Rayzed0224/Hot-Swappabl-LoRA-Adapters

FYP Workspace

This repository is organized by function rather than by loose root scripts.

Start Here

• Workspace layout: docs/workspace_guide.md
• Final report mapping: docs/final_report_map.md
• Report test matrix: docs/report_test_cases.md
• Quick commands: readme.txt

Main Folders

• fyp_core/: shared configuration, router logic, runtime guardrails, and dataset helpers
• apps/: active inference entrypoints
• training/: active LoRA training workflows
• evaluation/: active benchmark and probe scripts
• local_agent/: packaged desktop agent implementation
• launchers/: desktop/backend launcher entrypoints
• trained_local/: trained adapter folders
• results/: evaluation outputs and summaries
• archive/: legacy scripts and backups kept for reference

Common Commands

Run the Gradio app in WSL:

python3 -m apps.app

Train one niche adapter:

python3 -m training.train_expert --adapter minecraft_modding

Run the active niche benchmark:

python3 -m evaluation.benchmark_niche_adapters

Build the desktop executable on Windows:

.\build_local_agent_exe.ps1