Spaces:
Sleeping
Sleeping
| title: EvoPlatform | |
| emoji: π» | |
| colorFrom: indigo | |
| colorTo: purple | |
| sdk: gradio | |
| sdk_version: 5.38.1 | |
| app_file: app.py | |
| pinned: false | |
| license: mit | |
| # π§ EvoTransformer β The AI That Evolves | |
| **Why Evo?** | |
| Most AI models are frozen in time. Evo learns. Evo mutates. Evo survives. | |
| It evolves its architecture based on **your feedback** β live. | |
| ### 𧬠Core Features | |
| - Learns from your questions and feedback | |
| - Evolving Transformer architecture (layers, attention heads, FFN dim, memory) | |
| - GPT-3.5 comparison built in | |
| - No manual fine-tuning β Evo retrains live | |
| - Live architecture mutation + genome tracking | |
| ### π οΈ Tech Stack | |
| - PyTorch | |
| - Transformers (tokenizer only) | |
| - Gradio (interactive UI) | |
| - Hugging Face or Local/Colab | |
| - OpenAI API (for GPT-3.5 comparison) | |
| --- | |
| ### π Try it out | |
| > Ask Evo a question. Choose two options. It tells you which makes more sense β and why. | |
| > Then **vote on who was better: Evo or GPT**. Evo adapts live. | |
| --- | |
| ### π‘ Example: | |
| **Q**: Blue is a | |
| **Option 1**: Color | |
| **Option 2**: Fruit | |
| **Evo Answer**: Color (Reasoning: 0.497 vs 0.431) | |
| **GPT-3.5 Answer**: color | |
| You vote β Evo β Evo learns from that and retrains in real time. | |
| --- | |
| ### π§ Evo isn't trained. | |
| Evo **learns to train itself**. | |