README.md

---
license: mit
tags:
- llama-cpp-python
- prebuilt-wheels
- huggingface-spaces
- cpu-only
- python-3.13
---

# 🦙 Llama-CPP-Python Pre-built Wheels (Python 3.13)

### The solution for Hugging Face "Build Timeout" errors on the Free CPU Tier.

If you are using **Python 3.13** on a Hugging Face Free Space, compiling `llama-cpp-python` from source usually crashes or times out. This repository provides pre-compiled **manylinux** wheels that install in seconds.

---

## 🚀 Why use these wheels?

* **No Compilation:** Skips the 15+ minute build process.
* **Python 3.13 Support:** Specifically built for the latest Python version.
* **Generic CPU Optimization:** Compiled with `GGML_NATIVE=OFF`. This ensures the model runs on HF's shared CPUs without "Illegal Instruction" or "Core Dump" errors.
* **Lightweight:** Only ~4.3 MB compared to the massive overhead of building from source.

---

## 🛠️ How to use in your HF Space

### Option A: Using `requirements.txt`
Simply paste this direct link into your `requirements.txt` file:

```text
[https://huggingface.co/James040/llama-cpp-python-wheels/resolve/main/llama_cpp_python-0.3.16-cp313-cp313-linux_x86_64.whl](https://huggingface.co/James040/llama-cpp-python-wheels/resolve/main/llama_cpp_python-0.3.16-cp313-cp313-linux_x86_64.whl)


Option B: Using a Dockerfile
If you are using a custom Docker setup, add this line:
RUN pip install [https://huggingface.co/James040/llama-cpp-python-wheels/resolve/main/llama_cpp_python-0.3.16-cp313-cp313-linux_x86_64.whl](https://huggingface.co/James040/llama-cpp-python-wheels/resolve/main/llama_cpp_python-0.3.16-cp313-cp313-linux_x86_64.whl)


📦 Build SpecificationsThese wheels were built using a high-performance automated pipeline on GitHub.SpecificationValuePython Version3.13PlatformLinux x86_64 (Manylinux)Build FlagsGGML_NATIVE=OFF, GGML_BLAS=OFFBuild SourceJameson040/my_lama-wheels