TanujInsane
initial commit
8ddfad6
|
Raw History Blame Contribute Delete
3.51 kB
# QUICKSTART - Document Classification OpenEnv
## 30-Second Setup
```bash
cd c:\Users\91748\Desktop\metax
pip install -r requirements.txt
python test_environment.py
```
## 5-Minute Demo
```python
from environment import DocumentClassificationEnv
from grading import BaselineAgent
# Create environment
env = DocumentClassificationEnv("easy")
obs, _ = env.reset()
# Run agent
agent = BaselineAgent("easy")
done = False
total_reward = 0
while not done:
action = agent.decide(obs)
obs, reward, done, _, info = env.step(action)
total_reward += reward
print(f"Score: {total_reward:.2f}")
```
## File Overview
| File | Purpose | Size |
|------|---------|------|
| `environment.py` | Core OpenEnv implementation | 420 lines |
| `tasks.py` | Data generation & features | 400 lines |
| `grading.py` | Agent evaluation system | 370 lines |
| `baseline_inference.py` | Baseline evaluation script | 100 lines |
| `app.py` | Web interface (Gradio) | 300 lines |
| `test_environment.py` | Unit tests | 180 lines |
| `example_usage.py` | Usage examples | 260 lines |
## Key Features
✅ **Full OpenEnv Compliance**
- Typed observation/action spaces
- step() / reset() / state() API
- openenv.yaml specification
✅ **3 Difficulty Levels**
- Easy: 5 categories, 100 documents
- Medium: 10 categories, 500 documents
- Hard: 20 categories, 1000 documents
✅ **Production Ready**
- Docker containerization
- Hugging Face Spaces deployment
- Comprehensive documentation
- Full test coverage
## Tasks
### Easy (Baseline: 0.78)
- Classification into 5 categories
- Pre-extracted features
- No time constraints
- Perfect for learning the API
### Medium (Baseline: 0.65)
- 10 categories with variants
- 2 second per-action time limit
- Balanced accuracy/speed
### Hard (Baseline: 0.52)
- 20 fine-grained categories
- 1 second per-action time limit
- Speed-accuracy tradeoff
## Common Commands
```bash
# Test everything
python test_environment.py
# Baseline evaluation
python baseline_inference.py --task easy
python baseline_inference.py --task all
# Examples
python example_usage.py
# Web interface
python app.py
# Docker
docker build -t doc-classifier .
docker run -p 7860:7860 doc-classifier
```
## API Quick Reference
```python
from environment import DocumentClassificationEnv
# Create
env = DocumentClassificationEnv("easy")
# Reset
obs, info = env.reset()
# Step
obs, reward, done, truncated, info = env.step(action)
# State
state = env.state()
# Properties
env.action_space # Discrete(5/10/20)
env.observation_space # Dict
env.CATEGORY_MAPS # Category names
```
## Custom Agent Template
```python
def my_agent(observation):
# Your logic here
features = observation['features']
action = classify(features)
return action
# Evaluate
from grading import AgentGrader
grader = AgentGrader("easy")
score, metrics = grader.grade_agent(my_agent)
```
## Expected Results
Perfect agent:
- Easy: 1.0
- Medium: 0.95
- Hard: 0.90
Baseline agent:
- Easy: 0.78
- Medium: 0.65
- Hard: 0.52
## Next Steps
1. **Learn**: Read README.md
2. **Explore**: Run example_usage.py
3. **Test**: Run test_environment.py
4. **Build**: Create your agent
5. **Deploy**: Use Docker or HF Spaces
## Resources
- `README.md` - Full documentation
- `BUILD_SUMMARY.md` - What was built
- `DEPLOYMENT_GUIDE.md` - Deployment instructions
- `openenv.yaml` - OpenEnv specification
- `example_usage.py` - Code examples
---
**Ready to start?** → `python test_environment.py`