feature/mnn #17

Merged
blusnake merged 8 commits from feature/mnn into main 2026-10-25 02:58:18 +00:00
Owner

fine tuning of llm model

fine tuning of llm model
Integrate private-process CPU inference, pinned model installation and settings, and guarded host task execution. Prioritize Qwen3.5 2B with 0.8B as a resource fallback. Preserve implemented messaging integration groundwork. Native text probes pass; reliable 2B tool selection remains unresolved and model-written module success claims are rejected. Inference defaults off. Record final builds, 161 JVM tests and remaining acceptance work.
Expose current time, exact duration conversion and date offsets through a globally retrievable core skill. Preserve host approval and source-parameter validation. Add structured interpreter controls, native tool-result turns, optional-field narrowing and process recovery coverage. Keep version 1.2.1 and 2B primary. Record 170 JVM tests, 76 reported Android tests, three actual-2B lifecycle/routing checks and optimized-release startup; retain model-quality and device-performance limitations.
Sign in to join this conversation.
No reviewers
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
blusnake/project-wise-wolf!17
No description provided.