Why SpeziLLMLocal doesn't work on TestFlight but it works locally

[Turquoise-T]

In what area do you have a technical challenge?

Swift Package Manager

Description

I uitilized SpeziLLMLocal (llama) in my project.
In my feature, I want to give llama a prompt as the input and want to get the output based on the input.
I just did like this:
func getMealProtein(meal: String) async { await MainActor.run { self.proteinAmount = "" } let prompt = promptTemplate.constructPrompt(mealName: meal) let llmSession: LLMLocalSession = runner( with: LLMLocalSchema( model: .llama3_8B_4bit, parameters: .init( systemPrompt: prompt ) ) ) // let llmSession: LLMLocalSession = runner( // with: LLMLocalSchema( // model: .llama3_8B_4bit // ) // ) do { for try await token in try await llmSession.generate() { await MainActor.run { self.proteinAmount.append(token) } } print("Protein extracted is ", proteinAmount) } catch { print("Error generating protein: \(error)") } }
I can get a response immediately when I run locally(just some warning message but at least I know it works and it should be tested on the TestFlight.) But on Testflight, when I entered the mealName, just nothing happened.

Reproduction

• local:
  
• testflight

661_1741603756.mp4

Expected behavior

On testflight has instant feedback for protein when I enter the mealName

Additional context

No response

Code of Conduct

• I agree to follow this course's Code of Conduct and Contributing Guidelines

[PSchmiedmayer]

@felixschlegel @nriedman @AdritRao @vishnuravi Do you mind taking a look at this?

@Turquoise-T Can you verify that the local model has been successfully downloaded to your device? It might be that you have an older phone and the model is too large for this device.
What type of device are you using and could you test that it works by deploying to app to your phone using Xcode? You can learn more how to run the app on your phone here: https://github.com/orgs/CS342/discussions/156

[Turquoise-T]

@vishnuravi Thanks for your help! I have did that. However I found that when it is deployed on the testFlight, it just comes across some issue: see here: https://github.com/CS342/2025-Stanford-360/actions/runs/13821787969/job/38670188770. I'm not sure why this happened and how can I solve it.

[Turquoise-T]

@vishnuravi I think this is a more detailed reason.

Does it mean I can only change it to a smaller model as @PSchmiedmayer Paul suggests?

[vishnuravi]

@Turquoise-T can you try running it on your own phone first to see if it helps with the issue?

[Turquoise-T]

@vishnuravi @PSchmiedmayer Hi! I solved the problem by changing it to a smaller model! Thanks all for the help!🥳