Troubleshooting
Fix common BitNet and 1-bit LLM issues
Build Fails (e.g. std::chrono / log.cpp)
Often due to compiler or llama.cpp version. Update your compiler (e.g. Clang), ensure submodules are up to date (git submodule update), and check GitHub Discussions for the latest workarounds.
Windows: clang or build not found
Use a Developer Command Prompt or PowerShell with Visual Studio environment. Run VsDevCmd.bat or Enter-VsDevShell so that clang is on PATH. See Installation and FAQ.
Model file not found
Confirm the GGUF path passed to -m is correct and the file exists. Run setup_env.py first if needed. See Usage and Models.
Out of memory (OOM)
Use a smaller model, reduce context size (-c), or close other apps. 1-bit models need less memory than FP16 but large context or big models still need sufficient RAM/VRAM. See Models and Best Practices.
CUDA errors
Ensure NVIDIA drivers and CUDA toolkit match the version BitNet expects. Try CPU-only if GPU is not required. Check GitHub Issues for similar reports.