From 164bd2130ae2eb5f4bcd68d7fbda843d1caea98b Mon Sep 17 00:00:00 2001 From: juli Date: Sun, 9 Aug 2026 22:12:48 +0200 Subject: [PATCH] Switch model to Q6_K (~1.5 GB), lighter/faster than Q8 - ModelDownloader now pulls the Q6_K quant - README: reflect Q6_K sizes + note prebuilt APK on Releases Co-Authored-By: Claude Opus 4.8 --- README.md | 8 +++++--- .../java/monster/autisme/tsjetpiti/ModelDownloader.kt | 4 ++-- 2 files changed, 7 insertions(+), 5 deletions(-) diff --git a/README.md b/README.md index 0dc421b..a06e9fb 100644 --- a/README.md +++ b/README.md @@ -30,7 +30,7 @@ conversation and start fresh. - **Inference:** llama.cpp compiled from source (pinned tag `b10333`) through the NDK. Pulled automatically at build time via CMake `FetchContent` — no submodule to init. -- **Model:** `Qwen3.5-2B-Uncensored-HauhauCS-Aggressive-Q8_0.gguf` (~1.9 GB) is +- **Model:** `Qwen3.5-2B-Uncensored-HauhauCS-Aggressive-Q6_K.gguf` (~1.5 GB) is **downloaded on first launch** from Hugging Face into the app's private storage. It is *not* bundled in the APK. (Swap the quant in `ModelDownloader.kt`.) - **Persona:** set via `SYSTEM_PROMPT` in `MainActivity.kt`; sampling is a little @@ -47,7 +47,7 @@ conversation and start fresh. in **Android Studio** (it will offer to install the matching NDK `27.2.12479018` and CMake `3.22.1`). - **A phone:** 64-bit ARM (`arm64-v8a`), Android 8.0+ (minSdk 26), and enough free - RAM to hold a ~1.9 GB Q8 model (~4+ GB RAM device recommended). + RAM to hold a ~1.5 GB Q6 model (~4 GB RAM device recommended). - Network on first launch to download the model. ## Build @@ -63,9 +63,11 @@ The APK lands in `app/build/outputs/apk/debug/`. Or just Run ▶ from Android St ## Install (sideload) +A prebuilt `app-debug.apk` is attached to every [Release](../../releases). + The debug APK is signed with the debug key, so you can sideload it directly: copy `app-debug.apk` to the phone, enable "install unknown apps" for your file -manager, and tap it. First launch downloads the ~1.9 GB model over Wi-Fi. +manager, and tap it. First launch downloads the ~1.5 GB model over Wi-Fi. ## Where things live diff --git a/app/src/main/java/monster/autisme/tsjetpiti/ModelDownloader.kt b/app/src/main/java/monster/autisme/tsjetpiti/ModelDownloader.kt index 73cede1..26ed81c 100644 --- a/app/src/main/java/monster/autisme/tsjetpiti/ModelDownloader.kt +++ b/app/src/main/java/monster/autisme/tsjetpiti/ModelDownloader.kt @@ -7,11 +7,11 @@ import java.net.URL /** * Downloads the GGUF model on first launch and caches it in the app's private - * files dir. ~1.9 GB (Q8_0), so it only happens once. + * files dir. ~1.5 GB (Q6_K), so it only happens once. */ object ModelDownloader { - const val MODEL_FILENAME = "Qwen3.5-2B-Uncensored-HauhauCS-Aggressive-Q8_0.gguf" + const val MODEL_FILENAME = "Qwen3.5-2B-Uncensored-HauhauCS-Aggressive-Q6_K.gguf" private const val MODEL_URL = "https://huggingface.co/HauhauCS/Qwen3.5-2B-Uncensored-HauhauCS-Aggressive" +