Files
NOVA/configs/nova.yml
Dani a7f091aa45 Initial commit: NOVA - Neuro-Optimizing Versatile Agent
Complete transformer LLM built from scratch with:

Core Features:
- Full transformer architecture (RoPE, RMSNorm, SwiGLU, KV-cache)
- SentencePiece tokenizer (BPE/Unigram)
- Training pipeline (AMP, gradient checkpointing, DDP)
- Persona system with personality matrix (NO AI disclosure by default)
- Genetic evolution (NOVA-EVO) for hyperparameter optimization
- Legal-only data pipeline with license tracking
- Chat interface (CLI + REST API)
- Conversation memory (SQLite)

Model Sizes:
- 125M, 350M, 1.3B, 3B parameters
- Local-first, runs on CPU or GPU
- Python 3.10.6+, PyTorch 2.0+

Personas:
- girlfriend_gentle (high warmth, high empathy)
- girlfriend_playful (high humor, high playfulness)
- girlfriend_supportive (balanced, default)

Documentation:
- Complete README with quickstart
- Model card with ethical considerations
- Privacy documentation (local-first, zero telemetry)
- Data licenses and attribution
- Contributing guide

Infrastructure:
- GitHub Actions CI/CD
- Comprehensive test suite
- Quickstart script
- CLI tool

License: Apache 2.0

🤖 Generated with Claude Code
https://claude.com/claude-code

Co-Authored-By: Claude <noreply@anthropic.com>
2025-10-12 20:56:37 -04:00

75 lines
1.6 KiB
YAML

# NOVA Master Configuration
# Neuro-Optimizing Versatile Agent
# Hardware and compute
hardware:
device: auto # auto, cpu, cuda, cuda:0, etc.
num_threads: 4
allow_cuda: true
cuda_version_min: "12.0"
# Model selection
model:
size: 125m # 125m, 350m, 1.3b, 3b
config_path: null # Optional custom config
checkpoint_path: null # Path to load weights
# Persona and personality
persona:
default: girlfriend_supportive # girlfriend_gentle, girlfriend_playful, girlfriend_supportive
always_disclose: false # NO AI disclosure by default
custom_persona_path: null # Path to custom persona YAML
# Evolution (NOVA-EVO)
evolution:
enabled: false # Opt-in
budget: small # small, medium, large
population_size: 20
num_generations: 10
max_time_hours: 24.0
# Data
data:
legal_only: true # ONLY use properly licensed data
sources:
- toy-dataset # Default for offline demo
cache_dir: data/cache
preprocessed_dir: data/processed
# Training defaults
training:
learning_rate: 3.0e-4
batch_size: 8
gradient_accumulation_steps: 4
warmup_steps: 1000
max_steps: 100000
save_steps: 1000
eval_steps: 500
# Inference defaults
inference:
max_new_tokens: 200
temperature: 0.8
top_p: 0.9
top_k: 50
repetition_penalty: 1.1
# Memory
memory:
enabled: true
db_path: memory.db
max_context_length: 2048
# Logging and monitoring
logging:
level: INFO
wandb_enabled: false
wandb_project: null
tensorboard_enabled: false
# Safety
safety:
content_filter: true # Basic safety filters
max_generation_length: 500
timeout_seconds: 30