Skip to main content

Indie game storeFree gamesFun gamesHorror games
Game developmentAssetsComics
SalesBundles
Jobs
TagsGame Engines
(5 edits)

 Linux Mint 22, 64GB DDR5, AMD RYZEN 9 7950X , AMD Radeon 6900 XT w/16gb VRAM.

I know it's not my system because in older iterations 1.6.4 and 1.6.5 I was just fine.
Selected local GPU, but it seems like it attempts to connect to an external LLM service and errors out with:

The LLM service has encountered an Error: An error occurred while sending the request.


Below is what I pulled from the log as it ran.

I assume it's attempting to use a 3rd party LLM service instead of my local GPU because of the Tokens, but I'll eat humble pie if I'm wrong.

Tokens: 797

<CheckMemoryCompression>d__163:MoveNext()

System.Threading.ExecutionContext:RunInternal(ExecutionContext, ContextCallback, Object, Boolean)

System.Runtime.CompilerServices.MoveNextRunner:Run()

System.Threading.Tasks.AwaitTaskContinuation:RunCallback(ContextCallback, Object, Task&)

System.Threading.Tasks.Task:FinishContinuations()

System.Threading.Tasks.Task`1:TrySetResult(TResult)

System.Runtime.CompilerServices.AsyncTaskMethodBuilder`1:SetResult(TResult)

KoboldSharp.<GetTokenCount>d__7:MoveNext()

System.Threading.ExecutionContext:RunInternal(ExecutionContext, ContextCallback, Object, Boolean)

System.Runtime.CompilerServices.MoveNextRunner:Run()

UnityEngine.WorkRequest:Invoke()

UnityEngine.UnitySynchronizationContext:Exec()

System.Net.Http.HttpRequestException: An error occurred while sending the request ---> System.Net.WebException: Error getting response stream (ReadDoneAsync2): ReceiveFailure

I'm not trying to run it via 3rd Party Vendors, is there a setting I can change in the .config files to force it to utilize my GPU?

Steps to reproduce:
Download 1.6.6b without having run any other versions of Silverpine. Launch. After AI Model loads, attempt to talk to any NPC. Error should appear.

I'm pretty sure I'm not doing anything out of the ordinary for this.


Edit: Loading the Demo Server things behave as expected, but attempting to load via GPU using the Demo program ALSO triggers this LLM Warning.


Edit 02: Loading older QWEN-3.5 works. I Didn't both with Gemma-4-Dense-Large, I don't have the VRAM for it, but since the Sparse isn't behaving I doubt it's larger brother would either.

The game communicates with the locally hosted KoboldCPP server in the same way it would with the demo server or OpenRouter.

To me it seems like the backend is failing in some way without the game noticing, and then the request errors out.
KoboldCPP outputs are currently sent to /dev/null on Linux because I had some worries about compatibility with different distros, so there's no way to see why this is happening either. Since it's model specific, maybe the download got corrupted somehow. Try deleting the Gemma-4-Sparse .gguf and .meta files in Silverpine 1.6.6b\Silverpine_Data\StreamingAssets to redownload it.

Another thing it could be is that Gemma 4 uses SWA, which might be an issue with AMD somehow?
If anything, I will look into adding better logging for the outputs on Linux in 1.7.

(5 edits)

Oh! Okay, so we're literally talking to our own local server as if we would with a 3rd party provider. That makes sense why I was getting that error then. 

You could use the .config location that you use for saves to also record logs. I'd say defaulting to /dev/null with the option to enable using ~/.config/unity3d/Three Eyes Software/Silverpine/logs or allowing a user to specify a specific dir would be an option as well.

Linux generally has a number of dirs that are found regardless of distro. So, if we're not using the /home/user/.config/silverpine/logs we could utilize /var/log/silverpine or something else. 

Anyways, I'm going to try out your suggestion and repo0rt back! Thank you Three Eye!

Edit: Manually Downloaded Gemma-4-Sparse outside of the Game, looked into the .seal file and found the URL for it. Deleting and redownloading through the game didn't change behaviour. Did the same with Koboldcpp... We'll see what happens. 

Edit02: Right, so that didn't change the behavior either. Ah well, I'll just try it once we reach 1.7!