Gemma 3-1B — System Architecture

System Overview

graph LR
    A[Raw Dataset] -->|create_subset.py| B(Training Subset)
    B -->|main.py| C{Fine-Tuning}
    C -->|LoRA Adapters| D[Saved Checkpoints]
    D -->|merge_lora.py| E[Merged Model]
    E -->|export_gguf.py| F[GGUF Format]
    E -->|inference.py| G((Chat Interface))
    
    style C fill:#f9f,stroke:#333,stroke-width:2px
    style G fill:#bbf,stroke:#333,stroke-width:2px

Reading This Diagram

This diagram illustrates the end-to-end workflow for fine-tuning and deploying the Gemma 3-1B model. Data flows from the raw Alpaca dataset, through subset creation, fine-tuning, model merging, GGUF export, and finally to an interactive inference interface. Each script is a key step in the pipeline.

---

Data Flow

sequenceDiagram
    participant User
    participant System
    User->>System: Provide dataset & config
    System-->>User: Subset created
    User->>System: Start training
    System-->>User: Checkpoints, reports
    User->>System: Merge & export
    System-->>User: GGUF model, inference

Reading This Diagram

This sequence diagram shows the interaction between the user and the system at each stage: data preparation, training, merging, exporting, and inference.

---

🔗 Related Documents

• index
• system-architecture

---

Last Updated: 2026-04-29
Version: 1.0
Status: Complete