search

JSchmie / ScrAIbe-WebUI

WebUI for ScAIbe
https://github.com/JSchmie/ScrAIbe-WebUI
GNU General Public License v3.0
19 stars 2 forks source link

[BUG] docker: Error response from daemon: could not select device driver "" with capabilities: [[gpu]]. #42

Closed ithakaa closed 2 months ago

ithakaa commented 2 months ago

I am attempting to install ScrAIbe-WebUI inside an Ubuntu 22.04 LXC

docker run -d --name scraibe-webui -p 7860:7860 --gpus 'all' -v $(pwd)/data:/data hadr0n/scraibe-webui"

docker: Error response from daemon: could not select device driver "" with capabilities: [[gpu]].

I do not have a dedicated GPU so am attempting this with a CPU only setup

JSchmie commented 2 months ago

Hi, I think this issue originates in your docker run you are applying the --gpu 'all' option, which tells docker to use a GPU. Try to remove the --gpus 'all' Option from your docker run command. So your command would be: 

docker run -d --name scraibe-webui -p 7860:7860 -v $(pwd)/data:/data hadr0n/scraibe-webui"

If this works, please let us know, then we will try to enhance our docs to avoid this problem in the future. :)

ithakaa commented 2 months ago

Thank you, that did the trick, however...

A few issues:

  1. If I am in the process of transcribing an audio file and I reload the webpage, the task is no longer listed as being processed, what happens to the active task ?
  2. After the task is complete, is there a download transcription option ?
  3. Transcribing a large 1.5GB audio file results in an "error" on the webpage
JSchmie commented 2 months ago

Regarding your questions:

  1. If I am in the process of transcribing an audio file and I reload the webpage, the task is no longer listed as being processed. What happens to the active task?

The active task will continue to be processed in the background, but it will no longer appear in your front end after a page reload. Once the task is finished, the associated file should be deleted from your tmp folder. This behavior occurs because Gradio, as mentioned in this issue, does not persist session data across page reloads. Additionally, the app doesn't use cookies or similar methods to track sessions, so each page refresh is treated as a new session. However, by using our asynchronous backend, you can avoid this issue. In that case, the file and email are linked to the session data, and the processed file will be emailed to you once the task is complete.

  1. After the task is complete, is there a download transcription option?

Currently, there isn't a specific download button for the transcription. However, since the transcription is just plain text, you can use the copy button above each field to manually save it to a file. Alternatively, using our asynchronous backend will provide you with both a .txt and a .json file when the transcription is complete.

  1. Transcribing a large 1.5GB audio file results in an "error" on the webpage.

Thank you for reporting this issue. Unfortunately, I haven’t been able to reproduce the error with a large file. Could you please submit another issue using our bug report template, providing more details about your system, network configuration, and any relevant Docker logs? This will help us investigate the problem more effectively.

ithakaa commented 1 month ago

ok, thanks, great project i'll be back when it works and is more mature, great work so far !!