|
import os |
|
|
|
import datasets |
|
import json |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class RepoCodeGenConfig(datasets.BuilderConfig): |
|
|
|
def __init__(self, *args, **kwargs): |
|
super().__init__( |
|
*args, |
|
name= "repo_codegen", |
|
**kwargs, |
|
) |
|
|
|
|
|
class RepoCodeGen(datasets.GeneratorBasedBuilder): |
|
|
|
VERSION = datasets.Version("1.0.0") |
|
|
|
BUILDER_CONFIG_CLASS = RepoCodeGenConfig |
|
BUILDER_CONFIGS = [RepoCodeGenConfig()] |
|
DEFAULT_CONFIG_NAME = "repo_codegen" |
|
|
|
|
|
def _info(self): |
|
return datasets.DatasetInfo( |
|
features=datasets.Features({ |
|
"id": datasets.Value("int64"), |
|
"project": datasets.Value("string"), |
|
"module": datasets.Value("string"), |
|
"entry_point": datasets.Value("string"), |
|
"solution": datasets.Value("string"), |
|
"prompt": datasets.Value("string"), |
|
"target_function_prompt": datasets.Value("string"), |
|
"function_signature": datasets.Value("string"), |
|
"test": datasets.Value("string"), |
|
"test_list": datasets.Sequence(datasets.Value("string")), |
|
"docstring": datasets.Value("string"), |
|
"original_docstring": datasets.Value("string"), |
|
"docstring_tokens": datasets.Sequence(datasets.Value("string")), |
|
"check": datasets.Value("string"), |
|
"cross_context": datasets.Value("bool"), |
|
"isContained": datasets.Value("bool"), |
|
"line_coverage": datasets.Value("float"), |
|
}), |
|
supervised_keys=None, |
|
|
|
) |
|
|
|
def _split_generators(self, dl_manager): |
|
generators = [] |
|
|
|
|
|
for _files in [ |
|
"./short_context_False_shorten_contain_doc_False.json", |
|
"./short_context_True_shorten_contain_doc_False.json", |
|
"./short_context_True_shorten_contain_doc_True.json" |
|
]: |
|
if "short_context_True" in _files: |
|
name = "short_context" |
|
else: |
|
name = "long_context" |
|
|
|
if "shorten_contain_doc_False" in _files: |
|
name += "_nodoc" |
|
else: |
|
name += "_doc" |
|
files = dl_manager.download([_files]) |
|
|
|
generators.append( |
|
datasets.SplitGenerator( |
|
name=name, |
|
gen_kwargs={ |
|
"files": files, |
|
}, |
|
), |
|
) |
|
|
|
|
|
return generators |
|
|
|
def _generate_examples(self, files): |
|
key = 0 |
|
for file_idx, file in enumerate(files): |
|
with open(file, "r") as f: |
|
for dp in f: |
|
row = json.loads(dp) |
|
yield key, { |
|
"id": row["id"], |
|
"project": row["project"], |
|
"module": row["module"], |
|
"entry_point": row["entry_point"], |
|
"solution": row["solution"], |
|
"prompt": row["prompt"], |
|
"target_function_prompt": row["target_function_prompt"], |
|
"function_signature": row["function_signature"], |
|
"test": row["test"], |
|
"test_list": row["test_list"], |
|
"docstring": row["docstring"], |
|
"original_docstring": row["original_docstring"], |
|
"docstring_tokens": row["docstring_tokens"], |
|
"check": row["check"], |
|
"cross_context": row["cross_context"], |
|
"isContained": row["isContained"], |
|
"line_coverage": row["line_coverage"], |
|
} |
|
key += 1 |
