README.md

---
title: SafeClaw Coder LoRA
emoji: 🧑‍💻
colorFrom: indigo
colorTo: blue
sdk: gradio
sdk_version: "5.0.1"
app_file: app.py
pinned: false
license: apache-2.0
short_description: Qwen2.5-Coder-0.5B + SafeClaw LoRA on CPU.
---

# SafeClaw Coder LoRA — public CPU demo

This Space hosts a small Gradio chat UI for the SafeClaw fine-tuning PoC.
It loads:

- **Base model:** [`Qwen/Qwen2.5-Coder-0.5B`](https://huggingface.co/Qwen/Qwen2.5-Coder-0.5B)
- **LoRA adapter:** [`vladpp91/Tett`](https://huggingface.co/vladpp91/Tett)
- **Training data:** [`bigcode/the-stack-smol-xl`](https://huggingface.co/datasets/bigcode/the-stack-smol-xl)
  (python + javascript + typescript + go + rust subsets)

Source code for the SafeClaw integration and the training script lives
at [github.com/goodifro-pixel/safeclaw](https://github.com/goodifro-pixel/safeclaw).

## ⚠️ Free CPU Space — slow generation

This Space deliberately runs on the **free CPU tier** (16 GB RAM, no GPU).
Even at 0.5B parameters on CPU, generation is slow: short prompts come
back in roughly 10–30 seconds, longer ones in a minute or so. Output is
streamed token-by-token so you can see progress.

If you want sub-second responses, duplicate this Space and switch the
hardware to a GPU tier in the Space settings — no code changes are
needed.