machinewrapped / gpt-subtrans

Open Source project using LLMs to translate SRT subtitles
Other
346 stars 39 forks source link

gui-subtrans on Windows can't load SRT file #171

Closed zhoukuncheng closed 4 months ago

zhoukuncheng commented 4 months ago

: Error initialising project : Value '' not found in []

:

INFO: Initialising 
INFO: Loaded settings from C:\Users\jack\AppData\Roaming\MachineWrapped\GPTSubtrans\settings.json
INFO: Loading stylesheet from C:\tools\gui-subtrans\_internal\\subtrans.qss
INFO: GUI-Subtrans v0.7.0
INFO:  to C:\Users\jack\AppData\Roaming\MachineWrapped\GPTSubtrans\gui-subtrans.log
INFO: Project file C:\Users\jack\Downloads\You need $500. How should you get it.en-US.subtrans does not exist
ERROR: Error initialising project settings: Value '' not found in []

file: a random VTT subtitle downloaded from YouTube and converted to SRT:

You need $500. How should you get it.en-US.srt.zip

Version: v0.7.1

machinewrapped commented 4 months ago

It doesn't seem to be a problem with the SRT file, it looks like a configuration issue.

Which provider are you using? Did you run the install script for it?

machinewrapped commented 4 months ago

Try v0.7.3, I added some extra handling in the new project settings dialog, which might at least provide more information about what the problem is. https://github.com/machinewrapped/gpt-subtrans/releases/tag/v0.7.3

machinewrapped commented 4 months ago

I found the problem - it did happen if the provider is not configured correctly (e.g. an invalid API key). Should be fixed in v0.7.4. https://github.com/machinewrapped/gpt-subtrans/releases/tag/v0.7.4

You will still need to configure the provider before you can translate, but at least it should tell you that instead of crashing :-)

zhoukuncheng commented 4 months ago

Thanks. In v0.7.4, it works now.

I am using gui-subtrans with a local model (wangshenzhi/llama3-8b-chinese-chat-ollama-q4:latest) with Ollama, so I don’t need to configure the API key.

machinewrapped commented 4 months ago

Ah, good luck - I haven't had good results with local models so far, let me know how it goes!

On Mon, 13 May 2024 at 16:20, Zhoukun Cheng @.***> wrote:

Thanks. In v0.7.4, it works now.

I am using gui-subtrans with a local model ( wangshenzhi/llama3-8b-chinese-chat-ollama-q4:latest) with Ollama, so I don’t need to configure the API key.

— Reply to this email directly, view it on GitHub https://github.com/machinewrapped/gpt-subtrans/issues/171#issuecomment-2107713564, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACNLYBHSR4Q6XFRSVDWJEVLZCDD2TAVCNFSM6AAAAABHQ3NJJ6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMBXG4YTGNJWGQ . You are receiving this because you commented.Message ID: @.***>

zhoukuncheng commented 4 months ago

With a fine-tuned llama3-8b model, I see the translation results are acceptable. For example, wangshenzhi/llama3-8b-chinese-chat-ollama-q4:latest for English to Chinese translation.