Metadata-Version: 2.4
Name: retentionengine
Version: 0.0.1
Summary: A simple adapter implementation to transform pretrained Transformer-family models into the Titans architecture.
Author-email: RetentionLabs <retentionlabs.ai@gmail.com>
Requires-Python: >=3.9, <3.14
Description-Content-Type: text/markdown
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Science/Research
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.9
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Programming Language :: Python :: 3.13
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
License-File: LICENSE
Requires-Dist: numpy
Requires-Dist: torch>=2.0.0
Requires-Dist: transformers>=4.30.0
Requires-Dist: pytest>=7.0.0 ; extra == "dev"
Requires-Dist: black>=23.0.0 ; extra == "dev"
Requires-Dist: isort>=5.12.0 ; extra == "dev"
Requires-Dist: mypy>=1.0.0 ; extra == "dev"
Requires-Dist: sphinx>=6.0.0 ; extra == "docs"
Requires-Dist: sphinx-rtd-theme>=1.2.0 ; extra == "docs"
Project-URL: Bug Tracker, https://github.com/retentionlabs/retentionengine/issues
Project-URL: Documentation, https://retentionengine.github.io
Project-URL: Homepage, https://github.com/retentionlabs/retentionengine
Provides-Extra: dev
Provides-Extra: docs

# RetentionEngine
A simple adapter implementation to transform pretrained Transformer-family models into the Titans architecture.

## Experiments Setup
### Exp 1: Comparison with Base Models [3B]
#### Baseline
- Titans 4B (Gemma3 4B IT)
- Llama3.2 3B Instruct
- Qwen3 4B
- Gemma3 4B IT
- RAG + Llama3.2 3B Instruct
- RAG + Qwen3 4B
- RAG + Gemma3 4B IT

#### Suggested
- Atlas-L 3B (Llama3.2 3B Instruct)
- Atlas-Q 4B (Qwen3 4B)
- Atlas-G 4B (Gemma3 4B IT)

### Exp 2: Comparison with Base Models [8B]
#### Baseline
- Titans 9B (Gemma2 9B IT)
- Llama3.1 8B Instruct
- Qwen3 8B
- Gemma3 9B IT
- RAG + Llama3.1 8B Instruct
- RAG + Qwen3 8B
- RAG + Gemma3 9B IT

#### Suggested
- Atlas-L 8B (Llama3.1 8B Instruct)
- Atlas-Q 8B (Qwen3 8B)
- Atlas-G 9B (Gemma3 9B IT)

### Exp 3: Comparison with Recurrent Models [8B]
#### Baseline
- Titans 9B (Gemma2 9B IT)
- Mamba2 8B
- RWKV6 7B

#### Suggested
- Atlas-L 3B (Llama3.2 3B Instruct)
- Atlas-Q 4B (Qwen3 4B)
- Atlas-G 4B (Gemma3 4B IT)
- Atlas-L 8B (Llama3.1 8B Instruct)
- Atlas-Q 8B (Qwen3 8B)
- Atlas-G 9B (Gemma3 9B IT)


## Build
```bash
git clone https://github.com/retentionlabs/RetentionEngine.git
cd RetentionEngine
uv pip sync
```

## Usage
- See `usage.ipynb` for usage examples.

## Deploy
```bash
rm -rf dist/
rm -rf build/
rm -rf *.egg-info

python -m pip install --upgrade pip
pip install flit
python -m flit build

echo "Uploading..."
python -m flit publish

echo "Deployment complete!"
```

