Fix single text strings
This commit is contained in:
parent
6ca98c10d9
commit
0421c51bae
13 changed files with 1301 additions and 1477 deletions
305
modules/csv.py
305
modules/csv.py
|
|
@ -395,121 +395,6 @@ def getSpeaker(speaker):
|
|||
return [speaker, [0, 0]]
|
||||
|
||||
|
||||
def subVars(jaString):
|
||||
jaString = jaString.replace("\u3000", " ")
|
||||
|
||||
# Nested
|
||||
count = 0
|
||||
nestedList = re.findall(r"[\\]+[\w]+\[[\\]+[\w]+\[[0-9]+\]\]", jaString)
|
||||
nestedList = set(nestedList)
|
||||
if len(nestedList) != 0:
|
||||
for icon in nestedList:
|
||||
jaString = jaString.replace(icon, "[Nested_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Icons
|
||||
count = 0
|
||||
iconList = re.findall(r"[\\]+[iIkKwWaA]+\[[0-9]+\]", jaString)
|
||||
iconList = set(iconList)
|
||||
if len(iconList) != 0:
|
||||
for icon in iconList:
|
||||
jaString = jaString.replace(icon, "[Ascii_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Colors
|
||||
count = 0
|
||||
colorList = re.findall(r"([\\]+c\[\d+\][\\]+c|[\\]+c\[\d+\])", jaString)
|
||||
colorList = set(colorList)
|
||||
if len(colorList) != 0:
|
||||
for color in colorList:
|
||||
jaString = jaString.replace(color, "[Color_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Names
|
||||
count = 0
|
||||
nameList = re.findall(r"[\\]+[nN]\[.+?\]+", jaString)
|
||||
nameList = set(nameList)
|
||||
if len(nameList) != 0:
|
||||
for name in nameList:
|
||||
jaString = jaString.replace(name, "[Noun_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Variables
|
||||
count = 0
|
||||
varList = re.findall(r"[\\]+[vV]\[[0-9]+\]", jaString)
|
||||
varList = set(varList)
|
||||
if len(varList) != 0:
|
||||
for var in varList:
|
||||
jaString = jaString.replace(var, "[Var_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Formatting
|
||||
count = 0
|
||||
formatList = re.findall(r"[\\]+[\w]+\[[a-zA-Z0-9\\\[\]\_,\s-]+\]", jaString)
|
||||
formatList = set(formatList)
|
||||
if len(formatList) != 0:
|
||||
for var in formatList:
|
||||
jaString = jaString.replace(var, "[FCode_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Put all lists in list and return
|
||||
allList = [nestedList, iconList, colorList, nameList, varList, formatList]
|
||||
return [jaString, allList]
|
||||
|
||||
|
||||
def resubVars(translatedText, allList):
|
||||
# Fix Spacing and ChatGPT Nonsense
|
||||
matchList = re.findall(r"\[\s?.+?\s?\]", translatedText)
|
||||
if len(matchList) > 0:
|
||||
for match in matchList:
|
||||
text = match.strip()
|
||||
translatedText = translatedText.replace(match, text)
|
||||
|
||||
# Nested
|
||||
count = 0
|
||||
if len(allList[0]) != 0:
|
||||
for var in allList[0]:
|
||||
translatedText = translatedText.replace("[Nested_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Icons
|
||||
count = 0
|
||||
if len(allList[1]) != 0:
|
||||
for var in allList[1]:
|
||||
translatedText = translatedText.replace("[Ascii_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Colors
|
||||
count = 0
|
||||
if len(allList[2]) != 0:
|
||||
for var in allList[2]:
|
||||
translatedText = translatedText.replace("[Color_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Names
|
||||
count = 0
|
||||
if len(allList[3]) != 0:
|
||||
for var in allList[3]:
|
||||
translatedText = translatedText.replace("[Noun_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Vars
|
||||
count = 0
|
||||
if len(allList[4]) != 0:
|
||||
for var in allList[4]:
|
||||
translatedText = translatedText.replace("[Var_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Formatting
|
||||
count = 0
|
||||
if len(allList[5]) != 0:
|
||||
for var in allList[5]:
|
||||
translatedText = translatedText.replace("[FCode_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
return translatedText
|
||||
|
||||
|
||||
def batchList(input_list, batch_size):
|
||||
if not isinstance(batch_size, int) or batch_size <= 0:
|
||||
raise ValueError("batch_size must be a positive integer")
|
||||
|
|
@ -517,12 +402,7 @@ def batchList(input_list, batch_size):
|
|||
return [input_list[i : i + batch_size] for i in range(0, len(input_list), batch_size)]
|
||||
|
||||
|
||||
def createContext(fullPromptFlag, subbedT):
|
||||
characters = "Game Characters:\n\
|
||||
クリスティーナ (Christina) - Female\n\
|
||||
リズ (Liz) - Female\n\
|
||||
"
|
||||
|
||||
def createContext(fullPromptFlag, subbedT, format):
|
||||
system = (
|
||||
PROMPT + VOCAB
|
||||
if fullPromptFlag
|
||||
|
|
@ -539,19 +419,13 @@ Output ONLY the {LANGUAGE} translation in the following format: `Translation: <{
|
|||
{VOCAB}\n\
|
||||
"
|
||||
)
|
||||
if isinstance(subbedT, list):
|
||||
user = f"```json\n{subbedT}```"
|
||||
else:
|
||||
user = subbedT
|
||||
return characters, system, user
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
return system, user
|
||||
|
||||
|
||||
def translateText(characters, system, user, history, penalty, format):
|
||||
def translateText(system, user, history, penalty, format, model=MODEL):
|
||||
# Prompt
|
||||
msg = [{"role": "system", "content": system + characters}]
|
||||
|
||||
# Characters
|
||||
msg.append({"role": "system", "content": characters})
|
||||
msg = [{"role": "system", "content": system}]
|
||||
|
||||
# History
|
||||
if isinstance(history, list):
|
||||
|
|
@ -560,17 +434,14 @@ def translateText(characters, system, user, history, penalty, format):
|
|||
msg.append({"role": "system", "content": history})
|
||||
|
||||
# Response Format
|
||||
if format == "json":
|
||||
responseFormat = {"type": "json_object"}
|
||||
else:
|
||||
responseFormat = {"type": "text"}
|
||||
responseFormat = {"type": "json_object"}
|
||||
|
||||
# Content to TL
|
||||
msg.append({"role": "user", "content": f"{user}"})
|
||||
response = openai.chat.completions.create(
|
||||
temperature=0,
|
||||
frequency_penalty=penalty,
|
||||
model=MODEL,
|
||||
model=model,
|
||||
response_format=responseFormat,
|
||||
messages=msg,
|
||||
)
|
||||
|
|
@ -588,6 +459,8 @@ def cleanTranslatedText(translatedText, varResponse):
|
|||
"「": '\\"',
|
||||
"」": '\\"',
|
||||
"- ": "-",
|
||||
"】": "]",
|
||||
"【": "[",
|
||||
"Placeholder Text": "",
|
||||
# Add more replacements as needed
|
||||
}
|
||||
|
|
@ -596,7 +469,6 @@ def cleanTranslatedText(translatedText, varResponse):
|
|||
|
||||
# Elongate Long Dashes (Since GPT Ignores them...)
|
||||
translatedText = elongateCharacters(translatedText)
|
||||
translatedText = resubVars(translatedText, varResponse[1])
|
||||
return translatedText
|
||||
|
||||
|
||||
|
|
@ -618,6 +490,7 @@ def elongateCharacters(text):
|
|||
def extractTranslation(translatedTextList, is_list):
|
||||
try:
|
||||
translatedTextList = re.sub(r'\\"+\"([^,\n}])', r'\\"\1', translatedTextList)
|
||||
translatedTextList = re.sub(r"(?<![\\])\"+(?![\n,])", r'"', translatedTextList)
|
||||
line_dict = json.loads(translatedTextList)
|
||||
# If it's a batch (i.e., list), extract with tags; otherwise, return the single item.
|
||||
string_list = list(line_dict.values())
|
||||
|
|
@ -631,7 +504,7 @@ def extractTranslation(translatedTextList, is_list):
|
|||
return None
|
||||
|
||||
|
||||
def countTokens(characters, system, user, history):
|
||||
def countTokens(system, user, history):
|
||||
inputTotalTokens = 0
|
||||
outputTotalTokens = 0
|
||||
enc = tiktoken.encoding_for_model("gpt-4")
|
||||
|
|
@ -643,7 +516,6 @@ def countTokens(characters, system, user, history):
|
|||
else:
|
||||
inputTotalTokens += len(enc.encode(history))
|
||||
inputTotalTokens += len(enc.encode(system))
|
||||
inputTotalTokens += len(enc.encode(characters))
|
||||
inputTotalTokens += len(enc.encode(user))
|
||||
|
||||
# Output
|
||||
|
|
@ -660,83 +532,96 @@ def combineList(tlist, text):
|
|||
|
||||
@retry(exceptions=Exception, tries=5, delay=5)
|
||||
def translateGPT(text, history, fullPromptFlag):
|
||||
global PBAR
|
||||
global PBAR, MISMATCH, FILENAME
|
||||
with open("log/translationHistory.txt", "a+", encoding="utf-8") as logFile:
|
||||
mismatch = False
|
||||
totalTokens = [0, 0]
|
||||
if isinstance(text, list):
|
||||
format = "json"
|
||||
tList = batchList(text, BATCHSIZE)
|
||||
else:
|
||||
format = "text"
|
||||
tList = [text]
|
||||
|
||||
mismatch = False
|
||||
totalTokens = [0, 0]
|
||||
if isinstance(text, list):
|
||||
format = "json"
|
||||
tList = batchList(text, BATCHSIZE)
|
||||
else:
|
||||
format = "text"
|
||||
tList = [text]
|
||||
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if isinstance(tItem, list):
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if not isinstance(tItem, list):
|
||||
tItem = [tItem]
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = subVars(payload)
|
||||
subbedT = varResponse[0]
|
||||
else:
|
||||
varResponse = subVars(tItem)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９]+", subbedT):
|
||||
if PBAR is not None:
|
||||
PBAR.update(len(tItem))
|
||||
continue
|
||||
|
||||
# Create Message
|
||||
characters, system, user = createContext(fullPromptFlag, subbedT)
|
||||
|
||||
# Calculate Estimate
|
||||
if ESTIMATE:
|
||||
estimate = countTokens(characters, system, user, history)
|
||||
totalTokens[0] += estimate[0]
|
||||
totalTokens[1] += estimate[1]
|
||||
continue
|
||||
|
||||
# Translating
|
||||
response = translateText(characters, system, user, history, 0.2, format)
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Check Translation
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(extractedTranslations):
|
||||
# Mismatch. Try Again
|
||||
response = translateText(characters, system, user, history, 0.05, format)
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Formatting
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(extractedTranslations):
|
||||
mismatch = True # Just here for breakpoint
|
||||
|
||||
# Set if no mismatch
|
||||
if mismatch == False:
|
||||
tList[index] = extractedTranslations
|
||||
history = extractedTranslations[-10:] # Update history if we have a list
|
||||
else:
|
||||
history = text[-10:]
|
||||
mismatch = False
|
||||
|
||||
# Update Loading Bar
|
||||
with LOCK:
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９\uFF61-\uFF9F]+", subbedT):
|
||||
if PBAR is not None:
|
||||
PBAR.update(len(tItem))
|
||||
else:
|
||||
# Ensure we're passing a single string to extractTranslation
|
||||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
history = tItem[-MAXHISTORY:]
|
||||
continue
|
||||
|
||||
# Create Message
|
||||
system, user = createContext(fullPromptFlag, subbedT, format)
|
||||
|
||||
# Calculate Estimate
|
||||
if ESTIMATE:
|
||||
estimate = countTokens(system, user, history)
|
||||
totalTokens[0] += estimate[0]
|
||||
totalTokens[1] += estimate[1]
|
||||
continue
|
||||
|
||||
# Translating
|
||||
response = translateText(system, user, history, 0.05, format)
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Check Translation
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(extractedTranslations):
|
||||
# Mismatch. Try Again
|
||||
response = translateText(system, user, history, 0.05, format, "gpt-4o")
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Formatting
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(
|
||||
extractedTranslations
|
||||
):
|
||||
mismatch = True # Just here for breakpoint
|
||||
logFile.write(f"Input:\n{subbedT}\n")
|
||||
logFile.write(f"Output:\n{translatedText}\n")
|
||||
|
||||
# Set if no mismatch
|
||||
if mismatch == False:
|
||||
tList[index] = extractedTranslations
|
||||
history = extractedTranslations[
|
||||
-MAXHISTORY:
|
||||
] # Update history if we have a list
|
||||
else:
|
||||
history = text[-MAXHISTORY:]
|
||||
mismatch = False
|
||||
if FILENAME not in MISMATCH:
|
||||
MISMATCH.append(FILENAME)
|
||||
|
||||
# Update Loading Bar
|
||||
with LOCK:
|
||||
if PBAR is not None:
|
||||
PBAR.update(len(tItem))
|
||||
else:
|
||||
# Ensure we're passing a single string to extractTranslation
|
||||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
|
||||
finalList = combineList(tList, text)
|
||||
return [finalList, totalTokens]
|
||||
if format == "json":
|
||||
return [finalList, totalTokens]
|
||||
else:
|
||||
return [finalList[0], totalTokens]
|
||||
|
|
|
|||
405
modules/json.py
405
modules/json.py
|
|
@ -300,135 +300,40 @@ def translateJSON(data, pbar):
|
|||
# Save some money and enter the character before translation
|
||||
def getSpeaker(speaker):
|
||||
match speaker:
|
||||
case "セレナ":
|
||||
return ["Serena", [0, 0]]
|
||||
case "レナ":
|
||||
return ["Rena", [0, 0]]
|
||||
case "フィルス":
|
||||
return ["Phils", [0, 0]]
|
||||
case "レイン":
|
||||
return ["Meryl", [0, 0]]
|
||||
case "ファイン":
|
||||
return ["Fine", [0, 0]]
|
||||
case "":
|
||||
return ["", [0, 0]]
|
||||
case _:
|
||||
return translateGPT(
|
||||
speaker,
|
||||
"Reply with only the " + LANGUAGE + " translation of the NPC name.",
|
||||
False,
|
||||
# Find Speaker
|
||||
for i in range(len(NAMESLIST)):
|
||||
if speaker == NAMESLIST[i][0]:
|
||||
return [NAMESLIST[i][1], [0, 0]]
|
||||
|
||||
# Translate and Store Speaker
|
||||
response = translateGPT(
|
||||
f"{speaker}",
|
||||
"Reply with the " + LANGUAGE + " translation of the NPC name.",
|
||||
True,
|
||||
)
|
||||
response[0] = response[0].title()
|
||||
response[0] = response[0].replace("'S", "'s")
|
||||
response[0] = response[0].replace("Speaker: ", "")
|
||||
|
||||
# Retry if name doesn't translate for some reason
|
||||
if re.search(r"([a-zA-Z？?])", response[0]) == None:
|
||||
response = translateGPT(
|
||||
f"{speaker}",
|
||||
"Reply with the " + LANGUAGE + " translation of the NPC name.",
|
||||
False,
|
||||
)
|
||||
response[0] = response[0].title()
|
||||
response[0] = response[0].replace("'S", "'s")
|
||||
|
||||
def subVars(jaString):
|
||||
jaString = jaString.replace("\u3000", " ")
|
||||
|
||||
# Nested
|
||||
count = 0
|
||||
nestedList = re.findall(r"[\\]+[\w]+\[[\\]+[\w]+\[[0-9]+\]\]", jaString)
|
||||
nestedList = set(nestedList)
|
||||
if len(nestedList) != 0:
|
||||
for icon in nestedList:
|
||||
jaString = jaString.replace(icon, "{Nested_" + str(count) + "}")
|
||||
count += 1
|
||||
|
||||
# Icons
|
||||
count = 0
|
||||
iconList = re.findall(r"[\\]+[iIkKwWaA]+\[[0-9]+\]", jaString)
|
||||
iconList = set(iconList)
|
||||
if len(iconList) != 0:
|
||||
for icon in iconList:
|
||||
jaString = jaString.replace(icon, "{Ascii_" + str(count) + "}")
|
||||
count += 1
|
||||
|
||||
# Colors
|
||||
count = 0
|
||||
colorList = re.findall(r"[\\]+[cC]\[[0-9]+\]", jaString)
|
||||
colorList = set(colorList)
|
||||
if len(colorList) != 0:
|
||||
for color in colorList:
|
||||
jaString = jaString.replace(color, "{Color_" + str(count) + "}")
|
||||
count += 1
|
||||
|
||||
# Names
|
||||
count = 0
|
||||
nameList = re.findall(r"[\\]+[nN]\[.+?\]+", jaString)
|
||||
nameList = set(nameList)
|
||||
if len(nameList) != 0:
|
||||
for name in nameList:
|
||||
jaString = jaString.replace(name, "{Noun_" + str(count) + "}")
|
||||
count += 1
|
||||
|
||||
# Variables
|
||||
count = 0
|
||||
varList = re.findall(r"[\\]+[vV]\[[0-9]+\]", jaString)
|
||||
varList = set(varList)
|
||||
if len(varList) != 0:
|
||||
for var in varList:
|
||||
jaString = jaString.replace(var, "{Var_" + str(count) + "}")
|
||||
count += 1
|
||||
|
||||
# Formatting
|
||||
count = 0
|
||||
formatList = re.findall(r"[\\]+[\w]+\[.+?\]", jaString)
|
||||
formatList = set(formatList)
|
||||
if len(formatList) != 0:
|
||||
for var in formatList:
|
||||
jaString = jaString.replace(var, "{FCode_" + str(count) + "}")
|
||||
count += 1
|
||||
|
||||
# Put all lists in list and return
|
||||
allList = [nestedList, iconList, colorList, nameList, varList, formatList]
|
||||
return [jaString, allList]
|
||||
|
||||
|
||||
def resubVars(translatedText, allList):
|
||||
# Fix Spacing and ChatGPT Nonsense
|
||||
matchList = re.findall(r"\[\s?.+?\s?\]", translatedText)
|
||||
if len(matchList) > 0:
|
||||
for match in matchList:
|
||||
text = match.strip()
|
||||
translatedText = translatedText.replace(match, text)
|
||||
|
||||
# Nested
|
||||
count = 0
|
||||
if len(allList[0]) != 0:
|
||||
for var in allList[0]:
|
||||
translatedText = translatedText.replace("{Nested_" + str(count) + "}", var)
|
||||
count += 1
|
||||
|
||||
# Icons
|
||||
count = 0
|
||||
if len(allList[1]) != 0:
|
||||
for var in allList[1]:
|
||||
translatedText = translatedText.replace("{Ascii_" + str(count) + "}", var)
|
||||
count += 1
|
||||
|
||||
# Colors
|
||||
count = 0
|
||||
if len(allList[2]) != 0:
|
||||
for var in allList[2]:
|
||||
translatedText = translatedText.replace("{Color_" + str(count) + "}", var)
|
||||
count += 1
|
||||
|
||||
# Names
|
||||
count = 0
|
||||
if len(allList[3]) != 0:
|
||||
for var in allList[3]:
|
||||
translatedText = translatedText.replace("{Noun_" + str(count) + "}", var)
|
||||
count += 1
|
||||
|
||||
# Vars
|
||||
count = 0
|
||||
if len(allList[4]) != 0:
|
||||
for var in allList[4]:
|
||||
translatedText = translatedText.replace("{Var_" + str(count) + "}", var)
|
||||
count += 1
|
||||
|
||||
# Formatting
|
||||
count = 0
|
||||
if len(allList[5]) != 0:
|
||||
for var in allList[5]:
|
||||
translatedText = translatedText.replace("{FCode_" + str(count) + "}", var)
|
||||
count += 1
|
||||
|
||||
return translatedText
|
||||
speakerList = [speaker, response[0]]
|
||||
NAMESLIST.append(speakerList)
|
||||
return response
|
||||
return [speaker, [0, 0]]
|
||||
|
||||
|
||||
def batchList(input_list, batch_size):
|
||||
|
|
@ -438,69 +343,47 @@ def batchList(input_list, batch_size):
|
|||
return [input_list[i : i + batch_size] for i in range(0, len(input_list), batch_size)]
|
||||
|
||||
|
||||
def createContext(fullPromptFlag, subbedT):
|
||||
characters = "Game Characters:\n\
|
||||
ルナリア (Lunaria) - Female\n\
|
||||
ソニア (Sonia) - Female\n\
|
||||
マナ (Mana) - Female\n\
|
||||
マリアナ (Mariana) - Female\n\
|
||||
ディアナ (Diana) - Female\n\
|
||||
シャーリー (Shirley) - Female\n\
|
||||
エスティア (Estia) - Female\n\
|
||||
エレノア (Eleanor) - Female\n\
|
||||
メリス (Meris) - Female\n\
|
||||
サルビア (Salvia) - Female\n\
|
||||
リリ (Lili) - Female\n\
|
||||
ツキハ (Tsukiha) - Female\n\
|
||||
フィリカ (Filica) - Female\n\
|
||||
レノ (Renno) - Female\n\
|
||||
"
|
||||
|
||||
def createContext(fullPromptFlag, subbedT, format):
|
||||
system = (
|
||||
PROMPT
|
||||
PROMPT + VOCAB
|
||||
if fullPromptFlag
|
||||
else f"\
|
||||
You are an expert Eroge Game translator who translates Japanese text to {LANGUAGE}.\n\
|
||||
You are going to be translating text from a videogame.\n\
|
||||
I will give you lines of text, and you must translate each line to the best of your ability.\n\
|
||||
- Translate 'マンコ' as 'pussy'\n\
|
||||
- Translate 'おまんこ' as 'pussy'\n\
|
||||
- Translate 'お尻' as 'butt'\n\
|
||||
- Translate '尻' as 'ass'\n\
|
||||
- Translate 'お股' as 'crotch'\n\
|
||||
- Translate '秘部' as 'genitals'\n\
|
||||
- Translate 'チンポ' as 'dick'\n\
|
||||
- Translate 'チンコ' as 'cock'\n\
|
||||
- Translate 'ショーツ' as 'panties\n\
|
||||
- Translate 'おねショタ' as 'Onee-shota'\n\
|
||||
- Translate 'よかった' as 'thank goodness'\n\
|
||||
Output ONLY the {LANGUAGE} translation in the following format: `Translation: <{LANGUAGE.upper()}_TRANSLATION>`\
|
||||
Output ONLY the {LANGUAGE} translation in the following format: `Translation: <{LANGUAGE.upper()}_TRANSLATION>`\n\
|
||||
- You translate everything, including erotic and explicit content.\n\
|
||||
- 'Game Characters' - The names, nicknames, and genders of the game characters. Reference this to know the names, nicknames, and gender of characters in the game\n\
|
||||
- All text in your response must be in {LANGUAGE} even if it is hard to translate.\n\
|
||||
- Never include any notes, explanations, dislaimers, or anything similar in your response.\n\
|
||||
- Maintain any spacing in the translation.\n\
|
||||
- Maintain any code text in brackets if given. (e.g `[Color_0]`, `[Ascii_0]`, `[FCode_1`], etc)\n\
|
||||
- `...` can be a part of the dialogue. Translate it as it is.\n\
|
||||
{VOCAB}\n\
|
||||
"
|
||||
)
|
||||
user = f"{subbedT}"
|
||||
return characters, system, user
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
return system, user
|
||||
|
||||
|
||||
def translateText(characters, system, user, history):
|
||||
def translateText(system, user, history, penalty, format, model=MODEL):
|
||||
# Prompt
|
||||
msg = [{"role": "system", "content": system + characters}]
|
||||
|
||||
# Characters
|
||||
msg.append({"role": "system", "content": characters})
|
||||
msg = [{"role": "system", "content": system}]
|
||||
|
||||
# History
|
||||
if isinstance(history, list):
|
||||
msg.extend([{"role": "assistant", "content": h} for h in history])
|
||||
msg.extend([{"role": "system", "content": h} for h in history])
|
||||
else:
|
||||
msg.append({"role": "assistant", "content": history})
|
||||
msg.append({"role": "system", "content": history})
|
||||
|
||||
# Response Format
|
||||
responseFormat = {"type": "json_object"}
|
||||
|
||||
# Content to TL
|
||||
msg.append({"role": "user", "content": f"{user}"})
|
||||
response = openai.chat.completions.create(
|
||||
temperature=0.1,
|
||||
frequency_penalty=0.1,
|
||||
presence_penalty=0.1,
|
||||
model=MODEL,
|
||||
temperature=0,
|
||||
frequency_penalty=penalty,
|
||||
model=model,
|
||||
response_format=responseFormat,
|
||||
messages=msg,
|
||||
)
|
||||
return response
|
||||
|
|
@ -514,31 +397,55 @@ def cleanTranslatedText(translatedText, varResponse):
|
|||
"〜": "~",
|
||||
"ッ": "",
|
||||
"。": ".",
|
||||
"「": '\\"',
|
||||
"」": '\\"',
|
||||
"- ": "-",
|
||||
"】": "]",
|
||||
"【": "[",
|
||||
"Placeholder Text": "",
|
||||
# Add more replacements as needed
|
||||
}
|
||||
for target, replacement in placeholders.items():
|
||||
translatedText = translatedText.replace(target, replacement)
|
||||
|
||||
translatedText = resubVars(translatedText, varResponse[1])
|
||||
return [line for line in translatedText.split("\n") if line]
|
||||
# Elongate Long Dashes (Since GPT Ignores them...)
|
||||
translatedText = elongateCharacters(translatedText)
|
||||
return translatedText
|
||||
|
||||
|
||||
def elongateCharacters(text):
|
||||
# Define a pattern to match one character followed by one or more `ー` characters
|
||||
# Using a positive lookbehind assertion to capture the preceding character
|
||||
pattern = r"(?<=(.))ー+"
|
||||
|
||||
# Define a replacement function that elongates the captured character
|
||||
def repl(match):
|
||||
char = match.group(1) # The character before the ー sequence
|
||||
count = len(match.group(0)) - 1 # Number of ー characters
|
||||
return char * count # Replace ー sequence with the character repeated
|
||||
|
||||
# Use re.sub() to replace the pattern in the text
|
||||
return re.sub(pattern, repl, text)
|
||||
|
||||
|
||||
def extractTranslation(translatedTextList, is_list):
|
||||
pattern = r"`?<Line(\d+)>([\\]*.*?[\\]*?)<\/?Line\d+>`?"
|
||||
# If it's a batch (i.e., list), extract with tags; otherwise, return the single item.
|
||||
if is_list:
|
||||
return [
|
||||
re.findall(pattern, line)[0][1]
|
||||
for line in translatedTextList
|
||||
if re.search(pattern, line)
|
||||
]
|
||||
else:
|
||||
matchList = re.findall(pattern, translatedTextList)
|
||||
return matchList[0][1] if matchList else translatedTextList
|
||||
try:
|
||||
translatedTextList = re.sub(r'\\"+\"([^,\n}])', r'\\"\1', translatedTextList)
|
||||
translatedTextList = re.sub(r"(?<![\\])\"+(?![\n,])", r'"', translatedTextList)
|
||||
line_dict = json.loads(translatedTextList)
|
||||
# If it's a batch (i.e., list), extract with tags; otherwise, return the single item.
|
||||
string_list = list(line_dict.values())
|
||||
if is_list:
|
||||
return string_list
|
||||
else:
|
||||
return string_list[0]
|
||||
|
||||
except Exception as e:
|
||||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
|
||||
def countTokens(characters, system, user, history):
|
||||
def countTokens(system, user, history):
|
||||
inputTotalTokens = 0
|
||||
outputTotalTokens = 0
|
||||
enc = tiktoken.encoding_for_model("gpt-4")
|
||||
|
|
@ -550,7 +457,6 @@ def countTokens(characters, system, user, history):
|
|||
else:
|
||||
inputTotalTokens += len(enc.encode(history))
|
||||
inputTotalTokens += len(enc.encode(system))
|
||||
inputTotalTokens += len(enc.encode(characters))
|
||||
inputTotalTokens += len(enc.encode(user))
|
||||
|
||||
# Output
|
||||
|
|
@ -567,55 +473,96 @@ def combineList(tlist, text):
|
|||
|
||||
@retry(exceptions=Exception, tries=5, delay=5)
|
||||
def translateGPT(text, history, fullPromptFlag):
|
||||
totalTokens = [0, 0]
|
||||
if isinstance(text, list):
|
||||
tList = batchList(text, BATCHSIZE)
|
||||
else:
|
||||
tList = [text]
|
||||
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if isinstance(tItem, list):
|
||||
payload = "\n".join([f"`<Line{i}>{item}</Line{i}>`" for i, item in enumerate(tItem)])
|
||||
payload = payload.replace("``", "`Placeholder Text`")
|
||||
varResponse = subVars(payload)
|
||||
subbedT = varResponse[0]
|
||||
global PBAR, MISMATCH, FILENAME
|
||||
with open("log/translationHistory.txt", "a+", encoding="utf-8") as logFile:
|
||||
mismatch = False
|
||||
totalTokens = [0, 0]
|
||||
if isinstance(text, list):
|
||||
format = "json"
|
||||
tList = batchList(text, BATCHSIZE)
|
||||
else:
|
||||
varResponse = subVars(tItem)
|
||||
format = "text"
|
||||
tList = [text]
|
||||
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if not isinstance(tItem, list):
|
||||
tItem = [tItem]
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９]+", subbedT):
|
||||
continue
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９\uFF61-\uFF9F]+", subbedT):
|
||||
if PBAR is not None:
|
||||
PBAR.update(len(tItem))
|
||||
history = tItem[-MAXHISTORY:]
|
||||
continue
|
||||
|
||||
# Create Message
|
||||
characters, system, user = createContext(fullPromptFlag, subbedT)
|
||||
# Create Message
|
||||
system, user = createContext(fullPromptFlag, subbedT, format)
|
||||
|
||||
# Calculate Estimate
|
||||
if ESTIMATE:
|
||||
estimate = countTokens(characters, system, user, history)
|
||||
totalTokens[0] += estimate[0]
|
||||
totalTokens[1] += estimate[1]
|
||||
continue
|
||||
# Calculate Estimate
|
||||
if ESTIMATE:
|
||||
estimate = countTokens(system, user, history)
|
||||
totalTokens[0] += estimate[0]
|
||||
totalTokens[1] += estimate[1]
|
||||
continue
|
||||
|
||||
# Translating
|
||||
response = translateText(characters, system, user, history)
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
# Translating
|
||||
response = translateText(system, user, history, 0.05, format)
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Formatting
|
||||
translatedTextList = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedTextList, True)
|
||||
tList[index] = extractedTranslations
|
||||
if len(tItem) != len(translatedTextList):
|
||||
mismatch = True # Just here so breakpoint can be set
|
||||
history = extractedTranslations[-10:] # Update history if we have a list
|
||||
else:
|
||||
# Ensure we're passing a single string to extractTranslation
|
||||
extractedTranslations = extractTranslation("\n".join(translatedTextList), False)
|
||||
tList[index] = extractedTranslations
|
||||
# Check Translation
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(extractedTranslations):
|
||||
# Mismatch. Try Again
|
||||
response = translateText(system, user, history, 0.05, format, "gpt-4o")
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Formatting
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(
|
||||
extractedTranslations
|
||||
):
|
||||
mismatch = True # Just here for breakpoint
|
||||
logFile.write(f"Input:\n{subbedT}\n")
|
||||
logFile.write(f"Output:\n{translatedText}\n")
|
||||
|
||||
# Set if no mismatch
|
||||
if mismatch == False:
|
||||
tList[index] = extractedTranslations
|
||||
history = extractedTranslations[
|
||||
-MAXHISTORY:
|
||||
] # Update history if we have a list
|
||||
else:
|
||||
history = text[-MAXHISTORY:]
|
||||
mismatch = False
|
||||
if FILENAME not in MISMATCH:
|
||||
MISMATCH.append(FILENAME)
|
||||
|
||||
# Update Loading Bar
|
||||
with LOCK:
|
||||
if PBAR is not None:
|
||||
PBAR.update(len(tItem))
|
||||
else:
|
||||
# Ensure we're passing a single string to extractTranslation
|
||||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
|
||||
finalList = combineList(tList, text)
|
||||
return [finalList, totalTokens]
|
||||
if format == "json":
|
||||
return [finalList, totalTokens]
|
||||
else:
|
||||
return [finalList[0], totalTokens]
|
||||
|
|
|
|||
|
|
@ -379,34 +379,6 @@ def getSpeaker(speaker):
|
|||
return [speaker, [0, 0]]
|
||||
|
||||
|
||||
def subVars(jaString):
|
||||
jaString = jaString.replace("\u3000", " ")
|
||||
|
||||
# Formatting
|
||||
codeList = re.findall(r"([\\]*(\w+)\[(\d+)\])|([\\]*(\w+)\[[\\]*\\w+\[(\d+)\]\])", jaString)
|
||||
codeList = set(codeList)
|
||||
if len(codeList) != 0:
|
||||
for var in codeList:
|
||||
if var[2]:
|
||||
jaString = jaString.replace(var[0], f"[{var[1]}Code_" + f"{var[2]}]")
|
||||
else:
|
||||
jaString = jaString.replace(var[3], f"[{var[4]}Code_" + f"{var[5]}]")
|
||||
|
||||
# Put all lists in list and return
|
||||
return [jaString, codeList]
|
||||
|
||||
|
||||
def resubVars(translatedText, codeList):
|
||||
# Formatting
|
||||
for var in codeList:
|
||||
if var[2]:
|
||||
translatedText = translatedText.replace(f"[{var[1]}Code_" + f"{var[2]}]", var[0])
|
||||
else:
|
||||
translatedText = translatedText.replace(f"[{var[4]}Code_" + f"{var[5]}]", var[3])
|
||||
|
||||
return translatedText
|
||||
|
||||
|
||||
def batchList(input_list, batch_size):
|
||||
if not isinstance(batch_size, int) or batch_size <= 0:
|
||||
raise ValueError("batch_size must be a positive integer")
|
||||
|
|
@ -431,10 +403,7 @@ Output ONLY the {LANGUAGE} translation in the following format: `Translation: <{
|
|||
{VOCAB}\n\
|
||||
"
|
||||
)
|
||||
if format == "json":
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
else:
|
||||
user = subbedT
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
return system, user
|
||||
|
||||
|
||||
|
|
@ -449,10 +418,7 @@ def translateText(system, user, history, penalty, format, model=MODEL):
|
|||
msg.append({"role": "system", "content": history})
|
||||
|
||||
# Response Format
|
||||
if format == "json":
|
||||
responseFormat = {"type": "json_object"}
|
||||
else:
|
||||
responseFormat = {"type": "text"}
|
||||
responseFormat = {"type": "json_object"}
|
||||
|
||||
# Content to TL
|
||||
msg.append({"role": "user", "content": f"{user}"})
|
||||
|
|
@ -487,7 +453,6 @@ def cleanTranslatedText(translatedText, varResponse):
|
|||
|
||||
# Elongate Long Dashes (Since GPT Ignores them...)
|
||||
translatedText = elongateCharacters(translatedText)
|
||||
translatedText = resubVars(translatedText, varResponse[1])
|
||||
return translatedText
|
||||
|
||||
|
||||
|
|
@ -522,10 +487,6 @@ def extractTranslation(translatedTextList, is_list):
|
|||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
except Exception as e:
|
||||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
|
||||
def countTokens(system, user, history):
|
||||
inputTotalTokens = 0
|
||||
|
|
@ -568,14 +529,15 @@ def translateGPT(text, history, fullPromptFlag):
|
|||
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if isinstance(tItem, list):
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = subVars(payload)
|
||||
subbedT = varResponse[0]
|
||||
else:
|
||||
varResponse = [tItem, []]
|
||||
subbedT = varResponse[0]
|
||||
if not isinstance(tItem, list):
|
||||
tItem = [tItem]
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９\uFF61-\uFF9F]+", subbedT):
|
||||
|
|
@ -643,4 +605,7 @@ def translateGPT(text, history, fullPromptFlag):
|
|||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
|
||||
finalList = combineList(tList, text)
|
||||
return [finalList, totalTokens]
|
||||
if format == "json":
|
||||
return [finalList, totalTokens]
|
||||
else:
|
||||
return [finalList[0], totalTokens]
|
||||
|
|
|
|||
|
|
@ -395,121 +395,6 @@ def getSpeaker(speaker):
|
|||
return [speaker, [0, 0]]
|
||||
|
||||
|
||||
def subVars(jaString):
|
||||
jaString = jaString.replace("\u3000", " ")
|
||||
|
||||
# Nested
|
||||
count = 0
|
||||
nestedList = re.findall(r"[\\]+[\w]+\[[\\]+[\w]+\[[0-9]+\]\]", jaString)
|
||||
nestedList = set(nestedList)
|
||||
if len(nestedList) != 0:
|
||||
for icon in nestedList:
|
||||
jaString = jaString.replace(icon, "[Nested_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Icons
|
||||
count = 0
|
||||
iconList = re.findall(r"[\\]+[iIkKwWaA]+\[[0-9]+\]", jaString)
|
||||
iconList = set(iconList)
|
||||
if len(iconList) != 0:
|
||||
for icon in iconList:
|
||||
jaString = jaString.replace(icon, "[Ascii_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Colors
|
||||
count = 0
|
||||
colorList = re.findall(r"([\\]+c\[\d+\][\\]+c|[\\]+c\[\d+\])", jaString)
|
||||
colorList = set(colorList)
|
||||
if len(colorList) != 0:
|
||||
for color in colorList:
|
||||
jaString = jaString.replace(color, "[Color_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Names
|
||||
count = 0
|
||||
nameList = re.findall(r"[\\]+[nN]\[.+?\]+", jaString)
|
||||
nameList = set(nameList)
|
||||
if len(nameList) != 0:
|
||||
for name in nameList:
|
||||
jaString = jaString.replace(name, "[Noun_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Variables
|
||||
count = 0
|
||||
varList = re.findall(r"[\\]+[vV]\[[0-9]+\]", jaString)
|
||||
varList = set(varList)
|
||||
if len(varList) != 0:
|
||||
for var in varList:
|
||||
jaString = jaString.replace(var, "[Var_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Formatting
|
||||
count = 0
|
||||
formatList = re.findall(r"[\\]+[\w]+\[[a-zA-Z0-9\\\[\]\_,\s-]+\]", jaString)
|
||||
formatList = set(formatList)
|
||||
if len(formatList) != 0:
|
||||
for var in formatList:
|
||||
jaString = jaString.replace(var, "[FCode_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Put all lists in list and return
|
||||
allList = [nestedList, iconList, colorList, nameList, varList, formatList]
|
||||
return [jaString, allList]
|
||||
|
||||
|
||||
def resubVars(translatedText, allList):
|
||||
# Fix Spacing and ChatGPT Nonsense
|
||||
matchList = re.findall(r"\[\s?.+?\s?\]", translatedText)
|
||||
if len(matchList) > 0:
|
||||
for match in matchList:
|
||||
text = match.strip()
|
||||
translatedText = translatedText.replace(match, text)
|
||||
|
||||
# Nested
|
||||
count = 0
|
||||
if len(allList[0]) != 0:
|
||||
for var in allList[0]:
|
||||
translatedText = translatedText.replace("[Nested_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Icons
|
||||
count = 0
|
||||
if len(allList[1]) != 0:
|
||||
for var in allList[1]:
|
||||
translatedText = translatedText.replace("[Ascii_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Colors
|
||||
count = 0
|
||||
if len(allList[2]) != 0:
|
||||
for var in allList[2]:
|
||||
translatedText = translatedText.replace("[Color_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Names
|
||||
count = 0
|
||||
if len(allList[3]) != 0:
|
||||
for var in allList[3]:
|
||||
translatedText = translatedText.replace("[Noun_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Vars
|
||||
count = 0
|
||||
if len(allList[4]) != 0:
|
||||
for var in allList[4]:
|
||||
translatedText = translatedText.replace("[Var_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Formatting
|
||||
count = 0
|
||||
if len(allList[5]) != 0:
|
||||
for var in allList[5]:
|
||||
translatedText = translatedText.replace("[FCode_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
return translatedText
|
||||
|
||||
|
||||
def batchList(input_list, batch_size):
|
||||
if not isinstance(batch_size, int) or batch_size <= 0:
|
||||
raise ValueError("batch_size must be a positive integer")
|
||||
|
|
@ -517,37 +402,7 @@ def batchList(input_list, batch_size):
|
|||
return [input_list[i : i + batch_size] for i in range(0, len(input_list), batch_size)]
|
||||
|
||||
|
||||
def createContext(fullPromptFlag, subbedT):
|
||||
characters = "Game Characters:\n\
|
||||
レナリス (Renalith) - Female\n\
|
||||
スクルー (Sukuru) - Female\n\
|
||||
シスターミサ (Sister Misa) - Female\n\
|
||||
オリン (Orin) - Female\n\
|
||||
プローテ (Prote) - Female\n\
|
||||
夜霧 (Night Fog) - Female\n\
|
||||
ワウ (Wao) - Female\n\
|
||||
ファンナ (Fanna) - Female\n\
|
||||
精霊主スクルド (Spirit God Skuld) - Female\n\
|
||||
エキドナ (Echnida) - Female\n\
|
||||
マルス (Mars) - Male\n\
|
||||
ラヴィー (Lavi) - Unknown\n\
|
||||
魅音 (Mion) - Female\n\
|
||||
ヴィオラ (Viola) - Female\n\
|
||||
リンメイ (Lin Mei) - Female\n\
|
||||
リネット (Lynette) - Female\n\
|
||||
チェロル (Cheryl) - Female\n\
|
||||
カルーア姫 (Princess Karua) - Female\n\
|
||||
田姫 (Tajirme) - Female\n\
|
||||
リュート (Luto) - Male\n\
|
||||
ホルン (Horn) - Female\n\
|
||||
ルメラ (Lumera) - Female\n\
|
||||
末嬉 (Sueki) - Female\n\
|
||||
モニカ姫 (Princess Monica) - Female\n\
|
||||
エメルーラ (Emerald) - Female\n\
|
||||
フンシス (Funsis) - Male \n\
|
||||
バゼット (Bazzet) - Female\n\
|
||||
"
|
||||
|
||||
def createContext(fullPromptFlag, subbedT, format):
|
||||
system = (
|
||||
PROMPT + VOCAB
|
||||
if fullPromptFlag
|
||||
|
|
@ -564,16 +419,13 @@ Output ONLY the {LANGUAGE} translation in the following format: `Translation: <{
|
|||
{VOCAB}\n\
|
||||
"
|
||||
)
|
||||
user = f"```json\n{subbedT}```"
|
||||
return characters, system, user
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
return system, user
|
||||
|
||||
|
||||
def translateText(characters, system, user, history, penalty):
|
||||
def translateText(system, user, history, penalty, format, model=MODEL):
|
||||
# Prompt
|
||||
msg = [{"role": "system", "content": system + characters}]
|
||||
|
||||
# Characters
|
||||
msg.append({"role": "system", "content": characters})
|
||||
msg = [{"role": "system", "content": system}]
|
||||
|
||||
# History
|
||||
if isinstance(history, list):
|
||||
|
|
@ -581,13 +433,16 @@ def translateText(characters, system, user, history, penalty):
|
|||
else:
|
||||
msg.append({"role": "system", "content": history})
|
||||
|
||||
# Response Format
|
||||
responseFormat = {"type": "json_object"}
|
||||
|
||||
# Content to TL
|
||||
msg.append({"role": "user", "content": f"{user}"})
|
||||
response = openai.chat.completions.create(
|
||||
temperature=0,
|
||||
frequency_penalty=penalty,
|
||||
model=MODEL,
|
||||
response_format={"type": "json_object"},
|
||||
model=model,
|
||||
response_format=responseFormat,
|
||||
messages=msg,
|
||||
)
|
||||
return response
|
||||
|
|
@ -604,6 +459,8 @@ def cleanTranslatedText(translatedText, varResponse):
|
|||
"「": '\\"',
|
||||
"」": '\\"',
|
||||
"- ": "-",
|
||||
"】": "]",
|
||||
"【": "[",
|
||||
"Placeholder Text": "",
|
||||
# Add more replacements as needed
|
||||
}
|
||||
|
|
@ -612,7 +469,6 @@ def cleanTranslatedText(translatedText, varResponse):
|
|||
|
||||
# Elongate Long Dashes (Since GPT Ignores them...)
|
||||
translatedText = elongateCharacters(translatedText)
|
||||
translatedText = resubVars(translatedText, varResponse[1])
|
||||
return translatedText
|
||||
|
||||
|
||||
|
|
@ -633,17 +489,22 @@ def elongateCharacters(text):
|
|||
|
||||
def extractTranslation(translatedTextList, is_list):
|
||||
try:
|
||||
translatedTextList = re.sub(r'\\"+\"([^,\n}])', r'\\"\1', translatedTextList)
|
||||
translatedTextList = re.sub(r"(?<![\\])\"+(?![\n,])", r'"', translatedTextList)
|
||||
line_dict = json.loads(translatedTextList)
|
||||
# If it's a batch (i.e., list), extract with tags; otherwise, return the single item.
|
||||
string_list = list(line_dict.values())
|
||||
if is_list:
|
||||
string_list = list(line_dict.values())
|
||||
return string_list
|
||||
else:
|
||||
return string_list[0]
|
||||
|
||||
except Exception as e:
|
||||
print(e)
|
||||
return translatedTextList
|
||||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
|
||||
def countTokens(characters, system, user, history):
|
||||
def countTokens(system, user, history):
|
||||
inputTotalTokens = 0
|
||||
outputTotalTokens = 0
|
||||
enc = tiktoken.encoding_for_model("gpt-4")
|
||||
|
|
@ -655,7 +516,6 @@ def countTokens(characters, system, user, history):
|
|||
else:
|
||||
inputTotalTokens += len(enc.encode(history))
|
||||
inputTotalTokens += len(enc.encode(system))
|
||||
inputTotalTokens += len(enc.encode(characters))
|
||||
inputTotalTokens += len(enc.encode(user))
|
||||
|
||||
# Output
|
||||
|
|
@ -672,82 +532,96 @@ def combineList(tlist, text):
|
|||
|
||||
@retry(exceptions=Exception, tries=5, delay=5)
|
||||
def translateGPT(text, history, fullPromptFlag):
|
||||
global PBAR
|
||||
global PBAR, MISMATCH, FILENAME
|
||||
with open("log/translationHistory.txt", "a+", encoding="utf-8") as logFile:
|
||||
mismatch = False
|
||||
totalTokens = [0, 0]
|
||||
if isinstance(text, list):
|
||||
format = "json"
|
||||
tList = batchList(text, BATCHSIZE)
|
||||
else:
|
||||
format = "text"
|
||||
tList = [text]
|
||||
|
||||
mismatch = False
|
||||
totalTokens = [0, 0]
|
||||
if isinstance(text, list):
|
||||
tList = batchList(text, BATCHSIZE)
|
||||
else:
|
||||
tList = [text]
|
||||
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if isinstance(tItem, list):
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if not isinstance(tItem, list):
|
||||
tItem = [tItem]
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = subVars(payload)
|
||||
subbedT = varResponse[0]
|
||||
else:
|
||||
varResponse = subVars(tItem)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９]+", subbedT):
|
||||
if PBAR is not None:
|
||||
PBAR.update(len(tItem))
|
||||
continue
|
||||
|
||||
# Create Message
|
||||
characters, system, user = createContext(fullPromptFlag, subbedT)
|
||||
|
||||
# Calculate Estimate
|
||||
if ESTIMATE:
|
||||
estimate = countTokens(characters, system, user, history)
|
||||
totalTokens[0] += estimate[0]
|
||||
totalTokens[1] += estimate[1]
|
||||
continue
|
||||
|
||||
# Translating
|
||||
response = translateText(characters, system, user, history, 0.02)
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Check Translation
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if len(tItem) != len(extractedTranslations):
|
||||
# Mismatch. Try Again
|
||||
response = translateText(characters, system, user, history, 0.2)
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Formatting
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if len(tItem) != len(extractedTranslations):
|
||||
mismatch = True # Just here for breakpoint
|
||||
|
||||
# Set if no mismatch
|
||||
if mismatch == False:
|
||||
tList[index] = extractedTranslations
|
||||
history = extractedTranslations[-10:] # Update history if we have a list
|
||||
else:
|
||||
history = text[-10:]
|
||||
mismatch = False
|
||||
|
||||
# Update Loading Bar
|
||||
with LOCK:
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９\uFF61-\uFF9F]+", subbedT):
|
||||
if PBAR is not None:
|
||||
PBAR.update(len(tItem))
|
||||
else:
|
||||
# Ensure we're passing a single string to extractTranslation
|
||||
extractedTranslations = extractTranslation(translatedText, False)
|
||||
tList[index] = extractedTranslations
|
||||
history = tItem[-MAXHISTORY:]
|
||||
continue
|
||||
|
||||
# Create Message
|
||||
system, user = createContext(fullPromptFlag, subbedT, format)
|
||||
|
||||
# Calculate Estimate
|
||||
if ESTIMATE:
|
||||
estimate = countTokens(system, user, history)
|
||||
totalTokens[0] += estimate[0]
|
||||
totalTokens[1] += estimate[1]
|
||||
continue
|
||||
|
||||
# Translating
|
||||
response = translateText(system, user, history, 0.05, format)
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Check Translation
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(extractedTranslations):
|
||||
# Mismatch. Try Again
|
||||
response = translateText(system, user, history, 0.05, format, "gpt-4o")
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Formatting
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(
|
||||
extractedTranslations
|
||||
):
|
||||
mismatch = True # Just here for breakpoint
|
||||
logFile.write(f"Input:\n{subbedT}\n")
|
||||
logFile.write(f"Output:\n{translatedText}\n")
|
||||
|
||||
# Set if no mismatch
|
||||
if mismatch == False:
|
||||
tList[index] = extractedTranslations
|
||||
history = extractedTranslations[
|
||||
-MAXHISTORY:
|
||||
] # Update history if we have a list
|
||||
else:
|
||||
history = text[-MAXHISTORY:]
|
||||
mismatch = False
|
||||
if FILENAME not in MISMATCH:
|
||||
MISMATCH.append(FILENAME)
|
||||
|
||||
# Update Loading Bar
|
||||
with LOCK:
|
||||
if PBAR is not None:
|
||||
PBAR.update(len(tItem))
|
||||
else:
|
||||
# Ensure we're passing a single string to extractTranslation
|
||||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
|
||||
finalList = combineList(tList, text)
|
||||
return [finalList, totalTokens]
|
||||
if format == "json":
|
||||
return [finalList, totalTokens]
|
||||
else:
|
||||
return [finalList[0], totalTokens]
|
||||
|
|
|
|||
|
|
@ -392,10 +392,7 @@ Output ONLY the {LANGUAGE} translation in the following format: `Translation: <{
|
|||
{VOCAB}\n\
|
||||
"
|
||||
)
|
||||
if format == "json":
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
else:
|
||||
user = subbedT
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
return system, user
|
||||
|
||||
|
||||
|
|
@ -410,10 +407,7 @@ def translateText(system, user, history, penalty, format, model=MODEL):
|
|||
msg.append({"role": "system", "content": history})
|
||||
|
||||
# Response Format
|
||||
if format == "json":
|
||||
responseFormat = {"type": "json_object"}
|
||||
else:
|
||||
responseFormat = {"type": "text"}
|
||||
responseFormat = {"type": "json_object"}
|
||||
|
||||
# Content to TL
|
||||
msg.append({"role": "user", "content": f"{user}"})
|
||||
|
|
@ -438,7 +432,6 @@ def cleanTranslatedText(translatedText, varResponse):
|
|||
"「": '\\"',
|
||||
"」": '\\"',
|
||||
"- ": "-",
|
||||
"—": "-",
|
||||
"】": "]",
|
||||
"【": "[",
|
||||
"Placeholder Text": "",
|
||||
|
|
@ -483,10 +476,6 @@ def extractTranslation(translatedTextList, is_list):
|
|||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
except Exception as e:
|
||||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
|
||||
def countTokens(system, user, history):
|
||||
inputTotalTokens = 0
|
||||
|
|
@ -529,17 +518,15 @@ def translateGPT(text, history, fullPromptFlag):
|
|||
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if isinstance(tItem, list):
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
else:
|
||||
varResponse = [tItem, []]
|
||||
subbedT = varResponse[0]
|
||||
if not isinstance(tItem, list):
|
||||
tItem = [tItem]
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９\uFF61-\uFF9F]+", subbedT):
|
||||
|
|
@ -607,4 +594,7 @@ def translateGPT(text, history, fullPromptFlag):
|
|||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
|
||||
finalList = combineList(tList, text)
|
||||
return [finalList, totalTokens]
|
||||
if format == "json":
|
||||
return [finalList, totalTokens]
|
||||
else:
|
||||
return [finalList[0], totalTokens]
|
||||
|
|
|
|||
|
|
@ -283,20 +283,11 @@ def translateRenpy(data, translatedList):
|
|||
# Save some money and enter the character before translation
|
||||
def getSpeaker(speaker):
|
||||
match speaker:
|
||||
case "す":
|
||||
return ["Sumire", [0, 0]]
|
||||
case "ら":
|
||||
return ["Raul", [0, 0]]
|
||||
case "ぶ":
|
||||
return ["Gen", [0, 0]]
|
||||
case "さ":
|
||||
return ["Sasha", [0, 0]]
|
||||
case "む":
|
||||
return ["Villager", [0, 0]]
|
||||
case "ま":
|
||||
return ["Villager 2", [0, 0]]
|
||||
case "ファイン":
|
||||
return ["Fine", [0, 0]]
|
||||
case "":
|
||||
return ["", [0, 0]]
|
||||
case _:
|
||||
return [speaker, [0, 0]]
|
||||
# Find Speaker
|
||||
for i in range(len(NAMESLIST)):
|
||||
if speaker == NAMESLIST[i][0]:
|
||||
|
|
@ -352,10 +343,7 @@ Output ONLY the {LANGUAGE} translation in the following format: `Translation: <{
|
|||
{VOCAB}\n\
|
||||
"
|
||||
)
|
||||
if format == "json":
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
else:
|
||||
user = subbedT
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
return system, user
|
||||
|
||||
|
||||
|
|
@ -370,10 +358,7 @@ def translateText(system, user, history, penalty, format, model=MODEL):
|
|||
msg.append({"role": "system", "content": history})
|
||||
|
||||
# Response Format
|
||||
if format == "json":
|
||||
responseFormat = {"type": "json_object"}
|
||||
else:
|
||||
responseFormat = {"type": "text"}
|
||||
responseFormat = {"type": "json_object"}
|
||||
|
||||
# Content to TL
|
||||
msg.append({"role": "user", "content": f"{user}"})
|
||||
|
|
@ -442,10 +427,6 @@ def extractTranslation(translatedTextList, is_list):
|
|||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
except Exception as e:
|
||||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
|
||||
def countTokens(system, user, history):
|
||||
inputTotalTokens = 0
|
||||
|
|
@ -488,17 +469,15 @@ def translateGPT(text, history, fullPromptFlag):
|
|||
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if isinstance(tItem, list):
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
else:
|
||||
varResponse = [tItem, []]
|
||||
subbedT = varResponse[0]
|
||||
if not isinstance(tItem, list):
|
||||
tItem = [tItem]
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９\uFF61-\uFF9F]+", subbedT):
|
||||
|
|
@ -566,4 +545,7 @@ def translateGPT(text, history, fullPromptFlag):
|
|||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
|
||||
finalList = combineList(tList, text)
|
||||
return [finalList, totalTokens]
|
||||
if format == "json":
|
||||
return [finalList, totalTokens]
|
||||
else:
|
||||
return [finalList[0], totalTokens]
|
||||
|
|
|
|||
|
|
@ -2451,34 +2451,6 @@ def getSpeaker(speaker):
|
|||
return [speaker, [0, 0]]
|
||||
|
||||
|
||||
def subVars(jaString):
|
||||
jaString = jaString.replace("\u3000", " ")
|
||||
|
||||
# Formatting
|
||||
codeList = re.findall(r"([\\]*(\w+)\[(\d+)\])|([\\]*(\w+)\[[\\]*\\w+\[(\d+)\]\])", jaString)
|
||||
codeList = set(codeList)
|
||||
if len(codeList) != 0:
|
||||
for var in codeList:
|
||||
if var[2]:
|
||||
jaString = jaString.replace(var[0], f"[{var[1]}Code_" + f"{var[2]}]")
|
||||
else:
|
||||
jaString = jaString.replace(var[3], f"[{var[4]}Code_" + f"{var[5]}]")
|
||||
|
||||
# Put all lists in list and return
|
||||
return [jaString, codeList]
|
||||
|
||||
|
||||
def resubVars(translatedText, codeList):
|
||||
# Formatting
|
||||
for var in codeList:
|
||||
if var[2]:
|
||||
translatedText = translatedText.replace(f"[{var[1]}Code_" + f"{var[2]}]", var[0])
|
||||
else:
|
||||
translatedText = translatedText.replace(f"[{var[4]}Code_" + f"{var[5]}]", var[3])
|
||||
|
||||
return translatedText
|
||||
|
||||
|
||||
def batchList(input_list, batch_size):
|
||||
if not isinstance(batch_size, int) or batch_size <= 0:
|
||||
raise ValueError("batch_size must be a positive integer")
|
||||
|
|
@ -2503,10 +2475,7 @@ Output ONLY the {LANGUAGE} translation in the following format: `Translation: <{
|
|||
{VOCAB}\n\
|
||||
"
|
||||
)
|
||||
if format == "json":
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
else:
|
||||
user = subbedT
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
return system, user
|
||||
|
||||
|
||||
|
|
@ -2521,10 +2490,7 @@ def translateText(system, user, history, penalty, format, model=MODEL):
|
|||
msg.append({"role": "system", "content": history})
|
||||
|
||||
# Response Format
|
||||
if format == "json":
|
||||
responseFormat = {"type": "json_object"}
|
||||
else:
|
||||
responseFormat = {"type": "text"}
|
||||
responseFormat = {"type": "json_object"}
|
||||
|
||||
# Content to TL
|
||||
msg.append({"role": "user", "content": f"{user}"})
|
||||
|
|
@ -2559,7 +2525,6 @@ def cleanTranslatedText(translatedText, varResponse):
|
|||
|
||||
# Elongate Long Dashes (Since GPT Ignores them...)
|
||||
translatedText = elongateCharacters(translatedText)
|
||||
translatedText = resubVars(translatedText, varResponse[1])
|
||||
return translatedText
|
||||
|
||||
|
||||
|
|
@ -2594,10 +2559,6 @@ def extractTranslation(translatedTextList, is_list):
|
|||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
except Exception as e:
|
||||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
|
||||
def countTokens(system, user, history):
|
||||
inputTotalTokens = 0
|
||||
|
|
@ -2640,14 +2601,15 @@ def translateGPT(text, history, fullPromptFlag):
|
|||
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if isinstance(tItem, list):
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = subVars(payload)
|
||||
subbedT = varResponse[0]
|
||||
else:
|
||||
varResponse = [tItem, []]
|
||||
subbedT = varResponse[0]
|
||||
if not isinstance(tItem, list):
|
||||
tItem = [tItem]
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９\uFF61-\uFF9F]+", subbedT):
|
||||
|
|
@ -2715,4 +2677,7 @@ def translateGPT(text, history, fullPromptFlag):
|
|||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
|
||||
finalList = combineList(tList, text)
|
||||
return [finalList, totalTokens]
|
||||
if format == "json":
|
||||
return [finalList, totalTokens]
|
||||
else:
|
||||
return [finalList[0], totalTokens]
|
||||
|
|
|
|||
|
|
@ -2464,10 +2464,7 @@ Output ONLY the {LANGUAGE} translation in the following format: `Translation: <{
|
|||
{VOCAB}\n\
|
||||
"
|
||||
)
|
||||
if format == "json":
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
else:
|
||||
user = subbedT
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
return system, user
|
||||
|
||||
|
||||
|
|
@ -2482,10 +2479,7 @@ def translateText(system, user, history, penalty, format, model=MODEL):
|
|||
msg.append({"role": "system", "content": history})
|
||||
|
||||
# Response Format
|
||||
if format == "json":
|
||||
responseFormat = {"type": "json_object"}
|
||||
else:
|
||||
responseFormat = {"type": "text"}
|
||||
responseFormat = {"type": "json_object"}
|
||||
|
||||
# Content to TL
|
||||
msg.append({"role": "user", "content": f"{user}"})
|
||||
|
|
@ -2554,10 +2548,6 @@ def extractTranslation(translatedTextList, is_list):
|
|||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
except Exception as e:
|
||||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
|
||||
def countTokens(system, user, history):
|
||||
inputTotalTokens = 0
|
||||
|
|
@ -2600,17 +2590,15 @@ def translateGPT(text, history, fullPromptFlag):
|
|||
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if isinstance(tItem, list):
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
else:
|
||||
varResponse = [tItem, []]
|
||||
subbedT = varResponse[0]
|
||||
if not isinstance(tItem, list):
|
||||
tItem = [tItem]
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９\uFF61-\uFF9F]+", subbedT):
|
||||
|
|
@ -2678,4 +2666,7 @@ def translateGPT(text, history, fullPromptFlag):
|
|||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
|
||||
finalList = combineList(tList, text)
|
||||
return [finalList, totalTokens]
|
||||
if format == "json":
|
||||
return [finalList, totalTokens]
|
||||
else:
|
||||
return [finalList[0], totalTokens]
|
||||
|
|
|
|||
|
|
@ -310,121 +310,6 @@ def getSpeaker(speaker):
|
|||
return [speaker, [0, 0]]
|
||||
|
||||
|
||||
def subVars(jaString):
|
||||
jaString = jaString.replace("\u3000", " ")
|
||||
|
||||
# Nested
|
||||
count = 0
|
||||
nestedList = re.findall(r"[\\]+[\w]+\[[\\]+[\w]+\[[0-9]+\]\]", jaString)
|
||||
nestedList = set(nestedList)
|
||||
if len(nestedList) != 0:
|
||||
for icon in nestedList:
|
||||
jaString = jaString.replace(icon, "[Nested_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Icons
|
||||
count = 0
|
||||
iconList = re.findall(r"[\\]+[iIkKwWaA]+\[[0-9]+\]", jaString)
|
||||
iconList = set(iconList)
|
||||
if len(iconList) != 0:
|
||||
for icon in iconList:
|
||||
jaString = jaString.replace(icon, "[Ascii_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Colors
|
||||
count = 0
|
||||
colorList = re.findall(r"([\\]+c\[\d+\][\\]+c|[\\]+c\[\d+\])", jaString)
|
||||
colorList = set(colorList)
|
||||
if len(colorList) != 0:
|
||||
for color in colorList:
|
||||
jaString = jaString.replace(color, "[Color_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Names
|
||||
count = 0
|
||||
nameList = re.findall(r"[\\]+[nN]\[.+?\]+", jaString)
|
||||
nameList = set(nameList)
|
||||
if len(nameList) != 0:
|
||||
for name in nameList:
|
||||
jaString = jaString.replace(name, "[Noun_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Variables
|
||||
count = 0
|
||||
varList = re.findall(r"[\\]+[vV]\[[0-9]+\]", jaString)
|
||||
varList = set(varList)
|
||||
if len(varList) != 0:
|
||||
for var in varList:
|
||||
jaString = jaString.replace(var, "[Var_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Formatting
|
||||
count = 0
|
||||
formatList = re.findall(r"[\\]+[\w]+\[[a-zA-Z0-9\\\[\]\_,\s-]+\]", jaString)
|
||||
formatList = set(formatList)
|
||||
if len(formatList) != 0:
|
||||
for var in formatList:
|
||||
jaString = jaString.replace(var, "[FCode_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Put all lists in list and return
|
||||
allList = [nestedList, iconList, colorList, nameList, varList, formatList]
|
||||
return [jaString, allList]
|
||||
|
||||
|
||||
def resubVars(translatedText, allList):
|
||||
# Fix Spacing and ChatGPT Nonsense
|
||||
matchList = re.findall(r"\[\s?.+?\s?\]", translatedText)
|
||||
if len(matchList) > 0:
|
||||
for match in matchList:
|
||||
text = match.strip()
|
||||
translatedText = translatedText.replace(match, text)
|
||||
|
||||
# Nested
|
||||
count = 0
|
||||
if len(allList[0]) != 0:
|
||||
for var in allList[0]:
|
||||
translatedText = translatedText.replace("[Nested_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Icons
|
||||
count = 0
|
||||
if len(allList[1]) != 0:
|
||||
for var in allList[1]:
|
||||
translatedText = translatedText.replace("[Ascii_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Colors
|
||||
count = 0
|
||||
if len(allList[2]) != 0:
|
||||
for var in allList[2]:
|
||||
translatedText = translatedText.replace("[Color_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Names
|
||||
count = 0
|
||||
if len(allList[3]) != 0:
|
||||
for var in allList[3]:
|
||||
translatedText = translatedText.replace("[Noun_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Vars
|
||||
count = 0
|
||||
if len(allList[4]) != 0:
|
||||
for var in allList[4]:
|
||||
translatedText = translatedText.replace("[Var_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
# Formatting
|
||||
count = 0
|
||||
if len(allList[5]) != 0:
|
||||
for var in allList[5]:
|
||||
translatedText = translatedText.replace("[FCode_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
return translatedText
|
||||
|
||||
|
||||
def batchList(input_list, batch_size):
|
||||
if not isinstance(batch_size, int) or batch_size <= 0:
|
||||
raise ValueError("batch_size must be a positive integer")
|
||||
|
|
@ -432,18 +317,7 @@ def batchList(input_list, batch_size):
|
|||
return [input_list[i : i + batch_size] for i in range(0, len(input_list), batch_size)]
|
||||
|
||||
|
||||
def createContext(fullPromptFlag, subbedT):
|
||||
characters = "Game Characters:\n\
|
||||
シェーア (Shea) - Female\n\
|
||||
ミューテ (Mute) - Female\n\
|
||||
タビノ (Tabino) - Female\n\
|
||||
スラミー (Slamy) - Female\n\
|
||||
クリスタ (Christa) - Female\n\
|
||||
ソフィー (Sophie) - Female\n\
|
||||
ドーラ (Dora) - Female\n\
|
||||
ミューレ (Mule) - Female\n\
|
||||
"
|
||||
|
||||
def createContext(fullPromptFlag, subbedT, format):
|
||||
system = (
|
||||
PROMPT + VOCAB
|
||||
if fullPromptFlag
|
||||
|
|
@ -460,19 +334,13 @@ Output ONLY the {LANGUAGE} translation in the following format: `Translation: <{
|
|||
{VOCAB}\n\
|
||||
"
|
||||
)
|
||||
if isinstance(subbedT, list):
|
||||
user = f"```json\n{subbedT}```"
|
||||
else:
|
||||
user = subbedT
|
||||
return characters, system, user
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
return system, user
|
||||
|
||||
|
||||
def translateText(characters, system, user, history, penalty, format):
|
||||
def translateText(system, user, history, penalty, format, model=MODEL):
|
||||
# Prompt
|
||||
msg = [{"role": "system", "content": system + characters}]
|
||||
|
||||
# Characters
|
||||
msg.append({"role": "system", "content": characters})
|
||||
msg = [{"role": "system", "content": system}]
|
||||
|
||||
# History
|
||||
if isinstance(history, list):
|
||||
|
|
@ -481,17 +349,14 @@ def translateText(characters, system, user, history, penalty, format):
|
|||
msg.append({"role": "system", "content": history})
|
||||
|
||||
# Response Format
|
||||
if format == "json":
|
||||
responseFormat = {"type": "json_object"}
|
||||
else:
|
||||
responseFormat = {"type": "text"}
|
||||
responseFormat = {"type": "json_object"}
|
||||
|
||||
# Content to TL
|
||||
msg.append({"role": "user", "content": f"{user}"})
|
||||
response = openai.chat.completions.create(
|
||||
temperature=0,
|
||||
frequency_penalty=penalty,
|
||||
model=MODEL,
|
||||
model=model,
|
||||
response_format=responseFormat,
|
||||
messages=msg,
|
||||
)
|
||||
|
|
@ -509,6 +374,8 @@ def cleanTranslatedText(translatedText, varResponse):
|
|||
"「": '\\"',
|
||||
"」": '\\"',
|
||||
"- ": "-",
|
||||
"】": "]",
|
||||
"【": "[",
|
||||
"Placeholder Text": "",
|
||||
# Add more replacements as needed
|
||||
}
|
||||
|
|
@ -517,7 +384,6 @@ def cleanTranslatedText(translatedText, varResponse):
|
|||
|
||||
# Elongate Long Dashes (Since GPT Ignores them...)
|
||||
translatedText = elongateCharacters(translatedText)
|
||||
translatedText = resubVars(translatedText, varResponse[1])
|
||||
return translatedText
|
||||
|
||||
|
||||
|
|
@ -539,6 +405,7 @@ def elongateCharacters(text):
|
|||
def extractTranslation(translatedTextList, is_list):
|
||||
try:
|
||||
translatedTextList = re.sub(r'\\"+\"([^,\n}])', r'\\"\1', translatedTextList)
|
||||
translatedTextList = re.sub(r"(?<![\\])\"+(?![\n,])", r'"', translatedTextList)
|
||||
line_dict = json.loads(translatedTextList)
|
||||
# If it's a batch (i.e., list), extract with tags; otherwise, return the single item.
|
||||
string_list = list(line_dict.values())
|
||||
|
|
@ -552,7 +419,7 @@ def extractTranslation(translatedTextList, is_list):
|
|||
return None
|
||||
|
||||
|
||||
def countTokens(characters, system, user, history):
|
||||
def countTokens(system, user, history):
|
||||
inputTotalTokens = 0
|
||||
outputTotalTokens = 0
|
||||
enc = tiktoken.encoding_for_model("gpt-4")
|
||||
|
|
@ -564,7 +431,6 @@ def countTokens(characters, system, user, history):
|
|||
else:
|
||||
inputTotalTokens += len(enc.encode(history))
|
||||
inputTotalTokens += len(enc.encode(system))
|
||||
inputTotalTokens += len(enc.encode(characters))
|
||||
inputTotalTokens += len(enc.encode(user))
|
||||
|
||||
# Output
|
||||
|
|
@ -581,83 +447,96 @@ def combineList(tlist, text):
|
|||
|
||||
@retry(exceptions=Exception, tries=5, delay=5)
|
||||
def translateGPT(text, history, fullPromptFlag):
|
||||
global PBAR
|
||||
global PBAR, MISMATCH, FILENAME
|
||||
with open("log/translationHistory.txt", "a+", encoding="utf-8") as logFile:
|
||||
mismatch = False
|
||||
totalTokens = [0, 0]
|
||||
if isinstance(text, list):
|
||||
format = "json"
|
||||
tList = batchList(text, BATCHSIZE)
|
||||
else:
|
||||
format = "text"
|
||||
tList = [text]
|
||||
|
||||
mismatch = False
|
||||
totalTokens = [0, 0]
|
||||
if isinstance(text, list):
|
||||
format = "json"
|
||||
tList = batchList(text, BATCHSIZE)
|
||||
else:
|
||||
format = "text"
|
||||
tList = [text]
|
||||
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if isinstance(tItem, list):
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if not isinstance(tItem, list):
|
||||
tItem = [tItem]
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = subVars(payload)
|
||||
subbedT = varResponse[0]
|
||||
else:
|
||||
varResponse = subVars(tItem)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９]+", subbedT):
|
||||
if PBAR is not None:
|
||||
PBAR.update(len(tItem))
|
||||
continue
|
||||
|
||||
# Create Message
|
||||
characters, system, user = createContext(fullPromptFlag, subbedT)
|
||||
|
||||
# Calculate Estimate
|
||||
if ESTIMATE:
|
||||
estimate = countTokens(characters, system, user, history)
|
||||
totalTokens[0] += estimate[0]
|
||||
totalTokens[1] += estimate[1]
|
||||
continue
|
||||
|
||||
# Translating
|
||||
response = translateText(characters, system, user, history, 0.05, format)
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Check Translation
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(extractedTranslations):
|
||||
# Mismatch. Try Again
|
||||
response = translateText(characters, system, user, history, 0.05, format)
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Formatting
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(extractedTranslations):
|
||||
mismatch = True # Just here for breakpoint
|
||||
|
||||
# Set if no mismatch
|
||||
if mismatch == False:
|
||||
tList[index] = extractedTranslations
|
||||
history = extractedTranslations[-10:] # Update history if we have a list
|
||||
else:
|
||||
history = text[-10:]
|
||||
mismatch = False
|
||||
|
||||
# Update Loading Bar
|
||||
with LOCK:
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９\uFF61-\uFF9F]+", subbedT):
|
||||
if PBAR is not None:
|
||||
PBAR.update(len(tItem))
|
||||
else:
|
||||
# Ensure we're passing a single string to extractTranslation
|
||||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
history = tItem[-MAXHISTORY:]
|
||||
continue
|
||||
|
||||
# Create Message
|
||||
system, user = createContext(fullPromptFlag, subbedT, format)
|
||||
|
||||
# Calculate Estimate
|
||||
if ESTIMATE:
|
||||
estimate = countTokens(system, user, history)
|
||||
totalTokens[0] += estimate[0]
|
||||
totalTokens[1] += estimate[1]
|
||||
continue
|
||||
|
||||
# Translating
|
||||
response = translateText(system, user, history, 0.05, format)
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Check Translation
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(extractedTranslations):
|
||||
# Mismatch. Try Again
|
||||
response = translateText(system, user, history, 0.05, format, "gpt-4o")
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Formatting
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(
|
||||
extractedTranslations
|
||||
):
|
||||
mismatch = True # Just here for breakpoint
|
||||
logFile.write(f"Input:\n{subbedT}\n")
|
||||
logFile.write(f"Output:\n{translatedText}\n")
|
||||
|
||||
# Set if no mismatch
|
||||
if mismatch == False:
|
||||
tList[index] = extractedTranslations
|
||||
history = extractedTranslations[
|
||||
-MAXHISTORY:
|
||||
] # Update history if we have a list
|
||||
else:
|
||||
history = text[-MAXHISTORY:]
|
||||
mismatch = False
|
||||
if FILENAME not in MISMATCH:
|
||||
MISMATCH.append(FILENAME)
|
||||
|
||||
# Update Loading Bar
|
||||
with LOCK:
|
||||
if PBAR is not None:
|
||||
PBAR.update(len(tItem))
|
||||
else:
|
||||
# Ensure we're passing a single string to extractTranslation
|
||||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
|
||||
finalList = combineList(tList, text)
|
||||
return [finalList, totalTokens]
|
||||
if format == "json":
|
||||
return [finalList, totalTokens]
|
||||
else:
|
||||
return [finalList[0], totalTokens]
|
||||
|
|
|
|||
|
|
@ -334,10 +334,7 @@ Output ONLY the {LANGUAGE} translation in the following format: `Translation: <{
|
|||
{VOCAB}\n\
|
||||
"
|
||||
)
|
||||
if format == "json":
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
else:
|
||||
user = subbedT
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
return system, user
|
||||
|
||||
|
||||
|
|
@ -352,10 +349,7 @@ def translateText(system, user, history, penalty, format, model=MODEL):
|
|||
msg.append({"role": "system", "content": history})
|
||||
|
||||
# Response Format
|
||||
if format == "json":
|
||||
responseFormat = {"type": "json_object"}
|
||||
else:
|
||||
responseFormat = {"type": "text"}
|
||||
responseFormat = {"type": "json_object"}
|
||||
|
||||
# Content to TL
|
||||
msg.append({"role": "user", "content": f"{user}"})
|
||||
|
|
@ -380,7 +374,6 @@ def cleanTranslatedText(translatedText, varResponse):
|
|||
"「": '\\"',
|
||||
"」": '\\"',
|
||||
"- ": "-",
|
||||
"—": "-",
|
||||
"】": "]",
|
||||
"【": "[",
|
||||
"Placeholder Text": "",
|
||||
|
|
@ -425,10 +418,6 @@ def extractTranslation(translatedTextList, is_list):
|
|||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
except Exception as e:
|
||||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
|
||||
def countTokens(system, user, history):
|
||||
inputTotalTokens = 0
|
||||
|
|
@ -471,17 +460,15 @@ def translateGPT(text, history, fullPromptFlag):
|
|||
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if isinstance(tItem, list):
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
else:
|
||||
varResponse = [tItem, []]
|
||||
subbedT = varResponse[0]
|
||||
if not isinstance(tItem, list):
|
||||
tItem = [tItem]
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９\uFF61-\uFF9F]+", subbedT):
|
||||
|
|
@ -549,4 +536,7 @@ def translateGPT(text, history, fullPromptFlag):
|
|||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
|
||||
finalList = combineList(tList, text)
|
||||
return [finalList, totalTokens]
|
||||
if format == "json":
|
||||
return [finalList, totalTokens]
|
||||
else:
|
||||
return [finalList[0], totalTokens]
|
||||
|
|
|
|||
|
|
@ -410,10 +410,7 @@ Output ONLY the {LANGUAGE} translation in the following format: `Translation: <{
|
|||
{VOCAB}\n\
|
||||
"
|
||||
)
|
||||
if format == "json":
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
else:
|
||||
user = subbedT
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
return system, user
|
||||
|
||||
|
||||
|
|
@ -428,10 +425,7 @@ def translateText(system, user, history, penalty, format, model=MODEL):
|
|||
msg.append({"role": "system", "content": history})
|
||||
|
||||
# Response Format
|
||||
if format == "json":
|
||||
responseFormat = {"type": "json_object"}
|
||||
else:
|
||||
responseFormat = {"type": "text"}
|
||||
responseFormat = {"type": "json_object"}
|
||||
|
||||
# Content to TL
|
||||
msg.append({"role": "user", "content": f"{user}"})
|
||||
|
|
@ -500,10 +494,6 @@ def extractTranslation(translatedTextList, is_list):
|
|||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
except Exception as e:
|
||||
PBAR.write(f"extractTranslation Error: {e} on String {translatedTextList}")
|
||||
return None
|
||||
|
||||
|
||||
def countTokens(system, user, history):
|
||||
inputTotalTokens = 0
|
||||
|
|
@ -546,17 +536,15 @@ def translateGPT(text, history, fullPromptFlag):
|
|||
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if isinstance(tItem, list):
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
else:
|
||||
varResponse = [tItem, []]
|
||||
subbedT = varResponse[0]
|
||||
if not isinstance(tItem, list):
|
||||
tItem = [tItem]
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９\uFF61-\uFF9F]+", subbedT):
|
||||
|
|
@ -624,4 +612,7 @@ def translateGPT(text, history, fullPromptFlag):
|
|||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
|
||||
finalList = combineList(tList, text)
|
||||
return [finalList, totalTokens]
|
||||
if format == "json":
|
||||
return [finalList, totalTokens]
|
||||
else:
|
||||
return [finalList[0], totalTokens]
|
||||
|
|
|
|||
811
modules/wolf.py
811
modules/wolf.py
File diff suppressed because it is too large
Load diff
266
modules/wolf2.py
266
modules/wolf2.py
|
|
@ -353,40 +353,6 @@ def getSpeaker(speaker):
|
|||
return [speaker, [0, 0]]
|
||||
|
||||
|
||||
def subVars(jaString):
|
||||
jaString = jaString.replace("\u3000", " ")
|
||||
|
||||
# Formatting
|
||||
count = 0
|
||||
codeList = re.findall(r"[\\]+[\w]+\[[a-zA-Z0-9\\\[\]\_,\s-]+\]", jaString)
|
||||
codeList = set(codeList)
|
||||
if len(codeList) != 0:
|
||||
for var in codeList:
|
||||
jaString = jaString.replace(var, "[FCode_" + str(count) + "]")
|
||||
count += 1
|
||||
|
||||
# Put all lists in list and return
|
||||
return [jaString, codeList]
|
||||
|
||||
|
||||
def resubVars(translatedText, codeList):
|
||||
# Fix Spacing and ChatGPT Nonsense
|
||||
matchList = re.findall(r"\[\s?.+?\s?\]", translatedText)
|
||||
if len(matchList) > 0:
|
||||
for match in matchList:
|
||||
text = match.strip()
|
||||
translatedText = translatedText.replace(match, text)
|
||||
|
||||
# Formatting
|
||||
count = 0
|
||||
if len(codeList) != 0:
|
||||
for var in codeList:
|
||||
translatedText = translatedText.replace("[FCode_" + str(count) + "]", var)
|
||||
count += 1
|
||||
|
||||
return translatedText
|
||||
|
||||
|
||||
def batchList(input_list, batch_size):
|
||||
if not isinstance(batch_size, int) or batch_size <= 0:
|
||||
raise ValueError("batch_size must be a positive integer")
|
||||
|
|
@ -394,54 +360,7 @@ def batchList(input_list, batch_size):
|
|||
return [input_list[i : i + batch_size] for i in range(0, len(input_list), batch_size)]
|
||||
|
||||
|
||||
def createContext(fullPromptFlag, subbedT):
|
||||
characters = "Game Characters:\n\
|
||||
ロラン (Roland) - Male\n\
|
||||
リュカ (Ryuka) - Male\n\
|
||||
レックス (Rex) - Male\n\
|
||||
タバサ (Tabasa) - Female\n\
|
||||
アルス (Ars) - Male\n\
|
||||
アマカラ (Amakara) - Male\n\
|
||||
エリー (Eri) - Female\n\
|
||||
リオ (Rio) - Female\n\
|
||||
サマル (Samal) - Male\n\
|
||||
ムーン (Moon) - Female\n\
|
||||
アリーナ (Arina) - Female\n\
|
||||
クリフト (Cliff) - Male\n\
|
||||
マーニャ (Manya) - Female\n\
|
||||
ミネア (Minea) - Female\n\
|
||||
デボラ (Debora) - Female\n\
|
||||
ビアンカ (Bianca) - Female\n\
|
||||
フローラ (Flora) - Female\n\
|
||||
バーバラ (Barbara) - Female\n\
|
||||
ミレーユ (Mireyu) - Female\n\
|
||||
アイラ (Aira) - Female\n\
|
||||
フォズ (Foz) - Female\n\
|
||||
マリベル (Maribel) - Female\n\
|
||||
ククール (Kukool) - Male\n\
|
||||
ゲルダ (Gerda) - Female\n\
|
||||
ゼシカ (Jessica) - Female\n\
|
||||
ヤンガス (Yangus) - Male\n\
|
||||
ラヴィエル (Raviel) - Female\n\
|
||||
セティア (Setia) - Female\n\
|
||||
ダイ (Dai) - Male\n\
|
||||
ヒュンケル (Hyunckel) - Male\n\
|
||||
ポップ (Pop) - Male\n\
|
||||
マァム (Maam) - Female\n\
|
||||
レオナ (Leona) - Female\n\
|
||||
アステア (Astea) - Female\n\
|
||||
イヨ (Iyo) - Female\n\
|
||||
ジャガン (Jagan) - Male\n\
|
||||
ヤオ (Yao) - Female\n\
|
||||
デイジィ (Daisy) - Female\n\
|
||||
バイシュン (Baishun) - Male\n\
|
||||
ブライ (Buraimu) - Male\n\
|
||||
ハッサン (Hassan) - Male\n\
|
||||
アロマ (Aroma) - Female\n\
|
||||
ピッケ (Pikke) - Female\n\
|
||||
ドラオ (Dorao) - Male\n\
|
||||
"
|
||||
|
||||
def createContext(fullPromptFlag, subbedT, format):
|
||||
system = (
|
||||
PROMPT + VOCAB
|
||||
if fullPromptFlag
|
||||
|
|
@ -458,19 +377,13 @@ Output ONLY the {LANGUAGE} translation in the following format: `Translation: <{
|
|||
{VOCAB}\n\
|
||||
"
|
||||
)
|
||||
if isinstance(subbedT, list):
|
||||
user = f"```json\n{subbedT}```"
|
||||
else:
|
||||
user = subbedT
|
||||
return characters, system, user
|
||||
user = f"```json\n{subbedT}\n```"
|
||||
return system, user
|
||||
|
||||
|
||||
def translateText(characters, system, user, history, penalty, format):
|
||||
def translateText(system, user, history, penalty, format, model=MODEL):
|
||||
# Prompt
|
||||
msg = [{"role": "system", "content": system + characters}]
|
||||
|
||||
# Characters
|
||||
msg.append({"role": "system", "content": characters})
|
||||
msg = [{"role": "system", "content": system}]
|
||||
|
||||
# History
|
||||
if isinstance(history, list):
|
||||
|
|
@ -479,17 +392,14 @@ def translateText(characters, system, user, history, penalty, format):
|
|||
msg.append({"role": "system", "content": history})
|
||||
|
||||
# Response Format
|
||||
if format == "json":
|
||||
responseFormat = {"type": "json_object"}
|
||||
else:
|
||||
responseFormat = {"type": "text"}
|
||||
responseFormat = {"type": "json_object"}
|
||||
|
||||
# Content to TL
|
||||
msg.append({"role": "user", "content": f"{user}"})
|
||||
response = openai.chat.completions.create(
|
||||
temperature=0,
|
||||
frequency_penalty=penalty,
|
||||
model=MODEL,
|
||||
model=model,
|
||||
response_format=responseFormat,
|
||||
messages=msg,
|
||||
)
|
||||
|
|
@ -507,6 +417,8 @@ def cleanTranslatedText(translatedText, varResponse):
|
|||
"「": '\\"',
|
||||
"」": '\\"',
|
||||
"- ": "-",
|
||||
"】": "]",
|
||||
"【": "[",
|
||||
"Placeholder Text": "",
|
||||
# Add more replacements as needed
|
||||
}
|
||||
|
|
@ -515,7 +427,6 @@ def cleanTranslatedText(translatedText, varResponse):
|
|||
|
||||
# Elongate Long Dashes (Since GPT Ignores them...)
|
||||
translatedText = elongateCharacters(translatedText)
|
||||
translatedText = resubVars(translatedText, varResponse[1])
|
||||
return translatedText
|
||||
|
||||
|
||||
|
|
@ -537,6 +448,7 @@ def elongateCharacters(text):
|
|||
def extractTranslation(translatedTextList, is_list):
|
||||
try:
|
||||
translatedTextList = re.sub(r'\\"+\"([^,\n}])', r'\\"\1', translatedTextList)
|
||||
translatedTextList = re.sub(r"(?<![\\])\"+(?![\n,])", r'"', translatedTextList)
|
||||
line_dict = json.loads(translatedTextList)
|
||||
# If it's a batch (i.e., list), extract with tags; otherwise, return the single item.
|
||||
string_list = list(line_dict.values())
|
||||
|
|
@ -550,7 +462,7 @@ def extractTranslation(translatedTextList, is_list):
|
|||
return None
|
||||
|
||||
|
||||
def countTokens(characters, system, user, history):
|
||||
def countTokens(system, user, history):
|
||||
inputTotalTokens = 0
|
||||
outputTotalTokens = 0
|
||||
enc = tiktoken.encoding_for_model("gpt-4")
|
||||
|
|
@ -562,7 +474,6 @@ def countTokens(characters, system, user, history):
|
|||
else:
|
||||
inputTotalTokens += len(enc.encode(history))
|
||||
inputTotalTokens += len(enc.encode(system))
|
||||
inputTotalTokens += len(enc.encode(characters))
|
||||
inputTotalTokens += len(enc.encode(user))
|
||||
|
||||
# Output
|
||||
|
|
@ -579,83 +490,96 @@ def combineList(tlist, text):
|
|||
|
||||
@retry(exceptions=Exception, tries=5, delay=5)
|
||||
def translateGPT(text, history, fullPromptFlag):
|
||||
global PBAR
|
||||
global PBAR, MISMATCH, FILENAME
|
||||
with open("log/translationHistory.txt", "a+", encoding="utf-8") as logFile:
|
||||
mismatch = False
|
||||
totalTokens = [0, 0]
|
||||
if isinstance(text, list):
|
||||
format = "json"
|
||||
tList = batchList(text, BATCHSIZE)
|
||||
else:
|
||||
format = "text"
|
||||
tList = [text]
|
||||
|
||||
mismatch = False
|
||||
totalTokens = [0, 0]
|
||||
if isinstance(text, list):
|
||||
format = "json"
|
||||
tList = batchList(text, BATCHSIZE)
|
||||
else:
|
||||
format = "text"
|
||||
tList = [text]
|
||||
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if isinstance(tItem, list):
|
||||
for index, tItem in enumerate(tList):
|
||||
# Before sending to translation, if we have a list of items, add the formatting
|
||||
if not isinstance(tItem, list):
|
||||
tItem = [tItem]
|
||||
for j in range(len(tItem)):
|
||||
if not tItem[j]:
|
||||
tItem[j] = tItem[j].replace("", "Placeholder Text")
|
||||
payload = {f"Line{i+1}": string for i, string in enumerate(tItem)}
|
||||
payload = json.dumps(payload, indent=4, ensure_ascii=False)
|
||||
varResponse = subVars(payload)
|
||||
subbedT = varResponse[0]
|
||||
else:
|
||||
varResponse = subVars(tItem)
|
||||
varResponse = [payload, []]
|
||||
subbedT = varResponse[0]
|
||||
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９]+", subbedT):
|
||||
if PBAR is not None:
|
||||
PBAR.update(len(tItem))
|
||||
continue
|
||||
|
||||
# Create Message
|
||||
characters, system, user = createContext(fullPromptFlag, subbedT)
|
||||
|
||||
# Calculate Estimate
|
||||
if ESTIMATE:
|
||||
estimate = countTokens(characters, system, user, history)
|
||||
totalTokens[0] += estimate[0]
|
||||
totalTokens[1] += estimate[1]
|
||||
continue
|
||||
|
||||
# Translating
|
||||
response = translateText(characters, system, user, history, 0.05, format)
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Check Translation
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(extractedTranslations):
|
||||
# Mismatch. Try Again
|
||||
response = translateText(characters, system, user, history, 0.05, format)
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Formatting
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(extractedTranslations):
|
||||
mismatch = True # Just here for breakpoint
|
||||
|
||||
# Set if no mismatch
|
||||
if mismatch == False:
|
||||
tList[index] = extractedTranslations
|
||||
history = extractedTranslations[-10:] # Update history if we have a list
|
||||
else:
|
||||
history = text[-10:]
|
||||
mismatch = False
|
||||
|
||||
# Update Loading Bar
|
||||
with LOCK:
|
||||
# Things to Check before starting translation
|
||||
if not re.search(r"[一-龠ぁ-ゔァ-ヴーａ-ｚＡ-Ｚ０-９\uFF61-\uFF9F]+", subbedT):
|
||||
if PBAR is not None:
|
||||
PBAR.update(len(tItem))
|
||||
else:
|
||||
# Ensure we're passing a single string to extractTranslation
|
||||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
history = tItem[-MAXHISTORY:]
|
||||
continue
|
||||
|
||||
# Create Message
|
||||
system, user = createContext(fullPromptFlag, subbedT, format)
|
||||
|
||||
# Calculate Estimate
|
||||
if ESTIMATE:
|
||||
estimate = countTokens(system, user, history)
|
||||
totalTokens[0] += estimate[0]
|
||||
totalTokens[1] += estimate[1]
|
||||
continue
|
||||
|
||||
# Translating
|
||||
response = translateText(system, user, history, 0.05, format)
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Check Translation
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(extractedTranslations):
|
||||
# Mismatch. Try Again
|
||||
response = translateText(system, user, history, 0.05, format, "gpt-4o")
|
||||
translatedText = response.choices[0].message.content
|
||||
totalTokens[0] += response.usage.prompt_tokens
|
||||
totalTokens[1] += response.usage.completion_tokens
|
||||
|
||||
# Formatting
|
||||
translatedText = cleanTranslatedText(translatedText, varResponse)
|
||||
if isinstance(tItem, list):
|
||||
extractedTranslations = extractTranslation(translatedText, True)
|
||||
if extractedTranslations == None or len(tItem) != len(
|
||||
extractedTranslations
|
||||
):
|
||||
mismatch = True # Just here for breakpoint
|
||||
logFile.write(f"Input:\n{subbedT}\n")
|
||||
logFile.write(f"Output:\n{translatedText}\n")
|
||||
|
||||
# Set if no mismatch
|
||||
if mismatch == False:
|
||||
tList[index] = extractedTranslations
|
||||
history = extractedTranslations[
|
||||
-MAXHISTORY:
|
||||
] # Update history if we have a list
|
||||
else:
|
||||
history = text[-MAXHISTORY:]
|
||||
mismatch = False
|
||||
if FILENAME not in MISMATCH:
|
||||
MISMATCH.append(FILENAME)
|
||||
|
||||
# Update Loading Bar
|
||||
with LOCK:
|
||||
if PBAR is not None:
|
||||
PBAR.update(len(tItem))
|
||||
else:
|
||||
# Ensure we're passing a single string to extractTranslation
|
||||
tList[index] = translatedText.replace("Placeholder Text", "")
|
||||
|
||||
finalList = combineList(tList, text)
|
||||
return [finalList, totalTokens]
|
||||
if format == "json":
|
||||
return [finalList, totalTokens]
|
||||
else:
|
||||
return [finalList[0], totalTokens]
|
||||
|
|
|
|||
Loading…
Reference in a new issue