IssueTracker

I'm running Gemini Nano locally on Pixel 9 Pro, version com.google.ai.edge.aicore:aicore:0.0.1-exp01, and I'm getting this error more often than the actual response.

com.google.ai.edge.aicore.InferenceException: AICore failed with error type 2-INFERENCE_ERROR and error code 4-REQUEST_PROCESSING_ERROR: Inference failed.

The prompt details certainly affect this behavior, but I can't figure out what trips it exactly.

Model config:

        val generationConfig = generationConfig {
            context = App.instance
            temperature = 0.0f
            topK = 3
            maxOutputTokens = 1024
        }
        val model = GenerativeModel(generationConfig = generationConfig)

Prompt:

You will receive emails as input. Summarize the email content in 1-3 sentences.
Email: Hi, I'm writing to inform you that your salary has been deposited into your account. Your balance is $7,500. You can use your card 1234 4567 1234 8907 to withdraw it from ATM.

I shoehorned the card part because it seems like the model consistently breaks on similar inputs.

The web version (not Nano though) works fine. Removing You can use your card 1234 4567 1234 8907 to withdraw it from ATM. fixes the issue.

I've had this error on many other prompts as well.