akhil-jr/aknano-500m-base

🚨 IMPORTANT: DO NOT DOWNLOAD THESE WEIGHTS MANUALLY. This model uses a custom architecture. Standard inference scripts or GGUF converters will fail. To run this model, just clone this repo and run: 👉 [https://github.com/akhil-jr/aknano-custom-language-model.git]

---

Architecture & Attribution

This model is not a standard Llama or Mistral clone. It is a tiny 500 million model built from scratch using the incredibly lightweight and efficient nanochat framework created by Andrej Karpathy.

How to Use

If you are a machine learning engineer or data architect looking to test this:

1. Clone the central hub: git clone https://github.com/akhil-jr/aknano-custom-language-model.git
2. Install dependencies: pip install -r requirements.txt
3. Launch the dynamic UI: python chat_web.py --model base (The script will automatically fetch these weights via the Hugging Face Hub API).