CAN NOT RUN TURBOPILOT USING DOCKER

[tuancao216]

Hi,

I've followed your guideline by running turbopilot from docker image. I ran into the error:

I'm using Ubuntu WSL2 on Windows 11.

[ravenscroftj]

Very strange - it looks like you have the right arguments and config set up. Can you confirm that the model file you downloaded is from "the bloke" repository and that it is in tact (try running sha256sum ./models/stablecode-instruct-alpha-3b.ggmlv1.q4_0.bin - the hash you get should be 7143e0148e5c0101fb66ce6e233182c10db4a52b438cae557a7d368f98eb29a1). You may need to re-download the model.

If that's not the problem it may be something to do with the GGML loader behaviour on WSL but I won't get ahead of myself until we've confirmed that the files are ok.

[tuancao216]

Very strange - it looks like you have the right arguments and config set up. Can you confirm that the model file you downloaded is from "the bloke" repository and that it is in tact (try running sha256sum ./models/stablecode-instruct-alpha-3b.ggmlv1.q4_0.bin - the hash you get should be 7143e0148e5c0101fb66ce6e233182c10db4a52b438cae557a7d368f98eb29a1). You may need to re-download the model.

If that's not the problem it may be something to do with the GGML loader behaviour on WSL but I won't get ahead of myself until we've confirmed that the files are ok.

The issue go away, but I got another issue while I was using your extension to integrate with vscode.

I believe my laptop has enough RAM to run this model:

[tuancao216]

@ravenscroftj any update on this issue?

[ravenscroftj]

I'd encourage you to try again with the latest release and also try running with the --debug flag to get more verbose output (rather than using the env vars you can just completely override the docker command after you specify the image name)

Segfault isn't a particularly informative error message so if the changes that went in over the last few days didn't fix it we will have to find a way to get some better output.

[tuancao216]

I'd encourage you to try again with the latest release and also try running with the --debug flag to get more verbose output (rather than using the env vars you can just completely override the docker command after you specify the image name)

Segfault isn't a particularly informative error message so if the changes that went in over the last few days didn't fix it we will have to find a way to get some better output.

I try with the latest release and it doesn't work.