Files
alpertunga-bile-prompt-gene…/generator/preprocess.py
T

141 lines
4.4 KiB
Python

from re import compile
from collections import namedtuple
from functools import lru_cache
def get_unique_list(sequence: list[str]) -> list[str]:
seen = set()
return [
x.strip() for x in sequence if not (x.strip() in seen or seen.add(x.strip()))
]
def remove_exact_keywords(line: str) -> list[str]:
find_empty_parantheses_regex = compile(r"\(\s*\)")
remove_nonprompts_regex = compile(r"[^a-zA-Z_\-()\[\]{}]*")
remove_nonweighters_regex = compile(r"[()\[\]{}]*")
remove_inside_regex = compile(r"[^()\[\]{}]*")
# remove exact prompts
prompts = get_unique_list(line.split(","))
pure_prompts = [] # contains prompt values
extracted_pure_prompts = set() # order doesn't matters, it contains prompt keywords
# remove exact keyword
for prompt in prompts:
tempPrompt = remove_nonprompts_regex.sub(
"", prompt
).lstrip() # from -> ((masterpiece:1.2)) | to -> ((masterpiece))
if len(tempPrompt) == 0:
continue
tempPrompt = remove_nonweighters_regex.sub(
"", tempPrompt
) # from -> ((masterpiece)) | to -> masterpiece
if tempPrompt in extracted_pure_prompts:
tempPrompt = remove_inside_regex.sub(
"", prompt
) # from -> ((masterpiece:1.2)) | to -> (())
if (
len(find_empty_parantheses_regex.findall(tempPrompt))
> 0 # find balanced parantheses count
):
# check balanced parantheses
inner_parant_count = tempPrompt.count("(")
outer_parant_count = tempPrompt.count(")")
if inner_parant_count == outer_parant_count:
continue
lowest_count = min(inner_parant_count, outer_parant_count)
"""
remove balanced parantheses
because it is going to be appended to latter or former prompt
"""
for _ in range(lowest_count):
tempPrompt = tempPrompt.replace("()", "")
pure_prompts.append(tempPrompt)
continue
extracted_pure_prompts.add(tempPrompt)
pure_prompts.append(prompt)
return pure_prompts
def fix_commas(string: str) -> str:
remove_multiwhitespaces_regex = compile(r"\s+")
remove_nonpromptcommas_regex = compile(r"(,\s){2,}")
temp_str = remove_multiwhitespaces_regex.sub(" ", string)
return remove_nonpromptcommas_regex.sub(", ", temp_str)
def fix_artifacts(string: str) -> str:
temp_string = string
ArtifactFix = namedtuple("ArtifactFix", ["regex", "new_str"])
replace_list = [
ArtifactFix(regex=compile(r"\(\s*,"), new_str="("),
ArtifactFix(regex=compile(r"\[\s*,"), new_str="["),
ArtifactFix(regex=compile(r"{\s*,"), new_str="{"),
ArtifactFix(regex=compile(r",\s*\)"), new_str=")"),
ArtifactFix(regex=compile(r",\s*\]"), new_str="]"),
ArtifactFix(regex=compile(r",\s*}"), new_str="}"),
]
# fixing the artifacts
for artifact_fix in replace_list:
temp_string = artifact_fix.regex.sub(artifact_fix.new_str, temp_string)
return temp_string
@lru_cache
def preprocess(line: str, preprocess_mode: str) -> str:
remove_scalarweights_regex = compile(r",\s*:[0-9]*\.?[0-9]+")
remove_emptyprompts_regex = compile(r",\s+[()\[\]{}]+\s*,")
remove_danglingparantheses_regex = compile(r"\B\s+|\s+\B")
temp_line = line.encode("ascii", "xmlcharrefreplace").decode()
temp_line = temp_line.encode("utf-8", "xmlcharrefreplace").decode()
temp_line = temp_line.replace("\xa0", " ")
temp_line = temp_line.replace("\n", ", ")
temp_line = fix_commas(temp_line)
temp_line = remove_scalarweights_regex.sub(
"", temp_line
) # from -> , 0.6 | to -> *empty string*
if preprocess_mode == "exact_keyword":
temp_line = ", ".join(remove_exact_keywords(temp_line))
temp_line = fix_commas(temp_line)
temp_line = fix_artifacts(temp_line)
temp_line = remove_emptyprompts_regex.sub(
",", temp_line
) # from -> , (((, | to -> ,
temp_line = remove_danglingparantheses_regex.sub("", temp_line).replace(
",", ", "
) # from -> (( ((prompt)) | to -> ((prompt))
elif preprocess_mode == "exact_prompt":
temp_line = ", ".join(get_unique_list(temp_line.split(",")))
return temp_line
__all__ = [preprocess]