README.md

metadata

title: UVM Verification Framework
emoji: 🔬
colorFrom: blue
colorTo: indigo
sdk: docker
app_file: src/api/server.py
pinned: false

UVM Verification Framework

Automated UVM testbench generation from YAML / FuseSoC .core specifications — with protocol libraries, schema validation, coverage-driven auto-training, and CI/CD integration.

Supports: UART (16550), SPI, I2C, APB, AXI4-Lite, Wishbone

---

Quick Start

pip install -r requirements.txt

# Generate a UVM testbench
python -m src.main --spec configs/uart16550-1.5.core

# With AI auto-training
python -m src.main --spec configs/uart16550-1.5.core --auto-train --max-iterations 3

# Run regression
python regression/run_regression.py --spec configs/uart16550-1.5.core

Features

UVM Generation

• Complete UVM environment: agent, driver, monitor, scoreboard, coverage, RAL
• IEEE 1800.2-2020 compliant SystemVerilog
• Spec-driven register map with DLAB-aware addressing (UART 16550)
• Functional coverage groups with cross coverage
• SVA protocol assertions
• RAL model with adapter and predictor

AI/ML Coverage Optimization

• Reinforcement Learning: Double Q-learning with prioritized replay, n-step returns, eligibility traces
• Coverage Prediction: 3-model ensemble (RF + GBR + LR) with confidence calibration
• Auto-Training: Generate → simulate → analyze → improve loop
• Exploration Strategies: epsilon-greedy, softmax, UCB, Thompson, NoisyNet

CI/CD

• Docker container with FastAPI + React frontend
• Hugging Face Space deployment ready
• REST API for generation and evaluation
• ZIP export of generated testbenches

Documentation

Document	Description
User Guide	Full walkthrough: install, spec format, integration, regression
API Docs	FastAPI Swagger UI (when running)

Project Structure

├── configs/           # YAML / .core specification files
├── docs/              # Documentation
├── frontend/          # React UI (Vite + TypeScript)
├── protocols/         # Protocol definitions
├── regression/        # Regression scripts
├── src/
│   ├── features/      # Spec feature extraction
│   ├── generation/    # Jinja2 templates + engine
│   ├── models/        # RL, coverage prediction, spec adaptation
│   ├── evaluation/    # SV syntax check, quality scoring
│   ├── pipeline.py    # Auto-training pipeline
│   └── main.py        # CLI entry point
├── vip/               # Packaged VIP files
└── output/            # Generated testbenches

API Server

uvicorn src.api.server:app --host 0.0.0.0 --port 8000

Frontend

cd frontend && npm install && npm run build

Docker

docker-compose up --build
# Frontend: http://localhost:7860
# API:      http://localhost:8000/docs

Regression

# Quick smoke test
python regression/smoke.py

# Full regression
python regression/run_regression.py \
  --regression regression/uart_full.yaml

# Custom test + seeds
python regression/run_regression.py \
  --spec configs/uart16550-1.5.core \
  --tests smoke,loopback,interrupt \
  --seeds 5

Tests

pytest tests/ -v

License

MIT