This dataset consists of all of the data used for code style probing.
-evaluation set is included in the Data zip
-all results are untokenized to allow us to retokenize with different NL prompts or to not use prompts.
	-non tokenized versions will also allow addressing of truncation
		
curated_eval_set.zip
-this file contains eval sets used in testing and getting scores in paper
		
Labeled_Code.tar.zip
-raw data for parallel corpora, these csvs contain code samples with transformations applied to them
-code is not truncated in these examples
-filtering is not done for cases where input == label all bq files will have same length as result

	-'full_data_acl',
		-training data for codet5 cluster classifier label is cluster labels input is code 'content' column
		
	-'data_full_uncommented_no_outlier.csv',
		-do not use, uncommenting transformation on subset of data
		
	-'bq_data_outliers_acl',
		-big query data with outliers, transformations derive from this data originally
		
	-'bq_data_no_outliers_acl',
		-big query data with outliers removed
		
	-'bq_uncommented_no_outlier.csv',
		-do not use, uncommenting transformation on no outlier subset
		
	-'bq_uncommented_outlier.csv',
		-parallel corpora for uncommenting
		-label: 'content' input: 'uncommented_content'
		
	-'bq_data_no_outlier_no_class.csv',
		-do not use, class transformation on no outlier subset
		
	-'bq_data_outlier_no_class.csv',
		-parallel corpora for removing class
		-uncommented_content got deleted, get from bq_uncommented_outlier.csv
		-label: 'uncommented_content' input: 'no_class_content'
		
	-'bq_data_uncased.csv',
		-parallel_corpora for removing casing
		-uncommented_content got deleted, get from bq_uncommented_outlier.csv
		-label: 'uncommented_content' input: 'casing_content'
		
	-'eval_set_no_class.csv',
		-eval set for testing class parallel corpora with all super() removed
		
	-'bq_data_uncomp_fixed_outlier.csv',
		-list comp parallel corpora
		-label is 'uncommented_content' input is uncomped_content
		
	-'bq_data_outlier_no_class_no_super.csv',
		-training data for class parallel corpora with super() removed
		-gold label is 'content', input is 'no_class_content'
		
	-'parallel_no_decorators.csv'
		-prallel corpora with decorators removed in labels 
		-gold label is 'uncommented_content', input is 'decorator_modified'
		
	-'bq_no_docstring_outlier_acl.csv'
		-updated docstring file containing code with docstrings removed
		-full length code samples 
	
	-'docstring_method_split.csv'
		-updated docstring file containing code with docstrings removed in method format instead of full files
		-whenever using docstrings tokenize this file and split as needed