mlc-ai / mlc-llm

Universal LLM Deployment Engine with ML Compilation
https://llm.mlc.ai/
Apache License 2.0
19.26k stars 1.58k forks source link

[Question] Android app related sentence generation #2958

Closed j0h0k0i0m closed 3 weeks ago

j0h0k0i0m commented 1 month ago

❓ General Questions

Hello! I am experimenting on an Android device and I have a few questions. (Device : Galaxy s24 Ultra | Model : meta-llama/Llama-3.2-3B-Instruct)

If I do input text into the app continuously, there are moments when the phone briefly freezes and the output generates repeated sentences.

Additionally, GPU usage fluctuates between 80% and 90%, but it starts using 100% at a certain point.

During this time, I observed the following error repeatedly appearing through logcat.

2024-10-02 16:28:38.515 27976-27976 ThemeUtils              ai.mlc.mlcchat                       E  View class dev.jeziellago.compose.markdowntext.CustomTextView is an AppCompat widget that can only be used with a Theme.AppCompat theme (or descendant).

Additionally, I would like to limit the length of the output within Android, but it doesn’t seem to work through the context-window-size setting.

How can I control the length of the app’s output?