model.train_model(train,
pearson_corr=pearson_corr,
spearman_corr=spearman_corr,
mae=mean_absolute_error)
result, model_outputs, wrong_predictions = model.eval_model(
dev,
pearson_corr=pearson_corr,
spearman_corr=spearman_corr,
mae=mean_absolute_error)
predictions, raw_outputs = model.predict(test_sentence_pairs)
dev['predictions'] = model_outputs
test['predictions'] = predictions
dev = un_fit(dev, 'labels')
dev = un_fit(dev, 'predictions')
test = un_fit(test, 'predictions')
dev.to_csv(os.path.join(TEMP_DIRECTORY, RESULT_FILE),
header=True,
sep='\t',
index=False,
encoding='utf-8')
draw_scatterplot(dev, 'labels', 'predictions',
os.path.join(TEMP_DIRECTORY, RESULT_IMAGE), "Russian-English")
print_stat(dev, 'labels', 'predictions')
format_submission(df=test,
index=index,
language_pair="ru-en",
method="TransQuest",
path=os.path.join(TEMP_DIRECTORY, SUBMISSION_FILE),
index_type="Auto")
args=transformer_config)
model.train_model(train,
pearson_corr=pearson_corr,
spearman_corr=spearman_corr,
mae=mean_absolute_error)
result, model_outputs, wrong_predictions = model.eval_model(
dev,
pearson_corr=pearson_corr,
spearman_corr=spearman_corr,
mae=mean_absolute_error)
predictions, raw_outputs = model.predict(test_sentence_pairs)
dev['predictions'] = model_outputs
test['predictions'] = predictions
dev = un_fit(dev, 'labels')
dev = un_fit(dev, 'predictions')
test = un_fit(test, 'predictions')
dev.to_csv(os.path.join(TEMP_DIRECTORY, RESULT_FILE),
header=True,
sep='\t',
index=False,
encoding='utf-8')
draw_scatterplot(dev, 'labels', 'predictions',
os.path.join(TEMP_DIRECTORY, RESULT_IMAGE), "English-Chinese")
print_stat(dev, 'labels', 'predictions')
format_submission(df=test,
index=index,
language_pair="en-zh",
method="TransQuest",
path=os.path.join(TEMP_DIRECTORY, SUBMISSION_FILE))
for dev, test, index, language in zip(dev_list, test_list, index_list,
[*languages]):
dev = un_fit(dev, 'labels')
dev = un_fit(dev, 'predictions')
test = un_fit(test, 'predictions')
dev.to_csv(os.path.join(
TEMP_DIRECTORY,
RESULT_FILE.split(".")[0] + "_" + language + "." +
RESULT_FILE.split(".")[1]),
header=True,
sep='\t',
index=False,
encoding='utf-8')
draw_scatterplot(
dev, 'labels', 'predictions',
os.path.join(
TEMP_DIRECTORY,
RESULT_IMAGE.split(".")[0] + "_" + language + "." +
RESULT_IMAGE.split(".")[1]), language)
print_stat(dev, 'labels', 'predictions')
if language == "RU-EN":
format_submission(df=test,
index=index,
language_pair=language.lower(),
method="TransQuest",
path=os.path.join(
TEMP_DIRECTORY,
SUBMISSION_FILE.split(".")[0] + "_" + language +
"." + SUBMISSION_FILE.split(".")[1]),
index_type="Auto")