From 60712983cebca8bde64aa43790121cd3163c788c Mon Sep 17 00:00:00 2001 From: DazedAnon Date: Sat, 1 Jun 2024 13:20:41 -0500 Subject: [PATCH] Fixes to mvmz --- modules/rpgmakerace.py | 47 ++++++++++++++-------------------------- modules/rpgmakermvmz.py | 48 +++++++++++++++++++++++++++++------------ 2 files changed, 50 insertions(+), 45 deletions(-) diff --git a/modules/rpgmakerace.py b/modules/rpgmakerace.py index 01f3dec..a0fd540 100644 --- a/modules/rpgmakerace.py +++ b/modules/rpgmakerace.py @@ -50,7 +50,7 @@ if 'gpt-3.5' in MODEL: elif 'gpt-4o' in MODEL: INPUTAPICOST = .005 OUTPUTAPICOST = .015 - BATCHSIZE = 50 + BATCHSIZE = 20 FREQUENCY_PENALTY = 0.1 #tqdm Globals @@ -394,7 +394,6 @@ def parseSystem(data, filename): termList = data['terms'][term] totalLines += len(termList) totalLines += len(data['game_title']) - totalLines += len(data['terms']['messages']) totalLines += len(data['variables']) totalLines += len(data['weapon_types']) totalLines += len(data['armor_types']) @@ -1873,31 +1872,6 @@ def searchSystem(data, pbar): totalTokens[0] += response[1][0] totalTokens[1] += response[1][1] data['weapon_types'][i] = response[0].replace('\"', '').strip() - - - # Variables (Optional ususally) - # for i in range(len(data['variables'])): - # response = translateGPT(data['variables'][i], 'Reply with only the '+ LANGUAGE +' translation of the title', False) - # totalTokens[0] += response[1][0] - # totalTokens[1] += response[1][1] - # data['variables'][i] = response[0].replace('\"', '').strip() - # - - # Messages - messages = (data['terms']['messages']) - for key, value in messages.items(): - response = translateGPT(value, 'Reply with only the '+ LANGUAGE +' translation of the battle text.\nTranslate "常時ダッシュ" as "Always Dash"\nTranslate "次の%1まで" as Next %1.', False) - translatedText = response[0] - - # Remove characters that may break scripts - charList = ['.', '\"', '\\n'] - for char in charList: - translatedText = translatedText.replace(char, '') - - totalTokens[0] += response[1][0] - totalTokens[1] += response[1][1] - messages[key] = translatedText - return totalTokens @@ -2052,10 +2026,21 @@ 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\ -葵 (Aoi) - Female\n\ -豪山剛 (Tsuyoshi Gouyama) - Male\n\ -' + characters = "Game Characters:\n\ +朱音 (Akane) - Female\n\ +満留 (Maru) - Female\n\ +ルイム (Ruimu) - Female\n\ +都子 (Miyako) - Female\n\ +たかし (Takashi) - Male\n\ +レヴァン (Revan) - Male\n\ +大海 (Taikai) - Female\n\ +桜花 (Ouka) - Female\n\ +烈火 (Rekka) - Female\n\ +よみこ (Yomiko) - Female\n\ +新 (Arata) - Female\n\ +ヴィルトルクス (Viltrex) - Female\n\ +光男 (Mitsuo) - Male\n\ +" system = PROMPT + VOCAB if fullPromptFlag else \ f"\ diff --git a/modules/rpgmakermvmz.py b/modules/rpgmakermvmz.py index 3a501a7..51dcc11 100644 --- a/modules/rpgmakermvmz.py +++ b/modules/rpgmakermvmz.py @@ -48,7 +48,7 @@ if 'gpt-3.5' in MODEL: elif 'gpt-4o' in MODEL: INPUTAPICOST = .005 OUTPUTAPICOST = .015 - BATCHSIZE = 20 + BATCHSIZE = 40 FREQUENCY_PENALTY = 0.1 #tqdm Globals @@ -744,19 +744,33 @@ def searchCodes(page, pbar, jobList, filename): i += 1 continue - # Check for Speaker - coloredSpeakerList = re.findall(r'^[\\]+[cC]\[[\d]+\](.+?)[\\]+[Cc]\[[\d]\]\\?\\?$', jaString) - if len(coloredSpeakerList) == 0: - coloredSpeakerList = re.findall(r'^【(.*?)】$', jaString) - if len(coloredSpeakerList) != 0 and codeList[i+1]['code'] in [401, 405, -1]: + # Speaker Check + # Colors + speakerList = re.findall(r'^[\\]+[cC]\[[\d]+\](.+?)[\\]+[Cc]\[[\d]\]\\?\\?$', jaString) + + # Brackets + if len(speakerList) == 0: + speakerList = re.findall(r'^【(.*?)】$', jaString) + + # None + if len(speakerList) == 0: + if len(jaString) < 40 \ + and 'code' in codeList[i+1] \ + and codeList[i+1]['code'] in [401, 405, -1] \ + and len(codeList[i+1]['parameters']) > 0 \ + and len(codeList[i+1]['parameters'][0]) > 0: + if codeList[i+1]['parameters'][0][0] in ['「', '"', '(', '（', '*', '[']: + speakerList = re.findall(r'.+', jaString) + + if len(speakerList) != 0 and codeList[i+1]['code'] in [401, 405, -1]: # Get Speaker - response = getSpeaker(coloredSpeakerList[0]) + response = getSpeaker(speakerList[0]) speaker = response[0] totalTokens[0] += response[1][0] totalTokens[1] += response[1][1] # Set Data - codeList[i]['parameters'][0] = jaString.replace(coloredSpeakerList[0], speaker) + codeList[i]['parameters'][0] = jaString.replace(speakerList[0], speaker) # Iterate to next string i += 1 @@ -772,9 +786,11 @@ def searchCodes(page, pbar, jobList, filename): # Join Up 401's into single string if len(codeList) > i+1: while codeList[i+1]['code'] in [401, 405, -1]: - codeList[i]['parameters'] = [] - codeList[i]['code'] = -1 + if setData == True: + codeList[i]['parameters'] = [] + codeList[i]['code'] = -1 i += 1 + j = i # Only add if not empty if len(codeList[i]['parameters']) > 0: @@ -787,7 +803,7 @@ def searchCodes(page, pbar, jobList, filename): # Format String if len(currentGroup) > 0: - finalJAString = ''.join(currentGroup).replace('？', '?') + finalJAString = ' '.join(currentGroup).replace('？', '?') oldjaString = finalJAString # Check if Empty @@ -796,7 +812,8 @@ def searchCodes(page, pbar, jobList, filename): continue # Set Back - codeList[i]['parameters'] = [finalJAString] + if setData == True: + codeList[i]['parameters'] = [finalJAString] ### \\n nCase = None @@ -888,6 +905,8 @@ def searchCodes(page, pbar, jobList, filename): finalJAString = finalJAString.replace('。', '.') finalJAString = re.sub(r'(\.{3}\.+)', '...', finalJAString) finalJAString = finalJAString.replace('　', '') + finalJAString = finalJAString.replace('」', '\"') + finalJAString = finalJAString.replace('」', '\"') ### Remove format codes # Furigana @@ -2044,8 +2063,7 @@ def batchList(input_list, batch_size): def createContext(fullPromptFlag, subbedT): characters = 'Game Characters:\n\ -葵 (Aoi) - Female\n\ -豪山剛 (Tsuyoshi Gouyama) - Male\n\ +グレイス (Grace) - Female\n\ ' system = PROMPT + VOCAB if fullPromptFlag else \ @@ -2098,6 +2116,8 @@ def cleanTranslatedText(translatedText, varResponse): '< ': '<', '': '>', + '「': '\"', + ' 」': '\"', 'Placeholder Text': '', '- chan': '-chan', '- kun': '-kun',