Some changes to json

This commit is contained in:
Dazed 2023-11-07 11:01:10 -06:00
parent 9d7e8896a1
commit ddee81460e

View file

@ -1,4 +1,3 @@
from concurrent.futures import ThreadPoolExecutor, as_completed
import json
import os
from pathlib import Path
@ -21,16 +20,17 @@ load_dotenv()
openai.organization = os.getenv('org')
openai.api_key = os.getenv('key')
APICOST = .002 # Depends on the model https://openai.com/pricing
INPUTAPICOST = .002 # Depends on the model https://openai.com/pricing
OUTPUTAPICOST = .002
PROMPT = Path('prompt.txt').read_text(encoding='utf-8')
THREADS = 10 # For GPT4 rate limit will be hit if you have more than 1 thread.
THREADS = 10 # Controls how many threads are working on a single file (May have to drop this)
LOCK = threading.Lock()
WIDTH = 90
LISTWIDTH = 60
WIDTH = 50
LISTWIDTH = 90
NOTEWIDTH = 50
MAXHISTORY = 10
ESTIMATE = ''
TOTALCOST = 0
TOTALTOKENS = 0
totalTokens = [0, 0]
NAMESLIST = []
#tqdm Globals
@ -39,9 +39,10 @@ POSITION=0
LEAVE=False
BRFLAG = False # If the game uses <br> instead
FIXTEXTWRAP = True
IGNORETLTEXT = False
def handleJSON(filename, estimate):
global ESTIMATE, TOKENS, TOTALTOKENS, TOTALCOST
global ESTIMATE, totalTokens
ESTIMATE = estimate
if estimate:
@ -52,10 +53,10 @@ def handleJSON(filename, estimate):
end = time.time()
tqdm.write(getResultString(translatedData, end - start, filename))
with LOCK:
TOTALCOST += translatedData[1] * .001 * APICOST
TOTALTOKENS += translatedData[1]
totalTokens[0] += translatedData[1][0]
totalTokens[1] += translatedData[1][1]
return getResultString(['', TOTALTOKENS, None], end - start, 'TOTAL')
return getResultString(['', totalTokens, None], end - start, 'TOTAL')
else:
try:
@ -68,20 +69,19 @@ def handleJSON(filename, estimate):
json.dump(translatedData[0], outFile, ensure_ascii=False)
tqdm.write(getResultString(translatedData, end - start, filename))
with LOCK:
TOTALCOST += translatedData[1] * .001 * APICOST
TOTALTOKENS += translatedData[1]
totalTokens[0] += translatedData[1][0]
totalTokens[1] += translatedData[1][1]
except Exception as e:
traceback.print_exc()
return 'Fail'
return getResultString(['', TOTALTOKENS, None], end - start, 'TOTAL')
return getResultString(['', totalTokens, None], end - start, 'TOTAL')
def openFiles(filename):
with open('files/' + filename, 'r', encoding='UTF-8-sig') as f:
data = json.load(f)
# Map Files
if 'test' in filename:
if '.json' in filename:
translatedData = parseJSON(data, filename)
else:
@ -91,25 +91,30 @@ def openFiles(filename):
def getResultString(translatedData, translationTime, filename):
# File Print String
tokenString = Fore.YELLOW + '[' + str(translatedData[1]) + \
' Tokens/${:,.4f}'.format(translatedData[1] * .001 * APICOST) + ']'
totalTokenstring =\
Fore.YELLOW +\
'[Input: ' + str(translatedData[1][0]) + ']'\
'[Output: ' + str(translatedData[1][1]) + ']'\
'[Cost: ${:,.4f}'.format((translatedData[1][0] * .001 * INPUTAPICOST) +\
(translatedData[1][1] * .001 * OUTPUTAPICOST)) + ']'
timeString = Fore.BLUE + '[' + str(round(translationTime, 1)) + 's]'
if translatedData[2] == None:
# Success
return filename + ': ' + tokenString + timeString + Fore.GREEN + u' \u2713 ' + Fore.RESET
return filename + ': ' + totalTokenstring + timeString + Fore.GREEN + u' \u2713 ' + Fore.RESET
else:
# Fail
try:
raise translatedData[2]
except Exception as e:
traceback.print_exc()
errorString = str(e) + Fore.RED
return filename + ': ' + tokenString + timeString + Fore.RED + u' \u2717 ' +\
return filename + ': ' + totalTokenstring + timeString + Fore.RED + u' \u2717 ' +\
errorString + Fore.RESET
def parseJSON(data, filename):
totalTokens = 0
totalTokens = [0, 0]
totalLines = 0
totalLines = len(data)
global LOCK
@ -118,7 +123,9 @@ def parseJSON(data, filename):
pbar.desc=filename
pbar.total=totalLines
try:
totalTokens += translateJSON(data, pbar)
result = translateJSON(data, pbar)
totalTokens[0] += result[0]
totalTokens[1] += result[1]
except Exception as e:
return [data, totalTokens, e]
return [data, totalTokens, None]
@ -126,36 +133,55 @@ def parseJSON(data, filename):
def translateJSON(data, pbar):
textHistory = []
maxHistory = MAXHISTORY
tokens = 0
tokens = [0, 0]
speaker = 'None'
for item in data:
# Remove any textwrap
if FIXTEXTWRAP == True:
jaString = item['value']
jaString = jaString.replace('\n', '')
for item in data.items():
# Speaker
if 'name' in item[1]:
if item[1]['name'] not in [None, '-']:
response = translateGPT(item[1]['name'], 'Reply with only the english translation of the NPC name', False)
speaker = response[0]
tokens[0] += response[1][0]
tokens[1] += response[1][1]
item[1]['name'] = speaker
else:
speaker = 'None'
# Translate
if jaString != '':
response = translateGPT(jaString, 'Past Translated Text: ' + '|\n\n'.join(textHistory), True)
tokens += response[1]
translatedText = response[0]
textHistory.append('\"' + translatedText + '\"')
else:
translatedText = jaString
textHistory.append('\"' + translatedText + '\"')
# Text
if 'text' in item[1]:
if item[1]['text'] != None:
jaString = item[1]['text']
# Textwrap
translatedText = textwrap.fill(translatedText, width=WIDTH)
translatedText = translatedText.replace('\n', '@b')
# Remove any textwrap
if FIXTEXTWRAP == True:
jaString = jaString.replace('\n', '')
# Set Data
item['value'] = translatedText
# Translate
if jaString != '':
response = translateGPT(f'{speaker} | {jaString}', textHistory, True)
tokens[0] += response[1][0]
tokens[1] += response[1][1]
translatedText = response[0]
textHistory.append('\"' + translatedText + '\"')
else:
translatedText = jaString
textHistory.append('\"' + translatedText + '\"')
# Keep textHistory list at length maxHistory
if len(textHistory) > maxHistory:
textHistory.pop(0)
currentGroup = []
pbar.update(1)
# Remove added speaker
translatedText = re.sub(r'^.+?\s\|\s?', '', translatedText)
# Textwrap
translatedText = textwrap.fill(translatedText, width=WIDTH)
# Set Data
item[1]['text'] = translatedText
# Keep textHistory list at length maxHistory
if len(textHistory) > maxHistory:
textHistory.pop(0)
currentGroup = []
pbar.update(1)
return tokens
@ -164,11 +190,11 @@ def subVars(jaString):
# Icons
count = 0
iconList = re.findall(r'[\\]+[iI]\[[0-9]+\]', jaString)
iconList = re.findall(r'[\\]+[iIkKwW]+\[[0-9]+\]', jaString)
iconList = set(iconList)
if len(iconList) != 0:
for icon in iconList:
jaString = jaString.replace(icon, '<I' + str(count) + '>')
jaString = jaString.replace(icon, '[Ascii_' + str(count) + ']')
count += 1
# Colors
@ -177,16 +203,16 @@ def subVars(jaString):
colorList = set(colorList)
if len(colorList) != 0:
for color in colorList:
jaString = jaString.replace(color, '<C' + str(count) + '>')
jaString = jaString.replace(color, '[Color_' + str(count) + ']')
count += 1
# Names
count = 0
nameList = re.findall(r'[\\]+[nN]\[[0-9]+\]', jaString)
nameList = re.findall(r'[\\]+[nN]\[.+?\]+', jaString)
nameList = set(nameList)
if len(nameList) != 0:
for name in nameList:
jaString = jaString.replace(name, '<N' + str(count) + '>')
jaString = jaString.replace(name, '[N_' + str(count) + ']')
count += 1
# Variables
@ -195,16 +221,18 @@ def subVars(jaString):
varList = set(varList)
if len(varList) != 0:
for var in varList:
jaString = jaString.replace(var, '<V' + str(count) + '>')
jaString = jaString.replace(var, '[Var_' + str(count) + ']')
count += 1
# Formatting
count = 0
formatList = re.findall(r'[\\]+[!.]', jaString)
if '笑えるよね.' in jaString:
print('t')
formatList = re.findall(r'[\\]+CL', jaString)
formatList = set(formatList)
if len(formatList) != 0:
for format in formatList:
jaString = jaString.replace(format, '<F' + str(count) + '>')
for var in formatList:
jaString = jaString.replace(var, '[FCode_' + str(count) + ']')
count += 1
# Put all lists in list and return
@ -213,7 +241,7 @@ def subVars(jaString):
def resubVars(translatedText, allList):
# Fix Spacing and ChatGPT Nonsense
matchList = re.findall(r'<\s?.+?\s?>', translatedText)
matchList = re.findall(r'\[\s?.+?\s?\]', translatedText)
if len(matchList) > 0:
for match in matchList:
text = match.replace(' ', '')
@ -223,35 +251,35 @@ def resubVars(translatedText, allList):
count = 0
if len(allList[0]) != 0:
for var in allList[0]:
translatedText = translatedText.replace('<I' + str(count) + '>', var)
translatedText = translatedText.replace('[Ascii_' + str(count) + ']', var)
count += 1
# Colors
count = 0
if len(allList[1]) != 0:
for var in allList[1]:
translatedText = translatedText.replace('<C' + str(count) + '>', var)
translatedText = translatedText.replace('[Color_' + str(count) + ']', var)
count += 1
# Names
count = 0
if len(allList[2]) != 0:
for var in allList[2]:
translatedText = translatedText.replace('<N' + str(count) + '>', var)
translatedText = translatedText.replace('[N_' + str(count) + ']', var)
count += 1
# Vars
count = 0
if len(allList[3]) != 0:
for var in allList[3]:
translatedText = translatedText.replace('<V' + str(count) + '>', var)
translatedText = translatedText.replace('[Var_' + str(count) + ']', var)
count += 1
# Formatting
count = 0
if len(allList[4]) != 0:
for var in allList[4]:
translatedText = translatedText.replace('<F' + str(count) + '>', var)
translatedText = translatedText.replace('[FCode_' + str(count) + ']', var)
count += 1
# Remove Color Variables Spaces
@ -264,9 +292,18 @@ def resubVars(translatedText, allList):
def translateGPT(t, history, fullPromptFlag):
# If ESTIMATE is True just count this as an execution and return.
if ESTIMATE:
enc = tiktoken.encoding_for_model("gpt-4")
tokens = len(enc.encode(t)) * 2 + len(enc.encode(str(history))) + len(enc.encode(PROMPT))
return (t, tokens)
enc = tiktoken.encoding_for_model("gpt-3.5-turbo")
historyRaw = ''
if isinstance(history, list):
for line in history:
historyRaw += line
else:
historyRaw = history
inputTotalTokens = len(enc.encode(historyRaw)) + len(enc.encode(PROMPT))
outputTotalTokens = len(enc.encode(t)) * 2 # Estimating 2x the size of the original text
totalTokens = [inputTotalTokens, outputTotalTokens]
return (t, totalTokens)
# Sub Vars
varResponse = subVars(t)
@ -274,16 +311,15 @@ def translateGPT(t, history, fullPromptFlag):
# If there isn't any Japanese in the text just skip
if not re.search(r'[一-龠]+|[ぁ-ゔ]+|[ァ-ヴ]+|[\uFF00-\uFFEF]', subbedT):
return(t, 0)
return(t, [0,0])
# Characters
context = '```\
Game Characters:\
Character: 池ノ上 拓海 == Ikenoue Takumi - Gender: Male\
Character: 福永 こはる == Fukunaga Koharu - Gender: Female\
Character: ソル == Sol - Gender: Female\
Character: ェニ先生 == Eni-sensei - Gender: Female\
Character: 神泉 理央 == Kamiizumi Rio - Gender: Female\
Character: 吉祥寺 アリサ == Kisshouji Arisa - Gender: Female\
Character: 久我 友里子 == Kuga Yuriko - Gender: Female\
```'
# Prompt
@ -309,14 +345,14 @@ def translateGPT(t, history, fullPromptFlag):
temperature=0.1,
frequency_penalty=0.2,
presence_penalty=0.2,
model="gpt-3.5-turbo",
model="gpt-3.5-turbo-1106",
messages=msg,
request_timeout=30,
)
# Save Translated Text
translatedText = response.choices[0].message.content
tokens = response.usage.total_tokens
totalTokens = [response.usage.prompt_tokens, response.usage.completion_tokens]
# Resub Vars
translatedText = resubVars(translatedText, varResponse[1])
@ -339,4 +375,4 @@ def translateGPT(t, history, fullPromptFlag):
if len(translatedText) > 15 * len(t) or "I'm sorry, but I'm unable to assist with that translation" in translatedText:
raise Exception
else:
return [translatedText, tokens]
return [translatedText, totalTokens]