_features,
seq_length=MAX_SEQ_LENGTH,
is_training=True,
drop_remainder=False)
print('Beginning Training!')
current_time = datetime.now()
estimator.train(input_fn=train_input_fn, max_steps=num_train_steps)
print("Training took time ", datetime.now() - current_time)
def run(checkpoint_path, data_flie_path):
# get model (make sure to change checkpoint according to the model in the configurations file)
_, test, false_negatives_test_set = ContextualRelevance(
data_flie_path).get_data()
# get bert_code tokenizer form hub model
tokenizer = create_tokenizer_from_hub_module(BERT_MODEL_HUB)
test_InputExamples = test.apply(
lambda x: run_classifier.InputExample(guid=None,
text_a=x[DATA_COLUMN],
text_b=x[ANSWER_COLUMN],
label=x[LABEL_COLUMN]),
axis=1)
test_features = run_classifier.convert_examples_to_features(
test_InputExamples, label_list, MAX_SEQ_LENGTH, tokenizer)
test_input_fn = run_classifier.input_fn_builder(features=test_features,
seq_length=MAX_SEQ_LENGTH,
is_training=False,
drop_remainder=False)
model_fn = model_fn_builder(num_labels=len(label_list),
learning_rate=LEARNING_RATE,
num_train_steps=num_train_steps,
num_warmup_steps=num_warmup_steps,
bert_model_hub=BERT_MODEL_HUB)
estimator = tf.compat.v1.estimator.Estimator(
model_fn, params={"batch_size": BATCH_SIZE})
metric_result = estimator.evaluate(input_fn=test_input_fn,
steps=None,
checkpoint_path=checkpoint_path)
metric_result['false_negatives'] += false_negatives_test_set
metric_result['recall'] = metric_result['true_positives'] / (
metric_result['true_positives'] + metric_result['false_negatives'])
metric_result['eval_accuracy'] = (
metric_result['true_positives'] + metric_result['true_negatives']
) / (metric_result['true_positives'] + metric_result['false_negatives'] +
metric_result['true_negatives'] + metric_result['false_positives'])
precision = metric_result['precision']
recall = metric_result['recall']
metric_result['F1'] = 2 * (precision * recall) / (precision + recall)
return metric_result
lambda x: run_classifier.InputExample(
guid=
None, # Globally unique ID for bookkeeping, unused in this example
text_a=x[DATA_COLUMN],
text_b=None,
label=x[LABEL_COLUMN]),
axis=1)
# get bert_code tokenizer form hub model
tokenizer = create_tokenizer_from_hub_module(BERT_MODEL_HUB)
test_InputExamples = test.apply(lambda x: run_classifier.InputExample(
guid=None, text_a=x[DATA_COLUMN], text_b=None, label=x[LABEL_COLUMN]),
axis=1)
test_features = run_classifier.convert_examples_to_features(
test_InputExamples, label_list, MAX_SEQ_LENGTH, tokenizer)
test_input_fn = run_classifier.input_fn_builder(features=test_features,
seq_length=MAX_SEQ_LENGTH,
is_training=False,
drop_remainder=False)
model_fn = model_fn_builder(num_labels=len(label_list),
learning_rate=LEARNING_RATE,
num_train_steps=num_train_steps,
num_warmup_steps=num_warmup_steps,
bert_model_hub=BERT_MODEL_HUB)
estimator = tf.compat.v1.estimator.Estimator(model_fn,
params={"batch_size": BATCH_SIZE})
tokenizer = create_tokenizer_from_hub_module(BERT_MODEL_HUB)