Skip to content

Commit

Permalink
Merge pull request #807 from popcion/main
Browse files Browse the repository at this point in the history 
add split translation
  • Loading branch information
@zyddnys
zyddnys authored Feb 5, 2025
2 parents 126bb08 + 49dcd3d commit f75b669
Show file tree
Hide file tree
Showing 2 changed files with 248 additions and 125 deletions.
176 changes: 129 additions & 47 deletions manga_translator/translators/chatgpt.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -9,7 +9,7 @@
import asyncio
import time
from typing import List, Dict

from manga_translator.utils import is_valuable_text
from .common import CommonTranslator, MissingAPIKeyException
from .keys import OPENAI_API_KEY, OPENAI_HTTP_PROXY, OPENAI_API_BASE
CONFIG = None
Expand Down Expand Up @@ -42,7 +42,7 @@ class GPT3Translator(CommonTranslator):
'THA': 'Thai',
'IND': 'Indonesian'
}
_INVALID_REPEAT_COUNT = 2 # repeat up to 2 times if "invalid" translation was detected
_INVALID_REPEAT_COUNT = 0 # useless
_MAX_REQUESTS_PER_MINUTE = 20
_TIMEOUT = 40 # Seconds to wait for a response from the server before retrying
_RETRY_ATTEMPTS = 3 # Number of times to retry an errored request before giving up
Expand Down Expand Up @@ -130,95 +130,177 @@ def _format_prompt_log(self, to_lang: str, prompt: str) -> str:
async def _translate(self, from_lang: str, to_lang: str, queries: List[str]) -> List[str]:
translations = [''] * len(queries)
self.logger.debug(f'Temperature: {self.temperature}, TopP: {self.top_p}')
MAX_SPLIT_ATTEMPTS = 5 # Default max split attempts
RETRY_ATTEMPTS = self._RETRY_ATTEMPTS

query_index = 0
for prompt, query_size in self._assemble_prompts(from_lang, to_lang, queries):
self.logger.debug('-- GPT Prompt --\n' + self._format_prompt_log(to_lang, prompt))
async def translate_batch(prompt_queries, prompt_query_indices, split_level=0):
nonlocal MAX_SPLIT_ATTEMPTS
split_prefix = ' (split)' if split_level > 0 else ''

for attempt in range(self._RETRY_ATTEMPTS):
try:
response = await self._request_translation(to_lang, prompt)
self.logger.debug('-- GPT Response --\n' + response)
# Assemble prompt for the current batch
prompt, query_size = self._assemble_prompts(from_lang, to_lang, prompt_queries).__next__()
self.logger.debug(f'-- GPT Prompt{split_prefix} --\n' + self._format_prompt_log(to_lang, prompt))

# split
for attempt in range(RETRY_ATTEMPTS):
try:
# Start the translation request with timeout handling
request_task = asyncio.create_task(self._request_translation(to_lang, prompt))
started = time.time()
timeout_attempt = 0
ratelimit_attempt = 0
server_error_attempt = 0
while not request_task.done():
await asyncio.sleep(0.1)
if time.time() - started > self._TIMEOUT + (timeout_attempt * self._TIMEOUT / 2):
# Server takes too long to respond
if timeout_attempt >= self._TIMEOUT_RETRY_ATTEMPTS:
raise Exception('Openai servers did not respond quickly enough.')
timeout_attempt += 1
self.logger.warn(f'Restarting request due to timeout. Attempt: {timeout_attempt}')
request_task.cancel()
request_task = asyncio.create_task(self._request_translation(to_lang, prompt))
started = time.time()

# Get the response
response = await request_task
self.logger.debug(f'-- GPT Response{split_prefix} --\n' + response)

# Split response into translations
new_translations = re.split(r'<\|\d+\|>', response)
if not new_translations[0].strip():
new_translations = new_translations[1:]

if len(queries) == 1 and len(new_translations) == 1 and not re.match(r'^\s*<\|\d+\|>', response) :
self.logger.warn(f'Single query response does not contain prefix, retrying...')
continue
if len(prompt_queries) == 1 and len(new_translations) == 1 and not re.match(r'^\s*<\|\d+\|>', response):
self.logger.warn(f'Single query response does not contain prefix, retrying...(Attempt {attempt + 1})')
continue

#Check for error messages in translations
# Check for error messages in translations
ERROR_KEYWORDS = [
# ENG_KEYWORDS
# ENG_KEYWORDS
#"sorry,",
"I'm sorry, I can't assist with that.",
"I'm sorry, but I can't assist with that request.",
"I'm sorry, I can't assist with that.",
#"I apologize",
#"assist with",
"I cannot help with",
"I must decline",
"I am not comfortable about",
"I will not engage with",
#"not comfortable",
#"engage with",
"I cannot generate or create",
#"I'd prefer not to",
#"I must refrain from",
"I'd prefer not to",
"I must refrain from",
"This goes beyond what I can",
#"unable",
"That's not something I can help with",
#"appropriate",

# CHINESE_KEYWORDS
"抱歉，我不",
"我无法满足该请求",
# CHINESE_KEYWORDS
"抱歉，我不",
"我无法满足该请求",
"对不起，我不",
"我无法将",
"我无法把",
"我无法回答你",
"我无法将",
"我无法把",
"我无法回答你",
"这超出了我的范围",
"我不便回答",
"我不能提供相关建议",
"这类内容我不能处理",
"我需要婉拒",

# JAPANESE_KEYWORDS
"申し訳ありませんが",
# JAPANESE_KEYWORDS
"申し訳ありませんが",
]
if any(keyword in t for t in new_translations for keyword in ERROR_KEYWORDS):
self.logger.warn(f'Error message detected in response, retrying... (Attempt {attempt + 1})')
continue
if any(keyword in t for t in new_translations for keyword in ERROR_KEYWORDS):
remaining_attempts = RETRY_ATTEMPTS - attempt - 1
self.logger.warn(f'Error message detected in response, remaining {remaining_attempts} time(s) before splitting the translation.')
continue

if len(new_translations) < query_size:
# Try splitting by newlines instead
new_translations = re.split(r'\n', response)

if len(new_translations) < query_size:
self.logger.warn(f'Incomplete response, retrying... (Attempt {attempt + 1})')
remaining_attempts = RETRY_ATTEMPTS - attempt - 1
self.logger.warn(f'Incomplete response, remaining {remaining_attempts} time(s) before splitting the translation.')
continue

# Trim excess translations and pad if necessary
new_translations = new_translations[:query_size] + [''] * (query_size - len(new_translations))

# Clean translations by keeping only the content before the first newline
new_translations = [t.split('\n')[0].strip() for t in new_translations]

# Successfully obtained translations for the current batch
translations[query_index:query_index + query_size] = [t.strip() for t in new_translations]
query_index += query_size
break
new_translations = [t.split('\n')[0].strip() for t in new_translations]
# Remove any potential prefix markers
new_translations = [re.sub(r'^\s*<\|\d+\|>\s*', '', t) for t in new_translations]

#for i, translation in enumerate(new_translations):
# if not is_valuable_text(translation):
# self.logger.info(f'Filtered out: {translation}')
# self.logger.info('Reason: Text is not considered valuable.')
# new_translations[i] = ''

# Check if any translations are empty
if any(not t.strip() for t in new_translations):
self.logger.warn(f'Empty translations detected. Resplitting the batch.')
break # Exit retry loop and trigger split logic below

# Store the translations in the correct indices
for idx, translation in zip(prompt_query_indices, new_translations):
translations[idx] = translation

# Log progress
self.logger.info(f'Batch translated: {len([t for t in translations if t])}/{len(queries)} completed.')
self.logger.debug(f'Completed translations: {[t if t else queries[i] for i, t in enumerate(translations)]}')
return True # Successfully translated this batch

except openai.RateLimitError: # Server returned ratelimit response
ratelimit_attempt += 1
if ratelimit_attempt >= self._RATELIMIT_RETRY_ATTEMPTS:
raise
self.logger.warn(
f'Restarting request due to ratelimiting by openai servers. Attempt: {ratelimit_attempt}')
await asyncio.sleep(2)
except openai.APIError:
server_error_attempt += 1
if server_error_attempt >= self._RETRY_ATTEMPTS:
self.logger.error(
'Openai encountered a server error, possibly due to high server load. Use a different translator or try again later.')
raise
self.logger.warn(f'Restarting request due to a server error. Attempt: {server_error_attempt}')
await asyncio.sleep(1)
except Exception as e:
self.logger.error(f'Error during translation attempt: {e}')
if attempt == self._RETRY_ATTEMPTS - 1:
if attempt == RETRY_ATTEMPTS - 1:
raise
await asyncio.sleep(1)

# Remove any potential prefix markers before returning the results
translations = [re.sub(r'^\s*<\|\d+\|>\s*', '', t) for t in translations]
# If retries exhausted and still not successful, proceed to split if allowed
if split_level < MAX_SPLIT_ATTEMPTS:
if split_level == 0:
self.logger.warn('Retry limit reached. Starting to split the translation batch.')
else:
self.logger.warn('Further splitting the translation batch due to persistent errors.')
mid_index = len(prompt_queries) // 2
futures = []
# Split the batch into two halves
for sub_queries, sub_indices in [
(prompt_queries[:mid_index], prompt_query_indices[:mid_index]),
(prompt_queries[mid_index:], prompt_query_indices[mid_index:]),
]:
if sub_queries:
futures.append(translate_batch(sub_queries, sub_indices, split_level + 1))
results = await asyncio.gather(*futures)
return all(results)
else:
self.logger.error('Maximum split attempts reached. Unable to translate the following queries:')
for idx in prompt_query_indices:
self.logger.error(f'Query: {queries[idx]}')
return False # Indicate failure for this batch

# Begin translation process
prompt_queries = queries
prompt_query_indices = list(range(len(queries)))
await translate_batch(prompt_queries, prompt_query_indices)

self.logger.debug(translations)
if self.token_count_last:
self.logger.info(f'Used {self.token_count_last} tokens (Total: {self.token_count})')

return translations

async def _request_translation(self, to_lang: str, prompt: str) -> str:
Expand Down Expand Up @@ -334,7 +416,7 @@ async def _request_translation(self, to_lang: str, prompt: str) -> str:
class GPT4Translator(GPT35TurboTranslator):
_CONFIG_KEY = 'gpt4'
_MAX_REQUESTS_PER_MINUTE = 200
_RETRY_ATTEMPTS = 5
_RETRY_ATTEMPTS = 3
_MAX_TOKENS = 8192

@property
Expand All @@ -357,7 +439,7 @@ async def _request_translation(self, to_lang: str, prompt: str) -> str:

try:
response = await self.client.chat.completions.create(
model='gpt-4o',
model='chatgpt-4o-latest',
messages=messages,
max_tokens=self._MAX_TOKENS // 2,
temperature=self.temperature,
Expand Down
Loading

0 comments on commit f75b669

Please sign in to comment.