regression.py

from torch.utils.data import DataLoader
import math
from sentence_transformers import SentenceTransformer,  LoggingHandler, losses, models, util
from sentence_transformers.evaluation import EmbeddingSimilarityEvaluator
from sentence_transformers.readers import InputExample
import logging
from datetime import datetime
import sys
import os
import gzip
import csv


sts_dataset_path = './stsbenchmark.tsv.gz'


# Load bert-base-uncased as the model for the regression task
model_name = 'bert-base-uncased'

# Set the traning parameters
train_batch_size = 16
num_epochs = 100
model_save_path = 'regression_output/regression_training_stsbenchmark_'+model_name.replace("/", "-")

word_embedding_model = models.Transformer(model_name)

# Apply mean pooling to get one fixed sized sentence vector
pooling_model = models.Pooling(word_embedding_model.get_word_embedding_dimension(),
                               pooling_mode_mean_tokens=True,
                               pooling_mode_cls_token=False,
                               pooling_mode_max_tokens=False)

model = SentenceTransformer(modules=[word_embedding_model, pooling_model])


print('Starting to read training dataset...')
train_samples = []
dev_samples = []
test_samples = []
with gzip.open(sts_dataset_path, 'rt', encoding='utf8') as fIn:
    reader = csv.DictReader(fIn, delimiter='\t', quoting=csv.QUOTE_NONE)
    for row in reader:
        score = float(row['score']) / 5.0  # Normalize score to range 0 ... 1
        inp_example = InputExample(texts=[row['sentence1'], row['sentence2']], label=score)

        if row['split'] == 'dev':
            dev_samples.append(inp_example)
        elif row['split'] == 'test':
            test_samples.append(inp_example)
        else:
            train_samples.append(inp_example)


train_dataloader = DataLoader(train_samples, shuffle=True, batch_size=train_batch_size)
train_loss = losses.CosineSimilarityLoss(model=model)

evaluator = EmbeddingSimilarityEvaluator.from_input_examples(dev_samples, name='sts-dev')

warmup_steps = math.ceil(len(train_dataloader) * num_epochs  * 0.1) #10% of train data for warm-up

print('Starting to train the model')

model.fit(train_objectives=[(train_dataloader, train_loss)],
          evaluator=evaluator,
          epochs=num_epochs,
          evaluation_steps=1000,
          warmup_steps=warmup_steps,
          output_path=model_save_path)


print('Model has been trained successfully. Storing it in - '+ model_save_path)

#Load the saved model
model = SentenceTransformer(model_save_path)

print('Starting to evaluate the model on the STSBenchmark Dataset')
test_evaluator = EmbeddingSimilarityEvaluator.from_input_examples(test_samples, name='sts-test')
test_evaluator(model, output_path=model_save_path)