Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[VSCode] Empty chat panel in devcontainer #3424

Open
WatchMeBakinWaffles opened this issue Nov 14, 2024 · 4 comments
Open

[VSCode] Empty chat panel in devcontainer #3424

WatchMeBakinWaffles opened this issue Nov 14, 2024 · 4 comments

Comments

@WatchMeBakinWaffles
Copy link

Describe the bug
Tabby chat panel hangs and end up being empty when running in devcontainer.

First :
image

Few tens of seconds later :
image

I don't see any error in vscode tabby output.

Information about your version
Using tabby version : 0.20.0

Information about your GPU
image

Additional context
Running in devcontainer based on docker image debian:12.7-slim

@zwpaper
Copy link
Member

zwpaper commented Nov 27, 2024

Hello @WatchMeBakinWaffles, when running Tabby or other GPU-intensive applications in a container, you should use an NVIDIA-based image. Could you try using images such as nvidia/cuda:11.7.1-devel-ubuntu22.04?

Additionally, there is a state check on the System page in the Tabby admin homepage settings. Could you also try to verify the state of the models there?

CleanShot 2024-11-27 at 12 11 34@2x

@WatchMeBakinWaffles
Copy link
Author

WatchMeBakinWaffles commented Nov 28, 2024

Hello @zwpaper. I will try next monday. But I'm still wondering, why would chat feature require nvidia based image while code completion is working correctly ?

@zwpaper
Copy link
Member

zwpaper commented Dec 3, 2024

@WatchMeBakinWaffles, you can check the model's functionality on the System page using a Debian-based image.

However, in most cases, an NVIDIA-based image is required for a GPU application to utilize GPU capabilities.

@WatchMeBakinWaffles
Copy link
Author

WatchMeBakinWaffles commented Dec 3, 2024

Oh ok. I guess there is a misunderstanding there :

My VSCode extension is running inside a devcontainer whereas tabby server is actually running on my Ubuntu WSL (and clearly running on the GPU, I can confirm).

My issue is that the chat feature in VSC works well when VSC is running on host while it stops running when switching to devcontainer (and actually, the chat through web GUI is nice and smooth).

Additional infos:

Here is my run command:
tabby-serve='tabby serve --model DeepseekCoder-1.3B --chat-model Qwen2-1.5B-Instruct --device cuda --port 8089'

EDIT:

image

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants