Errors when trying to translate to Simplified Chinese

[yuxi-liu-wired]

First attempt:

• .\gpt-subtrans test.srt --project True --scenethreshold 10 --target_language "Simplified Chinese"

Error: Can't write project file, no scenes
Traceback (most recent call last):
  File "C:\Users\DeadScholar\Desktop\gpt-subtrans\gpt-subtrans.py", line 75, in <module>
    project.TranslateSubtitles()
  File "C:\Users\DeadScholar\Desktop\gpt-subtrans\PySubtitle\SubtitleProject.py", line 107, in TranslateSubtitles
    self.WriteProjectFile()
  File "C:\Users\DeadScholar\Desktop\gpt-subtrans\PySubtitle\SubtitleProject.py", line 215, in WriteProjectFile
    raise Exception("Can't write project file, no scenes")
Exception: Can't write project file, no scenes

Second attempt:

• Used the GUI by running gui-subtrans. It translated about 100 lines correctly, but then started throwing errors like this on the command line, and after throwing this error for a while, the GUI abruptly died.

INFO: HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
INFO: Scene 1 batch 33: 0 lines and 0 untranslated.
INFO: Summary: 这一批字幕讨论了GPT（生成预训练变换器）的概念，以及它如何被用于翻译语言。同时，也提到了所有这些工具使用的训练数据都是被抓取的。
--- Logging error ---
Traceback (most recent call last):
  File "C:\Users\DeadScholar\Miniconda3\lib\logging\__init__.py", line 1103, in emit
    stream.write(msg + self.terminator)
  File "C:\Users\DeadScholar\Miniconda3\lib\encodings\cp1252.py", line 19, in encode
    return codecs.charmap_encode(input,self.errors,encoding_table)[0]
UnicodeEncodeError: 'charmap' codec can't encode characters in position 15-22: character maps to <undefined>
Call stack:
  File "C:\Users\DeadScholar\Desktop\gpt-subtrans\GUI\Command.py", line 57, in run
    success = self.execute()
  File "C:\Users\DeadScholar\Desktop\gpt-subtrans\GUI\ProjectCommands.py", line 293, in execute
    scene = project.TranslateScene(self.scene_number, batch_numbers=self.batch_numbers, line_numbers=self.line_numbers, translator = self.translator)
  File "C:\Users\DeadScholar\Desktop\gpt-subtrans\PySubtitle\SubtitleProject.py", line 158, in TranslateScene
    translator.TranslateScene(scene, batch_numbers=batch_numbers, line_numbers=line_numbers)
  File "C:\Users\DeadScholar\Desktop\gpt-subtrans\PySubtitle\SubtitleTranslator.py", line 139, in TranslateScene
    self.TranslateBatches(self.client, batches, line_numbers, context, remaining_lines)
  File "C:\Users\DeadScholar\Desktop\gpt-subtrans\PySubtitle\SubtitleTranslator.py", line 239, in TranslateBatches
    self.ProcessTranslation(batch, line_numbers, context, client)
  File "C:\Users\DeadScholar\Desktop\gpt-subtrans\PySubtitle\SubtitleTranslator.py", line 349, in ProcessTranslation
    logging.info(f"Summary: {batch.summary}")
Message: 'Summary: 这一批字幕讨论了GPT（生成预训练变换器）的概念，以及它如何被用于翻译语言。同时，也提到了所有这些工具使用的训练数据都是被抓取的。'
Arguments: ()

Third attempt:

• Reopened the GUI and tried loading the savefile test.subtrans. Got error

INFO: Reading project data from test.subtrans
ERROR: Unable to load test.srt ('NoneType' object cannot be interpreted as an integer)

Loading test.srt directly gives the same error. Deleting the test.subtrans would allow loading test.srt again, but starting anew.

[yuxi-liu-wired]

I found the solution, I think. I just need to add the encoding='utf-8', option every time I import logging.

logging.basicConfig(
    format='%(levelname)s: %(message)s',
    encoding='utf-8', 
    level=logging_level
    )

[machinewrapped]

Thanks for the report, I'll take a look. Adding encoding to the logger config makes sense, but it should only be necessary to do it in the main scripts (gpt-subtrans.py, gui-subtrans.py) as they configure a global logger instance.

[machinewrapped]

Actually now that I think about it I have vague memories of Windows causing problems when I tried to specify the encoding for the console logger, so I ended up leaving it with default encoding... it might be necessary to add a command line argument to set the encoding, or to write a test message in a try-catch block and fall back to default encoding if it fails.

[machinewrapped]

Can you try this? It attempts to log as utf-8 and falls back to default encoding if it fails: https://github.com/machinewrapped/gpt-subtrans/commit/c60d8fb5ebdfae8251df00d7adf52f12f318b163

I haven't seen it fail yet so the fallback is untested, I'll see if I can remember under what conditions utf-8 wasn't supported.

[yuxi-liu-wired]

I have tried this and it works as expected. There are no error messages in the command line or the GUI, and the translation works, and the savefile works.