46 lines
1.6 KiB
Markdown
46 lines
1.6 KiB
Markdown
# The Trial Literary Analysis SLM
|
|
|
|
## Project Overview
|
|
Building a specialized Small Language Model (SLM) for comprehensive analysis of "The Trial" by Franz Kafka.
|
|
|
|
## Capabilities
|
|
- Factual Q&A about plot, characters, and timeline
|
|
- Literary analysis (themes, symbolism, narrative techniques)
|
|
- Creative content generation in Kafka's style
|
|
- Contextual conversation with cross-references
|
|
|
|
## Hardware Requirements
|
|
- GPU: 8-16GB VRAM (RTX 3080/4080 recommended)
|
|
- Storage: 50GB for models and data
|
|
- RAM: 16GB+ recommended
|
|
|
|
## Project Structure
|
|
```
|
|
the-trial-slm/
|
|
├── data/ # Training datasets
|
|
├── models/ # Model files and adapters
|
|
├── scripts/ # Training and utility scripts
|
|
├── notebooks/ # Jupyter notebooks for development
|
|
├── tests/ # Evaluation and testing scripts
|
|
└── deployment/ # Ollama integration files
|
|
```
|
|
|
|
## Training Phases
|
|
1. **Data Preparation** (Weeks 1-2)
|
|
2. **Infrastructure Setup** (Week 3)
|
|
3. **Model Training** (Weeks 4-5)
|
|
4. **Ollama Integration** (Week 6)
|
|
5. **Testing & Refinement** (Week 7)
|
|
|
|
## Base Model
|
|
- **Primary**: Llama 3.2 3B Instruct
|
|
- **Method**: QLoRA fine-tuning
|
|
- **Target**: Consumer GPU deployment
|
|
|
|
## Key Differences from Monte Cristo Project
|
|
- **Source Text**: "The Trial" by Franz Kafka (Project Gutenberg #7849)
|
|
- **Literary Period**: Early 20th century existential literature
|
|
- **Key Themes**: Bureaucratic absurdity, alienation, guilt vs innocence
|
|
- **Style**: Absurdist, nightmarish, psychological realism
|
|
- **Protagonist**: Josef K. (bank clerk arrested without cause)
|